Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Congressional research reportJun 4, 2010
Ask Donna
What actually matters in this document.
Text
Detection of Nuclear Weapons and Materials:
Science, Technologies, Observations
Jonathan Medalia
Specialist in Nuclear Weapons Policy
June 4, 2010
Congressional Research Service
7-....
www.crs.gov
R40154
CRS Report for Congress
Prepared for Members and Committees of Congress
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Summary
Detection of nuclear weapons and special nuclear material (SNM, plutonium, and certain types of
uranium) is crucial to thwarting nuclear proliferation and terrorism and to securing weapons and
materials worldwide. Congress has funded a portfolio of detection R&D and acquisition
programs, and has mandated inspection at foreign ports of all U.S.-bound cargo containers using
two types of detection equipment.
Nuclear weapons contain SNM, which produces suspect signatures that can be detected. It emits
radiation, notably gamma rays (high-energy photons) and neutrons. SNM is dense, so it produces
a bright image on a radiograph (a picture like a medical x-ray) when x-rays or gamma rays are
beamed through a container in which it is hidden. Using lead or other shielding to attenuate
gamma rays would make that image larger. Nuclear weapons produce detectable signatures, such
as radiation or a noticeable image on a radiograph. Other detection techniques are also available.
Nine technologies illustrate the detection portfolio: (1) A new scintillator material to improve
detector performance and lower cost. This project was terminated in January 2010. (2) GADRAS,
an application using multiple algorithms to determine the materials in a container by analyzing
gamma-ray spectra. If materials are the “eyes and ears” of detectors, algorithms are the “brains.”
(3) A project to simulate large numbers of experiments to improve detection system performance.
(4, 5) Two Cargo Advanced Automated Radiography Systems (CAARS) to detect high-density
material based on the principle that it becomes less transparent to photons of higher energy,
unlike other material. (6) A third CAARS to detect material with high atomic number (Z, number
of protons in an atom’s nucleus) based on the principle that Z affects how material scatters
photons. This project was terminated in March 2009. (7) A system to generate a 3-D image of the
contents of a container based on the principle that Z and density strongly affect the degree to
which muons (a subatomic particle) scatter. (8) Nuclear resonance fluorescence imaging to
identify materials based on the spectrum of gamma rays a nucleus emits when struck by photons
of a specific energy. (9) The Photonuclear Inspection and Threat Assessment System to detect
SNM up to 1 km away, unlike other systems that operate at very close range. It would beam highenergy photons at distant targets to stimulate fission in SNM, producing characteristic signatures
that may be detected. These technologies are selected not because they are necessarily the “best”
in their categories, but rather to show a variety of approaches, in differing stages of maturity,
performed by different types of organizations, relying on different physical principles, and
covering building blocks (materials, algorithms, models) as well as systems, so as to convey
many points on the spectrum of detection technology development.
This analysis leads to several observations for Congress. It is difficult to predict the schedule or
capabilities of new detection technologies. It is easier and less costly to accelerate a program in
R&D than in production. “Concept of operations” is crucial to detection system effectiveness.
Congress may wish to address gaps and synergisms in the technology portfolio. Congress need
not depend solely on procedures developed by executive agencies to test detection technologies,
but may specify tests an agency is to conduct. Ongoing improvement in detection capabilities
produces uncertainties for terrorists that will increase over time, adding deterrence beyond that of
the capabilities themselves.
This report will be updated occasionally.
Congressional Research Service
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Contents
Chapter 1. Nuclear Weapons and Materials: Signatures and Detection .........................................2
What Is to Be Detected?........................................................................................................2
Photons 101 ....................................................................................................................2
What signatures show the presence of nuclear weapons and SNM? .................................3
Opacity to photons ..........................................................................................................3
How Does Detection Work? ..................................................................................................5
How are signatures gathered, processed, and used?..........................................................5
Principles of detection.....................................................................................................6
Means of detection........................................................................................................ 10
Evasion of detection technologies ................................................................................. 12
Current Detection Technologies .......................................................................................... 13
Radiation “pagers” ........................................................................................................ 13
Radiation portal monitors.............................................................................................. 13
Radioactive isotope identification devices ..................................................................... 13
Radiographic imaging systems ...................................................................................... 13
Chapter 2. Advanced Technologies............................................................................................ 13
Nanocomposite Scintillators................................................................................................ 15
The problem.................................................................................................................. 15
Background .................................................................................................................. 15
Technology description ................................................................................................. 16
Potential advantages...................................................................................................... 16
Status, schedule, and funding ........................................................................................ 17
Risks and concerns........................................................................................................ 17
Potential gains by increased funding.............................................................................. 18
Potential synergisms and related applications ................................................................ 19
GADRAS: A Gamma-Ray Spectrum Analysis Application Using Multiple Algorithms........ 19
The problem.................................................................................................................. 19
Background .................................................................................................................. 19
Technology description ................................................................................................. 20
Potential advantages...................................................................................................... 22
Status, schedule, and funding ........................................................................................ 22
Risks and concerns........................................................................................................ 23
Potential gains by increased funding.............................................................................. 24
Potential synergisms and related applications ................................................................ 24
Computer Modeling to Evaluate Detection Capability ......................................................... 24
The problem.................................................................................................................. 24
Background .................................................................................................................. 25
Technology description ................................................................................................. 27
Potential advantages...................................................................................................... 28
Status, schedule, and funding ........................................................................................ 29
Risks and concerns........................................................................................................ 29
Potential gains by increased funding.............................................................................. 31
Potential synergisms and related applications ................................................................ 31
L-3 CAARS: A Low-Risk Dual-Energy Radiography System.............................................. 31
The problem.................................................................................................................. 31
Background .................................................................................................................. 32
Technology description ................................................................................................. 35
Congressional Research Service
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Potential benefits........................................................................................................... 36
Status, schedule, and funding ........................................................................................ 37
Risks and concerns........................................................................................................ 39
Potential gains by increased funding.............................................................................. 41
Potential synergisms and related applications ................................................................ 42
SAIC CAARS: A Higher-Risk, Higher-Benefit Dual-Energy Radiography System .............. 42
Technology description ................................................................................................. 42
Potential benefits........................................................................................................... 46
Status, schedule, and funding ........................................................................................ 47
Risks and concerns........................................................................................................ 48
Potential gains by increased funding.............................................................................. 49
Potential synergisms and related applications ................................................................ 49
AS&E CAARS: Using Backscattered X-Rays to Detect Dense Material ............................. 49
Background .................................................................................................................. 49
Technology description ................................................................................................. 51
Potential benefits........................................................................................................... 53
Status, schedule, and funding ........................................................................................ 53
Risks and concerns........................................................................................................ 54
Potential gains by increased funding.............................................................................. 55
Potential synergisms and related applications ................................................................ 56
Muon Tomography.............................................................................................................. 56
The problem.................................................................................................................. 56
Background .................................................................................................................. 57
Technology description ................................................................................................. 57
Potential benefits........................................................................................................... 64
Status, schedule, and funding ........................................................................................ 64
Risks and concerns........................................................................................................ 65
Potential gains by increased funding.............................................................................. 69
Potential synergisms and other applications................................................................... 69
Scanning Cargo or Analyzing a Terrorist Nuclear Weapon with Nuclear Resonance
Fluorescence .................................................................................................................... 70
Two problems ............................................................................................................... 70
Background .................................................................................................................. 71
Technology description ................................................................................................. 72
Potential benefits........................................................................................................... 73
Status, schedule, and funding ........................................................................................ 74
Risks and concerns........................................................................................................ 75
Potential gains by increased funding.............................................................................. 77
Potential synergisms and related applications ................................................................ 77
Detecting SNM at a Distance .............................................................................................. 78
The problem.................................................................................................................. 78
Background .................................................................................................................. 78
Technology description ................................................................................................. 79
Potential benefits........................................................................................................... 80
Status, schedule, and funding ........................................................................................ 81
Risks and concerns........................................................................................................ 81
Potential gains by increased funding.............................................................................. 82
Potential synergisms and related applications ................................................................ 83
Chapter 3. Observations ............................................................................................................ 83
Observations on Progress in Detection Technology.............................................................. 83
Congressional Research Service
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Observations on Technical Progress and Congress............................................................... 87
Observations on Technical Progress and Terrorism .............................................................. 89
What Is to Be Detected?...................................................................................................... 91
Background ........................................................................................................................ 91
Photons......................................................................................................................... 91
Radioactivity................................................................................................................. 92
Fissile material.............................................................................................................. 93
Detection ...................................................................................................................... 93
Shielding and background radiation............................................................................... 94
Signatures of Plutonium, Highly Enriched Uranium, and Nuclear Weapons ......................... 95
Atomic number and density........................................................................................... 95
Opacity to photons ........................................................................................................ 96
Presence of gamma rays beyond background levels ....................................................... 96
Presence of neutrons beyond background levels ............................................................ 96
Gamma ray spectra ....................................................................................................... 96
Time pattern of neutrons and gamma rays ..................................................................... 97
Fission chain time signature .......................................................................................... 99
Detecting Signatures of a Nuclear Weapon or SNM............................................................. 99
Overview: How are signatures gathered, processed, and used? ...................................... 99
How detectors work .................................................................................................... 100
Detecting gamma rays................................................................................................. 103
Detecting neutrons ...................................................................................................... 104
Detecting absorption or scattering of high-energy photons........................................... 105
Evasion of Detection Technologies.................................................................................... 105
Figures
Figure 1. Gamma-Ray Spectra: 90% Uranium-235 vs. Background .............................................4
Figure 2. Gamma-Ray Spectra: Weapons-Grade Plutonium vs. Background ...............................5
Figure 3. Gamma-Ray Spectra: 90% Uranium-235 vs. Background, Taken with a PVT
Detector ...................................................................................................................................7
Figure 4. Gamma-Ray Spectra of Plutonium-239 ........................................................................8
Figure 5. Resolution of the Cesium-137 Gamma-Ray Spectrum by Different CZT
Detectors Has Improved Over Time ....................................................................................... 10
Figure 6. How GADRAS Identifies Special Nuclear Material.................................................... 21
Figure 7. A Notional Receiver Operating Characteristic (ROC) Curve ....................................... 25
Figure 8. Three Notional ROC Curves....................................................................................... 26
Figure 9. Dual-Energy Radiography .......................................................................................... 34
Figure 10. L-3 CAARS Schematic Drawings............................................................................. 36
Figure 11. SAIC CAARS Prototype .......................................................................................... 43
Figure 12. Dual-Energy Radiography with False Colors Assigned ............................................. 45
Figure 13. EZ-3DTM Differentiates Between Elements by Atomic Number (Z)........................... 50
Figure 14. Schematic Diagram of EZ-3D Technique.................................................................. 52
Congressional Research Service
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Figure 15. Schematic Drawing of Muon Tomography Inspection Station Configurations:
Tunnel and Top/Bottom.......................................................................................................... 58
Figure 16. Muon Tomography Resolution Increases with Scan Time ......................................... 60
Figure 17. Muon Tomography Creates Three-Dimensional Images ........................................... 61
Figure 18. Schematic Diagram of Nuclear Resonance Fluorescence System ............................. 72
Figure 19. Schematic Diagram of Photonuclear Inspection and Threat Assessment System
(PITAS) ................................................................................................................................. 80
Appendixes
Appendix. The Physics of Nuclear Detection............................................................................. 91
Contacts
Author Contact Information .................................................................................................... 106
Congressional Research Service
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
C
ongress has been interested in detecting nuclear weapons and the special materials needed
to make them for many years, especially since 9/11. Nuclear detection has many
applications for countering nuclear terrorism and nuclear proliferation, such as securing
nuclear weapons and materials in U.S., Russian, and other nuclear facilities, tracking materials at
border crossings and choke points, screening maritime cargo containers, and examining actual or
suspect nuclear sites.
The United States currently uses several types of nuclear detection equipment. All have
significant shortcomings. Some work only at very short range; some cannot identify the material
emitting radiation, which can lead to false alarms and interrupt commerce; some depend on
operator skill, and may be defeated by a clever smuggler or a sleepy operator; and some are easily
defeated by shielding. In an effort to overcome such problems, Congress has funded a pipeline of
advanced-technology research, development, and acquisition.
This report seeks to help Congress understand this technology. It discusses the science of
detecting nuclear weapons and materials, describes nine advanced U.S. technologies selected to
illustrate the range of projects in the pipeline, and offers observations for Congress. The report
does not compare technologies. 1 The inclusion of the nine technologies should not be taken to
mean that CRS judges them to be better than the hundreds of others not considered here. The
report does not discuss the controversial Advanced Spectroscopic Portal because a detailed
discussion of it could draw attention from the other technologies considered here.
The scope of this report excludes the organization of the government for dealing with nuclear
detection, the role of intelligence and law enforcement in detecting terrorist nuclear weapons,
detection of radiological dispersal devices (such as “dirty bombs”), the role of nuclear forensics
in attributing an attack to its perpetrator, response to a nuclear attack, and the architecture of a
national nuclear detection system. 2 Nor does it discuss possible means by which terrorists might
acquire a bomb, or whether they could make a bomb on their own. Much has been written on
these topics. 3 While many who are concerned with nuclear detection focus on thwarting nuclear
terrorism, this report focuses on technology per se. It avoids extensive discussion of means to
counter detection to avoid classified information.
Nuclear detection technology has a dual role in thwarting a terrorist nuclear attack—deterrence
and defense. Deterrence means dissuasion from an action by threat of unacceptable consequences.
Terrorists may be deterred from a nuclear strike by one of the few consequences unacceptable to
them: failure. Detection systems would raise that risk. These systems could also make a terrorist
nuclear strike too complex to succeed. But other factors would also have these effects: the
difficulty of fabricating a bomb, the chance that law enforcement or intelligence would detect
efforts to obtain a bomb, the possible inability to detonate a purloined bomb, and the risk that
1
Comparison would require a detailed review of hundreds of technology projects to determine which are most worthy
of further examination; creating metrics to compare the selected projects; and obtaining accurate data for use in the
metrics. Each of these tasks would require the work of many experts over many months.
2
For further information, see CRS Report RL34574, The Global Nuclear Detection Architecture: Issues for Congress,
by (name redacted).
3
See Charles Ferguson and William Potter, The Four Faces of Nuclear Terrorism, Monterey, CA, Center for
Nonproliferation Studies, 2004; Michael Levi, On Nuclear Terrorism, Cambridge, MA, Harvard University Press,
2007; and Carson Mark et al., “Can Terrorists Build Nuclear Weapons?,” in Paul Leventhal and Yohan Alexander,
Preventing Nuclear Terrorism: The Report and Papers of the International Task Force on Prevention of Nuclear
Terrorism, a Nuclear Control Institute book, Lexington, MA, Lexington Books, 1987, pp. 55-65.
Congressional Research Service
1
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
scientists recruited for the plot would defect. Such risks would disappear, however, if terrorists
were given a bomb and operating instructions. They would then only need to mount a smuggling
operation. In that case, the role of nuclear detection systems would change: they would become
the main defense.
