1. Introduction

Non-destructive visualization of internal defects is an important process in industrial production. Various imaging technologies using electromagnetic waves of different frequencies such as millimeter wave, infrared wave, and X-rays have been used in non-destructive testing and analysis by exploiting their unique characteristics [1,2]. However, each frequency range has advantages and disadvantages over alternative systems, and none is suitable for all samples. For example, millimeter waves can penetrate a wide variety of materials; however, image resolution is limited by the long wavelength. Infrared waves do not penetrate deep into samples and are easily scattered, whereas images of soft materials obtained using X-rays have poor contrast due to the high photon energies involved. The terahertz (THz) wave, between millimeter and infrared waves in the electromagnetic spectrum, is potentially useful for non-destructive, non-invasive sensing and imaging applications owing to the nonionizing nature of the radiation, and its moderate transmission through materials such as plastics, papers, ceramics, and dielectrics. Various applications of THz waves, such as the detection of illicit drugs [3], explosives [4], weapons [5], quality control of pharmaceutical products [6], and human bone imaging [7] have been demonstrated. Previously, two-dimensional (2-D) THz imaging has been demonstrated based on transmission or reflection measurements from the sample [8,9]. However, it is desirable to obtain three-dimensional (3-D) images to reveal further detail of a sample’s internal structure. THz tomography provides 3-D imagery, and various techniques (e.g., time-of-flight tomography and computed tomography [10–12]) have been reported. The majority of tomographic systems use terahertz time domain spectroscopy (THz-TDS) in which the emission and detection of pulsed THz waves is achieved using photoconductive antennas or non-linear optical crystals via excitation by a femtosecond laser. Such systems cover the frequency range from hundreds of gigahertz (GHz) to several THz, with a maximum dynamic range of ~10⁶ (around a few hundred GHz). However, the dynamic range decreases rapidly with an increase in frequency, thereby decreasing the signal-to-noise ratio and consequently increasing the uncertainty in optical constants for high-frequency measurements [13–15]. Moreover, such systems suffer from problems such as high scattering losses; therefore, it can be difficult to obtain an image of samples hidden in packages due to the insufficient dynamic range of the system. In contrast to such systems, an imaging system with a high-power, wavelength-tunable THz source and ultrasensitive detector based on an optical parametric process provides an excellent alternative for addressing such issues.

We developed a 3-D THz computed tomography system where the emission of frequency-tunable, high-power THz waves is achieved using an injection-seeded THz-wave parametric source. Ultrasensitive detection of these waves is accomplished using a novel scheme known as THz heterodyne detection. The unique emission and detection combination of this method yields a spectrally flat THz tomography system with a high dynamic range, which allows us to obtain high-quality, high-resolution 3-D tomographic images. We present 3-D images of a pencil, showing the lead inside it, and a plastic product with an internal defect to demonstrate the potential industrial applications of our system in non-destructive testing and evaluation.

2. Experimental setup

The THz tomography system is shown in Fig. 1. It comprises three sections: the THz emitter, a THz detector, and a sample holder mounted on translational and rotational stages. In our experiment, THz waves were emitted from an injection-seeded THz wave parametric generator [16–18]. We used a single-mode, linearly polarized, diode-end-pumped microchip Nd:YAG laser (Hamamatsu Photonics Co. Ltd.; λ = 1064 nm; pulse width: 450 ps; pulse energy: 1.2 mJ; repetition rate: 50 Hz) as a pump source that was passively Q-switched using a [110]-cut Cr⁴⁺:YAG saturable absorber [19]. The pump laser was amplified by an optical amplifier (PHLUXi. Inc.) to an energy of 18 mJ/pulse in a double-path configuration. The amplified laser was split, such that the emitter and detector crystals were excited separately. Congruent 5 mol% magnesium oxide-doped lithium niobate (MgO:LiNbO₃) crystals were used for emission and detection of THz waves (size: 50 mm × 6 mm × 5 mm and 65 mm × 6 mm × 5 mm, respectively). These crystals are suitable for emission and detection of THz waves owing to their high figure of merit, transparency over a wide wavelength range (λ = 0.4–5.5 μm), high non-linear coefficient (d₃₃ = 25.2 pmV⁻¹ at λ = 1064 nm), and high laser-induced damage threshold. A continuous-wave tunable external cavity laser diode (ECLD velocity 6300, New Focus, λ = 1068–1075 nm) was used as the injection seed for the idler beam, which has an average power of 500 mW. The seed beam was passed through an achromatic phase-matching optical setup comprising a grating and confocal optical arrangement [20]. The polarization of both the pump and seed beams was parallel to the z-axis of the non-linear crystal. When the angle between these beams was maintained so that thenon-collinear phase-matching condition was met, a single-frequency, coherent THz wave was emitted, the frequency of which could be tuned by changing the wavelength of the seed beam. The THz wave was efficiently coupled in free space with a silicon prism [21].

