- Email: [email protected]
Time-correlated pulse-height measurements of low-multiplying nuclear materials
Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Time-correlated pulse-height measurements of low-multiplying nuclear materials E.C. Miller a,n, J.L. Dolan a, S.D. Clarke a, S.A. Pozzi a, A. Tomanin b, P. Peerani b, P. Marleau c, J.K. Mattingly d a
Department of Nuclear Engineering and Radiological Sciences, University of Michigan, Ann Arbor, MI, USA European Commission EC-JRC-IPSC, Ispra, Italy c Sandia National Laboratories, Livermore, CA, USA d North Carolina State University, Raleigh, NC, USA b
art ic l e i nf o
a b s t r a c t
Article history: Received 12 February 2013 Received in revised form 16 June 2013 Accepted 17 June 2013 Available online 22 June 2013
Methods for the determination of the subcritical neutron multiplication of nuclear materials are of interest in the field of nuclear nonproliferation and safeguards. A series of measurements were performed at the Joint Research Center facility in Ispra, Italy to investigate the possibility of using a time-correlated pulse-height (TCPH) analysis to estimate the sub-critical multiplication of nuclear material. The objective of the measurements was to evaluate the effectiveness of this technique, and to benchmark the simulation capabilities of MCNPX-PoliMi/MPPost. In this campaign, two lowmultiplication samples were measured: a 1-kg mixed oxide (MOX) powder sample and several lowmass plutonium–gallium (PuGa) disks. The measured results demonstrated that the sensitivity of the TCPH technique could not clearly distinguish samples with very-low levels of multiplication. However, the simulated TCPH distributions agree well with the measured data, within 12% for all cases, validating the simulation capabilities of MCNPX-PoliMi/MPPost. To investigate the potential of the TCPH method for identifying high-multiplication samples, the validated MCNPX-PoliMi/MPPost codes were used to simulate sources of higher multiplications. Lastly, a characterization metric, the cumulative region integral (CRI), was introduced to estimate the level of multiplication in a source. However, this response was shown to be insensitive over the range of multiplications of interest. & 2013 Elsevier B.V. All rights reserved.
Keywords: Liquid scintillators Sub-critical multiplication Neutron–gamma correlation MCNPX-PoliMi
1. Introduction The ability to rapidly determine the characteristics of an unknown nuclear sample and classify it as a threat or non-threat is of great interest in the areas of non-proliferation and national security. One signature that can be used to identify a threat source is multiplication. High source multiplications are typically indicative of special nuclear material (SNM). Multiplication measurements are typically performed using large assemblies of 3He detectors. While effective, these 3He detectors have several disadvantages. First, they are increasingly expensive due to the shortfall in the supply of 3He. Second, 3He detectors rely on the detection of thermalized neutrons which obscures potentially useful timing and energy information from the source. Liquid scintillator detectors can be used to directly measure the gamma-ray and fast neutron spectrum from a source. The nanosecond time response of liquid scintillator detectors is much faster
n
Corresponding author. Tel.: +1 443 778 5176. E-mail address: [email protected] (E.C. Miller).
0168-9002/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.nima.2013.06.062
than the microsecond response for a 3He detector. This fast timing allows for the analysis of detected particles on the time scale of individual fission events and not over the combined response of multiple fissions. This additional information could be potentially used to develop a detection system that can offer better information about the source in question. Past work has demonstrated that cross-correlation measurements with liquid scintillator detectors can be used to characterize source material [1]. Cross-correlation measurements identify time-correlated events using multiple detectors. Using pulseshape discrimination (PSD) subtle differences in the shape of the pulses produced by neutrons and gamma-ray interactions can be identified and the detected event can be attributed to the interacting particle type. This allows the time-correlated events to be classified by interacting particle: neutron–neutron (n–n), neutron– gamma (n–γ), gamma-neutron (γ–n), or gamma–gamma (γ–γ). The ability to classify the types of correlated events can yield additional information that can be used to characterize an unknown sample. Time-correlated pulse-height (TCPH) is an expansion of these cross-correlation measurements that include pulse-height information
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
in the analysis of the γ–n events. A γ–n event is defined as the arrival of a gamma-ray in an arbitrarily defined “start” detector followed by the detection of a neutron event in a second “stop” detector. Using the time difference between the events (stop-start), and the pulse-height information from the arriving neutron, a surface plot of pulse height at specific arrival times can be created. This surface can be used to identify events from fission chains. If the gamma and neutron contributing to a γ–n event are released from the same fission event, the expected arrival time, t (s), for a neutron with energy En (MeV), can be determined using d d t ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi − 2En =M n c
ð1Þ
where d (m) is the source–detector distance, Mn (MeV) is the mass of the neutron, and c (m/s) is the speed of light. The maximum amount of light generated in the detector can be estimated for a neutron of a given energy using a detector specific energy-to-light conversion shown in Eq. 2 [2]. L ¼ aEp −bð1−expðcEp ÞÞ
ð2Þ
where L (MeVee) is the light out, Ep (MeV) is the energy deposited by the incident neutron, and a, b, and c are detector specific coefficients. The coefficients used for this paper were: a ¼0.817, b¼ 2.63 and c ¼−0.297 [3]. Using Eqs. 1 and 2 the maximum light produced at a given arrival time produces a discrimination line for the TCPH distribution. All true correlated, non-multiplying source events must fall below this discrimination line. However, if the measured material has some level of multiplication, it becomes possible to detect correlated events from the same fission chain. Neutrons from later generation events in the fission chain can be correlated with earlier generation gamma-rays. This effect results in the possibility of detecting truly correlated events that arrive at greater times and generate more light than predicted by the discrimination line. By characterizing the amount of events past the discrimination line it may be possible to estimate the level of multiplication in a measured material. The TCPH technique was modeled extensively using the Monte Carlo code MCNPX-PoliMi. MCNPX-PoliMi is an enhanced version of MCNPX that allows for the simulation of correlated neutrons and gamma-rays [4,5]. The MCNPX-PoliMi post processing code MPPost was used to simulate the detector response to the particle transport data [6,7].
2. Experiment and analysis A series of measurements were performed at the Joint Research Center in Ispra, Italy. The campaign focused on measurements of mixed oxide (MOX) powder and plutonium–gallium (PuGa) disks to evaluate the sensitivity of the TCPH method to low multiplication samples. Each measurement conjuration was later simulated to validate the capabilities of MCNPX-PoliMi and MPPost. 2.1. Setup This measurement campaign used four EJ-309 liquid scintillator detectors measuring 7.62 cm in diameter by 7.62-cm long. The signals from the detector cells were digitized using a CAEN DT5720 digitizer with University of Michigan (UM) developed acquisition software. The DT5720 is a 12-bit, 250 MHz, 4-channel digitizer that can transfer data to a computer via a USB connection [8]. This allows for data collection with a laptop, greatly improving portability of the acquisition system. The measurement system was set up on two tables that were approximately 75 cm from a concrete floor. The tables were
109
Table 1 The composition of the MOX canister as of April 2012 Isotope
Mass (g)
Weight Percent
234
0.05 4.79 0.05 670.50 0.24 111.81 47.00 1.67 3.38 5.12 166.22 1010.83
0.0001 0.0047 0.0001 0.6633 0.0002 0.1106 0.0465 0.0017 0.0033 0.0051 0.1644 1.0000
U U 236 U 238 U 238 Pu 239 Pu 240 Pu 241 Pu 242 Pu 241 Am O2 Total 235
constructed of 2.5-cm by 2.5-cm aluminum rods with a 4-mm thick aluminum surface. The four detectors were arranged in an arc with a 40-cm source–detector distance. The average spacing between the detector-centerlines was 15 cm. 2.2. MOX source The MOX powder source was a 1.011-kg powder with an estimated density of 0.7 g/cm3. The MOX powder was encased in a cylindrical canister that is 32.2 cm tall with a 10.4 cm diameter. The estimated MOX density was determined by using the wellknown mass and the volume of the storage container. While this estimated value is lower than typical powder samples, it is believed to be close to correct based on profile measurements of the source. The source information is presented in Table 1 [9]. The aged source intensity for the MOX sample was 8.22 104 neutrons per second. A majority of the source neutrons, 59%, come from the spontaneous fission of 240Pu. The remaining neutrons come from the spontaneous fission of 242Pu and a variety of (α, n) reactions. For the MOX measurements, a thin, 1-cm shell of lead was placed directly around the canister. Some amount of lead shielding was required for this measurement because the gamma-ray count rate, without the lead, was sufficient to overload the data acquisition setup at the 50-keVee measurement threshold. However, adding too much lead will negatively impact the ability to detect the correlated fission gamma-rays required for the TCPH technique. The 1-cm thickness was chosen in an attempt to balance these two variables. This represents an inherent challenge when for the TCPH technique, detecting correlated gamma-rays and neutrons while not overloading the detection system. Two measurements of the MOX source (ENEA-1) were performed, in a reflected (polyethylene added around the lead shielding) and unreflected (only lead shielding) configuration. 2.2.1. Unreflected measurement The MOX source is interesting because the material has some level of multiplication. The keff for the unreflected MOX source was estimated using MCNPX-PoliMi to be 0.014 which results in a source multiplication of 1.014 where the multiplication is defined as M¼
1 1−kef f
ð3Þ
The unreflected MOX canister was measured overnight for a total of 9.66 h. A picture of the measurement setup and the simulated geometry is shown in Fig. 1. The simulation did not include the detector PMTs, detector stands, or the stand used to support the MOX canister. These omissions will have a negligible effect on the results.