Chapter 1. Nuclear Weapons and Materials:
Signatures and Detection
As background for understanding the detection technologies in Chapter 2, this chapter outlines
nuclear detection science. The Appendix provides more detail.
What Is to Be Detected?
Detection focuses on weapons and the nuclear materials that fuel them. Weapons can be small. In
the Cold War, the United States built high-yield weapons several feet long, atomic demolition
munitions that a soldier could carry, and nuclear artillery shells. A terrorist-made weapon would
probably be larger.
Nuclear weapons require fissile material, atoms of which can fission (split) when struck by fast or
slow neutrons4; pieces of this material can support a nuclear chain reaction. The fissile materials
used in nuclear weapons are uranium, isotope 235 (U-235), and plutonium, isotope 239 (Pu-239).
The Atomic Energy Act of 1954 designates them as “special nuclear material” (SNM).5 Uranium
in nature is 99.3% U-238 and 0.7% U-235; U-235 must be concentrated, or “enriched.” Uranium
enriched to 20% U-235 is termed highly enriched uranium, or HEU, but nuclear weapons
typically use uranium enriched to 90% or so. In this report, HEU refers to this weapons-grade
enrichment level. Weapons-grade plutonium, or WGPu, is also a mix of isotopes, at least 93% Pu239. According to one account by five nuclear weapon scientists from Los Alamos National
Laboratory, it would take 26 kg of HEU or 5 kg of WGPu to fuel a bomb,6 amounts that would fit
into cubes 11 cm or 6 cm, respectively, on a side.
Photons 101
Nuclear detection makes extensive use of photons, packets of energy with no rest mass and no
electrical charge. Electromagnetic radiation consists of photons, and may be measured as
wavelength, frequency, or energy; for consistency, this report uses only energy, expressed in units
of electron volts (eV). 7 Levels of energy commonly used in nuclear detection are thousands or
4
Some materials can fission only when struck by fast neutrons; fissile materials are the only materials that can fission
when struck by slow as well as fast neutrons.
5
The Atomic Energy Act of 1954, 42 U.S.C. 2014, defines SNM as uranium enriched in the isotopes 233 or 235 or
plutonium. The Nuclear Regulatory Commission has not declared any other material to be SNM even though the Act
permits it to do so. U.S. Nuclear Regulatory Commission. “Special Nuclear Material,” http://www.nrc.gov/materials/
sp-nucmaterials.html.
6
Mark et al., “Can Terrorists Build Nuclear Weapons?”
7
An electron volt is a very small unit of energy, “a unit of energy equal to the work done by an electron accelerated
through a potential difference of 1 volt.” http://wordnetweb.princeton.edu/perl/webwn?s=electron%20volt.
Congressional Research Service
2
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
millions of electron volts, keV and MeV, respectively. The electromagnetic spectrum ranges from
radio waves (some of which have photon energies of 10-9 eV), through visible light (a few eV), to
higher-energy x-rays (10 keV and up) and gamma rays (mostly 100 keV and up).
X-ray photons and gamma-ray photons of the same energy are identical. However, they are
generated in different ways. Gamma rays originate in processes in an atom’s nucleus. Each
radioactive isotope that emits gamma rays does so in a unique energy spectrum, as in Figure 1,
which is the only way to identify an isotope outside a laboratory. A detector with a form of
“identify” or “spectrum” in its name, such as Advanced Spectroscopic Portal or radioactive
isotope identification device, identifies isotopes by their gamma-ray spectra. X-rays originate in
interactions with an atom’s electrons. Many detection systems use x-ray beams, which can have
higher energies than gamma rays and thus are more penetrating. X-ray beams are often generated
through the bremsstrahlung process, German for “braking radiation,” which works as follows. An
accelerator creates a magnetic field that accelerates charged particles, such as electrons, which
slam into a target of heavy metal. When they slow or change direction as a result of interactions
with atoms, they release energy as x-rays whose energy levels are distributed from near zero to
the energy of the electron beam. They do not exhibit spectral peaks like gamma rays.
What signatures show the presence of nuclear weapons and SNM?
For purposes of this report, a signature is a property by which a substance (in particular, SNM)
may be detected or identified. This section presents several signatures. The Appendix and
Chapter 2 discuss others.
Atomic number and density
Atomic number, abbreviated “Z,” is the number of protons in an atom’s nucleus. It is a property
of individual atoms. In contrast, density is a bulk property, expressed as mass per unit volume. In
general, the densest materials are those of high Z. These properties may be used to detect uranium
and plutonium. Uranium is the densest and highest-Z element found in nature (other than in trace
quantities); plutonium has a slightly higher Z (94), and its density varies from slightly more to
slightly less than that of uranium. Some detection methods discussed in Chapter 2, such as
effective Z, make use of Z, and some, such as radiography and muon tomography, make use of Z
and density combined.
Opacity to photons
An object’s opacity to a photon beam depends on its Z and density, the amount of material in the
path of the beam, and the energy of the photons. Gamma rays and x-rays can penetrate more
matter than can lower-energy photons, but dense, high-Z material absorbs or scatters them. Thus a
way to detect an object, such as a bomb, in a container is to beam in x-rays or gamma rays to
create a radiograph (an opacity map) like a medical x-ray.
Radioactivity
Radioactive atoms are unstable and give off various types of radiation; the types of use for
nuclear detection are gamma rays and neutrons.
Congressional Research Service
3
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Gamma rays . Gamma-ray spectra are well characterized for each isotope. Figure 1 and Figure 2
show spectra for HEU and WGPu. Each point on the spectrum shows the number of photons
emitted (vertical axis) at each energy level (horizontal axis). Background gamma radiation is
ubiquitous. Since many materials, including SNM, emit gamma radiation, elevated levels of
gamma radiation may or may not indicate the presence of SNM, but careful analysis of the total
gamma-ray spectrum, as discussed in Chapter 2 under GADRAS, may reveal the presence of
SNM. Of particular interest, uranium that has been through a nuclear reactor picks up a very
small amount of uranium-232, which decays through intermediate steps to thallium-208. The
latter has a half-life of 3 minutes, and its decay produces a gamma ray of 2.614 MeV (as shown in
Figure 1), one of the highest-energy gamma rays. As a result, it is distinctive as well as highly
penetrating, facilitating detection.8
Figure 1. Gamma-Ray Spectra: 90% Uranium-235 vs. Background
Source: Lawrence Livermore National Laboratory.
Notes: Spectra were taken with a high-purity germanium detector at a distance of 1 m and computed with the
GADRAS algorithm, discussed below. Uranium is unshielded, background is dirt. The uranium-235 line starts as
the upper line (in red) at far left.
8
For further information on gamma rays generated by uranium, as well as shielding and detection of uranium, see
Bernard Phlips et al., “Comparison of Shielded Uranium Passive Gamma-Ray Detection Methods,” Proceedings of
SPIE, vol. 6213, 62130H (2006); doi:10.1117/12.666342, online publication date May 24, 2006; abstract available at
http://spiedl.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=PSISDG00621300000162130H000001&idtype=
cvips&gifs=yes&ref=no; and J.R. Lemley et al., “Confirmatory Measurements for Uranium in Nuclear Weapons by
High-Resolution Gamma-Ray Spectrometry (HRGS),” Brookhaven National Laboratory, BNL-66293, July 25, 1999,
http://www.osti.gov/bridge/product.biblio.jsp?osti_id=750764.
Congressional Research Service
4
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Figure 2. Gamma-Ray Spectra: Weapons-Grade Plutonium
vs. Background
Source: Lawrence Livermore National Laboratory.
Notes: Spectra were taken with a high-purity germanium detector at a distance of 1 m and computer with
GADRAS. Plutonium is unshielded, background is dirt. The plutonium line starts as the upper line (in red) at the
far left.
Neutrons . Atoms of some heavy elements fission. Of the naturally occurring elements, only U238 spontaneously fissions at an appreciable rate. Fission releases neutrons. U-235 emits few
neutrons, but because U-238, the other main component of HEU, emits some, 1 kg of HEU emits
3 neutrons per second,9 so it provides a weak signature. Plutonium emits on the order of 60,000
neutrons per kg per second depending on the mix of plutonium isotopes.10 Unlike gamma rays,
neutrons do not have a characteristic energy spectrum by which an isotope can be identified.
How Does Detection Work?
How are signatures gathered, processed, and used?
Detection involves using detector elements to obtain data, converting data to usable information
through algorithms, and acting on that information through concept of operations, or CONOPS.
9
Roger Byrd et al., “Nuclear Detection to Prevent or Defeat Clandestine Nuclear Attack,” IEEE Sensors Journal,
August 2005, p. 594.
10
Ibid.
Congressional Research Service
5
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Detectors, algorithms, and CONOPS are the eyes and ears, brains, and hands of nuclear detection:
effective detection requires all three.
Since photons and neutrons have no electrical charge, their energy is converted to electrical
pulses that can be measured. This is the task of detectors, discussed next. The pulses are fed to
algorithms. An algorithm, such as a computer program, is a finite set of logical steps for solving a
problem. For nuclear detection, an algorithm must process data into usable information fast
enough to be of use to an operator. It receives data from a detector’s hardware, such as pulses
representing the time and energy of each photon arriving at the detector. It converts the pulses to a
format that a user can understand, such as displaying a gamma ray spectrum or flashing “alarm.”
Every detector uses one or more algorithms. Improvements to algorithms can contribute as much
as hardware improvements to detector capability.
CONOPS may be divided into two parts. One specifies how a detection unit is to be operated to
obtain data. Elements include: How many containers must the unit scan per hour? How close
would a detector be to a container? Shall the unit screen cargo in a single pass, or shall it be used
for primary screening, with suspicious cargo sent for a more detailed secondary screening? A
second part details how the data are to be used. Elements include: What happens if the equipment
detects a possible threat? Which alarms are to be resolved on-site and which are to be referred to
off-site experts? Under what circumstances would a port or border crossing be closed? More
generally, how is the flow of data managed, in both directions?11 What types of intelligence
information do inspectors receive, and how do data from detection systems flow to federal, state,
and local officials for analysis or action? While this report does not focus on CONOPS because it
is not a technology, it is an essential part of nuclear detection.
Principles of detection
Detectors require a signal-to-noise ratio high enough to permit detection. That is, they must
extract the true signal (such as a gamma-ray spectrum) from noise (such as spurious signals
caused by background radiation). Two concepts are central to gamma-ray detector sensitivity:
detection efficiency and spectral resolution. Efficiency refers to the amount of signal a detector
records. Radiation intensity (e.g., number of photons per unit of area) diminishes with distance.
Since a lump of SNM emits radiation in all directions, using a detector that is larger, or that is
closer to the SNM, increases the fraction of radiation from the source that impinges on the
detector and thereby increases efficiency. Another aspect is the fraction of the radiation striking
the detector that creates a detectable signal. A more efficient detector collects data faster, reducing
the time to screen a cargo container.
Spectral resolution refers to the sharpness of peaks in a gamma-ray spectrum. A perfect detector
would record a spectrum as vertical “needles” because each radioactive isotope releases gamma
rays only at specific energies. Since detectors are not perfect, each energy peak is recorded as a
bell curve. The narrower the curve, the more useful the data. Polyvinyl toluene (PVT), a plastic
used in radiation detectors that can be fielded in large sheets at low cost, is efficient but has poor
resolution. It can detect radiation, but peaks from gamma rays of different energies blur together,
which can make it impossible to identify an isotope. Figure 3 shows the spectra of 90% U-235
and background radiation as recorded by a PVT detector. In contrast, high-purity germanium
11
For further analysis of this topic, see CRS Report RL34070, Fusion Centers: Issues and Options for Congress, by
John Rollins.
Congressional Research Service
6
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
(HPGe) produces sharp peaks, permitting clear identification of specific isotopes. These detectors
are expensive, heavy, have a small detector area, and must be cooled to extremely low
temperatures with liquid nitrogen or a mechanical system, making them less than ideal for use in
the field. Figure 4 shows the spectrum of Pu-239 as recorded by detectors with better resolution
than PVT.
Figure 3. Gamma-Ray Spectra: 90% Uranium-235 vs. Background,Taken with a PVT
Detector
Source: Lawrence Livermore National Laboratory.
Notes: Spectra were taken with a PVT detector at a distance of 1 m and computed with GADRAS. Uranium is
unshielded, background is dirt. The uranium-235 line starts as the upper line (in red) at far left.
Congressional Research Service
7
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Figure 4. Gamma-Ray Spectra of Plutonium-239
Source: Los Alamos National Laboratory.
Notes: These are spectra of 64% plutonium-239. They were taken using different detector materials to show
differences in resolution: from top to bottom, sodium iodide mixed (“doped”) with thallium; cadmium-zinctelluride; cadmium telluride; and high-purity germanium. This figure was created some years ago; the sensitivity
of detector materials has improved since then.
Sensitivity can be improved. (1) One type of detector is cadmium-zinc-telluride (CZT) crystals.
Better crystals and better ways to overcome their limitations have improved sensitivity. The peak
on the right of each spectrum in Figure 5 shows the cesium-137 spectrum taken with CZT
detectors that, for the years indicated, were at the high end of sensitivity. (2) Detectors build
radiography images or gamma-ray spectra over time. With more time, a detector can collect more
data, in the form of gamma rays or neutrons. More data improve separation of signal from noise
and reduce false alarms. Longer scan times improve accuracy but impede the flow of commerce,
costing money, so a balance is sought between these two opposed goals. (3) Increasing the spatial
resolution of a detector improves sensitivity:
Congressional Research Service
8
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
This is easily demonstrated in the example of a shielded versus unshielded radiation detector.
Unshielded detectors are sensitive to radiation impinging on it in all directions, which is
characteristic of the nature of naturally-occurring background radiation. By adding shielding,
a detector’s field-of-view can be controlled, and background radiation levels reduced,
increasing the signal-to-noise ratio for the detector in the direction the detector is aimed.12
12
Personal communication, Defense Threat Reduction Agency, August 8, 2008.
Congressional Research Service
9
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Figure 5. Resolution of the Cesium-137 Gamma-Ray Spectrum by Different CZT
Detectors Has Improved Over Time
Source: Brookhaven National Laboratory.
Notes: Spectra were taken with cadmium-zinc-telluride (CZT)-based detectors that were high-end in each
period listed. The peak on the far right of each spectrum is of particular value for identification of radioisotopes;
it becomes sharpers (narrower and higher) with each successive detector. As discussed in the Appendix, many
factors were responsible for this improvement.
Means of detection
Nuclear detection uses neutrons and photons in various ways. Because either neutrons or photons
can readily penetrate most materials, they are the main forms of radiation used to detect
Congressional Research Service
10
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
radioactive material passively, such as by sensing radiation coming out of a cargo container.
Gamma rays and x-rays can be used in an active mode to probe a container for dense material
through radiography, which creates an x-ray-type image. Neutrons of any energy level, and
photons above about 5.6 MeV, can be beamed into a container to induce fission in SNM. Fission
results in the emission of neutrons and gamma rays, which can be detected.
Detecting gamma rays
Gamma rays do not have an electrical charge, but an electrical signal is needed to measure them.