 

Fig. 1 (a) Three-dimensional (3-D) terahertz (THz) wave computed tomography configuration. Inset shows a THz image taken at the focal point using a THz imager (NEC Corp., IR/V-T0831). (b) Principle of the THz wave heterodyne detection scheme under non-collinear phase-matching conditions.

Download Full Size | PDF

In this imaging system, THz waves were detected by a THz heterodyne detection technique, a novel approach in the THz research field, although well-established in microwave and optical applications. In this detection method, the THz wave was efficiently coupled into the MgO:LiNbO₃ crystal using silicon prism and mixed with pump beam as shown in Fig. 1(b). Because non-collinear phase-matching conditions must be satisfied in this process, the THz signal is incident upon the non-linear crystal at an angle such that the phase-matching condition is met. Moreover, it is important to note that the THz pulse must temporally overlap with the pump laser pulse for the efficient detection of THz waves. To satisfy this condition, the optical path length between the beam splitter and the detector was equal to the total path length between the beam splitter and detector via THz emitter. As a result of this non-linear frequency mixing process, primarily caused by stimulated polariton scattering, an idler wave was generated with a wavelength determined by the wavelengths of the pump beam and the THz wave (ω_idler = ω_pump – ω_THz) [16–18]. The idler wave intensity is proportional to the incident THz wave intensity. As the detection scheme is based on non-collinear phase matching, the pump beam and the idler beam are spatially separated by an angle that is determined by the frequency of the THz wave. Therefore, filters are not needed to separate the pump and idler beams, as often required in alternative collinear phase-matching schemes [22–25], where the idler beam is extracted using spatial optical filters. In this experiment, the idler beam, whose wavelength lies in infrared region, is detected using a pyroelectric detector (Gentec-EO Inc., ED100AUV). As detection methodologies in the infrared frequency region are well-established, and highly sensitive detectors are widely available, a small change in the THz wave intensity can be monitored by measuring the infrared intensity. This detection scheme is one of the main system features that contribute to its high dynamic range.

3. Results

To measure dynamic range of the system, the THz wave intensity was varied using attenuators with transmittances of 30%, 10%, 3%, and 1% (TFA, Tokyo Instruments Inc.), and the idler beam intensity was measured. Figure 2(a) shows the idler beam intensity with respect to the intensity of the THz wave at 1.58 THz. A dynamic range greater than 90 dB was obtained, which is the highest recorded dynamic range using a single detector, in contrast to our previous work in which two different kinds of detector with differing sensitivities were used to obtain a dynamic range of 100 dB [18]. During the detection process, the intensity of the idler beam was observed to vary non-linearly with respect to the THz wave intensity. The variation in idler intensity was within three orders of magnitude, compared with the nine orders of magnitude variation in THz intensity, as shown in Fig. 2(a). This is one of the notable features of the system which enables to use the readily available measurement equipment to measure more than nine orders of magnitude variation in THz intensity. Figure 2(b) shows the THz wave spectra obtained, which is approximately flat and extends from 0.95 THz to 2.7 THz. The noise floor is clearly visible outside of this frequency region. The emission of low-frequency THz waves is restricted due to difficulty in optical alignment because the phase matching angle between the pump beam and seed beam is less than 1°. Above 2.7 THz, emission and detection are limited by the strong THz wave absorption in the LiNbO₃ crystals [26].

 

Fig. 2 (a) Relationship between the idler beam power and the THz wave intensity at 1.58 THz. (b) Semilog plot of intensity spectrum between 0.95 THz to 2.7 THz. The noise floor outside of this frequency range is clearly visible.

Download Full Size | PDF

In the imaging system, the emitted THz wave was collimated by cylindrical lens L1 (Tsurupica, f = 100 mm) and focused at the sample by lens L2 (Tsurupica, f = 100 mm). The beam profile of the THz wave at the focus was measured by an uncooled THz imager (NEC Corp., IR/V-T0831), and it has a diameter of approximately 1 mm, as shown in the inset of Fig. 1(a). In THz tomography, the Rayleigh length is an important parameter that determines the measurable sample size and plays a critical role in image quality. The Rayleigh length is estimated to be 30 mm in the system. The sample under investigation was mounted on a stage that could rotate (θ) and translate the sample in the horizontal (x) and vertical directions (z). The THz wave transmitted through the sample was collimated by lens L3 (f = 100 mm) and was then focused at the non-linear crystal by lens L4 (f = 100 mm) for heterodyne detection, as discussed in the previous section. The various optical effects that introduce noise in THz images have been reported previously [27–29]. To suppress such effects, the refracted beam was collected using lenses (L3 and L4) with a larger diameter (63.5 mm). During the imaging process, 2-D sinograms were recorded by rotating the sample in the θ-axis, and moving horizontally along the x-axis. Filtered backward projection using an inverse Radon transform was used to reconstruct the 2-D image of a sample signal layer [30–32]. The process was repeated for each layer by moving the sample vertically along the z-axis. A 3-D computed tomography image was reconstructed using all of the 2-D images.

