Initial Evaluation for a Combined Neutron and Gamma1

Ray Multiplicity Counter2

Andreas Enqvista,∗, Sara A. Pozzib, Marek Flaskab, Imre Pazsita3

aDepartment of Nuclear Engineering, Chalmers University of Technology,4

SE-412 96 Goteborg, Sweden5

bDepartment of Nuclear Engineering and Radiological Sciences, University of Michigan,6

Ann Arbor, MI 48109, USA7

Abstract8

Multiplicity counters for neutron assay have been extensively used in ma-9

terials control and accountability for nonproliferation and nuclear safeguards.10

Typically, neutron coincidence counters are utilized in these fields. In this pa-11

per we present a measurement system that makes use not only of neutron (n)12

multiplicity counting but also of gamma ray (γ) multiplicity counting and the13

combined higher-order multiplets containing both neutrons and gamma rays.14

The benefit of this approach is in using both particle types available from15

the sample, leading to a reduction in measurement time when using more16

measurables. We present measurement results of n, γ, nn, nγ, γγ, nnn, nnγ,17

nγγ and γγγ multiplets emitted by a 252Cf source and a 239Pu-Be source. A18

dual radiation measuring system proposed here could use extra measurables19

either to improve the statistics when compared to a neutron-only system,20

or alternatively be used for extended analysis and interpretation of sample21

parameters.22

The study presented here provides an answer to the questions regarding23

the ability to measure both neutron and gamma ray coincidences at once.24

The results show that the measurements system proposed in this study has a25

potential to be valuable in the area of nuclear nonproliferation and homeland26

security.27

Key words: nuclear nonproliferation, pulse shape discrimination, liquid28

scintillator detectors, multiplicity counting, coincidences.29

∗Corresponding author.Email addresses: andreas@nephy.chalmers.se (Andreas Enqvist),

pozzisa@umich.edu (Sara A. Pozzi), mflaska@umich.edu (Marek Flaska),imre@chalmers.se (Imre Pazsit)Preprint submitted to Nuclear Instr. & Meth. A May 11, 2010

1. Introduction30

Neutron multiplicity counters have been used for several decades, mostly31

as thermal multiplicity counters based on 3He detectors [1] or as counters32

with slightly wider neutron energy ranges [2]. These systems have very good33

detection efficiency; however, the time resolution suffers from the required34

moderation of the neutrons. A number of new measurement options have35

recently become available with the advancements in applying organic liquid36

scintillation detectors within the areas of nuclear nonproliferation and mate-37

rials control and accountability and with the developments in fast digitizer38

technology.39

Organic scintillation detectors can provide some energy spectroscopy [3]40

as well as excellent die-away properties, thus reducing the number of ac-41

cidentals in measurements. Furthermore, the pulse shape can be used to42

discriminate between different types of incoming radiation. For example43

neutron pulses can be separated out from the gamma ray pulses. This means44

that data analysis techniques such as neutron spectrum unfolding from the45

measured pulse height distributions (PHDs) [4, 5] can be applied, even in46

the presence of gamma ray radiation.47

Recently, liquid scintillation detectors have been investigated for use in48

neutron multiplicity counters [6]. The main benefit of such a design is that it49

is much less sensitive to accidental coincidences than the traditional approach50

based on 3He detectors. Thus, the liquid scintillators are suitable for use51

even on samples with high single-neutron background, such as (α, n) rates52

exhibited in some samples containing the isotopes of plutonium. However,53

Frame et al. [6] estimated that such a system will have problems with neutron54

classification at energies below 2 MeV, as well as having a low total detection55

efficiency. Such problems might be corrected with alternative designs of the56

detector. It can be noted that all gamma ray pulses would be ignored in this57

type of design.58

Fission multiplicity counters [7] use systems of plastic scintillation detec-59

tors sensitive to both fast neutrons and gamma rays. However, these coun-60

ters offer no possibility of discriminating the gamma rays from the neutrons,61

thus only the total fission multiplicity is measured, rather than a neutron62

or gamma ray multiplicity. More recent developments, especially those using63

the Nuclear Materials Identification System [8], allowed temporal discrimina-64

tion of gamma rays from neutrons to further enhance the analysis of nuclear65

materials.66

2

We suggest a novel approach for finding sample characteristics based on67

the use of multiple liquid scintillators, a fast digitizing system, and advanced68

