CHARACTERIZATION AND APPLICATION OF 4HE FAST NEUTRON SCINTILLATION DETECTORS TO NUCLEAR MATERIALS AND RADIATION DETECTION

By

YINONG LIANG

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2018

© 2018 Yinong Liang

I dedicate this work to my wonderful parents.

4

ACKNOWLEDGMENTS

I would like to express my sincere gratitude to my advisor, the chair of my thesis

committee, Prof. Andreas Enqvist. Without his supports and encouragements, I could

not have finished the Ph.D study. He spent endless hours proofreading my research

papers and giving me excellent suggestions, which always result in improved versions

of documents. English is my second language, and I’m grateful for all his edits for

making my writing precise and clear.

Working with Prof. Enqvist is inspirational. He gives me many opportunities to try

new ideas, explore different fields, and present in professional conferences. When I

mention about him to my friends, they always tell me how lucky I am. I am so lucky to

having you as my advisor! I would like to thank Kelly Jordan, for accepting me to the

4He project in the first place. I’ll never forget your encouragement when I’m suffering

from homesick and the loss of my grandmother. Another person who is special to me is

Ting, who did her postdoctoral study in our group. From her, I learnt how to research a

problem and achieve goals.

Additional thanks are due to all my committee members: Prof. James Baciak,

Prof. Yong Yang, and Prof. Heather Ray, thanks for your academic guidance and

supports throughout the research. I would also like to say thanks my colleagues:

Haitang, Xianfei, Ira, Kelsey, Noah, Taylor, Yuan, Surafel, and many others who I have

worked with. I did enjoy every group meeting and lab measurement with you!

Pursuing a Ph.D. is a difficult and sometimes a lonely path. I would like to say

thanks to these sweet friends I met in UF: Chenyi, Di, Mo, Jing, Hui, Yue, Yuning, and

Qian. Thanks for lighting up my life here. I miss every party we held.

5

Thank you, my dear friends in China: Yujie, Yanxue, Xiaojing, Yu, Ruiqi, Yanxin,

Hongling, Xiang, Tairan, and the lovely roommates I met during undergrad. I know you

for years. Spending time with you always makes me so happy, if only for a short time,

and I love every minute of it.

Also, I’d like to express my deepest thanks to my family for their love,

understanding, and encouragement. Without your understanding, I would not have

pursed an education in the U.S. At the end, I sincerely wish everyone I know all the best

in the future.

6

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS .................................................................................................. 4

LIST OF TABLES ............................................................................................................ 8

LIST OF FIGURES .......................................................................................................... 9

ABSTRACT ................................................................................................................... 13

CHAPTER

1 INTRODUCTION .................................................................................................... 15

1.1 Problem Description and Motivation ................................................................. 17 1.2 Project Overview and Deliverables ................................................................... 21

1.3 Dissertation Layout ........................................................................................... 25

2 NEUTRON DETECTION ........................................................................................ 27

2.1 Neutrons from Spent Nuclear Materials ............................................................ 28

2.2 Review of Neutron Interactions ......................................................................... 31 2.3 Review of Neutron Detection Methods .............................................................. 34

2.4 4He Fast Neutron Scintillation Detectors ........................................................... 38

3 4He DETECTORS RESPONSE CHARACTERIZATIONS AND SPECTRAL UNFOLDING ALGORITHM .................................................................................... 42

3.1 4He Detectors Characteristics ........................................................................... 43

3.2 Time-of-Flight Measurement and Detector Response Matrix ............................ 49 3.3 Spectral Unfolding with 4He Detectors .............................................................. 59

3.3.1 Review of Current Spectrum Unfolding Methods ..................................... 60

3.3.2 Iterative Least Squares Unfolding Algorithm and Uncertainty Estimation ..................................................................................................... 61

4 TIMING RESOLUTION MEASUREMENT OF PMT AND SIPM-BASED 4HE DETECTORS .......................................................................................................... 70

4.1 Why Characterize the Timing Resolution .......................................................... 71

4.2 Experiments ...................................................................................................... 72 4.2.1 PMT and SiPM Based 4He Detectors ...................................................... 73 4.2.2 SIS3316 and PSI Digitizers ..................................................................... 74 4.2.3 Timing Resolution Measurements and Calculation ................................. 76

4.3 Timing Resolution of PMT-Based 4He Detectors .............................................. 79 4.4 Timing Resolution of SiPM-Based 4He Detectors ............................................. 84 4.5 Conclusions ...................................................................................................... 87

7

5 NEUTRON AND GAMMA-RAY CROSS-CORRELATION FUNCTIONS MEASUREMENT WITH 4HE DETECTORS ........................................................... 89

5.1 Introduction and Advantages of Measuring Cross-Correlation Functions with 4He Detectors ...................................................................................................... 89

5.2 Cross-Correlation Functions with 252Cf Spontaneous Neutron Source ............. 91 5.3 Cross-Correlation Functions with Pu-Be (α, n) Source ..................................... 98 5.4 Conclusion ...................................................................................................... 101

6 SECONDARY NEUTRON MEASUREMENT WITH 4HE DETECTORS AT UF-HEALTH PROTON THERAPY INSTITUTE .......................................................... 103

6.1 Introduction and Literature Review of Proton Therapy and Its Risks .............. 104

6.2 Experimental Setups and Dose Estimation Method ........................................ 105 6.3 Results and Discussions ................................................................................. 108 6.4 Conclusions .................................................................................................... 118

7 CONCLUSION AND FUTURE WORK .................................................................. 121

LIST OF REFERENCES ............................................................................................. 123

BIOGRAPHICAL SKETCH .......................................................................................... 131

8

LIST OF TABLES

Table page 4-1 Relatively energy resolution of the TOF measurement (10 m path), given a

4He detector time resolution of 3.229 ns. ............................................................ 72

5-1 The ratio of the four category-pairs from 252Cf and Pu-Be measurements. ........ 99

6-1 Neutron characteristics from various solid water thicknesses from measurement. ................................................................................................... 114

6-2 Neutron dose estimation for various solid water thicknesses. .......................... 116

6-3 Neutron dose estimation at different room locations. ........................................ 118

9

LIST OF FIGURES

Figure page 2-1 Change in total neutron spectra as a function of cooling time (left) and

change in neutron source strength as a function of cooling time (right). While the total neutron and spontaneous fission neutron flux both decrease, the (α,n) neutron flux increases. ............................................................................... 31

2-2 Typical capture cross-sections of fissile material. Slowing down neutrons results in an increased interaction probability. .................................................... 32

2-3 Neutron capture and elastic scattering cross-sections of 3He and 4He, respectively. The elastic scattering cross-section of 4He exhibits a peak at around 1 MeV, matching the emission spectrum of fission neutrons quite well [43]. .................................................................................................................... 38

2-4 A 4He fast neutron detector. ............................................................................... 40

3-1 Typical digitized PMT outputs of a gamma-ray (left) and neutron (right) detection event. The fast component lasts approximately 50 ns, followed immediately by a slow component till the end of the 4.27 μs pulse event window. .............................................................................................................. 44

3-2 Scatter plot of fast component against slow component of the TOF measurement. Nonlinearities in electronics were reduced by utilizing a low gain settings and the pulse post-processing method to appear outside of the energy range of the TOF measurement (10 MeV). ............................................. 46

3-3 Scatter plot of fast component against slow component of the TOF measurement within different neutron energy ranges (10,000 number of pulses in each plot). ............................................................................................ 47

3-4 Induced fission neutrons (by neutron generator active interrogation on natural U samples) are distinguishable from generator neutrons in pulse height spectrum comparisons. ...................................................................................... 48

3-5 The scintillation light (in terms of the slow component) vs. the deposited energy. ............................................................................................................... 49

3-6 The TOF spectrum with about 6⨯105 events (left), and the experimental neutron flux from the 9Be (d, n) reaction (right), incident on the detector volume. ............................................................................................................... 52

3-7 Black line: measured 4He detector intrinsic efficiency as a function of incident neutron energy from TOF measurement. Red line: simulated 4He detector intrinsic efficiency as a function of incident neutron energy from MCNPX. ......... 54

10

3-8 Slow component only vs. TOF (left), and total scintillation light vs. TOF (right) after PSD. Visually, the summation of fast component and slow component (i.e. total scintillation light outputs) does not improve the correlation. ................. 54

3-9 Response matrix for 4He detector from TOF measurement. ............................... 56

3-10 The simulated neutron energy-deposition distribution (left), and the final response matrix for 4He detector from MCNPX simulations (right). .................... 57

3-11 The TOF spectrum measured with a 73.7 Ci 252Cf spontaneous fission source containing about 2⨯106 events (left), and the calculated incident neutron spectrum (right). The spurious peak at the very beginning of the calculated spectrum results from random scatters. ............................................ 58

3-12 Response matrix measured from a 252Cf spontaneous fission source with 0.25 MeV bins (left), and the unfolded spectrum for the same 252Cf source (right). ................................................................................................................. 59

3-13 The unfolded D-D spectra when using 250 (left) and 2000 (right) scintillation light output bins. ................................................................................................. 63

3-14 A flowchart of the iterative unfolding algorithm. .................................................. 64

3-15 The light output spectra of the three measurements. ......................................... 68

3-16 The unfolding results (left), and a zoom in of the unfolded spectra (right), of the three measurements. .................................................................................... 68

4-1 Arktis Radiation Detectors Ltd. The SiPM-based 4He detector. 2016. ................ 74

4-2 Linear interpolation by using three MAW values. ................................................ 75

4-3 Experimental setup for studying the PMT-based 4He fast neutron detectors. .... 79

4-4 Gaussian fits for the timestamp difference from summed as well individual PMTs measured at 250 MHz, 50% default CFD. Time difference is calculated as the EJ-309 scintillator’s timestamp subtracted by the 4He detector’s timestamp. The source is placed at the center of the detector. .......................... 81

4-5 Gaussian fits for the timestamp difference from summed, as well individual, PMTs measured at 5 GHz, 20% CFD. The time difference was calculated as the EJ-309 scintillator’s timestamp subtracted by the 4He detector’s timestamp. The 60Co source is placed at the center of the detector. .................. 82

4-6 FWHM as a function of CFD ratio at 5 GHz. The 60Co source is placed at the center of the detector. ......................................................................................... 83

11

4-7 FWHM as a function of source location at 5 GHz. 3 source locations are measured. The middle data point is the center-point between PMT1 and PMT2. ................................................................................................................. 84

4-8 Examples of 4He SiPM detector’s output pulses. The pulses have a long rising time (about 150 ns) along with significant signal noise. ............................ 85

4-9 Gaussian fit for the timestamp difference from summed as well individual channels within segment 1 at 250 MHz sampling frequency. Time difference is calculated as the EJ-309 scintillator’s timestamp subtracted by the 4He detector’s timestamp. The source is placed at the center of segment 1. ............ 85

4-10 The FWHM of each segment when source is placed at the center of that segment, measured with the 250 MHz Struck digitizer. ...................................... 86

5-1 Experimental setup of the cross-correlation functions measurements. ............... 92

5-2 Scatter plot of the integration of fast versus slow component of the cross-correlation measurement. ................................................................................... 93

5-3 Measured 252Cf cross-correlation functions at “15cm-15cm” source-detector distance (left) and “10cm-20cm” source-detector distance (right). The total cross-correlation function is obtained by summing all the correlated pairs together. Uncertainty is shown on the “total” curve, and is of identical magnitude in the individual component curves for data points at the same vertical position (“Normalized counts”-amplitude) as the total curve. .................. 94

5-4 Measured and Gaussian-fitted time delay distributions for 252Cf (n, n) pairs at various source-detector distances (left), and the (n, n) peak position as a function of source location (right) after gamma peak correction. ........................ 96

5-5 Measured and Gaussian-fitted time delay distributions for 252Cf (n, n) pairs at various ADC thresholds at “15cm-15cm” source-detector distance (left), and the FWHM of Gaussian fitting as a function of digitizer threshold at “15cm-15cm” source-detector distance (right). .............................................................. 97

5-6 The (n, n) peak position as a function of the 252Cf location along the length-dimension of the detector after gamma peak correction. .................................... 98

5-7 Measured Pu-Be and 252Cf cross-correlation functions at “15cm-15cm” source-detector distance. Data from Pu-Be and 252Cf is normalized by “per ns” Uncertainty is showed on the “total” Pu-Be curve, and is of identical magnitude in the individual curves for data points at the same vertical position (“Normalized counts”-amplitude) as the total curve. .............................. 99

5-8 Measured and Gaussian fitted time delay distributions for Pu-Be (γ, n) pairs at various source-detector distances (left), and the (γ, n) peak position as a function of source location (right) after gamma peak correction. ...................... 101

12

6-1 The measured depth-dose curve (SOBP) for this field at UFPTI, for a 180 MeV proton beam. ............................................................................................ 106

6-2 The simplified diagram of the nozzle. ............................................................... 107

6-3 The experimental setup at UFPTI. .................................................................... 108

6-4 PSD plots at various solid water thicknesses (upper left: 15 cm, upper right: 20 cm, lower left: 25 cm, lower right: 30 cm). ................................................... 110

6-5 Measured scintillation light outputs (left) and the exiting neutron spectra (right) from solid water with various thicknesses. The detector is placed 30 cm away from the solid water. .......................................................................... 111

6-6 Three range modulation wheels (RMW) with nine tracks modeled within TOPAS. ............................................................................................................ 112

6-7 Simulated neutron spectra emitted from various solid water thicknesses. ........ 113

6-8 Experimental setups to study the effect of measurement location. ................... 116

6-9 Measured scintillation outputs (left) and the exiting neutron spectra (right) at different room locations. ................................................................................... 118

13

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

CHARACTERIZATION AND APPLICATION OF 4HE FAST NEUTRON SCINTILLATION

DETECTORS TO NUCLEAR MATERIALS AND RADIATION DETECTION

By

Yinong Liang

December 2018

Chair: Andreas Enqvist Major: Nuclear Engineering Sciences

Spent nuclear fuel (SNF) will likely be stored in dry storage for an extended

period of time in the United States, and many other locations around the world.

Currently there are no reliable methods to verify the content of sealed dry casks. This

research will utilize 4He pressurized gas fast neutron scintillation detection technology to

address this technology gap.

The novel 4He detectors have unique advantages over traditional neutron

scintillation detectors, such as low gamma-ray interaction probability and direct

detection of fast neutrons. A high cross section of the 4He elastic scattering can be

found at fast neutron energies, particularly around 1 MeV. Around the dry casks, the

emission from fissile materials have energies ranging from thermal neutrons to un-

collided high-energy neutrons, which closely match the 4He elastic scattering cross

section.

The goal of the project is to develop a detection system which can provide a

direct, non-invasive, accurate, and independent measurement of the neutron spectrum

emitted from spent nuclear fuel.

14

It can be used for verification of nuclear material as well as giving information on

the content of the used fuel. In addition to using the 4He scintillation detectors as an

analysis tool to monitor the SNF dry casks, the detector also shows promising potentials

and advantages in related areas such as nuclear safeguards and medical physics.

Indeed, the detectors are suitable for a wide range of applications where the neutron

spectrum itself needs to be determined or the neutron spectrum can provide valuable

insights for characterization and quantification purposes.

15

CHAPTER 1 INTRODUCTION

The fuel assemblies in commercial nuclear power plants are made up of uranium

oxide, which is formed into solid cylindrical ceramic pellets and contained in a rigid

metal framework of zirconium alloy tubes. Due to accumulation of neutron absorbers

and depletion of the fuel, the nuclear chain reaction can no longer be self-sustained and

the power output will be reduced. Roughly every 18 months, the plant is shut down so

that approximately one-third of the fuel assemblies can be removed and replaced with

fresh fuel. The removed fuel assemblies are so-called “spent nuclear fuel” (SNF). They

are high-level nuclear waste and require specialized storage.

Upon removal from the core, SNF is initially stored in deep spent fuel cooling

pools on-site. SNF cooling pools, also known as “wet storage”, are designed as a short-

term storage location. The SNF is allowed to be cooled both thermally and radioactively

by the decay of the short-lived fission products over time. Then it can be safely

relocated to a dry storage system, if necessary. For long term storage, a geological

repository is the preferred method, but in the United States, the repository location has

been debated for decades. Currently, there is no permanent repository. The dry cask

storage is designed as an interim solution and is now expected to store the SNF for

longer periods of time than previously anticipated.

The amount of fuel in dry storage increases at a rate of roughly 2,000 metric tons

(MT) each year [1], and by the end of 2018, for the roughly 100 currently licensed

operating reactors, the total amount of the SNF is estimated over 80,000 MT. The on-

site SNF cooling pools are reaching capacity, and currently there are no reliable

methods to verify the content of sealed dry casks. The International Atomic Energy

16

Agency (IAEA) has expressed a need for robust safeguards and verification

technologies to ensure a continuous knowledge and maintain the integrity of spent

nuclear fuel inside the dry casks. The IAEA also needs technologies addressing other

problems arising from the SNF storage, including safety, security, transportation, costs,

and politics.

Our research aims to address the above-mentioned technology gap in the

safeguarding and monitoring of spent nuclear fuel storages by utilizing the novel 4He

fast neutron scintillation detectors. The goal of the project is to develop a detection

system capable of detecting the measurable changes in neutron emission and use

these specific features to verify the contents in the storage casks. Due to the sheer

amount of fissile materials in a single dry cask, the emission from fissile materials have

energies ranging from thermal neutrons to un-collided high-energy neutrons suitable for

the detection capability of the 4He detectors. Included and related studies include time-

of-flight (TOF) based neutron response function characterization [2], iterative least

square fitting based unfolding algorithm development [3] and detector’s timing resolution

characterization based uncertainty estimation, and Monte-Carlo computer modeling

based neutron spectrum prediction [4]. The knowledge of which will be used in creating

a fully developed and verified cask-measuring prototype system will be built and used to

measure suitable spent fuel materials or casks for system evaluation.

In this chapter we will begin with a brief introduction of the SNF storage methods

and its current status in the United States, followed by the reasons that motivated the

need of monitoring the SNF dry storage facilities. Common methods and radiation

detectors for nuclear materials verification are then evaluated, and the unique features

17

of the novel 4He fast neutron scintillation detectors will be presented and evaluated,

showing its promising potentials for the applications above. Other detector capabilities

such as coincidence measurements and radiation dose evaluations are also explored

and regarded as the additional deliverables of the project. The chapter concludes with

an overview of the thesis layout, to give the reader a clear view of the outline and

structure of the whole research.

1.1 Problem Description and Motivation

There are two acceptable storage methods for spent nuclear fuel after it is

removed from the reactor core: SNF cooling pools and dry storage casks. The cooling

pools are very deep, with several meters of water above the top of the long fuel

assemblies for radiation shielding purposes. These pools are lined with stainless steel

to prevent leaking [5], and can accommodate aluminum racks which hold the fuel

assemblies. Deionized water is usually used in the pools to mitigate the corrosion and

decrease the reactivity of zircaloy fuel cladding and stainless steel canisters with the

water [6]. In addition to using pumps to circulate water to remove the heat produced by

the spent fuel assemblies, natural circulation of air is another way to cool the SNF in a

worst case scenario [7].

For most United States nuclear reactors, wet storage alone is not able to meet all

the storage needs. Without a large-scale certified repository, commercial nuclear power

plants in the United States have been forced to use their on-site SNF pools for long-

term storage. As on-site SNF water pool storage capacity fills up, the demand for a dry

storage technology has increased [5]. Therefore, in recent years, cask storage has

become a more popular method of SNF storage.

18

Dry cask storage is a method of storing high-level radioactive waste, such as

SNF that has already been cooled in the spent fuel pools for several years to become

less radioactive. The casks are typically steel cylinders that are either welded or bolted

closed, providing containment of the spent fuel. Additional steel, concrete, or other

material surrounds each cylinder to provide further radiation shielding. The dry casks

are all sealed and designed for safety proposes, but that also brings limitations in ability

to verify the contents of the casks [8].

The amount of fuel in dry storage increases each year. On average, each of the

operating reactors produces 20 metric tons of plutonium per year [9]. For comparison 8

kg of plutonium is refered to a significant quantity (SQ), due it being the consensus

minimum amount needed to construct a successful nuclear weapon. The United States

has commercial spent fuel stored up at 78 sites in 34 states, which currently accounts

for about 200,000 significant quantities of plutonium in dry cask storage [10]. The design

of the SNF dry storage is a well-established technology that has been licensed in the

United States since 1985. It has a good safety record. No major safety incidents with

radiation release have occurred so far.

However, a problem unique to the nuclear energy field is the underlying

connection between nuclear energy and nuclear weapons. Fissile materials such as

plutonium and 235U are contained in the fuel of light-water reactors, which could be used

to make nuclear weapons. If terrorists manage to steal the SNF, the radioactive

materials would be released. Although the DOE claims that only 34 g of respirable

irradiated fuel could be released from an attack, and Nevada State reported that eight

acres would be contaminated for every 2,000–10,000 Ci released [11] [12]. Yet little

19

information is available on the potential loss of SNF materials. The fission products and

higher actinides are highly toxic and could contaminate humans and the environment.

Only a few studies have considered the consequences of the release of the SNF [11]

[12]. In addition, dry storage casks tend to be more vulnerable if stored outside reactor

facilities due to relatively less protections and guards. Therefore, SNF must be

adequately protected and monitored, and people must have a continuous knowledge of

the contents through the lifetime of the dry casks.

For SNF verification, determining the quantity of fissile material in spent nuclear

fuel, is one typical technology used in nuclear nonproliferation and safeguards [13].