To demonstrate the visualization of internal structures using the system, a pencil (outer diameter: 7.4 mm; lead diameter: 2 mm) was used as a test sample to obtain a 3-D image recorded at 1.5 THz. Figure 3(a) shows a photograph of the pencil, and Fig. 2(b) shows a sinogram that was obtained by rotating the sample from 0° to 360° for each linear stageposition. Here, the angular resolution was 6° and the linear stage was moved in 250-μm increments. It took about 18 minutes to collect the data for a single slice of the sample. This sinogram was obtained from the absorbance calculated as α = - ln (I_sam/I_ref), where, I_sam and I_ref are the THz signal transmitted through the sample and air respectively. Here, the red color indicates the high absorbance whereas the blue color indicates the low absorbance. An inverse Radon transform was used to reconstruct the image from the projection data. Figure 3(c) depicts a cross-sectional image of a single layer of the pencil, clearly showing the graphite with high contrast due to its strong absorption of THz waves [33]. The size of the reconstructed image was 15 mm × 15 mm with 3600 pixels. The size of the pencil obtained from the image is 7.25 mm whereas the diameter of the pencil lead is 2.74 mm. This demonstrate the ability of our system to reconstruct the image of a sample with considerable accuracy. A 2-D image was recorded every 2 mm from the top of the sample. Finally, the 2-D images were combined using Voxler (Golden Software Inc.) to reconstruct a 3-D computed tomography image, as shown in Fig. 3(d). An animated view of the internal structure is provided in Visualization 1.

 

Fig. 3 (a) Pencil used as a test sample, mounted in a sample holder: (b) sinogram, (c) image of a single cross-sectional slice of the pencil, and (d) a 3-D computed tomography image of the pencil.

Download Full Size | PDF

To demonstrate the potential application of this system to the non-destructive inspection of industrial products, plastic sheets were fabricated with an intentionally made hole (diameter: 7 mm) as a defect. We used two such sheets and made a structure as shown in Fig. 4(a), which was then placed in a polyethylene package. The inset shows a photograph of the sample. Figure 4(b) shows a sinogram of a single slice of the sample. Here the sample was rotated with an angular resolution of 4° for each linear stage position. The linear stage was stepped in 500-µm increments. In this case, it took around an hour to collect all angle dependent data required to obtain an image of a single slice. Similar to previous sample, the red and blue color indicate high and low absorbance respectively. Figure 4(c) shows a cross-sectional image obtained using an inverse Radon transform. The size of this single slice is 37.5 mm × 37.5 mm with 5626 pixels. The hole is clearly visible with almost no absorption, as depicted by the blue color, whereas the remaining portion and the cover material are clearlyidentified as regions of high absorption. The diameter of the hole estimated from the image is about 7.5 mm. The high THz absorption around the edges and joints of the parts is possibly due to the diffraction of the THz waves. Despite the use of large-diameter lenses L3 and L4, complete suppression of the loss caused by THz-wave diffraction was difficult. The 2-D images obtained were combined with a 1-mm vertical resolution, and the 3-D image shown in Fig. 4(d) was reconstructed. The defect in the sample is clearly visible as shown with the red-dotted circle in Figs. 4(c) and 4(d). Visualization 2 provides an animated view of the structure.

 

Fig. 4 (a) Plastic sample (black) with a hole placed inside a polyethylene cylinder. Inset shows the front view of the sample. (b) Sinogram of a single sample layer. (c) 2-D images identifying the hole as a region of low THz absorption (blue color). (d) Vertical cross-section of the 3-D reconstructed image showing the hole. The inset shows a 3-D image of the plastic sample. The red dotted circles in (c) and (d) identify the hole (defect) in the sample.

Download Full Size | PDF

4. Conclusion

In this paper, we presented a novel THz-CT system with a dynamic range greater than 90 dB, in which the emission and detection of THz waves were achieved by non-linear optical parametric processes and heterodyne detection, respectively. The system was used to obtain a 3-D image of a pencil, in which the internal structures revealed the hidden lead of the pencil. As a potential industrial application of this system, a plastic product with an internal defect was measured and identified in both 2-D and 3-D images. Here, it worthwhile to note that the visualization of internal defects in soft materials, such as plastics, using X-ray imaging can be difficult due to the poor image contrast. We believe that 3-D THz computed tomography systems can be used in non-destructive testing and evaluation of a wide variety of industrial products, such as semiconductors, pharmaceuticals, plastics, and ceramics.