The generation and characterization of narrowband THz pulses by means of chirped pulse difference frequency generation in Auston-switch type photoconductive antennas is reported. Using optical pulses with energies in the range from 1 nJ to 1µJ, we generate THz pulses with up to 50 pJ in energy and electric field strengths on the order of 1 kV/cm. Two emitter concepts are investigated and circumvention of the fast saturation for small area excitation by scaling of the THz emitter is demonstrated.
© 2011 OSA
Terahertz (THz) radiation in the range from 0.2 – 10 THz, generated by ultrafast lasers via optical rectification or highly excited dipole antennas typically consists of 0.5 to 2 cycles of light and therefore covers a broad frequency range. This makes these pulses ideal for broadband spectroscopic applications, where spectra can be acquired at the highest possible time-resolution. For applications such as free space imaging such THz pulses are less useful due to absorption of water vapor. Moreover, broadband pulses typically excite a whole series of transitions, whereas intense narrowband pulses allow for nonlinear excitation of specific resonances. For the latter cases intense narrowband sources such as free-electron lasers have been the ideal light source, and for imaging true continuous-wave sources such as quantum cascade lasers and photomixers proved to be highly suitable.
Utilizing the broad spectra of ultrafast lasers, one possibility to generate narrowband THz radiation is based on mixing two time-delayed, linearly chirped pulses in a nonlinear medium. With this method narrowband, tunable THz pulses up to 1 THz were successfully generated in a dipole antenna [1,2] via a fast oscillating current induced by the beat frequency of the mixed laser pulses. Higher-frequency pulses were recently demonstrated in ZnTe, where a second-order process is utilized .
For single-cycle THz pulse generation with ultrafast lasers, significant progress has been made over the last decade , but so far no efficient way for narrowband pulse generation or tunable filtering of a desired frequency in the range from 200 GHz to 5 THz exists, even with the rapid development of metamaterials. Today, the most intense narrowband THz pulses are provided by free-electron lasers and other sources driven by linear electron accelerators . In the past, a number of techniques have been developed which utilize intense femtosecond laser pulses for tunable narrowband pulse generation [6–17]. Common to many approaches is active pulse shaping [6–14], where in the simplest case spectral filtering is applied to the NIR pulse in order to achieve a narrower THz spectrum. Another technique is based on generating NIR pulse sequences consisting of up to eight optical pulses with varying time-delay . With these methods, tunable THz pulses can be generated in the range from 0.5 to 3 THz with typical bandwidths of 100 GHz. Employing specially designed periodically poled lithium niobate (PP LiNbO3) waveguides, narrowband THz pulses can be generated with bandwidths as low as 20 GHz, but only at very low temperatures and with the requirement of a new emitter for each THz frequency [16,17]. The THz generation takes place mostly in nonlinear crystals and therefore requires large optical intensities for sufficiently high conversion efficiency. The method of chirped pulse difference frequency generation (DFG) – first introduced by Weling et al.  – represents a technique that only requires a minimum of pulse shaping, namely simply introducing a linear chirp to the optical pulse and splitting it into two parts. With this method, tunable THz pulses have been generated with amplifier systems and nonlinear crystals, and frequencies in the range of 200-700 GHz were demonstrated using a femtosecond oscillator and photoconductive antennas (PCAs) . In fact, the most narrowband (Δf/f = 10%) and simultaneously tunable pulses up to now were generated applying this method. While for optical pulse energies in the few nJ to µJ regime, PCAs give the highest efficiencies for generating single-cycle THz pulses, nonlinear crystals such as ZnTe and LiNbO3 are often used when energies of 1 mJ or higher are available in the optical pulse. Nonlinear crystals, however, do not suffer from the rapid saturation of the THz generation process when going to higher fluences, since they do not require the generation of photogenerated carriers. The problem of high carrier concentration becomes even more severe when exciting photoconductive switches with two pulses, as it is needed for the chirped pulse difference frequency generation process. Therefore, cw-THz generation via difference frequency mixing in PCAs is mostly done using low-temperature grown or ion-implanted GaAs  because of the short carrier lifetime . In addition, the devices need to be sufficiently small in order to avoid problems with the RC time constants, which in turn limits the optical power that can be used. So far no difference frequency generation in large-area PCAs has been reported in the range of 1 THz and above. Applying the method of difference frequency mixing with two chirped optical pulses to our recently introduced large area PCA design [19,20] we demonstrate the generation of tunable THz pulse generation in the range from 300 GHz up to 2.4 THz, with maximum efficiency around 1 THz.
