Broadband, high power THz source at 540 kHz using organic crystal BNA

Few-cycle terahertz (THz) sources driven by ultrafast lasers are widely used in THz time domain spectroscopy (THz-TDS), which is a well-established tool for many applications in THz science and technology.^1–4 For applications where ultra-broad THz bandwidths are desired, most commonly spectroscopy, various techniques are available to generate broadband THz emission, such as two-color plasma filaments, optical rectification (OR), and spintronic THz emitters. Among these techniques, collinear optical rectification in nonlinear crystals with second order nonlinearity is one of the simplest and most commonly used methods, albeit well-known difficulties in achieving very broad bandwidths due to phonon absorption peaks in typically used inorganic materials such as gallium phosphide (GaP) or zinc telluride (ZnTe). In this respect, organic crystals such as DAST (4-N, N-dimethylamino- 4 ′ - N ′ -methyl-stilbazolium tosylate), DSTMS (4-N,N-dimethylamino- 4 ′ - N ′ -methyl-stilbazolium 2,4,6-trimethylbenzenesulfonate), OH1 (2-(3-(4-hydroxystyryl)-5,5-dimethylcyclohex-2-enylidene) malononitrile), HMQ-TMS (2-(4-hydroxy-3-methoxystyryl)-1-methylquinolinium 2,-4,-6-trimethylbenzenesulfonate), and BNA provide a promising platform to overcome this drawback. Due to the intrinsically lower dispersion of refractive index in organic nonlinear crystals, they are collinearly phase-matched in a much broader THz bandwidth compared to inorganic crystals.^5,6 For example, Puc et al.⁷ detected a signal bandwidth of more than 20 THz using DSTMS pumped by 38 fs laser pulses. In addition, a typically high nonlinear coefficient results in very high optical to THz conversion efficiencies, reaching up to the percent level. Using DSTMS, Vicario et al.⁸ achieved a conversion efficiency of 3%, which led to an ultra-high THz pulse energy of 0.9 mJ at a repetition rate of 10 Hz. For the organic crystal OH1, a conversion efficiency of 3.2% was demonstrated, again at a low repetition rate of 10 Hz.⁹ By pumping a DAST crystal with mid-IR pulses, Gollner et al. achieved a record value of 6% conversion efficiency and 1 mW of THz average power at a repetition rate of 20 Hz.¹⁰ In addition, BNA is highly suitable for efficient THz generation¹¹ and has already been shown to operate efficiently with ytterbium-based lasers.¹² A direct comparison between efficiencies obtained using different organic crystals used in previous experiments is difficult to perform in a meaningful way, since the pump lasers and excitation conditions are very different. However, it is worth comparing other organic crystals pumped at a similar wavelength of 1030 nm to place our results into context. For example, using HMQ-TMS, an efficiency of 0.26% at a repetition rate of 500 Hz was reported in Ref. 13, but Buchmann et al.¹⁴ reported an efficiency of 0.055% at the same wavelength but a repetition rate of 10 MHz. This illustrates the difficulty in comparing the efficiencies in the literature even for same crystals. Using BNA at this pump wavelength, 0.04% efficiency was achieved at a repetition rate of 13.3 MHz¹⁵ and 0.6% at a repetition rate of 1 kHz.¹⁶ 

However, due to limited crystal quality and poor thermal properties (compared to inorganic crystals), most results using these crystals have so far been limited to operation with low repetition rate, i.e., <1 kHz, pump lasers with a high pulse energy and low average power. Only very recently, these crystals have started to be investigated for high repetition rate regimes: HMQ-TMS and BNA have recently shown the first promising results with a THz average power of around 1 mW at >10 MHz repetition rates.^14,15 In these results, the thermal limitations can be circumvented by operating in burst mode, leaving enough time between bursts for the crystal to cool down.¹⁵ 

However, even if the thermal effects can be controlled, the optical rectification process is challenging to drive efficiently at MHz repetition rates. This is due to the comparatively low pulse energies being available, unless prohibitively high average powers are used, which are hardly commercially available. As a consequence, the demonstrated multi-MHz repetition rate sources still operate with a moderate conversion efficiency and yield rather moderate THz fields, which limits many applications in spectroscopy. Furthermore, repetition rates of tens of MHz could, for some systems, be too fast to allow the samples to relax from pulse to pulse, which is a prerequisite for pump-probe experiments. In this respect, repetition rates of hundreds of kHz offer an attractive middle ground for high-dynamic range and intense THz generation using well-established industrial-grade lasers. So far, only one recent result has reported optical rectification in this excitation regime using GaP and GaSe as emitters,¹⁷ but no reports have been made using organic crystals.