There are two main ways to turn a gamma ray into electrical energy. One is with a scintillator
material, such as PVT. When a gamma ray interacts with this material, it emits many lowerenergy photons of visible light. A photomultiplier tube converts them to electrons, then multiplies
the electrons to generate a measurable pulse of electricity whose voltage is proportional to the
number of lower-energy photons, which in turn is proportional to the energy deposited by the
gamma ray. An electronic device called a multi-channel analyzer sorts the pulse into a “bin”
depending on its energy and increases the number of counts in that bin by one. A software
package draws a histogram with energy on the horizontal axis and counts on the vertical axis. The
histogram is the gamma-ray spectrum for that isotope. In contrast, a semiconductor material, such
as HPGe, turns gamma rays directly into an electrical signal proportional to the gamma-ray
energy deposited. A voltage is applied across the material, with one side of the material the
positive electrode and the other the negative electrode. When a gamma ray interacts with the
material, it knocks electrons loose from the semiconductor’s crystal lattice. The voltage sweeps
them to the positive electrode. Their motion produces an electric current whose voltage is
proportional to the energy of each gamma ray. Each pulse of current is then sorted into a bin
depending on its voltage and the spectrum is computed as described above.
Detecting neutrons
A common neutron detector is a tube of helium-3 gas linked to a power supply, with positively
and negatively charged plates or wires in the tube. In its rest state, current cannot pass through the
helium because it acts as an insulator. When a low-energy neutron passes through the tube, a
helium-3 atom absorbs it, producing energetic charged particles that lose their energy by
knocking electrons off other helium-3 atoms. Positively charged particles move to the negative
plate; electrons move to the positive plate. Since electric current is the movement of charged
particles, these particles generate a tiny electric current that is counted. Neutron count is an
important way to identify SNM because SNM and U-238 emit neutrons spontaneously in
significant numbers. Few other sources do. Neutron spectra are of little value for identifying
isotopes. They do not have lines representing discrete energies, and neutrons lose energy as they
collide with low-Z material, blurring their spectra. However, helium-3 has become extremely
scarce, and neutron detection systems for homeland security would require so much of it that
alternatives are being sought, such as boron-10.13
13
The American Association for the Advancement of Science held a workshop on the helium-3 shortage on April 6,
2010. Briefing slides are available at http://cstsp.aaas.org/agenda_meeting.html.
Congressional Research Service
11
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Detecting dense material
Photons of high enough energy can penetrate solid objects, but are scattered or absorbed by dense
objects (or a sufficient thickness of less-dense material). This is the basis for radiography. For
cargo scanning, x-rays or gamma rays are beamed through a container, and a detector on the other
side records the number of photons received in each pixel. An algorithm then creates an opacity
map of the contents. While a bomb would present a sizable image, dense objects in a container
might mask a piece of SNM.
Evasion of detection technologies
An enemy could use various means in an effort to defeat detection systems. One such means is
shielding. Gamma rays may travel many feet through such low-Z material as wood, food, and
plastic, but high-Z material absorbs and deflects them. Conversely, low-Z material absorbs and
scatters neutrons, but they pass more readily through higher-Z material. Different amounts of
material are needed to attenuate gamma rays depending on their quantity and energy level.
Gamma rays from WGPu are sufficiently energetic and plentiful that they are difficult to shield,
while a layer of lead would shield gamma rays from HEU, especially if it had not been through a
reactor and, in consequence, had not picked up uranium-232, as discussed above. This distinction
is of practical significance. Press reports indicate that Iran is using centrifuges to enrich uranium
from chemical forms derived from uranium ore, which have not been through a reactor.14
Sources of radiation other than SNM complicate detection. Background radiation from naturally
occurring radioactive material, such as thorium and uranium, is present everywhere, often in trace
amounts. Cosmic rays generate low levels of neutrons. Some commercial goods contain
radioactive material, such as ceramics (which may contain uranium) and kitty litter (which may
contain thorium and uranium). Other radioactive isotopes are widely used in medicine and
industry.
Enemy attempts to defeat one type of detection system may complicate a plot or make it more
detectable. (1) It is harder to defeat systems that detect multiple phenomena than a system
detecting one phenomenon only. For example, shielding a bomb with lead to attenuate gamma
rays would create a large, opaque image on a radiograph. For this reason, Congress mandated that
U.S.-bound containers loaded in foreign ports be “scanned by nonintrusive imaging equipment
and radiation detection equipment at a foreign port before it was loaded on a vessel.”15 (2) An
enemy could attempt salvage fuzing, which would detonate a weapon that sensed attempts to
detect it, such as with photon beams, or to open it. However, salvage fuzing could detonate a
weapon by accident; if it were scanned overseas; or in a U.S. port, where it would do much less
damage than in a city center. (3) Attempts to smuggle HEU into the United States to avoid
detection of a complete bomb would require fabricating the weapon inside this nation, which
could require such activities as smuggling in other weapon components and purchasing
specialized equipment, and could run the risk of accidents, any of which could provide clues to
law enforcement personnel.
14
See, for example, Joby Warrick, “Iran’s New Centrifuge Raises Concerns about Nuclear Aims,” Washington Post,
May 2, 2010, p. 11.
15
P.L. 110-53, Implementing Recommendations of the 9/11 Commission Act of 2007, Section 1701, 121 Stat. 489.
Congressional Research Service
12
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Current Detection Technologies
The United States deploys various technologies, such as the following, to detect nuclear weapons
or SNM. They are available but have important drawbacks.
Radiation “pagers”
These devices, about the size and shape of a pager, can detect radiation at close distance to alert
individuals to the presence of elevated levels of radiation. They may use any of several types of
detector material. They are lightweight and inexpensive, but cannot identify the material causing
an alarm.
Radiation portal monitors
Many of these devices use large sheets of plastic scintillator material, such as PVT, to detect
radiation coming from a vehicle. RPMs were deployed soon after 9/11 because they were
available at moderate cost. However, PVT cannot identify the source of radiation. Yet many items
in everyday commerce contain radioactive material. As a result, some RPMs produce many false
alarms, which may require considerable effort to resolve, delaying the flow of commerce. Newer
versions have some isotope identification capability.
Radioactive isotope identification devices
These devices are typically hand-held. They have software that can identify a radioisotope by its
gamma-ray spectrum. The most capable of these devices use a crystal of high-purity germanium,
a semiconductor material, and are considered the “gold standard” of all identification devices.
Such devices are heavy and delicate, and must be cooled with liquid nitrogen or by mechanical
means, limiting their usability in the field. They have a relatively short range for detecting
radiation sources with low radioactivity, notably shielded HEU, making them unsuitable as the
primary method of screening cargo containers.
Radiographic imaging systems
These devices send high-energy photons through cargo containers to create a radiographic image
of the contents. The radiograph is scanned, either automatically or by an operator, to search for
nuclear weapons, contraband, stowaways, and other illicit cargo. While a nuclear weapon would
show up as a white (or black) image on the radiograph and would be clearly visible if hidden in a
shipment of low-Z material like food or paper, an operator might overlook it if it were in a
shipment of other large, dense objects or jumbled items of various sizes and densities. A small
piece of SNM might also be overlooked.
Chapter 2. Advanced Technologies
Many nuclear detection technology projects are under way in the United States and elsewhere.
This section discusses nine U.S. technologies selected to include different (1) agencies
sponsoring projects (the Defense Threat Reduction Agency (DTRA), an agency of the
Department of Defense (DOD); the Domestic Nuclear Detection Office (DNDO), an agency of
Congressional Research Service
13
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
the Department of Homeland Security (DHS); and the National Nuclear Security Administration
(NNSA), an agency of the Department of Energy (DOE)), (2) organizations performing the work
(national laboratories, industry, universities, and collaborations between them), (3) types of
technology (materials, algorithms, simulation, systems), (4) physical principles (muon
tomography, radiography, stimulated emission of radiation, nuclear resonance fluorescence), (5)
distances between the detector and the object being scanned (near and far), and (6) levels of
maturity (in use for many years but constantly updated, near deployment, and anticipated to be
available for deployment in several years). This section does not consider technologies in the
earliest stages of development because it is too soon to tell how they will pan out. The discussion
of each technology includes several categories:
•
The problem that the technology addresses
•
Technology background
•
Description of the technology
•
Potential benefits that the technology offers
•
Status, schedule, and funding
•
Scientific, engineering, cost and schedule, and operational risks
•
Potential gains by increased funding
•
Potential synergisms
The last three categories require some explanation.
Risks: The discussion presents potential benefits of the technologies. It does not present “cons.”
That would be premature because development programs address problems. Instead, each section
discusses risk. There are several categories of risk. Scientific problems may thwart a technology.
Even if it is scientifically sound, it may be hard to engineer into a workable system. Even if it can
be engineered, it may be unaffordable, or take too long to develop. Even if it can surmount these
hurdles, problems encountered in field use may render it unattractive.
Potential gains by increased funding: In preliminary discussions, project managers asserted
they would, if given more funds, use the added funds to solve problems or exploit opportunities.
CRS therefore asked managers of all nine technologies considered in this report how they would
spend more money as a way to probe for problems and opportunities with their projects. Would
they pursue several promising routes to a technology instead of pursuing only one at the outset?
Would they buy more equipment so they could speed up the work? Would they hire more staff?
This analysis applies only to the nine technologies discussed in this report. It is not intended to
focus authorization or appropriation consideration solely on them. Other technologies not
considered here might realize larger (or smaller or no) gains through increased funding.
Potential synergisms: As the technology descriptions show, many systems have common
elements, such as certain types of detectors or algorithms, and work on one technology may
contribute to others or have applications beyond current plans.
This report now discusses each of the nine technologies using the above format.
Congressional Research Service
14
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Nanocomposite Scintillators16
As of January 2010, DTRA and DNDO terminated the nanocomposite scintillator project. This
section, therefore, will not be updated further, but continues in this report because it contains
valuable information on nuclear detection. Two small parts of the project, both described below,
are continuing as separate projects: the development of application-specific integrated circuits for
detecting “thermal” (very low energy) neutrons and identifying gamma-ray spectra
simultaneously, and an effort to use nanocomposite scintillator material for neutron detection.
The problem
Scintillator materials are used to detect, and in some cases identify, gamma rays. Higherperformance scintillators are more expensive, harder to manufacture, and fragile; scintillators that
are less costly, easier to manufacture, and more rugged offer lower performance. At issue: can the
desirable qualities of each type be combined to achieve better performance at lower cost?
Background
Scintillators are materials that, when struck by photons of higher energy, such as gamma rays,
capture this energy and release it as photons of lower energy, usually visible light. The material
should capture as much of the energy of each photon striking it as possible in order to build an
accurate photon energy spectrum and thus identify the material emitting the photons. Ideally, a
photon should deposit its full energy in the scintillator material, a so-called full energy
interaction. It is also important that the deposited energy be efficiently converted into photons of
visible light, which are then counted to determine energy.
There are two main types of scintillators. Inorganic scintillators (those not containing carbon) are
typically single crystals, such as sodium iodide (NaI) with a small amount of thallium added. The
probability of full energy interaction increases sharply with atomic number (Z) of the scintillator
material, 17 and is high for inorganic crystals. The more energy from each photon a scintillator
absorbs and then gives off, the better the correlation between energy input and output, and the
more precise the spectrum that can be constructed. As a result, a device using an inorganic crystal
has a good ability to identify the radioactive material producing a gamma-ray spectrum. There are
several drawbacks. The area of a detector that is sensitive to gamma rays is small (limited to the
size of a crystal), so the detector must be close to the object to be searched or must scan for
longer time so it can receive more gamma rays. They are fragile; dropping one can destroy it.
Many inorganic crystals absorb water and are sensitive to light, so they must be protected from
environmental conditions. NaI crystals are easy to grow, but cost about $3 per cubic centimeter
(cc) of crystal. Other, higher-resolution scintillators are harder to grow, are more costly, and are
sensitive to moisture. For NaI, light output varies strongly with temperature, so the temperature
must be stabilized or the data corrected.
16
The principal investigator for this project, Edward McKigney, Senior Project Leader, Safeguards Science and
Technology Group, Nuclear Nonproliferation Division, Los Alamos National Laboratory, provided detailed
information for this section, personal communications, April-August 2008. Others have commented as well to provide
alternative perspectives.
17
Specifically, the number of gamma rays depositing all their energy increases as Z to the 4.5th power.
Congressional Research Service
15
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Organic scintillators have the opposite set of properties. They can be made of plastic, such as
PVT. As such, they are easy and cheap to make, and are much less fragile than crystals. They can
be produced in bulk, making them suitable for deployment in large sheets, such as for radiation
portal monitors. On the other hand, since they are composed mostly of hydrogen and carbon, both
very low Z elements, they are very inefficient at absorbing the total energy of gamma rays. As a
result, as Figure 3 shows, peaks in PVT-produced gamma ray spectra are indistinct at best,
making such spectra of little or no value for identifying radioisotopes.
Technology description
A research project under way at Los Alamos National Laboratory, the Nanocomposite Scintillator
Project, seeks to combine the advantages of both types of scintillator materials to overcome the
disadvantages of each. The principle is that “nanocrystals,” crystals 2 to 5 nanometers in diameter
(1 nanometer = 1 billionth of a meter), of certain inorganic scintillator materials can capture most
of the energy from photons, thus offering nearly the performance of single large crystals, if
packed densely enough in plastic. The resulting mixture also has the desirable features of plastic.
The crystals are lanthanum bromide mixed with cerium, or cerium bromide. The modified
polystyrene plastic is a scintillator material, so it increases the amount of energy converted to a
detectable signal.
In operation, when a gamma-ray photon strikes this material, its energy is absorbed by
nanocrystals and the plastic, raising some atoms to a higher energy level. These atoms give off
this energy as photons in the visible and near-visible regions of the electromagnetic spectrum
(“optical photons”). A photodetector, an electronic component that generates many electrons for
each photon it receives, amplifies the signal and converts it from an optical signal to an electronic
signal that can be counted. The number of optical photons generated corresponds to the energy
level of the photon striking the material. A multi-channel analyzer counts the optical photons,
determines the energy level of the photon striking the material, and increases the count of photons
of that energy level by one, by this process creating a gamma-ray spectrum.
Edward McKigney, the principal investigator, believes that nanocomposite scintillator material
will be able to discriminate between neutrons and gamma rays. He asserts that simulations
support this case. The project has obtained data from experiments using the plastic without
nanocrystals and nanocrystals without the plastic, and from the literature, and has used these data
in simulations. Basic physics calculations also support this case. Neutrons generate protons when
they strike the plastic, and gamma rays generate electrons. The plastic responds differently to
protons and to electrons; the same is true of the nanocrystals. However, as of August 2008 the
project had not conducted experiments demonstrating the ability of the material to detect neutrons
and to differentiate between them and gamma rays.
Potential advantages
McKigney states that because the crystals are tiny, growing them is not difficult and is much less
costly than growing large whole crystals of these materials. The material can be fabricated at an
industrial scale, he says, further reducing cost. He estimates that this material potentially offers
the performance of $300/cc material (e.g., lanthanum bromide) at a cost of 50 cents/cc. Because
the material acts as a plastic, it is rugged and flexible, and can be made in large sheets, according
to McKigney, increasing the sensitivity of a detector using it. It would operate at ambient
temperatures.