2. Principle and experimental setup
To describe the principle of chirped pulse difference frequency generation, we consider two identical optical pulses that are delayed in time by τ. The electric field of the two linearly chirped Gaussian pulses with pulse length can be written as2] that the chirp parameter also determines the spectral width (FWHM) of the generated pulses2]. A schematic of the described process is shown in Fig. 1(a) and illustrates the mechanisms and important parameters involved.
The experimental realization was based on a regenerative Titanium Sapphire laser amplifier (Coherent RegA) that outputs near-infrared pulses with energies of 8 µJ at a repetition rate of 250 kHz. The pulses are centered at 800 nm and have a typical pulse length of 35 fs. Using the built-in grating-based compressor, the output pulses can efficiently be stretched up to a pulse length of 3.3 ps. Generating longer pulses would require a modification of the compressor or an additional external stretcher. The chirped pulses are then divided into two parts using a beamsplitter: the intense part (90%) is sent to the difference frequency generation unit and the second part (10%) is used to characterize the generated THz radiation via electro-optic sampling . To this end, the probe pulse can be delayed in time employing a mechanical delay stage and is further recompressed to a pulse length of 45 fs using a home-built grating compressor before entering the electro-optic sensor crystal. The intense, chirped pulse is sent to an interferometric unit, where the pulse is devided in a 50:50 beamsplitter (see Fig. 1(b)). The length of one arm can be adjusted using a motorized translation stage with respect to the other and its polarization is rotated by 90° before both beams are recombined on a polarizing beam splitter. The now perpendicularly polarized beams are then sent through a half-wave plate and a polarizer before they are focused onto the THz emitter. The generated THz radiation is collected with a 2” diameter gold-coated off-axis parabolic mirror and is sent to the electro-optic sampling unit (not shown), where it is recombined with the probe pulse on a 500 µm thick <110> oriented ZnTe crystal. The setup is under ambient air condition.
As THz emitters we employ two large-area photoconductive antennas with interdigitated finger structures [20,21]. While one emitter is based on standard semi-insulating GaAs (SI-GaAs), the second one is based on a low-temperature grown GaAs (LT-GaAs). The SI-GaAs emitter had an active area of 10 × 10 mm2 while the LT-GaAs based was only 1 × 1 mm2 in area. Both emitters have a gap spacing of 5 µm. The diameter of the optical beam for low optical pulse energies (<50 nJ) was set to 300 µm and for high pulse energies to 6 mm, in order to avoid saturation of the THz emitter. Due to the different sizes of the two emitter structures, the latter configuration could only be investigated using the much larger SI-GaAs emitter. The combined optical power on the emitters was set to 10 mW (5 mW in each part). The emitters where biased using a synchronized pulse generator, providing an acceleration field of 40 kV/cm. For acquiring the THz transients, Lock-In detection referenced to the electrical pulses was employed and for power measurements, a liquid He-cooled silicon bolometer served as a detector.
In this section we will discuss our experimental results on the narrowband THz pulse generation. In Fig. 2 we compare the generation of THz pulses from a large-area PCA driven by a single femtosecond laser pulse with those generated by the two-pulse DFG method. For the latter, a chirped optical pulse length of 3.3 ps is chosen and we use the LT-GaAs based emitter. Figure 2(a) shows the electric field transients for both cases and one clearly makes out the differences between them: while with single fs-excitation one observes a half-cycle THz pulse whose trace mainly follows the electron dynamics, we generate multi-cycle pulses with a fixed period in case of the chirped pulse DFG. The inset of Fig. 2(a) depicts the spectral power of both time-domain traces, both having their maximum around 1 THz. Unlike with the fs-excitation where a broad spectrum is seen (FWHM = 1.2 THz), we can generate much narrower (FWHM = 200 GHz) pulses with the chirped pulse DFG technique and it demonstrates the usability of “normal” PCA structures for this.
It is now most interesting to investigate the tuning range of the narrowband pulses using these emitters as they will have – in contrast to nonlinear crystals – a very pronounced roll-off towards higher frequencies. In order to change the central frequency of the pulses, one simply has to change the time delay in the interferometer. For the data shown in Fig. 2(b), we have varied the time delay between 30 and 580 fs, respectively and plotted the spectral power for each spectrum. Also shown is the spectrum when single fs-excitation is applied. At small delays, i.e. small difference frequencies, we observe a rising of the power which is also seen in fs-excitation. However, starting from about 1 THz, the spectral power decays much faster than in the single pulse experiment and a strong roll-off is observed. This stronger roll-off with about 30dB/octave is mainly caused by the capacitance and the intrinsically small photocurrent of the LT-GaAs antenna, the carrier lifetime, but also by the cubic phase modulation that is present in our system. At the maximum of 1 THz, however, we obtain a quarter of the peak power compared to the fs-excitation at identical pump fluence and bias.