In this letter, we demonstrate a THz-TDS based on collinear OR in the organic crystal BNA (Swiss Terahertz GmbH) at room temperature, driven by a commercial laser system delivering up to 47 W of average power, operating at a 540 kHz repetition rate. The pulses from the driving laser are temporally compressed from 240 fs down to 45 fs using a home-built Herriott-type multi-pass cell (MPC) compressor. Using 4.7 W of driving power on the crystal and a duty cycle of 50%, we reach 5.6 mW of THz average power with a broad bandwidth extending up to 7.5 THz at a high dynamic range of 75 dB, with a conversion efficiency of 0.12%. To the best of our knowledge, this is the highest average power THz source achieved so far in BNA, improving the current state-of-the-art reported in Ref. 15 by a factor of more than 5 and an order of magnitude higher conversion efficiency. Moreover, it is the first organic crystal-based source operating in the attractive multi-hundred kHz repetition rate regime.

The experimental setup is shown in Fig. 1. The driving laser is a commercial ytterbium-doped laser (Carbide, Light Conversion) with a central wavelength of 1035 nm, a maximum average power of 50 W, and a pulse duration of 240 fs. The current experiment is performed at a repetition rate of 540 kHz; however, the laser could potentially be tuned in repetition rate between 0.1 and 1 MHz. In order to both increase the efficiency of THz generation in BNA and generate a broader bandwidth THz radiation, an external compressor is beneficial to shorten the temporal duration of the laser pulses. The Herriott-type MPC, indicated by a dashed-contour box, has been designed and built in-house. It consists of two highly reflective plano–concave mirrors with different radii of curvature (ROC) of 300 and 500 mm and a cavity length of 750 mm between these mirrors, providing 13 round trips (26 passes) through a 9.5 mm anti-reflection coated fused silica (FS) plate, which represents the nonlinear medium where spectral broadening of our laser takes place via self-phase modulation. The plate is carefully positioned along the caustic of the beam to achieve sufficient broadening without spatiotemporal couplings that can degrade the output beam quality. One of the two mirrors has a group delay dispersion (GDD) of −350 fs² per bounce, which compensates for the material dispersion of the nonlinear medium. After the spectral broadening in the MPC, the beam undergoes 12 reflections on dispersive mirrors with a GDD of −200 fs² per bounce to remove the chirp from the spectrally broadened pulse and compress it temporally down to a 45 fs pulse duration, which is very close to the Fourier limit of the broadened spectrum.

We characterize the compressed pulses at the aforementioned repetition rate utilizing second harmonic generation frequency resolved optical gating (SHG-FROG). Figures 2(a) and 2(b) show the measured and reconstructed SHG-FROG traces, exhibiting a retrieval error of 0.3% on a 512 × 512 grid, confirming the good fidelity of the retrieved trace. Figure 2(c) shows the temporal pulse profile of the retrieved pulses. The achieved peak power at this repetition rate is 1.4 GW, making this a very attractive high-repetition rate laser system not only for optical rectification, but also for other nonlinear conversion applications. Figure 2(d) shows the measured and reconstructed spectra and spectral phase after MPC, which are compared to the spectrum measured using an optical spectrum analyzer (OSA).

