Development of a high-energy-resolution EDXRD system with a CdTe detector for security inspection

In recent years, the illicit trafficking of drugs, explosives and other contraband has seriously threatened social security. Most of the trafficking occurs via bulk ground transportation modes, air travel and rail transportation. Inspecting dangerous materials hidden inside parcels and luggage rapidly is becoming necessary for combating the spread of contraband. Traditional techniques include imaging methods such as dual-energy X-ray transmission and X-ray Compton backscattering. These techniques are based on measuring the amount of X-ray absorption or Compton scattering due to the different hidden materials. Both absorption and Compton scattering are related to the effective atomic number (Z) and the material’s density (ρ).^1,2 Dual-energy X-ray transmission can easily inspect metal items because of the high Z in metal objects. X-ray Compton backscattering is strong at detecting organics because of the large scattering cross section. However, the Z and ρ of drugs and explosives are similar to those of harmless, commonly used items; therefore, it becomes very difficult to distinguish these substances from illegal substances. It is well known that X-ray diffraction can measure the crystalline structure of materials.³ This technique has been widely used in the fields of materials science, biomaterials and so on. Comparing to angular dispersive X-ray diffraction (ADXRD) techniques for substance analysis in many laboratories, energy dispersive X-ray diffraction (EDXRD) uses a polychromatic source instead of a quasi-monochromatic source. Polychromatic sources are easily accessible and produce more photons, nearly covering the entire available spectrum of wavelengths. Moreover, the components of an EDXRD system negate the need for rotating round the samples, which can simplify the equipment setup and shorten detection times. Thus, EDXRD has become an important technology for accurate identification of contraband in the field of security inspection.^4,5

For practical applications, fast and accurate security inspection requires a high-resolution and high-sensitivity EDXRD system. At present, several organisations have conducted research on fabricating systems based on the EDXRD technique, such as XRD 3500^TM from Morpho Detection^6,7 and DILAX from UCL.⁸ XRD 3500^TM uses a germanium detector and has a high energy resolution. However, the detector requires liquid nitrogen cooling, resulting in a bulky and expensive instrument.⁹ To ensure high sensitivity, a customised rotating anode X-ray tube is used for achieving 15 kW of power. DILAX employs a room-temperature semiconductor detector based on CdZnTe. However, the resolution of this system is about 23%, resulting in only two to three broad peaks for illicit drugs or explosives can be obtained. And each broad peak results from a summation of multiple diffraction peaks due to the low resolution. This seriously reduces the accuracy of identification. Therefore, for successful application of EDXRD for security inspection, it is necessary to develop an EDXRD system with both high resolution and high sensitivity using a conventional source and conventional detector.

This paper focus on the development of an EDXRD system with high energy resolution and high sensitivity. A simulation model was established, and the effect of the geometrical parameters on the energy resolution and collecting efficiency was studied. Using the model, an optimised EDXRD system with high energy resolution and high sensitivity was constructed. Experiments show that the system’s energy resolution was approximately 5% with a detection time of 3 s. Additional sampling time was needed to obtain an effective spectrum of the concealed samples because of the stronger attenuation. However, the detection time needed was less than 10 s. Combined with imaging techniques, our EDXRD system can accurately identify contraband hidden inside parcels or baggage within specific regions of interest (ROI).

EDXRD is based on Bragg’s diffraction law:

where $λ$ is the wavelength of the X-ray, d is the inter-planar spacing, $θ$ is the diffraction angle and n is a positive integer specifying the order of diffraction. Since most of the illicit drugs and explosives are crystals and every crystal has a unique structure, their crystalline structure can be used as a “fingerprint” for identification. If a sufficient subset of d in a particular material can be derived, identification of that material becomes possible.¹⁰ In the EDXRD method, Bragg’s law can be converted to Eq. (2) in terms of the X-ray energy:

where h is the Planck’s constant, c is the velocity of light and E is the energy of the incident X-ray. Diffraction occurs only when the energy of the incident beam satisfies Eq. (2). In the EDXRD spectrum, every peak corresponds to a d of the crystal. Therefore, the major factor for accurately acquiring the value of d is the resolution of the spectrum, which is decided by the angular resolution of the system. In this case, the energy resolution can be obtained using Eq. (3):¹¹ 

where ∆E is the full width at half maximum (FWHM) of the spectrum and ∆θ is the FWHM of the distribution of the diffraction angle.

Rapidly inspecting is required for security inspection. Thus, the other important property is the system’s collecting efficiency. However, both energy resolution and collecting efficiency are decided by the geometrical parameters of the system. To achieve a high-resolution system, a strict collimation of the incident and diffracted X-ray beams is necessary, which leads to a drastic decrease in the flux of the detected X-ray. Hence, a high-performance EDXRD system should balance energy resolution with collecting efficiency by optimising all of the variable parameters.

