1. Introduction

Ultrashort laser pulses comprising only a few optical cycles in the spectral range of 6–16 $\mu$m (MID-IR) offer lots of opportunities for research in materials and the life sciences. The development of versatile femtosecond (fs) sources helps to unravel complex physical, chemical and biological processes by means of advanced experimental schemes, such as 2-dimensional vibrational spectroscopy [1–4], label-free microscopy [5,6], nonlinear optics [7,8] and nonlinear phononics [9]. Stimulating and probing low-energy excitation is key to shed light on a vast range of very different phenomena spanning from high-temperature superconductivity in solid-state materials to protein structure and dynamics in liquid phase.

Many important biomolecules have their characteristic vibrational absorption bands within this wavelength range (6–16 $\mu$m); the so-called fingerprint region. The amide vibrations ranging from 1200–1700 cm$^{-1}$, aromatic breathing and out-of-plane CH-bending modes (in the 900–1000 cm$^{-1}$) can be directly accessed via wavelength-tunable MID-IR sources [1,2]. In addition to well-known one-photon excitations, the high peak power of ultrashort laser pulses with high pulse energy induces nonlinear absorption and thereby exciting higher lying vibrational states in multiphoton processes. These final states might not be accessible by using continuous-wave IR sources, because of dipole selection rules in single-photon transitions. Furthermore, the combination of long-wavelength sources with pulse-shaping capabilities for tailored waveform generation adds a new dimension to the coherent-control toolbox typically aiming at steering photophysical and photochemical processes with light [10–12]. For instance, enhanced selective excitation of ground-state vibrational modes in the fingerprint region could be used for identifying different cellular components in an image using their intrinsic vibrations in a non-invasive manner. Two important aspects of this idea have been demonstrated in the past. First, MID-IR pulse shaping has been used for efficient stepwise excitation of vibrational energy levels ("vibrational ladder climbing") in NO molecules by means of tailored electromagnetic waveforms accounting for the anharmonic molecular potential [13]. Second, MID-IR induced vibrational absorption has been used in photothermal imaging applications, where the photothermal effect is spatially resolved using a visible laser beam. In this scheme the lack of spatial resolution due to longer-wavelength MID-IR sources is overcome by the visible laser and images are obtained with chemical-specific contrast [5].

In recent years the commercial availability of nonlinear crystals (NLC) that are transparent in the MID-IR along with high-power pump lasers operating at high repetition rate and high average power reaching 1000 W with sub-ps pulse duration contributed to the research and development of novel fs MID-IR laser architectures. Several research teams worldwide have pushed this development with power-scaling schemes using different pump-laser technology including fiber [14,15], thin-disk [16], and Innoslab lasers [17] at a pump wavelength of 1 $\mu$m and 2 $\mu$m, respectively [18,19]. The state-of-the-art laser parameters from recent literature using lithium gallium sulfide (LGS) crystals in the final DFG stage in comparison to the present work is summarized in Table 1. The choice of NLC that can be efficiently pumped at 1 $\mu$m is mainly limited by the two-photon absorption as recently shown in thermal studies [20]. It turns out that in the case, where two-photon absorption is dominating, LGS is performing best with its large comparative bandgap, effective nonlinear coefficient and high thermal conductivity not showing any thermal degradation or Kerr lensing under realistic operational conditions [20].

Table 1. Performance parameters of MID-IR laser systems based on lithium gallium sulfide (LGS) nonlinear crystals

View Table

The two major approaches for down-converting the pump photon into the MID-IR idler involve either an intra- or inter-pulse DFG process [30]. Intra-pulse DFG employs a laser pulse with broad spectral bandwidth. In this case, the difference-frequency mixing of the blue side (pump for DFG) and red side (signal for DFG) of its spectrum generates the idler upon nonlinear interaction (wave-mixing) inside the NLC. In the case of inter-pulse DFG, two independent signal and pump laser pulses are required. The main advantage of intra-pulse DFG is the simplicity of the optical setup, generating a broadband MID-IR of similar pulse duration to the pump pulse. The wavelength-tunability of intra-pulse DFG is mainly defined by the phase-matching angle of the crystal itself. The inter-pulse DFG offers a wide tunable spectral range compared to intra-pulse DFG, because independent wavelength tuning of the seed is possible along with the NLC angle tuning. The beampath has to be collinear in case of inter-pulse DFG to obtain angular-dispersion-free idler pulses. This poses a major challenge in the MID-IR OPA development as filtering of the high-intensity pump pulse along with the signal pulse calls for special filters and coatings with high damage threshold and low group-delay-dispersion (GDD) materials [24,31]. The goal of the present MID-IR laser research and development project was to build a versatile source providing both, broadband, few-cycle, CEP-stable, high-energy pulses on the one hand and wavelength-tunable, high-energy pulses on the other hand. The result is a powerful OPCPA system with two complementary modes of operation at 200 kHz repetition rate with minimum re-adjustment necessary upon switching.