110
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
Fig. 1. The measurement setup for the 40-cm measurement of the unreflected MOX canister and the modeled geometry.
Fig. 2. TCPH distributions showing the log of the number of counts per second, (top) the raw measured γ–n spectrum showing the background radiation in the negative direction, (bottom) the true measured spectrum with the background removed.
An example of a TCPH distribution is shown in Fig. 2. The y-axis shows the pulse height of the neutron in MeVee. The temperature plot represents the log of counts-per-second. The x-axis represents the time of flight determined using the arrival time of neutron in the “stop” detector and the arrival of a γ in the “start” detector. The negative time values arise because the γ–n pairing is defined by which detector the particles arrive in. If a gamma-ray is detected in the start detector and the neutron is detected in the stop detector, a positive time difference is recorded. If the neutron arrives in the start detector and the gamma-ray in the stop detector then the time difference will be negative. In an effort to reduce the intense gammaray background produced by the MOX powder, the background distribution was subtracted from the TCPH distributions, as shown in Fig. 2. The accidental background was determined by taking the average of the γ–n correlations in the negative time direction and subtracting this from the TCPH distribution. This background subtraction was applied to all of the TCPH distributions in shown. Fig. 3 shows a comparison between a measured and simulated TCPH distribution. The solid black line overlaid on the distribution represents the maximum pulse height for a neutron arriving at the
front face of the detector determined using Eqs. 1 and 2. The dashed line represents the maximum pulse height for a neutron reaching the back face of the detector. This convention is followed for all TCPH figures shown in this paper. For a non-multiplying point source all non-accidental events detected should arrive below these two discrimination lines. The simulation has an excellent agreement with the measured distributions considering that this response requires that the simulation correctly predict the energy and arrival times with very fine resolution. The main observable difference between the two results can be found at very short-times and low-light values, where there are increased events in the measured distribution. This concentration of events is the result of misclassified particles, where the PSD incorrectly identified an arriving gamma-ray as a neutron. This is not unexpected, when using a traditional PSD approach it is especially difficult to distinguish neutrons and gamma-rays with low-energy. However, using arrival time information, misclassified gamma-ray events can be identified. The results in Fig. 3 show that there is a noticeable concentration of events that are beginning to move past the discrimination
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
111
Fig. 3. TCPH distributions for the unreflected MOX source showing the log of counts per second, (a) measured and (b) simulated. The solid overlaid line represents the maximum pulse height for a neutron arriving at the front face of the detector. The dashed line represents the maximum pulse-height for a neutron reaching the back face of the detector.
101 Measured MCNPX-PoliMi 100 Counts per Second
lines. This effect is partially due to the size of the extended source, and complicates the identification of late arriving neutrons. To provide a direct comparison between the measurement and simulated TCPH distributions the results were projected on the y-axis to produce a pulse-height distribution (PHD) and the projection on the x-axis produces a time-of-flight (TOF). These distributions are shown in Fig. 4. The error was calculated in two ways. First the total percent difference between the simulated and measured results was 10.56%. This value helps show the level of agreement between the source strength and geometry (i.e. isotopic composition, solid angle). The second quantity, the average point-by-point error identifies how well the transport physics and energy-to-light conversion of particles in the detector are handled. These errors are higher, but are also influenced by the total percent difference. For this case the average point-by-point errors are, 17.53% for the PHD and 19.96% for the TOF.
10-1
10-2
10-3 0
1 Light (MeVee)
1.5
2
0.05 Measured MCNPX-PoliMi 0.04 Counts per Second
2.2.2. Reflected measurement A polyethylene reflector was added to increase the multiplication of the MOX canister; MCNPX-PoliMi simulations for this configuration predicted a keff of 0.08 and a multiplication of 1.087. This measurement was performed overnight and a total of 14.9 h of data was collected. The polyethylene reflector was constructed in a “U” shape around the MOX cylinder. The reflector was used to increase the level of multiplication in the source without shielding the detected signal. The MOX canister was surrounded by 1 cm of lead and was raised from the table surface by 7 cm of polyethylene. The reflector was constructed of polyethylene slabs 60 8.0 2 cm. The measurement setup is displayed in Fig. 5 with the relevant dimensions for the polyethylene structure. The TCPH distributions are compared in Fig. 6. A background subtraction was performed on the measured data. There is very good agreement in the shapes of the distributions. There is a slight increase in the number of events falling above the discrimination lines than was observed in the previous case. As with the unreflected measurement, there is a large concentration of misclassified events in the measured distribution at short-time and low-light values. The total percent error for this measurement setup was −2.59%. The average errors were 16.88% for the PHD and 22.16% for the TOF. In both simulations of the MOX canister the peak value of the TOF distribution is slightly under predicted. This effect could be caused by a small change in the source–detector distance, either in the detector placements or as a result of an internal structure in the MOX sample.
0.5
0.03
0.02
0.01
0 0
10
20 30 Delta Time (ns)
40
50
Fig. 4. (a) PHD comparison for the MOX distribution and (b) TOF comparison for the MOX distribution.
2.3. PuGa source Three PuGa metal disks were measured together in both an unreflected and reflected configuration. There were several PuGa disks available that ranged in mass from 0.01 g to 9.81 g of plutonium. For this measurement, the three largest samples were used, accounting for 86% of the available plutonium mass. All of
112
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
Fig. 5. The reflected MOX measurement setup and polyethylene dimensions.
Fig. 6. TCPH comparison for the reflected MOX measurement, (a) measured and (b) simulated. The solid overlaid line represents the maximum pulse height for a neutron arriving at the front face of the detector. The dashed line represents the maximum pulse-height for a neutron reaching the back face of the detector.
Table 2 Isotopic composition and masses for the three PuGa samples measured Isotope
238
Pu Pu Pu 241 Pu 242 Pu 241 Am Ga Total 239 240
Mass (g)
Weight %
Disk 209
Disk 210
Disk 211
0.003 1.416 0.402 0.037 0.014 0.035 0.035 1.941
0.006 3.627 1.030 0.094 0.037 0.089 0.089 4.972
0.013 7.154 2.032 0.184 0.072 0.176 0.176 9.808
0.001 0.729 0.207 0.019 0.007 0.018 0.018
the disks had an identical source composition, shown in Table 2. The samples were 73% 239Pu by total mass. Combined, the three disks had a source intensity of only 3789 neutrons per second. The spontaneous fission of 240Pu, accounts for 93% of all source neutrons created. The next largest contribution comes from the spontaneous fission of 242Pu. Additional neutrons will be created from induced fission events in the 239Pu.
