APPLIED PHYSICS LETTERS
VOLUME 80, NUMBER 16
22 APRIL 2002
Continuous-wave all-optoelectronic terahertz imaging Karsten J. Siebert,a) Holger Quast, Rainer Leonhardt, Torsten Lo¨ffler, Mark Thomson, Tobias Bauer, and Hartmut G. Roskos Physikalisches Institut, Johann Wolfgang Goethe-Universita¨t, Robert-Mayer-Str. 2-4, D-60054 Frankfurt (M), Germany
Stephanie Czasch Institut fu¨r Veterina¨r-Pathologie, Frankfurter Str. 96, D-35392 Gießen, University of Gießen, Germany
共Received 20 November 2001; accepted for publication 22 February 2002兲 We present an all-optoelectronic THz imaging system based on photomixing of two continuous-wave laser beams using photoconductive antennas. For a specific biological sample, we compare continuous-wave THz imaging and pulsed THz imaging at 1 THz with respect to data-acquisition time and signal-to-noise ratio, and discuss image formation from both amplitude and phase data. In addition, we introduce the application of hyperboloidal lenses which allow tighter focusing and a corresponding improvement in spatial resolution compared to off-axis paraboloidal mirrors. © 2002 American Institute of Physics. 关DOI: 10.1063/1.1469679兴
Since its demonstration by Hu and Nuss1 in 1995, optoelectronic THz imaging has become a rapidly expanding field of research.2 The potential for applications such as in package inspection and in the biomedical field was demonstrated.3–5 The technology has reached a maturity which allows the adaption of more sophisticated imaging techniques known from the optical region of the spectrum, e.g., near-field6,7 and dark-field imaging.8 Typical optoelectronic THz measurement systems are based on expensive femtosecond laser systems and have problems generating narrow-linewidth spectral data. For these reasons, there is an interest in finding conceptual alternatives without giving up the advantages of the coherent detection scheme, namely the capability to measure both the amplitude and phase of THz signals over a large frequency range 共several THz兲 with high signal-to-noise ratio 共SNR兲. Photomixing of two continuous-wave 共cw兲 laser beams holds much promise here because 共i兲 the THz frequency and linewidth are pre-selected by the choice of the laser wavelengths and linewidths, and 共ii兲 compact and inexpensive semiconductor lasers are gradually becoming accessible for such applications. The first implementation of an all-optoelectronic THz measurement system was demonstrated in 1998 by Verghese et al.9 who employed the optical beat signal from two cw Ti:sapphire lasers operating at different wavelengths to generate and detect the THz signal using photoconductive antennas. Mixing of the radiation of semiconductor lasers has been demonstrated10,11 and the THz output power of photoconductive antennas on low-temperature grown GaAs 共LTGaAs兲 has been improved by optimizing both the antenna design12–14 and the LT-GaAs material properties.15 Optoelectronic cw THz imaging, however, has not been demonstrated and studied yet. In this letter, we present an implementation of a cw THz optoelectronic imaging system and discuss its properties in comparison with those of a femtosecond imaging system. Figure 1 shows the layout of our system. Two optical a兲
Electronic mail:
[email protected]
waves, with independently tunable wavelengths centered around 800 nm, are generated in a dual-color cw Ti:sapphire laser16 consisting of two unidirectional, single-longitudinalmode ring cavities sharing a single Ti:sapphire gain medium which is pumped by two beams from a 5 W Coherent VERDI all-solid-state laser. The emitted laser beams are spatially combined with a 50/50 beam-splitter cube yielding two beams which are intensity modulated at the difference frequency 共0–10 THz兲 of the optical waves. The subsequent optical system is similar to that of femtosecond THz transmission measurement systems. One optical beam is guided via a computer-controlled optical delay line to the emitter antenna, the other is used to gate the receiver antenna. We employ two H-shaped photomixer antennas both with a 50m-long dipole and 5⫻10 m2 photoconductive gap on LTGaAs grown at 270 °C 共emitter兲 and 200 °C 共detector兲, respectively, exhibiting measured carrier trapping times of 1.2 ps 共emitter兲 and 0.35 ps 共detector兲. The variation of the growth temperatures allows one to achieve higher output power from the emitter and broader detection bandwidth at the detector. Each photomixer is illuminated at 100 mW of optical power. The emitter is biased with a 25-kHz-square wave signal between ⫾12.5 V. This modulation allows lock-in detection of the THz signal while avoiding detrimen-
FIG. 1. Setup of the all-optoelectronic cw THz imaging system.