Initial implementations sought to use gamma-ray spectrometry to solve this problem.

However, due to self-shielding effects, gamma-ray spectrometry is only able to analyze

the “skin of a sample”, rather than providing information about the “core of a large

sample”. In contrast, measuring the neutron radiation emitted from the canister has the

potential to give significant information about the fuel stored inside. Therefore, neutron

techniques are commonly chosen over gamma-ray techniques. The 3He proportional

counters are regarded as the “gold standard” of neutron detection, which has high

intrinsic neutron detection efficiency and is insensitive to gamma-ray radiation.

However, 3He detectors suffer from a supply shortage and for the loss of neutron

energy information when used with moderator material to achieve acceptable detector

efficiency. Liquid scintillation detectors, NE-213 and EJ-309 for example, could be a

replacement for 3He detectors. They are time-tested and have a relatively high intrinsic

reaction rate. No moderators are needed during fast neutron detection, and the fast

timing performance makes them especially attractive for coincidence measurements

20

[14]. But in a high gamma-ray intensity field, gamma-ray saturation becomes an issue

and adding gamma-ray shielding materials might introduce additional uncertainties. The

material safety could be another concern since liquid scintillators are commonly toxic

and/or flammable, which makes them less deployable in a field scenario. Another

solution may be Bonner spheres, which consist of several thermal neutron detectors

surrounded by various thicknesses of moderating material [15]. However, it is time

consuming and the energy resolution of the network is limited by the number of detector

configurations used. In addition, neutron spectrum unfolding may be another problem,

because it requires a priori spectral shape informationcontaining physical information

about the neutron field, and the selection of the priori spectrum could affect the quality

of the obtained solution. Therefore, there is a need to search for other technologies

which are capable of detecting neutrons over the i neutron energy range of interest

(from thermal to un-collided high-energy neutrons) with reasonable detection efficiency,

good neutron/gamma-ray discrimination ability, and spectrum unfolding feasibility [16].

The novel 4He fast neutron scintillation detector, developed by Arktis Radiation

Detectors, is a relatively new tool which fulfils the above-mentioned requirements for an

alternative neutron detector. To begin with, 4He gas is significantly more available than

3He gas. In addition, the detector has some unique advantages such as high neutron

detection efficiency, low gamma-ray interaction probability, and direct detection of fast

neutrons. In particular, the gamma-ray rejection capability, is crucial and ensures that

the 4He detector is an ideal tool for measuring the neutrons emitted from spent nuclear

fuel storage casks, where significant interfering gamma-ray background is present.

More details regarding the detector characteristics will be discussed in following

21

chapters. There is currently no reliable methods to verify the content of sealed dry casks

[17]. Therefore, this project was purposed to evaluate the feasibility to build a prototype

detection system with the 4He detectors, which is capable of unambiguously verifying

the contents of dry storage casks in a non-intrusive manner for the safeguarding and

monitoring of SNF storage installations.

The research is important because with an increasing quantity of spent nuclear

fuel being stored in dry casks for an extended period of time in the United States [4], it is

important to improve the methods of dry casks monitoring. There is also a need to

certify these dry casks for long-term storage and transportation, as well as to minimize

the risk of nuclear proliferation and terrorism [18]. Additionally, the research is beneficial

in terms of the next generation nuclear materials management, nuclear safeguards

(multiplicity and coincidence counting techniques, etc.), and the continuous verification

of cask contents during shipping and receiving operations. The detection system will

prove a valuable tool for identifying cask-specific features to ensure no material is

diverted during these higher risk operations or throughout cask life in storage. Last, the

project can also help to determine the potential needs for built-in measurement

capabilities for future dry storage casks designs.

1.2 Project Overview and Deliverables

This project aims to address the technology gap in the safeguarding and

monitoring of used fuel storage installations by developing a 4He-based system capable

of such monitoring. The novel 4He gas scintillation fast neutron detectors, with the

unique advantages over current neutron detectors, have been proposed as the

foundation to develop a neutron spectrometer for the applications above.

22

Previous measurements [19] have shown that the 4He detector have different

responses (scintillation light outputs) to a variety of neutron energy spectra. The

detectors can measure a wide range of neutron energies up to 10 MeV and the energy-

dependent light output distributions can be used as the basic tool to identify neutron

sources. Thus, determining the response matrix is the very first and the most important

step to utilize the 4He detector efficiently. As a property of the 4He detector, the

response matrix, which includes dependence on detector-electronics like the

photomultiplier tubes, can be used to predict the light output as well as be used to

unfold the incoming neutron spectrum from the measured light distribution. The latter is

of greater interest for this project of spent fuel monitoring where the neutron spectrum is

unknown and is to be estimated from 4He detector outputs. The determination of the

detector’s response matrix can be achieved through time-of-flight (TOF) measurements

[2], where the incident neutron energy can be determined by knowing its time-of-flight

(TOF) and travelled distance. The detector response to both mono-energetic (e.g.

neutron generators) and quasi-mono-energetic (e.g. accelerator based) sources from

the direct measurement of a neutron signal will be obtained and compared through

various experiments. In addition, the Monte Carlo simulation code MCNP-PoliMi will be

used, which predicts the deposited neutron energy for individual neutron interactions so

as to build the “kinetic response matrix” [3]. By combining the matrix with the measured

and modeled non-linear scintillation response (a function of deposited energy), a

simulation-based response matrix can be obtained for verification of the measured

response matrix.

23

The next step of the project is to develop a spectrum unfolding algorithm. The

detector light output is a function of both the incident neutron energy and the detector’s

internal conversion between energy and light. This internal conversion is defined by the

detector’s response matrix, which is a probabilistic mapping between incident neutron

energy and the scintillation light. When both the detector’s response matrix and

scintillation light response are known, the incident neutron spectrum can be calculated

backwards through the unfolding algorithm, which is developed upon iterative least-

squares fitting. The uncertainties of the unfolding algorithm will then be estimated.

Different neutron sources, such as spontaneous fissions and (α, n) reactions, have been

used to test the robustness and effectiveness of the unfolding algorithm.

In addition to detector characterization and unfolding algorithm development, the

other part of this work is focused on Monte-Carlo computer modeling. As from previous

studies [4], the majority of neutrons produced in SNF are generated by spontaneous

fission or from alpha particles interacting with oxygen, and a significant fraction of these

neutrons will be transmitted through the cask which can be measured by the 4He

detectors. Spontaneous fission neutrons are emitted following the Watt spectrum with a

maximum probability around 1 MeV and decreasing gradually afterwards, ending

around 10 MeV [15]. Neutrons from (α, n) reactions usually have higher energies and

the shape of the spectrum differ, depending on the reactions. For (α, n) neutrons from

SNF, previous simulation works indicate a higher yield between 2.5 MeV and 3 MeV [4].

The final task is to develop a cask-measuring prototype system. This work

evaluates all the detector capabilities and characterizes the performance and feasibility

to be utilized in a final prototype measurement system. Simulation demonstration of a

24

prototype cask measuring system is covered in another PhD Thesis [20]. The spectral

features mentioned above are the basic metric for performance prediction and design

optimization of the monitoring system. Suitable realistic SNF libraries will be used in the

modeling of this task to ensure that the results adhere with realistic scenarios. Existing

computer models (in MCNP6 and ORIGEN-S) will be expanded to precisely simulate

passive neutron spectra emitted from the spent fuel, transport in the fuel and cask, and

the predicted cask measurement system response. The fully developed and verified

prototype will then be used to measure suitable spent fuel materials or casks for system

evaluation.

In addition to the above-mentioned deliverables, other applications related to

nuclear materials detection and characterization will also benefit from the new 4He-

based measurement techniques. As part of a passive non-destructive assay (NDA)

special nuclear material detection and characterization tool, the 4He detectors can use

their good timing resolution for coincidence measurements. The cross-correlation

functions can be obtained from coincidence measurements of neutrons and/or gamma

rays. The cross-correlation functions represent signatures allowing typical neutron

source identification (i.e. spontaneous-fission or alpha-n) and radioactive material-

geometry configuration assessment [14]. In the case of mixed or heavily gamma-ray

emitting materials, most fast-neutron detectors will suffer from gamma-ray saturation

and pulse pile-up, and may require large amounts of shielding which adds additional

uncertainties to the measurement. In contrast, the gamma-ray insensitivity enables the

4He detectors to operate functionally without such issues.

25

The application of the 4He detectors can also be extended to the medical physics

area, for example measuring the secondary neutron spectrum and dose from proton

therapy. In proton therapy, high-energy protons deposit almost all their energy within a

short range, resulting in minimal dose outside the range of the proton beam. The finite

range and sharp distal fall-off means proton therapy offers dosimetric advantages by

focusing on the tumor more precisely than x-ray radiation therapy (XRT) [21]. High-

energy protons from an on-site cyclotron are guided through delivery systems and

collimated to provide proton beam profiles suitable for cancer patient treatment. It is

known that secondary radiation such as neutrons and gamma rays can be generated

within the beam shaping process as well as in the patient body. The secondary

neutrons are of great interest since they could be a significant contribution to the overall

dose and could cause potential biologic effects. Exiting neutron dose measurements

with Bonner spheres is time consuming and are not able to cover the whole energy

range of the secondary neutrons. With the cooperation with UF-Health Proton Therapy

Institute (UFPTI), we are able to measure the secondary neutron spectrum and estimate

dose using 4He detectors. This provides an efficient and straight-forward way of

estimating dose due to secondary neutrons in order to optimize treatment designs and

implement treatment verification.

1.3 Dissertation Layout

In this thesis, Chapter 1 introduces the motivations of this research and an

overview of the project. Chapter 2 provides a background knowledge of the spent

nuclear fuel materials where the detectable neutrons come from, and a literature review

of common neutron detection methods.

26

At the end of this chapter, the unique advantages of the 4He detectors are listed

as the reason why they are suitable for this research. Chapter 3 and Chapter 4 describe

the experiments for detector characterization. Both the detector’s response to various

neutron energies and the detector’s timing performance were studied, which allows a

better understanding of the detector and a further exploration of the detectors

applications. Chapter 5 and Chapter 6 present the additional applications of the

detectors beside spent fuel casks monitoring. In Chapter 5, a cross-correlation

measurement is discussed to show the potential applications of the detectors, which are

capable of identifying different materials and material-geometry configurations. In

Chapter 6, the detectors application is extended to medical physics. The optimization

and verification of treatment designs of proton therapy can be achieved via the

measurement of secondary neutron spectrum produced during proton therapy with 4He

detectors, and this work was conducted with the cooperation with UF Health Proton

Therapy Institute (UFHPTI). Finally, Chapter 7 summarizes the work performed in this

thesis, and suggests studies for future investigation.

27

CHAPTER 2 NEUTRON DETECTION

Neutrons emitted from the spent nuclear fuel storage casks can provide more

content-related information than gamma rays due to higher penetrability. The neutron

spectrum serves as an indicator of the nuclear materials present and provides a

fingerprint for monitoring and verifying the storage casks. Therefore, it is worth

understanding the mechanism of neutron interactions in materials that are commonly

used for neutron detectors, such as those common neutron detectors are built from.

In this chapter, a background introduction of the spent nuclear materials is

presented at first, describing different types of fuel and the two primary processes

responsible for the production of neutrons: spontaneous fissions and (α, n) reactions.

The next two sections discuss how neutrons interact in matter and provide a review of

commonly used thermal and fast neutron detectors. Based on the types of detection

medium and mechanism, each detector has its own advantages and limits. Therefore, it

is important to choose a suitable detector for the intended application. The SNF exhibits

a high gamma-ray radiation environment, where common neutron detectors are easily

saturated, and the output pulses can pile up and lead to artificially increased or

decreased measured observables. Use of 4He detectors can somewhat mitigate these

problems, and in the last section of the chapter we provide a thorough discussion of 4He

detectors, showing how they work, what makes them unique, and why they are chosen

over other neutron detectors for this project.

The nuclear power plants in the world today can be broken into three major

categories: light water reactors (LWR), heavy water reactors (HWR), and other less

versions including graphite moderatedgas-cooled, and fast breeder reactor types.

28

Among which, LWRs are the most common. In LWRs, low-enriched uranium (3%-5%

235U) pellets are used as fuels, and water is used as a moderator to slow down the

neutrons as well as a coolant to remove heat produced in nuclear reactions in the

reactor. Although the types of fuel vary with the types of reactors, currently all U.S.

commercial reactors are LWRs. Therefore, the description in this chapter is primarily

focused on the SNF from LWRs. The idea of using 4He fast neutron detectors to

characterize and verify SNF is also applicable to other types of reactors.

2.1 Neutrons from Spent Nuclear Materials

The primary waste form resulting from nuclear energy production is spent nuclear

fuel. Spent fuel is fuel that has been used in commercial nuclear reactors that is no

longer able to economically sustain a nuclear chain reaction due to accumulation of

neutron absorbers and reduction in fissile material. It is worthwhile to introduce the term

“burnup”, which is expressed as the actual energy released per mass of initial fuel in

gigawatt-days/metric ton of heavy metal (GWd/tHM). In another words, burnup is a way

to measure how much of the fuel was burned in the reactor. In general, the longer the

fuel can sustain a chain reaction and therefore remain in the reactor, the higher the

burnup.

Approximately every 12-18 months, the plant is shut down so that a portion of

these fuel assemblies can be removed and replaced with fresh fuel [22]. In the SNF,

approximately 3% of the mass is fission and activation products. Fission products are

formed when heavy atomic nuclei, such as uranium or plutonium, are split in the fission

process in the nuclear reactor. Examples of fission products include iodine (129I, 131I),

cesium (134Cs, 135Cs, 137Cs), and strontium (90Sr).

29

They are regarded as radioactive waste and can be separated further for many

industrial and medical uses. Activation products are formed when neutrons are

absorbed in the fuel assemblies. Cobalt, nickel, and niobium are activated in this way.

About 1% of the mass comes from transuranic elements, which are produced by

neutron absorption in uranium. The most important of the transuranic elements is

plutonium (239Pu, 240Pu, 241Pu, etc.). It is a both a useful byproduct or a dangerous and

inconvenient waste, due to long half-lives. One of the main goals regarding nuclear

proliferation is preventing the plutonium from being used by terrorists as nuclear

dispersion devices, to produce nuclear weapons. Around 96% of the mass is the

remaining uranium: most of the original 238U and a little 235U [23].

After cooling down in spent fuel pools for several years, SNF can be transferred

to dry casks for medium term storage. In the dry storage casks, neutrons can be

produced through spontaneous fission of fissile materials. In the dry storage casks,

242Cm and 244Cm, with half-lives of 162.8 days and 18.10 years respectively [24],

contribute a significant portion of the spontaneous fission neutrons. Fissile material is

not the only source of neutrons in sample such as SNF; contributions from (a, n)

reactions cannot be omitted. The α-particles are produced by the decay of a myriad of

actinides created in nuclear fuel during power operations. These α-particles interact with

the oxygen in the fuel matrix to produce neutrons. Therefore, the primary neutron

sources of spent nuclear fuel are from spontaneous fissions and (α, n) reactions. Simply

looking at the gross neutron count does not reveal much information about the nuclear

materials, other techniques capable of distinguishing spontaneous fission and (α, n)

neutrons are required. One differentiator is that the spontaneous fission neutrons exhibit

30

a multiplicity feature, while the (α, n) reaction emits only a single neutron per reaction

[25]. The multiplicity counting method via coincidence measurement is inspired by this

feature. Another way to differentiate the neutrons coming from spent nuclear fuel is by

looking at the neutron spectrum. The fission neutrons are emitted approximately in a

Watt spectrum distribution with an emission peak around 1 MeV [26] while, depending

on the reactions, neutrons from (α, n) reactions are emitted predominantly at higher

energies.

Previous work [4] has modeled the neutron spectral characteristics in a 17X17

pressurized water reactor (PWR) assemblies via the ORIGEN-S code, the Monte Carlo

N-Particle code MCNP6, and the Next Generation Safeguards Initiative (NGSI) Spent

Fuel Library. Figure 2-1 shows that the simulated total neutron flux on the outside of an

assembly varies as a function of the cooling time, and the flux of different neutron

sources shows varying trends as cooling time changes. In the first couple of years most

neutrons are produced from spontaneous fission of heavy nuclides where 244Cm is the

biggest contributor. As the cooling time increases, (α, n) reactions then take over, where

241Am (produced by beta decay of 241Pu) is the largest contributor to neutron production.

There are significant differences between the neutrons from different reactions which

can be used for neutron source separation. From the simulation, the spontaneous

fission neutrons have a peak between 0.8 to 1 MeV and the maximum energy can go up

to above 10 MeV. The neutron spectrum from (α, n) reactions exhibits a harder

spectrum with multiple peaks between 2.5 and 3.0 MeV, and the tail of the neutron

spectrum stops much earlier at approximately 5 MeV. Therefore, it is possible to use the

neutron spectrum measured outside the spent fuel storages to achieve “cask-

31

fingerprints”. If unexpected changes are detected beyond the natural aging of the

material, it could indicate a diversion of nuclear materials.

Figure 2-1. Change in total neutron spectra as a function of cooling time (left) and change in neutron source strength as a function of cooling time (right). While the total neutron and spontaneous fission neutron flux both decrease, the (α,n) neutron flux increases [4].

2.2 Review of Neutron Interactions

In the previous section, we explained why using the neutron spectrum emitted

from SNF is vital for this project. Therefore, the first step is to detect the neutrons.

Because neutrons are neutral particles, the detection of neutrons is based on indirect

interactions. The two basic interactions that common detectors are built from are

scattering and nuclear reactions [27], and we will discuss them in details later in this

section.

Neutrons must pass close to a nucleus to interact, since the nuclear force which

leads to these interactions is very short-ranged. Due to the small size of the nucleus

(10-15 m) comparing to the atom (usually 0.3 to 3 angstroms), neutrons have a low

interaction probability [28]. To express the likelihood of an incident neutron interacting

with a target nucleus, a quantitative term known as the cross-section (σ) is used. The

32

cross-section has unit of area and has traditionally been measured in unit of barn (10-28

m2). The neutron cross-section varies and depends on: target nucleus, type of

interactions, neutron energy and target energy. For a given target and reaction type, the

cross-section is strongly dependent on the neutron energy as seen in Figure 2-2.

Therefore, we will divide neutrons into two categories on the basis of their energy, either

“fast neutrons” or “slow neutrons” and we will discuss them separately.

Figure 2-2. Typical capture cross-sections of fissile material. Slowing down neutrons results in an increased interaction probability [29].

Different techniques exist for different detection regions, for slow/low-energy

neutron detection, detectors are especially constructed to leverage suitable materials to

detect neutrons with energy below 0.5 eV. Slow neutrons frequently undergo elastic

scatterings and can transfer a fraction of their energy to the nucleus they interact with.

However, because the kinetic energy of a neutron is so low, only a small amount of

energy can be transferred during those elastic scatterings, and the resulted recoil

nucleus does not have enough energy to be regarded as an ionizing particle. Elastic

scattering often causes neutrons to come to a thermal equilibrium with the absorber

material before a different type of interaction takes place. Most slow neutrons will

33

therefore be confined to thermal energy region (around 0.025eV, at room temperature).

In addition, slow neutrons can undergo nuclear reactions, where neutrons are absorbed

and charged particles are then formed. The products of the nuclear reactions include

gamma rays, protons, alphas, and fission fragments, which can create secondary

radiation or ionization of sufficient energy to be detected directly. In most materials,

radiative capture ((n, γ) reaction) is the most probable reaction[30], but charged particle

nuclear reactions are more readily used in the indirect detection.

Fast neutrons have energies above 1 keV. With increasing neutron energy, the

probability of most nuclear reactions drops off rapidly while elastic scattering becomes

dominating. In contrast to slow neutrons, the recoiling nuclei from elastic scatterings

with fast neutrons can ionize surrounding materials and therefore, can be directly

detected. For an elastic collision, through the conservation of momentum and kinetic

energy, the energy of the recoil nucleus is given by:

𝐸𝑅 = [4𝐴

(1 + 𝐴)2] (𝑐𝑜𝑠2𝜃)𝐸𝑛.

(2-1)

In equation (2-1), is the scattering angle of the recoil nucleus in the lab

coordinate system (target nucleus is at rest), En is the initial energy of the incoming

neutron, and A is the ratio of the target nucleus mass and neutron mass. The maximum

energy transfer happens when is 0°, which is calculated as in equation (2-2) as below:

𝐸𝑅𝑚𝑎𝑥 = [4𝐴

(1 + 𝐴)2] 𝐸𝑛.

(2-2)

From the equations above, a neutron can transfer a fraction or all of its kinetic

energy to the recoil nuclei depending on the target mass, resulting in the neutron being

moderated down to lower energies. The most efficient moderator is hydrogen because a

34

neutron can lose all its energy in a single collision with a hydrogen nucleus. With

heavier nuclei, the maximum amount of transferred energy is significantly less.

Inelastic scattering will take place if the energy of the fast neutron is sufficiently

high. During an inelastic scattering, the recoil nucleus is elevated to one of its excited

states and decays by gamma ray or particle emission. Each excited state (1st, 2nd, 3rd,

etc.) associates with a discrete Q-value, however, if a gamma ray escapes from the

detector volume, it is hard to assess the original neutron energy. Therefore inelastic

scattering is an undesirable reaction in fast neutron recoil detectors [31].