2. Experimental setup

The optical parametric chirped-pulse amplifier scheme with a final difference-frequency-generation stage (OPCPA-DFG) discussed here uses an industrial Yb-YAG pump laser (Amphos) with a central wavelength $\lambda _0 = 1030$ nm and a pulse duration $\tau \approx 1$ ps. The pump laser has a tunable repetition rate from 200 kHz up to 1 MHz. Similar average power levels can be obtained at all repetition rates with maximum output power of 200 W. In the present study 40 W from the Amphos laser corresponding to a pulse energy of 200 $\mu$J at a 200 kHz repetition rate is used to pump the OPCPA-DFG without thermal issues when filtering the idler. Thereof, 4–5 $\mu$J is used to generate the seed pulse by means of white-light generation (WLG) in an undoped bulk YAG crystal of 5 mm length, which was not scanned. The seed is further amplified in a non-collinear optical parametric amplifier (NOPA). The NOPA consists of a type I phase-matched 3 mm BBO crystal and a second-harmonic-generation (SHG) pump from a 1 mm thick BBO (type I phase-matching). 25–30 $\mu$J are sent to the SHG stage. We reach about $53\%$ conversion efficiency in the SHG and about $11\%$ at the NOPA position. The remaining pulse energy ($\approx 155$ $\mu$J) is used as the pump in the final DFG stage at a peak intensity of $\sim$15 GW/cm$^2$, which consists of an LGS crystal (type II phase-matching in the XY plane, $\theta$ = 90$^{\circ }$; $\phi$ = 37.5$^{\circ }$). This setup provides the aforementioned flexibility allowing for either few-cycle broadband or tunable-wavelength operation modes.

3. Results and discussion

The energies of both seed and pump pulses can be tuned individually for efficient energy extraction. The pump and the amplified seed are made collinear using a dichroic beam combiner (Semrock). The versatile laser architecture employs two different chirp schemes of opposite sign for the seed, which is stretched to $\approx 0.65$ ps after white-light generation. A positive GDD is introduced by a bulk material (SF11, 100 mm) and results in a narrow-band, high-energy and wavelength-tunable MID-IR source depicted in Fig. 1. Characteristic power stability and beam profile measurements of the broadband MID-IR OPCPA centered at $\approx 9$ $\mu$m with an M$^2\approx 1.4$ using a commercial beam profiler (WinCamD-IR-BB) are included in the figure. These measurements were performed over a period of 1 hour under ambient laboratory conditions without any OPCPA enclosure or vibrational damping, which along with quantum noise contributions result in fluctuations on the order of $\approx 9\%$ standard deviation. The independently tunable NOPA spectrum and the energies before the final pumped-DFG stage with the LGS crystal and the amplified seed after the LGS are shown in Fig. 2.

 

Fig. 1. Schematic of the OPCPA-DFG with positive GDD seed including characteristic MID-IR power stability and beam profile measurements of the broadband idler with reduced transmission in the diagnostic mode (see main text). The laser system can be broadband at a central wavelength of $\approx 9$ $\mu$m and is wavelength tunable. BS - beam separator, SHG - second-harmonic generation, WLG - white-light generation, NOPA - non-collinear optical parametric amplifier, DFG - difference-frequency generation, SF11 - stretcher.

Download Full Size | PDF

 

Fig. 2. Seed pulse energy curves before and after the final pumped-DFG stage employing an LGS crystal (a) and the wavelength-tunable NOPA spectra (b). Solid lines in (a) are to guide the eye.

Download Full Size | PDF

In the negative-signal GDD scheme the SF11 stretcher is replaced by custom-made chirped mirrors (Layertec) for broadband amplification in both the NOPA and the final pumped-DFG stage. The LGS crystal (having a GDD per mm of $-3813.2$ fs$^2$ at 8.5 $\mu$m [32]) itself acts as the main compressor of the final idler output with a positive GDD. The setup uses a custom-made beam separator (Laser Optik) with coating on a 3 mm-thick ZnSe substrate to remove the pump and the seed pulses along with the residual chirp in the idler pulse. The beam separator must be water-cooled for long-term operation. Without cooling we have observed that the coating damages after 30 min exposure due to the high-power pump and the amplified seed pulses. An alternative separation method that we have successfully applied is to take advantage of the larger divergence of the longer-wavelength idler. We used a collimation method involving a two-inch gold-coated Herriott cell mirror ($f = 200$ mm, center hole diameter of 4 mm) and a protected gold convex mirror ($f = -150$ mm). The center hole bypasses the high-power pump and the seed pulses along with $\approx 65\%$ MID-IR idler as well. Due to the high losses, this method is only useful for MID-IR pulse-characterization and setup-optimization purpose. However, the measured energy of the collimated MID-IR idler with this method (diagnostic mode) was still $\approx 0.35$ $\mu$J.