2.3.1. Unreflected measurement For the unreflected measurement the PuGa disks were placed in a Plexiglas stand that held the disks vertically. The large number of gamma-rays emitted from these disks was enough to overload the maximum data transfer rate of the data acquisition system. To reduce the count rate, a thin (2.5 mm) sheet of lead was added in front of the PuGa sources. Even with the lead shield the detected neutron/gamma-ray ratio was 0.0036.
The setup for the unreflected PuGa measurement is shown in Fig. 7. As with previous simulations the PMTs and source stands were not included. This measurement setup had a very low multiplication of 1.05 and a simulated keff of 0.0476. The measured and simulated TCPH distributions are shown in Fig. 8. There is very good agreement for the two distributions. Again the large amount of misclassified events is observed at short-times and low-light. This misclassification is exceptionally pronounced in this measurement because of the large amount of gamma-rays emitted from the source. There is an interesting feature in the simulated distribution where there is a small concentration of low-energy late-time events that are not seen in the measured result. These events will be from neutrons scattering in the near-environment and arriving in the detector. The high gamma-background obscures this feature in the measured result. Fig. 9 shows the comparison of the PHD and TOF. With the strong gamma-ray source even a very good PSD discrimination ratio will begin to have a significant amount of misclassified events. Fortunately, for this analysis these events can be removed by time-of-flight data. To remove these misclassified events and obtain a “true” measurement distribution, timing information was used to remove neutrons that arrive at unphysical times. The “true” measurement data was then compared to the simulated results. The total percent difference was −11.56%, and the average pointby-point error for the PHD and TOF were 33.29% and 33.94%, respectively. There is a clear shift in the measured TOF distribution compared to the simulated result. This shift is more significant than the one that was observed in the MOX case. The shift in the position of the TOF distribution is likely caused by a small shift in
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
113
Fig. 7. A photograph of the unreflected PuGa measurement showing the Plexiglas holder with the thin lead compared to the MCNPX-PoliMi simulated geometry.
Fig. 8. TCPH distributions for the unreflected PuGa source, (a) measured and (b) simulated. The solid overlaid line represents the maximum pulse height for a neutron arriving at the front face of the detector. The dashed line represents the maximum pulse-height for a neutron reaching the back face of the detector.
the actual source–detector distance for the four detectors compared to the simulation. It would be more effective in future measurements to have the detectors in a fixed assembly to more easily control the source–detector distance. 2.3.2. Reflected measurement The PuGa disks were also measured in a reflected configuration. To increase the overall multiplication of the measured system a polyethylene structure was built around the Plexiglas source holder. The polyethylene structure was 60 cm long, 35 cm tall, and 16 cm thick. The addition of the polyethylene increased the multiplication from 1.05 to 1.068. The source–detector distance remained 40 cm. The measurement setup is shown with the simulated geometry in Fig. 10. A 2.5 mm sheet of lead was again required to reduce the number of detected gamma-rays. Even with the lead shielding the detected neutron-to-gamma ratio was 0.004. Fig. 11 shows good agreement between the measured and simulated TCPH distributions. There are a small number of events past the discrimination line in the simulation that are not observed in the measurement. These low-count fringe events were obscured by the background in the measured data and were ultimately removed when the background subtraction was applied. The other observable difference is the misclassified gamma-ray events, clearly visible in the measured result at lowlight and short-times. For this case the total percent difference was 8.56%. The average point-by-point error for the PHD is 30.37% and 30.87% for the TOF.
are summed together creating a cumulative region integral (CRI) distribution. This approach measures the gradient of the distribution. Accidental events should influence the entire distribution evenly and should not have a significant impact on the gradient of the distribution. The CRI distribution results for the Ispra measurement are compared in Fig. 13. As the source multiplication increases the gradient over the non-multiplying discrimination line becomes increasingly flat. As a result, the CRI distribution will have a decreasing slope with increasing multiplication. This trend can be observed for the 252Cf and PuGa results shown in Fig. 13. The CRI distribution of the 252Cf source, with a multiplication of 1, is farthest to the left. As expected, the unreflected and reflected PuGa CRI distributions are located to the right of the 252Cf case. The two MOX CRI distributions appear with a much lower slope than expected compared to the other cases; however, they are positioned correctly relative to each other. This change in the slope of the MOX cases is the result of the extended source geometry. The extended source obscures the source–detector distance, decreasing the gradient around the discrimination line. The CRI distributions for all of the Ispra measurements were simulated with MCNPX-PoliMi, normalized to the integral number of counts, and compared to the measured results. The average point-by-point error for all of the Ispra cases is shown in Table 3. All of the cases agree within 3% for both the total error and average point-by-point error. These results show that MCNPX-PoliMi can accurately predict the shape of the CRI distribution. 3.1. Simulation of highly multiplying samples
3. Characterization of source multiplication The objective of the TCPH method is to provide a means for estimating the multiplication of an unknown sample. To approach this objective, multiple counting regions, evenly spaced, starting at the front face of the detector, are used. Fig. 12 shows the regions applied to the unreflected MOX source. All events in each region
All of the samples measured in the Ispra campaign have very low levels of multiplication. To examine the effectiveness of this characterization technique for sources with larger multiplication values the CRI discrimination region approach was applied to models of a reflected 4.5-kg plutonium sphere with 94% 239Pu and 6% 240Pu. The reflected plutonium sphere setup was chosen because it has been successfully modeled with MCNPX-PoliMi in
114
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
3.2. Estimating an unknown source
101 Measured Measured ("True" Events) MCNPX-PoliMi
Counts per second
100
10-1
10-2
10-3
10-4 0
8
0.5
1 1.5 Pulse Height (MeVee)
2
x 10-3 Measured Measured ("True" Events) MCNPX-PoliMi
7
Counts per second
6 5 4 3 2 1 0
0
10
20 30 Time of Flight (ns)
40
50
Fig. 9. A comparison of the unreflected PuGa measured data to simulated results. The “true” measured results have misclassified neutrons removed by timing. (a) PHD and (b) TOF.
previous work [10,11]. A complete description of the source can be found in the paper by Mattingly [11]. The simulation of the plutonium sphere measurement used two 12.7-cm diameter by 12.7-cm long EJ-309 scintillator cells placed side-by-side. The center of the plutonium sphere was placed 50 cm from the front face of the detectors. The plutonium sphere was surrounded by up to 15.24 cm of polyethylene reflectors to increase the multiplication. The table and floor were also included in the model. As the level of multiplication increases, the TCPH distribution becomes increasingly more in level in the region around the theoretical discrimination line. To clearly observe the effects of highly multiplying samples the number of regions must be increased until the edge of the distribution is found. The CRI distributions fall in the expected pattern when the number of regions is set to 250. Fig. 14 shows a clearly increasing trend with multiplication. The CRI distribution for the unreflected plutonium sphere case shown in Fig. 14 has a distinct bump around region 50. This feature is the result of neutrons scattering off of the floor reaching the detector. This effect is obscured in the reflected cases as more true coincident events begin to fall in these regions.
The CRI distributions provide a visual comparison of the level of multiplication in a measured sample, but make quantitative comparisons difficult. To provide a direct method of comparison the integral of the CRI distribution was taken. Fig. 15 shows a comparison of the CRI integrals for three sets of data. The first set of data is the results from the Ispra measurements of MOX powder and PuGa. The second set is the simulated reflected plutonium sphere cases. Third, an additional set of results for a plutonium sphere with varying radii was added to provide a baseline reference for multiplications over a reasonable range without other complicating factors. To create the data set for the plutonium sphere with a varying radius the plutonium sphere used the same model that was used in the reflected case with radii ranging from 2 cm to 4.8 cm. The radii investigated correspond to plutonium masses ranging from 0.6 kg to 9.1 kg of plutonium. This provides data points both above and below the 5 kg plutonium mass the IAEA defines as a significant quantity. The error bars on the figure represent the counting error for each case. It should be noted that the large errors bars for the Ispra data are the result of poor counting statistics limited by the available measurement times. The error for the plutonium source of varying radii increases as multiplication decreases because the same number of source events were simulated in each case. The samples with large multiplications resulted in many induced fission events resulting more detected events, reducing the error. When then CRI integrals for all three data sets are compared and plotted against the multiplication, a weak trend is apparent: as the level of multiplication increases, the area under the curve decreases. However, the level of sensitivity is very low. Very small changes in the CRI integral can result in significant changes in the predicted amount of multiplication. The large error bars resulting from poor counting statistics in the PuGa measurements correspond to a dramatic range of potential multiplications. This result shows that extracting multiplication data using the CRI integral method is not sensitive enough for the identification of an unknown source. However, this result does show that a trend exists and it may be possible, with a different approach, to extract more sensitive multiplication information from the TCPH distributions.