The tuning range has also been investigated at shorter pulse lengths and simultaneously the total power was recorded with a silicon bolometer. The results are shown in Fig. 3 , where the top graph displays the total optical power measured and the lower part gives the tuning curves for different input pulse lengths. For all pulse lengths, we get a linear behavior of the frequency increase with the time delay in the interferometer, and from a linear fit to these curves we can extract the chirp parameter. It decreases from 29.6 ps−2 for a pulse length of 1.4 ps to 13.2 ps−2 for a pulse length of 3.3 ps, respectively. The maximum frequency that can be generated lies around 2.4 THz. As can been seen in the top part of Fig. 3, the highest average powers are generated applying the shortest optical pulses, i.e. 1.4 ps, where we can generate an average power of 12 µW, resulting in a pulse energy of about 50 pJ. This gives a conversion efficiency of 1 × 10−3. The maximum optical power is found to be at a frequency of 0.7-0.8 THz, consistent to what we observed via the electro-optic sampling method (compare Fig. 2(b)). Note that close to zero time delay the THz emission vanishes, indicating that only DFG based radiation is generated and the contribution from each of the two individual pulses to the THz is negligible. If we go to longer pulses, the peak shifts to longer delay times (but remains at 0.75 THz) and becomes smaller. Starting from a pulse length of 2.1 ps, we always find two distinctive dips in the power curves. These dips are not due to multiple pulses, but reflect the main water vapor absorption lines (at around 1.15 and 1.7 THz ) and become deeper and sharper as we reduce the bandwidth of the THz pulses. The reason why water vapor absorption is present is that the system is under ambient air and except for the case of fs-excitation with a single pulse (see Fig. 2), it is not purged with dry N2 gas.
To investigate the dependence of the bandwidth of the emitted radiation we have set the emission frequency to 0.98 THz for two different input pulse lengths and evaluated the spectral width. The result is shown in Fig. 4 . Here we clearly see a decrease of the bandwidth from 360 GHz for the case of 1.4 ps long pulses to around 200 GHz when 3.3 ps long pulses are applied in the DFG process. The measured values agree quite well with the expected values from Eq. (3), which predict a FWHM of 450 GHz for a chirp parameter of 29.6 ps−2 (1.4 ps pulse) and 208 GHz for the case of the long pulses (b = 13.2 ps−2). While the latter shows very good agreement with the calculated value, the measured value for 1.4 ps is smaller than expected. A reason for this is most likely that the generated pulse is not symmetric and falls off towards lower frequencies. Also, the optical pulses are not entirely bandwidth limited (Δτ × Δυ = 0.66) as it is assumed for the derivation of Eq. (3).
We now want to explore the dependence of the generated narrowband THz radiation on the employed optical power. Therefore, we have varied the optical pulse energy over a wide range, i.e. from nJ as a typical optical pulse energy of Ti:Sapphire oscillators, up to the full pulse energy of a few µJ that is available from our amplified system. The results of that study are summarized in Fig. 5 . There, the DFG frequency was set to 1 THz and full electric field transients were recorded at each power level. The extracted power value is derived from integration over the power spectrum obtained by a fast Fourier transform. For all measurements, the optical pulse length was set to 3.3 ps. We first begin with the LT-GaAs based antenna that we have used so far and whose power dependence is shown as the blue triangles. While we start with quadratic power dependence at low pulse energies, we quickly run into the saturation of the emitted power and above 40 nJ, the emitted power increases very slowly. This behavior, however, is well known from photoconductive antennas where for very large carrier concentrations the ultrafast photocurrent is reduced by the strong local THz field resulting in strong saturation of the THz emission.