After the compressor, the laser beam is guided toward the THz-TDS setup and is split in two parts: 99% of the incoming laser power is used to pump a 0.65 mm thick BNA crystal, which, according to the phase-matching calculations, is highly suitable for our pulse duration. It is glued on a sapphire substrate to generate THz radiation by OR. The remaining 1% power is used as probe beam for electro-optic sampling (EOS). The crystal is mounted in such a way that the pump beam passes through the sapphire before reaching the BNA. Therefore, the THz radiation is not affected by the sapphire heat sink, which is only placed on one side. The reflection of the near-infrared (NIR) pump is taken into account (loss of about 7.5%) in the calculation of THz conversion efficiency. The 1/e² diameter of the laser beam at the position of BNA is 1.6 mm, generated by a focusing lens with a focal length of 300 mm, placed before the crystal. As we show in our recent exploration at higher repetition rates,¹⁵ it is critical for the operation of such crystals at a high repetition rate to use a chopper wheel to approach as much as possible the thermal relaxation time of the crystal during the off-time of a burst. For this experiment, we use an optical chopper wheel with a duty cycle of 50%, which was the only one available at the time of this experiment. Note that the duty cycle could be reduced even further to reach possibly higher efficiencies and/or use the full power of the laser system. The pump power is adjusted on the BNA crystal using a combination of a thin film polarizer (TFP) and a waveplate (λ/2). The generated THz radiation is collected and refocused on the detector using two off-axis parabolic (OAP) mirrors with a diameter of 50.8 mm and focal lengths of 50.8 and 101.6 mm, respectively. In order to fully characterize the THz radiation, three different methods are implemented: a calibrated pyroelectric power meter (THz20, SLT GmbH), a standard electro-optic sampling setup, or a sensitive THz camera (RIGI Camera, Swiss Terahertz). To filter out the residual laser radiation and the generated green light after BNA, polytetrafluoroethylene (PTFE) sheets (with 89% averaged THz transmission) and black pieces of cloth (with 48% averaged THz transmission) are used.

In the first measurement set, the power meter is placed at the focus of the second OAP to measure the THz average power. The power meter is calibrated and optimized at a modulation frequency of 18 Hz at the German metrology institute (Physikalisch-Technische Bundesanstalt, PTB). Therefore, the frequency of the chopper before BNA is set to 18 Hz. In Fig. 3, the left axis shows the THz power and the right axis indicates the THz conversion efficiency vs pump power. We can pump the crystal up to 4.7 W without any irreversible damage to the crystal. This maximum pump power corresponds to an average intensity of 470 W/cm² and a peak intensity of 17 GW/cm². We reach a maximum THz average power of 5.6 mW. The calculated efficiency at the maximum THz power is 0.12%.

In order to detect the THz electric field, the power meter is replaced with a 0.2 mm GaP detection crystal in the EOS setup to sample the generated THz trace using ∼200 mW of laser power. The data are acquired using a lock-in amplifier, which records the signal out of the balanced photodetector and the digitized position of the shaker. The modulation frequency of the pump beam is used as a reference for the lock-in amplifier, and it is set to 2.6 kHz. A bandwidth of 300 Hz is chosen for the low-pass filter of the lock-in amplifier, and the frequency of the shaker to sample the THz trace is set to 0.5 Hz.

Figure 4(a) shows the THz trace in the time domain averaged over 140 traces and recorded in 70 s in unpurged conditions. The corresponding power spectrum on a logarithmic scale is obtained by Fourier transform from the measured THz trace and is shown in Fig. 4(b). The spectrum has a wide bandwidth that spans up to about 7.5 THz with a high dynamic range of about 75 dB. The smooth and dense wideband spectrum is facilitated by favorable phase-matching conditions at a 1035 nm driving wavelength.

To verify the actual generated spectral bandwidth in BNA, we numerically model the THz generation process by solving the coupled wave equations in 1 + 1D, considering the temporal dimension and the propagation direction. The simulation considers phase matching, pump depletion, and the nonlinear susceptibility of this material.¹⁸ The refractive index and absorption coefficient of BNA in the THz regime are taken from Ref. 19, and the nonlinear susceptibility is taken from Ref. 5. Additionally, we take into account the transmission of the used PTFE filters,²⁰ the low-pass filter of the lock-in amplifier in combination with the shaker, and the response function of the 0.2 mm GaP detection crystal calculated according to Ref. 21. The blue shaded area in Fig. 4(b) represents the simulated spectrum of BNA as generated in 0.65 mm BNA (labeled as “simulation”). By accounting for the low pass filtering effects mentioned above, the simulated spectrum of BNA times all filters is shown as a dashed dark blue line and is labeled as “simulation × LPFs.” It shows excellent agreement with the measured spectrum at frequencies below 4 THz. The small mismatch (in logarithmic scale) of the higher frequencies can be explained by the difficulties in the detection of higher frequencies in electro-optic sampling. The high frequency components are focused tighter, and the spatial overlap between THz and probe beam to sample the THz trace is maybe not as good as for lower frequencies, resulting in lower amplitudes for higher THz frequencies.

