1. Introduction

Intense, few-cycle and CEP stabilized MIR laser pulses are of great interest for the ultrafast community especially in the realm of the attosecond physics [1]. MIR pulses are exploited to increase the cutoff energy for high harmonic generation [2–4], a sub-cycle process sensible to the electric waveform of few-cycle pulses [5]. Thus, a lot of effort is undertaken to develop ultrafast CEP stable MIR laser [6–14]. As for any laser amplifier chain, the highest possible seed level is desirable as the starting point, especially in the case of high power lasers where the picosecond contrast is often affected in the first high gain stage.

To achieve tunable seed pulses in the MIR with femtosecond bandwidths, non-linear down conversion is typically employed. It rests upon two ingredients: an ultra-broad bandwidth and subsequent DFG. The enormous importance of this approach arises from the ability of passive CEP stabilization even though the driving laser is not stabilized [15]. The closer the DFG output wavelength lies to the driving laser, the more bandwidth is required to find the matching frequency pairs for DFG. If the initial driving laser does not meet this bandwidth requirement, it can be spectrally broadened through continuum generation in various ways: in nonlinear fibers [16,17] thin glass plates [8,18,19], hollow-core fibers [9,10,20,21] or filamentation in gas [11,12,22]. The second step, the DFG, is usually carried out as three wave mixing in a χ⁽²⁾ medium [6–14,23], sometimes as four wave mixing scheme [24]. If needed, these MIR seed pulses can be further amplified in subsequent optical parametric amplifier (OPA) stages [8,10–14].

The DFG schemes can be classified into two categories: (i) intra-pulse, where wave mixing occurs within one pulse and (ii) inter-pulse, where two pulses interact in a pump-probe setup [7]. In the intra-pulse scheme, the entire continuum is directly sent to the non-linear medium [9–14] with the advantage of simplicity and inherent phase stability. The disadvantages are restricted degrees of freedom and some amount of “wasted intensity” incident on target without contributing to the desired DFG process. In some cases, notch filters were used to reduce undesired wavelengths [25]. On the other-hand, inter-pulse schemes allow for full manipulation of pulse properties like pump-probe delay, phase shifting or polarization control and the ability to tailor the spectral content with beam splitters [8]. The down sides are increased complexity and the risk of phase jitter or drifts between different optical paths.

Our aim was to design a setup that combines the advantages of both methods while omitting their disadvantages and to achieve CEP stable, multi µJ level DFG output. Therefore, we propose an “all-inline pump-probe setup” by aid of a 4-f configuration shown in Fig. 1, where all spectral components can be reflected off the same optics. In addition to a modified experimental setup we try to make advantage of a particular scheme for spectral broadening. In many experiments, symmetric spectral broadening is obtained via self-phase modulation (SPM) [9,11,12,14] and the two outermost spectral edges are used for DFG. This way, the energy around the center wavelength remains unused. In contrast, we employed asymmetric spectral broadening in a filament where the central part around 800nm served as the pump for the DFG process while the spectral wing around 1.4µm is used as the seed. We expect a twofold gain, namely to keep the most energetic part of the spectrum for pumping the parametric process and to simultaneously compress it after passing through the 4-f setup. In this manner, we achieved CEP stable, 26.5 fs MIR pulses at 2 µm carrying an energy of 4.8 µJ starting from a 35 fs TiSa laser. The DFG efficiency (pump energy / MIR energy) was 7.7%. In many experiments, µJ level pulses are achieved only after the first amplification stage [8,11,12,14]. This is a typical energy level required for application in condensed matter physics.

 

Fig. 1 Experimental setup for all-inline inter-pulse DFG. A spectrally broadened pulse by filamentation is dispersed by a grating (G), collimated by a cylindrical mirror (CM) and finally back-reflected in the Fourier plane by a plane mirror (M). A slight vertical tilt of this mirror ensures vertical separation of the output beams. The 4-f setup enables to (i) compress the pulses, (ii) adjust the delay between pump and signal wavelengths, (iii) control the phase of the CEP stable DFG output and (iv) control the polarization state of the signal or the pump beam. The output is focused into a nonlinear crystal (NLC) to generate MIR pulses by DFG. Pump and the signal beams are separated by a dichroic filter (F).