analysis algorithms. The proposed system uses pulse shape discrimination69

(PSD) [9] to determine what particle caused the detection pulse. The excel-70

lent timing properties of liquid scintillators are used to find out what pulses71

are correlated in time and the particle types that created them. This ap-72

proach has the benefit of providing more measurables compared to the case73

of pure neutron multiplicity counting, while having a few extra parameters -74

but not more than the amount of added measurables - which could be used75

in a number of ways to determine the sample characteristics [10, 11, 12].76

This paper describes the measurement system used for performing multi-77

plicity measurements using neutron and gamma ray sources such as 252Cf and78

239Pu-Be. The approach builds upon previously described cross-correlation79

measurements [13, 14]. The PSD will be described in detail as one of the80

key parameters for the investigation of the proposed approach. The results81

of initial assessment will then be discussed and future possible improvements82

along with potential drawbacks of the proposed system.83

2. Description of Experimental Setup84

The measurements were performed at the University of Michigan, Depart-85

ment of Nuclear Engineering and Radiological Sciences - where the measure-86

ment system was developed - in July 2009. The measurement system consists87

of four EJ-309 cylindrical scintillation detectors with a height of 13.3 cm and88

a diameter of 13 cm. This liquid has a high flash point (144 ◦C), and low89

chemical toxicity.90

In the measurements (Fig. 1), the detectors were placed at 90 degrees91

with the source-detector distance kept constant. The data acquisition is92

performed with a 250-MHz, 12-bit, 8-channel CAEN V1720 digitizer. Four93

channels were used, each being able to trigger independently on the incoming94

detector pulses. The data from all channels are stored and then subsequently95

analyzed to determine if any additional detectors, in addition to the trigger96

detector, contain time-correlated pulses. The detectors were calibrated using97

a 137Cs source.98

A threshold of approximately 70 keVee corresponding to a neutron energy99

deposited of approximately 450 keV, was applied to achieve excellent PSD100

performance. The 70-keVee threshold was chosen based on previous investi-101

gations focused on the accuracy of particle discrimination. At this threshold,102

3

Figure 1: Measurement setup.

a gamma rejection of approximately 1/1000 is expected. The 252Cf source103

was placed at different detector distances to determine the effect of changing104

absolute detector efficiency, while the Pu-Be source was only used at a larger105

distance with 5 cm lead shielding to suppress high gamma ray production.106

The count rates when using all four detectors varied from approximately107

4.5 kHz to 15 kHz, with the background being approximately 1 kHz. The108

measurement setups were as follows:109

•252Cf, 15 cm source–detector distance,110

•252Cf, 20 cm source–detector distance,111

•252Cf, 30 cm source–detector distance,112

• Pu-Be shielded with 5 cm of lead, 50 cm source–detector distance.113

The measurement times varied between 10 and 30 minutes resulting in ap-114

proximately 10 million recorded events per setup.115

3. Pulse Shape Discrimination and Pulse Timing116

PSD and pulse timing are performed with an offline algorithm. Tradi-117

tionally, PSD can be performed based on charge integration or differentiation118

[15, 16]. In previous work, we have used charge integration and analyzed the119

tail-to-total integral ratio versus pulse height [14]. In this work, the total120

4

integral is calculated from the start of the pulse to 220 ns past the pulse121

maximum, while the tail integral is calculated from 20 ns to 220 ns past the122

pulse maximum. These values were previously determined by analyzing the123

figure of merit for the PSD as a function of the time ranges for the pulse124

integration.125

A linear discrimination boundary is used to discriminate neutrons from126

gamma rays (Fig. 2). This method show high accuracy, but can unfor-127

tunately not be quantified with simple figures of merit (FOM) due to the128

nonlinear discrimination curve chosen. The FOM is defined as follows: FOM129

= distance between peaks of neutron and gamma distributions / (FWHMγ130

+ FWHMneutron)131

Pulse timing is performed on the digitized pulse data by simulating a132

constant fraction discriminator where the the start of the pulse is defined as133