Congressional Research Service
16
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Status, schedule, and funding
The project has several components. The largest is to develop nanocomposite scintillator material.
There are several smaller components: process scale-up; electronics development; detector
design; and basic research for a more advanced material. The budget for the entire project is as
follows. Early research started in 2004. Los Alamos provided $65,000 in initial funds in FY2005.
The laboratory and DNDO provided $1.2 million for FY2006 and $1.6 million for FY2007. Los
Alamos, DNDO, and DTRA provided $4.6 million for FY2008 and are projected to provide $5.5
million for FY2009. The project seeks to deliver a small cylinder (1 inch in diameter and 1 inch
high) of the material, and to characterize its performance, by December 2008. Anticipated costs
are $4 million for FY2009 for the nanocomposite scintillator component. The goal for the end of
FY2009 is to have a pilot-scale demonstration of the scintillator material and to start transferring
its technology to industry. This material would be optimized for gamma-ray detection. As of July
2009, the schedule for this demonstration had slipped by 6 to 12 months.18
Risks and concerns
Scientific risks and concerns
Fabrication of this material requires finding chemicals that can coat the surface of the
nanocrystals so they can disperse properly in the plastic while optimizing other properties.19 The
nanocrystals must be packed densely in the plastic to increase sensitivity and resolution. As of
May 2008, researchers had reached a packing level of 6% (by volume), with a goal of 50%. It
remains to be seen if they can meet this goal. Packing crystals densely in plastic will change some
properties of the plastic, 20 so care must be taken to minimize this problem. Another source of risk
is that, as of August 2008, the project had not reached high enough packing levels to allow for
measurements to determine the sharpness of gamma-ray peaks in spectra generated by this
material. The project plans to conduct experiments on this point by December 2008. Such
measurements are expected to reduce scientific risk. However, unexpected nanoscale physics
could impair energy resolution. In that case, the particle structure would have to be engineered to
mitigate the effects; at worst, these effects might degrade performance.
Engineering risks and concerns
(1) When developing the plastic-crystal composite, attention must be paid to ensuring that the
material can be made with industrial processes used to manufacture other plastics. (2) The
chemical reaction that occurs in manufacturing plastic gives off heat, potentially creating hot
spots that would impair the performance of the material. This is not a problem for very small
quantities. At the other end of the scale, for industrial production, the problem is well understood
18
Information provided by Edward McKigney, Los Alamos National Laboratory, e-mail, July 21, 2009.
19
These properties include probability of a full energy deposition interaction; efficiency of converting deposited energy
to light, the signal that is measured; transparency to photons generated in this way so the detector volume responds
uniformly; and physical robustness.
20
For example, dense packing of nanocrystals may change the material’s mechanical flow. That would make it harder
to process the material by the lowest-cost method, extrusion, but it could still be cast and machined to shape. PVT
scintillator is cast and can be machined, suggesting that that approach offers low cost while producing large sheets of
material.
Congressional Research Service
17
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
and chemical engineering solutions have been implemented for decades. A concern is whether a
solution can be devised for pilot-scale production.
Cost and schedule risks and concerns
The project is not mature enough to provide a firm estimate of the cost of the material produced
on an industrial scale. The cost estimate cited above is based on the cost of procuring the raw
materials and assumes that the energy cost for processing is low. Inflation in energy would not be
expected to increase cost sharply, but cost of the product is sensitive to the cost of cerium, a rare
earth element. The manufacturing processes are similar to those used for such consumer goods as
fabric softeners and disposable plastic water bottles, so unit cost arguably should not be high.
However, the cost estimate excludes the cost of fixed infrastructure; how much that would add to
unit cost depends on infrastructure cost and the number of units produced. Transferring the
technology to a commercial partner by the end of FY2009 depends on resolving potential
scientific and engineering problems in a timely manner and finding the right partner. The project
is not far enough along to make a firm estimate of schedule beyond FY2009.
Operational risks and concerns
In any project, it is possible to develop a product only to have it fail in everyday use. This project
is trying to minimize this risk. It is trying to design robustness into the material, such as by (1)
using a plastic that is soft and rubbery rather than brittle, (2) ensuring that the material will be
effective across the temperature range to which it will be exposed, and (3) ensuring that the
detector and packaging are compatible so that, for example, the detector material will not expand
so much as to crack its casing.
Potential gains by increased funding
The project’s total budget is about $5.5 million per year. McKigney states that he could “usefully
employ a total budget of up to $12M/year to reduce the time to deployment of these technologies”
in several ways. (1) The project is pursuing, with about one-third of its total funding, a separate
basic research project to develop a scintillator material approaching the resolution of high-purity
germanium detectors with the cost and processing characteristics of plastic. More funds could
accelerate this project, according to McKigney. He cautions that this project might take a decade
or more and has much greater scientific risk than the current nanocomposite scintillator project.
(2) The project uses equipment available at Los Alamos, but he states that staff could save time if
they had their own equipment. 21 (3) It would be difficult to fabricate a single large panel (e.g., 50
cm wide by 20 cm thick by 90 cm long) of detector material. Further, a panel segmented into tens
to thousands of tiny panels produces the optimum tradeoff between cost and effectiveness and can
provide data on the position of a radioactive source. Each panel element would need its own
electronics channel, so application-specific integrated circuits must be custom-designed, which
could take two or three years. With more funding, he asserts, the project could develop these
21
For example, the project needs an understanding of the structural features of the materials being synthesized that an
instrument called a resonant Raman spectrometer could provide. McKigney states that this instrument would speed up
development of the process to synthesize the scintillator material by providing direct information about what material
has been synthesized. Currently, the project infers such information from measurements that McKigney states are
harder to interpret.
Congressional Research Service
18
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
chips and the scintillator material concurrently, saving time. Recognizing this leverage, DTRA
provided about $150,000 for this purpose spread over 2½ years beginning in May 2008.
McKigney states that additional funding would accelerate this schedule. (4) Hiring more
chemists, electrical engineers, and others would accelerate these projects, according to
McKigney.
Potential synergisms and related applications
(1) This material offers the greatest benefit in detectors that use large panels of scintillator
material, such as some under development as discussed below. (2) By offering greater sensitivity
and greater resolution, this material could provide better data for algorithms to process,
permitting the development of simpler, more capable algorithms. (3) Nanocomposite scintillator
material may be able to measure neutron and gamma ray energies. As such, it might be possible
to use it in a detection system instead of separate means of detecting each particle type. (4) Some
companies are considering systems that use tubes filled with helium-3 gas to detect neutrons.
However, helium-3 is scarce, and there may not be enough to support large-scale use of these
tubes. Nanocomposite scintillator material might be an alternative for neutron detection.
GADRAS: A Gamma-Ray Spectrum Analysis Application Using
Multiple Algorithms22
The problem
In a carefully controlled laboratory environment, a radioisotope can be readily identified by
matching its gamma ray spectrum to one in a library of spectra. At a port or border crossing, a
spectrum would be complicated by radiation from many sources and by attenuation caused by
cargo and other materials. At issue: How can the signal from SNM be extracted from a gamma
ray spectrum, and how can this capability be improved?
Background
Nuclear detection hardware receives much attention, but data from the hardware—e.g., pulses of
various energies from gamma rays—are unintelligible until processed by software. Indeed,
software in the form of algorithms is the “brains” of a detector. An algorithm, such as a computer
program, is a finite set of logical steps for solving a problem. For nuclear detection, an algorithm
must process data into usable information fast enough to be of use to an operator.
Algorithms are designed to assure a low rate of false positives, which impede commerce, and a
near-zero rate of false negatives, which could let a terrorist weapon into the United States. For
gamma-ray identification, an algorithm creates equations to model “radiation transport,” the
movement of radiation through material. A detector will record gamma rays from all sources—
background radiation, items in ordinary commerce, radioisotopes that terrorists might include in a
shipment to confuse analysis, and a nuclear weapon or SNM. Many uncertainties affect an
22
The principal developer of this algorithm, Dean Mitchell, Distinguished Member of the Technical Staff, Contraband
Detection Organization, Sandia National Laboratories, provided detailed information for this section, personal
communications, April-August 2008. Others have commented as well to provide alternative perspectives.
Congressional Research Service
19
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
algorithm. Gamma rays lose energy as they interact with matter, altering their spectra depending
on the type and thickness of the materials they pass through, and the initial energy of a gamma
ray. Random fluctuations in the number of counts in each gamma-ray energy bin create statistical
uncertainties, especially if the number of counts is low. The composition and thickness of
materials between the radiation source and the detector (cargo, shielding, container wall, air, etc.)
is not known. The equations may not exactly represent the detector dimensions or the shielding
configuration because approximations are made to reduce complex environments to a set of
equations that can be more readily computed.
Technology description
Dean Mitchell of Sandia National Laboratories has developed an applications package, “Gamma
Detector Response and Analysis Software,” or GADRAS.23 Development of GADRAS started in
1985 for use in the Remote Atmospheric Monitoring Project (RAMP), which used low-resolution
detectors to analyze airborne radionuclides. 24 Beginning in 1996, automated isotope identification
was developed within GADRAS to process spectroscopic data collected at border crossings to
support interdiction of radioactive materials. 25 Work on the current version of GADRAS started in
2001. In earlier versions, it was seen as acceptable to spend a month analyzing an individual
spectrum; in the wake of 9/11, in order to be of value for screening commerce, GADRAS had to
be modified to process data quickly while minimizing false positives and false negatives. As of
April 2010, the GADRAS application included six radiation analysis algorithms, gamma ray and
neutron detector response functions, and support for radiation transport calculations.26
To identify radioactive sources creating a gamma-ray spectrum, GADRAS matches an entire
gamma ray spectrum against one or more known spectra. Figure 6 illustrates this approach. Many
other algorithms focus on peaks in gamma ray spectra because they are the most obvious features.
In contrast, GADRAS analyzes the full spectrum for several reasons. (1) Peaks may overlap,
making source identification ambiguous. (2) Most counts in a gamma-ray spectrum are often
outside the peaks, in which case using only peak data would ignore most of the data. For
example, less than 3% of the counts in a spectrum for U-238 occur in the 1.001-MeV peak, the
most prominent feature of its spectrum. (3) Counts outside the peaks help characterize the
composition and thickness of intervening material. Since gamma rays interact with these
materials, characterizing the materials improves the accuracy with which the gamma-ray
spectrum as read by a detector can be linked back to the gamma-ray source. Arriving at a solution
consistent with all the data increases confidence in the result. (4) The absence of counts in a
region of a spectrum can be a clue to the identity of radioactive materials. (5) Using the entire
spectrum helps analyze data from scintillators having low energy resolution because low
resolution often precludes identification of peaks in the spectrum, and helps analyze spectra of
weak sources.
23
For a technical discussion of GADRAS, see Dean J. Mitchell, “Variance Estimation for Analysis of Radiation
Measurements,” Sandia Report SAND2008-2302, April 2008.
24
Dean J. Mitchell, “Analysis of Chernobyl Fallout Measured with a RAMP Detector,” SAND87-0743-UC-32, Sandia
National Laboratories, 1987.
25
Dean J. Mitchell, “Analysis of Low-Resolution Gamma-Ray Spectra by Using the Unscattered Flux Estimate to
Search an Isotope Database,” Systems Research Report, Sandia National Laboratories, 1997.
26
Information provided by Dean Mitchell, e-mail, July 7, 2009, and April 1, 2010.
Congressional Research Service
20
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Figure 6. How GADRAS Identifies Special Nuclear Material
Source: Figure by Dean Mitchell, Sandia National Laboratories, April 2010.
Notes: This figure shows how GADRAS can identify various materials (in this case uranium) despite background
radiation. The bottom scale shows gamma-ray energy in thousands of electron volts (keV); the left scale shows
number of gamma-ray counts at each keV level. Black bars at the upper edge of the blue area are the raw data
that a gamma-ray detector provides. (Figures in the online version of this report are in color.) Data are
presented as bars rather than points to indicate one standard deviation uncertainties. The raw data cannot
differentiate between gamma rays produced by different materials, as a 200-keV gamma ray from one substance
is identical to a 200-keV gamma ray from another substance, and background radiation may hide gamma-ray
peaks from a material of interest. A count of gamma rays from a cargo container might produce a data set like
this.
GADRAS analyzes the entire spectrum to determine which combination of materials might produce it. In this
example, GADRAS finds uranium-238 (red area, middle band) and uranium-235 (green area at bottom left) as
well as background radiation (blue area, top band). The key point is that analysis of the entire spectrum provides
more accurate identification of the materials under investigation than would analyzing only gamma-ray peaks
because much the information in the spectrum is outside the peaks. For example, uranium-238 has a peak at
1,001 keV but in this figure only a very small part of the uranium-238 gamma-ray spectrum (less than 3 percent),
and an even smaller fraction of the total, is at that energy level.
GADRAS also uses neutron flux data. Since neutrons pass more readily through high-Z material
and gamma rays pass more readily through low-Z material, different materials, such as in a
container, will affect the total radiation output differently. As a result, using both gamma ray and
neutron data improves the analysis.
GADRAS has been in use since 1986. Since 9/11, more operators have used it in a wider range of
applications. In response, the software is continually upgraded to support new types of radiation
sensors, provide new capabilities, and meet new performance requirements. According to
Mitchell, “One of the new features is the ability to support the analysis of data that is collected
with neutron multiplicity counters. This capability enables inclusion all of the data collected by
Congressional Research Service
21
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
gamma-ray detectors and various types of neutron detectors into a unified analysis algorithm.”27
One goal is to make GADRAS more automated so that less skill is required to operate it. Another
goal is to make it faster. A current effort focuses on improving sensitivity to SNM and reducing
the false alarm rate. In the past, some sacrifices were made to fidelity of the analysis in order to
gain speed; now, with faster computers, it should be possible to improve the analysis while
increasing speed; this approach is being investigated. Another approach to reducing false alarms
is to examine radiation sensor data collected in 2005 on cars entering the Lincoln Tunnel between
New York City and New Jersey. According to Mitchell, the data included 50 to 60 false alarms.
Mitchell and other researchers at Sandia are trying to determine what caused the false alarms in
order to modify GADRAS to correct for these problems.28
The main application of GADRAS is to support “Triage/Reachback” analysis. A radiation
detection operator in the field, such as a Customs and Border Protection (CBP) officer, who finds
a vehicle or cargo container that presents a suspicious radiation signature that cannot be easily
resolved, can send the detection data (such as a gamma ray spectrum) to the Laboratories and
Scientific Services section of CBP for a more detailed analysis.29 That analysis uses GADRAS.
Similarly, if that service is unable to resolve the matter, it can send the data to a secondary
Reachback at the nuclear weapons laboratories, which also use GADRAS.
Potential advantages
By analyzing a complete gamma ray spectrum, GADRAS increases the accuracy of determining
whether a cargo container or other item is carrying SNM, reducing the risk of false positives and
false negatives. Using neutron count data in addition to gamma ray spectra further improves
capability.