2.3 Review of Neutron Detection Methods

In order to design and built a detector capable of detecting neutrons, several

factors must be taken into account when searching for nuclear reactions that may be

useful in neutron detection. First, the cross section for the reaction must be as large as

possible so that an efficient detector can be built in a small size (larger detectors are

costly and harder to construct and operate consistently). For the same reason, the

target nuclide should also have high isotopic abundance. The Q-value must be as large

as possible if using a radioactive capture reaction, since it determines the energy

released in the reaction following neutron capture. The higher the Q-value, the easier it

is to discriminate neutron events against undesirable background such as gamma rays

[27], via pulse shape discrimination or detection amplitude threshold. Most low-energy

(thermal) neutron detection is based on neutron-induced reactions in which the

detection of neutrons can be achieved through the detection of directly detectable

neutron-induced charged particles. In theory, all these reactions can be applied for fast

neutron detection as well.

35

However, the probability of the neutron-induced reactions decrease rapidly with

increasing neutron energy [32]. Therefore, fast neutron detection devices must employ

a modified or completely different detection scheme to achieve acceptable detection

efficiency.

There are three basic desirable attributes for neutron detectors for safeguards

application involving detection and monitoring of SNF: 1) high absolute detection

efficiency, 2) low intrinsic gamma-ray sensitivity of the detector and 3) an ability to

maintain a high neutron detection efficiency when simultaneously exposed to high

gamma-ray flux rates [33].

3He is known for its high thermal neutron absorption cross section, which results

in a very high detection efficiency as the main advantage of the 3He detectors. The 3He

proportional counter is very insensitive to gamma rays due to the low electron density of

the 3He gas, and can be used for both slow and fast neutron detection, which makes it

the most common tool for neutron detection [34]. In addition, the detectors are

inflammable, non-toxicity, physically robust, and easily-implemented. All these

characteristics allow them to be used for field measurements. For the detection

mechanism, neutrons interact with the 3He gas via the (n, p) reaction as shown in

equation (2-3):

𝐻𝑒23 + 𝑛 →0

1 𝐻13 + 𝑝1

1 + 0.764 𝑀𝑒𝑉. (2-3)

The thermal neutron cross-section for this reaction is 5330 barns, significantly

higher than that for the boron neutron capture reaction. Mostly, the 3He detectors are

used for thermal neutron gross counting, but the 3He (n, p) reaction can also be applied

to fast neutron detection and spectroscopy. As the incident neutron energy goes up, the

36

elastic scattering and (n, d) reaction begin to play a significant role. Those different

types of reactions account for multiple features of 3He detector response. For example,

the maximum energy deposition during an elastic scatter (the head-on collision) is 75%

of the neutron initial energy. However, one major shortcoming when characterizing 3He

detectors for neutron spectroscopy application is that only the pulses corresponding to

the total energy release in the (n, p) reaction should be retained and other pulses due to

the wall effect and 3He recoil should be eliminated from the measured pulse height

distribution. Additionally, due to the limited 3He supply, it is economically beneficial to

explore possible alternative neutron detectors for SNF measurements.

Boron trifluoride (BF3) filled proportional counters are similar to 3He detectors.

They allow good neutron/gamma-ray discrimination and are commonly used in non-

neutron detectors. However, the relative low detection efficiency (about 2-5 times lower

than 3He detectors) limits the detector’s application in nuclear safeguards, which is the

same limiting factor for the 10B lined proportional counters (about 7 times less efficient

than 3He detectors per detector).

Another alternative neutron detector is organic scintillators. For example, the

organic scintillation detectors in liquid or plastic form are widely used in the areas of

homeland security and nuclear nonproliferation recently [35]. These detectors are

sensitive to both fast neutrons and gamma rays [27], and electronic pulse shape

discrimination techniques can be applied later when differentiating the incoming

particles. Spectrum unfolding techniques are required and recently studied [36] [37] [38]

for fast neutron spectroscopy applications. However, the gamma-ray sensitivity could

also jeopardize the detector’s application in large volume scenarios where the goal is to

37

suppress the gamma-ray response. In this case, high-Z materials may be essential in

order to shield overwhelming gamma-ray count rates and achieve a better ratio of the

recorded neutrons. The shielding could introduce additional uncertainties that result

from induced and scattered neutrons. Other concerns also include low thermal neutron

detection efficiency, and long-term safety issues.

The last potential alternative neutron detector type for nuclear safeguards

application covered here is elpasolites like the Cs2LiYCl6:Ce3+ (CLYC) scintillator, which

has 95% 6Li enrichment and 0.5% Ce3+ doping. It can simultaneously detect gamma

rays, thermal neutrons and fast neutrons at relative high gamma-ray resolution (4% at

662 keV) with high light yield and a high gamma-ray energy equivalence of the neutron

absorption peak (>3MeV). For thermal neutrons, the detector has over two times the

cross-section of 3He (10 atmospheres), compared on a volume basis [39]. For fast

neutron spectroscopy, no unfolding techniques is required for neutrons below 4 MeV

[40][41]. CLYC has good pulse shape discrimination ability, which can be achieved via

both light integration and time profiles of light production [42]. But the long scintillation

decay time (several μs) could lead to significant pulse pile-ups at moderate input count

rate (e.g., kHz) which makes them less suitable for spent fuel monitoring applications.

The above-mentioned alternative detectors have their own limits in nuclear

safeguards applications. Therefore, this work looks at the novel 4He fast neutron

scintillation detectors, which could be regarded as a replacement for 3He detectors and

applied to field measurement.

38

2.4 4He Fast Neutron Scintillation Detectors

The detectors that are proposed to be used in this research are 4He scintillation

detectors. They are filled with high-pressure 4He gas, which is used as a fast neutron

detection medium.

As in Figure 2-3, the neutron capture cross-section of 3He is compared with the

elastic scattering cross-section of 4He (natural helium). For low-energy neutrons, 3He

has a high capture cross-section. However, when using the 3He detectors to detect fast

neutrons, it is best implemented in conjunction with moderating materials to make use

of the high cross-section at lower energies. In comparison, the elastic scattering cross-

section of 4He is substantially smaller in the low-energy region. However, for fast

neutron detection, the 4He cross section exhibits a peak that is located at roughly 1

MeV, nicely matching the peak emission of fission neutrons.

Figure 2-3. Neutron capture and elastic scattering cross-sections of 3He and 4He, respectively. The elastic scattering cross-section of 4He exhibits a peak at around 1 MeV, matching the emission spectrum of fission neutrons quite well [43].

39

In an elastic scattering interaction, energy is transferred from the incoming

neutron to a 4He nucleus. The neutron is not absorbed in this process. It can keep

traveling through the medium (4He gas) while the direction and speed are changed.

The maximum energy transfer from an incoming neutron to a 4He nucleus is 64% of the

neutron’s energy prior to the interaction due to neutron scatter kinematics.

The kinetic energy transferred to the nucleus is sufficient to strip the electrons

from the nucleus that then moves as a recoil alpha particle and excites (equation (2-4))

or ionizes (equation (2-5)) other helium atoms along its path within the detector volume.

𝛼 + 𝐻𝑒 → 𝐻𝑒∗ + 𝛼′. (2-4)

𝛼 + 𝐻𝑒 → 𝐻𝑒+ + 𝑒− + 𝛼′. (2-5)

When ionization takes place, the free electrons often have enough energy to

induce a secondary ionization as in equation (2-6).

𝑒− + 𝐻𝑒 → 𝐻𝑒+ + 2𝑒−. (2-6)

These free electrons can recombine with the ionized 4He atoms as shown in equation

(2-7), which will produce additional excited states [44].

𝑒− + 𝐻𝑒+ → 𝐻𝑒∗. (2-7)

These excitations will lead to the production of singlet (more probable) or triplet

(much less probable) excimer states [45]. The decay of these excimers to the ground

state is an important step to produce scintillation photons, which are either shifted in

spectrum through wavelength shifting interactions, or directly detected by

photomultiplier tubes (PMTs) at either end, as shown in Figure 2-4. The next chapter

covers additional details about this scintillation process.

40

Figure 2-4. Yinong Liang. A 4He fast neutron detector. 2016.

The 4He detector technology has several advantages over current neutron and

gamma-ray detectors used for nuclear security and safeguards. When comparing with

the above-mentioned 3He detector and its potential alternatives, 4He is much more

available and economical than 3He gas [46], and has a better ability to retain energy

information of the neutron interactions. Due to its low electron density, 4He has limited

sensitivity to gamma-ray radiation, which in combination with pulse shape discrimination

results in excellent gamma-ray rejection. Furthermore, gamma-ray interactions deposit

lower energies and have lower light yield in gaseous helium scintillation when compared

to neutron interactions, which can be utilized for pulse shape discrimination. Finally, the

detector’s performance has been shown not to degrade over time due to high intensity

radiation, such as that from spent nuclear fuel. Overall, the superior gamma-ray

rejection capability and a more rugged design makes 4He detectors a potentially very

useful tool in spent nuclear fuel monitoring and neutron spectroscopy analysis. The

detector can take the measured scintillation light as input and through unfolding

reproduce the incident neutron spectrum, which serves a way for “fingerprinting” the

spent fuel storage casks. In addition, the 4He detectors allow for coincidence

measurements of neutrons and gamma rays, which is well-known in terms of identifying

41

[47] and characterizing [35] nuclear materials. Other related applications in the areas of

radiation detection and instrumentation, including fusion diagnostics [48], nuclear

safeguards [49], and medical physics [50], can all benefit from the 4He-based neutron

detection techniques.

42

CHAPTER 3 4He DETECTORS RESPONSE CHARACTERIZATIONS AND SPECTRAL

UNFOLDING ALGORITHM

Neutron spectrometry is a desirable measurement technique for nuclear material

verification and safeguards. A historic review by Brooks and Klein presented seven

methods of neutron spectrometry [51], such as kinematic measurements of recoil

particles, charged particles released in neutron-induced reactions, time-of-flight (TOF)

measurement of neutron velocity, the mathematical unfolding of the responses of a

neutron detector which are neutron energy-dependent, etc. Before utilizing neutron

spectroscopic techniques with the 4He detectors, it is necessary to perform detector

characterizations as the first step.

In this chapter, the detector’s scintillation process, PSD performance, and light

responses are covered first. The aforementioned methods for neutron spectroscopy

applications are all employed. Through the TOF measurements at the John E. Edwards

accelerator laboratory at Ohio University (OU), the detectors response matrix was

obtained. Then a literature review of current spectrum unfolding methods is discussed,

followed by a description of the iterative-improvement-based quadrature method

developed for this project. Finally, three measurements are conducted with a 2.45

MeV mono-energetic (D-D neutron generator), a 252Cf spontaneous fission, and a Pu-Be

(α,n) neutron source. The unfolding algorithm is then tested with the experimentally

measured response matrix. The results demonstrated the detectors potential ability to

differentiate various neutron sources and predict the incoming neutron spectra based on

the measured light output distributions.

43

3.1 4He Detectors Characteristics

As previously mentioned, when a neutron scatters with the 4He nucleus, the

kinetic energy transferred to the 4He nucleus is sufficient to strip away the electrons.

The recoiling nucleus (i.e. α-particle) will then either excite or ionize other helium atoms

along its path within the detector volume. 4He is a relatively efficient scintillator which

can produce approximately 15,000 scintillation photons per MeV of energy deposition

by neutrons [13]. The detector body is made of stainless steel, and the gas pressure of

the 4He detectors is about 150 bar. Based on the scaling law of range calculation, for a

2 MeV, 4 MeV, and 6 MeV alpha particle, the range in the detector is about 0.03 cm,

0.07 cm, and 0.13 cm respectively, which is much smaller than the radius of the gas

chamber (2.2 cm). This results in most neutron scatter events depositing their energy in

the active volume as opposed to suffering from wall-effects. Scintillation photons are

generated from the de-excitation from either single or triplet states of the helium

excimers, and the output signals can be separated into two components: a fast

component which lasts on the scale of tens of nanoseconds and a slow component with

a much longer time scale of microseconds (see Figure 3-1) due to the differences in the

excimers states. The two electrons of 4He are occupied at its 1s orbital with opposite

spins in the ground state. In the single excimer state, one of the electrons is prompted

to a higher orbital without changing its spin orientation, while in the triplet excimer state,

the electron at higher energy level changes its spin orientation, therefore the two

electrons are now occupying different orbitals and have the same spin state [52]. During

de-excitation, the fast component generates a sharp light pulse which results from the

decay process of the singlet excimer states, while the slow component generates

44

multiple discrete short pulses which are produced during the decay process of the triplet

excimer states.

Figure 3-1. Typical digitized PMT outputs of a gamma-ray (left) and neutron (right) detection event. The fast component lasts approximately 50 ns, followed immediately by a slow component till the end of the 4.27 μs pulse event window.

The scintillation photon wavelength is approximately 80 nm, but the light is

collected by two Hamamatsu R580 photomultiplier tubes (PMT), with spectral sensitivity

mostly in the range of 200 to 900 nm [53]. Therefore, a wavelength shifting (WLS)

material is coated on the inner wall of the gas chamber to change the wavelength of the

emitted light. The WLS material absorbs the scintillation light coming directly from the

helium gas and re-emits it at a longer wavelength (lower energy), which better matches

the range of sensitivity of the PMT photocathode (Bialkali). The two PMTs on both ends

of the gas chamber are set up in coincidence mode so as to suppress the recording of

PMT noise signals. PMT calibration was performed to ensure the two PMTs have

matching gain factors and the amplified scintillation signals do not exceed the limited

dynamic range of the digitizer [54]. Field programmable gate arrays (FPGAs) within the

DAQ search for coincident signals within 32 ns from the two PMTs. The PMT signals

are digitized by bespoke data acquisition cards with two different sampling frequencies,

45

125 MHz (for the whole pulse) and 1024 MHz (for the fast component only), respectively

[55]. When a particle interaction generates such a coincidence the FPGA reads out a

switched capacitor array [56] with the stored waveform.

The 4He detectors exhibit excellent gamma-ray rejection due to several physical

aspects [19]. First, gamma rays have a relatively low interaction probability. Contrary to

neutrons that interact with the nucleus of the atom, gamma rays interact with the atom's

orbiting electrons. Helium's low atomic number (Z=2) results in a low electron density,

which reduces the probability of gamma-ray interactions [57]. Second, gamma rays can

only deposit relatively small amount of energy in the detector volume. Gamma-ray

interactions in the active volume produce recoil electrons, and the energy loss of

recoiling Compton electrons in the 150 bar 4He gas is about 40 times lower than in an

organic scintillator. The electron will travel much farther while slowing down and will be

more likely to hit a wall before depositing all of its energy [58]. Third, unlike liquid or

plastic scintillators, gamma rays also have a lower light yield in gaseous helium when

comparing to neutron interactions, due to inefficient recombination of electron-ion pairs.

Previous results [19], show that the number of photoelectrons detected in the slow

component is approximately a factor 3-4 larger than gamma-ray events, allowing for

efficient pulse shape discrimination. This is illustrated in Figure 3-1, which shows the

signal trace from a neutron event and a gamma-ray event for comparison. By

comparing the integrated slow and the fast component of each event, pulse shape

discrimination (PSD) can be performed. Figure 3-2 presents a scatter plot of the fast

component against slow component of the TOF measurement at OU. Events above the

black cut-off line will be regarded as neutrons and used for building the detector neutron

46

response matrix. Pulse filtering algorithms [54] are applied to remove pile-up events (i.e.

two events within the same event window) and other undesirable hard-to-analyze

events.

Figure 3-2. Scatter plot of fast component against slow component of the TOF measurement. Nonlinearities in electronics were reduced by utilizing a low gain settings and the pulse post-processing method to appear outside of the energy range of the TOF measurement (10 MeV).

In addition, in Figure 3-3, the scatter plots are analyzed within different neutron

energy ranges, showing the same number of pulses in each plot. The integrated slow

component increases as the neutron energy increases, while most gamma rays are

observed after neutrons reach 2 MeV. For neutrons below 1 MeV, the strengths of both

slow component and fast component are comparable with gamma rays, therefore no

clear separation in the PSD plot can be observed. As previously mentioned, the

detector has the highest neutron detection efficiency from 1 MeV to 2 MeV, thus only a

few gamma rays are detected within that energy range. The neutron detection efficiency

then drops after 2 MeV, and that is the reason for observing a larger fraction of gamma

rays. On the other hand, differentiating events by their TOF is a reasonable way to filter

47

out most gamma rays, and the charge integration based PSD can serve to additionally

strengthen the results.

Figure 3-3. Scatter plot of fast component against slow component of the TOF

measurement within different neutron energy ranges (10,000 number of pulses in each plot).

No moderating material is needed for detecting fast neutrons. Therefore, the

incident neutron energy can be retained. This energy deposition governs the quantity of

photons emitted, and the amount of light detected from a neutron event is related to the

incident neutron energy. Figure 3-4 demonstrates the 4He detectors energy

discrimination ability in which generator-produced 2.45 MeV neutrons are used to

induce uranium fission from natural uranium samples. Given the higher energies of

induced fission neutrons than the interrogating 2.45 MeV neutrons, the 4He detectors

are able to detect these high-energy neutrons unambiguously. The results highlight the

48

detector’s potential application in active interrogation of cargo containers for the

detection of fissile material.

Figure 3-4. Induced fission neutrons (by neutron generator active interrogation on

natural U samples) are distinguishable from generator neutrons in pulse height spectrum comparisons.

Through elastic scattering, an incident neutron transfers up to 64% of its energy

to a recoil α-particle depending on the elastic scattering angle distribution due to

kinematics. However, the relationship between energy deposition and scintillation light

production is not linear due to scintillator and electronics effects, which are known as

the electronics nonlinearity and scintillator nonlinearity [59]. From previous

measurements [19], the curved shape of the neutron bend in Figure 3-5 is likely caused

by nonlinearities in the data acquisition system of large signals such as saturation and

after-pulsing, which happen more frequently for neutrons of relative high energy. The

underlying relationship between the deposited neutron energy and the scintillation light

in terms of the slow component is shown in Figure 3-5 as obtained from the same

measurement as the response matrix measurement below. An empirical power function

fitting is chosen here over the linear function fitting, since the former is better anchored

49

in physics (i.e. zero energy deposition leads to zero production of scintillation light),

while the linear function fitting results in a non-zero y-intercept.

Figure 3-5. The scintillation light (in terms of the slow component) vs. the deposited energy.

Given the deposited energy is probabilistic and proportional to the scintillation

pulse heights, the next step is to determine the underlying relationship. The so-called

detector response matrix (or response function) can be constructed to map the incident

neutron energy to the detector scintillation light output, where the incident neutron

energy is determined through time-of-fight measurements. The TOF measurement is

critical to this research and will be further discussed in the following sections.

3.2 Time-of-Flight Measurement and Detector Response Matrix

The incident neutron energy can be calculated as in Equation 3-1 as below:

𝑣 =𝐿

𝑡 𝑎𝑛𝑑 𝐸 =

𝑚𝑛 ∙ 𝑣2

2,

(3-1)

where the time (𝑡) taken by neutrons to travel a known distance (𝐿) is tracked and used

to calculate the neutron velocity (𝑣) and energy (𝐸).

50

The TOF measurements were performed using the Tandem Van de Graaf

accelerator at the John E. Edwards accelerator laboratory at Ohio University. The Van

de Graaf accelerator can accelerate protons, deuterons, or heavy ions on to various

targets, such as 10B, 27Al, and 9Be, producing monoenergetic neutrons as well as

neutrons with continuous spectra. In this measurement, a 7.5 MeV deuteron beam was

pulsed and bunched by a double klystron buncher to produce pulse widths of < 2 ns at

the beryllium metal target at 60 [60]. Neutrons were produced via Be (d, xn) reactions

with a continuous neutron spectrum up to 10 MeV. The generation of neutrons followed

the pulsing frequency of the deuteron beam, with 1600 ns between pulses (0.625 MHz).

For the chosen flight path, this corresponds to the slowest neutrons from each pulse

that can reach the detector before the next pulse, to have an energy around 200 keV,

below which the 4He (n, elastic) interaction probability is lower and the light pulses

produced are nearly indistinguishable from signal noise. Associated with the deuteron

beam operation, gamma rays are also detected from natural gamma-ray background as

well as gamma rays from beam activation of the target. Gamma rays travel at the speed

of light and therefore, as shown on the left of Figure 3-6, the TOF spectrum has a main

peak (so-called gamma flash), followed by neutron events with various energies along

the 1600 ns window. Ideally, for travelling a known distance of 10 m to the detectors,

these gamma rays should be observed at 33 ns in the TOF spectrum. However, the

beam pick-off signal that represents the start of each deuteron pulse (the time when

deuterons hit the target) is in fact triggered from a certain position earlier in the

accelerator beam-line. Combined with the delays from cables of various lengths and the

uncertainties and electronic noises from data acquisition modules, the result is a

51

constant time deviation from the theoretical gamma flash position at 33 ns. This time

deviation can be corrected by applying a constant shift to the each TOF value. As in

Figure 3-6, the timestamp of each event is corrected, and is used as the reference

timestamp of neutrons originating from the target. From each deuteron beam pulse,

neutrons of different energies are generated and collimated down the TOF tunnel,

reaching the detector at different times. The TOF is calculated by subtracting the

reference timestamps of neutron events in the detectors and the beam pick-off signal.