4. Conclusions TCPH measurements of low-multiplying source materials were successfully performed at the JRC in Ispra, Italy. The multiplications of the measured MOX and PuGa samples ranged from 1.05 to 1.087. The TCPH method was unable to clearly differentiate the source multiplications over this narrow range. However, these results were essential in validating the simulation capabilities of MCNPX-PoliMi and MPPost. The simulated results were within −11.56% of the measured values. With the simulation tools validated, the TCPH analysis was extended to simulate multiplying samples. Using MCNPX-PoliMi a highly multiplying 4.5-kg plutonium sphere was analyzed using TCPH. In an attempt to characterize the level of multiplication in the sample, a multi-region approach was introduced. CRI distributions for the Ispra measurements agreed within 3% of the simulated results. Using this validation the CRI simulations, this approach was applied to simulations of the plutonium sphere. These results were used to fit a line to the behavior of the CRI values over a range of multiplications with the intent of establishing a means of determining an unknown source multiplication. While
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
115
Fig. 10. A photograph of the reflected PuGa measurement showing the polyethylene structure compared to the simulated MCNPX-PoliMi geometry.
Fig. 11. The reflected PuGa TCPH distributions for the (a) measured data and (b) for the simulated results. The solid overlaid line represents the maximum pulse height for a neutron arriving at the front face of the detector. The dashed line represents the maximum pulse-height for a neutron reaching the back face of the detector.
1 0.9 Cf-252 PuGa (bare) PuGa (mod) MOX (bare) MOX (mod)
Normalized Counts
0.8 0.7 0.6
Increasing Multiplication
0.5 0.4 0.3 0.2 0.1 0 0
Fig. 12. A TCPH for the unreflected MOX case with 20 dividing regions (dashed lines) used to evaluate the level of multiplication.
the measurements trend weakly with source multiplication the CRI approach is not sensitive enough to the multiplication of the source to be used for the identification of an unknown multiplication. This measurement campaign successfully identified several problems with applying the TCPH approach in the field. First, the high gamma-ray rates from multiplying sources. The high gammaray rates create an extremely high accidental background that makes identifying late arriving neutrons difficult. The low neutron-to-gamma-ray ratio also negatively impacts the ability to do PSD by increasing the amount of misclassified events. Lastly, high gamma-ray backgrounds require adjustments to the measurement design to ensure that the digitizing setup is not
5
10 Region
15
20
Fig. 13. CRI distributions for the Ispra measurements.
Table 3 The percent difference between the measured and simulated CRI distributions
252
Cf MOX (unreflected) MOX (reflected) PuGa (unreflected) PuGa (reflected)
Percent difference (total)
Percent difference (average)
−0.45 −2.81 −1.89 0.75 0.33
0.57 −2.97 −1.66 2.89 2.71
overloaded. This problem can be mitigated by increasing the threshold of the measurement system, thereby reducing the lowenergy events that are more frequently misclassified and reduce the overall count rate.
116
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
TCPH distributions change in shape with different sources and geometries. Template matching the TCPH distribution to a library of sources and configurations may be a way to improve the applicability of this technique. These problems are not insurmountable, and with additional development it is possible that the TCPH technique could be used to identify the multiplication of an unknown source. However, significant development progress will be needed to address the issues identified in this measurement campaign.
1 0.9
Normalized Counts
0.8 0.7 0.6 0.5 0.4
Pu Sphere (bare) Pu Sphere (1.27 cm poly) Pu Sphere (2.54 cm poly) Pu Sphere (3.81 cm poly) Pu Sphere (7.62 cm poly) Pu Sphere (15.24 cm poly)
0.3 0.2 0.1
Acknowledgments
0 0
50
100
150 Region
200
250
Fig. 14. CRI distributions for the plutonium sphere using 250 regions to clearly resolve the increasing multiplication of the simulated samples.
Ispra Reflected Pu Sphere Different Radius Pu Sphere
290 270
CRI Integral
250
The authors would like to thank Santino Frison for his assistance in completing the measurements. This research was funded in part by the Nuclear Forensics Graduate Fellowship Program which is sponsored by the U.S. Department of Homeland Security's Domestic Nuclear Detection Office and the U.S. Department of Defense's Defense Threat Reduction Agency. This research was also funded by the National Science Foundation and the Domestic Nuclear Detection Office of the Department of Homeland Security through the Academic Research Initiative Award # CMMI 0938909 Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-AC04-94AL85000. SAND 2013-0690J
230
References
210 190 170 150 0
5
10 Multiplication
15
20
Fig. 15. CRI integral values for the Ispra measurements, reflected plutonium spheres, and plutonium spheres of varying radii.
Second, measurements of extended sources at close range, such as the MOX canister, introduce significant geometry effects to the TCPH distribution. It is difficult to balance the need for a short counting time and the need to be far enough from the source to have a point-like behavior. Third, the classification of the source multiplication is difficult. The placements of the discrimination lines are sensitive to distance, and so this value needs to be well known, and controlled. Using the gradient of events past the discrimination line (the CRI method) is not sensitive to the changes in multiplication. A more advanced approach will be needed. This work did show that the
[1] E.C. Miller, J.L. Dolan, S.D. Clarke, M. Flaska, S.A. Pozzi, E. Padovani, P. Peerani, P. Schillebeeckx, Experiments and simulation of cross-correlations on MOX fuel, in: Proceedings of the Institute of Nuclear Materials Management Annual Meeting, Palm Desert, CA, USA, July 17–22, 2011. [2] A. Enqvist, C.C. Lawrence, T.N. Massey, S.A. Pozzi, Neutron light output functions measured for EJ309 liquid scintillation detectors,in: Proceedings of INMM Annual Meeting, Orlando, FL, July 15–19, 2012. [3] A. Enqvist, C.C. Lawrence, B.M. Wieger, S.A. Pozzi, T.N. Massey, Nuclear Instruments and Methods A 715 (2013) 79. [4] E. Padovani, S.A. Pozzi, S.D. Clarke, E.C. Miller, MCNPX-PoliMi, RSICC Package number C00791 MNYCP 00, 〈http:/www.rsicc.ornl.gov〉. [5] S.A. Pozzi, S.D. Clarke, W. Walsh, E.C. Miller, J. Dolan, M. Flaska, B.M. Wieger, A. Enqvist, E. Padovani, J.K. Mattingly, D. Chichester, P. Peerani, Nuclear Instruments and Methods in Physics Research A (2012) 119. [6] E.C. Miller, A. Poitrasson-Riviere, A. Enqvist, J.L. Dolan, S. Prasad, M.M. Bourne, K. Weinfurther, S.D. Clarke, M. Flaska, S.A. Pozzi, E. Padovani, J.K. Mattingly, MCNPX-PoliMi Post-Processor (MPPost) Manual, RSICC Package number C00791 MNYCP 00, 〈www.rsicc.ornl.gov〉. [7] E.C. Miller, S.D. Clarke, S. Prasad, M. Flaska, S.A. Pozzi, E. Padovani, Journal of Nuclear Materials Management XL (2) (2012). [8] Technical Information Manual, MOD. DT5720, Revision n.5, CAEN, 〈www.caen. it〉, February 2012. [9] P. Peerani, M. Swinhoe, A.L. Weber, L.G. Evans, ESARDA Multiplicity Benchmark Exercise Phases III and IV, Esarda Bulletin 42 (2009). [10] E.C. Miller, B. Dennis, S.D. Clarke, S.A. Pozzi, J.K. Mattingly, Nuclear Instruments and Methods in Physics Research A 652 (2011) 540. [11] J. Mattingly, Polyethylene-Reflected Plutonium Metal Sphere: Subcritical Neutron and Gamma Measurements, SAND2009-5804-R2, Sandia National Laboratories 2009.