0003-6951/2002/80(16)/3003/3/$19.00 3003 © 2002 American Institute of Physics Downloaded 16 Apr 2002 to 128.113.8.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
3004
Appl. Phys. Lett., Vol. 80, No. 16, 22 April 2002
tal feedback into the two-color laser source which is found to occur when the optical beam is mechanically chopped. The lock-in time constant is set to 20 ms. The THz radiation is coupled out of the emitter via a hyperhemispherical Si substrate lens of 2 mm diameter, collimated by an off-axis paraboloidal mirror and focused by a plano-hyperboloidal lens. It is then passed through the sample and guided onto the detector antenna with the reverse sequence of optics. The signal-to-noise ratio 共SNR兲 obtained at 1 THz without a sample in the beam path is 100:1. The SNR is defined in terms of the mean amplitude versus its standard deviation. The plano-hyperboloidal lenses were manufactured in house from high-density polyethylene and allow for very short focal lengths while producing aberration-free foci. Moreover, such lenses have the additional advantages that they are relatively thin even for large apertures, and can be positioned easily because their principal plane is always at the lens apex. The lenses used have a focal length of 2.0 cm while having a free aperture of 6.0 cm and a thickness of 2.0 cm at the apex. For THz imaging the sample is mounted on a computercontrolled x–y translation stage and can be moved continuously along a meandering path through the focus of the THz beam. The relative phase of the THz signal is varied by translating the optical delay line. In order to obtain the shortest possible data acquisition time for the imaging system, we take advantage of the extremely long coherence length of the two optical single-mode frequencies. During the spatial scan over a row 共x axis兲 of the object, the optical delay line is moved with constant velocity without changing its direction. For each pixel of the image, the phase of the detected THz wave is varied over two THz periods. Twelve temporal data points per period are sampled and employed for the evaluation of both amplitude and phase of the sinusoidal THz field. At the end of each row of pixels both the optical time delay and the object’s translation stage change directions simultaneously while the object is moved to a new y position. For each object, a reference scan of the time delay is taken without the sample or at a transparent point on the sample covering the same optical pathlength as during a scan of a row of pixels. It is taken both in forward and backward direction to compensate for any differences in the phase of the signal due to the change of the scanning direction. In order to determine the focal diameter of the cw THz imaging system and to determine the blurring of the image due to the continuous-scanning technique, the edge of a 30m-thick steel foil was moved across the focus of the THz beam. The experiment was carried out at a frequency of 1.1 THz. In Fig. 2, the relative transmitted THz power is plotted versus the spatial position of the edge. The gray curve denotes a stepped translation of the edge, whereas the black curve shows the data for the edge being moved across the beam with constant velocity. In both cases, the pixel period was 10 m. The two curves are in close agreement with each other which shows that the continuous scanning does not blur the edge sharpness of the image. The 10%–90% rise in the transmitted power occurs over 320 m for both scanning techniques, which provides a measure for the spatial resolution of the imaging system. It is worth noting that the signal
Siebert et al.
FIG. 2. Normalized power transmission dring a scan of a steel edge across the THz focus. Note that the signal fluctuations are found to be mainly due to Fresnel diffraction.
fluctuations seen in the data stem mainly from interference effects and only to a minor amount from noise. In the following, we demonstrate cw imaging with a more complex sample and compare the results with images taken with a pulsed system based on a Ti:sapphire amplifier laser operating at 1 kHz repetition rate.8 The sample was a formalin-fixed, dehydrated, and wax-mounted slice through a canary’s head 关see photograph in Fig. 3共a兲兴. The size of the sample was 32 mm⫻24 mm⫻3 mm. Figure 3 shows logarithmic power transmission images taken both with the cw system at 1 THz 关Fig. 3共b兲兴 and the pulsed system at 1 THz 关Fig. 3共c兲兴. The images consist of 3072 pixels each, with a spacing between neighboring pixels of 0.5 mm. To compare the image quality, the relative power transmission for line scans at fixed y positions is plotted in Fig. 3共d兲. The two solid curves correspond to scans of lines at the same position of the object 共line 16 in the cw image and line 18 in the pulse image as indicated in the plots兲. The contrast achieved with the cw system compares well with that obtained using the pulsed system. Information about the noise in the data is obtained from the two dashed curves which were measured through pure wax 共line 46 in the cw image and line 3 in the pulse image兲. For both systems, the ratio of mean relative power transmission to the standard deviation is 14:1 共corresponding to 28:1 for the amplitude兲. In addition to the amplitude information, the relative phase of the cw wave at each pixel can be exploited for image formation as shown in Fig. 4. Nearly all features observable in the power picture can be seen here as well. For thin samples where the optical thickness varies by less than one wavelength as in our case, or for samples with a sufficiently homogeneous optical thickness 共i.e., without changes larger than one wavelength兲 and a constant geometrical thickness this information is equivalent to the information obtained by time-of-flight measurements with pulsed systems. For samples with larger variations in the optical thickness, the modulo-2 ambiguity has to be taken into account. The time it takes for single-scan image acquisition differs significantly between the cw and the pulsed systems.
Downloaded 16 Apr 2002 to 128.113.8.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp
Siebert et al.