Status, schedule, and funding
GADRAS has been used for cargo inspection since 1998. Software upgrades are released every
two months or so. There is no line item for GADRAS development. Mitchell estimates that
Sandia is spending perhaps $600,000 in FY2010 to continue to develop GADRAS. DTRA stated
in 2008, “The DTRA in concert with NNSA is currently proposing development of the next
generation of GADRAS as part of a potential [memorandum of understanding] between the
organizations. The effort would emphasize the revision algorithms, update to Vista compliant
software, and increasing portability for field use.”30
27
Information provided by Dean Mitchell, e-mail, April 1, 2010. A neutron multiplicity counter detects SNM by
detecting the time pattern of neutron generation. A subcritical mass of highly enriched uranium or weapons-grade
plutonium can support a fission chain reaction producing an increasing number of neutrons. (Since it cannot support
enough fissions to create a nuclear explosion, such chain reactions die out.) These neutrons are generated in a closely
spaced pattern over a brief time. In contrast, background neutrons occur in a random time pattern.
28
Information provided by Dean Mitchell, e-mail, April 1, 2010. A neutron multiplicity counter detects SNM by
detecting the time pattern of neutron generation. A subcritical mass of highly enriched uranium or weapons-grade
plutonium can support a fission chain reaction producing an increasing number of neutrons, though it cannot support
enough fissions to create a nuclear explosion, so such chain reactions die out. These neutrons are generated in a closely
spaced pattern over a brief time. In contrast, background neutrons occur in a random time pattern.
29
For further information, see U.S. Department of Homeland Security. Customs and Border Protection. “Laboratories
and Scientific Services.” Available at http://www.cbp.gov/xp/cgov/trade/automated/labs_scientific_svcs/.
30
Personal communication, August 5, 2008.
Congressional Research Service
22
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Risks and concerns
Scientific risks and concerns
Since upgrades are ongoing, developers face such scientific problems as how to improve
equations to characterize the response of detector material to radiation. Such problems are not a
major risk to continued development of GADRAS, Mitchell states. Another risk is that the
gamma-ray signal from shielded HEU may be so low that even a high-quality algorithm cannot
identify the HEU.
Engineering risks and concerns
The major risk to GADRAS development comes from the programming language that is used for
the graphical user interface (GUI). The GUI for the current version of GADRAS is written in a
Microsoft language called Visual Basic Version 6 (VB6). VB6 functions under current versions of
Windows, including the most recent, Vista. Microsoft, however, no longer maintains VB6, so it is
not necessarily compatible with new compilers,31 often leaving no effective upgrade path for large
application programs like GADRAS. User feedback indicates that some GADRAS components
do not function properly in Vista. While minor changes may resolve the latter problem, the
current GUI may not run under a future version of Windows.32 GADRAS is a large program, and
current funding does not support the effort that would be required to update it.
Cost and schedule risks and concerns
Since upgrading GADRAS is a low-budget activity, it entails little cost risk. No near-term
requirements impose significant schedule risks. However, it would take one to two years to revise
the GUI. Making this revision in advance of a change of operating systems would enable users to
use GADRAS without interruption when they switch to the new operating system.
Operational risks and concerns
Since GADRAS has been in existence for many years, and since it is used mainly by scientists
and technicians rather than by operators in the field, there is little risk that it will fail in everyday
use. The risk that an upgrade will cause a problem is reduced by careful testing. Other risks are
the possibility that GADRAS would not run on future versions of Windows operating systems,
and that conversion would not be made before an operating system that would not run GADRAS
is introduced. Some have said that GADRAS is a complicated program to use; accordingly, the
CONOPS for GADRAS is that it is used mainly for Reachback rather than in the field.
31
A compiler is a computer program that translates an application (such as GADRAS) into instructions that a computer
can process.
32
As another example, because of incompatibilities between VB6 and current generations of Fortran, components of
GADRAS that are written in Fortran must be compiled using Compaq Fortran, another discontinued and unsupported
programming language.
Congressional Research Service
23
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Potential gains by increased funding
Mitchell states that perhaps $600,000 a year for two years would allow Sandia to hire a few
programmers who could convert GADRAS to use current compilers in order to avoid potential
disruptions associated with new operating systems. A related task is documentation of the
Application Programs Interface. This documentation would enable other users, such as at other
laboratories, to develop new applications that can access capabilities that are incorporated into the
Dynamic Link Library, which performs most of the computations in GADRAS. This increased
access would reduce the cost and development time for new applications.
Potential synergisms and related applications
(1) GADRAS might improve the performance of systems that induce fission and detect the
resulting radiation. (2) By gaining more information from the gamma ray spectrum, it might
reduce the gamma ray or neutron flux needed to induce fission and the dose resulting from
fission, thereby increasing worker safety. (3) More computation power would permit more
sophisticated iterations of GADRAS to be developed, or would permit GADRAS to run faster, or
both. (4) Increased computation power, especially smaller and more capable computers, might
enable GADRAS to be modified for use in the field with radiation detection equipment,
permitting quicker resolution of suspicious containers and vehicles. Reachback would then be
used only for the hardest-to-resolve cases. (5) GADRAS could be modified to improve the
performance of radioisotope identification devices (RIIDs). DTRA states that it “is currently
funding development of a more portable version of GADRAS ... that is intended to be resident on
certain RIID systems, such as handheld HPGe system produced by Ortec.”33 (6) Improved
scintillator materials can be expected to provide better data inputs to GADRAS, enabling a
further reduction in false negatives and false positives. Universities, companies, and government
laboratories are working to develop such materials.
Computer Modeling to Evaluate Detection Capability34
The problem
Developing equipment to detect terrorist nuclear weapons and SNM requires many choices. It
would be of great value to evaluate how they affect cost and performance before committing to a
system. However, many combinations and tradeoffs are possible, and it would be prohibitively
expensive and time-consuming to run enough trials for each to make a valid comparison among
them. When faced with similar choices, such as in designing a car, corporations typically run
huge numbers of simulated trials using computer models that take significant investment to
develop and maintain. At issue: How can computer modeling contribute to the development of
nuclear detectors, and what are its limitations?
33
Personal communication, August 5, 2008.
34
Richard Wheeler, Lead for Homeland Security Analysis, Global Security Directorate, Lawrence Livermore National
Laboratory, provided detailed information for this section, personal communications, April-August 2008. Others have
commented as well to provide alternative perspectives.
Congressional Research Service
24
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Background
A detector should maximize the probability of detecting an actual threat (a true positive) while
minimizing the probability of detecting a nonexistent threat (a false alarm, or false positive). For
a given technology, these objectives cannot be achieved simultaneously—an improvement in
either one comes only at the expense of the other. A receiver operator characteristic (ROC) curve,
such as Figure 7, illustrates this relationship.35 By relating the true positive and false positive
rates, the curve defines the performance of a receiver (in signal processing, where the term
“ROC” originated) or of nuclear detection equipment. ROC curves show that the probability of a
true positive and a false positive go up together. This is logical; one could eliminate false alarms
by turning off the detector, or could be sure of detecting every threat by having the detector alarm
whenever the slightest trace of radiation is detected, which, given omnipresent background
radiation, would be all the time. The peril of failing to detect an actual threat is clear. At the same
time, law enforcement and commercial interests are intolerant of false alarms because these
alarms require a major effort, diverting officers to the scene and possibly unloading a cargo
container or closing a border crossing. Further, in the real world, numerous false alarms may
cause operators to ignore all alarms or set the detector to be less sensitive, reducing the false
alarm rate but also increasing the likelihood of missing an actual threat. This tradeoff is shown in
Figure 7 by moving from point C, with a high probability of detection but a high false alarm rate,
to point B, with intermediate values for both, to point A, with a low false alarm rate but a low
probability of detection.
Figure 7. A Notional Receiver Operating Characteristic (ROC) Curve
Source: Prepared by CRS with the assistance of Lawrence Livermore National Laboratory.
35
For a helpful interactive tutorial on ROC curves, see Anaesthetist.com, “The Magnificent ROC,” at
http://www.anaesthetist.com/mnm/stats/roc/Findex.htm.
Congressional Research Service
25
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Figure 8 shows three ROC curves to illustrate several concepts. (1) Moving from curve A to
curve B to curve C shows the performance of a hypothetical detector improving over time,
perhaps as a different detector material is used or an algorithm is modified. The improvement can
be visualized by moving upward (line 1), which shows an increase in the true positive rate for a
given false positive rate, or by moving from right to left (line 2), which shows a reduction in the
false positive rate for a given true positive rate. (A diagonal line from lower right to upper left
would show improvement in both.) (2) A, B, and C could represent differences in performance of
one detector under different conditions, such as changes in the background, different operating
conditions (e.g., scan time), or different benign materials in a container. (3) The curves could
characterize the performance of three competing detectors. Note that actual ROC curves have
more complex shapes than the notional curves shown. They may even cross over each other,
indicating that one option is not uniformly better than another, requiring consideration of further
tradeoffs.
Figure 8.Three Notional ROC Curves
Source: Prepared by CRS with the assistance of Lawrence Livermore National Laboratory.
Many variables affect detector performance. Some are operational, such as scan time, the
detector’s target (containers, cars, trucks, trains), the distance between detector and target, and
environmental conditions (background radiation, season, temperature). The detector is to detect
SNM or nuclear weapons, yet the signatures on which it will focus may be accompanied by
radiation from innocent sources and background radiation, and may be shielded inadvertently or
deliberately. There are choices for the active elements of a detector, the algorithm used, and
specific subroutines. These choices affect detector performance. Many combinations of these
variables are possible. To gain enough data to make a ROC curve statistically valid, many trials
would need to be performed for each combination of controllable variables (operating setup,
detector, and algorithm) against many targets, each generating its own radiation signature. Each
event in which a vehicle or container passes through the detection system is called an
Congressional Research Service
26
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
“encounter.” It would be prohibitively expensive and time-consuming to run many encounters for
each of thousands of combination of controllable variables, but it would be of great value to have
the resulting data in order to compare, improve, or choose between detection systems and their
components.
Computer modeling can help. Modeling creates mathematical representations of the real world,
varies the inputs, and calculates the outputs. In the case of nuclear detection, the real world
consists of controllable variables (operational scenario, detector, and algorithm) and uncontrolled
variables (signals from radioactive material). The performance of a modeled detector can be
illustrated using ROC curves as described above. Running the model to simulate many encounters
for each combination of controllable variables provides enough observation points to generate a
statistically significant ROC curve. This process is repeated for many combinations of variables.
The resulting data show the user how change to variables affects performance, and permit
comparing one detector against another. Computer-generated “data” for nuclear detection
encounters are always imperfect, as discussed below. As a result, the adequacy with which the
models represent reality is always at issue, and model developers devote great effort to improving
that representation.
Modeling can also be used to evaluate requirements for elements of a detection system. For
example, a model was used to study the spectral resolution (see “Principles of detection,” above)
required of a detector material to distinguish the gamma-ray spectra of threat sources from nonthreat sources. According to a report on this project, “To capture the range of gamma-ray sources
and shielding configurations found in commerce, we generated simulated populations of 1000 or
more spectra, each with 3000 energy channels.”36 Clearly, it would have been costly and timeconsuming to generate the data experimentally.
Technology description
DNDO has established an ongoing program, Detection Modeling and Operational Analysis
(DMOA), that the national laboratories and private sector contractors carry out. It builds models
that characterize the variables noted above, i.e., the operational scenario, the radiation signals, the
detector, and the signal-processing algorithm. Creating mathematical representations of the first
three takes an immense amount of work because each is so complex. Alternative algorithms are
simulated as part of the overall detection performance modeling. (Algorithm development
requires a great deal of work, but is not within the scope of DMOA, which focuses on
simulations.) For example, modeling a gamma-ray spectrum requires taking into account various
sources of radiation and types of shielding. A DMOA study of real-world data found that the
spectra from cargo differed between spring-summer and fall-winter; the study speculated that a
different mix of products shipped in the two periods caused the difference. 37
36
Karl Nelson, Thomas Gosnell, and David Knapp, The Effect of Gamma-Ray Detector Energy Resolution on the
Ability to Identify Radioactive Sources, Lawrence Livermore National Laboratory, Radiological and Nuclear
Countermeasures Program, LLNL-TR-411374, February 2009, p. i, https://e-reports-ext.llnl.gov/pdf/370769.pdf.
37
Lawrence Livermore National Laboratory, Radiological and Nuclear Countermeasures Program, “Radiation
Detection Modeling and Operational Analysis: Benchmarks, Nuisance Source, Algorithm and Resolution Studies,”
LLNL-TR-401531, December 31, 2007, by Padmini Sokkappa et al., pp. 21-22 This source is a report marked by
Lawrence Livermore National Laboratory, the originating organization, “For Official Use Only.” That laboratory has
approved CRS use of the information cited by this footnote.
Congressional Research Service
27
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
The model processes data on the operational scenario, radiation signals, and the detector to create
a gamma-ray spectrum that is sent to the algorithm. The algorithm does not “know” the difference
between a spectrum generated in the real world or by computer, and processes both in the same
way. Since the difference is the source of the data, the key to simulation is generating the gammaray spectrum (signal plus noise) that goes to the algorithm. The greater the fidelity with which the
model mimics real-world inputs, the better it represents system performance. Modeled
performance, as measured by detection probability and false alarm rate, can be summarized in a
ROC curve in the same way as empirically measured performance. The modeling effort also
includes assessing the sensitivity of detector performance to changes in operational scenarios,
detector hardware, and algorithms.
One DMOA effort in 2006 focused on four main areas to improve its models.38 This effort
illustrates the work undertaken to improve the fidelity of models and how this work requires
detailed knowledge of the components being modeled.
•
Evaluation benchmarks. Providing a basis for comparing systems requires a
reference set of threat objects and shielding. Previously, this set included
plutonium, HEU, and other threats, as well as different levels of shielding. In
2006, DMOA added new threats and shieldings to the reference set. It also
developed a reference set of objects for detection by radiography.
•
Background and nuisance source modeling. According to DMOA, “Background
radiation and nuisance source population characteristics generally dictate
detection threshold settings through their impact on innocent alarm rates.
Characterization of these factors is critical to evaluating the performance of
radiation detection systems.” DMOA used real-world data to develop a model of
the distribution of naturally occurring radioactive material. Data from the model
were then compared with another set of real-world data to check the validity of
the model.
•
Operational analysis. DMOA used real-world data to compare the performance of
several algorithms. One result was to “highlight the sensitivity of system
performance to the detection algorithm used.”
•
Detector resolution study. DMOA studied the resolution needed to distinguish
threats from other sources of radiation. The analysis, though preliminary, found
that there is a benefit by having better resolution than that offered by sodium
iodide, but that improving resolution beyond a certain point would offer marginal
benefit.
Potential advantages
Modeling offers advantages compared to obtaining data from the real world . (1) A model permits
construction of statistically significant ROC curves. It is not uncommon to require a false alarm
rate of one in ten thousand or lower; statistical validation of performance at this level would
require hundreds of thousands or millions of experiments. It would be costly and time-consuming
to conduct field trials using a single combination of variables in order to build one statistically
significant ROC curve, let alone conducting field trials for thousands of combinations of variables
38
This paragraph is based on ibid., pp. i-ii.