In view of future applications of this system in spectroscopy, we provide here an estimate for the THz peak electric field reached in our setup at the maximum average power of 5.6 mW using the approximation of a Gaussian THz beam and by measuring the spot size with a bolometric THz camera.²². It is, however, to be noted that this method is only an estimated value given various uncertainties in the measurement,²³ particularly for such broadband beams. In this regard, the power meter is replaced by the camera to measure the THz spot size. To characterize the THz pulse, the Gaussian envelope of the main peak of the trace shown in Fig. 4(a) is used. The intensity of the main peak can be calculated as follows:

where A_eff is the effective area calculated using the diameter at 1/e² level in the focal plane of the second parabolic mirror and τ_THz is the THz pulse duration at full width at half maximum (FWHM). For a multi-period trace that we have from the EOS data [see Fig. 4(a)], the peak intensity (I_THz in the formula) should be estimated for the main half-period [see dashed Gaussian fit in Fig. 5(a)]. Moreover, W_THz indicates the effective energy of this main peak. The calculated half period temporal width is about 0.13 ps. Figure 5(b) shows the THz spot in the focus of the second OAP mirror at the 1/e² diameter level of 2.65 mm × 2.25 mm.

The THz electric field strength is then given by

yielding a THz peak electric field of 29 kV/cm.

We discuss here the obtained power efficiency of 0.12% in comparison with the recent results and discuss the possible limitations and how this can be improved in future experiments. The conversion efficiency reached at our high repetition rate is remarkably comparable to the previously obtained values at a lower repetition rate: 0.2% at a repetition rate of 10 Hz in Ref. 24 and 0.8% at 1 kHz in Ref. 25. This can be attributed to the good heat management by using an optical chopper, which we know from a previous work is the key point to reach the high efficiency. According to our previous investigation in Ref. 15, a low duty cycle pump chopping allows us to effectively suppress the thermal effects that reduce the conversion efficiency at a high repetition rate. We expect the same scaling laws to apply for 540 kHz, since, in both cases, the pulse-to-pulse time is significantly smaller than the thermal relaxation time of the crystal—the only difference is that we, here, have a significantly higher peak power available. This results both in a fundamentally more efficient optical rectification process and, since larger pump spots can be used, in an additional relaxation of the thermal effects. Moreover, we highlight that the compressed pulse at the output of the multi-pass cell delivers a good spatial-spectral homogeneity and a clean temporal profile with low temporal pedestals compared to typical low repetition rate amplifier systems used in other studies,^24,25 additionally helping to achieve a higher efficiency.

With respect to what can be expected without considering thermal effects, the first fundamental limitation to efficiency is the nonlinear conversion process itself. The Manley–Rowe relation for a nonlinear optical process such as optical rectification states that one THz photon is created via annihilation of one pump photon and, thus, the emitted THz photons cannot exceed the photon number of the pump light.¹⁸ With a very rough assumption of a central THz frequency of about 2.5 THz (∼10.3 meV) in our case, since the THz photon energy is about two orders of magnitude smaller than the NIR photon energy at 1 µm (∼1.2 eV), an energy conversion efficiency of 1% at maximum is expected. Higher conversion efficiencies are possible via cascading, but it is very hard to estimate the efficiency of this process as it strongly depends on linear and nonlinear pulse propagation of the pump pulses as well as the phase-matching properties of the new-red-shifted photons and the generated THz light—all values that require material properties that are unknown for these crystals. However, we know from our previous work,¹⁵ in which we used 13 MHz of repetition rate, that thermal effects are the main limitation to scale efficiency at high repetition rates. In order to circumvent this, a low duty cycle pump chopping is required to approximately match the off time of the pump pulse burst to the relaxation time of the crystal, with shorter pulse bursts helping to reach a overall lower temperature. In this experiment, we only had available a 50% duty cycle chopping wheel; we can expect to get closer to the theoretical limit and approach the percent level by using a chopping wheel with a 10% duty cycle and a correspondingly higher power laser at the input. We note that in this experiment, we were only using 20% of the available pump power; thus, lower chopping duty cycles would be a straightforward path to improve our result.

