1. Introduction

Terahertz (THz) waves fall in the boundary region between radio and light waves, thus exhibiting characteristics of both. Similar to radio waves, THz waves can penetrate substances such as plastic and paper; however, like light waves, they can be easily manipulated by lenses and mirrors. Moreover, they have wavelengths of hundreds of µm, and thus a spatial resolution, which is determined by their diffraction limit, sufficient for imaging to be accomplished in non-destructive drug inspections. Unlike the millimeter and microwave regions, the THz-wave region contains unique absorption spectra (fingerprint spectra) for a range of reagents, which makes them ideal for conducting non-destructive and contactless inspections [1, 2].

Currently, non-destructive inspection techniques for illegal drugs in envelopes have not been fully developed, allowing rampant smuggling of such substances in and out of the country. Laws in many countries do not allow opening of suspicious letters or packages without a search warrant, and delays in acquiring warrants have led to higher rates of illegal drug trafficking. Presently, deterrent measures include X-ray examination, drug-sniffing dogs, and swipes for tracing drugs in packages. While X-ray examination can identify the shape of a bag in an envelope, it cannot identify the substance within and therefore cannot provide sufficient evidence to justify opening the package. Drug-sniffing dogs and swipes have further limitations as they are only effective in cases where trace amounts of illegal drugs are attached to the surface of the package. Consequently, inspection methods using millimeter waves and infrared light have been developed. Millimeter waves have high penetration but do not contain characteristic fingerprint spectra, making it difficult to identify drugs inside packages. Alternatively, while infrared light can obtain chemical fingerprint spectra, it is highly attenuated by thick coverings, making it unfit for inspection of illegal drugs inside thicker envelopes.

In contrast, THz waves can penetrate various cover materials and also contain fingerprint spectra characteristics for illegal drugs, which make them a highly viable candidate for conducting non-destructive, contactless drug inspections. Since the 1990s, we have been developing a wide band, frequency-tunable THz wave source using nonlinear optical effects [3–5]. In 2003, we reported the first instance of a non-destructive illegal drug inspection inside envelopes using THz spectral imaging with a THz Parametric Oscillator (TPO) [2, 6]; the system used then had a dynamic range below four orders of magnitude. Nevertheless, drug detection through 0.1-mm-thick envelopes was demonstrated for international postal services. However, real practice requires the inspection of mail in which substances are hidden by thicker coverings. For such purposes, several studies have been reported worldwide using THz time domain spectroscopy (TDS) [7, 8]. On request from Japanese Customs, our research groups have also worked on developing a spectral detection system using TDS for identifying drugs in thick envelopes in collaboration with industry between 2005 and 2009. Our investigations at that time were only able to identify spectra in thin envelopes [9]. These studies were not successful in detecting illegal substances in mail inspections of packages with thicknesses as great as those reported in this study. In TDS, the detection area of photoconductive antenna is as small as a few tens of µm, hence its performance is easily affected by scattering and refraction by the target. Therefore, a number of issues must be resolved before successful implementation. Currently, studies are being conducted to make improvements through signal processing [10–13] and development of high-performance system [14].

Our recent studies have succeeded in improving the output power and sensitivity of this THz spectral imaging system by injection-seeded THz parametric generation (is-TPG) and detection, which has achieved a dynamic range of 100 dB [15–17]. Thus, is-TPG has a generation principle fundamentally different from that of TDS. It is a unique, broadband THz wave source that can produce a wavelength tunable, single wavelength Fourier transform-limited pulse. This study reports the development of this THz spectral imaging system, which uses generation and detection techniques with a high dynamic range. The comparison between the newly developed imaging system and the previous one is shown in Table 1.