The timing resolution can be determined by the FWHM of the gamma flash. It is

calculated as 14 ns and can then be used to estimate the energy resolution of the TOF

measurement. Due to the relative low sampling resolution (8 ns) of the DRS4 waveform

digitizing chip of the data acquisition system [19], interpolation between two adjacent

timestamps is chosen over Gaussian fitting which may contain large uncertainties. The

relative energy resolution is estimated to be 9.5% at 6 MeV, 5.6% at 2 MeV, and 4.0%

at 1 MeV, which accounts for the uncertainties in both neutron flight time and its travel

distance. More details regarding the detector settings and accelerator facility

descriptions can be found in previous measurements [2].

52

Figure 3-6. The TOF spectrum with about 6⨯105 events (left), and the experimental neutron flux from the 9Be (d, n) reaction (right), incident on the detector volume.

The target used in this work was 9Be, which produces neutrons of energies up to

10 MeV. Figure 3-6 (right graph) shows the expected incident neutron flux (integrated

over the detector solid angle) from the 9Be (d, n) reaction. The expected neutron flux is

determined in previous work [61], using existing, calibrated neutron detectors at the

accelerator facility.

By calculating the ratio between the number of detected neutrons and expected

neutrons as in Figure 3-6, the energy-dependent efficiency is shown as in Figure 3-7.

The maximum efficiency was 6.8% at 1.1 MeV, due to the large neutron scatter cross

section for 4He around that energy (shown in Figure 2-3). The bump in the efficiency

curve around 3 MeV may arise from fluctuations in the reference neutron spectrum

which were measured by the calibrated detectors as well. In addition, the efficiency

obtained from MCNPX PoliMi Monte-Carlo simulation is plotted and compared with the

measured efficiency. A beam of 20 groups of monoenergetic neutrons ranging from 0.5

MeV to 10 MeV (increased by 0.5 MeV) were simulated, where neutrons will travel 10 m

in air before hitting the detector. At 150 bar, the density of 4He gas is about 0.02464

53

g/cm3, and default MODE: N is used in the simulations. The ENDF-based MCNP

libraries is chosen for neutron evaluations up to 20 MeV, which contains information on

neutron collisions, such as cross-sections and outgoing neutron energy [62]. Elastic

scattering reactions only result in energy transfer without any other secondary effects

[63]. Moreover, in this simple check, we are only concerned about the numbers of

neutrons interacting with 4He nuclei and their energy depositions (used for the

construction of the kinetic response matrix), therefore it is acceptable that the emission

data for charged particles or recoil nuclei is not produced and tracked when using

standard neutron libraries. The outputs from the simulations contains statistical errors

corresponding to one standard deviation [64], and the uncertainties in measured

efficiency are calculated via error propagation, which result in the error bars shown in

Figure 3-7. Both curves have a peak around 1-2 MeV, while the simulated one has

higher efficiency values below 2 MeV. It is likely the result from the conservative PSD

algorithm, which sacrifice a considerable amount of low energy neutrons due to their

relatively small pulses. In addition, the simulation simplified the geometry configurations

by assuming air only between the neutron source and the detector, while in reality,

neutrons may suffer from random scatters within the collimator or tunnel. The relative

large differences between measured and simulated efficiency agree with a previous

study [2], at 1 MeV, approximately 6.6% detection efficiency was measured while the

simulation predicted a 9% detection efficiency.

54

Figure 3-7. Black line: measured 4He detector intrinsic efficiency as a function of incident neutron energy from TOF measurement. Red line: simulated 4He detector intrinsic efficiency as a function of incident neutron energy from MCNPX.

The detector response matrix is constructed on a 2-D histogram where the two

base axes are the incident neutron energy and the detector scintillation light outputs.

The relationship in terms of either total scintillation light output (i.e., slow component

plus fast component) or just slow component only is shown as in Figure 3-8. Large TOF

values correspond to low energy neutrons, and as expected, low energy neutrons tend

to produce smaller scintillation light pulses.

Figure 3-8. Slow component only vs. TOF (left), and total scintillation light vs. TOF (right) after PSD. Visually, the summation of fast component and slow component (i.e. total scintillation light outputs) does not improve the correlation.

55

It was reported in [19] and confirmed in Figure 3-8, that the distribution of slow

component values is correlated with neutron energies. Additionally, Ting [54] addressed

that the fast components of the detector scintillation lights may introduce large

uncertainties when doing the pulse integration. Fast components have short time scales

on the order of nanoseconds and tend to exceed the dynamic range of the digitizer,

depending on both gain settings and the radiation energy. Therefore, the response

matrix is constructed by using slow component values only. Each neutron event can be

defined by its initial energy which is calculated from its TOF, and by the scintillation light

(slow component) it produced, which is measured by the 4He detector. The slow

component is divided into 1000 bins to reduce the statistical fluctuations since even

small variations in the measured light output distribution, such as interfering background

radiation, will result in wide fluctuations which can be observed in the unfolded neutron

spectrum. The incident energy is divided into 20 bins (0.5 MeV bin width) to account for

the detector’s energy resolution (in another words, timing and flight-path distance). All

the events are then grouped into corresponding bins which creates a bivariate

histogram of 1000-by-20 bins with normalized counts. It describes the unnormalized

distribution function of events, “R(L, E)”, as a function of the energy (E) and slow

component of the scintillation light (L). The final detector response matrix is given in

Figure 3-9.

56

Figure 3-9. Response matrix for 4He detector from TOF measurement.

In addition to the TOF measurement, the detector response matrix can also be

obtained from simulations. The same simulation model is used as in the efficiency

simulation. The deposited energy in one elastic scattering reaction is a function of

scatter angle and is calculated as the difference between the incoming and outgoing

neutron energy. For each quasi-monoenergetic neutron group, the distribution of energy

deposition is plotted with 0.5 MeV bin width. By combining all groups of neutrons, the

kinetic response matrix can be constructed, which shows a probabilistic mapping

between the incident neutron energy and the deposited neutron energy as in Figure

3-10 (left) as the first step. The deposit energy can then be converted to scintillation

light by the relationship as shown in Figure 3-5, and the final simulated response matrix

is shown in Figure 3-10 (right). The simulated response matrix has the same diagonal

features as the measured response matrix and tends to yield larger scintillation light

along the whole energy range. The reason is that in real measurements, because of

photon-statistics and detectors resolution, the scintillation light output distributions will

be broadened and smoothed. While in simulations, the resolution of the detector has to

57

be based on empirical data [19], and has not yet been added to the simulation process

because we are not able to quantify it at this point, thus the simulated response matrix

represents an ideal light distribution. Nonetheless, it is a good rough comparison and

serves as a verification of the measured response matrix. The following unfolding

process will therefore still use the measured response matrix.

Figure 3-10. The simulated neutron energy-deposition distribution (left), and the final response matrix for 4He detector from MCNPX simulations (right).

Similar TOF measurements were also taken at our lab at the University of Florida

with a 73.7 Ci 252Cf spontaneous fission source. A Struck digitizer system with a 250

MHz sampling rate (4 ns sampling frequency) was used to increase the timing

resolution, and therefore the calculated energy resolution. As shown in Figure 3-11

(left), the FWHM of the gamma flash is narrowed to 6 ns, and the energy resolution for 1

MeV neutrons is enhanced to 1.9%. However, significant number of neutrons suffered

from random scatters in the lab environment, and these neutrons account for the

spurious peak at the very beginning of the calculated neutron spectrum as in Figure

3-11 (right).

58

Figure 3-11. The TOF spectrum measured with a 73.7 Ci 252Cf spontaneous fission source containing about 2⨯106 events (left), and the calculated incident neutron spectrum (right). The spurious peak at the very beginning of the calculated spectrum results from random scatters.

The final response matrix is created with 2500-by-36 bins as shown on the left of

Figure 3-12. Due to the improved timing resolution, the energy bin width is reduced to

0.25 MeV. All the neutrons lower than 0.25 MeV are rejected and are not used when

constructing the response matrix, since we are now able to identify true neutron events

from the random scatter events. In addition, the detection efficiency for neutrons below

0.25 MeV is very low (below 0.1%), therefore even if the true neutron events within that

energy range can be successfully identified, the count would be quite low. The lab-

measured response matrix is tested for unfolding the same 252Cf source. As on the right

of Figure 3-12, no significant improvement of the unfolded spectrum is found comparing

with the results obtained from the response matrix measured at the accelerator facility

(see Figure 3-16 in next section). Thus, even though the energy resolution is improved,

when considering the long measurement time for response matrix construction

(approximately a factor 4 longer than the Ohio measurement for achieving same

59

number of total events) and the artificial terms in the response matrix, we decide to

keep using the response matrix measured at the accelerator laboratory at OU.

Figure 3-12. Response matrix measured from a 252Cf spontaneous fission source with 0.25 MeV bins (left), and the unfolded spectrum for the same 252Cf source (right).

3.3 Spectral Unfolding with 4He Detectors

The detector light output is determined by both the incident neutron energy and

the detector’s internal conversion between energy and light. This information is

contained within the detector’s response function. Spectrum unfolding is a de-

convolution process, using both the detection response function and the measured

scintillation light output [65]. However, two main problems limit the accuracy of

unfolding. First, the output response function requires comprehensive measurements

and simulations to determine the response function between an incident neutron of any

given energy and the corresponding light outputs. Secondly, these measurements and

simulations have unavoidable uncertainties, which are included in the response

function. Due to these inherent uncertainties, wide fluctuations can be observed in the

unfolded neutron energy spectrum due to even small variations in the measured light

output distribution, such as interfering background radiation.

60

The relationship between the incident neutron spectrum 𝑥(𝐸𝑛) and the resulting

light output spectrum 𝑁(𝐿) in a scintillator can be written as a Fredholm integral

equation as in Equation 3-2:

𝑁(𝐿) = ∫ 𝑅(𝐿, 𝐸𝑛)𝑥(𝐸𝑛)𝑑𝐸𝑛 (3-2)

Where 𝑅(𝐿, 𝐸𝑛) is the detector response function, which represents a correlation

between the neutron spectrum and the light output spectrum. For a known 𝑅(𝐿, 𝐸𝑛), an

estimate of 𝑥(𝐸𝑛) can be obtained.

3.3.1 Review of Current Spectrum Unfolding Methods

Unfolding procedures have been widely used to compute neutron and gamma-

ray spectra from experimental detector outputs. Any practitioner must face the question

of how to determinate whether the solution is correct within reasonable uncertainties

[51]. The main problems when doing spectrum unfolding are: a). how to standardize and

optimize the measurement system. b.) how to standardize and optimize the unfolding

procedures. In order to addressing these problems, many unfolding methods have been

developed and can be classified into four broad categories [66].

The first category is based upon Monte Carlo, in which a “candidate spectrum” is

randomly selected from specific distributions. The selected spectrum is then convolved

with the detector response function and will produce an estimated detector light output.

If this estimated detector light output is in sufficient agreement with measured detector

output, then the “candidate spectrum” is accepted for use in computing an average

spectrum. If the deviation between the estimated detector output and the measured

detector output exceeds a certain tolerance, then the current “candidate spectrum” will

61

be rejected and another will be selected. Only one RSICC code neutron spectrum

unfolding code, known as SWIFT [67] is in this category.

The second category is labeled parametric representation and contains

procedures which may be used when a functional representation of the neutron

spectrum is available (e.g. the fission spectrum). Users then can determine the

parameters by matching the measured detector output. This procedure has many

restrictions on computing the spectrum and is used only when those restrictions can be

justified.

The third category is the derivative methods. It uses an approximation to cause

the neutron spectrum to become a readily determined explicit function of the measured

detector output by assuming the detector’s response to a mono-energetic source to be

a step function at first. DUFOLD, NUTSPEC, and STUNFO [68] are based on this

method.

The last category for unfolding is the quadrature method, upon which the

algorithm used for this project was built. The integral Equation (3-2) is replaced with a

quadrature form, breaking each parameter into several groups. The problem then

becomes solving an ill-conditioned linear system.

Linear estimation methods, mathematical programming methods, and iterative

improvement methods are the three main classes in the category. SPECTRA and

CRYSTALL BALL [69] are examples.

3.3.2 Iterative Least Squares Unfolding Algorithm and Uncertainty Estimation

As shown in Equation (3-2), when given a R(L, En ), an estimation of x(En) can be

obtained. However, the resolution of x(En) is limited by discrete binning of R(L, En).

62

Equation 3-3 shows the matrix form of Equation 3-2, with K groups discrete intervals of

energy E, and M groups of light output N(L).

[𝑁1

⋮𝑁𝑀

] = [

𝑅1,1 ⋯ 𝑅1,𝐾

⋮ ⋱ ⋮𝑅𝑀,1 ⋯ 𝑅𝑀,𝐾

] × [

𝑥1

⋮𝑥𝐾

] (3-3)

The bin width of the light output can be adjusted. Based on our experiences, it is

important to have suitable K and M in order to have optimal solutions. When the number

of light output bins (M) is large, the response matrix is nearly singular (since R is not

square, we multiplied RT on each side of Equation 3-2, then RRT can be regarded as

symmetrical), where one or more of the singular values are very small. Therefore, the

results from matrix inverse could contain large numerical errors due to the small

singular values [70] and the weak convergence [71]. On the other hand, if the number of

light output bin (M) is comparable to the number of the energy bin (K), the changes in

N(L) from various neutron sources will not be sufficient enough to reflect the

characteristics of the incident spectra. In Figure 3-13, we tried to unfold the 2.45 MeV

monoenergetic neutron source with different light output bins due to the prior knowledge

of the spectrum. For small M (M=250), we obtained a larger amount of zero solutions,

which indicates that for different neutron energy groups, the detector’s light responses

are roughly the same. Therefore, once the group with the highest probability for

producing such N(L) is identified, the rest of the energy groups will be of zero weight.

While for large M (M=2000), the unfolded spectra contain certain spurious peaks.

Considerable fraction of neutrons is predicted with energy higher than 2.45 MeV, which

disagrees with reality (in absence of pulse pile-up or high-energy background events).

After applyingd possible combinations of M and K, we construct the response matrix

63

with 1000-by-20 as in previous section, and each event from the TOF measurements is

then sorted into the corresponding energy and light output bin.

Figure 3-13. The unfolded D-D spectra when using 250 (left) and 2000 (right) scintillation light output bins.

The basic solution to the matrix inverse problem can be obtained by commonly

known computational algorithms. However, the direct inversion can produce negative

solutions of flux, which are nonphysical in our unfolding problem of real neutron spectra.

Therefore, it becomes necessary to develop an optimization method with constraints.

Matlab can solve least squares problems with bounds or linear constraints,

providing a solution that will be regarded as an initial “guess spectrum”. The “guess

spectrum” is non-negative and obtained directly from matrix inversion. Upon which, the

iterative Least Square Method (LSM) algorithm [72] will be applied to find the best 𝑥𝑗 by

iteratively performing the following (a flowchart is also shown in Figure 3-14):

1. A reasonable “guess spectrum” xj is found, resulting from the direct inverse of the least square solution.

2. Set L= 0. Minimize the value of Equation (3-4) below by using conjugate gradient

method. Wi are set as the inverse square roots of the counts of each light output

bin, the gradient of Equation 3-4 is simplified as 𝛻𝑓(𝑥) = 𝑁𝑖 − 𝑅𝑥.

64

𝑓(𝑥) = ∑ 𝑤𝑖(𝑁𝑖 − ∑ 𝑅𝑖,𝑗𝑥𝑗)

𝐾

𝑗=1

2𝑀

𝑖=1

(3-4)

3. Set 𝑝0 = −𝑟0, 𝑟0 = 𝑅𝑥0 − 𝑁, and iteratively find the next direction (pL) where the gradient is negative and step (αL) where the residual reduction is minimized by using previous pL and αL.

4. Terminate when the residual reduction is smaller than the change in the solution.

Figure 3-14. A flowchart of the iterative unfolding algorithm.

The solution of xj will eventually converge to a most optimal solution that is both

positive and having a minimized f(x). This method uses the solution of the matrix direct

inversion as a first “guess”, then optimizes it via the iterative improvement-based

quadrature method which is widely known for solving ill-conditioned problems [66].

The uncertainties in the unfolded spectra could come from both the measured

response matrix R(L, En) and the light output spectrum N(L). The uncertainties

associated with N(L) are from detector readings and can be determined by Poisson

statistics [73]. While the uncertainties in the response matrix have to be either

65

calculated from error propagations, which requires matrix derivatives [74] and the

knowledge of the uncertainty matrix (covariance matrix) [73], or computed from existing

unfolding codes such as MAXED [75] and FERDO [76].

In this work, we applied a stochastic method in order to overcome the lack of

knowledge of the detector’s energy resolution and the unavailability of the a priori

information of the response matrix and particle fluence in the above methods. When

building the response matrix, neutrons are sorted into corresponding energy and light

output bins, and their energies are calculated based on the TOF. Therefore,

uncertainties in the response matrix are in part caused by timing uncertainties of the

TOF measurement. Our work [77] shows that the 4He detector has a 3.229 ns full width

at half maximum (FWHM) time resolution (modelled by a Gaussian fit) at similar

detector settings as the TOF measurement. Consequently, a time spread following a

Gaussian distribution with 3.229 ns FWHM is randomly added to each TOF value and

propagated to the calculated neutron energy and the resulting response matrix. The

uncertainties of the counts in each light output bins are added to N(L) in the same way,

based on the assumption of Poisson statistics in the nuclear detection counting system

[27]. 10,000 trials were conducted, resulting 10,000 response matrixes and therefore

10,000 unfolding results. The uncertainties are calculated as the maximum and the

minimum values of the unfolded spectra.

Three measurements were conducted with a 252Cf spontaneous fission neutron

source, a Pu-Be (α, n) neutron source, and a deuterium-deuterium (D-D) fusion-based

neutron generator to test the iterative LSM with the experimentally measured response

matrix.

66

The sample data is collected separately by using a 115 μCi 252Cf source and a 10

Ci Pu-Be source, which are placed at 35 cm from the center of the detector's active

volume. Within 2 minutes, a total of 5.02⨯104 and 1.11⨯105 counts (before PSD) are

collected from the 252Cf source and Pu-Be source respectively. Monoenergetic neutrons

of energy 2.45 MeV from the University of Florida's D-D neutron generator were also

measured. The detector is placed 1 m from the generator, and 7.65*104 counts are

collected within 2 minutes. The same gain setting (relatively small gain was used to

avoid pulse saturation) and the same pulses filters are applied as the TOF

measurement at OU to maintain consistency. All the measurements are taken at room

temperature. No additional gamma-ray shielding materials are used during the

measurements due to the gamma-ray insensitivity of the detectors.

Figure 3-15 shows the light output spectra after pulse shape discrimination. The

D-D monoenergetic 2.45 MeV neutron source produces overall the smallest amount of

scintillation light, while for the (α, n) and spontaneous fission neutron source, one could

not tell much information from the detector pulse height spectrum directly, yet the

unfolded neutron spectra revealed unique features as shown in Figure 3-16. As

mentioned above, no a-priori information about the incident neutron spectrum is

required, which mimics the blind-measuring case during nuclear materials monitoring.

During the spectrum unfolding, non-linear bin width was chosen to lower the impact of

the statistical fluctuations especially for large scintillation light amplitudes. More

fluctuations are observed for amplitudes over 30,000 (a.u.) in the light output spectra,

but they only account for 6% of the total events. Overall, the unfolded 252Cf spectrum

exhibits a factor of 1.38 discrepancy from the Watt distribution of a typical spontaneous

67

fission source [15], which has a maximum probability around 1 MeV and decreases

gradually afterwards. These characteristics can be used for distinguishing a fission

spectrum from (α, n) neutrons, which usually have a higher yield between 2.5 MeV and

3 MeV based on previous work using simulations [4], as well as shown in the unfolded

spectrum. Information for the specific Pu-Be source is limited, therefore we are not able

to plot a reference spectrum. The unfolded spectrum of the D-D measurement has a

peak at 2.5 MeV with a relatively small uncertainty. 53% of the neutrons are lower than

2.5 MeV, which may have resulted from some of the neutrons being slowed down by

scatter and room return, as well as the neutron energy bin edge being relatively close to

the actual 2.45 MeV neutron energy. And there is only a minor predicted fraction of

neutrons with energy higher than 3 MeV which agrees with the expectation. In addition,

there is another way to evaluate the unfolded spectra in terms of their average energy

and high energy neutron ratio. The Pu-Be unfolded spectrum yields an average energy

of 4.35 MeV, and 50.77% neutrons are higher than 4 MeV. While the 252Cf unfolded

spectrum has a 2.63 MeV average energy, and 21.91% neutrons are higher than 4

MeV. These features provide a proof-of-concept to identify and verify various neutron

source types. This could be used for example for spent nuclear fuel monitoring, where

the spontaneous fission neutrons and the (α, n) neutrons have their unique contributions

to the overall neutron emission spectrum, as a function of fuel burnup and cooling time

[4].

68

Figure 3-15. The light output spectra of the three measurements.

Figure 3-16. The unfolding results (left), and a zoom in of the unfolded spectra (right), of the three measurements.

To conclude, the work here shows the 4He detector's characterization and its

potential applications as a neutron spectrometer. An experimentally determined neutron

response function from the TOF experiments was used and an iterative least square

unfolding algorithm was developed to obtain the neutron spectra from a 252Cf

spontaneous fission source, a Pu-Be (α, n) neutron source, and a 2.45 MeV

monoenergetic D-D neutron generator. Notable discrepancies and uncertainties in the

unfolded spectra were found, yet expected characteristics for different types of neutron

sources were observed and readily distinguished. The recoil spectrum unfolding is

69

unlikely to match the incoming spectrum as perfectly as certain detectors designed for

spectroscopy applications [78], but it has the benefit of directly using the existing

detectors without complex electronics. In this work we are focusing on differentiating

various neutron sources for spent fuel monitoring purpose, where the main contributions

of neutrons are from spontaneous fissions and (α, n) reactions. While separate neutron

sources looked significantly different, the results and discrepancy of unfolded spectra

show that mixed neutron sources of small difference would be very hard to differentiate

using the detectors used here, likely significantly impacted by a relative poor resolution

of the He-4 detectors. The results addressed the advantages of the novel 4He fast

neutron scintillation detectors and supported the proof-of-concept idea of using the

detectors to verify the content of the spent fuel dry casks. Other applications in the

areas of nuclear nonproliferation and homeland security can also benefit from this work.