A possible estimation of the corresponding efficiency we could reach in such a future experiment has already been given by our previous results: the optical to THz conversion efficiency is directly proportional to the pump peak intensity and inversely proportional to the pulse duration squared, $η∝Ipeakτ2$⁠. The pulse duration in the current study is 40 fs, which is approximately half of the pulse duration in our previous case,¹⁵ which was 88 fs. Theoretically, this results in an ∼4 times higher efficiency. The maximum peak intensity in this experiment is 17 GW/cm², and in the old study, it was 8 GW/cm², which results in an additional ∼2 times higher conversion efficiency. Therefore, the conversion efficiency can be expected to be eight times higher in our case than in our previous results, thus reaching 0.32%. In practice, we measured a threefold increase in efficiency (0.12%), which is most likely lower due to the above-mentioned influence of thermal effects and due to a not perfectly optimized beam diameter on the crystal in the current case since the accessing time to the laser was limited. However, the good conversion efficiency and the linear power trend indicate operation in close to optimal conditions. It should be noted that the BNA substrate in Ref. 15 was diamond, which had a much higher thermal conductivity than the one used here (sapphire), which resulted also in a better heat dissipation.

Concerning the reported damage threshold values of BNA, it was found to be ∼10 mJ/cm² using a pump laser at a 1200 nm wavelength, with a repetition rate of 1 kHz at a chopping frequency of 500 Hz.²⁵ In our experiment, we measured a lower threshold of 1.7 mJ/cm² at a much higher laser repetition rate of 540 kHz, but at a lower chopping frequency and a different excitation wavelength. However, an exact comparison between these numbers obtained in two very different excitation conditions is not straightforward; the lower damage threshold can be rationalized with the detailed investigation of thermal effects reported in our previous work,¹⁵ where we showed that a trade-off between the repetition rate and conversion efficiency needs to be met due to thermal effects and the correspondingly reduced damage threshold.

In conclusion, we show a high average power, broadband, and high dynamic range THz-TDS pumped with a 1035 nm laser at a 540 kHz repetition rate with a pulse duration of 45 fs. Using a driving power of 4.7 W, the maximum measured THz power is 5.6 mW, which is the highest THz power obtained using BNA to the best of our knowledge. The maximum calculated conversion efficiency is 0.12%, which is comparable to values obtained at much lower repetition rates using this crystal. This source represents a unique tool for a variety of time-resolved THz spectroscopy experiments, currently limited by dynamic range and bandwidth, potentially for nonlinear THz spectroscopy. We believe that further THz power upscaling well into tens of mW is possible by further optimizing the cooling of the crystal, operating in purged conditions, and finely tuning focusing conditions and beam chopping duty cycle.

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), Project-ID No. 287022738 – TRR 196 MARIE, and, in part, by the Alexander von Humboldt Stiftung (Sofja Kovalevskaja Preis). It was further funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy, Grant No. EXC-2033 – Project No. 390677874 – RESOLV. Additionally, we acknowledge the support by the DFG Open Access Publication Funds of the Ruhr-Universität Bochum and by the MERCUR Kooperation project “Towards an UA Ruhr ultrafast laser science center: tailored fs-XUV beam line for photoemission spectroscopy.”

M.S. is an employee of Swiss Terahertz GmbH.

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

Samira Mansourzadeh: Conceptualization (equal); Data curation (lead); Investigation (equal); Writing – original draft (lead); Writing – review & editing (equal). Tim Vogel: Conceptualization (equal); Data curation (supporting); Investigation (equal); Software (lead); Writing – review & editing (equal). Alan Omar: Conceptualization (supporting); Investigation (equal); Writing – review & editing (equal). Mostafa Shalaby: Resources (supporting); Writing – review & editing (equal). Mirko Cinchetti: Funding acquisition (supporting); Resources (supporting); Writing – review & editing (equal). Clara J. Saraceno: Funding acquisition (lead); Project administration (lead); Resources (lead); Supervision (lead); Validation (lead); Writing – review & editing (equal).