Spectroscopic Imaging with an Ultra-Broadband (1–4 THz) Compact Terahertz Difference-Frequency Generation Source

Abstract

We demonstrate spectroscopic imaging using a compact ultra-broadband terahertz semiconductor source with a high-power, mid-infrared quantum cascade laser. The electrically pumped monolithic source is based on intra-cavity difference-frequency generation and can be designed to achieve an ultra-broadband multi-mode terahertz emission spectrum extending from 1–4 THz without any external optical setup. Spectroscopic imaging was performed with three frequency bands, 2.0 THz, 2.5 THz and 3.0 THz, and as a result, this imaging technique clearly identified three different tablet components (polyethylene, D-histidine and DL-histidine). This method may be highly suitable for quality monitoring of pharmaceutical materials.

1. Introduction

Terahertz (THz) wave (0.3–10 THz) has been used to demonstrate imaging of objects that are opaque at optical frequencies. There are many imaging applications in this frequency region [1,2], including medical [3,4], security screening [5], nondestructive testing for industrial materials [6,7,8,9] and historical arts objects [10,11,12]. THz spectroscopic imaging is expected to be a useful technique for chemical recognition, such as quality control in the pharmaceutical industry, because intermolecular vibrations in some chemicals and organic molecules appears in the THz region. There are many fingerprint frequencies in the 1–4 THz range, and therefore, this frequency range is suitable for chemical recognition.

Spectroscopic imaging has been widely demonstrated by using THz time-domain spectroscopy (THz-TDS) [13]. THz-TDS techniques typically employ a femtosecond laser source for both THz generation and detection. Although THz-TDS techniques can provide spectral information over a broad bandwidth (several THz), they require expensive and bulky laser sources. Additionally, acquiring such images takes a long time because the frequency spectrum at each imaging pixel is obtained by Fourier transformation of the time-domain THz waveform.

On the other hand, a range of narrowband tunable sources, including THz parametric generators [14] and optically pumped gas lasers [15], have been applied to frequency-domain spectroscopic imaging. However, these techniques are based on optical generation of THz waves using lasers, and they require complex and large-footprint laser systems. Furthermore, uni-travelling carrier photodiodes (UTC-PDs) [16] and electronic multiplier chains [17] are tunable THz sources, and spectroscopic imaging with these sources has been demonstrated. However, they typically are restricted to sub-THz frequencies due to the limited output power in the frequency range above 1 THz.

As a powerful platform for frequencies above 2 THz, stand-alone, portable systems for imaging based on THz quantum cascade lasers (THz-QCLs) have been reported, and spectroscopic imaging using a THz-QCL has been demonstrated [18]. However, these imaging experiments using such THz-QCL devices have been performed only below liquid-nitrogen temperatures, T_max < 60 K [19,20]. Considering the above circumstances, therefore, broadband and compact light sources are necessary for realizing THz spectroscopic imaging.

Another compact THz source based on intra-cavity difference-frequency generation (DFG) in a two-color mid-infrared QCL integrated with two different laser active regions [21] uses mid-infrared active regions engineered to exhibit a large optical nonlinearity χ⁽²⁾ for an efficient THz DFG process. Presently, these are the only electrically pumped monolithic semiconductor and room temperature operable sources at frequencies between 0.6 THz and 6 THz [22,23,24,25]. The emission efficiency of a compact THz source has been improved by adopting a Cherenkov scheme [26,27]. In addition, recent developments in nonlinear QCLs with a dual-upper-state (DAU) design approach [28,29] that has a wide gain spectrum resulted in a drastic improvement in the nonlinear susceptibility of the active region, for efficient THz generation, as well as device performance. Applying the improved nonlinear susceptibility, this active region concept expanded the frequency range down to sub-THz, for the first time [25].

These features of the compact THz DFG sources provide a possibility for imaging applications, and, in fact, we have presented high-quality non-destructive imaging using a THz DFG source [30]. However, despite the possibility of spectroscopic imaging, in our previous work, the distinct characteristics of the ultra-broadband emission spectrum were not fully exploited. In the case of a broadband THz DFG source, frequency tuning and calibration are unnecessary because frequency selectivity could be performed by changing the frequency bandpass filter (BPF) in front of the detector. In this work, we demonstrated the feasibility of spectroscopic imaging with a compact THz DFG source, using multiple frequency bands (2.0 THz, 2.5 THz and 3.0 THz).