Congressional Research Service
28
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
that represent the range of conditions. A model permits running many thousands of simulated
encounters in less than a day to explore the range of encounters a detector may face. Once the
model has been constructed, validated, and implemented, the cost of these computer runs is very
low. (2) A model and a test program are complementary. The test program generates data for the
model, and the model can steer the test program by indicating what tests would be of greatest
value for improving the predictive power of the model. (3) A model permits exploration of
encounters that might occur in the real world but that could not be conducted because of cost,
safety, or difficulty. For example, it would be unthinkable to place a nuclear weapon in
commercial traffic to test detectors, but the only facilities where such testing could be done,
notably the Nevada Test Site, have characteristics very different than those of ports or border
crossings. As another example, it might be desirable to see whether a protracted period of heat or
high humidity would impair how a detector would function, but it would be much easier to
replicate such conditions through modeling. (4) A model permits comparison of components of an
encounter, such as which algorithm better processes the data for a given gamma-ray spectrum. (5)
A model permits analysis of alternatives and clarification of tradeoffs before committing to a
specific design, helping to inform billion-dollar decisions.
Status, schedule, and funding
DMOA is an ongoing program. When DNDO was established, DMOA became an explicit
element within the System Architecture program. It has been funded within that program at
approximately $2 million per year. The amount funded by all DNDO offices on related detection
modeling is on the order of five times that of the Architecture program. Other detection modeling
is done for other purposes by other offices within DNDO, as well as by DOE and DOD. However,
since modeling activities are inherently cross-cutting and support many technology development
and assessment projects, it is difficult to estimate total spending on modeling in the federal
budget.
Risks and concerns
Scientific risks and concerns
While no model can be perfect, the key risk is that the model’s output (for example, gamma ray
spectra that the model generates) might differ significantly from data that would have been
obtained from actual field trials. Part of this risk concerns modeling the underlying science. Some
aspects are known in detail, such as the radiation spectrum of U-235. But there are uncertainties.
HEU is not pure U-235, so the spectrum will be somewhat different from that of U-235. Some
shielding can be modeled precisely, such as a centimeter of lead. But the defense (the modeling
design team) does not know what shielding, if any, the attacker might use. A cargo container can
hold many types of cargo, each of which interferes with radiation transport differently. Different
arrangements of the cargo place different amounts and types of material between source and
detector, affecting gamma ray spectra. Any model makes approximations and simplifications,
sometimes to allow the simulations to run faster, or simply because more fidelity for certain
phenomena are judged unnecessary or unwarranted. This creates further risk that the model might
not sufficiently represent reality. There are other risks. Items included in the model may not be
selected accurately. Systematic errors may arise, such as differences between spring-summer and
fall-winter cargo. Adding detailed radiation sources with additional shielding, types of cargo, and
detector details may increase the realism of the model, but also add complexity, opening the door
to additional errors.
Congressional Research Service
29
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Engineering risks and concerns
Models are complex mathematical approximations of reality. Yet real-world data are often
limited, perhaps covering too narrow a range of conditions. DMOA uses statistical methods to
transform a real-world data set into data for different detectors and conditions. As a DMOA report
states, “LLNL [Lawrence Livermore National Laboratory] has developed a procedure for
generating a statistical model of a nuisance source population ... based on measured data. ... The
statistical model developed provides a basis for simulating an unlimited number of random
samples in nuisance sources for a population assumed to be similar to that underlying the
measured data.”39 A difficulty is in validating the model. Various isotopes cause background
radiation, and DMOA observes that “there is no simple standard procedure to compare
multivariate populations.”40
Models may also manipulate data by adding in, or “injecting,” spectra from well-characterized
radiation sources like HEU or WGPu to computed spectra of cargo typical of normal commerce
to see how well an algorithm can detect threat material in cargo. However, that approach may be
an inadequate representation of reality. It is arguably unlikely that terrorists would include a
nuclear weapon or SNM in a random cargo container if they chose that vector; rather, they might
arrange the cargo to help evade detection. Simulation developers have not modeled such
deliberate arrangements of cargo because that would require thinking about how to model
adversary behavior, something outside their expertise. It might be desirable to have terrorism
experts modify some injections to reflect adversary behavior.
Cost risks and concerns
Most funds spent on DMOA are for staff salaries, so the likelihood of unanticipated cost increases
appears low. However, this assumes that computing hardware required to run the models quickly
is already paid for.
Schedule risks and concerns
If terrorists planned to bring a nuclear weapon or SNM into the United States, they would
presumably try to evade detection, leading to an offense-defense competition. It would be
essential for this nation to stay ahead. That task involves much effort by many entities. For
modeling, the task would be to upgrade means of evasion included in the model in time to stay
ahead of the threat. This would depend in part on inputs from intelligence services, such as data
on specific threats, but also on advances in characterizing background radiation, developing
algorithms, developing statistical means of transforming or validating data, and the like. The
program is paced by resources available. The risk is that progress in nuclear detection, including
modeling, will not be fast enough to defeat the threat.
39
Radiological and Nuclear Countermeasures Program, “Radiation Detection Modeling and Operational Analysis,” p.
17. This source is a report marked by Lawrence Livermore National Laboratory, the originating organization, “For
Official Use Only.” That laboratory has approved CRS use of the information cited by this footnote.
40
Ibid., pp. 20-21.
Congressional Research Service
30
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Operational risks and concerns
Law enforcement and commercial interests are intolerant of false alarms. The false alarm rate
often drives detector settings and choice of detection systems, and increasing the true positive rate
also increases the false positive (false alarm) rate. While a model can explore the tradeoff and
optimum balance between true and false positive rates, a risk is that systematic errors, incorrect
data transformations, and incomplete accounting for background radiation would cause the model
to generate a setting that is less than optimal. In some cases, this problem can be overcome by
making adjustments in the field, but in other cases, such as the choice of detector material,
adjustments may not be possible.
Potential gains by increased funding
At present, the DMOA efforts at the national laboratories and private sector contractors are
distributed throughout their organizations. Most staff members who work on detection modeling
do so only part time. Some modeling tools (simulation codes, databases, and algorithms) tend to
be somewhat informal, developed by various groups for specific needs. According to Richard
Wheeler, Lead for Homeland Security Analysis, Global Security Directorate, Lawrence
Livermore National Laboratory, added funding in this area could enable the designation of fulltime staff, development of standardized modeling tools, acquisition of dedicated computing
resources, establishment of rigorous peer review of the models, formal coordination of related
efforts sponsored by DOE, DOD, and other agencies, and generation of new databases essential
for model validation. While analysis of detector performance against realistic threats is sensitive
or classified, many advances in modeling and modeling tools could be made in an open
environment, including the academic community. Wheeler states that more funds could support a
more substantial engagement of university researchers.
Potential synergisms and related applications
Simulation does not detect anything by itself, but is by its nature synergistic with other aspects of
nuclear detection. It has as its purposes improving many elements of detection systems, whether
deployed or under development; helping to integrate hardware and software into a system;
optimizing systems and CONOPS; and informing decisions on the most cost-effective mix of
systems to acquire.
L-3 CAARS: A Low-Risk Dual-Energy Radiography System41
The problem
Current radiography systems have important limitations in their ability to detect SNM in cargo
containers. A small amount of dense material may be inconspicuous if shipped in a container
filled with dense or random objects. SNM could also be placed inside dense objects for
camouflage. Another difficulty of detecting SNM in containers is that an operator may fail to see
41
Joel Rynes, Program Manager, CAARS Program, Domestic Nuclear Detection Office, Department of Homeland
Security, provided detailed information for this section, personal communications, April-August 2008. Others,
including L-3 Communications Corporation, have commented as well to provide alternative perspectives.
Congressional Research Service
31
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
a threat. At issue: Can radiography take advantage of different characteristics of SNM to detect it
in cargo? And can a system evaluate radiographic images automatically to reduce dependence on
operator judgment?
In an effort to overcome these limitations, DNDO is conducting R&D on advanced radiography
systems under the Cargo Advanced Automated Radiography System (CAARS) program. CAARS
involves three technical approaches to radiography, each with a different contractor. (As noted
below, DNDO terminated the contract for the American Science and Engineering, Inc., system in
March 2009.) This section and the next two discuss these systems. It must be emphasized at the
outset that these systems are under development. As a result, no deployable CAARS systems
exist, so diagrams and performance specifications presented in these sections must be viewed as
goals that are yet to be demonstrated in commercially available equipment. Customs and Border
Protection (CBP) raises several concerns about CAARS, as described under “operational risks
and concerns,” below. CBP, a component of the Department of Homeland Security, is a front-line
agency whose officers perform such missions as operating border crossings and inspecting cargo
entering the United States at seaports, airports, and land crossings.
Background
Two types of commercially available equipment for scanning cargo containers have been widely
used since 2002.42 One type, radiation portal monitors, passively detects radiation coming from a
container. Another type, radiography equipment, creates x-ray-type images of a container;
examples include the Rapiscan Classic Eagle and the SAIC VACIS.43 This type is more relevant
to CAARS. Radiography equipment works as follows. A cargo container is driven between two
components of the equipment, or the equipment moves over a stationary container. One
component produces gamma rays from a highly radioactive substance like cobalt-60 or cesium137, or x-rays from a linear accelerator. These photons emerge in a thin vertical fan-shaped beam
and pass through the side of the container.44 On the other side of the container, an array of
detector material records photon intensity levels, which correlate to opacity, pixel by pixel. A
computer assembles the pixels into a radiograph. A black or white spot indicates an area that is
opaque to photons of the energy used. Key limitations of such systems are that they cannot
42
“Prior to 9/11, not a single radiation portal monitor [RPM] and only 64 large-scale non-intrusive inspection [i.e.,
radiography] systems were deployed to our nation’s borders. By October of 2002, CBP had deployed the first RPM at
the Ambassador Bridge in Detroit.” Testimony of Thomas Winkowski, Assistant Commissioner, Office of Field
Operations, U.S. Customs and Border Protection, before the Senate Homeland Security and Governmental Affairs
Committee, September 25, 2008. CBP has 1,145 RPMs and 209 large-scale radiography systems deployed as of
October 23, 2008. Personal communication, Patrick Simmons, Director, Non-Intrusive Inspection, Customs and Border
Protection, October 23, 2008. Radiography has been used for industrial applications (e.g., inspecting metal parts for
cracks and voids) for decades.
43
For information on these systems, see Rapiscan Systems, “Rapiscan Eagle P6000,” at http://www.rapiscan.com/
eagle-P6000.html, and Science Applications International Corporation (SAIC), “Safety & Security: VACIS Cargo,
Vehicle, and Contraband Inspection Systems,” at http://www.saic.com/products/security/.
44
Jonathan Katz et al. argue that it would be more effective to have the beam interrogate a cargo container vertically
rather than horizontally. “In innocent cargo long slender dense objects are packed with their longest axes horizontal,
and dense cargoes are spread on the floor of the container. Therefore, near-vertical irradiation will only rarely show
regions of intense absorption [of photons] in innocent cargo. In contrast, horizontal irradiation would often find this
‘false positive’ result, requiring manual unloading and inspection. Another advantage of downward near-vertical
illumination is that the Earth is an effective beam-stop, combined with a thin lead ground plane, its albedo [the fraction
of the beam that is reflected] is negligible and additional shielding would not be required.” J.I. Katz et al., “XRadiography of Cargo Containers,” Science and Global Security, Vol, 5, No. 1, January 2007, pp. 49-56.
Congressional Research Service
32
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
differentiate between different types of material and cannot pick out threats from clutter. While a
complete nuclear weapon might or might not be noticed, detection probability decreases as the
threat object becomes smaller, so it would be difficult for current radiography systems to detect a
small piece of HEU.45
A different physical principle enables a radiography system to search for materials with high
atomic number (Z). Both the L-3 and SAIC CAARS systems utilize this principle. They use one
or two linear accelerators to generate 6-MeV electrons and x-rays with energies from 0 to 6 MeV
and, in the same manner, x-rays with energies from 0 to 9 MeV (abbreviated in this section as 6MeV x-rays and 9-MeV x-rays). The former have the greatest number of photons at around 2
MeV; the latter, at about 3 MeV. (The third CAARS system utilizes a different principle.)
Photons of these two energies interact with matter differently. Six-MeV x-rays scatter when they
strike electrons.46 The amount of scattering is a function of the electron density of the material, so
it increases with Z. As a result, high-Z material is more opaque to 6-MeV x-rays than is low-Z
material, and creates a bright or dark spot on a radiograph. Figure 9, top panel, is a radiograph
taken with 6-MeV x-rays. Nine-MeV x-rays interact more strongly with an atom’s nucleus. When
a nucleus absorbs a photon, it releases energy in the form of an electron-positron pair. This effect
is proportional to Z squared. Thus high-Z material is much more opaque to 9-MeV x-rays (many
fewer get through the material being interrogated because they are absorbed) than to 6-MeV xrays.
45
Katz states, “The real issue in radiographing thick targets is not the source strength, but scattering in the target. If you
don’t have a narrow Bucky collimator matched to the geometry, the signal (increase of attenuation of unscattered
photons showing the high-Z object ...) is swamped by forward-scattered photons from the mass of solid material. The
small feature disappears from the image (rather like exposing undeveloped camera film to the light).” Personal
communication, August 8, 2008. For cargo screening using radiography, a “Bucky collimator” is a piece of high-Z
metal (such as tungsten) placed just in front of a photon detector element. The metal has a small hole precisely aligned
with the direction of the interrogation beam in order to eliminate most photons that have been scattered by the cargo
and that could obliterate small features of the image, such as a piece of SNM. The resulting increase in signal-to-noise
ratio greatly improves the detector’s ability to pick out suspicious cargo.
46
The scattering mechanisms are much more complicated than can be described here. For details, see Glenn Knoll,
Radiation Detection and Measurement, third edition (New York, John Wiley & Sons, 2000), pp. 48-53, 308-312.
Congressional Research Service
33
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Figure 9. Dual-Energy Radiography
Source: These images were taken in January 2006 with an SAIC proof-of-concept laboratory prototype cargo
imaging system called VACIS-Z: Detection of High-Z Material in Cargo. The VACIS-Z contract was funded by
the Homeland Security Advanced Research Projects Agency of the Department of Homeland Security through a
contract that predates DNDO (which was established in April 2005). The contract number is N41756-04-C4200, awarded under Technical Support Working Group (TSWG) Broad Agency Announcement DAAD-03-T0024. For further information on TSWG, see http://www.tswg.gov/.
Notes: This graphic shows three radiographs of the same objects. (Abbreviations: DU, depleted uranium, i.e.,
natural uranium with most uranium-235 removed, used here as a surrogate for highly enriched uranium; Pb, lead;
Al, aluminum; W, tungsten; HE, chemical high explosive) The top image was taken using bremsstrahlung photons
with a maximum energy of 6 MeV. The middle image shows, pixel by pixel, the ratio of a 9-MeV and a 6-MeV
radiograph. Higher-Z objects are more opaque to photons with energies up to 9 MeV than to those with
energies up to 6 MeV, and pixels of such objects are represented in darker colors. The bottom image shows only
the high-Z material from the middle image, which would automatically generate an alarm. In this image, the
system would alarm on depleted uranium, lead, tungsten, and lead surrounded by a simulated chemical high
explosive.