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research A journal homepage: www.elsevier.com/locate/nima
Time-correlated pulse-height measurements of low-multiplying nuclear materials E.C. Miller a,n, J.L. Dolan a, S.D. Clarke a, S.A. Pozzi a, A. Tomanin b, P. Peerani b, P. Marleau c, J.K. Mattingly d a
Department of Nuclear Engineering and Radiological Sciences, University of Michigan, Ann Arbor, MI, USA European Commission EC-JRC-IPSC, Ispra, Italy c Sandia National Laboratories, Livermore, CA, USA d North Carolina State University, Raleigh, NC, USA b
art ic l e i nf o
a b s t r a c t
Article history: Received 12 February 2013 Received in revised form 16 June 2013 Accepted 17 June 2013 Available online 22 June 2013
Methods for the determination of the subcritical neutron multiplication of nuclear materials are of interest in the field of nuclear nonproliferation and safeguards. A series of measurements were performed at the Joint Research Center facility in Ispra, Italy to investigate the possibility of using a time-correlated pulse-height (TCPH) analysis to estimate the sub-critical multiplication of nuclear material. The objective of the measurements was to evaluate the effectiveness of this technique, and to benchmark the simulation capabilities of MCNPX-PoliMi/MPPost. In this campaign, two lowmultiplication samples were measured: a 1-kg mixed oxide (MOX) powder sample and several lowmass plutonium–gallium (PuGa) disks. The measured results demonstrated that the sensitivity of the TCPH technique could not clearly distinguish samples with very-low levels of multiplication. However, the simulated TCPH distributions agree well with the measured data, within 12% for all cases, validating the simulation capabilities of MCNPX-PoliMi/MPPost. To investigate the potential of the TCPH method for identifying high-multiplication samples, the validated MCNPX-PoliMi/MPPost codes were used to simulate sources of higher multiplications. Lastly, a characterization metric, the cumulative region integral (CRI), was introduced to estimate the level of multiplication in a source. However, this response was shown to be insensitive over the range of multiplications of interest. & 2013 Elsevier B.V. All rights reserved.
Keywords: Liquid scintillators Sub-critical multiplication Neutron–gamma correlation MCNPX-PoliMi
1. Introduction The ability to rapidly determine the characteristics of an unknown nuclear sample and classify it as a threat or non-threat is of great interest in the areas of non-proliferation and national security. One signature that can be used to identify a threat source is multiplication. High source multiplications are typically indicative of special nuclear material (SNM). Multiplication measurements are typically performed using large assemblies of 3He detectors. While effective, these 3He detectors have several disadvantages. First, they are increasingly expensive due to the shortfall in the supply of 3He. Second, 3He detectors rely on the detection of thermalized neutrons which obscures potentially useful timing and energy information from the source. Liquid scintillator detectors can be used to directly measure the gamma-ray and fast neutron spectrum from a source. The nanosecond time response of liquid scintillator detectors is much faster
n
Corresponding author. Tel.: +1 443 778 5176. E-mail address: [email protected] (E.C. Miller).
0168-9002/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.nima.2013.06.062
than the microsecond response for a 3He detector. This fast timing allows for the analysis of detected particles on the time scale of individual fission events and not over the combined response of multiple fissions. This additional information could be potentially used to develop a detection system that can offer better information about the source in question. Past work has demonstrated that cross-correlation measurements with liquid scintillator detectors can be used to characterize source material [1]. Cross-correlation measurements identify time-correlated events using multiple detectors. Using pulseshape discrimination (PSD) subtle differences in the shape of the pulses produced by neutrons and gamma-ray interactions can be identified and the detected event can be attributed to the interacting particle type. This allows the time-correlated events to be classified by interacting particle: neutron–neutron (n–n), neutron– gamma (n–γ), gamma-neutron (γ–n), or gamma–gamma (γ–γ). The ability to classify the types of correlated events can yield additional information that can be used to characterize an unknown sample. Time-correlated pulse-height (TCPH) is an expansion of these cross-correlation measurements that include pulse-height information
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
in the analysis of the γ–n events. A γ–n event is defined as the arrival of a gamma-ray in an arbitrarily defined “start” detector followed by the detection of a neutron event in a second “stop” detector. Using the time difference between the events (stop-start), and the pulse-height information from the arriving neutron, a surface plot of pulse height at specific arrival times can be created. This surface can be used to identify events from fission chains. If the gamma and neutron contributing to a γ–n event are released from the same fission event, the expected arrival time, t (s), for a neutron with energy En (MeV), can be determined using d d t ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi − 2En =M n c
ð1Þ
where d (m) is the source–detector distance, Mn (MeV) is the mass of the neutron, and c (m/s) is the speed of light. The maximum amount of light generated in the detector can be estimated for a neutron of a given energy using a detector specific energy-to-light conversion shown in Eq. 2 [2]. L ¼ aEp −bð1−expðcEp ÞÞ
ð2Þ
where L (MeVee) is the light out, Ep (MeV) is the energy deposited by the incident neutron, and a, b, and c are detector specific coefficients. The coefficients used for this paper were: a ¼0.817, b¼ 2.63 and c ¼−0.297 [3]. Using Eqs. 1 and 2 the maximum light produced at a given arrival time produces a discrimination line for the TCPH distribution. All true correlated, non-multiplying source events must fall below this discrimination line. However, if the measured material has some level of multiplication, it becomes possible to detect correlated events from the same fission chain. Neutrons from later generation events in the fission chain can be correlated with earlier generation gamma-rays. This effect results in the possibility of detecting truly correlated events that arrive at greater times and generate more light than predicted by the discrimination line. By characterizing the amount of events past the discrimination line it may be possible to estimate the level of multiplication in a measured material. The TCPH technique was modeled extensively using the Monte Carlo code MCNPX-PoliMi. MCNPX-PoliMi is an enhanced version of MCNPX that allows for the simulation of correlated neutrons and gamma-rays [4,5]. The MCNPX-PoliMi post processing code MPPost was used to simulate the detector response to the particle transport data [6,7].
2. Experiment and analysis A series of measurements were performed at the Joint Research Center in Ispra, Italy. The campaign focused on measurements of mixed oxide (MOX) powder and plutonium–gallium (PuGa) disks to evaluate the sensitivity of the TCPH method to low multiplication samples. Each measurement conjuration was later simulated to validate the capabilities of MCNPX-PoliMi and MPPost. 2.1. Setup This measurement campaign used four EJ-309 liquid scintillator detectors measuring 7.62 cm in diameter by 7.62-cm long. The signals from the detector cells were digitized using a CAEN DT5720 digitizer with University of Michigan (UM) developed acquisition software. The DT5720 is a 12-bit, 250 MHz, 4-channel digitizer that can transfer data to a computer via a USB connection [8]. This allows for data collection with a laptop, greatly improving portability of the acquisition system. The measurement system was set up on two tables that were approximately 75 cm from a concrete floor. The tables were
109
Table 1 The composition of the MOX canister as of April 2012 Isotope
Mass (g)
Weight Percent
234
0.05 4.79 0.05 670.50 0.24 111.81 47.00 1.67 3.38 5.12 166.22 1010.83
0.0001 0.0047 0.0001 0.6633 0.0002 0.1106 0.0465 0.0017 0.0033 0.0051 0.1644 1.0000
U U 236 U 238 U 238 Pu 239 Pu 240 Pu 241 Pu 242 Pu 241 Am O2 Total 235
constructed of 2.5-cm by 2.5-cm aluminum rods with a 4-mm thick aluminum surface. The four detectors were arranged in an arc with a 40-cm source–detector distance. The average spacing between the detector-centerlines was 15 cm. 2.2. MOX source The MOX powder source was a 1.011-kg powder with an estimated density of 0.7 g/cm3. The MOX powder was encased in a cylindrical canister that is 32.2 cm tall with a 10.4 cm diameter. The estimated MOX density was determined by using the wellknown mass and the volume of the storage container. While this estimated value is lower than typical powder samples, it is believed to be close to correct based on profile measurements of the source. The source information is presented in Table 1 [9]. The aged source intensity for the MOX sample was 8.22 104 neutrons per second. A majority of the source neutrons, 59%, come from the spontaneous fission of 240Pu. The remaining neutrons come from the spontaneous fission of 242Pu and a variety of (α, n) reactions. For the MOX measurements, a thin, 1-cm shell of lead was placed directly around the canister. Some amount of lead shielding was required for this measurement because the gamma-ray count rate, without the lead, was sufficient to overload the data acquisition setup at the 50-keVee measurement threshold. However, adding too much lead will negatively impact the ability to detect the correlated fission gamma-rays required for the TCPH technique. The 1-cm thickness was chosen in an attempt to balance these two variables. This represents an inherent challenge when for the TCPH technique, detecting correlated gamma-rays and neutrons while not overloading the detection system. Two measurements of the MOX source (ENEA-1) were performed, in a reflected (polyethylene added around the lead shielding) and unreflected (only lead shielding) configuration. 2.2.1. Unreflected measurement The MOX source is interesting because the material has some level of multiplication. The keff for the unreflected MOX source was estimated using MCNPX-PoliMi to be 0.014 which results in a source multiplication of 1.014 where the multiplication is defined as M¼
1 1−kef f
ð3Þ
The unreflected MOX canister was measured overnight for a total of 9.66 h. A picture of the measurement setup and the simulated geometry is shown in Fig. 1. The simulation did not include the detector PMTs, detector stands, or the stand used to support the MOX canister. These omissions will have a negligible effect on the results.