As already mentioned above, we have also studied to what degree emitters based on SI-GaAs are suitable for this method of narrowband THz generation. We first tested these emitters, which are identical in structure, with fs-excitation and found that the maximum electric field of the THz pulses was about a factor of three higher than that of the LT-GaAs emitter, which results in almost a factor of ten in power. For a direct comparison with the LT-GaAs emitter, we studied the power dependence also for the case of a small optical focus, shown as the black squares in Fig. 5. Similar to the LT-GaAs emitter, we observe a quadratic increase of the emitted power with increasing pulse energy (solid black line in Fig. 5). While showing the same saturation behavior with increasing pulse energy as for the case of the LT-GaAs emitter, its overall output power is about one fourth throughout all pulse energies. Hence we find that these emitters – much cheaper in the substrate price and readily available – still function as suited emitters for this generation scheme, despite the fact that in conventional photomixers, SI-GaAs-based emitters show very poor performances. One can conclude that the higher mobility can partly overcome the deficits arising from the very long recovery time in these emitters.
One way of circumventing the early saturation with rising pump pulse energy is to employ a wider optical focus on the antenna. Due to the perfect scalability, these emitters provide an ideal platform to test this approach. In fact, it has recently been shown for these emitters that with excitation areas of about 20 mm2, electric field transients of more than 30 kV/cm can be obtained using µJ femtosecond optical pulses . In order to test this regime in our setup, a 10 × 10 mm2 large SI-GaAs antenna was placed far before the focus of the optical beam (right sketch in Fig. 5) in order to achieve an excitation spot of 28 mm2 (beam diameter of 6 mm). With such a large optical excitation, the generated THz radiation will have its focus at the same position as the generating optical pulse. Just as in the case of small-focus excitation, the radiation is collected with an off-axis parabolic mirror and is sent to electro-optic sampling. Using this method, one simply has to move the emitter in or out of focus and none of the optical elements need alignment. The result of the large-area excitation is shown as the red dots in Fig. 5. Just as for the other two emitters, we observe a quadratic increase and in the regime of small pulse energies (< 100 nJ), the generated THz power is ten times smaller when compared to the focused SI-GaAs based emitter and 40 times smaller than of the LT-GaAs emitter. However, when increasing the pump pulse energies further, saturation sets in much later than for the other two and eventually higher power – even higher than that of the LT-GaAs emitter – can be generated using the widened optical excitation scheme. Using a combined pulse energy of 1 µJ, the highest narrowband (FWHM 200 GHz) power was obtained and resulted in a peak electric field of 0.7 kV/cm. At this point, however, the efficiency has dropped to 3 × 10−5, compared to 1.5 × 10−4 in the small-focus regime of the SI-GaAs emitter. Comparing these values to recently published data where THz pulses were generated in a similar way using a 1 kHz Ti:Sapphire amplifier , this value is still more than an order of magnitude higher, demonstrating that the proposed difference frequency generating scheme with large area photoconductive antennas is an efficient tool for narrowband THz generation.
Based on a regenerative Ti:Sapphire amplifier and large area photoconductive switches, we have demonstrated the efficient generation of narrowband THz pulses, easily tunable in the frequency range from 0.3 to 2.4 THz. For the narrowest THz pulses, the highest conversion efficiency of 5 × 10−4 was obtained employing a large area LT-GaAs based emitter. For pulses exhibiting a 360 GHz bandwidth, a conversion efficiency of more than 1 × 10−3 is achieved. Additionally, we have shown functionality of the system with commercially available SI-GaAs based emitter structures that exhibited excellent behavior and demonstrated the scalability of our generation scheme, which allowed the generation of 0.7 kV/cm strong narrowband THz pulses. For future development, even larger THz powers can be expected when instead of SI-GaAs, either LT-GaAs or ion-implanted GaAs  is chosen as substrate for the case of large area excitation.
The authors thank H. Schneider for fruitful discussion and K. Köhler for providing the LT-GaAs substrate.
References and links
1. A. S. Weling, B. B. Hu, N. M. Froberg, and D. H. Auston, “Generation of tunable narrow-band THz radiation from large aperture photoconducting antennas,” Appl. Phys. Lett. 64(2), 137–139 (1994). [CrossRef]
2. A. S. Weling and D. H. Auston, “Novel sources and detectors for coherent tunable narrow-band terahertz radiation in free space,” J. Opt. Soc. Am. B 13(12), 2783–2791 (1996). [CrossRef]
3. J. R. Danielson, A. D. Jameson, J. L. Tomaino, H. Hui, J. D. Wetzel, Y.-S. Lee, and K. L. Vodopyanov, “Intense narrow band terahertz generation via type-II difference-frequency generation in ZnTe using chirped optical pulses,” J. Appl. Phys. 104(3), 033111 (2008). [CrossRef]