In order to design a system with high energy resolution and high sensitivity, a simulation model was established to investigate the effect of geometrical parameters on the energy resolution and collecting efficiency. The geometrical parameters of the system are listed as follows:¹² 

1. The diffraction angle θ;

2. The aperture width of the primary collimators P₁ and P₂ (W_P1 and W_P2) and the secondary collimators S₁ and S₂ (W_S) (assuming that S₁ and S₂ have the same aperture width);

3. The distance between the primary collimators P₁ and P₂ (L_P); the distance between the primary collimator P₂ and the secondary collimator S₁ (L_PS); the distance between the secondary collimators S₁ and S₂ (L_S). The detector was placed close to S₂.

Owing to the collimator aperture width, the incident and diffracted beams are not perfectly narrow. As shown in Fig. 1, the effective diffraction area is defined by the intersection of the incident beam and the diffracted beam. To calculate the angular spread of the system, the source is divided into segments, and the sample contained within the effective area is divided into voxels. For each source point, each voxel has a specific diffraction angle and acceptance angle. The accumulation of the diffraction angles achieved from all voxels and segments is the angular spread of the EDXRD system. Thus, ∆θ can be obtained, and the energy resolution caused by the system parameters can be calculated using Eq. (3).

The collecting efficiency of the EDXRD system can be represented by the ratio of the amount of photons received by the detector to the total number of photons radiated from the source. The sensitivity of the system and the time required for detection both depend on collecting efficiency. To compute the efficiency, a ray-tracing method was used. As shown in Fig. 1, assuming that each segment of the source irradiates photons uniformly within a certain angle, the angle of each photon emitted from the source determines whether the photon can arrive at the surface of the sample. After interfering with the sample, the photon at specific angle will arrive at the detector via the secondary collimator aperture. By repeating this procedure for each segment of the source and each angle of irradiated and diffracted photon, the ratio of the amount of photons received by the detector to the total number of photons radiated from the source can be acquired.

To design an optimised EDXRD system, we have to weigh every component of the system in the calculation of energy resolution and collecting efficiency. By controlling the variables and changing one parameter at a time, we calculated the energy resolution and collecting efficiency. The value of each parameter can be seen in Table I. For the simulation of efficiency, each calculation used the same total number of emitted photons.

The effects of W_P1 and W_P2 were simulated and displayed in Fig. 2. As the width increased, the effect of W_P2 on the angular spread of the EDXRD system was greater than W_P1, but the collecting efficiency was more influenced by W_P1 when the width was greater than 1 mm. Therefore, a wide P₁ aperture was chosen to increase the collection efficiency, and a small P₂ aperture was chosen to limit the angular spread for improving energy resolution.

In addition, the effect of L_P on the angular spread and collection efficiency is shown in Fig. 3(a). When the distance was less than 300 mm, both the angular spread and the efficiency varied drastically. The angular spread was approximately reduced from 0.9° to 0.4°, and the count of received photons decreased from 850 to 400. Very few variations after 300 mm were observed, leading to a nearly flat count. The sample should be placed between P₂ and S₁, so L_PS should not be less than 300 mm for inspecting a piece of luggage or a parcel. The effect of L_PS on the angular spread and the collection efficiency can be seen in Fig. 3(b). As the distance increased, the angular spread decreased slowly from 0.55° to 0.3°, and the efficiency decreased from 400 to 200. The value of L_PS was mainly based on the size of the detected samples.

The width of the secondary collimators (W_S) affected the acceptance angle of the detector. As shown in Fig. 4(a), the angular spread and the collection efficiency increased linearly as the secondary collimator aperture width enlarged, and the effect on the efficiency is more significant. As the width increased from 0.2 to 1.5 mm, the angular spread increased by 0.15°, and the received photon count increased by 750. Fig. 4(b) shows the change of the angular spread and collection efficiency with L_S. The angular spread was reduced by less than 0.1°, and the count was less than 100. L_S had little effect on both the energy resolution and the collecting efficiency. The most important parameter related to the secondary collimators was the aperture width.

At a diffraction angle of 5°, most drugs and explosives have two to six main peaks in the energy range of 20–50 keV. For preventing the detected diffraction peaks from overlapping, ∆E should be better than about 2.5 keV at 50 keV. In order to design a high-energy-resolution system, we limited $Δθ$ to no more than 0.25°. Within the limitations, the appropriate parameters were selected based on the size of the actual security equipment.

The geometry of the experimental system is shown in Fig. 5. The experiment was conducted using a tungsten target X-ray tube (NDI-225-22, Varian). The diameter of the source’s focal spot was 5.5 mm, and the radiation coverage was 40°. The X-ray source was operated at 80 kV and 25 mA to obtain a polychromatic beam with energy range from 0 to 80 keV. The diffraction photons were detected at a nominal 5° using a CdTe diode detector (XR-100T-CdTe, Amptek) with a 5 × 5 × 1 mm³ crystal and a two-stage thermoelectric cooler. The energy resolution of the detector was less than 1.5 keV FWHM at 122 keV (⁵⁷Co).