20% of the peak value. The timing is then less sensitive to the pulse am-134

plitude. By using interpolation between the data values allows for increased135

time resolution from the initial 4 ns to approximately 1 ns. The sharp rise136

of the pulse edge means linear interpolation can be used without large losses137

of accuracy. The gamma ray–gamma ray peak in the cross-correlation mea-138

surements was used to test the timing accuracy and a FWHM of less than139

1.5 ns was observed.140

4. Neutron and Gamma Ray Multiplicities141

The benefit of the proposed system is its speed and robustness which is142

due to the fact that several quantities are measured (nine multiplets up to143

the third order). Further, the system operates on short time scales (300 ns144

is the width of a typical pulse) due to the removal of neutron moderation145

when compared to a 3He neutron multiplicity counter. While a 3He counter146

design usually encompasses the entire sample to maximize the neutron mod-147

eration and consequently the detection efficiency, a scintillation system could148

be made with a sparse layout and thus be a candidate for a portable system.149

Such a design requires that detection rates of multiplets are sufficiently high150

even with relatively low absolute detector efficiency. For allowing the unfold-151

ing of sample parameters such as proposed in [11], successful measurement152

systems for all the multiplets need to be devised.153

5

0 2 4 6 8 100

0.5

1

1.5

2

2.5

Total integral

Tai

l int

egra

l

(a) Full range.

0 0.2 0.4 0.6 0.8 1 1.20

0.1

0.2

0.3

0.4

0.5

Total integral

Tai

l int

egra

l

(b) Zoomed in.

Figure 2: Tail vs. total integrals for a 252Cf measurement. The data shown are for asingle EJ-309 detector, and the discrimination line indicates which pulses are created byneutrons (above the line) and by gamma rays (below the line).

6

0 100 200 300 400 500 600 700−0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Time (ns)

Am

plitu

de (

Vol

ts)

Pulse 1Pulse 2Pulse 3

Figure 3: Sample pulses from the data-acquisition system.

4.1. Measurement Results and Discussion154

The measurements were performed with a 252Cf source producing 19500155

fissions/second [17] and a shielded 1-Ci Pu-Be source [18]. All pulses that156

seem to be a real coincidence, considering time of flight for the neutrons and157

the total size of the system (up to approximately 50-ns difference), are fully158

registered with the digitizer in a window of 400 ns (100 sampling points).159

Table 1 shows the measured rates of multiplets for various source–detector160

distances from the 252Cf source. The results indicate that for measurement161

times in the order of a few minutes good accuracy for the singlet and doublet162

rates can be achieved.163

4.2. Measurements with 252Cf164

The rates of the different multiplets for the 252Cf source are shown in Fig.165

4. It is evident that the source–detector distance has a large impact on the166

higher-order multiplets, which increase by more than an order of magnitude167

when the distance is decreased from 30 cm to 15 cm. With reduced distance,168

the rate of the higher-order multiplets increases more than the total count169

rate at the largest distance. Optimal rates are considered those where the170

measurement uncertainty for the high-order multiplets is low (acceptable)171

for reasonably short measurement times. With detection rates for gamma172

rays being higher, this might indicate that the minimum detectable activity173

could be reduced when also accounting for gamma multiplets compared to174

pure neutron multiplicity.175

7

Count rates (s−1) and statistical errors:30cm ±1σ 20cm ±1σ 15cm ±1σ

Rn 690.4 0.63 1338.6 0.95 1947.4 1.36Rγ 3119.9 1.34 5335.6 1.90 7256.6 2.62Rnn 13.3 0.09 51.5 0.19 114.3 0.33Rnγ 52.3 0.17 197.3 0.37 409.3 0.62Rγγ 63.4 0.19 227.7 0.39 455.0 0.66Rnnn 0.1 0.01 0.8 0.02 3.0 0.05Rnnγ 0.6 0.02 5.5 0.06 18.7 0.13Rnγγ 1.5 0.03 12.0 0.09 38.7 0.19Rγγγ 1.2 0.03 8.8 0.08 27.0 0.16