Future work could include the comparison of the iterative LSM with other neutron

spectrum unfolding codes such as FERDO [76] and GRAVEL [79], multi-source

measurement and unfolding, and ultimately, the development of neutron analysis

system with quantifiable signatures outputs for spent nuclear fuel monitoring [18].

70

CHAPTER 4 TIMING RESOLUTION MEASUREMENT OF PMT AND SIPM-BASED 4HE

DETECTORS

In addition to the detector’s light output, the time response is another important

property for scintillation detectors. The location where the scintillation photons are

generated, the differences in scintillation mechanisms, and the effects of self-

absorption, reemission, and multiple light reflections can all lead to the spread of the

arrival time of the photons at photomultiplier tubes. In this chapter, we discuss the

motivation of characterizing timing resolution of the 4He detectors at first, followed by

the details of the materials and setup of the measurement. A second version of the 4He

detector is introduced, which uses silicon photomultipliers (SiPMs) to detect the light.

The SiPM-based 4He detectors have a longer gas chamber (with 3 segments) than the

PMT-based version, and almost all the volume is active fill gas that enables neutron

detection.

The timing resolutions of both types of 4He detectors are measured, and the

effects of sampling rate and constant fraction discriminator’s ratio of the digitizer

systems are evaluated. For the PMT-based detectors, measuring timing resolution is

essential in order to estimate the accuracy of the TOF affecting the uncertainties of the

response matrix and the unfolding results. For the SiPM-based 4He detectors, even

though they are not the focus of this thesis, the improved detection efficiency and

spatial-dependent light responses can benefit advanced SNF monitoring such as

nondestructive imaging measurements. Therefore, measuring their time responses can

be useful for further research. Even though they have significantly worse detector timing

performance as shown later.

71

4.1 Why Characterize the Timing Resolution

Gamma rays interact with electrons and produce ionization within the scintillator.

Neutrons can produce ionization by interacting with atomic nuclei through a variety of

elastic and inelastic processes. The result of these processes is the production of

energetic electrons-ion pairs with excitation energy, and scintillation photons will be

produced during de-excitation. Generically these processes are fast, therefore

scintillation detectors are chosen over gas-filled detectors and semiconductor detectors

when fast timing is required. As for the 4He detectors, recently they have been

considered for the use in time-sensitive measurements, such as TOF and fast neutron

multiplicity counting. In these measurements, the accuracy of the data is highly

influenced by the detector’s timing performance.

The TOF measurement can be used as an example. This technique is a method

to determine the kinetic energy of a neutron, based on how long it takes to travel a

known distance. Both pulsed neutron beams generated from accelerators and

spontaneous fission neutron sources can be used in TOF measurements [80]. The

kinetic energy of the neutrons is determined by calculating the difference between the

timestamp reported by the analog-digital-converter (ADC) of each neutron event in the

4He detector and the start time of the trigger signal synchronized with the

accelerator/source. With a known distance between the emission source and the

detector, the TOF is converted to determine the incident neutron energy. The outputs of

the TOF measurement can be used in applications such as detector light response

calibration and mathematical spectrum unfolding [2]. The relative energy resolution of a

TOF measurement is shown in equation (4-1) as a first approximation:

72

∆𝐸𝑛

𝐸𝑛= 2 × √(

∆𝑡

𝑡)

2

+ (∆𝐿

𝐿)

2

(4-1)

As calculated from equation (4-1), a high-accuracy TOF result can be achieved

by the use of long flight paths (large 𝐿 and 𝑡) and the relative energy resolution can be

estimated by knowing the time resolution of the detection system. Previous studies [2],

were mainly focused on the TOF-based detector response characterization and the

spectrum unfolding algorithm development. The accuracy of the TOF affects the

construction of the detector’s response function and therefore the precision of the

unfolding results. For a better understanding, Table 4-1 shows the relatively energy

resolution of a TOF measurement from a previous study [2], given a 3.229 ns timing

resolution. As shown in the table, even with a 10 m long flight path the detectors timing

performance has a significant impact on the accuracy of the calculated incident neutron

energies, especially for high-energy neutrons. Furthermore, for many applications it may

be impractical or impossible to select a long flight-path to reduce energy-uncertainty.

Therefore, it is crucial to characterize the detector’s timing resolution.

Table 4-1. Relatively energy resolution of the TOF measurement (10 m path), given a 4He detector time resolution of 3.229 ns.

TOF (ns) Energy (MeV) ∆E/E (%) ∆E (MeV)

295.2 6 4.25 0.26 361.5 4 3.47 0.14 511.1 2 2.46 0.05 723.0 1 0.87 0.0087

4.2 Experiments

In the timing characterization measurement, a set of digitized pulse signals are

acquired in a coincidence measurement using 4He detectors and EJ-309 liquid

scintillators. Two digitizer systems with different sampling rates are used to fully

73

understand the detectors timing features. The Struck digitizer system has a fixed

sampling rate at 250 MHz with an implemented trigger moving average window (MAW)

algorithm. The constant fraction discrimination (CFD) technique is used to mitigate “time

walks” and correct the timestamp values. Therefore, the timing of the trigger has

minimized pulse amplitude dependence. The DRS4 waveform digitizer system on the

other hand is capable of sampling signals at a frequency ranging from 0.7 GHz to 5

GHz with 1024 sampling points. The sampling rate of 5 GHz is chosen in order to

explore the best timing resolution that the detectors can achieve. The full-width at half-

maximum (FWHM) from Gaussian fits is used to examine the timing resolution. The 4He

detectors have two distinctly different versions (PMT-based and SensL silicon-

photomultipliers (SiPM)-based), both of which are analyzed here. Additionally, the

selection of readout-electronics in form of digitizers can significantly impact especially

the timing accuracy of the detectors. The following sections details the specific setups

used in this work.

4.2.1 PMT and SiPM Based 4He Detectors

As shown in Figure 2-4, the first type of the 4He detector has two Hamamatsu

R580 PMTs installed at both ends of the cylindrical gas chamber (20 cm length). The

PMTs are set up in coincidence mode to suppress the recording of PMT noise signals.

The second version of the 4He detector (detector model: S-670) has a series of

SiPMs, also known as Geiger mode avalanche photodiodes (GAPD) or solid state

photomultiplier (SSPM), immersed inside a longer (60 cm) segmented, 180-bar

pressurized, 4He gas chamber [81]. Figure 4-1 shows a SiPM-based 4He detector, its

segment arrangement, and the readout board location.

74

The detector has 3 segments in total, with 4 pairs of SiPM units in each segment.

The paired-signals are positioned equilaterally across the detector and are summed and

fed into the same output channel. The detector is equipped with a fully integrated

readout electronics board (AD011 readout board). Neutron/gamma-ray discrimination,

detector calibration parameters, and detector control functionality are embedded in the

readout board. Optionally, the detectors can be equipped with an AA121 analog readout

board. It gives access to the individual analogue pulse-shaped signals of the SiPM pairs

and is used in this paper. There are 12 channels (channel 1~4 for segment 0, 5~8 for

segment 1, and 9~12 for segment 2) from the AA121 analog readout board.

Figure 4-1. Arktis Radiation Detectors Ltd. The SiPM-based 4He detector. 2016.

Typically, PMTs have faster signal rise time and lower dark current rates than

SiPMs, but SiPMs have a high quantum efficiency (QE) and gain, are also very compact

and rugged, insensitive to magnetic fields, and can be cost-efficient when fabricated on

a large scale [82].

4.2.2 SIS3316 and PSI Digitizers

The SIS3316 16-channel digitizer card (developed by Struck Innovative System)

has a 14-bit resolution and 250 MHz sampling rate. The ADC signal goes in to a MAW

first, and then a moving average (MA) will be performed over the programmable

peaking time and delayed by adding the peaking time and gap time together.

75

When the CFD feature is enabled, a trigger pulse is generated when the actual

trapezoidal value goes below the half of its maximum value as shown in Figure 4-2 [83].

A more accurate timestamp value is calculated via the digitizer’s built-in algorithm by

interpolating the three MAW values ((maximum value, value after trigger, value before

trigger) from the finite impulse response (FIR) trigger trapezoidal at a 50% CFD ratio

[84]. In addition, during post-measurement pulse analysis, a time interpolation based

Matlab script is also used to enable customizable CFD ratio values.

Figure 4-2. Linear interpolation by using three MAW values.

The sampling rate of the DRS4 Evaluation Board (developed by Paul Scherrer

institute (PSI)) can be modified from 0.7 GHz to 5 GHz as mention above [85]. A 5 GHz

sampling rate was chosen in this measurement, where the time difference in two

adjacent sampling points is 0.2 ns on average. The high sampling rate is used to

investigate sampling rate influence on detectors timing performance. The digitizer is

able to record the timestamps associated with the 1024 sampling points, and the actual

timestamps of each event are calculated from a linear interpolation based Matlab script

at optimal CFD ratio. At 5 GHz, the digitizer is only able to record pulses with length up

to 200 ns.

76

In addition, the pre-shaped slower pulses from the SiPM-based detectors lead to

a much wider (10 times at least) FWHM comparing to the PMT-based detectors. Simply

increasing the sampling rate is not able to narrow the FWHM down noticeably.

Therefore, only the Struck digitizer is used for the SiPM-based detectors.

4.2.3 Timing Resolution Measurements and Calculation

The organic liquid scintillator EJ-309 is used here as the reference. It has a

cylindrical shape of 3” diameter ⨯ 3” height, and is coupled to a 3" diameter ETEL

9821KB PMT. The PMT has a single-electron FWHM of 3.2 ns and transit time of 42 ns.

EJ-309 scintillators are widely used in both neutron and gamma-ray detections. They

are well studied [36] [86] [87], and can be used as a reference when characterizing 4He

detectors. The time resolution of the EJ-309 scintillators is measured first. Two EJ-309

scintillators with the same dimensions were oppositely located at an equal distance to

the source (3 cm from the 60Co gamma-ray source to the face of the detectors). The two

gamma rays from 60Co beta-decay are angularly correlated, with a maximum probability

at 180 degrees [88]. Therefore, the detectors are placed at opposite directions with

respect to the source. Both detectors are set up in time-coincidence, such that each

detector is more likely to detect one each of the two gamma rays emitted from the same

beta-decay.

The experimental setup for measuring PMT-based 4He detectors is basically the

same, except for substituting one EJ-309 to one PMT-based 4He detector. In addition, a

linear fan-in/fan-out unit is implemented. The anode pulses from two PMTs are either

fed into the digitizers directly, or are connected to a Phillips Scientific 740 quad

linear/logic fan-in/fan-out, and summed into one pulse.

77

The specified intrinsic time resolution of the fan-in/fan-out unit is relatively small,

thus it is negligible when comparing with the nanosecond range timing resolution of the

detectors. The output pulses are digitized with Struck and PSI digitizers, respectively,

and are processed off-line to obtain the timestamps of each pulse. The differences of

the timestamp values between the 4He detector and the EJ-309 scintillator is calculated

to obtain the detectors timing resolution. With pulse amplitudes ranging from 50 mV to

1.5 V, the CFD ratio are set from 10% to 60%, with increments of 5%

The SiPM-based 4He detector has a unique design of three optically separated

segments, and for each segment, it has 4 channels each consisting of a SiPM pair.

Similarly, pulses from the 4 channels are either sent into the digitizer directly, or are

summed through the fan-in/fan-out unit first. Three source locations are chosen evenly

along the detector’s active volume, and is moved from the center of “segment 0” to the

center of “segment 2” (see Figure 4-1). The setup is the same as for the PMT-based

detectors, but only the Struck digitizer is used to record the pulses due to longer pulse

shapes. The timing resolution is then calculated and compared at the same CFD ratios

as for the PMT-based 4He detector.

The FWHM of the Gaussian-fitted distribution between the timestamp differences

of the two detectors is used to estimate the 4He detector’s time resolution. The electron

transit time of a PM tube is defined as the average time difference between the arrival of

a photon at the photocathode and the final collection of the subsequent electrons at the

anode. In most timing applications, however, the transit time itself is not the primary

interest, because if it was always a constant value, it would introduce only a fixed delay

78

in the derived signal and can be corrected later in pulse processing. Instead, the spread

in transit time is the more important characteristic.

The distribution of this transit time is called the transit time spread (TTS) and is

an important factor in time-resolved measurements. With the large number of photo-

electrons generated from the interacting 60Co gamma ray, the effect is a more

consistent PMT timing than for a single photo-electron. Depending on when and where

the gamma-ray interaction occurs, there will be a slight difference in the reported

“timestamp” for each pair of coincident gamma rays from the 60Co cascade decay. The

time difference is calculated as:

∆𝑇 = 𝑇𝑖𝑚𝑒𝑠𝑡𝑎𝑚𝑝𝐷𝑒𝑡1 − 𝑇𝑖𝑚𝑒𝑠𝑡𝑎𝑚𝑝𝐷𝑒𝑡2. (4-2)

The time difference can be fitted by a Gaussian function. For peaks whose shape

is Gaussian with standard deviation σ, the full width at half maximum (FWHM) is given

by:

𝐹𝑊𝐻𝑀 = 2σ√2ln(2) = 2.35𝜎. (4-3)

The standard deviation of the timing difference is calculated as below based on

error propagation:

𝜎∆𝑇2 = 𝜎∆𝐷𝑒𝑡1

2 + 𝜎∆𝐷𝑒𝑡22.

(4-4)

Combing all the equations above together, the FWHM for each type of detection

systems can be obtained. In this work, we mainly focused on the “CFD-ratio-dependent”

time resolution and the “location-dependent” time resolution for both summed and

individual signals from the 4He detectors.

79

4.3 Timing Resolution of PMT-Based 4He Detectors

Figure 4-3 shows a schematic of the experimental setups for measuring the

PMT-based 4He detector’s timing resolution. The PMT high voltage is set as 1356V for

PMT1 and 1538V for PMT2, respectively to match the pulse amplitude based on

previous calibrations [54]. A 60Co gamma-ray source is placed 3 cm from the face of

both detectors at the center of the detector’s active volume in order to achieve a higher

count rate for the relatively gamma-ray insensitive 4He detector. Lead bricks are used

for collimation and efficiency purpose. The impact of scattering interactions of gamma

rays with the lead bricks cannot be neglected, and will be regarded as an embedded

uncertainty of the measurement. However, due to the short travel path, the speed of

light, and the overall measured timing resolution, it is near negligible in impact.

Figure 4-3. Experimental setup for studying the PMT-based 4He fast neutron detectors.

30,000 coincidence events are recorded. Pulses from the 4He detectors are

summed through the fan/in-fan/out and then measured in coincidence mode with the

EJ-309 output signals. The time resolution of EJ-309 scintillators is measured at first

with a 1.175 ns FWHM in the region of interest (0.1~1.3 MeVee) at the digitizer’s default

50% CFD, upon which, the FWHM of the 4He detector is calculated as 3.229 ns for the

summed PMTs, 3.338 ns for PMT1, and 3.261 ns from PMT2 as shown in Figure 4-4.

80

The photoelectrons recorded per energy unit with the summed PMT pulses are

doubled, therefore the trigger threshold in ADC units for the individual PMTs is chosen

as 60% (in order to be conservative) of the summed PMTs. When the trigger threshold

is set as the same as the summed signals, 2.736 ns and 2.591 ns FWHM are observed

for PMT1 and PMT2 respectively due to elimination of most small pulse events.

Increasing the photoelectron production tends to lead better time resolution.

However, from the measured data, no major improvement is observed. It could be due

to the slight transit time difference in the two PMTs (about 2 ns). Adding the two signals

without proper alignment could compromise the time resolution. In a test measurement,

when adding cable length to PMT1 to compensate the time-delays between the two

PMTs, a better time resolution (3.118 ns) is achieved. While adding cable length to

PMT2 (the one with longer transit time) to achieve larger time-delay, a worse time

resolution (4.559 ns) is obtained. Perfect timing alignment of signals in terms of cable

length and connectors will be unpractical in the real measurement, thus the 3.229 ns

time resolution of the summed PMTs is used for further reference. In addition to the

unaligned signals, other aspects affecting the timing includes: the selection of the trigger

threshold between the individual and summed signals; the slight transit timing

differences between the two PMTs in terms of the traveling time for photoelectrons to

reach the first dynode; and the time for secondary electrons multiplication.

81

Figure 4-4. Gaussian fits for the timestamp difference from summed as well individual PMTs measured at 250 MHz, 50% default CFD. Time difference is calculated as the EJ-309 scintillator’s timestamp subtracted by the 4He detector’s timestamp. The source is placed at the center of the detector.

Figure 4-4 provides a good starting point for the detector’s timing analysis. Going

forward the 5 GHz PSI digitizer is used, where 35,000 coincidence events are recorded.

The summed PMTs yields a 2.103 ns FWHM, and at 60% of its trigger threshold, the

timing resolutions for individual signals are calculated as 2.220 ns FWHM (PMT1) and

2.206 ns FWHM (PMT2) as shown in Figure 4-5. When the same trigger threshold is

applied for the individual signals as was used for the summed signal, 2.085 ns and

1.972 ns FWHM are obtained for PMT1 and PMT2, respectively. As expected, due to

the higher sampling rate of the PSI digitizer, the measured timing resolution is improved

by about 1 ns, and the summed signal still yields slightly better timing resolution than

individual signals.

82

Figure 4-5. Gaussian fits for the timestamp difference from summed, as well individual, PMTs measured at 5 GHz, 20% CFD. The time difference was calculated as the EJ-309 scintillator’s timestamp subtracted by the 4He detector’s timestamp. The 60Co source is placed at the center of the detector.

Figure 4-6 shows the calculated FWHM as a function of CFD ratio. Overall, the

detectors time resolution decreases as CFD ratio increases, and PMT2 presents a

slightly better timing performance at 20% CFD ratio than PMT1. In order to compare

with the time resolution measured by the Struck digitizer, the same 20% CFD ratio is

tested during post pulse processing, resulting in a 6.274 ns FWHM for summed PMTs,

4.934 ns FWHM for PMT1, and 4.474 ns FWHM for PMT2. When enabling the Struck

digitizer’s CFD feature (along with the 50% default CFD ratio), three MAW values are

used during time interpolation. While computing the timing resolutions at user-defined

CFD ratio, we have to use the adjacent timestamps with the inherent 4 ns timestamp

intervals, which results in a twice wider FWHM. Thus, the 50% default MAW CFD ratio

is used for timing analyses on the Struck digitizer.

83

Figure 4-6. FWHM as a function of CFD ratio at 5 GHz. The 60Co source is placed at the center of the detector.

20% CFD ratio is chosen for analyzing the “location-dependent” time resolution

at 5 GHz with the PSI digitizer. Although the 15% CFD ratio provides the best result, for

the events with relative low energy deposition, the data point at 15% of the maximum

pulse height tends to be located at the very beginning of the rise of the pulse, which

might contain certain amount of uncertainties. Even though increasing the applied

threshold could mitigate this effect, it comes with the drawback of reduced statistics. As

shown in Figure 4-7, PMT1 and PMT2 achieved their best timing resolution when the

source is closest to the PMT under study. This is because of the increase of the

photoelectrons generated in that PMT, as well as the short photon travel length and

photon transit time. Same as the result obtained from the Struck digitizer, the summed

PMT has the overall best time resolution due to the increase in photoelectrons recorded

per energy unit.

84

Figure 4-7. FWHM as a function of source location at 5 GHz. 3 source locations are measured. The middle data point is the center-point between PMT1 and PMT2.

4.4 Timing Resolution of SiPM-Based 4He Detectors

The experimental setup for measuring SiPM-based detectors is similar to Figure

4-3, except for substituting the PMT-based detector to a SiPM-based one. 20,000

coincidence events are measured. The SiPM bias voltage adjustment is attempted to

match the pulse amplitudes for each segment, but the electronics noise is highly

affected by tiny changes of the applied biasing voltage, and in normal operation the

SiPMs are all powered at the same voltage. The trigger thresholds in ADC units in the

post-measurement pulse processing script are adjusted for each segment in order to

suppress small pulses. Any pulse below 50 mV threshold will be regarded as noise and

rejected, and 30.5 V voltage is used during the measurement. As in Figure 4-8, the

SiPM output pulses have a relatively long length along with significant noise.

85

Figure 4-8. Examples of 4He SiPM detector’s output pulses. The pulses have a long rising time (about 150 ns) along with significant signal noise.

The pulses from SiPMs are noisy, and tend to be small comparing to PMT pulses

[54], making it’s hard to differentiate true events from electronics noise. Those

combined effects lead to a much wider time spread as shown in Figure 4-9.

Figure 4-9. Gaussian fit for the timestamp difference from summed as well individual channels within segment 1 at 250 MHz sampling frequency. Time difference is calculated as the EJ-309 scintillator’s timestamp subtracted by the 4He detector’s timestamp. The source is placed at the center of segment 1.