110
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
Fig. 1. The measurement setup for the 40-cm measurement of the unreflected MOX canister and the modeled geometry.
Fig. 2. TCPH distributions showing the log of the number of counts per second, (top) the raw measured γ–n spectrum showing the background radiation in the negative direction, (bottom) the true measured spectrum with the background removed.
An example of a TCPH distribution is shown in Fig. 2. The y-axis shows the pulse height of the neutron in MeVee. The temperature plot represents the log of counts-per-second. The x-axis represents the time of flight determined using the arrival time of neutron in the “stop” detector and the arrival of a γ in the “start” detector. The negative time values arise because the γ–n pairing is defined by which detector the particles arrive in. If a gamma-ray is detected in the start detector and the neutron is detected in the stop detector, a positive time difference is recorded. If the neutron arrives in the start detector and the gamma-ray in the stop detector then the time difference will be negative. In an effort to reduce the intense gammaray background produced by the MOX powder, the background distribution was subtracted from the TCPH distributions, as shown in Fig. 2. The accidental background was determined by taking the average of the γ–n correlations in the negative time direction and subtracting this from the TCPH distribution. This background subtraction was applied to all of the TCPH distributions in shown. Fig. 3 shows a comparison between a measured and simulated TCPH distribution. The solid black line overlaid on the distribution represents the maximum pulse height for a neutron arriving at the
front face of the detector determined using Eqs. 1 and 2. The dashed line represents the maximum pulse height for a neutron reaching the back face of the detector. This convention is followed for all TCPH figures shown in this paper. For a non-multiplying point source all non-accidental events detected should arrive below these two discrimination lines. The simulation has an excellent agreement with the measured distributions considering that this response requires that the simulation correctly predict the energy and arrival times with very fine resolution. The main observable difference between the two results can be found at very short-times and low-light values, where there are increased events in the measured distribution. This concentration of events is the result of misclassified particles, where the PSD incorrectly identified an arriving gamma-ray as a neutron. This is not unexpected, when using a traditional PSD approach it is especially difficult to distinguish neutrons and gamma-rays with low-energy. However, using arrival time information, misclassified gamma-ray events can be identified. The results in Fig. 3 show that there is a noticeable concentration of events that are beginning to move past the discrimination
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
111
Fig. 3. TCPH distributions for the unreflected MOX source showing the log of counts per second, (a) measured and (b) simulated. The solid overlaid line represents the maximum pulse height for a neutron arriving at the front face of the detector. The dashed line represents the maximum pulse-height for a neutron reaching the back face of the detector.
101 Measured MCNPX-PoliMi 100 Counts per Second
lines. This effect is partially due to the size of the extended source, and complicates the identification of late arriving neutrons. To provide a direct comparison between the measurement and simulated TCPH distributions the results were projected on the y-axis to produce a pulse-height distribution (PHD) and the projection on the x-axis produces a time-of-flight (TOF). These distributions are shown in Fig. 4. The error was calculated in two ways. First the total percent difference between the simulated and measured results was 10.56%. This value helps show the level of agreement between the source strength and geometry (i.e. isotopic composition, solid angle). The second quantity, the average point-by-point error identifies how well the transport physics and energy-to-light conversion of particles in the detector are handled. These errors are higher, but are also influenced by the total percent difference. For this case the average point-by-point errors are, 17.53% for the PHD and 19.96% for the TOF.
10-1
10-2
10-3 0
1 Light (MeVee)
1.5
2
0.05 Measured MCNPX-PoliMi 0.04 Counts per Second
2.2.2. Reflected measurement A polyethylene reflector was added to increase the multiplication of the MOX canister; MCNPX-PoliMi simulations for this configuration predicted a keff of 0.08 and a multiplication of 1.087. This measurement was performed overnight and a total of 14.9 h of data was collected. The polyethylene reflector was constructed in a “U” shape around the MOX cylinder. The reflector was used to increase the level of multiplication in the source without shielding the detected signal. The MOX canister was surrounded by 1 cm of lead and was raised from the table surface by 7 cm of polyethylene. The reflector was constructed of polyethylene slabs 60 8.0 2 cm. The measurement setup is displayed in Fig. 5 with the relevant dimensions for the polyethylene structure. The TCPH distributions are compared in Fig. 6. A background subtraction was performed on the measured data. There is very good agreement in the shapes of the distributions. There is a slight increase in the number of events falling above the discrimination lines than was observed in the previous case. As with the unreflected measurement, there is a large concentration of misclassified events in the measured distribution at short-time and low-light values. The total percent error for this measurement setup was −2.59%. The average errors were 16.88% for the PHD and 22.16% for the TOF. In both simulations of the MOX canister the peak value of the TOF distribution is slightly under predicted. This effect could be caused by a small change in the source–detector distance, either in the detector placements or as a result of an internal structure in the MOX sample.
0.5
0.03
0.02
0.01
0 0
10
20 30 Delta Time (ns)
40
50
Fig. 4. (a) PHD comparison for the MOX distribution and (b) TOF comparison for the MOX distribution.
2.3. PuGa source Three PuGa metal disks were measured together in both an unreflected and reflected configuration. There were several PuGa disks available that ranged in mass from 0.01 g to 9.81 g of plutonium. For this measurement, the three largest samples were used, accounting for 86% of the available plutonium mass. All of
112
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
Fig. 5. The reflected MOX measurement setup and polyethylene dimensions.
Fig. 6. TCPH comparison for the reflected MOX measurement, (a) measured and (b) simulated. The solid overlaid line represents the maximum pulse height for a neutron arriving at the front face of the detector. The dashed line represents the maximum pulse-height for a neutron reaching the back face of the detector.
Table 2 Isotopic composition and masses for the three PuGa samples measured Isotope
238
Pu Pu Pu 241 Pu 242 Pu 241 Am Ga Total 239 240
Mass (g)
Weight %
Disk 209
Disk 210
Disk 211
0.003 1.416 0.402 0.037 0.014 0.035 0.035 1.941
0.006 3.627 1.030 0.094 0.037 0.089 0.089 4.972
0.013 7.154 2.032 0.184 0.072 0.176 0.176 9.808
0.001 0.729 0.207 0.019 0.007 0.018 0.018
the disks had an identical source composition, shown in Table 2. The samples were 73% 239Pu by total mass. Combined, the three disks had a source intensity of only 3789 neutrons per second. The spontaneous fission of 240Pu, accounts for 93% of all source neutrons created. The next largest contribution comes from the spontaneous fission of 242Pu. Additional neutrons will be created from induced fission events in the 239Pu.
2.3.1. Unreflected measurement For the unreflected measurement the PuGa disks were placed in a Plexiglas stand that held the disks vertically. The large number of gamma-rays emitted from these disks was enough to overload the maximum data transfer rate of the data acquisition system. To reduce the count rate, a thin (2.5 mm) sheet of lead was added in front of the PuGa sources. Even with the lead shield the detected neutron/gamma-ray ratio was 0.0036.
