Recently terahertz imaging using two-dimensional E-O sampling has attracted much interest because it can acquire real-time terahertz images unlike a conventional raster scan method. We are applying this technique to the non-destructive measurement of opaque materials in a visible range. We acquired 10-fps consecutive terahertz transmission images: dripping water in a plastic pipe and metal included in a piece of gum. Since the obtained images were confirmed to be proportional to the electric field of the terahertz waves, the images in the present paper are useful for quantitative analysis. We also showed the signal-to-noise ratio of the terahertz images.
© 2007 Optical Society of America
Terahertz (THz) electro-magnetic waves (0.1–10 THz) have various features, such as sufficient transmission, non-invasion, and little scattering. Moreover, in the THz region peaks of intermolecular or phonon vibrations exist. Therefore, studies of THz spectroscopy  and imaging [2, 3] for basic and industry applications have already been reported.
Since a THz wave poses no threat of contamination or invasion, as opposed to X-rays, THz imaging is useful for non-destruction and non-contact tests; THz imaging, which can also acquire two-dimensional (2D) transmission images, is generally performed using the raster scan method  that obtains THz images by moving the sample on the focus of the THz wave. However, this way has a disadvantage: it requires a long time to acquire the THz image due to sample movement.
Accordingly, real-time THz imaging based on the following principle has been developed . An expanded THz wave illuminates a large area of a sample and penetrates it. As a result, the THz wave has a 2D transmission image of the sample. The image is transcribed onto an expanded probe pulse in a nonlinear optical crystal by the Pockels effect. Since the probe pulse is captured on camera, we can obtain a 2D transmission image of the sample without movement. In real-time THz imaging measurement, the THz image is acquired in the following steps: 1) a probe pulse image without THz radiation as a background is measured beforehand; 2) a THz image is obtained by subtracting the background from a probe pulse with THz radiation.
However, the background image becomes different by the laser fluctuations with the passage of time. Due to the above phenomena, the signal-to-noise ratio (SNR) of the image by real-time THz images was worse than the raster scan method. Therefore, dynamic subtraction with two-dimensional electro-optic sampling (2DEOS) was suggested by Z. Jiang et al. , who used one frame with THz radiation and the next frame without THz radiation and acquired THz images by subtracting one frame from the next one. In this way, they reduced the fluencies of the laser fluctuations and acquired THz images of sufficient SNR using the latest background on the above calculation.
However under the present conditions, real-time 2D transmission images of an opaque object in a visible range have not been reported, despite the possibility of being performed by a real-time THz imaging method. In this paper, we demonstrate real-time 2D transmission imaging of opaque objects by dynamic subtraction with 2DEOS. Furthermore, we discuss SNR and the linearity of THz images.
2. Experimental Setup
Figure 1 shows a schematic setup for real-time THz imaging that employs a regenerative amplifier Ti:Sapphire laser system with a repetition rate of 1 kHz, a pulse duration of 30 fs, an output power of 1 mJ, and a center wavelength of 800 nm.
The femtosecond laser pulse is divided into a pump pulse and a probe pulse by a beam splitter (BS). The pump pulse is incident on a 1-mm-thick (110) ZnTe crystal as an emitter through a time-delay stage. A THz pulse is generated by optical rectification in the emitter. Since the pump pulse is modulated at 500 Hz by an optical chopper synchronized to the laser, a generated THz pulse is also modulated at 500 Hz that illuminates a sample. The THz pulse, which penetrated the sample, is imaged onto a (110) ZnTe crystal (1-mm thick) as a receiver by two plastic lenses (f = 100 mm).
This imaging formation that uses two lenses can rectify distortion. The probe pulse copropagates with the THz pulse after reflection on a pellicle beam splitter and is incident on the receiver. If the THz and probe pulses are simultaneously incident on the receiver, the polarization of the probe pulse is rotated by the Pockels effect.
The rotation angle of the polarization of the probe pulse is proportional to the electric field of the THz pulse, whose intensity is converted to the intensity of the probe pulse by analyzer (A). Therefore, the THz image is transcribed onto a probe pulse that is imaged onto a CMOS camera (Intelligent Vision System: IVS, Hamamatsu C8210-50, 232×232 pixel, 1000 fps)  that synchronized with the 1-kHz repetition rate of the laser.
In turns, the CMOS camera captures image 1, which is a probe pulse with a THz pulse, and image 2, which is a probe pulse without a THz image. Next, the CMOS camera outputs an image that subtracted image 2 from image 1. As a result, we can obtain THz images in real time .
3.1 Intensity linearity of THz images
Since the intensity linearity of THz images (i.e., THz images with quantities) has not been examined in detail, we evaluated linearity.
First, we created three kinds of attenuation filters for THz pulses made from metal and controlled their attenuation ratios by changing the open area ratio. Therefore, we decrease the amplitude of the THz pulse using these filters without time delay.
Second, we corrected the attenuation ratios of these filters by measuring the temporal waveforms of the THz pulse with and without filters using the general THz-TDS system . We obtained the corrected values of the attenuation ratios.
Finally, we acquired the THz images of these filters by real-time THz imaging and calculated the measurement values of the attenuation factors using the average amplitude of the center area (20×20 pixels) of these images. Figure 2 shows the relationship between the attenuation ratio of the measurement value and the corrected values.
As a result, since the graph of Fig. 2 is linear, we confirmed that the THz images have good intensity linearity because the rotation angle of the polarization of the probe pulse by THz pulse in the receiver is very small (about 0.5°).