Measurement time (s):T 1735.8 1475.3 1054.6

Table 1: 252Cf count rates for different source–detector distances and all possible neutron,gamma ray and mixed multiplets up to the third order. The statistical error is calculatedas the square root of the total counts divided by the total measurement time.

Ratios:30cm 20cm 15cm

Rn/Rnn 52.1 26.0 17.0Rn/Rnnn 6809.3 1669.3 658.2Rnn/Rnnn 130.7 64.2 38.6Rγ/Rγγ 49.2 23.4 15.9Rγ/Rγγγ 2646.8 606.1 268.4Rγγ/Rγγγ 53.7 25.9 16.8Rn/Rγ 0.22 0.25 0.27Rnn/Rγγ 0.21 0.23 0.25Rnnn/Rγγγ 0.09 0.09 0.11

Table 2: Different 252Cf ratios of the count rates could be used to obtain geometricalinformation.

8

n γ nn nγ γ γ nnn nnγ nγ γ γ γ γ10

−1

100

101

102

103

104

Multiplet

Rat

es

15cm20cm30cm

Figure 4: Detection rates of different multiplets at different distances for the 252Cf source.

Table 2 reports a number of the ratios of rates that could be of interest,176

such as ratios between different multiplets of neutrons, which show the strong177

dependence on absolute detector efficiency with varying distance. More intri-178

cate ratios such as those between the same multiplet of neutrons and gamma179

rays could be used for possible PSD error analysis (see below). In the mea-180

surements it has been noted that the neutron doublets are clearly depending181

on the orientation of the detectors, with opposing detectors recording more182

than 50% more doublets compared to adjacent detectors. This corresponds183

well with the findings in Ref. [19].184

The values of the ratios Rn/Rγ , Rnn/Rγγ , Rnnn/Rγγγ should be also em-185

phasized, because they are almost fully independent of the source–detector186

distance. However, they do change depending on if they are the singlet, dou-187

blet, or triplet neutron to gamma ray ratio, since the number of gamma-rays188

and neutrons produced per single fission significantly varies. Specifically, up189

to 20 gamma rays, 10 neutrons can be emitted from a single fission, which190

can significantly change the above mentioned ratios. With the spontaneous-191

fission distribution of different isotopes being different, this set of ratio values192

could be used for isotopic determination. The singlet ratios also depend on193

the existence of (α, n) events, which needs to be taken into account for the194

potential identification of the isotopes present in the measured sample.195

One particular character of the multiplets is the behaviour of their mag-196

9

nitude when neutrons are successively replaced by gamma photons in the197

same order of multiplet. This can be easily seen in Table 1 and especially198

in Fig. 4, since the values are arranged just in this order. For the doublets,199

starting with the pure neutron doublet, replacing a neutron with a gamma200

always leads to a higher value, although the increase is smaller in the second201

step than in the first. For the triplets, the change of the magnitude is not202

monotonic; the magnitude of the pure gamma triplet rate is smaller than203

that of the nγγ triplet rate. This shows the special value of the use of the204

mixed rates, as it was also mentioned in [11, 12].205

4.3. Measurements with Pu-Be206

Measurements were performed with a strong Pu-Be source, which emits207

one neutron per (α, n) reaction accompanied with one or more gamma rays.208

The source also emits a large number of uncorrelated gamma rays, which was209

the reason for placing 5 cm of lead shielding around the source. Measurement210

results are given in Table 3: as can be noted, very few multiplet above the first211