86

The middle segment yields a 46.7 ns FWHM. For the individual signals (SiPM

pairs) inside that segment, two channels are measured for comparison purpose,

yielding a 77.0 ns and 73.7 ns FWHM, respectively. Measurements are repeated for

other source locations (at center of segment 0 and segment 2), and Figure 4-10 shows

the FWHM within each segment. Comparing to the PMT-based detectors, SiPMs have

poor timing due to the longer and more inconsistent rise time of the shaped output

pulses.

Figure 4-10. The FWHM of each segment when source is placed at the center of that segment, measured with the 250 MHz Struck digitizer.

Segment 2 is the farthest segment away from the analog readout board.

Surprisingly, it has the best timing, which is about 42.2 ns FWHM. Small pulses are

more likely to be electronics noise rather than true coincidence events. Among those

three segments, segment 2 tends to yield the largest and cleanest pulses. It enables

more pulses to pass through the pulse amplitude filter (50 mV). Therefore segment 2

has the best statistics and timing resolution. Variation between segments is likely due to

variation between individual SiPM chips with regards to needed bias-voltage to have the

same signal multiplication and sensitivity.

87

4.5 Conclusions

To conclude the aforementioned work: it is the first time that the timing resolution

of both PMT-based and SiPM-based 4He fast neutron scintillation detectors have been

measured. For the SiPM-based 4He detector, a 42.2 ns FWHM (the best one) is

obtained at 250 MHz for segment 2 when summing all SiPM-signals in the segment. For

the PMT-based 4He detector, the timing resolution is calculated as 3.229 ns FWHM at

250 MHz at 50% CFD ratio, and 2.103 ns FWHM at 5 GHz at 50% CFD ratio. The

threshold is chosen as 95 (mV⨯4ns). We are not able to provide the pulse amplitude (in

V) to MeVee conversion due to the lack of gamma-ray spectroscopy features from the

detector’s outputs.

As from the previous chapter, for a neutron that deposits 1 MeV energy within the

detector, it corresponds to a tail light output of 3800 (V⨯4ns). Neutrons tend to deposit

higher energy therefore result in better timing resolution (due to the increased

production of photoelectrons), and the 3.229 ns timing resolution can be regarded as a

conservative estimation for time-sensitive neutron measurements. Both individual as

well as summed output signals are measured and analyzed. Only slight timing

differences are observed, and the summed signals tend to have better time resolution

than individual ones. In addition, the location-dependent time resolution is studied in this

work, and the results show the detector’s timing responses vary when changing the

source (interaction) locations. Therefore, when taking time-sensitive measurements, it is

important to take the source location and detector placement into account. When using

the summed signals, the location dependence can somewhat be reduced, and an

overall better timing resolution can be achieved.

88

In general, as observed from this measurement, time resolution can be improved

by means of increasing sampling rate of the digitizer system, optimizing the CFD ratio,

and varying the source locations. For neutron detection, as mentioned above, the

measured timing resolution serves as a conservative estimation if measuring

predominately high-energy neutron scatters, and as a reference to correct the

uncertainties associated with TOF measurements and can be used to examine the

detector’s TOF-energy resolution in later work.

89

CHAPTER 5 NEUTRON AND GAMMA-RAY CROSS-CORRELATION FUNCTIONS

MEASUREMENT WITH 4HE DETECTORS

In the field of nuclear nonproliferation and safeguards, neutron multiplicity

distributions and detection of neutron and/or gamma-ray coincidence measurements

are useful signatures for source identification and characterization. Liquid and plastic

types of organic scintillators are widely used, since they are sensitive to both gamma

rays and fast neutrons without using moderating materials. However, gamma-ray

sensitivity can be a limitation in applications of high gamma-ray fields when measuring

neutrons. In some of those situations, alternative detectors are needed.

The 4He detectors are known for their gamma-ray rejection capability and

therefore, as detailed in this chapter, the feasibility of using them to measure neutron

and gamma-ray time-dependent cross-correlation distributions is explored. The chapter

begins with an introduction of the coincidence measurement and its applications,

followed by a description of how to categorize various coincidence pairs, and concludes

with an evaluation of the measured cross-correlation functions with different

experimental setups. The shape of the cross-correlation functions can be an indicator of

which type, and where the neutron source is. Thus, from this work, the application of the

4He detectors is extended to an area of nuclear safeguards where the identification of

nuclear materials and material-geometry configuration assessment are highly desired.

5.1 Introduction and Advantages of Measuring Cross-Correlation Functions with 4He Detectors

Recent techniques based on coincidence measurement of neutrons and gamma

rays, usually within a time window on the order of a few tens of nanoseconds, can be

used for identifying and characterizing nuclear materials. The cross-correlation functions

90

represent signatures allowing identification of typical neutron sources (i.e. spontaneous-

fission or (α, n)) [14], radioactive material-geometry configuration [35], and special

nuclear material (uranium, plutonium) quantification [47]. Most researchers have been

focused on organic liquid or plastic scintillators. However, during cross-correlation

measurements, high-Z materials have to be employed to shield overwhelming gamma

rays and achieve better ratio of different cross-correlation functions. That results in

additional uncertainties that are introduced from scattered and induced neutrons.

To obtain comparable numbers of correlated neutrons and gamma rays without

extended measurement times or gamma-ray shielding, the novel 4He fast neutron

scintillation detectors are used to fulfil the above-mentioned requirements. In this work,

the charge-integration based pulse shape discrimination technique is utilized to

discriminate between neutrons and gamma rays to identify the four possible correlated

pairs: (γ, γ), (n, γ), (γ, n), and (n, n), where the first index corresponds to detector #1

and the second corresponds to detector #2 in any given cross correlation function. Two

neutron sources, a 252Cf spontaneous-fission and a Pu-Be (α, n) neutron source, are

used to measure the cross-correlation functions at various source-detector distances.

Both the total and individual cross-correlation functions are obtained, and their shapes

can be easily distinguished among different neutron sources. The (n, n) and (γ, n)

correlations are selectively analyzed and the peak position of the (n, n) and (γ, n) pairs

shows a linear correlation with the source-detector distance. In particular, the (n, n)

correlated data can be very clearly assayed even in presence of significant gamma-ray

fields without the use of shielding.

91

By analyzing the signatures in the cross-correlation distributions, the 4He

detectors can be used as a tool to characterize potentially unknown nuclear materials

with simplified setups and reduced uncertainties.

5.2 Cross-Correlation Functions with 252Cf Spontaneous Neutron Source

The experimental setup of the cross-correlation functions measurement involves

two identical 4He detectors and two neutron sources, a 73.7 μCi 252Cf spontaneous-

fission source and a 1 Ci Pu-Be (α, n) source. In this section, we will start with the

spontaneous-fission source.

The detectors are placed in parallel with a detector-to-detector distance of 30 cm.

Five source locations are selected. The source is initially placed between the detectors

at three source-detector distances. Both symmetric and asymmetric configurations are

investigated as follows: “15cm-15cm”, “10cm-20cm”, and “5cm-25cm”. Additionally, two

more source locations are chosen along the y-axis (see the experimental setup in

Figure 5-1). The source is placed at the centroid of the two detectors to explore the

effect of detectors locus-dependent response on the measured cross-correlation

functions. During all the measurements, no gamma-ray shielding is used, yet

reasonable neutron coincidence counts are achieved without additional uncertainties

being introduced.

92

Figure 5-1. Yinong Liang. Experimental setup of the cross-correlation functions measurements. 2018.

Each 4He detector has two output signals from the photomultiplier tubes (PMTs).

Therefore four PMT anode signals are fed into the Struck 14-bit 250MHz multi-channel

waveform digitizer in total. One thousand data points are recorded for each pulse, giving

a total pulse acquisition time window of 4 μs. The digitizer’s constant fraction

discriminator is enabled, and a trigger is generated when the actual trapezoidal value

goes below the half of its maximum value (default setting). Thus a better pulse timing

resolution (~2-3 ns) could be obtained by using the linear interpolation method, which is

a better resolution than the digitizer’s inherent 4 ns sampling interval. The average

timestamp of the two PMTs outputs from one detector is calculated and will be used to

determine the time differences between coincidence events between the two detectors.

More details of this digitizer and its constant fraction discriminator can be found in

Chapter 4.

The above-mentioned four categories of correlated pairs: (γ, γ), (n, γ), (γ, n), and

(n, n) are identified through the charge-integration based PSD. As discussed in Chapter

3, neutrons tend to have significantly larger delayed scintillation light yield than gamma

rays, and the final PSD plot is shown in Figure 5-2. For low-energy neutrons, the output

93

pulses usually exhibit low amplitude, therefore those cannot be discriminated as readily

from the gamma-ray events. However, they only account for 11% of the total number of

pulses. Events above the PSD line are regarded as neutrons while events that fall into

the lower region are regarded as gamma rays. The same aforementioned pulse filtering

algorithms [89] were applied to remove pile-up events and other anomalous events to

improve the neutron-gamma-ray identification accuracy. Cross-correlation functions are

calculated by determining the time differences between the two detectors within a

coincidence time window (80 ns utilized here) at a specific pulse amplitude threshold

(30 mV), corresponding to incoming neutrons of approximately above 0.5 MeV to be

detectable.

Figure 5-2. Scatter plot of the integration of fast versus slow component of the cross-correlation measurement.

For the 73.7 μCi 252Cf spontaneous-fission source, during a 3.2-hour

measurement, 9,326 correlated events are recorded by the two detectors. If using four

detectors instead of two, the same statistics would be achieved in 1/6th of the time due

to higher correlation possibilities. Figure 5-3 shows the measured cross-correlation

functions at various source-detector distances (“15cm-15cm” on the left and “10cm-

94

20cm” on the right). Neutrons are emitted with a characteristic fission energy distribution

resulting in a wider neutron time-of-flight (TOF) spread, thus the (γ, γ) peak has a

smaller time spread than the (n, n) peak. In addition, significantly less (γ, γ) pairs are

counted than the (n, n) pairs, which validates the low gamma-ray sensitivity of the 4He

detectors. The two side peaks are produced when a fission neutron arrives in

coincidence with a gamma ray, which is categorized as (n, γ) or (γ, n) pairs.

Figure 5-3. Measured 252Cf cross-correlation functions at “15cm-15cm” source-detector distance (left) and “10cm-20cm” source-detector distance (right). The total cross-correlation function is obtained by summing all the correlated pairs together. Uncertainty is shown on the “total” curve, and is of identical magnitude in the individual component curves for data points at the same vertical position (“Normalized counts”-amplitude) as the total curve.

Ideally, the (γ, γ) peak should occur at time zero, however due to variance in the

scintillation photon transport time in the detectors and electronic delays in the PMTs and

cables, a 1 ns shift of the (γ, γ) distribution is observed. The time shift is corrected

during later analysis. Comparing the two figures in Figure 5-3, when the detectors have

the same distance to the source, the cross-correlation distributions are symmetric

around time zero. When the source is moved between the detectors (i.e. along the x-

axis indicated in Figure 5-2), notable changes in terms of peak position and time spread

95

are observed due to the differences in neutron travel path length. These changes

provide the basis to characterize material-geometry configurations.

Noted in previous cross-correlation measurements with organic scintillators [35],

some spurious peaks are observed in (n, γ) and (γ, n) cross-correlation functions. These

spurious peaks are due to misclassified events, for example, an additional neutron

interacting with one of the detectors when a (γ, γ) pair was already being counted. The

pulse from the above mentioned double-event has gamma-ray timing features but is

more likely to be classified as a neutron due its large scintillation light output. Increasing

detector distance or applying high-Z gamma-ray shielding materials could be one

possible way to reduce the accidentals, by sacrificing detection efficiency and accuracy.

In contrast, the 4He detector produced relatively clean (n, γ) and (γ, n) correlation

curves, void of such misclassification features justifying its advantages for this

application.

Gamma rays are near instantaneous with low detection efficiency and are not

significantly affected by the source location, and it is also hard to characterize the peak

position of the coincident mixed neutrons and gamma rays due to their reduced

statistics, and their energy-dependent combined detection efficiency. Thus, the (n, n)

cross-correlation function is the best source to characterize the source location. Figure

5-4 shows the measured and Gaussian-fitted time delay distributions for 252Cf (n, n)

pairs at various source-detector distances (left), and the relationship between the (n, n)

peak position and source location (right).

96

Figure 5-4. Measured and Gaussian-fitted time delay distributions for 252Cf (n, n) pairs at various source-detector distances (left), and the (n, n) peak position as a function of source location (right) after gamma peak correction.

In addition to changing the source locations, five thresholds in analog-digital-

converter (ADC) unit are tested during pulse post-processing as 0 V (no increased

threshold), 0.05 V, 0.1 V, 0.15 V, and 0.2 V. As shown in Figure 5-5, increasing the

threshold would reject more low-amplitude events, and these events are also harder to

correctly identify through PSD. As a result, more “true neutron events” are preserved

and the time delay distribution of the (n, n) pairs tends towards a Gaussian distribution,

indicated by the decreasing full width at half maximum (FWHM). However, as the

threshold keeps going up, there will be less events of sufficient amplitude to be kept and

the filter becomes too conservative to ensure good statistics. Therefore, it is necessary

to choose a suitable threshold based on the required goodness of fit and the

coincidence count rate of each measurement. For this work, 0.05 V threshold is

selected and used in all the analysis.

97

Figure 5-5. Measured and Gaussian-fitted time delay distributions for 252Cf (n, n) pairs at various ADC thresholds at “15cm-15cm” source-detector distance (left), and the FWHM of Gaussian fitting as a function of digitizer threshold at “15cm-15cm” source-detector distance (right).

When moving the 252Cf along y-axis, theoretically there should be no shift in the

measured cross-correlation functions. However, as illustrated in Chapter 4, there are

differences in the PMT timing characteristics, which are associated with photon

transport, source locations, and the dual-PMT read-out. Therefore, slight shifts are

observed depending on where in the detector the interaction took place. As shown in

Figure 5-6, the (n, n) peak position shows a linear trend when the source is moved

along the detector’s active volume. The results indicate this type of cross-correlation

measurement may potentially provide source depth information which allows a

comprehensive geometry characterization when leveraging multiple detector locations

of volumetric distributed sources such as small mixed waste drums. The y-axis

dependence could also be derived from detection statistics if moving the detectors

along the sample-location in the y-axis in a controlled manner. Yet as observed from

Figure 5-6, the linear relationship is relative weak and the change in the (n, n) peak

position is relatively small. Therefore, if measurement condition allows, we recommend

98

rotating the source container to replicate the same behavior as in Figure 5-4. Otherwise,

simply rotating the detectors would not allow a full mapping of the source container.

Figure 5-6. The (n, n) peak position as a function of the 252Cf location along the length-dimension of the detector after gamma peak correction.

5.3 Cross-Correlation Functions with Pu-Be (α, n) Source

The experiment with the Pu-Be (α, n) source is similar as the 252Cf measurement,

and the 1-Ci Pu-Be source is placed at the same source locations. For reference

purpose, source characterization is performed at first with an EJ-309 liquid scintillator.

For this specific Pu-Be source, within 3 seconds, a total of 50,526 events are recorded

at an 80 keVee threshold within the 3” diameter ⨯ 3” height detection volume. Among

those, 50% events are neutrons and 50% events are gamma rays after PSD analysis.

The same neutron/gamma-ray ratio is obtained at all suitable thresholds, indicating a

relative strong and consistent spectrum response between the two types of radiation.

The Pu-Be source emits only one neutron per (α, n) reaction, while 252Cf has an

average neutron multiplicity of 3.76 from spontaneous-fission events [25]. Thus, the

fraction of measured neutron-correlations will be lower than for 252Cf, as one can tell

from the measured cross-correlation functions visually.

99

Figure 5-7 shows the measured cross-correlation functions at “15cm-15cm”

source-detector distance, where 21,206 correlated events are measured within 0.87

hours. The total cross-correlation function from the 252Cf measurement is also plotted

here for comparison. In contrast to the 252Cf cross-correlation functions, less (n, n) and

(γ, γ) pairs are recorded in the Pu-Be measurement, while (n, γ) and (γ, n) pairs mainly

contribute to the shape of the total cross-correlation function. In addition to observing

the peaks from plots, Table 5-1 below shows the ratio of different pairs at “15cm-15cm”

source-detector distance for 252Cf and Pu-Be. The differences in the contribution of each

type of pairs can be used when identifying nuclear materials [90], and future work will be

focused on developing advanced source identification/classification methods.

Figure 5-7. Measured Pu-Be and 252Cf cross-correlation functions at “15cm-15cm” source-detector distance. Data from Pu-Be and 252Cf is normalized by “per ns” Uncertainty is showed on the “total” Pu-Be curve, and is of identical magnitude in the individual curves for data points at the same vertical position (“Normalized counts”-amplitude) as the total curve.

Table 5-1. The ratio of the four category-pairs from 252Cf and Pu-Be measurements.

Pairs (%) 252Cf Pu-Be

(n, n) 50.87 ± 0.11 34.64 ± 0.94 (γ, γ) 9.92 ± 0.40 10.82 ± 0.48

(γ, n) 19.63 ± 0.59 26.02 ± 0.80

(n, γ) 19.58 ± 0.59 28.52 ± 0.84

100

Similarly to the 252Cf measurements, the Pu-Be measurement shows (γ, γ) pairs

as being near instantaneous and of insignificant magnitude, and the (n, n) pairs have a

relative even distribution since only one neutron is emitted per (α, n) reaction. On the

other hand, (γ, n) pairs (or (n, γ) pairs) have enough counts with relatively small

spurious peaks, therefore they can be used when characterizing the source location.

Similar to the 252Cf measurement, when moving the source between the two detectors,

it still shows a linear correlation (within uncertainty) between the (γ, n) peak position and

source location. The lack of clear features in the (n, n) cross-correlation means that the

(γ, n) data was a better option for source location determination in case of the Pu-Be

source. There is still a small (n, n) cross-correlation bump near Δt=0. It originates from

induced fissions in the plutonium. Certain discrepancies could be found when trying to

fit the (γ, n) peak position and source location linearly, while such discrepancies are not

observed in the 252Cf measurement. For the (γ, n) pairs. When moving the source, the

differences in Δt are relatively small, since only one neutron’s travel time varies with the

change in distance. In addition, gamma rays have higher yield and energy for the Pu-Be

source than 252Cf [91] [92], and are slightly more likely to be misclassified as neutrons

due to pile-up and other factors. Therefore, a slight nonlinearity is observed as in the

right part of Figure 5-8.

101

Figure 5-8. Measured and Gaussian fitted time delay distributions for Pu-Be (γ, n) pairs

at various source-detector distances (left), and the (γ, n) peak position as a function of source location (right) after gamma peak correction.

5.4 Conclusions

To conclude the results and discussions above, for the first time, the time-

dependent cross-correlation distributions from a 252Cf spontaneous-fission source and a

Pu-Be (α, n) sources are measured using the 4He fast neutron detectors. Pulses are

recorded using a Struck digitizer with a trigger timing algorithm enabled to achieve an

enhanced timing resolution. An offline charge-integration based PSD technique is

utilized to discriminate between neutrons and gamma rays. Both separate ((γ, γ), (n, γ),

(γ, n), and (n, n)) and total cross-correlation functions are measured at various source-

detector distances.

At equal source-detector distance, the cross-correlation distributions are

symmetric around time zero. When moving the source between the two 4He detectors,

notable detectable changes are observed. The peak position of Gaussian-fitted time

delay distributions ((n, n) for 252Cf and (γ, n) for Pu-Be) shows a clear linear correlation

with the source location, which provides the basis for locating nuclear materials in larger

geometrical samples.

102

The same trends are observed when the sources are moved along the detector’s

active chambers, indicating the 3D characterization potentials. In addition, the shapes of

the cross-correlation functions for 252Cf and Pu-Be shows distinct features due to

different contributions from the correlated pairs. Source identification can be achieved

by leveraging that difference.

Comparing with previous cross-correlation measurements, no shielding materials

are needed due to the 4He detector’s acceptable fast neutron detection efficiency and

lower gamma-ray sensitivity than organic materials. Therefore, introducing 4He

detectors to cross-correlation measurements can result in reduced uncertainties, fast

measurement times, and expanding the applicability to high gamma-ray fields. The

results demonstrate the feasibility of using 4He detectors to measure total and individual

cross-correlation functions from both spontaneous-fission and (α, n) neutron sources.

The detector exhibits its unique characteristics that can be leveraged as advantages

over the widely-used organic scintillators in identifying and characterizing nuclear

materials using the cross-correlation functions. Future work includes measuring actual

radioactive waste barrels, the development of source identification and location-

mapping methods, as well as other applications in nuclear nonproliferation and

homeland security.

.

103

CHAPTER 6 SECONDARY NEUTRON MEASUREMENT WITH 4HE DETECTORS AT UF-HEALTH

PROTON THERAPY INSTITUTE

A major concern of proton therapy is the secondary neutrons produced during

treatment. Neutrons have high quality factor, therefore can be a significant contribution

to the overall dose and cause potential biologic effects to patients or facility staff. Since

neutron dose highly depends on the incoming neutron energy, it is necessary to obtain

the energy information of the secondary neutrons.

Motivated by this concern, we conducted this work in collaboration with UF-

Health Proton Therapy Institute (UFPTI), where the secondary neutron spectrum is

measured, and the dose is then estimated. In addition to nuclear-related applications,

the spectroscopic capability (up to 10 MeV) and gamma-ray insensitivity enables the

4He detectors to extend their application to medical physics. The 10 MeV unfolding limit

is due to the limitations during neutron response function characterization as described

in Chapter 3, where the neutron spectrum used for building the response matrix only

goes up to approximately 10 MeV.