Appl. Phys. Lett., Vol. 80, No. 16, 22 April 2002
3005
image was obtained during a time period of 14 h, 15 min corresponding to a total measurement time of 8.5 s per pixel. Comparing the data acquisition times, it has to be taken into account that the pulsed system was not optimized for short data acquisition times and that spectroscopic information was obtained simultaneously over a frequency range from 0.1 to 3 THz with a frequency resolution of 312 GHz, whereas the cw system only addresses one single frequency 共albeit with a much smaller linewidth ⬍2 GHz兲. A fairer comparison would be one with high-repetition-rate pulsed systems where the record for the data acquisition time is at 10 ms per pixel.17 Given the present performance of our cw system, we have little doubt that the same acquisition time is achievable when the lock-in detection scheme is replaced by digital signal processing approaches like those of the stateof-the-art pulsed systems. In summary, we have demonstrated a cw alloptoelectronic THz imaging system at 1 THz. The signal-tonoise ratio is better than 100:1, and the spatial resolution is about one wavelength. With the same antenna the system is suited to record images in the region between 0.2 and 1.5 THz. For image formation, both amplitude and phase information can be exploited. Concerning image quality, our system compares well with a state-of-the-art imaging system based on a 1 kHz laser while having a better frequency resolution. The authors thank G. Strasser, TU Wien, for providing the LT-GaAs and Volkmar Hock, U Wu¨rzburg, for the Ti/Au metallization of the antennas. The authors acknowledge financial support from the Deutsche Forschungsgemeinschaft and the EU project TERAVISION within the IST framework. B. B. Hu and M. C. Nuss, Opt. Lett. 20, 1716 共1995兲. D. M. Mittleman, M. Gupta, R. Neelamani, R. G. Baraniuk, J. V. Rudd, and M. Koch, Appl. Phys. B: Photophys. Laser Chem. 68, 1085 共1999兲. 3 R. M. Woodward, B. Cole, V. P. Wallace, D. D. Arnone, R. Pye, E. H. Linfield, M. Pepper, and A. G. Davies, Proceedings of CLEO, 2001, p. 329. 4 D. Arnone, Optics Lasers Europe, March 2000, p. 13. 5 D. M. Mittleman, S. Hunsche, L. Boivin, and M. C. Nuss, Opt. Lett. 22, 904 共1997兲. 6 S. Hunsche, M. Koch, I. Brener, and M. C. Nuss, Opt. Commun. 150, 22 共1998兲. 7 K. Wynne and D. A. Jaroszynski, Opt. Lett. 24, 25 共1999兲; 150, 22 共1998兲. 8 T. Lo¨ffler, T. Bauer, K. Siebert, H. G. Roskos, A. Fitzgerald, and S. Czasch, Opt. Express 9, 616 共2001兲. 9 S. Verghese, K. A. McIntosh, S. Calawa, W. F. Dinatale, E. K. Duerr, and K. A. Molvar, Appl. Phys. Lett. 73, 3824 共1998兲. 10 K. A. McIntosh, E. R. Brown, K. B. Nichols, O. B. McMahon, W. F. DiNatale, and T. M. Lyszczarz, Appl. Phys. Lett. 67, 3844 共1995兲. 11 P. Gu, F. Chang, M. Tani, K. Sakai, and C.-L. Pan, Jpn. J. Appl. Phys., Part 2 38, L1246 共1999兲. 12 S. Matsuura, G. A. Blake, R. A. Wyss, J. C. Pearson, C. Kadow, A. W. Jackson, and A. C. Gossard, Appl. Phys. Lett. 74, 2872 共1999兲. 13 E. K. Duerr, K. A. McIntosh, and S. Verghese, Proceedings of CLEO, 2000, p. 382. 14 S. M. Duffy, S. Verghese, K. A. McIntosh, A. Jackson, A. C. Gossard, and S. Matsuura, IEEE J. Micr. Theory Technol. 49, 1032 共2001兲. 15 A. W. Jackson, C. Kadow, A. C. Gossard, S. Matsuura, G. Blake, E. Duerr, and S. Verghese, Proceedings of the Symposium on Non-Stoichiometric III-V Compounds in Erlangen, 1999, p. 19. 16 F. Siebe, K. Siebert, R. Leonhardt, and H. G. Roskos, IEEE J. Quantum Electron. 35, 1731 共1999兲. 17 D. M. Mittleman, R. H. Jacobsen, and M. C. Nuss, IEEE J. Sel. Top. Quantum Electron. 2, 679 共1996兲. 1 2
FIG. 3. 共a兲 Photograph of the sample, a wax-mounted thin-cut canary’s head; object size: 32 mm⫻24 mm⫻3 mm. 共b兲 Cw THz power transmission image at 1 THz. 共c兲 Corresponding pulsed THz image at 1 THz. 共d兲 Relative transmission measured over a row of pixels with the cw THz system 共black curves兲 and the pulsed system 共gray curves兲, respectively. The difference in the relative power transmission result from normalization to different power values.
The cw scan took 11 min corresponding to an overall measurement time of 200 ms per pixel plus a short time at the end of each row to switch to the next one and the 30 s required for the reference scan. In comparison, the pulsed
FIG. 4. Cw THz image obtained from the phase information at 1 THz.
Downloaded 16 Apr 2002 to 128.113.8.139. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/aplo/aplcr.jsp