The L-3 and SAIC CAARS exploit this difference to create a so-called dual-energy radiograph, in
which each pixel represents the ratio of opacity of that pixel to 9-MeV and 6-MeV x-rays. To
create the radiograph, an algorithm assigns different colors to different ratios; in the middle panel
of Figure 9, higher-ratio pixels are darker. Pixels with ratios above a certain value indicate high-Z
material. While there is no absolute physical threshold between medium- and high-Z material,
CAARS uses Z>72 as the threshold for high-Z material;47 elements with Z>72 include tungsten,
gold, lead, uranium, and plutonium. 48 Pixels of such material could be presented in a separate
radiograph, as in the bottom panel of Figure 9. However, these data by themselves are not
47
The threshold value of Z>72 is used because elements between Z=57 (lanthanum) and Z=72 (hafnium), inclusive, are
very rare in commerce, making 72 a reasonable boundary between high Z and lower Z elements for the purposes of
CAARS.
48
For a periodic table of the elements, see “WebElements” at http://www.webelements.com/.
Congressional Research Service
34
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
sufficient to trigger an alarm. A typical cargo container has many overlapping objects, so another
algorithm is needed to separate them from each other. Still another algorithm utilizes the
foregoing data to calculate the size and Z of individual objects, and on that basis determines
whether to trigger an alarm.49
High-energy x-rays could also exploit a characteristic of U-235 and Pu-239 (as well as of U-238
and thorium-232, non-threat materials that are relatively uncommon in commerce): they fission
when struck by photons with an energy above approximately 5.6 MeV. As discussed in the
Appendix, the resulting fission products decay over many seconds, producing prompt and
delayed neutrons and gamma rays. High-energy x-rays may thus be used to detect high-Z material
in general and SNM in particular. However, CAARS does not utilize this characteristic.
Technology description
This section focuses on the least complex and lowest-risk CAARS system, which is being
developed by L-3 Communications Corporation. As Figure 10 shows, it would use a concrete
enclosure to minimize the radiation exclusion zone. The enclosure is 160 ft long in order to
process two trucks at a time in 3 minutes so as to meet the CAARS throughput requirement of 40
containers per hour. The system would use one linear accelerator to generate 6-MeV electrons and
(through the bremsstrahlung process) x-rays with energies from 0 to 6 MeV and, in the same
manner, another accelerator to generate x-rays with energies from 0 to 9 MeV. A container would
remain stationary as the beam is moved on a gantry over the truck. On the other side of the
container are two detector arrays, one particularly sensitive to 6-MeV x-rays and the other to 9MeV x-rays. Each array would record successive vertical slices of an opacity map of the
container. The slices would be combined into a dual-energy radiograph as described above.
49
For a technical discussion of dual-energy radiography, see S. Ogorodnikov and V. Petrunin, “Processing of
Interlaced Images in 4B10 MeV Dual Energy Customs System for Material Recognition,” Physical Review Special
Topics—Accelerators and Beams, Volume 5, 104701 (2002).
Congressional Research Service
35
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Figure 10. L-3 CAARS Schematic Drawings
Source: Domestic Nuclear Detection Office.
Notes: The detection system is mounted on rails and scans by moving over the trucks. A concrete enclosure
minimizes the radiation exclusion zone and thus the footprint. (middle) The unit has two accelerators (right) and
two detector arrays (left). (bottom) This is a simulation of the image the system would return when it detects
high-Z material in a cargo container.
Potential benefits
If CAARS works as DNDO anticipates, it could automatically detect high-Z materials while
providing standard radiographs so that it would have potentially little or no impact on CBP
operations, such as efforts to identify “traditional” contraband (drugs, guns, explosives). As such,
CAARS could, if successful, integrate the SNM detection mission with CBP’s historical mission
of detecting other illegal materials. DNDO anticipates that the automated detection of high-Z
materials would not slow down the pace of screening, which could continue to be determined by
Congressional Research Service
36
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
the rate at which operators can examine radiographic images for detection of normal contraband,
though CBP expresses concerns on this point, as discussed below. CAARS technology is being
designed to scan at least 40 40-foot containers an hour.50 DNDO anticipates that it would have a
90% probability of detecting 100 cc of high-Z material (such as a cube 4.6 cm, or 1.8 inches, on a
side), and a false alarm probability less than 3 percent, both with 95% confidence. 51 DNDO does
not count detection of high-Z material as a false alarm because such material may be used to
shield SNM. Another possible benefit of the CAARS program is that it is developing novel
technologies and advancing the state of the art; even if no CAARS systems were to be deployed,
scientific and technical advances made through this program could be of value to future detection
systems. Comparing the L-3 and SAIC systems, the latter takes up less space, an important
consideration for CBP at seaports where space is limited, but the latter has finer resolution, an
important consideration for CBP in searching for traditional contraband.52
Status, schedule, and funding
In September 2008, Vayl Oxford, Director of DNDO, described developments with the CAARS
program as follows:
Consistent with any rigorous development and acquisition program, DNDO conducted
system requirement reviews in November 2006 and preliminary design reviews in late May
and June 2007 to assess the maturity of the CAARS technology. As a result, DNDO found
that the technology was more difficult to implement than originally anticipated and
determined that the technology should be demonstrated so that its full performance capability
could be established prior to acquisition. It was also determined that the CAARS units, as
currently designed, are too large and complex to be operationally effective. Finally, since
2006, there have been several technical advances in currently-deployed or soon-to-bedeployed NII systems that might provide some, but not all, of the desired capability.
Accordingly, DNDO undertook a “course correction” in April 2008 and modified the three
CAARS contracts to remove the “acquisition” component of the contracts, yet retain the
demonstration and the test and evaluation (T&E) components of the contracts to allow
collection of the required performance data.53
According to Joel Rynes, Program Manager, CAARS Program, DNDO, the L-3 CAARS project
completed its Critical Design Review (CDR) milestone in July 2008.54 The CDR marks the point
at which the design has been completed and DNDO can give the contractor approval to begin
fabrication of the prototype. As of February 2009, the prototype was being assembled in Las
Vegas. It produced its first image in May 2009. The plan called for L-3 to collect data with the
50
DNDO dropped its original goal of scanning 120 containers per hour. Testimony of Vayl Oxford, Director, Domestic
Nuclear Detection Office, before the Senate Homeland Security and Governmental Affairs Committee, September 25,
2008.
51
U.S. Department of Homeland Security. Domestic Nuclear Detection Office. “Cargo Advanced Automated
Radiography (CAARS) Program Update Brief,” presented to the Senate Committee on Homeland Security and
Government [sic] Affairs, February 25, 2008, by William Hagan, Assistant Director, DNDO, Transformational and
Applied Research, slide 4. This source is a briefing marked by the Domestic Nuclear Detection Office, the originating
organization, “For Official Use Only.” That office has approved CRS use of the information cited by this footnote.
52
Personal communication, Joel Rynes, Program Manager, CAARS Program, DNDO, July 8, 2009.
53
“Opening Statement of Mr. Vayl S. Oxford, Director, Domestic Nuclear Detection Office, Department of Homeland
Security, before the Senate Homeland Security and Governmental Affairs Committee,” September 25, 2008, p. 4.
54
Information in this paragraph was provided by Joel Rynes, Program Manager, CAARS Program, DNDO, February 9,
2009.
Congressional Research Service
37
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
system for four or five months to develop the algorithm for discriminating between high-Z and
lower-Z material. DNDO conducted Technology Demonstration and Characterization (TD&C) in
February and March 2010 to characterize the performance of the prototype. 55 The other CAARS
programs are discussed in their respective sections below.
Part of the course correction was establishment of the Joint Integrated Non-Intrusive Inspection
(JINII) Program—”joint” because the project is a collaboration between DNDO, CBP, and the
DHS Directorate for Science and Technology, and “integrated” because it seeks to detect both
traditional contraband and high-Z material. It is “non-intrusive inspection” (NII) in the sense CBP
uses the term, namely, CBP personnel could clear containers, without physically having to open
them, with high confidence that they do not contain contraband or SNM.
JINII has two main components. One is CAARS. The other is a test campaign by DNDO to
examine the ability of existing, commercially available radiography systems to detect high-Z
material by means of operators visually inspecting radiographs. This is distinct from CAARS,
which is intended to highlight suspicious areas automatically. In addition, systems have been
developed outside of the CAARS program that have a limited capability—which DNDO
anticipates would be less than that of CAARS—to detect high-Z objects in cargo automatically.
One of these systems, the Rapiscan Eagle Portal, completed TD&C in September 2009.56 DNDO
states that if tests demonstrate this limited capability with commercial off-the-shelf systems,
deploying such systems could place capability in the field sooner than would be the case by using
only CAARS systems.
CBP already deploys the Rapiscan Eagle, which uses an accelerator (6 MeV in some versions, 4
MeV in others) to generate a radiographic image of a cargo container. The version that was tested
through JINII has an added algorithm, “Auto-Z,” that is designed to detect high-Z material in
cargo and indicate the location of such material on a radiograph automatically. In contrast to
either CAARS candidate, it would cost less, would use the current supporting infrastructure, and
would require little added operator training. According to Rynes, the algorithm should be able to
detect a 400-cc cube of high-Z material (7.37 cm, or 2.9 inches, on a side) nearly as well as
CAARS candidates, but could not detect a 100-cc cube of high-Z material (4.65 cm, or 1.8
inches, on a side) as well as they could. 57
In March 2010, DNDO completed its Technology Demonstration and Characterization (TD&C)
testing for the L-3, Rapiscan, and SAIC systems and put the CAARS follow-on program on hold.
DNDO expects the TD&C to provide data to quantify the performance and cost of these three
systems so that it can perform a cost-benefit analysis that would consider both high-Z and
traditional contraband detection, helping it decide how to proceed. DNDO could recommend
future development, operational testing, acquisition, or some combination, of the various systems.
If DNDO, CBP, and Congress judge the L-3 or SAIC systems to be more cost-effective than
existing systems, one or more systems with CAARS-level capability could be further developed
55
“Developmental Test and Evaluation” (DT&E) and “Technology Demonstration and Characterization” (TD&C) are
both tests of systems conducted by the government (as opposed to contractors). DT&E is a term used in the acquisition
process; DT&E tests systems before beginning procurement. As noted earlier, DNDO undertook a “course correction”
in April 2008, moving CAARS from an acquisition program to an R&D program. Some felt it was confusing to use the
term DT&E for a non-acquisition program, so DNDO decided to call the government tests of CAARS candidates
TD&C.
56
Information provided by Joel Rynes, April 16, 2010.
57
Information provided by Joel Rynes, July 8, 2009.
Congressional Research Service
38
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
or acquired through a competitive process yet to be determined. With TD&C completed, the
status of the test equipment as of May 2010 was as follows, according to Rynes,:
Significant portions of the AS&E CAARS material have been distributed to other DNDO
projects. The SAIC CAARS system is being dismantled and significant portions of the
material are being distributed to other DNDO projects. The L-3 CAARS system is being
dismantled and stored for potential installation at a later date. This would be [done] outside
of the CAARS program. The Rapiscan Eagle remains installed at Moffett Field [CA]. It has
funding to stay there through FY11. 58
The budget for the entire CAARS program is: FY2006, $16.3 million; FY2007, $26.4 million;
FY2008, $31.8 million; and FY2009, $26.1 million. The CAARS program per se ended in
FY2009, to be replaced by a followon DNDO-CBP program to advance CAARS technology so
that it can be deployed in the field. FY2010 funding for this program is $15.2 million.
Risks and concerns
Scientific risks and concerns
The L-3 system is intended to be the least complex and lowest-risk CAARS system. The
scientific risk to its hardware is low because it uses commercially available accelerators and
detectors. The software risk is potentially higher because algorithms to sort high-Z and low-Z
materials on a radiograph automatically have not been fully developed and have only been
modeled in a simulated environment, not tested in an actual operational one. The other two
CAARS technologies are more complicated, so their scientific risk may be greater.
Engineering risks and concerns
The concept has been demonstrated in the laboratory, but it remains to be shown that it can be
scaled up to the size CBP needs to scan containers. For example, can the algorithm to sort pixels
into high-Z and lower-Z bins handle large enough quantities of data in a timely manner? At high
enough energy levels, an accelerator can produce neutrons that require a large amount of
shielding. The x-rays generated by the accelerator can also scatter from interactions with the
cargo, also requiring shielding. A large footprint for shielding could preclude deployment of
detection systems at some ports. At issue: Can radiation be kept low enough that the footprint
does not become excessively large?
Cost and schedule risks and concerns
Slipping the CAARS schedule through the “course correction” to permit more time for R&D
could make it easier to meet the new schedule, reducing schedule risk. More R&D could also
reduce cost risk. On the other hand, scientific risk could result in cost and schedule risk.
Another cost and schedule risk is that work on three projects would reduce effort devoted to any
one system, delaying each system (as compared to the schedule possible if the full funding had
been applied to only one system) and increasing the cost of the total program. For example, it
58
Personal communication, e-mail, May 12, 2010.
Congressional Research Service
39
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
could be argued that a more efficient use of funds would be to focus only on the L-3 system
because its development is furthest along. On the other hand, a multi-pronged approach may offer
countervailing advantages. The simpler technology could (presumably) be deployed quickly,
adding capability quickly and providing a hedge against failure of more advanced systems. Morecapable systems could be deployed later, reducing the time pressure to develop them. Conversely,
pursuing several approaches hedges against the prospect that the simpler system might encounter
unanticipated problems that delay it to the point where it and a more advanced system could be
deployed at about the same time. Whether these advantages justify the higher cost is always a
matter of debate.
CBP notes that other companies are developing systems to detect high-Z material outside of the
CAARS program, and argues that, because of the operational concerns discussed next, the money
spent on CAARS could be better spent on such systems.59 DNDO points out that the JINII
program is developing and evaluating some of these systems. 60
Operational risks and concerns
(1) A 3% false alarm rate, if achieved, would place a large burden on CBP CONOPS, as it would
require responding to many false alarms. Alternatively, as discussed in the section on computer
modeling, a high false alarm rate could lead operators to ignore alarms or to raise the threshold
for alarms, increasing the likelihood of missing an actual threat. (2) Joel Rynes, Program
Manager, CAARS Program, DNDO, states that an operator could clear some false alarms without
intrusive inspection, such as by inspection of radiography images, reducing the rate of alarms
requiring intrusive inspections to well below 3%. At issue is whether there is enough confidence
in operator judgment with these methods to avoid intrusive inspections. (3) The system does not
differentiate between SNM and other high-Z material; adding non-SNM high-Z alarms to false
alarms boosts the non-SNM alarm rate and, presumably, the rate of alarms requiring intrusive
inspections. Without a way to differentiate between SNM and other high-Z material, can CAARS
meet its goal of having “little or no impact on CBP operations”? On the other hand, is any high-Z
material suspicious and worthy of inspection, so that non-SNM high-Z alarms should not be
counted as false alarms? (4) A CAARS program goal is automated detection of small amounts
(100 cc) of high-Z material through 10 inches of steel, with a follow-on goal of penetrating 16
inches of steel. 61 Is that a reasonable goal? If not, what is the maximum thickness of steel
commonly found in cargo containers and what energy level of photons would be required to
penetrate it? How would the added shielding required by equipment that generates photons
greater than 9 MeV affect deployment at ports, where space is at a premium? Or would no
practical energy level suffice to penetrate that much steel? Alternatively, would a simple “Bucky
collimator,” as described under “Background” in this section, make radiography much more
effective? (5) The radiation produced by CAARS, or other high energy imaging systems, even at
low levels, may cause interference with radiation portal monitors that have already been installed.