This chapter begins with an introduction of proton therapy, showing both its

advantages and potential risks. Details of the UFPTI experiments and the method of

neutron dose estimation are then described. Due to the lack of suitable detectors and

inability to do in-vivo measurements in proton therapy facilities, most research on

secondary radiation analysis are based on computer simulations. We compare our

novel measurement results with previous simulation results [93], which are shown in the

last section.

104

6.1 Introduction and Literature Review of Proton Therapy and Its Risks

The fundamental goal in radiation therapy treatment planning is to maximize the

dose delivered to the tumor while minimizing the dose to the surrounding healthy tissue.

High energy x-rays (XRT), have been effectively used for treating cancers for many

decades and formed the backbone of radiation oncology. Recently, proton therapy has

garnered great interest, and seen expanded implementation. Comparing to XRT,

protons deposit almost all their energy within a short distance. The depth dose curve

follows the so-called “Bragg curve”, and the dose outside the range of the proton beam

can be regarded as minimal. The finite range and sharp distal fall-off provide proton

therapy dosimetric advantages such as precisely focusing on the tumor and minimizing

the radiation dose on healthy tissue. Therefore, it is more appealing than traditional XRT

[21].

Secondary photon and neutron dose in conventional XRT is well-studied. Monte

Carlo simulation models have been commonly used, consisting of detailed geometry

configurations such as beam-line components, structural components, gantry, etc. [94]

[95]. Thermoluminescent dosimeters (TLDs), Bonner Sphere, and moderated gold foil

activation [96] are widely used when measuring secondary neutrons which are mostly

within the thermal range.

However, when it comes to the secondaries produced in proton therapy, the

references are rare. The production of secondaries strongly depends on the geometry

of beam head and on the materials of the proton beam delivery system and the patient

tissue in the beam pathway. Thus, it is almost impossible to establish or refer to a

standard configuration.

105

In proton therapy, secondary neutrons are generated by nuclear reactions when

highly energetic protons strike various nuclei targets in locations such as the patient, the

beam-collimator head, and other surrounding support structures [97]. The unwanted

neutrons can then interact with healthy tissues and lead to energy deposition.

A few reports on secondary neutron measurements are published, where the

secondary neutrons are measured at medical accelerator facilities using either rem-

counter [97], Bonner sphere [98] [99], or indium-foil activation [93].

On one hand, moderators or fast neutron capture materials are essential in order

to slow down high-energy neutrons. However, adding moderators can cause the loss of

neutron energy information, longer experimental setup times, and additional

uncertainties. Without prior knowledge of the neutron energy spectrum, the presumed

energy-dependent detection efficiency is also highly uncertain. The above-mentioned

measurements are simply count-rate based and lack the ability to obtain an accurate

neutron spectrum covering a relative wide energy range. In addition, significant

secondary photons are produced in proton therapy at the same time [99]. Traditional

detectors are easily saturated by high photon flux and applying photon shielding

materials can result in additional secondary neutrons via (,n)- and (n,2n)-reactions.

Therefore, 4He detectors are used in this work due to their unique advantages to

overcome many of these challenges.

6.2 Experimental Setups and Dose Estimation Method

The experiment is conducted at a treatment room in UFPTI. Protons are

accelerated up to 180 MeV by a cyclotron and transported to the treatment room. A

gantry room 1 (gantry 1) weekly quality assurance (QA) proton-beam field is used

106

where the field size is 20x20 cm2. The peak of the “Bragg curve” is not wide enough to

cover most treatment volumes, therefore, the incident proton beam will undergo a

rotating range modulator, which has absorbers of variable thickness in order to cover

the tumor volume with high accuracy. The individual Bragg curves are weighted in this

way and can produce a spread-out Bragg peak (SOBP) as shown in Figure 6-1. The

SOBP modulation width is 10.4 cm, where most protons will stop after 15 cm water.

Figure 6-1. The measured depth-dose curve (SOBP) for this field at UFPTI, for a 180 MeV proton beam.

The double scatter (DS) nozzle system presented in this study is based on the

universal nozzle installed at UFPTI. Figure 6-2 shows the simplified diagram of the

nozzle according to blueprints by the manufacturer and previous treatment planning

system (TPS) commissioning data. Major components such as the first scatter, ion

chambers, and range modulation wheel (RMW), second scatter, variable collimators,

snout and aperture are modeled in detail during the Monte Carlo simulations.

107

Figure 6-2. Hongdong Liu. The simplified diagram of the nozzle. 2018.

From the depth-dose measurement as shown in Figure 6-1, most protons stop

after traveling through 15 cm water, and high-energy secondary neutrons are produced

along the path. Several 30x30x5 cm solid water slabs are mounted on the treatment

bed. The 4He fast neutron detector is placed behind the solid water (with a distance of

30 cm) and is 80 cm away from the snout as shown in Figure 6-3. The solid water is

made of epoxy resins and powders with a density of 1.04 g/cm3 [100]. It is designed for

radiation beam calibration without the inconvenience of transporting, setting up and

filling water tanks. Incident particles can scatter and attenuate with solid water the same

way as with water (human body) without the charge storage problems. The solid water

has modular size and is manufactured with different thicknesses, which can be easily

adjusted with respect to various experimental setups. Starting from 15 cm, the thickness

of the solid water is added up to 30 cm (with an increment of 5 cm) by varying the

numbers of slabs being used. Output signals from the PMTs are fed into the Struck 14-

bit 250 MHz multi-channel waveform digitizer and are analyzed off-line later through a

Matlab script. At each setup, 80,000 counts are recorded within a couple of seconds.

The maximum neutron dose is estimated via count rate-to-dose conversion and

normalized by the therapeutic proton-absorbed dose reported by the dose monitor.

108

Figure 6-3. Yinong Liang. The experimental setup at UFPTI. 2018.

The incident neutron spectrum φ(En) is estimated via the iterative least square

algorithm as described in Chapter 3, which can predict the incident neutron spectrum

step by step with reasonable uncertainties. In this measurement, the 4He detector is

placed right behind the solid water. Therefore, it is reasonable to assume that most of

the recorded neutrons are directly from the solid water slabs, and the contribution of

random scattered neutrons (room returns) is insignificant. The neutron fluence can be

estimated when knowing the detection area, average detection efficiency, and

measured count rate. Based on the U.S. Nuclear Regulatory Commission or U.S.

Agreement state regulations, the fluence can be converted to dose via the fluence-to-

dose conversion factors (given by Table 1004(b).2 of paragraph 20.1004 of 10 CFR 20)

[101], where the quality factor Q is chosen depending on the incoming neutron energy

predicted by the unfolding algorithm. Finally, the absorbed dose per therapy unit (Gray,

Gy) is calculated and used when comparing different setups.

6.3 Results and Discussions

A charged-integration based PSD algorithm is used for differentiating neutrons

from gamma rays, and Figure 6-4 shows the PSD plots at various solid water

109

thicknesses. The 15 cm solid water case is associated with the highest neutron yield

and clearest separation as expected, since most neutrons are produced within the

range of protons and then enter the detector with minimized chance of further

interacting with the solid water. As the solid water thickness increases, more neutrons

will be slowed down within the solid water, therefore less neutrons and smaller slow

component integration (i.e., smaller energy deposition) are recorded in the 4He

detectors. These neutrons tend to look more similar to gamma rays, and the

conservative PSD cut-off can ensure that the vast majority of events in the upper region

are neutrons. Unlike the slow component, the reduction in fast component is not

obvious. The differences in fast component for neutron and gamma-ray event are

relatively small when integrating over a few data points (3 to 5 data points), therefore we

cannot observe obvious changes in fast component when increasing solid water

thickness. Overall, even though considerable amounts of gamma rays are produced at

the same time, the detector shows a good gamma-ray rejection capability and is able to

record enough neutrons to later perform spectral analysis.

110

Figure 6-4. PSD plots at various solid water thicknesses (upper left: 15 cm, upper right: 20 cm, lower left: 25 cm, lower right: 30 cm).

Figure 6-5 shows the measured scintillation light outputs and the exiting neutron

spectra from solid water with various thicknesses. The measured neutron spectrum has

a peak at 6 MeV from 15 cm solid water, while for 30 cm solid water the highest neutron

yield is around 2 MeV. The differences in the measured spectra are mainly due to the

increased probability of moderation as the solid water thickness increases, and the

spectra of neutrons emerged from 20 cm and 25 cm solid water vary in between the 15

cm and 30 cm cases. The measured scintillation outputs can also shed some light on

the incoming neutron energy, and therefore can somehow serve as a verification of the

measured results. High energy neutrons tend to produce larger scintillation lights (as

111

illustrated in Chapter 3), and as seen on the left of Figure 6-5, the 15 cm solid water has

the highest ratio of relatively large light outputs.

Figure 6-5. Measured scintillation light outputs (left) and the exiting neutron spectra (right) from solid water with various thicknesses. The detector is placed 30 cm away from the solid water.

The Geant4 v4.10.03 simulation toolkit-based Monte Carlo code TOPAS (version

3.1.p3) [102] is used for the simulations that are conducted by another student who

works at UFPTI. Major components as shown in Figure 6-2 are incorporated into the

modelling. The simulation starts at the window before proton beams enter the nozzle. A

reference beam with 15.1cm range, 10.4 cm width and 20 cmx20 cm field size are

employed. The nozzle employs a rotating RMW synchronized with beam current

modulation to obtain a uniform SOBP [103]. Figure 6-6 shows the RMW model in

TOPAS. The 25 cm size snout and 6.5 cm thick aperture are used to provide desired

field sizes. The conversion algorithm (Convalgo, IBA) is used to determine the beam

set-up, such as selections of the first and second scatter, track of the RMW, beam

energy, variable collimator opening, and the modulated beam current, according to the

range and modulation width of the SOBP.

112

Figure 6-6. Hongdong Liu. Three range modulation wheels (RMW) with nine tracks modeled within TOPAS. 2018.

The simulation is first conducted to record the phase space at the end of the

nozzle, i.e. a virtual plate downstream of the aperture. The phase space refers to the

broadened proton beam coming out from the secondary scattering systems, and is

confined to the required size via aperture, then the phase space file is used as the

source for multiple times to improve the statistics. The total number of particles

recorded in this phase space file (particles emitting from the nozzle head.) is 2.8x107.

All the simulations ran on a Linux workstation with Xeon E5-3667 V3 (Intel) of 3.2 GHz.

Calculation time required for recording the phase space file is about 20 CPU-hours, and

the computation time needed for reusing the phase space file 10 times, is about 7 CPU-

hours. The simulation results of the exiting neutron spectra from various solid water

thicknesses are shown in Figure 6-7 as below.

113

Figure 6-7. Simulated neutron spectra emitted from various solid water thicknesses.

For all the setups, the peak of the simulated spectrum is at 0.5 MeV. For

neutrons higher than 0.5 MeV, varying the thickness of the solid water will not affect the

energy distribution significantly. The 4He detectors have a relative low detection

efficiency (below 2%) for neutrons below 0.5 MeV. Additionally, the PSD algorithm is

conservative and tend to reject low energy neutrons. Therefore, the 0.5 MeV peak in the

simulated spectrum is not observed in the measured spectra. Additionally, in simulation

a higher fraction of high energy neutrons is predicted in the 30 cm solid water thickness

case, along with other signatures which may be hidden by the overwhelming number of

neutrons of energy < 0.5 MeV. This disagrees with our measurement results and we will

focus on improving the simulation models in future work. Nevertheless, since the

measured spectra agree with the detector’s light outputs, we kept using the

measurement result for later analysis and will work on improving and understanding the

simulation model in the future.

The recorded neutrons per second reflects the intensity of the field and serves as

a starting point for dose estimation. It should be noted that the neutron count rate is not

114

only depending on the overall production rate, but also the detector’s efficiency. For 4He

detectors, the energy-dependent detection efficiency was both measured and simulated

as in Figure 3-7, which follows the same trend as the elastic scattering cross-section of

the 4He nucleus. Thus, as shown in Table 6-1, neutron count rate is increased when

more neutrons around 1 MeV are produced. As high energy neutrons are moderated

down to the energy range (1 MeV ~ 2 MeV) where the detection efficiency is highest,

the highest count rate is observed at 25 cm solid water thickness. However, a

disagreement happens at 30 cm solid water thickness, where less neutrons are

recorded than expected. It may be because even though most measured neutrons are

within 1 MeV to 2 MeV range, the rest of them are mostly low energy neutrons (with the

smallest measured average energy as 3.9 MeV). Their corresponding low-amplitude

pulses are potentially misclassified as gamma rays and rejected. Therefore, the count

rate drops slightly at 30 cm solid water thickness even though more neutrons are

moderated to within the energy range of the detector’s maximum efficiency.

Table 6-1. Neutron characteristics from various solid water thicknesses from measurement.

Solid water thickness

Average neutron energy (MeV)

Fraction of neutrons within 1MeV ~ 2MeV (%)

Neutron count rate (cps)

15cm 4.95 9.21 6170 20cm 4.78 10.34 7182 25cm 4.05 16.92 15141 30cm 3.90 19.33 9357

As the proton beam is completely stopped in the solid water, only secondary

particles (neutrons and gamma rays) deliver a dose to the remaining regions. In this

work, we are mainly focused on estimating secondary neutron dose due to the scarcity

of prior studies and the spectroscopic ability of the 4He detectors. A robust neutron

115

count rate to absorbed dose conversion can be performed with the knowledge of

detection efficiency, detection area (88 cm2), and fluence-to-dose conversion factors. To

normalize the measured dose from various solid water thicknesses, the term “absorbed

dose per therapy Gy” is used, where the therapy Gy is given as 0.0243. The unfolding

algorithm divides neutrons into 20 energy groups and provides the probability of

occurrence for each group. The probability can then be used as weights during dose

estimation where the detection efficiency and fluence-to-dose conversion factor are both

energy-dependent. Instead of averaging all the neutrons into a certain value, the

weighted neutron dose takes every group of neutrons into account and the results are

shown in Table 6-2 below.

The maximum dose is achieved at 25 cm thickness, which is about twice higher

than the lowest one, and all the measured doses have the same magnitude as the

simulation results reported by Agosteo et al. [93] from a previous study. However, one

thing to keep in mind is that due to the limitation in the measurement of the detector’s

response characterization, we are not able to provide the spectrum for neutrons above

10 MeV, and the true neutron dose may be larger than measured. Additionally, the

detection efficiency of the 4He detectors for neutrons above 10 MeV is lower due

reduced interaction probability of neutrons of that high energy. Therefore, instead of

expanding the detector’s response matrix, looking for other suitable detectors may be

worthwhile. One distinct benefit in the high-energy domain for 4He-detctors is the

inherent low probability of break-up reactors for the helium-nuclei, most other nuclei

(other detector mediums) have a significant lower threshold for neutron-induced break-

up reactions which can sometime be hard to classify or evaluate correctly. This is the

116

first time demonstrating directly measuring the secondary neutron spectrum and

estimating neutron dose without time-consuming runs and complicated shielding setups.

From the measurement results, the influence of the solid water thickness on the

corresponded neutron dose is clear and can provide valuable insights in later treatment

evaluation.

Table 6-2. Neutron dose estimation for various solid water thicknesses.

Solid water thickness

Weighted secondary neutron fluence (cm-2/s)

Weighted secondary neutron dose (Gy per therapy Gy)

15cm 2078 4.93 × 10-6 20cm 2403 5.59 × 10-6 25cm 4730 1.04 × 10-5 30cm 2910 6.27 × 10-6

To study the effect of measurement location, in a separate trip to UFPTI, the

detector is placed 60 cm from the solid water with an angular deviation rather than in

the forward direction (as shown in Figure 6-8). Only 30 cm solid water is used due to the

emphasis on measuring location-dependent secondary neutron spectrum.

Figure 6-8. Experimental setups to study the effect of measurement location.

117

The results are shown as in Figure 6-9 and Table 6-3. Notably, in this

measurement campaign, the detector and data acquisition system settings are different.

A factor of five higher ADC threshold is used here, therefore a higher fraction of

recorded events are from high energy neutrons than in Figure 6-5. The average neutron

energy at 60 cm detector-solid water distance is about 1.3 times lower than the one

from 30 cm detector-solid water due to the detector being placed at the side of the

beam head. High energy neutrons would be predominantly generated in the forward

direction. As one can observe from the spectra, at 60 cm, few neutrons are detected

below 1 MeV due to the increased measurement threshold. However, the peak we

observed around 2 MeV may not be true, since the probability for detecting various

energy neutrons are modified with the increased ADC threshold. In other words, high

energy neutrons have higher weights, and it would not be reflected when applying the

un-modified response matrix during spectrum unfolding. When the detector is close to

the solid water, most neutrons are free from random scatter and energy loss, therefore

carrying relative higher energy. The weighted neutron dose is estimated as 1.39 × 10-7

Gy per therapy Gy at the side of beam head, which is about two orders of magnitude

lower than the maximum dose (1.04 × 10-5) as calculated in Table 6-2. Thus, when

conducting treatment evaluation, the effect of measuring point should not be ignored.

118

Figure 6-9. Measured scintillation outputs (left) and the exiting neutron spectra (right) at different room locations.

Table 6-3. Neutron dose estimation at different room locations.

Detector-solid water distance

Average neutron energy (MeV)

Neutron count rate (cps)

Weighted secondary neutron fluence (cm-2/s)

Weighted secondary neutron dose (Gy per therapy Gy)

30cm 4.28 3416 1003 2.33 × 10-6 60cm 3.15 65 2403 1.39 × 10-7

6.4 Conclusions

To conclude this work, the secondary neutron spectrum and dose from UFPTI

are measured directly with 4He fast neutron detectors for the first time. The measured

light responses and the unfolded spectra vary with the thickness of the solid water. The

term “absorbed dose per therapy Gy” is used for dose comparison and normalization. It

is weighted by taking every group of neutrons into account to accurately accommodate

the energy-dependent factors during count rate to dose conversion. The maximum dose

is calculated as 1.04 × 10-5 Gy per therapy Gy at 25 cm solid water thickness, and the

minimum dose is 4.93 × 10-6 Gy per therapy Gy which happens at 15 cm solid water

thickness. The measured dose is comparable to the simulations conducted by previous

research, while the measured spectra show some disagreements with the TOPAS

119

simulations for low energy neutrons due to the low detection efficiency. Two detector

locations: 30 cm and 60 cm away from the solid water are selected to study the impact

of measurement location. Due to the change in measurement direction (at the side of

beam head) and distance (60 cm) detector-solid water distance, less neutrons are

recorded, the average neutron energy is reduced, and the estimated dose is about one

order of magnitude lower than the 30 cm case (comparing at the same high threshold).

However, during the measurement, an increased ADC threshold is applied, which may

lead to spurious peaks during spectrum unfolding. Additional measurements at regular

threshold may be required for further analysis of the location-dependency of the dose.

Nevertheless, for secondary dose treatment evaluation, both location-dependency and

variations in solid water thickness should be considered.

The 4He detector has been evaluated in the field of fast neutron spectroscopy,

while in this work, we take the neutron energy information as input and further explored

its potentials to be used as a dosimeter. The secondary neutron spectrum is directly

measured without any high-Z gamma-ray shielding materials which can introduce

additional uncertainties. Current spectrum unfolding capability of our 4He detectors is

within 10 MeV due to the limitations during neutron response function characterization,

In addition, the recoil alpha particle from high energy neutrons may not able to deposit

all its energy within the detector gas chamber therefore the corresponding light output

could be compromised. For example, for a 25 MeV alpha particle (that corresponds to a

40 MeV neutron transferring the maximum fraction of 64% of its kinetic energy), the

range in the detector is about 2.3 cm, which exceeds the radius of the detector’s gas

chamber (2.2 cm). Therefore for neutrons above 10 MeV, the detector’s response is

120

complicated and usage of the currently-sized 4He detector may not be a good choice.

Future work includes finding suitable detectors for high-energy neutrons, measuring

neutron dose within various phantoms (eyes, lungs, etc.), improving the simulation

models, and comparing neutron dose along the proton beam line for various apertures

and field sizes.

121

CHAPTER 7 CONCLUSION AND FUTURE WORK

This work is motived by the need as expressed by IAEA for monitoring and

verifying the nuclear spent fuel stored in dry casks. Traditional detectors have their

limitations, and many are not deployable in a field scenario. The novel 4He fast neutron

scintillation detectors have caught our attention for multiple reasons, such as gamma-

ray rejection capability, spectroscopy potential, and reasonable timing resolution. These

unique features result in the detectors being an ideal tool to perform neutron

spectrometry in high gamma-ray intensity environment. And by doing so, could provide

significant information about the fuel stored inside the dry casks.

The first task we conducted was to characterize the detector via TOF

measurements at the Ohio University accelerator facilities. The detector’s light response

to neutrons with continuous energy (up to 10 MeV) was measured, upon which, the

detector’s response matrix is built. By knowing the response matrix, an iterative least-

squares based unfolding algorithm have been developed and the associated

uncertainties are estimated. The unfolded results roughly agree with the reference

spectrums of the known test sources, and the detector’s capability of differentiating

spontaneous fission neutrons and (α, n) neutrons (those are the two main neutron

sources of the SNF) are verified. In addition to energy response characterization, the

detector’s timing resolution is measured. Since a large fraction of the uncertainties in

the TOF measurement are resulting from timing uncertainties of the detector, this

measurement serves as a starting point for estimating the uncertainties for the unfolding

algorithm approach.

