2
Principle of Bulk Explosives Detection
In this chapter, the principles behind the technologies developed for detecting explosives concealed in baggage are discussed. The two major categories are x-ray-based technologies and neutron transmission-based technologies. The discussion of x-ray-based technologies covers those currently deployed in airports, as well as some modified implementations. The discussion of neutron-based technologies is limited to PFNTS—the focus of this study.
X-Ray-Based Technologies
Many of the explosives-detection technologies that are based on x-ray techniques measure the x-ray attenuation of the materials that make up the baggage. Attenuation is a function of energy, density, and average atomic number. Because x-rays interact primarily with electrons, the attenuation coefficient is strongly correlated with the electron density of the material under investigation (NRC, 1998).
The mechanisms primarily responsible for x-ray attenuation in materials at the x-ray energy ranges typically used by explosives-detection equipment are Compton scattering and photoelectric absorption. The photoelectric effect results in x-ray absorption, whereas Compton scattering merely scatters x-rays, altering the path and energy of the scattered photons (x-rays). The significance of the photoelectric effect is greater for materials composed of elements with a high atomic number (Z), such as metals or other inorganic materials. However, this cross section drops off rapidly (i.e., the attenuation caused by the photoelectric effect becomes less relevant) with increasing x-ray energy. For organic materials (low Z), Compton scattering is the dominant x-ray attenuation process. The attenuation cross section caused by Compton scattering varies less with x-ray energy than does the attenuation cross section caused by the photoelectric effect. Materials can be distinguished from each another based on the relative importance of the photoelectric and Compton cross sections. For example, inorganic materials can be identified by rapidly changing x-ray attenuation with changing x-ray energy, whereas organic materials display a more subtle change (NRC, 1998).
Multi-energy x-ray-based detection equipment suitable for distinguishing organic from inorganic materials and for measuring densities semiquantitatively has been developed. Combining the measurement of transmission and backscatter x-rays improves the detection of light (low Z) elements as they are found in explosives; however, it does not specifically identify explosives. Dual-energy CT (computed tomography) is capable of providing geometrical information, as well as information pertaining to both the physical density and the effective atomic number of a material. Although the effective atomic number is not enough to completely characterize a material, it does provide discrimination capability above and beyond characterization by a physical density metric alone.
A common imaging method, x-ray radiography (or projection imaging) is a collection of x-ray attenuation line integrals over two dimensions. This method does not resolve the third dimension along the incident x-ray direction. CT adds the capability of visually displaying the physical appearance of the materials in question from all three dimensions. Reconstructing two-dimensional cross-sectional images (tomographs) and then full three-dimensional volumes can greatly improve the detection of explosive threats by identifying certain shapes or patterns, such as wires, batteries, or detonators, as well as by measuring the volume of the suspect material. The additional geometrical information supplements the material x-ray attenuation information and results in a more specific discrimination of explosive materials.
To date, four EDSs have been certified by the FAA, all of which are x-ray CT-based systems. Three are manufactured by InVision Technologies, Inc.: the CTX-5000, CTX-5000 SP, and CTX-5500 DS; a fourth system, the 3DX-6000, which recently passed the FAA certification test, was developed by L3 Communications, Inc.
Other x-ray-based methods using high-energy photons (between 10 and 30 MeV) have been discussed in the litera-
ture (Hussein, 1992; Gozani, 1988). Because of low-reaction cross sections, they require high x-ray flux rates produced by powerful accelerators that may not be suitable for airport use. These high-energy techniques are based on photon interactions with the nuclear properties of nitrogen, carbon, and oxygen. They provide spatial information, but because of the small-reaction cross sections, it is difficult to distinguish between elements.
Pulsed Fast Neutron Transmission Spectroscopy
In the PFNTS method, a collimated broad-energy (0.5-8 MeV) or "white" neutron beam is passed through the bag and the energy-dependent neutron transmission measured. By comparing the energy-dependent attenuation of the source neutron spectrum, the ratios of hydrogen, oxygen, carbon, and nitrogen can be integrated over a path (line) in the bag, and multiple lines can be used to produce a radiographic image (Overley, 1985). A fictitious element (X) with a smooth energy-dependent cross section is often included to normalize the transmitted number density (Overley et al., 1997). X is intended to represent a smooth neutron attenuation, which can be attributed to elements not specifically represented in hydrogen, oxygen, carbon, and nitrogen decomposition. For every pixel in the target, the energy dependence in the transmitted neutron spectrum is used to determine the relative amounts of these five elements. Figures 2-1a and 2-1b (Chmelik et al., 1997) show how projections in these five dimensions can be used to distinguish the presence of an explosive and often (~ 72 percent of the time [Lefevre and Overley, 1998]) to identify the type of explosive. The points in Figure 2-1a and 2-1b represent 38,000 measurements from actual airline suitcases, with and without explosives. Because of the scatter in the plotted points for explosive and nonexplosive paths, it is difficult to apply projected-path nitrogen-only detection schemes effectively. Because of the outliers in the distribution of "nonexplosive" black points, it is difficult to eliminate false alarms without affecting the probability of detecting the explosive.
In the basic PFNTS method, a five-dimensional representation of the elemental composition and the spatial distributions of "potentially explosive" adjacent pixels are used to support the detection algorithm. Various algorithms can be used to reduce the five-dimensional elemental information to an "explosive potential" for a single pixel. Variations in the detection algorithm can also increase the base set beyond the nominal five elements. The Tensor algorithm includes another element (Y), which is changed from element to element within a specified set of cross sections during a regression calculation until a best fit is obtained (Tensor Technology, 1998b). Different spatial correlation algorithms can be used to reduce the map of "explosive potential" metrics to a yes or no decision on the presence of an explosive in the test article. The University of Oregon refers to its detection algorithm as a "B-matrix'' and bases the "explosive potential" metric on a comparison with the explosive/nonexplosive probability observed in a simulation database. Separate B-matrices are maintained for each explosive class the algorithm is designed to detect. Tensor uses a neural network trained on a set of explosive and nonexplosive bags.
PFNTS requires the use of a tightly bunched, pulsed neutron source. Time-of-flight is used to determine the energy of the transmitted neutrons. The temporal width of the initial neutron pulse and the time resolution of the time-of-flight measurement limit the energy resolution of the transmitted spectrum. Flight paths of 4 to 10 m (13-33 ft) are commonly used. Figure 2-2 shows the total cross section for neutrons on 16O, 14N, and 12C. The narrow peaks in the interaction cross sections do not appear in a typical PFNTS measurement because they are smeared out by the energy resolution of the detectors (Miller and Makky, 1993). Thus, the element identification depends on the broad energy-dependent structures in the cross sections, rather than on the narrow resonances.
Because this method measures the energy-dependent neutron attenuation, it is critical that a broad energy neutron source be used. This rules out 2H(2H,n)3He (referred to as a deuterium-deuterium or DD reaction) and 2H(3H,n)4He (referred to as a deuterium-tritium or DT reaction) sources, which have a monoenergetic or restricted energy range. In order to get a reasonable neutron flux for high-energy neutrons (up to 8 MeV), accelerators are generally required. 9Be(d,n)10B or 9Be(p,n)9B reactions are candidate neutron sources (Micklich et al., 1996). Accelerators that can exploit these neutron-producing reactions need a high current (~10 mA time-averaged current, 1-ns pulse width, and 1-ms repetition frequency) and a high-energy deuteron source (> 4 MeV). Most laboratory experiments with PFNTS have used a deuteron linear accelerator and the 9Be(d,n)10B reaction.