2. Experimental Setup

A THz DFG source based on a QCL was driven at 240 K, which is a temperature that can be achieved with a TEC (thermoelectric cooler) to obtain higher THz output power, which can be achieved by increasing the mid-infrared pump power product when the device is cooled. In this experiment, the device was cooled to 240 K using liquid nitrogen cooled cryostat. This device exhibited a broadband THz emission spectrum between about 1.5 THz and about 3.3 THz [31]. Since the THz beam from the DFG source was TM polarized, the THz beam was horizontally polarized [32]. We performed spectroscopic imaging using commercially available frequency BPFs. The DFG source was driven by a current source at 1.7 A with a 2% duty cycle and a 100 kHz repetition rate. The current-voltage-THz output power characteristics at 240 K and 293 K is shown in Figure 1a. We confirmed a THz peak output power of ~0.3 mW at 240 K in front of the device, though the THz power was attenuated in the imaging system. The attenuated THz power at the detector due to transmission loss of the lenses was estimated to be <0.1 mW. The THz spectra obtained at 240 K and 293 K shown in Figure 1b exhibited broadband multi-mode THz emission spanning over more than 2.5 octaves, from 1.0 to 3.5 THz. More importantly, we demonstrated a bandwidth of ~1.9 THz for a frequency QCL comb, in which the ultra-broadband emission spectra were observed at 20 K. Thus, the broadband THz DFG technology based on a QCL, which can operate at room temperature, may be important for spectroscopic applications.

Figure 2 shows transmission spectra of BPFs having different central transmission frequencies, namely, 2.0 THz, 2.5 THz and 3.0 THz (FB19M150, FB19M120 and FB19M100, respectively, Thorlabs, Inc., Newton, NJ, USA), obtained by a THz time-domain spectroscopy system. The average frequency bandwidth was about 0.4 THz (FWHM). To obtain the far-field profile of the devices, we used two-axis motorized XY translation stages. A Golay-cell detector (active area diameter of 6 mm) was mounted on the translation stages at a distance of about 59 mm from the DFG source. We obtained far-field profile along the emission plane. Two-dimensional (2D) scans (62.5 mm × 62.5 mm) were performed with intervals of 1.25 mm. To map the far-field profile at each frequency, we mounted one of the BPFs onto the detector. Total power transmittance of BPF was estimated to be about 21%, 27% and 38% (2.0 THz, 2.5 THz and 3.0 THz, respectively).

Figure 3 shows a schematic diagram of our transmission imaging system. The THz radiation from our device was reflected and collimated with an off-axis parabolic mirror (f = 50 mm, 3 inch diameter) and was focused onto a test object by using an aspheric polymer lens (f = 40 mm, 45 mm clear aperture; Tsurupica^®, Pax Co., Ltd., Sendai, Japan). Then, the THz beam transmitted through the object was collimated with another aspheric polymer lens (Tsurupica^®, f = 40 mm, 45 mm clear aperture) and was collected and coupled into the Golay-cell detector using another off-axis parabolic mirror (f = 100 mm, 3 inch diameter). THz image acquisition was demonstrated by the raster-scanning method. The object was mounted on a computer-controlled two-axis (XY) translation stage. To obtain spectroscopic images at a specific frequency, a BPF of that frequency was mounted onto the detector. Tsurupica^® is made from cyclic olefin copolymer, which has negligible dispersion in refractive index and absorption coefficient across the frequency range of 2 to 15 THz [33].

3. Results and Discussion

In this work, we obtained a false-color RGB composite image from images acquired with BPFs at three frequencies: 2.0 THz, 2.5 THz, and 3.0 THz. We evaluated the imaging characteristics (far-field profile and imaging result for frequency-independent sample) when using multiple frequency BPFs with central transmission frequencies of 2.0 THz, 2.5 THz and 3.0 THz.

Figure 4a shows the far-field profile of the THz beam from the DFG source without any BPF. Figure 4b–d show far-field profiles of the THz beam from the DFG source with BPFs (central transmission frequencies of 2.0 THz, 2.5 THz and 3.0 THz, respectively). As shown in Figure 4a–d, no speckle pattern or inhomogeneity was observed on the image at all frequencies. In addition, Figure 5a–d show profiles along the vertical broken line and horizontal broken line in Figure 4a–d and their Gaussian fits. From the vertical and horizontal profiles (Figure 5a–d), we estimated beam FWHM values along the slow axis (y-axis) and along the fast axis (x-axis).

Table 1. FWHM of far-field profile of DFG source and the corresponding divergence angle.

Frequency	FWHM of Far-Field Pattern (mm)		Divergence Angle (deg.)
	Slow Axis	Fast Axis	Slow Axis	Fast Axis
Without BPF	29.1	18.0	27.7	17.3
2.0 THz	22.7	18.1	21.8	17.4
2.5 THz	22.5	17.4	21.6	16.8
3.0 THz	30.1	16.4	28.6	15.8

shows the FWHM values of the far-field patterns and the corresponding divergence angles. The far-field profiles at all frequencies were single-lobed Gaussian-like profiles. In addition, the vertical beam waist at 3.0 THz was broad compared to the other frequencies. Furthermore, we found that the center of the THz beam was shifted along the x-axis as the frequency was changed: x = 33.3 mm (at 2.0 THz), x = 32.4 mm (at 2.5 THz) and x = 30.5 mm (at 3.0 THz). Due to the difference in position between the center of the beam at 2.0 THz, and that at 3.0 THz was 2.8 mm, the difference in emission angle for 2.0 THz and 3.0 THz was estimated to be 2.7°. We consider that the difference in emission angle was due to the angular phase matching (Cherenkov emission scheme); emission angle depends on the frequency of the converted wave.