The setup for the unreflected PuGa measurement is shown in Fig. 7. As with previous simulations the PMTs and source stands were not included. This measurement setup had a very low multiplication of 1.05 and a simulated keff of 0.0476. The measured and simulated TCPH distributions are shown in Fig. 8. There is very good agreement for the two distributions. Again the large amount of misclassified events is observed at short-times and low-light. This misclassification is exceptionally pronounced in this measurement because of the large amount of gamma-rays emitted from the source. There is an interesting feature in the simulated distribution where there is a small concentration of low-energy late-time events that are not seen in the measured result. These events will be from neutrons scattering in the near-environment and arriving in the detector. The high gamma-background obscures this feature in the measured result. Fig. 9 shows the comparison of the PHD and TOF. With the strong gamma-ray source even a very good PSD discrimination ratio will begin to have a significant amount of misclassified events. Fortunately, for this analysis these events can be removed by time-of-flight data. To remove these misclassified events and obtain a “true” measurement distribution, timing information was used to remove neutrons that arrive at unphysical times. The “true” measurement data was then compared to the simulated results. The total percent difference was −11.56%, and the average pointby-point error for the PHD and TOF were 33.29% and 33.94%, respectively. There is a clear shift in the measured TOF distribution compared to the simulated result. This shift is more significant than the one that was observed in the MOX case. The shift in the position of the TOF distribution is likely caused by a small shift in
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
113
Fig. 7. A photograph of the unreflected PuGa measurement showing the Plexiglas holder with the thin lead compared to the MCNPX-PoliMi simulated geometry.
Fig. 8. TCPH distributions for the unreflected PuGa source, (a) measured and (b) simulated. The solid overlaid line represents the maximum pulse height for a neutron arriving at the front face of the detector. The dashed line represents the maximum pulse-height for a neutron reaching the back face of the detector.
the actual source–detector distance for the four detectors compared to the simulation. It would be more effective in future measurements to have the detectors in a fixed assembly to more easily control the source–detector distance. 2.3.2. Reflected measurement The PuGa disks were also measured in a reflected configuration. To increase the overall multiplication of the measured system a polyethylene structure was built around the Plexiglas source holder. The polyethylene structure was 60 cm long, 35 cm tall, and 16 cm thick. The addition of the polyethylene increased the multiplication from 1.05 to 1.068. The source–detector distance remained 40 cm. The measurement setup is shown with the simulated geometry in Fig. 10. A 2.5 mm sheet of lead was again required to reduce the number of detected gamma-rays. Even with the lead shielding the detected neutron-to-gamma ratio was 0.004. Fig. 11 shows good agreement between the measured and simulated TCPH distributions. There are a small number of events past the discrimination line in the simulation that are not observed in the measurement. These low-count fringe events were obscured by the background in the measured data and were ultimately removed when the background subtraction was applied. The other observable difference is the misclassified gamma-ray events, clearly visible in the measured result at lowlight and short-times. For this case the total percent difference was 8.56%. The average point-by-point error for the PHD is 30.37% and 30.87% for the TOF.
are summed together creating a cumulative region integral (CRI) distribution. This approach measures the gradient of the distribution. Accidental events should influence the entire distribution evenly and should not have a significant impact on the gradient of the distribution. The CRI distribution results for the Ispra measurement are compared in Fig. 13. As the source multiplication increases the gradient over the non-multiplying discrimination line becomes increasingly flat. As a result, the CRI distribution will have a decreasing slope with increasing multiplication. This trend can be observed for the 252Cf and PuGa results shown in Fig. 13. The CRI distribution of the 252Cf source, with a multiplication of 1, is farthest to the left. As expected, the unreflected and reflected PuGa CRI distributions are located to the right of the 252Cf case. The two MOX CRI distributions appear with a much lower slope than expected compared to the other cases; however, they are positioned correctly relative to each other. This change in the slope of the MOX cases is the result of the extended source geometry. The extended source obscures the source–detector distance, decreasing the gradient around the discrimination line. The CRI distributions for all of the Ispra measurements were simulated with MCNPX-PoliMi, normalized to the integral number of counts, and compared to the measured results. The average point-by-point error for all of the Ispra cases is shown in Table 3. All of the cases agree within 3% for both the total error and average point-by-point error. These results show that MCNPX-PoliMi can accurately predict the shape of the CRI distribution. 3.1. Simulation of highly multiplying samples
3. Characterization of source multiplication The objective of the TCPH method is to provide a means for estimating the multiplication of an unknown sample. To approach this objective, multiple counting regions, evenly spaced, starting at the front face of the detector, are used. Fig. 12 shows the regions applied to the unreflected MOX source. All events in each region
All of the samples measured in the Ispra campaign have very low levels of multiplication. To examine the effectiveness of this characterization technique for sources with larger multiplication values the CRI discrimination region approach was applied to models of a reflected 4.5-kg plutonium sphere with 94% 239Pu and 6% 240Pu. The reflected plutonium sphere setup was chosen because it has been successfully modeled with MCNPX-PoliMi in
114
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
3.2. Estimating an unknown source
101 Measured Measured ("True" Events) MCNPX-PoliMi
Counts per second
100
10-1
10-2
10-3
10-4 0
8
0.5
1 1.5 Pulse Height (MeVee)
2
x 10-3 Measured Measured ("True" Events) MCNPX-PoliMi
7
Counts per second
6 5 4 3 2 1 0
0
10
20 30 Time of Flight (ns)
40
50
Fig. 9. A comparison of the unreflected PuGa measured data to simulated results. The “true” measured results have misclassified neutrons removed by timing. (a) PHD and (b) TOF.
previous work [10,11]. A complete description of the source can be found in the paper by Mattingly [11]. The simulation of the plutonium sphere measurement used two 12.7-cm diameter by 12.7-cm long EJ-309 scintillator cells placed side-by-side. The center of the plutonium sphere was placed 50 cm from the front face of the detectors. The plutonium sphere was surrounded by up to 15.24 cm of polyethylene reflectors to increase the multiplication. The table and floor were also included in the model. As the level of multiplication increases, the TCPH distribution becomes increasingly more in level in the region around the theoretical discrimination line. To clearly observe the effects of highly multiplying samples the number of regions must be increased until the edge of the distribution is found. The CRI distributions fall in the expected pattern when the number of regions is set to 250. Fig. 14 shows a clearly increasing trend with multiplication. The CRI distribution for the unreflected plutonium sphere case shown in Fig. 14 has a distinct bump around region 50. This feature is the result of neutrons scattering off of the floor reaching the detector. This effect is obscured in the reflected cases as more true coincident events begin to fall in these regions.
The CRI distributions provide a visual comparison of the level of multiplication in a measured sample, but make quantitative comparisons difficult. To provide a direct method of comparison the integral of the CRI distribution was taken. Fig. 15 shows a comparison of the CRI integrals for three sets of data. The first set of data is the results from the Ispra measurements of MOX powder and PuGa. The second set is the simulated reflected plutonium sphere cases. Third, an additional set of results for a plutonium sphere with varying radii was added to provide a baseline reference for multiplications over a reasonable range without other complicating factors. To create the data set for the plutonium sphere with a varying radius the plutonium sphere used the same model that was used in the reflected case with radii ranging from 2 cm to 4.8 cm. The radii investigated correspond to plutonium masses ranging from 0.6 kg to 9.1 kg of plutonium. This provides data points both above and below the 5 kg plutonium mass the IAEA defines as a significant quantity. The error bars on the figure represent the counting error for each case. It should be noted that the large errors bars for the Ispra data are the result of poor counting statistics limited by the available measurement times. The error for the plutonium source of varying radii increases as multiplication decreases because the same number of source events were simulated in each case. The samples with large multiplications resulted in many induced fission events resulting more detected events, reducing the error. When then CRI integrals for all three data sets are compared and plotted against the multiplication, a weak trend is apparent: as the level of multiplication increases, the area under the curve decreases. However, the level of sensitivity is very low. Very small changes in the CRI integral can result in significant changes in the predicted amount of multiplication. The large error bars resulting from poor counting statistics in the PuGa measurements correspond to a dramatic range of potential multiplications. This result shows that extracting multiplication data using the CRI integral method is not sensitive enough for the identification of an unknown source. However, this result does show that a trend exists and it may be possible, with a different approach, to extract more sensitive multiplication information from the TCPH distributions.