122

During the study of the detectors, we find that their applications are not limited to

SNF monitoring. Indeed, the detectors can be useful whenever detecting neutrons is of

great interest and when spectral-related information is needed. Cross-correlation

measurements were conducted, and the detectors show potential in radioactive source

identification and material-geometry configuration assessments. The detectors are also

utilized for measuring the secondary neutron spectrum and estimating the

corresponding secondary-neutron dose at a proton therapy facility. The measured

results can help verify dose simulations and provide valuable information for treatment

evaluations.

Further work is suggested as the development of a compact prototype detection

system for SNF monitoring, either as bare fuel assemblies or in spent fuel storage

containers such as canisters and casks. Additional future tasks include conducting

actual SNF measurement and comparing it with simulations and other available

detection systems, as well as focusing on the SiPM-based 4He detectors for spatial

characterizations.

.

123

LIST OF REFERENCES

[1] J. D. Werner, “U . S . Spent Nuclear Fuel Storage,” 2012.

[2] R. P. Kelley et al., “Neutron response function characterization of helium-4 scintillation detectors,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 793, pp. 101–107, 2015.

[3] Y. Liang et al., “Neutron Response Function Characterization and Spectrum Unfolding with He Fast Neutron Detectors,” 2018.

[4] I. Harkness, T. Zhu, Y. Liang, E. Rauch, A. Enqvist, and K. A. Jordan, “Development of Neutron Energy Spectral Signatures for Passive Monitoring of Spent Nuclear Fuels in Dry Cask Storage,” EPJ Web Conf., vol. 170, pp. 4–7, 2018.

[5] D. Lochbaum, Nuclear Waste Disposal Crisis. PennWell Books, 1996.

[6] R. B. Rebak, “Environmental degradation of materials for nuclear waste repositories engineered barriers,” J. Mater. Eng. Perform., vol. 9, pp. 317–323, 2007.

[7] F. Nimander, “Investigation of Spent Nuclear Fuel Pool Coolability,” Royal Institute of Technology, 2011.

[8] L.S. Romanato, “Safe advantage on dry interim spent nuclear fuel storage,” in HLW, TRU, LLW/ILW, Mixed, Hazardous Wastes & Environmental Management, 2008, vol. 137.

[9] M. Paul et al., “Extended dry storage of used nuclear fuel: technical issues: a USA perspective.,” J. Nucl. Eng. Technol., vol. 43, pp. 405–412, 2011.

[10] H. Feiveson, Z. Mian, M. V. Ramana, and F. von Hippel, “Managing Spent Fuel from Nuclear Power Reactors,” in 2011 International Panel on Fissile Materials Managing, 2011.

[11] R. Halstead and J. Ballard, “Nuclear Waste Transportation Security and Safety Issues: The Risk of Terrorism and Sabotage Against Repository Shipments.,” in NV. Agency Nucl. Proj., 1997.

[12] R. Sandoval, J. Weber, H. Levine, A. Romig, and J. Johnson, “An Assessmentof the Safety of Spent Fuel Transportation in Urban Environs,” 1983.

[13] D. Murer et al., “4He detectors for Mixed Oxide (MOX) fuel measurements,” IEEE Nucl. Sci. Symp. Conf. Rec., pp. 4858–4864, 2012.

124

[14] M. Flaska, A. Enqvist, and S. a Pozzi, “Measurement of Fast Neutron / Gamma-Ray Cross- Correlation Functions with Cf-252 and Pu-Be Neutron Sources,” Ieee Nucl. Sci. Symp. Conf. Rec., pp. 961–963, 2009.

[15] W. Mannhart, “PROPERTIES OF NEOTRON SOURCES: Evlauation of the Cf-252 fission neutron spectrum between 0 MeV and 20 MeV,” in ADVISORY GROUP MEETING ON PROPERTIES OF NEUTRON SOURCES ORGANIZED BY THE INTERNATIONAL ATOMIC ENERGY AGENCY, 1986, no. June 1986, pp. 158–170.

[16] C. C. Lawrence et al., “Time-of-flight measurement for energy-dependent intrinsic neutron detection efficiency,” IEEE Nucl. Sci. Symp. Conf. Rec., vol. 315, pp. 110–113, 2010.

[17] T. Saegusa, G. Yagawa, and M. Aritomi, “Topics of research and development on concrete cask storage of spent nuclear fuel,” Nucl. Eng. Des., vol. 238, no. 5, pp. 1168–1174, 2008.

[18] H. Chung, R. P. Kelley, W. Lee, Y. H. Chung, and K. A. Jordan, “Spent nuclear fuel cask and storage monitoring with Helium-4 scintillation fast neutron detectors,” Proc. Korean Nucl. Soc. 2014 Fall Meet., 2014.

[19] R. Chandra, G. Davatz, H. Friederich, U. Gendotti, and D. Murer, “Fast neutron detection with pressurized Helium-4 scintillation detectors,” J. Instrum., vol. 7, no. 03, pp. C03035–C03035, 2012.

[20] I. Harkness, “Safeguards Approaches for Spent Nuclear Fuel in Dry Cask Storage,” University of Florida, 2018.

[21] B. Hoppe et al., “Phase II trial of concurrent chemotherapy and proton therapy for stage 3 non-small cell lung cancer.,” Int J Part. Ther., vol. 2, no. 1 SRC-GoogleScholar FG-0, p. 58, 2015.

[22] L. C. Walters, “Thirty years of fuels and materials information from EBR-II,” J. Nucl. Mater., vol. 270, no. 1, pp. 39–48, 1999.

[23] C. Willman, A. Håkansson, O. Osifo, A. Bäcklin, and S. J. Svärd, “Nondestructive assay of spent nuclear fuel with gamma-ray spectroscopy,” Ann. Nucl. Energy, vol. 33, no. 5, pp. 427–438, 2006.

[24] O. W. Hermann and C. W. Alexander, “A Review of Spent-Fuel Photon and Neutron Source Spectra,” 1986.

[25] H. Nifenecker, C. Signarbieux, R. Babinet, and J. Poitou, “Neutron and Gamma Emission in Fission,” vol. 2, p. 117, 1973.

125

[26] Arndt Rimpler, “Bonner sphere neutron spectrometry at spent fuel casks,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers Detect. Assoc. Equip., vol. 476, pp. 468–473, 2002.

[27] G. F. Knoll, Radiation detection and measurement. Don Fowley, 2010.

[28] R. Society and P. Sciences, “On the nuclear forces and the magnetic moments of the neutron and the proton,” Proc. R. Soc. Lond. A. Math. Phys. Sci., vol. 166, no. 924, pp. 154–177, 1938.

[29] M. B. Chadwick et al., “ENDF/B-VII.1 Nuclear data for science and technology: cross sections, covariances, fission product yields and decay data,” Nucl. Data Sheets, vol. 112, no. 12, pp. 2887–2996, 2011.

[30] C. K. Bockelman, “Slow neutron capture and the (d, p) reaction,” Nucl. Phys., vol. 13, pp. 205–223, 1959.

[31] Lothar Koester, Neutron scattering lengths and fundamental neutron interactions. Springer Berlin Heidelberg, 1977.

[32] G. F. Chew, “The inelastic scattering of high energy neutrons by deuterons according to the impulse approximation,” Phys. Rev., vol. 80, no. 2, pp. 196–202, 1950.

[33] A. P. Simpson, S. Jones, M. J. Clapham, and S. A. McElhaney, “A Review of Neutron Detectino Technology Alternatives to Helium-3 for Safeguards Applications,” in INMM 52 Annual Meeting, 2011, pp. 1–10.

[34] R. Batchelor, R. Aves, and T. H. R. Skyrme, “Helium-3 filled proportional counter for neutron spectroscopy,” Rev. Sci. Instrum., vol. 26, no. 11, pp. 1037–1047, 1955.

[35] A. Enqvist, M. Flaska, and S. Pozzi, “Measurement and simulation of neutron/gamma-ray cross-correlation functions from spontaneous fission,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 595, no. 2, pp. 426–430, 2008.

[36] C. C. Lawrence, A. Enqvist, M. Flaska, S. A. Pozzi, and F. D. Becchetti, “Comparison of spectrum-unfolding performance of (EJ315) and (EJ309) liquid scintillators on measured 252Cf pulse-height spectra,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 729, pp. 924–929, 2013.

[37] L. Cartegni and S. A. Pozzi, “Simulation of the neutron response matrix for a liquid scintillator and spectrum unfolding,” Oak Ridge Natl. Lab. Intern. Rep., no. ORNL/TM-2004/315, 2005.

126

[38] H. Wang, D. Carter, T. N. Massey, and A. Enqvist, “Neutron light output function and resolution investigation of the deuterated organic liquid scintillator EJ-315,” Radiat. Meas., vol. 89, pp. 99–106, 2016.

[39] A. L. Guckes, “Applications of Elpasolites as a Multimode Radiation Sensor,” 2017.

[40] N. Dolympia, P. Chowdhury, E. G. Jackson, and C. J. Lister, “Fast neutron response of 6Li-depleted CLYC detectors up to 20 MeV,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 763, pp. 433–441, 2014.

[41] X. Wen and A. Enqvist, “Pulse shape discrimination of Cs2LiYCl6:Ce3+ detectors at high count rate based on triangular and trapezoidal filters,” Nucl. Inst. Methods Phys. Res. A, vol. 866, pp. 129–133, 2017.

[42] C. W. E. van Eijk, “Inorganic scintillators for thermal-neutron detection,” in IEEE Nuclear Science Symposium and Medical Imaging Conferenc, 2010.

[43] R. Chandra, G. D. Lewis, H. Friederich, U. Gendotti, and D. Murer, “Fast neutron detection with pressurized 4He scintillation detectors,” J. Instrum., vol. 7, p. C03035, 2012.

[44] W.R. Bennett Jr., “Optical spectra excited in high pressure noble gases by alpha impact,” Ann. Phys. (N. Y)., vol. 18, no. 2, 1962.

[45] J. Feist et al., “Neutron impact ionization of helium,” in XXVII Conference on Photonic, Electronic and Atomic Collisions, 2012.

[46] R. T. Kouzes, A. T. Lintereur, and E. R. Siciliano, “Progress in alternative neutron detection to address the Helium-3 shortage,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 784, pp. 172–175, 2015.

[47] S. D. Clarke, M. Flaska, S. A. Pozzi, and P. Peerani, “Neutron and gamma-ray cross-correlation measurements of plutonium oxide powder,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 604, no. 3, pp. 618–623, 2009.

[48] K. A. Yu, B. Esposito, L. A. Trykov, and V. P. Semenov, “Fast neutron spectrometry with organic scintillators applied to magnetic fusion experiments,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 476, pp. 511–515, 2002.

[49] J. M. Lewis, R. P. Kelley, D. Murer, and K. A. Jordan, “Fission signal detection using helium-4 gas fast neutron scintillation detectors,” Appl. Phys. Lett., vol. 105, no. 1, pp. 1–4, 2014.

127

[50] D. Shin et al., “Secondary Neutron Doses for Several Beam Configurations for Proton Therapy,” Int. J. Radiat. Oncol. Biol. Phys., vol. 74, no. 1, pp. 260–265, 2009.

[51] F. Brooks and H. Klein, “Neutron spectrometry—historical review and present status,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 476, no. 1–2, pp. 1–11, 2002.

[52] T. Fujisawa et al., “Allowed and forbidden transitions in artificial hydrogen and helium atoms,” Nature, vol. 419, 2002.

[53] Photomultiplier Handbook: Theory, Design, and Application. Burle Industries, 1980.

[54] T. Zhu et al., “Improved fission neutron energy discrimination with 4He detectors through pulse filtering,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 848, no. September 2016, pp. 137–143, 2017.

[55] H. Friederich et al., “A scalable DAQ system based on the DRS4 waveform digitizing chip,” IEEE Trans. Nucl. Sci., vol. 58, no. 4 PART 1, pp. 1652–1656, 2011.

[56] S. Ritt, “Design and performance of the 6 GHz waveform digitizing chip DRS4,” IEEE Nucl. Sci. Symp. Conf. Rec., pp. 1512–1515, 2008.

[57] J.M. Hubbell and S.M. Seltzer, “Tables of X-Ray mass attenuation coefficients and mass energy-absorption coefficients (version 1.4).,” National Institute of Standards and Technology, 2004. .

[58] “Arktis Radiation Detectors Ltd.,” personal communication. .

[59] M. Yang and Y. Cheng, “Energy non-linearity studies at Daya Bay,” Int. J. Mod. Phys. Conf. Ser., vol. 31, p. 1460312, 2014.

[60] J. W. Meadows, “The 9Be(d, n) thick-target neutron spectra for deuteron energies between 2.6 and 7.0 MeV,” Nucl. Inst. Methods Phys. Res. A, vol. 324, no. 1–2, pp. 239–246, 1993.

[61] W. B. Howard et al., “Measurement of the thick-target Beryllium-9(p,n) neutron energy spectra,” Niclear Sci. Eng., vol. 138, pp. 145–160, 2001.

[62] S. A. Pozzi, E. Padovani, and M. Marseguerra, “MCNP-PoliMi: A Monte-Carlo code for correlation measurements,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 513, no. 3, pp. 550–558, 2003.

128

[63] Michael F. L’Annunziata, “Handbook of Radioactivity Analysis,” no. 3rd Edition, 2012.

[64] MCNP 2.4.0 User Manual. Los Alamos National Laboratory, 2002.

[65] Y. Xu, M. Flaska, S. Pozzi, V. Protopopescu, and T. Downar, “A sequential least-squares algorithm for neutron spectrum unfolding from pulse-height distributions measured with liquid scintillators,” M&C+ SNA, p. 11, 2007.

[66] C. A. Oster, “REVIEW OF UNFOLDING METHODS USED IN THE U.S. AMD THEIR STANDARDIZATION FOR DOSIMETRY,” in Current status of neutron spectrum, Thechnical Commitee Meeting, 1977, no. International Atomic Energy Agency.

[67] R. H. Ritchie and H. B. Eldridge, “Thermal neutron flux depression by absorbing foils,” Nucl. Sci. Eng., vol. 8, no. 4, pp. 300–311, 1960.

[68] W. N. Selander and C. River, “Theoretical evaluation of self-shielding factors due to scattering resonances in foils,” 1960.

[69] S. Pearlstein and E. V Weinstock, “Scattering and self-shielding corrections in Cadmium-filtered gold , Indium , and l / v foil-activation measurements,” vol. 42, pp. 28–42, 1967.

[70] J. Wilkinson, The Algebraic Eigenvalue Problem. Oxford: Clarendon Press, 1965.

[71] F. W. Sun, Y. Jiang, and J. S. Baras, “On the convergence of the inverses of Toeplitz matrices and its applications,” IEEE Trans. Inf. Theory, vol. 49, no. 1, pp. 180–190, 2003.

[72] F. G. Perey, “Spectrum unfolding by the least-squares methods,” 1977.

[73] M. Matzke, “Propagation of uncertainties in unfolding procedures,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 476, no. 1, pp. 230–241, 2002.

[74] J. P. and M. Krâlik, “The unfolding of neutron spectra based on the singular value decomposition of the response matrix,” Nucl. Instruments Methods to Phys. Res., vol. 325, pp. 314–318, 1993.

[75] M. Reginatto, P. Goldhagen, and S. Neumann, “Spectrum unfolding, sensitivity analysis and propagation of uncertainties with the maximum entropy deconvolution code MAXED,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 476, no. 1–2, pp. 242–246, 2002.

[76] K. Shin, Y. Uwamino, and T. Hyodo, “Propagation of Errors from Response Functions to Unfolded Spectrum,” Nucl. Technol., vol. 53, no. 1, pp. 78–85, 1981.

129

[77] Y. Liang, T. Zhu, and A. Enqvist, Timing Characterization of Helium-4 Fast Neutron Detector with EJ-309 Organic Liquid Scintillator, vol. 170. 2018.

[78] D. L. Chichester, J. T. Johnson, and E. H. Seabury, “Fast-neutron spectrometry using a 3He ionization chamber and digital pulse shape analysis,” Appl. Radiat. Isot., vol. 70, no. 8, pp. 1457–1463, 2011.

[79] M. Matzke, Unfolding of pulse height spectra : the HEPRO program system. 1994.

[80] C. Guerrero et al., “Performance of the neutron time-of-flight facility n_TOF at CERN,” Eur. Phys. J. A, vol. 49, no. 2, pp. 1–15, 2013.

[81] R. P. Kelley et al., “Neutron Measurements with Extended Range Helium-4 Detectors in High Gamma Environments,” in 2016 IEEE Symposium on Radiation Measurements and Applications (SORMA West), 2016.

[82] J. Y. Yeom, R. Vinke, and C. S. Levin, “Optimizing timing performance of silicon photomultiplier based scintillation detectors,” IEEE Nucl. Sci. Symp. Conf. Rec., pp. 3119–3121, 2012.

[83] “16 Channel VME Digitizer User Manual,” vol. 49, no. 0, 2015.

[84] T.J. Paulus, “Timing electronics and fast timing methods scintillation detectors,” IEEE Trans. Nucl. Sci., vol. NS-32/3, p. 1242, 1985.

[85] B. Revision and S. Ritt, “DRS4 Evaluation Board User’s Manual,” no. January, 2014.

[86] A. Enqvist, C. C. Lawrence, B. M. Wieger, S. A. Pozzi, T. N. Massey, and A. Section, “Neutron light output response and resolution functions in EJ-309 liquid scintillation detectors,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 715, no. 2013 SRC-GoogleScholar FG-0, pp. 79–86, 2013.

[87] F. Pino, L. Stevanato, D. Cester, G. Nebbia, L. Sajo-Bohus, and G. Viesti, “The light output and the detection efficiency of the liquid scintillator EJ-309,” Appl. Radiat. Isot., vol. 89, pp. 79–84, 2014.

[88] F. A. Verser, “Gamma-gamma angular correlation in the decay of cobalt 60,” 1960.

[89] T. Zhu et al., “Improved fission neutron energy discrimination with Helium-4 detectors through pulse filtering,” Nucl. Instruments Methods Phys. Res. Sect. A Accel. Spectrometers, Detect. Assoc. Equip., vol. 848, pp. 137–143, 2016.

[90] M. Flaska and S. A. Pozzi, “Identification of shielded neutron sources with the liquid scintillator BC-501A using a digital pulse shape discrimination method,” NIM:A, vol. 577, no. 3, pp. 654–663, 2007.

130

[91] J. Scherzinger et al., “A comparison of untagged gamma-ray and tagged-neutron yields from241AmBe and238PuBe sources,” Appl. Radiat. Isot., vol. 127, pp. 98–102, 2017.

[92] S. D. Clarke, M. C. Hamel, A. Di fulvio, and S. A. Pozzi, “Neutron and gamma-ray energy reconstruction for characterization of special nuclear material,” Nucl. Eng. Technol., vol. 49, no. 6, pp. 1354–1357, 2017.

[93] S. Agosteo, C. Birattari, M. Caravaggio, M. Silari, and G. Tosi, “Secondary neutron and photon dose in proton therapy,” Radiother. Oncol., vol. 48, no. 3, pp. 293–305, 1998.

[94] S. F. Kry et al., “A Monte Carlo model for calculating out-of-field dose from a Varian 6 MV beam,” Med. Phys., vol. 33, no. 11, pp. 4405–4413, 2006.

[95] S. F. Kry et al., “A Monte Carlo model for out-of-field dose calculation from high-energy photon therapy,” Med. Phys., vol. 34, no. 9, pp. 3489–3499, 2007.

[96] S. F. Kry et al., “Out-of-field photon and neutron dose equivalents from step-and-shoot intensity-modulated radiation therapy,” Int. J. Radiat. Oncol. Biol. Phys., vol. 62, no. 4, pp. 1204–1216, 2005.

[97] R. Tayama et al., “Measurement of neutron dose distribution for a passive scattering nozzle at the Proton Medical research Center (PMRC).,” Nucl. Instr. Meth A., vol. 564, pp. 532–536, 2006.

[98] X. Yan, U. Titt, A. . Koehler, and W. . Newhauser, “Measurement of neutron dose equivalent to proton therapy patients outside of the proton radiation field,” Nucl. Instruments Methods Phys. Res., vol. 476, no. 1–2, pp. 429–434, 2002.

[99] U. Schneider, S. Agosteo, E. Pedroni, and J. Besserer, “Secondary neutron dose during proton therapy using spot scanning.,” J Radiat Oncol., vol. 53, no. 1 SRC-GoogleScholar FG-0, pp. 244–451, 2002.

[100] CNMC Company, “Solid Water® Phantom Materials Sheet.” .

[101] U.S. NRC, “20.1004 Units of radiation dose.”Available: https://www.nrc.gov/reading-rm/doc-collections/cfr/part020/part020-1004.html#N_3_201004.

[102] TOPAS MC Inc., “TOPAS Documentation,” 2018. .

[103] H. M. Lu and H. Kooy, “Optimization of current modulation function for proton spread-out Bragg peak fields,” Med. Phys., vol. 33, no. 5, pp. 1281–1287, 2006.

131

BIOGRAPHICAL SKETCH

Yinong Liang received her Bachelor of Science degree in nuclear science in

2014 from North China Electric Power University (NCEPU). Afterward, she came to the

University of Florida (UF) and joined Dr. Enqvist’s group to purse her master’s and

Ph.D. in nuclear engineering. Her interest areas include neutron detection, spectrum

unfolding, and medical physics.