We demonstrated imaging of a metal test object having bar pattern slits, at each frequency band (2.0 THz, 2.5 THz and 3.0 THz). We fabricated a test object presented in Figure 6. The upper part of the test object had 1 mm-wide slits, and the lower part had 0.5 mm-wide slits. THz images were obtained with scanning steps 0.2 mm. Figure 7a shows a THz image obtained using the DFG source alone, without a BPF. The imaging results at the individual frequencies (2.0 THz, 2.5 THz and 3.0 THz) are shown in Figure 7b–d. At all the frequencies, both the vertical and horizontal slits could be resolved. By using the knife edge method, the beam diameter at the focusing point in the transmission imaging system was found to be <0.7 mm at all frequencies, with negligible astigmatic difference at each frequency. Although the spatial resolution should be about <0.35 mm considering the beam spot size, we considered that the actual spatial resolution was about 0.4 mm because the image was obtained with a 0.2 mm scan pitch. Therefore, the 0.5 mm-wide slits could be resolved.

Figure 8a–c depict the line scans extracted from the image along the horizontal solid lines, as shown in Figure 7 (1-1′, 3-3′, 5-5′). Figure 8d–f depict the line scans extracted from the image along the vertical solid lines, as shown in Figure 7 (2-2′, 4-4′, 6-6′). The modulation depth M is calculated according to

$$M=\frac{I_{Max}-I_{Min}}{I_{Max}+I_{Min}}$$

(1)

with I_Max and I_Min denoting the maximum and minimum intensities of the image profile. The modulation depths M for 0.5 mm width at all frequencies were found to be >0.25.

Next, we performed spectroscopic imaging. First, to demonstrate false-color RGB composite imaging, we obtained a spectroscopic image of a standard test object having frequency-dependency. Figure 9a schematically depicts a test object having three different frequency-dependencies: BPF films (center frequencies of 2.0 THz, 2.5 THz, and 3.0 THz) were formed on three through-holes having diameters of 3 mm. We demonstrated imaging of the test object using our transmission imaging system (50 pixels × 75 pixels, 0.2 mm steps). A THz image of the standard test object obtained without a BPF is shown in Figure 9b. In this measurement, acquisition for the spectroscopic image (Figure 9b) took ~2 h because measurement time was ~2 s for each pixel (50 pixels × 75 pixels). Golay-cell detector is a commercially available, high-sensitivity and uncooled detector, however, the response time is slow because of thermal detection. As shown in Figure 9b, without a BPF, all frequency components were detected. Figure 10a–c show images of the test object obtained by the detector with a BPF (center frequencies of 2.0 THz, 2.5 THz, and 3.0 THz, respectively: red channel, green channel and blue channel). As shown in Figure 10a–c, with a BPF, another frequency component appeared on the images obtained in each frequency band. If we use BPFs with narrow bandwidths, like double-layer BPFs [34], each frequency component would be distinguished clearly. Then, we obtained a false-color image by combining the R, G, and B (imaging results at 2.0 THz, 2.5 THz, and 3.0 THz, respectively), as shown in Figure 10d. As shown in Figure 10d, the bottom-left zone with the 2.0 THz component appears as red, the bottom-right zone with the 2.5 THz component appears as green, and the top zone with the 3.0 THz component appears as blue.

We made three kinds of tablets with thicknesses of 0.2 mm: polyethylene (PE), 10 wt% D-histidine and 10 wt% DL-histidine. The D-histidine tablet and DL-histidine tablet were diluted in polyethylene. Figure 11a shows a photograph of tablet test objects; each tablet was placed on three through-holes having diameters of 3 mm. D-histidine has a typical absorption peak at about 2.8 THz, and DL-histidine has a typical absorption peak at 2.0 THz [35].

As shown in Figure 12, PE appears as white since it has almost equal transparency in all channels. D-histidine appears as red-white because it has high absorption in the green and blue channels. Additionally, DL-histidine appears as blue-green because it has absorption in the red channel. Thus, from the false-color RGB image obtained with our spectroscopic imaging system, we could distinguish a crystalline sample and its racemate sample.

4. Conclusions

In this work, we demonstrated, for the first time, spectroscopic imaging with an ultra-broadband THz DFG source based on a QCL. Spectroscopic imaging was realized by using BPFs with central transmission frequencies of 2.0 THz, 2.5 THz and 3.0 THz; spectroscopic imaging in a frequency band between 2 THz and 3 THz was demonstrated. The spatial resolution showed negligible astigmatic difference at each frequency. We performed false-color RGB imaging by using the spectroscopic imaging system with a simple procedure (calibration-free). From the false-color RGB image, we could identify D-histidine and DL-histidine. Thus, we could distinguish crystalline material and its racemate. Spectral imaging based on our broadband THz light source is an attractive technology for the pharmaceutical industry and biomedical applications. Recently, real-time spectroscopic imaging using a diffraction grating and a THz 2D camera have been reported [4]. DFG sources could also be applied to such a spectroscopic imaging system. This would enable rapid spectroscopic imaging. Thus, DFG sources are useful light sources for spectroscopic imaging.