Download Full Size | PDF

2. Experimental setup

The experimental setup presented in Fig. 1 was realized at the Advanced Laser Light Source - ALLS. First, the output of a Ti:Sa amplifier at 2.5 kHz, with 300 µJ pulse energy and 35 fs is focused in air by a 2 m lens to produce a stable filament for spectral broadening [26]. The broadening is performed by filamentation because it gives the highest throughput around the fundamental wavelength. Then, the filament’s output is collimated and sent into a folded 4-f setup.

It comprises a 235 l/mm grating to disperse light, a 0.5 m cylindrical mirror to collimate this dispersion and a plane mirror placed in the Fourier plane. The blue beams denote the pump part and the red beams the signal wavelengths, respectively. A cylindrical mirror is used instead of a spherical one to reduce the intensity in the Fourier plane (FP) which enables operation up to the multi-mJ regime. In this FP, all the spectral contents of the pulse are spatially dispersed and can be controlled independently. A quarter-wave plate can be added in front of the FP to rotate the polarization of the pump with respect to the signal. To compensate its material dispersion, a compensation plate at Brewster angle may be used in the other arm. Furthermore, angular tilting of the compensation plate ensures timing synchronization between both arms. Just like in a regular 4-f setup, small grating translations can be used to tune second order dispersion of few hundreds of fs^2 without introducing noticeable spatial chirp. All other DFG schemes rely on separate pulse compression steps. With a small amount of vertical tilt on the FP mirror it is possible to separate the input and the output beams after the 4-f setup vertically without introducing spatial chirp or significant aberration. Finally, the output beam is focused by a 2 m lens into a non-linear crystal to perform DFG. It is then collimated by a 1 m lens and it passes through a long-pass filter to separate the pump and the signal from the generated idler (LP1400 nm, Spectrogon). Furthermore, one could control the CEP by fine-tuning the phase delay between the pump and the signal by tilting the compensation plate or by tiny horizontal tilts of the mirror in the FP. Small horizontal mirror tilts correspond to a linear phase ramp in the spectral domain. We note, that for the sake flexibility in this proof of concept experiment, we used two different mirrors in the FP for the signal and the pump arm in most of the cases.

3. Results and discussion

First, we present the spectral characterization. The filamentation output was measured to carry a total energy of 240 µJ and its spectrum is shown in Fig. 2 as the bold black curve. We carefully adjusted the laser input parameters (chirp, energy, iris setting, focal length) to have the most stable filament. To measure this spectrum, two calibrated spectrometers have been used: one in the visible region (0.2 - 1.1 µm) and the other one in the IR (0.9 – 2.2 µm). The spectral parts transmitted through the 4-f setup are shown as the shaded areas in Fig. 2 and have a total transmission of 50%. The total spatial spread of the frequencies from 0.6 to 1.4 µm in the FP is 91 mm. In Fig. 2, the blue shaded part of the spectrum in the visible region denotes the bandwidth that passes through the quarter wave-plate. Only this part of the spectrum acts as the pump in the DFG process and the final energy incident onto the uncoated crystal was 70 µJ after all transportation optics. The spectral clipping of the waveplate could lead to a reduction of picosecond contrast. If the desired application requires a high contrast, one could move the wave-plate away from the FP to achieve an apodization effect or use a nonlinear scheme that doesn’t require a wave-plate in the 4-f. The red shaded part of the spectrum corresponds to the signal wavelengths. Both the pump and the signal show similar divergence properties after the 4-f setup and can be focused into one focal spot, shown as the insert in Fig. 2.

 

Fig. 2 Filament spectrum of the 4-f setup. Blue shaded area shows the pump bandwidth limited by the aperture of the wave plate in the 4-f setup. Red shaded area shows spectral content on the signal. The inset is the measured focal intensity distribution after the 4-f shaping unit.

Download Full Size | PDF

To demonstrate tailoring of the DFG we employed a 2 mm type II BBO (θ = 26.4°) and a 0.3 mm KTA (θ = 39.7°) crystals. Pulses at center wavelengths of 2 µm and 3.2 µm, respectively, were generated as is shown in Fig. 3. Due to the thin KTA crystal the DFG energy could not be measured while the BBO output reached up to 4.8µJ. This denotes a conversion efficiency of 7.7% (not accounting for Fresnel losses on the uncoated BBO). This efficiency compares well to other high energy µJ level DFG setups. Fattahi et al. achieved 12.4% in type I BBO and 4.4% in a type II BBO [15].