order are detected. Even nγ doublets which should have high frequency in212

connection to the (α, n) events, are severely suppressed by the Pb shielding.213

Table 3 shows that the singlets of neutrons and gamma rays are very dom-214

inating, while triplets are essentially non-existent due to the low number of215

correlated events. As expected, measured singlet-to-doublet and singlet-to-216

triplet ratios are much lower for the Pu-Be source than for the 252Cf source,217

which is caused by lack of time-correlated detection events. These quanti-218

ties could be used as a quick estimate of sample types, such as fissile versus219

non-fissile. In the case of the fissile 252Cf source, the triplet rates are approx-220

imately a factor 1000 lower than the singlet rates (see Rn/Rnnn and Rγ/Rγγγ221

in Table 2), while for the Pu-Be source, high-order correlated events are222

almost nonexistent, and occur a factor 1,000,000 times more rarely than a223

single event. In addition, other source properties such as ratios of gamma224

rays to neutrons for different radiation shields that would indicate the source225

type could be obtained. The multiplicity system discussed here could be226

‘calibrated’ to allow identification of various neutron and gamma ray sources227

and nuclear materials. The additional benefit of measuring also gamma rays228

could depend on the assayed material, making the approach more suitable229

to some types of measurements.230

10

PuBu source:Rate ±1σ

Rn 8160.3 3.2494Rγ 3283.3 2.0611Rnn 7.0647 0.0956Rnγ 10.571 0.1170Rγγ 3.9179 0.0712Rnnn 0.0013 0.0013Rnnγ 0.0116 0.0039Rnγγ 0.0078 0.0032Rγγγ 0.0103 0.0037

Table 3: Measured multiplet rates for the Pu-Be source. The measurement time was 772seconds.

n γ nn nγ γγ nnn nnγ nγγ γ γ γ10

−4

10−2

100

102

104

Multiplet

Rat

es

Pu−Be, 50cm

Figure 5: Detection rates for different multiplets for the Pu-Be source.

4.4. Analysis of PSD error231

Assume that the count rates are related as Rnnn = 0.1Rnnγ = 0.1Rnγγ =232

0.1Rγγγ , which is only a rough estimate of what was observed in measure-233

ments. It means that if 1% of all pulses are misclassified the value of Rγγγ234

will be underestimated by 2%, while the Rnnn value will be overestimated by235

7%. Similarly, if the PSD error is 5% then the value of Rγγγ could be off by236

9.5%, while the Rnnn value could be off by 33%. The values are related as237

11

follows:238

Rnnn = R′

nnn(1 − ǫ)3 + R′

nnγǫ(1 − ǫ)2 + R′

nγγǫ2(1 − ǫ) + R′

γγγǫ3, (1)

239

Rγγγ = R′

γγγ(1 − ǫ)3 + R′

nγγǫ(1 − ǫ)2 + R′

nnγǫ2(1 − ǫ) + R′

nnnǫ3, (2)

where ǫ is the PSD error and R′

xxx are the real rates and Rxxx are the mea-240

sured rates.241

When considering a pure neutron multiplicity assay, the resulting errors242

on the measured rates would be high if the PSD error was very high, but243

with the additional information available, such as the additional count rates244

and ratios from the proposed method, it is possible to calculate correction245

factors for the Rn, Rnn, Rnnn values. Designing calibration measurements246

that analyze the PSD error would be beneficial for further development of247

the applicability and accuracy of the proposed system.248

5. Summary and Conclusions249

The measurement system presented in this paper uses techniques that,250

when combined together, create a novel way of characterizing materials using251

both neutron and gamma ray multiplets. A study was performed to investi-252

gate the feasibility of the method, and to gain some insight on whether it is253

an applicable approach to materials accounting and identification.254

The measurement system also shows potential in accuracy of correctly255

differentiating the neutron pulses from the gamma ray pulses. This capa-256

bility is vital for the correct classification of mixed multiplets. The initial257

tests have shown that the system can acquire rates of n, γ, nn, nγ, γγ, nnn,258

nnγ, nγγ and γγγ multiplets in a relatively short measurement time for a259

252Cf source with 19500 fissions/s and a 1-Ci Pu-Be source. In the future,260

several aspects will be further investigated, including the number and con-261

figuration of detectors, and the possible utilization of shielding to avoid cross262