The EDXRD system contained three sets of collimators: a source collimator, primary collimators (P₁ and P₂) and a secondary collimator. Source collimation was used to prevent any off-focal light from interacting with system components and producing background noise. This was provided by an iron-lead collimator (85 mm long and 5 mm in diameter). P₁ was placed next to the source collimator. The diameter of the P₁ aperture was 1.5 mm. P₂ was placed at a distance of 400 mm from P₁, and the width of the aperture was 0.5 mm. In order to utilise the crystal of the detector fully, a set of Soller-slit collimators was used for the secondary collimation. According to the size of the detected samples, such as luggage and small parcels, the Soller-slit collimators were placed at a distance of 500 mm from P₂. The Soller-slit collimators contained 10 internal tungsten-steel plates (100 mm long, 0.5 mm thick and spaced at 0.5 mm intervals). With the Soller-slit collimators, the detector could cover a range of about 55 mm on the optical axis. On the basis of the simulation studies, the diffraction angle distribution of the EDXRD system is shown in Fig. 6. The FWHM of the angular spread ∆θ was 0.23°, and the theoretical energy resolution of the system, calculated using Eq. (3), was 4.6%.

The performance of the EDXRD system was first tested using a SiO₂ sample. The nominal diffraction angle θ was 5°. On the basis of the Joint Committee on Powder Diffraction Standards (JCPDS)¹³ database, the EDXRD spectrum of SiO₂ has two distinct peaks between 20 and 50 keV. As shown in Figure 7, the EDXRD system acquired experimental data that matched well with the JCPDS data. The ordinate “count” indicates the intensity count of the measured spectra, and the ordinate “intensity” indicates the intensity of normalized JCPDS data. The highest peak count for the data measured within 10 s was 60, and the count measured within 3 s was 20. The diffraction profile measured within 3 s was sufficient for identification and more sampling time makes the profile smoother. The FWHM of the 3 s data was 1.89 keV at 45.0 keV, and that of the 10 s data was 1.98 keV.

A second experiment was carried out using paracetamol (C₈H₉NO₂). Paracetamol has a complex crystalline structure with seven main peaks in the energy range of 20 to 50 keV, which are prone to overlap if the energy resolution is insufficient. The diffraction profiles of paracetamol with different detection times are shown in Fig. 8. The seven peaks were visible from the data measured within 3s. The FWHM of the 3 s data was 1.24 keV at 24.8 keV and 1.72 keV at 29.0 keV. The FWHM of the 10 s data was 1.43 keV at 24.8 keV and 1.91 keV at 29.0 keV. These results indicated that the energy resolution of the EDXRD system was 5%-6% depending on different integration times. The difference between the system resolution and the theoretical value may have been caused by the combined effects of detector resolution, alignment error, incoherent scatter and background noise.

A simulated traveller’s luggage experiment was performed using six different items placed in a wooden box to test the feasibility of the system for hidden contraband detection. As shown in Fig. 9(a), item A was a bag of paracetamol with a mass of about 25 g and a thickness of about 25 mm, item E was a bag of SiO₂, and item F was a pair of scissors. The remaining three items B, C and D were amorphous materials: a bag of flour, sunscreen and a “T” shape formed from PMMA. The wooden box was 7.5 mm thick. Because the effective detection area of the EDXRD system is small, about 1.3 × 12 × 55 mm³, it is necessary to first identify the regions of interest (ROI) for the detection of large objects. In this experiment, the X-ray Compton backscattering image was obtained to identify the ROI of the wooden box. Fig. 9(b) is the backscattering image of the box. Owing to the difference in density (ρ) and atomic number (Z), the different items appear at different brightness. The metallic item could be recognised, but it was impossible to distinguish the items with low Z from the image further. Moreover, it was even more difficult to identify the specific material composition of the items. Figs. 9(c)–(d) are the EDXRD spectra of ROI A–E with 10 s detection time. The profiles of the amorphous materials were continuous, and the profiles of the crystalline materials had obvious diffraction peaks. Moreover, the diffraction spectra of paracetamol and SiO₂ had significant distinctions in terms of the number of peaks and peak positions. The FWHM of the measured data was 1.7 keV at 24.8 keV and 2.0 keV at 29.0 keV. The counts on the spectra were higher than those in the previous experiments because the samples were larger. The difference in resolution was due to the stronger absorption and scattering from the wooden box.

Therefore, the developed system was capable of accurately measuring the diffraction spectrum of crystalline materials within a short integration time. Combined with X-ray transmission or Compton backscattering techniques, accurate material identification was possible for every region of interest in Fig. 9(b).

In this paper, a simulation model was established to study the effect of geometrical parameters on the energy resolution and collecting efficiency of an EDXRD system. The performance of the developed system demonstrated the capability of the EDXRD technique for identifying specific materials. The energy resolution of the system was approximately 5% with a detection time of 3 s. In practical applications, a bag of paracetamol hidden inside a wooden box could be distinguished effectively from other similar daily necessities. These results show that the developed EDXRD system can satisfy the requirement of acquiring a distinguishable spectrum for each material within a short detection time. This system can be used for the inspection of specific regions of interest detected by imaging techniques.

This work was supported by the National Key Research and Development Plan (NKRDP) (2016YFC0800904-Z03).