4. Conclusions TCPH measurements of low-multiplying source materials were successfully performed at the JRC in Ispra, Italy. The multiplications of the measured MOX and PuGa samples ranged from 1.05 to 1.087. The TCPH method was unable to clearly differentiate the source multiplications over this narrow range. However, these results were essential in validating the simulation capabilities of MCNPX-PoliMi and MPPost. The simulated results were within −11.56% of the measured values. With the simulation tools validated, the TCPH analysis was extended to simulate multiplying samples. Using MCNPX-PoliMi a highly multiplying 4.5-kg plutonium sphere was analyzed using TCPH. In an attempt to characterize the level of multiplication in the sample, a multi-region approach was introduced. CRI distributions for the Ispra measurements agreed within 3% of the simulated results. Using this validation the CRI simulations, this approach was applied to simulations of the plutonium sphere. These results were used to fit a line to the behavior of the CRI values over a range of multiplications with the intent of establishing a means of determining an unknown source multiplication. While
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
115
Fig. 10. A photograph of the reflected PuGa measurement showing the polyethylene structure compared to the simulated MCNPX-PoliMi geometry.
Fig. 11. The reflected PuGa TCPH distributions for the (a) measured data and (b) for the simulated results. The solid overlaid line represents the maximum pulse height for a neutron arriving at the front face of the detector. The dashed line represents the maximum pulse-height for a neutron reaching the back face of the detector.
1 0.9 Cf-252 PuGa (bare) PuGa (mod) MOX (bare) MOX (mod)
Normalized Counts
0.8 0.7 0.6
Increasing Multiplication
0.5 0.4 0.3 0.2 0.1 0 0
Fig. 12. A TCPH for the unreflected MOX case with 20 dividing regions (dashed lines) used to evaluate the level of multiplication.
the measurements trend weakly with source multiplication the CRI approach is not sensitive enough to the multiplication of the source to be used for the identification of an unknown multiplication. This measurement campaign successfully identified several problems with applying the TCPH approach in the field. First, the high gamma-ray rates from multiplying sources. The high gammaray rates create an extremely high accidental background that makes identifying late arriving neutrons difficult. The low neutron-to-gamma-ray ratio also negatively impacts the ability to do PSD by increasing the amount of misclassified events. Lastly, high gamma-ray backgrounds require adjustments to the measurement design to ensure that the digitizing setup is not
5
10 Region
15
20
Fig. 13. CRI distributions for the Ispra measurements.
Table 3 The percent difference between the measured and simulated CRI distributions
252
Cf MOX (unreflected) MOX (reflected) PuGa (unreflected) PuGa (reflected)
Percent difference (total)
Percent difference (average)
−0.45 −2.81 −1.89 0.75 0.33
0.57 −2.97 −1.66 2.89 2.71
overloaded. This problem can be mitigated by increasing the threshold of the measurement system, thereby reducing the lowenergy events that are more frequently misclassified and reduce the overall count rate.
116
E.C. Miller et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 108–116
TCPH distributions change in shape with different sources and geometries. Template matching the TCPH distribution to a library of sources and configurations may be a way to improve the applicability of this technique. These problems are not insurmountable, and with additional development it is possible that the TCPH technique could be used to identify the multiplication of an unknown source. However, significant development progress will be needed to address the issues identified in this measurement campaign.
1 0.9
Normalized Counts
0.8 0.7 0.6 0.5 0.4
Pu Sphere (bare) Pu Sphere (1.27 cm poly) Pu Sphere (2.54 cm poly) Pu Sphere (3.81 cm poly) Pu Sphere (7.62 cm poly) Pu Sphere (15.24 cm poly)
0.3 0.2 0.1
Acknowledgments
0 0
50
100
150 Region
200
250
Fig. 14. CRI distributions for the plutonium sphere using 250 regions to clearly resolve the increasing multiplication of the simulated samples.
Ispra Reflected Pu Sphere Different Radius Pu Sphere
290 270
CRI Integral
250
The authors would like to thank Santino Frison for his assistance in completing the measurements. This research was funded in part by the Nuclear Forensics Graduate Fellowship Program which is sponsored by the U.S. Department of Homeland Security's Domestic Nuclear Detection Office and the U.S. Department of Defense's Defense Threat Reduction Agency. This research was also funded by the National Science Foundation and the Domestic Nuclear Detection Office of the Department of Homeland Security through the Academic Research Initiative Award # CMMI 0938909 Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-AC04-94AL85000. SAND 2013-0690J
230
References
210 190 170 150 0
5
10 Multiplication
15
20
Fig. 15. CRI integral values for the Ispra measurements, reflected plutonium spheres, and plutonium spheres of varying radii.
Second, measurements of extended sources at close range, such as the MOX canister, introduce significant geometry effects to the TCPH distribution. It is difficult to balance the need for a short counting time and the need to be far enough from the source to have a point-like behavior. Third, the classification of the source multiplication is difficult. The placements of the discrimination lines are sensitive to distance, and so this value needs to be well known, and controlled. Using the gradient of events past the discrimination line (the CRI method) is not sensitive to the changes in multiplication. A more advanced approach will be needed. This work did show that the
[1] E.C. Miller, J.L. Dolan, S.D. Clarke, M. Flaska, S.A. Pozzi, E. Padovani, P. Peerani, P. Schillebeeckx, Experiments and simulation of cross-correlations on MOX fuel, in: Proceedings of the Institute of Nuclear Materials Management Annual Meeting, Palm Desert, CA, USA, July 17–22, 2011. [2] A. Enqvist, C.C. Lawrence, T.N. Massey, S.A. Pozzi, Neutron light output functions measured for EJ309 liquid scintillation detectors,in: Proceedings of INMM Annual Meeting, Orlando, FL, July 15–19, 2012. [3] A. Enqvist, C.C. Lawrence, B.M. Wieger, S.A. Pozzi, T.N. Massey, Nuclear Instruments and Methods A 715 (2013) 79. [4] E. Padovani, S.A. Pozzi, S.D. Clarke, E.C. Miller, MCNPX-PoliMi, RSICC Package number C00791 MNYCP 00, 〈http:/www.rsicc.ornl.gov〉. [5] S.A. Pozzi, S.D. Clarke, W. Walsh, E.C. Miller, J. Dolan, M. Flaska, B.M. Wieger, A. Enqvist, E. Padovani, J.K. Mattingly, D. Chichester, P. Peerani, Nuclear Instruments and Methods in Physics Research A (2012) 119. [6] E.C. Miller, A. Poitrasson-Riviere, A. Enqvist, J.L. Dolan, S. Prasad, M.M. Bourne, K. Weinfurther, S.D. Clarke, M. Flaska, S.A. Pozzi, E. Padovani, J.K. Mattingly, MCNPX-PoliMi Post-Processor (MPPost) Manual, RSICC Package number C00791 MNYCP 00, 〈www.rsicc.ornl.gov〉. [7] E.C. Miller, S.D. Clarke, S. Prasad, M. Flaska, S.A. Pozzi, E. Padovani, Journal of Nuclear Materials Management XL (2) (2012). [8] Technical Information Manual, MOD. DT5720, Revision n.5, CAEN, 〈www.caen. it〉, February 2012. [9] P. Peerani, M. Swinhoe, A.L. Weber, L.G. Evans, ESARDA Multiplicity Benchmark Exercise Phases III and IV, Esarda Bulletin 42 (2009). [10] E.C. Miller, B. Dennis, S.D. Clarke, S.A. Pozzi, J.K. Mattingly, Nuclear Instruments and Methods in Physics Research A 652 (2011) 540. [11] J. Mattingly, Polyethylene-Reflected Plutonium Metal Sphere: Subcritical Neutron and Gamma Measurements, SAND2009-5804-R2, Sandia National Laboratories 2009.