talk between detectors.263

It has been shown that a number of characteristic values can be obtained264

and ratios of the measured values can be used to assess the type of reactions265

causing the particle emission such as fission versus (α, n) reaction. The266

measured rates obtained in the measurements described in this paper are of267

satisfactory level for the simple design of the current system. The analysis268

shows that the ratios between rates could also be utilized as an indicator of269

source type and possibly also source isotope composition, once the analysis270

is extended further in connection to multiplicity theory.271

12

6. Acknowledgments272

The work of Andreas Enqvist was supported by the Swedish Radiation273

Safety Authority.274

References275

[1] N. Ensslin, Application Guide to Neutron Multiplicity Counting. Los276

Alamos Report LA-13422-M (1998).277

[2] J.E. Stewart, H.O. Menlove, D.R. Mayo, W.H. Geist, L.A. Carrillo, G.D.278

Herrera, Los Alamos Report LA-13743-M (2000).279

[3] H. Klein, S. Neumann, Nucl. Instr. Meth. A 476, (2002) 132.280

[4] Y.Xu, T.J.Downar, S.Avdic, V.Protopopescu, S.A.Pozzi, Techniques for281

neutron spectrum unfolding from pulse height distributions measured282

with liquid scintillators, in:Proceedings of the Institute of Nuclear Ma-283

terials Management 47th Annual Meeting, Nashville, TN, July 1620,284

2006.285

[5] M.Reginatto, P.Goldhagen, S.Neumann, Nucl. Instr. and Meth. A, 476286

(2002) 242.287

[6] K. Frame, W. Clay, T. Elmont, E. Esch, P. Karpius, D. MacArthur,288

E. McKigney, P. Santi, M. Smith, J. Thron, R. Williams, Nucl. Instr.289

Meth. A, 579 (2007) 192.290

[7] R. Sher and S. Undermyer, ”The detection of fissionable materials by291

nondestructive means,” American Nuclear Society (1980).292

[8] J.T. Mihalczo, J.A. Mullens, J.K. Mattingly, T.E. Valentine, Nucl. Instr.293

Meth. A, 450 (2000) 531.294

[9] M. Flaska S.A. Pozzi, Nucl. Instr. Meth. A, 577 (2007) 654.295

[10] I. Pazsit and L. Pal, “Neutron Fluctuations - a Treatise on the Physics of296

Branching Processes”, Elsevier Science Ltd, London, New York, Tokyo,297

2008.298

[11] S. Avdic, A. Enqvist, I. Pazsit, ESARDA Bulletin Vol. 43 (2009) 21.299

13

[12] A. Enqvist, I. Pazsit, S. Avdic, doi:10.1016/j.nima.2010.01.022.300

[13] S.A. Pozzi, S.D. Clarke, M. Flaska, P. Peerani, Nucl. Instr. and Meth.301

A, 608 (2009) 310.302

[14] A. Enqvist, M. Flaska, S.A.Pozzi, Nucl. Instr. and Meth. A 595 (2008)303

426.304

[15] J.M Adams, G. White, Nucl. Instr. and Meth. 156 (1978) 459.305

[16] H. Sperr, H Spieler, M.R. Maier, D Evers, Nucl. Instr. and Meth. 116306

(1974) 55.307

[17] T. E. Valentine, “Evaluation of prompt fission gamma rays for use in308

simulating nuclear safeguard measurements,” ORNL/TM-1999/300.309

[18] H. Vega-Carrillo, E. Manzanares-Acuna, A. Becerra-Ferreiro, A.310

Carrillo-Nunez, Applied Radiation and Isotopes 57 (2002) 167.311

[19] A. M. Gagarski, I. S. Guseva, V. E. Sokolov, G. V. Valski et al. Bulletin312

of the Russian Academy of Sciences: Physics 72 No. 6 (2008) 773.313

14