3.2 Deletion of a ghost image
Figure 3 shows the probe pulse image without the THz pulse. The area in the circle of the dashed line in Fig. 3 is a ghost image, caused by the reflection of the probe pulse in the following process. First, the probe pulse is reflected at the surface of the ZnTe crystal (receiver). Then, it is reflected at the rear face of the plastic lens. Finally, it is returned to the CMOS camera and becomes a ghost image.
This image reduces the SNR of the THz image. Thus, we removed it by pasting a black polyethylene film at the rear surface of the lens (Fig. 4) to reduce the probe pulse reflection; it has good transmittance for the THz pulse.
3.3 SNR of THz image
Figure 5 shows a photo of a sample: Al tape on a plastic substrate cut in the shape of an “H”. Although the THz pulse can penetrate the plastic, it cannot penetrate the Al tape. Thus, we can acquire an “H” - shaped THz image, as shown in Fig. 6, by the setup (Fig. 1).
Figure 7 shows a profile extracted at 116 pixels of the vertical axis in Fig. 6. The THz signal in the region from 100 to 180 pixels in Fig. 7 has vibration that is mainly caused by the distribution of the electro-optical constant in the ZnTe crystal.
Thus, we defined SNR as the ratio of the average intensity in the region of the THz signal to the standard deviation in the same region. We extracted 11 profiles between 97 and 107 pixels of the vertical axis and calculated the SNR of each profile. As a result, the average SNR of the 10 profiles is estimated to be about 11. This value is sufficient for applying the THz image.
3.4 THz movie
We acquired two movies of THz transmittance images (THz movies) at 10 fps (i.e., we integrated 50 frames) to demonstrate applications of non-destructive tests.
3.4.1. Dripping water in a pipe
Figure 8 is an optical image of a sample. We dripped water into a pipe whose inside is opaque in a visible range and measured the situation with a real-time THz imaging setup. As a result, we obtained a THz movie of the situation shown in Fig. 9. Here, the THz pulse is green. Water is black in Fig. 9 because it has strong absorption in the THz region, and a time delay occurred. We verified the situation in which water was dripping in a pipe.
3.4.2 Staple in a piece of gum
We believe applications of non-destructive food verifications by THz imaging are effective. Therefore, we measured a piece of gum with good transmittance in the THz region.
Figure 10 shows a sample of a piece of gum stuck with a staple. We moved the gum to a translation stage vertically to the propagation of the THz pulse and acquired a 10-fps THz movie (Fig. 11) of the state in which the gum was moving. The staple is black in Fig. 11 because it prevents the THz pulse. In Fig. 11, we confirmed the situation in which the staple was moving.
Our THz imaging method has two potential weaknesses. One, the THz image may disappear if the sample is changed. The reason is that the time delay of the THz pulse is changed by the different refractive index and the thickness of the new sample. Hence, for the new sample, we have to set a correct time-delay position where the intensity of the THz image is maximum. Our THz imaging method can acquire images by 500 fps without integration. Accordingly, we can easily adjust the correct time-delay position in real time, while moving the delay stage and simultaneously acquiring the THz image.
Another weakness is that the spatial resolution of the THz image is relatively low. For example, the resolution of our THz image is estimated at 1 mm because the THz pulse has broad spectrum including low frequency. This problem can be solved by metal hole array (MHA) . Since MHA works as a highpass filter in the terahertz region, a THz image with sufficient spatial resolution will be obtained.
We acquired two 10-fps THz movies of states in which water was dripping into a pipe and a staple in gum was moving. Therefore, we believe this THz imaging technique is very useful to non-destructive and non-contact tests for food, buildings, security, and so on.
Moreover, we confirmed that THz imaging has good intensity linearity because the modulation of the probe pulse by the THz pulse was small. Accordingly, quantitative evaluation using THz images, for example, absorption, was attained. We also defined the SNR of the THz image to be about 11.
The authors are grateful to T. Hiruma and Y. Suzuki for their support and S. Aoshima and M. Fujimoto for fruitful discussions.
1. D. Grischkowsky, S. Keiding, M. Exter, and Ch. Fattinger, “Far-infrared time-domain spectroscopy with terahertz beams of dielectrics and semiconductors,” J. Opt. Soc. Am. B 7, 2006–2015 (1990). [CrossRef]
3. D. Creeden, J. C. McCarthy, P. A. Ketteridge, P. G. Schunemann, T. Southward, J. J. Komiak, and E. P. Chicklis, “Compact, high average power, fiber-pumped terahertz source for active real-time imaging of concealed objectcs,” Opt. Express 15, 6478–6483 (2007). [CrossRef] [PubMed]
4. Q. Wu, T. D. Hewitt, and X.-C. Zhang, “Two-dimensional electro-optic imaging of THz beams,” Appl. Phys. Lett. 69, 1026–1028 (1996). [CrossRef]
5. Z. Jiang, X. G. Xu, and X.-C. Zhang, “Improvement of terahertz imaging with a dynamic subtraction technique,” Appl. Opt. 39, 2982–2987 (2000). [CrossRef]
6. Y. Sugiyama, M. Takumi, H. Toyoda, N. Mukozaka, A. Ihori, T. Kurashina, Y. Nakamura, T. Tonbe, and S. Mizuno, “A high-speed CMOS image sensor with profile data acquiring function,” IEEE J. Solid-State Circuits, 40, 2816–2823 (2005). [CrossRef]
7. F. Miyamaru, T. Yonera, M. Tani, and M. Hangyo, “Terahertz two-dimensional electrooptic sampling using high speed complementary metal-oxide semiconductor camera,” Jpn. J. Appl. Phys. 43, L489–L491 (2004). [CrossRef]
8. F. Miyamaru and M. Hangyo, “Finite size effect of transmission property for metal hole arrays in subterahertz region,” Appl. Phys. Lett. 84, 2742–2744 (2004). [CrossRef]