59
Information provided by Ira Reese, Executive Director, Laboratories and Scientific Services, and Patrick Simmons,
Director, Non-Intrusive Interrogation, both of Customs and Border Protection, personal communication, October 23,
2008.
60
Information provided by Joel Rynes, personal communication, October 28, 2008.
61
Domestic Nuclear Detection Office. “Cargo Advanced Automated Radiography (CAARS) Program Update Brief,”
slides 3, 18. This source is a briefing marked by the Domestic Nuclear Detection Office, the originating organization,
“For Official Use Only.” That office has approved CRS use of the information cited by this footnote.
Congressional Research Service
40
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
It is important to minimize this interference, though it can readily be eliminated by turning off the
accelerator when portal monitors are in use.
CBP raises other operational concerns. The first concern applies to the L-3 and AS&E CAARS
systems; the others apply to all three. (1) Because of an apparent miscommunication between
DNDO and CBP, DNDO thought that the dimensions of a one-of-a-kind CBP radiography
system, 60 x 160 ft, were acceptable to CBP for large-scale acquisition, some 300 to 500 units
nationwide. The L-3 and AS&E CAARS systems would use a concrete enclosure of that size (see
Figure 10) for radiation containment. However, CBP states that no unit that size would fit in U.S.
seaports, and that only four or five ports of entry on the U.S.-Mexican border could accommodate
the units. (2) CBP expresses concerns that the radiation emitted by radiography equipment such
as CAARS could require a large exclusion zone to protect workers. In contrast, radiation portal
monitors emit no radiation, but passively sense radiation emitted by radioactive material. (3) CBP
is concerned that total scan time would create an immense delay for container traffic entering the
United States from seaports or land border crossings. While DNDO’s goal is to have CAARS
scan a container in 15 seconds, CBP notes that the time when the scanning unit is on is only a
small part of the total time a scan requires. It points to the following sequence: a driver would
pull a truck with a container into the scanning enclosure (depicted in Figure 10); the driver would
leave the truck and go to a radiation-protected facility; the scanning equipment would move over
the stationary container at a precisely controlled speed; the truck and container would be scanned;
and the driver would reenter the truck and drive it out of the enclosure. This sequence, CBP
estimates, could take 5 minutes.62
Thomas Winkowski, Assistant Commissioner, Office of Field Operations, CBP, provided was
emphatic on the need to minimize delays. Not referring specifically to CAARS, he said that
rebooting and reinstalling software on some systems “can take seven or eight minutes. That’s the
kiss of death in my business from the standpoint of delays.” Regarding CAARS, he said that CBP
was “at the table” with DNDO in discussing requirements.
But what our concern was was that the footprint was too big ... and the throughput, you
know, for all your trucks to go through a car wash type system, as I call it, and the driver
comes out, and you do your scan—that realistically presents a tremendous amount of
problems from cycle time. So our position was that we really needed a different technology
that was more flexible and that didn’t have such a big footprint and require so much
handling.63
Rynes stated that concerns such as the foregoing raised by CBP were among the reasons for the
CAARS course correction and the establishment of the JINII program.64
Potential gains by increased funding
Rynes states that added funds could improve CAARS in several ways: (1) Further development
could reduce the size of CAARS systems so they would be more readily deployable at ports. (2)
62
Ira Reese, Executive Director, Laboratories and Scientific Services, and Patrick Simmons, Director, Non-Intrusive
Interrogation, both of Customs and Border Protection, provided information in this paragraph, personal
communication, October 23, 2008.
63
Testimony of Thomas Winkowski, Assistant Commissioner, Office of Field Operations, U.S. Customs and Border
Protection, before the Senate Homeland Security and Governmental Affairs Committee, September 25, 2008.
64
Information provided by Joel Rynes, personal communication, October 28, 2008.
Congressional Research Service
41
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
Money spent (by CAARS or another project) to improve detector efficiency could permit a lesspowerful electron source to suffice, further reducing the requirements for shielding. (3) Added
funds could improve algorithms to automate the processing of radiographs. (4) JINII is just
beginning to develop algorithms to detect contraband automatically to increase the rate at which
containers are inspected. Added funds would accelerate this development. (5) Added funds could
expedite better integration and data fusion of CAARS with already-deployed radiation portal
monitors.65
Potential synergisms and related applications
The preceding paragraph discussed synergisms by which technology developments could
improve CAARS capabilities. Another synergism concerns the use of neutrons or high-energy
photons to induce fission to discriminate SNM from other high-Z material. DNDO issued a broad
agency announcement in March 2008 to develop this capability.66 67 The CAARS candidates do
not include technology for this purpose. For example, they do not count neutrons or gamma rays
that would be generated by photofission even though 9-MeV photons can cause that effect.
However, CAARS might possibly draw on such work in the future.
SAIC CAARS: A Higher-Risk, Higher-Benefit Dual-Energy
Radiography System68
The “problem” and “background” information for the L-3 system, as described above, are the
same as for the SAIC system.
Technology description
Science Applications International Corporation (SAIC) developed a system (unnamed) as part of
the CAARS program. That program completed Technology Demonstration and Characterization
of the competing systems in March 2010, and as of April 2010 the future of CAARS was unclear.
However, SAIC is competing to apply the technology it developed under CAARS to another DHS
program as described below.
SAIC’s basic CAARS system uses an electron accelerator developed by Accuray Corporation. 69
The accelerator is “interleaved”—a single unit produces pulses of electron beams alternating
between 6 and 9 MeV. Each pulse lasts 3 microseconds, with 2.5 milliseconds between pulses.
65
Personal communication, July 11, 2008.
66
U.S. Department of Homeland Security. “Advanced Technology Demonstration for Shielded Nuclear Alarm
Resolution.” Broad Agency Announcement 08-102 for the Domestic Nuclear Detection Office, Transformational and
Applied Research Directorate, March 2008, p. 22, available at https://www.fbo.gov/files/6f6/
6f6f82cc03fe6fcbf298a7e3903a15b7.doc?i=640d9818cdf76fa884b8c064b8f19376.
67
As of April 2010, all Shielded Nuclear Alarm Resolution contractors had completed their preliminary design reviews
by September 2009 and are working toward their critical design reviews, in which DNDO would approve the design.
The first unit is scheduled to begin testing early in CY2011. Information provided by Joel Rynes, April 16, 2010.
68
Rex Richardson, Vice President and Principal Scientist, Science Applications International Corporation, provided
detailed information for this section, personal communications, June-August 2008. Others have commented as well to
provide alternative perspectives.
69
See http://www.accuray.com/.
Congressional Research Service
42
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
These beams strike a copper target and generate photons with a spectrum of energies up to the
highest energy of the electron beam, though lower-energy photons are filtered out.70 During a
scan, the photons pass through a cargo container in a vertical fan-like beam. On the other side of
the container is a detector array composed of multiple detector elements arranged in thin vertical
bands. Each element records an opacity image of a narrow horizontal “stripe” of the container. An
algorithm combines the stripes into a dual-energy radiograph, as described in the L-3 section. The
system automatically flags for further inspection areas of high-Z material and areas with too
much dense material for the photons to penetrate. Figure 11 illustrates the configuration of a unit.
Figure 11. SAIC CAARS Prototype
Source: Photograph provided by SAIC, May 2010
Notes: A bridge-like structure is mounted on rails some 20 ft apart. The unit on the far left houses the dualenergy accelerator and electronics. The vertical housing to the right of the truck is in line with x-rays generated
by the accelerator. It contains detectors that record the radiograph, pixel by pixel. To the right of the housing is
a steel plate 12 ft high by 20 ft wide by 1 to 7 in thick that absorbs x-rays that scatter off the cargo in the
direction of the beam. It weighs some 70,000 lb but is part of the unit, so the unit is self-contained. A thinner
plate with the same length and width to the left of the truck stops lower-energy photons that scatter backward
70
Electron beams of the levels discussed here generate bremsstrahlung photons of a wide spectrum of energies. Most
of the photons are of relatively low energy, well below 1 MeV. They contribute almost nothing to the radiographic
image but can represent a major portion of the radiation exposure to personnel. To remove or “filter” them, a piece of
copper is placed in front of the photon beam. The material is thick enough to stop lower-energy photons but not higherenergy ones.
Congressional Research Service
43
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
from the cargo. In operation, the driver exits the truck, and the structure—including accelerator, detector, and
steel plates—moves over the truck at 3 ft/sec. The control booth on the right is stationary; the tower on the far
right is not part of the system. This photograph was taken at SAIC’s San Diego factory in September 2009 during
readiness testing in preparation for DNDO’s Technology Demonstration and Characterization event.
One key to this system is that the proprietary detector operates at a photon flux (number of
photons per unit time) one-hundredth that of conventional cargo imaging systems. A lower flux
enables the accelerator to be much more compact. Accelerators with high photon flux require a
high-Z material, typically tungsten, for the bremsstrahlung target because it can withstand the
high heat from the electron beam. Such accelerators also use high-Z material to shield the photon
beam. However, photons with energies greater than approximately 6 MeV produce photoneutrons
(neutrons knocked off atoms by high-energy photons) when they strike high-Z materials, with the
number of such neutrons increasing as energy increases. The heat generated by the beam with
lower photon flux is low enough that copper can be used instead of tungsten. Copper is one of a
few metals with a threshold greater than 9 MeV for producing photoneutrons. As a result, SAIC
states, its system produces virtually no photoneutrons, eliminating the need for large concrete
structures for neutron shielding. Further, since the system has a low beam flux, SAIC states that
the need for shielding from x-rays that scatter in the cargo is greatly reduced.71
Another key to the system is the dual-energy interleaved accelerator. Three aspects are especially
consequential. First is the ability to construct this type of accelerator; previous efforts by Varian
Medical Systems, funded by SAIC in 2004, had poor X-ray beam stability.72 Second, this
accelerator uses a lower beam flux, with the advantages just discussed. Third, the number of
photons per pulse (“dose repeatability”) varies by a very small amount, less than 0.4%, far below
the required variance of less than 5 %.
While the goal of CAARS is to find high-Z material while not interfering with efforts by CBP
inspectors to find traditional contraband (drugs, guns, money, etc.), the better-than-expected dose
repeatability enabled the system to differentiate approximately 15 bands of Z from carbon (Z=6)
to uranium (Z=92), an atomic number resolution of about six (e.g., the difference between
oxygen, Z=8, and silicon, Z=14). This contrasts with the ability to find only materials with Z
greater than 72, as in Figure 9. On a radiograph in various shades of gray, the difference may not
be apparent at all, but if each Z band is represented by a different color, different materials display
much more clearly, as Figure 12 shows. Colors are assigned to Z bands arbitrarily, producing a
“false color” or “pseudocolor” radiograph, so called because the colors bear no relationship to the
color of materials being radiographed.
How does reducing the variance increase the number of Z-bands? Dual-energy radiography finds
the thickness, as measured by each of the two beams, of the material recorded for each pixel. For
example, it finds the number of photons (in this case, x-rays) transmitted through a cargo
container, and then received at the detector that measures the number of photons at each pixel. An
algorithm calculates the ratio of the two thicknesses and displays this ratio as an image in colors
or grayscale, pixel by pixel. Thickness ratios are calibrated before each scan. Because the number
71
The beam flux is further reduced by filtering. Electron beams of the levels discussed here generate bremsstrahlung
photons of a wide spectrum of energies. Most of the photons are of relatively low energy, well below 1 MeV. They
contribute almost nothing to the radiographic image but can represent a major portion of the radiation exposure to
personnel. To remove or “filter” them, a piece of copper is placed in front of the photon beam. The material is thick
enough to stop lower-energy photons but not higher-energy ones. After filtering, 6- and 9-MeVbeams generate the
greatest number of photons at about 2 and 3 MeV, respectively.
72
Beam stability refers to uniformity in both the duration of an x-ray pulse and the number of x-rays in a pulse.
Congressional Research Service
44
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
of x-rays varies from pulse to pulse, the output end of the accelerator has an x-ray dose meter to
correct for this variance. However, this correction can introduce large errors in the ratios
measured at some pixels because the correction applies only to the beam output and not to the
beam as received at different points. If there were no variance at all in the output (number of xrays per 6-MeV beam and per 9-MeV beam), this correction would not be needed, eliminating
this error source and increasing the number of Z bands that can be displayed. At the same time,
even with zero variance, the number of Z-bands that can be displayed is limited, and can be
reduced, by such factors as the size and thickness of objects being radiographed, and by the mix
of objects of different Z numbers that a beam might pass through for a given pixel.
Figure 12. Dual-Energy Radiography with False Colors Assigned
Bottom Image Can Differentiate Between Approximately15 Bands of Atomic Numbers (Z)
Columns have same density/area
Pb
Sugar
Motor Oil
Fe
Al
HD Poly
U uranium
Pb lead
Cu copper
Fe iron
High-Z
U
Pb
Cu
Fe
Al
Low-Z
Handgun
Drug Simulant
SAIC CAARS* 6 MeV / 9 MeV dual energy separation of materials by atomic number
(*work funded by DHS DNDO)
Source: Image provided by SAIC, April 2010.
Notes: This figure shows two radiographs of the same objects in a HINO delivery truck. The top image is a
conventional radiograph made using the 9 MeV beam. The bottom image assigns a color arbitrarily to each pixel
according to the ratio of (a) the thickness as measured by the 9-MeV beam to (b) the thickness as measured by
the 6-MeV beam. Higher ratios (higher Z) are displayed in this image as being toward the red end of the
spectrum while lower ratios (lower Z) are displayed more toward the blue end, though the algorithm can display
selected Z bands in whatever color the operator chooses so as to facilitate searching for items of concern for a
particular cargo container.. As a result, the colors bear no physical relationship to the material being
Congressional Research Service
45
Detection of Nuclear Weapons and Materials: Science, Technologies, Observations
radiographed. Because the 6- and 9-MeV beams have little variance in the number of photons per pulse, it is
possible to maintain a precise calibration of the relative intensities, leading to the 15-band atomic number
resolution. An operator could use the colors not only to find high-Z material, but also to find contraband hidden
in a material of slightly different Z. The algorithm can highlight anomalous areas, in this case a handgun and a drug
stimulant.
Text reading “columns have the same density/area” refers to the four stacks, or columns, in the middle of each
image. For each stack, the three blocks were select
This text is long and has been trimmed here. Open the source document for the complete record.
This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.