 

Fig. 3 Normalized DFG spectra. The spectrum at 2µm (solid line) was obtained with a BBO crystal and at 3.2µm (dashed line) with a KTA crystal

Download Full Size | PDF

For the temporal characterization, a homebuilt SHG-FROG [27] was used to characterize the MIR pulse as well as the pump and are shown in Fig. 4. The rms error values on the retrieved FROG spectrograms were 1.4% and 1.8% for the pump and the idler respectively. The pump and the DFG output have a FWHM duration of 20.6 ± 0.9 fs and 26.5 ± 1.7 fs, respectively. After filamentation, the 35 fs Ti:Sa pulses were compressed to 20.6 fs by shifting the grating position in the 4-f setup while the DFG output was compressed using 5 mm of Fused Silica (FS) to compensate the positive dispersion of the silicon IR filter. This pump duration of 20.6 fs is limited by the narrow spectral bandwidth due to the aperture of the wave plate in the 4-f. Using a less dispersive grating or a wave plate with a bigger aperture would give a shorter pump and possibly shorter DFG output. The DFG spectrum supports a transform limit of 21 fs corresponding to the 20.6 fs of the pump beam. However, uncompressed higher order dispersion from the silicon IR filter, the lens and the FS increase its duration to a 4 cycle pulse of 26 fs. Since the SHG-FROG measurement is performed by spatially splitting the beam horizontally, i.e. in the direction of the grating dispersion and the measured spectrograms are symmetric relative to delay 0, we can affirm that our beams didn’t carry significant spatial chirp.

 

Fig. 4 (a & e) Measured and (b & f) reconstructed FROG spectrograms; (c & g) independently measured (black) and retrieved spectra (red) and retrieved spectral phase (green); (d & h) retrieved pulse (black) and temporal phase (green) for the pump and the idler.

Download Full Size | PDF

We like to mention that our scheme does not suffer from possible CEP fluctuations introduced by the 4f setup since DFG is performed subsequently. Only relative phase differences between pump and signal are relevant. We characterized this phase jitter by linear interference between the first and the second order diffraction around 600 nm after the 4-f setup. The first order of diffraction follows the same path as the pump while the second order behaves like 1200 nm which is the signal path. Figure 5(a) displays the single shot interferogram and the corresponding phase stability of 60 mrad which corresponds to a timing jitter of 20 attoseconds. We point out that this jitter was measured in the presence of noticeable pointing fluctuations of the white light filament and without employing any active stabilization. Afterward, the CEP stability of the DFG was measured with an f-to-2f setup [28] by interfering the continuum generated in a sapphire plate with the second harmonic from a BBO crystal. Figure 5(b) shows the interferograms without feed-back stabilization and two mirrors in the FP corresponding to a CEP stability of 230 mrad. Each interferogram corresponds to 25 laser shots, which would correspond to a single shot rms value of 1150 mrad assuming purely statistical noise. We like to mention that the CEP stability was derived from white light generation in sapphire exhibiting strong peak-to-peak fluctuations of up to 50% while the DFG energy fluctuations were only about 1.25% rms. It is know from literature [29,30], that power fluctuations of the whit light lead to a reduced CEP stability. We believe that this energy-phase coupling in the measurement strongly contributes to our reported CEP error.

 

Fig. 5 Upper-panels, sequence of interferograms acquired (a) by using the 4-f setup as a single shot linear interferometer between two different diffraction orders at 600nm and (b) with an f-to-2f interferometer to measure the relative CEP jitter of the DFG output (25 shot average). Lower panels, reconstructed phase fluctuations. Since DFG is performed after the 4f setup, only the linear phase jitter in (a) should translate to CEP jitter.

Download Full Size | PDF

4. Conclusion

In conclusion, we report the generation of a 26.5 fs, 4.8 µJ and CEP stable pulse at the center wavelength of 2 µm in a single DFG stage. This was achieved by using a 4-f approach that combined the advantages of the inter-pulse and the intra-pulse schemes: independent control over the different wavelengths, spectral chirp, timing synchronization and polarization with an all inline setup. The 4-f setup served to compress the spectrally broadened pump pulse and the DFG output bandwidth was equal to the pump bandwidth.