The wavelength-tunable MID-IR spectra recorded in the positive-signal GDD scheme are shown in Fig. 3. The measured pulse energies at different MID-IR central wavelengths and the corresponding total efficiencies of the system in the tuning range are given in Fig. 3(a). The energy is measured by slightly introducing a non-collinear angle (0.2$^{\circ }$) between the pump and seed beams. The maximum pulse energy of $\approx 2.2$ $\mu$J and $\approx 450$ mW average power is obtained at 8.5 $\mu$m. The dip in the energy curve at 6–7 $\mu$m is mainly due to the water-vapor absorption in air. The MID-IR spectra were recorded using a home-built FTIR spectrometer.

 

Fig. 3. MID-IR pulse energy curve (a) and the total efficiencies of the wavelength-tunable MID-IR spectra (b) in the positive-signal GDD scheme. The solid line in (a) is to guide the eye.

Download Full Size | PDF

In the case of the negative-signal GDD scheme, a broadband NOPA seed pulse with a pulse energy of 1.2 $\mu$J is frequency-mixed with the 1030 nm pump pulse generating a broadband idler pulse in the MID-IR. The spectra are displayed in Fig. 4(a) and Fig. 4(b), respectively. The measured MID-IR pulse energy in this complementary broad-spectral-bandwidth operation mode was $\approx 1.2$ $\mu$J resulting in an average power of $\approx 240$ mW at a central wavelength of 8.5 $\mu$m. The MIR spectral bandwidth supports a theoretical Fourier-transform-limited pulse duration of 68 fs. Note the LGS transmission near 9 $\mu$m is only $\approx 20\%$ for a crystal thickness of 1.75 mm [33], which is clearly visible as a dip in the spectrum.

 

Fig. 4. The negative-signal GDD scheme generates a broadband NOPA spectrum, which seeds the DFG stage (a). This scheme results in a broadband MID-IR spectrum (b). The cross-correlation trace of the MID-IR pulse fitted with a Gaussian envelope (c) is shown in the inset.

Download Full Size | PDF

The temporal characterization of the MID-IR idler was performed with a home-built SHG autocorrelator employing an AgSe (AgGaSe$_2$) crystal of 0.1 mm thickness (type I phase-matching, $\theta = 44.2^{\circ }$; $\phi = 45^{\circ }$). The beam splitter used in the temporal-diagnostics setup was a KBr substrate (thickness, 4 mm) with a germanium coating on one side (Spectral Systems). The transmitted part of the idler gets chirped due to the GDD introduced by the KBr substrate (17 mm path length gives $-3953.74$ fs$^2$) converting the autocorrelator into a cross-correlator. The cross-correlation trace fitted with a Gaussian envelope is shown in the inset Fig. 4(c). The deconvolution of the cross-correlation width $\tau _{\mathrm {CC}}$ (fwhm) results in a MID-IR pulse width of $\tau _{\mathrm {p}}\approx 74\pm$2 fs (fwhm).

4. Summary

We have presented a versatile design of a MID-IR OPCPA laser system using an LGS crystal as nonlinear medium for efficient frequency-down conversion of high-power 1030 nm pump pulses. The two discussed operation modes differ in the group-delay dispersion of the signal pulse in the final DFG stage. Depending on the application, the laser architecture can be easily switched between complementary MID-IR pulse parameters in frequency and time domain, which were characterized in the present study by means of FTIR spectroscopy and SHG autocorrelation, respectively. On the one hand, the scheme with positive-signal GDD provides a wavelength-tunable source (4.2–8.5 $\mu$m) that is for instance well-suited for vibrational spectroscopy of gas phase molecules and liquids. The vibrational line widths of molecules are in the few $\approx 1$–10 ps region, therefore the system without the final compression is an ideal tool for vibrational imaging and related applications including microscopy [34]. On the other hand, the negative-signal GDD scheme provides broadband MID-IR pulses that can be compressed down to the Fourier limit. We could demonstrate ultrashort pulses with a duration of 74$\pm$2 fs at a central wavelength of 8.5 $\mu$m, which corresponds to only $\approx 3$ optical cycles. These waveforms are of considerable interest for coherent nonlinear spectroscopy and imaging applications involving multiphoton physics and time-resolved studies. It is important to note that both schemes generate pulses with 1–2 $\mu$J energy at 200 kHz repetition rate.