4. Y.-S. Lee, Principles of Terahertz Science and Technology (Springer, 2009).
5. A. M. Cook, R. Tikhoplav, S. Y. Tochitsky, G. Travish, O. B. Williams, and J. B. Rosenzweig, “Observation of narrow-band terahertz coherent Cherenkov radiation from a cylindrical dielectric-lined waveguide,” Phys. Rev. Lett. 103(9), 095003 (2009). [CrossRef] [PubMed]
6. W. C. Hurlbut, B. J. Norton, N. Amer, and Y.-S. Lee, “Manipulation of terahertz waveforms in nonlinear optical crystals by shaped optical pulses,” J. Opt. Soc. Am. B 23(1), 90–93 (2006). [CrossRef]
7. Y. Liu, S.-G. Park, and A. M. Weiner, “Terahertz waveform synthesis via optical pulse shaping,” IEEE J. Sel. Top. Quantum Electron. 2(3), 709–719 (1996). [CrossRef]
8. J. Y. Sohn, Y. H. Ahn, D. J. Park, E. Oh, and D. S. Kim, “Tunable terahertz generation using femtosecond pulse shaping,” Appl. Phys. Lett. 81(1), 13–15 (2002). [CrossRef]
11. C. D’Amico, M. Tondusson, J. Degert, and E. Freysz, “Tuning and focusing THz pulses by shaping the pump laser beam profile in a nonlinear crystal,” Opt. Express 17(2), 592–597 (2009). [CrossRef] [PubMed]
12. S. Vidal, J. Degert, J. Oberlé, and E. Freysz, “Femtosecond optical pulse shaping for tunable terahertz pulse generation,” J. Opt. Soc. Am. B 27(5), 1044–1050 (2010). [CrossRef]
13. M. Yamaguchi and J. Das, “Terahertz wave generation in nitrogen gas using shaped optical pulses,” J. Opt. Soc. Am. B 26(9), A90–A94 (2009). [CrossRef]
14. K. K. Kohli, A. Vaupel, S. Chatterjee, and W. W. Rühle, “Adaptive shaping of THz-pulses generated in <110> ZnTe crystals,” J. Opt. Soc. Am. B 26(9), A74–A78 (2009). [CrossRef]
15. J. Ahn, A. V. Efimov, R. D. Averitt, and A. J. Taylor, “Terahertz waveform synthesis via optical rectification of shaped ultrafast laser pulses,” Opt. Express 11(20), 2486–2496 (2003). [CrossRef] [PubMed]
16. Y.-S. Lee, T. Meade, T. B. Norris, and A. Galvanauskas, “Tunable narrow-band terahertz generation from periodically poled lithium niobate,” Appl. Phys. Lett. 78(23), 3583–3585 (2001). [CrossRef]
17. Y.-S. Lee, N. Amer, and W. C. Hurlbut, “Terahertz pulse shaping via optical rectification in poled lithium niobate,” Appl. Phys. Lett. 82(2), 170–172 (2003). [CrossRef]
18. S. Preu, G. H. Döhler, S. Malzer, L. J. Wang, and A. C. Gossard, “Tunable, continuous-wave Terahertz photomixer sources and applications,” J. Appl. Phys. 109(6), 061301 (2011). [CrossRef]
19. S. Winnerl, F. Peter, S. Nitsche, A. Dreyhaupt, B. Zimmermann, M. Wagner, H. Schneider, M. Helm, and K. Köhler, “Generation and detection of THz radiation with scalable antennas based on GaAs substrates with different carrier lifetimes,” IEEE J. Sel. Top. Quantum Electron. 14(2), 449–457 (2008). [CrossRef]
20. A. Dreyhaupt, S. Winnerl, T. Dekorsy, and M. Helm, “High-intensity terahertz radiation from a microstructured large-area photoconductor,” Appl. Phys. Lett. 86(12), 121114 (2005). [CrossRef]
21. Z. G. Lu, P. Campbell, and X.-C. Zhang, “Free-space electro-optic sampling with a high-repetition-rate regenerative amplified laser,” Appl. Phys. Lett. 71(5), 593–595 (1997). [CrossRef]
22. High-resolution transmission molecular absorption database (HITRAN), http://cfa-www.harvard.edu/HITRAN/.
23. M. Beck, H. Schäfer, G. Klatt, J. Demsar, S. Winnerl, M. Helm, and T. Dekorsy, “Impulsive terahertz radiation with high electric fields from an amplifier-driven large-area photoconductive antenna,” Opt. Express 18(9), 9251–9257 (2010). [CrossRef] [PubMed]