Table 1. Comparison between the TPO used in 2003 and the is-TPG used in this study

View Table

2. Experimental setup

As shown in Fig. 1, the THz spectral imaging system with is-TPG used in this study consists of a microchip Nd:YAG laser (Hamamatsu Photonics Co. Ltd., Hamamatsu, Japan), an external cavity diode laser (New Focus Inc., San Jose, CA, USA), two nonlinear crystals, an XY stage, a lock-in amplifier, and a computer for data acquisition. The nonlinear crystals are lithium niobate doped with 5-mol% MgO (MgO:LiNbO₃). Their sizes are 50 × 6 × 5 mm³ on the generation side and 65 × 6 × 5 mm³ on the detection side, to increase the parametric gain. The pump beam source has an output energy of 18 mJ/pulse by optical amplification of the microchip Nd:YAG laser, characterised by a wavelength of 1,064.4 nm, a pulse width of 450 ps, a pulse energy of 1.2 mJ/pulse, and repetition frequency of 50 Hz. The amplified pump beam is split in two by a beam splitter, and the two beams are incident on the crystals of the generation and detection sides [17]. Moreover, the beam from the continuous and variable-wavelength external cavity diode laser was amplified to 500 mW as the seed beam. This seed beam was incident on the LiNbO₃ crystal after passing through an achromatic optical system with a grating and a telescope system to satisfy phase matching condition when the wavelength is changed [18]. A single-frequency, high-power THz wave can be generated by directing the pump beam and seed beam into the LiNbO₃ crystal in a way that satisfies the non-collinear phase-matching angle, which enables a change in THz wavelength across a wide band solely by adjusting the wavelength of the seed beam. The generated THz wave is focused using a lens with a focal length of 100 mm, and the beam diameter at the focal point is approximately 1 mm. The sample is scanned in steps of 1 mm at the focal point using an XY stage. The scanning region used was 12 × 42 mm², thus creating 12 × 42 = 504 pixels. The THz wave transmitted through the sample was incident upon another LiNbO₃ crystal for detection. Owing to the parametric process, a near infrared idler beam was generated at a wavelength corresponding to the difference between the frequencies of the incident THz beam and the pump beam in the detection crystal [15]. This idler beam is detected by a near-infrared pyroelectric detector (ED100A^UV by Gentec-EO [Québec, Canada]) and recorded through a lock-in amplifier. The intensity of an idler beam is proportional to the THz wave intensity, therefore an extremely small change in THz wave intensity can be monitored by measuring the infrared idler beam. This enabled us to significantly improve the dynamic range of this new system in comparison with the previous one in 2003.

 

Fig. 1 Terahertz (THz) spectral imaging system.

Download Full Size | PDF

Figure 2 shows the region of the variable frequency of the generated THz wave, which is achieved by changing the frequency of the seed beam and the dynamic range of each frequency. In this experiment, to measure the dynamic range of the measurement system, the THz waves were attenuated in steps of 10 dB by sets of attenuators with transmittances of 30%, 10%, 3%, and 1% (Microtech Instruments, Inc., Eugene, OR, USA). Figure 2 shows that this system has a spectral range of 1.0–2.6 THz and a dynamic range of over eight orders of magnitude. Furthermore, the wide range between 1.2 and 2.2 THz has a dynamic range of over six orders of magnitude, making this system a practical spectrometer.

 

Fig. 2 Variable wavelength range and dynamic range of the injection-seeded THz parametric generation (is-TPG) system.

Download Full Size | PDF

3. Data analysis

This section provides an explanation of component spatial pattern analysis method used in this study [19]. First, the image of a sample comprising M separate substances with different absorption spectra from N different wavelengths can be expressed using the following linear matrix equation:

4. Experimental results

The samples used in this study were three saccharides: maltose, glucose, and fructose. In the 2003 report [2], we were able to use actual illegal drugs for the experiment in the presence of agents from the National Research Institute of Police Science. However, restrictions have been tightened recently, making it difficult to use actual illegal drugs for experimentation; hence, saccharides were substituted. Previous experience has shown the possibility of substituting illegal drugs with saccharides. During the sample preparation, particle sizes for the maltose, glucose, and fructose powders were selected to be 30 –130 µm and the powders were placed in plastic bags to prepare 1-mm-thick samples. The size of each sample was approximately 10 mm × 10 mm. The three saccharide samples were positioned in the order of maltose, glucose, and fructose, as shown in Fig. 3(a). To create a denser wrapping, three types of covering material were used: EMS envelopes, which are used for international mail, cardboard, and bubble wrap. Two EMS envelopes, two pieces of cardboard, and four sheets of bubble wrap were placed on the front and back of the saccharide samples as shown in Fig. 3(b)-3(d). The thickness after affixing the coverings was approximately 23 mm, much thicker than the 0.1-mm-thick envelopes used in the 2003 study [2]. The thick envelope used in this study is typical of that used for smuggling illegal drugs, since it can be used to mail contents within the specified weight limits.

 

Fig. 3 Sample preparation (a) Saccharide powders, (b) Covering materials (EMS envelope (top), cardboard (left), and bubble wrap (right)), (c) covered form (front), and (d) covered form (side).

Download Full Size | PDF

In this experiment, all the measurement were done in Nitrogen purged environment. We varied the frequency from 1.4 to 1.9 THz, as shown in Fig. 4, and measured N = 12 multi-spectral images, which were then used to make an N × L = 12 × 504 matrix [X]. The color scale of the image is expressed as the logarithm of the ratio -ln(I_t/I₀) where I_t is the transmission intensity of the sample and I₀ is that of the wrapping materials. Therefore, a whiter image indicates higher absorption. In the multi-spectral images, the glucose has relatively lower absorption and thus a darker image; however, the component pattern analysis compensates for the difference in absorption between the substances and expresses only the spatial pattern of each component. In other words, detection is possible even for low absorption.

 

Fig. 4 Multi-spectral images.

Download Full Size | PDF

Figure 5 shows the absorption spectrum of each saccharide sample measured using this system; the absorption peaks of each component can be seen (maltose: 1.1 THz and 1.6 THz, glucose: 1.4 THz, and fructose: 1.7 THz) [20]. It took about 3 minutes to measure each spectrum. The matrix [S] was constructed by obtaining the absorbance corresponding to the 12 frequencies used in the multi-spectral images. Here, considering that the frequency dependences of noise components such as paper, plastic and scattering are smaller than that of saccharides, the matrix [S] is constructed from the frequency-independent components to eliminate noise [6]; that is, [S] is an N × M = 12 × 4 matrix.

 

Fig. 5 Absorption spectrum of each saccharide sample measured by the is-TPG system.

Download Full Size | PDF

By substituting [X] and [S] in Eq. (3), we obtained the M × L = 4 × 504 matrix [C]. The results of extracting each saccharide component from the matrix [C] are shown in Fig. 6, which illustrates that the three saccharide samples were identified separately despite the thick wrapping materials. Moreover, the density of the white color captures the spatial distribution of the powders. Figure 7 shows the one-pixel absorption spectrum of each saccharide sample showing the absorption peaks, thereby correctly identifying each saccharide sample.

 

Fig. 6 Spectral imaging results (from the top: maltose, glucose, and fructose).

Download Full Size | PDF

 

Fig. 7 One-pixel absorption spectrum of each saccharide sample. The arrow head shows the absorption peak of each sample.

Download Full Size | PDF

5. Conclusion

A THz spectral imaging system was constructed that used an injection-seeded THz parametric generation and detection, and successfully identified images of saccharides through wrapping material much thicker than those used in the 2003 trials. Using this system, spectral imaging is possible for not only saccharides but also cases where illegal drugs and multiple reagents are mixed together. This method can be used to inspect mail, check prescription accuracy at pharmacies, inspect pharmaceutical manufacturing processes, and determine illegal possession of stimulants and explosives. The dynamic range can be improved by increasing the resolution of the idler beam detector, so spectral imaging can be achieved through even thicker covering materials.