A compact, self-compression-based sub-3 optical cycle source in the $3\mbox{--}4\,\mu {\rm{m}}$ spectral range

Development of laser sources, which emit ultrashort pulses in the mid-infrared spectral range is dictated by the emerging applications in ultrafast vibrational spectroscopy, mid-infrared nonlinear optics and strong-field physics [1]. Generation of mid-infrared pulses with durations as short as few optical cycles is exclusively based on the optical parametric amplification, which supports large amplification bandwidths necessary to maintain short pulse widths, see e.g. [2].

Technically simple, low cost and easily implementable setups for generation of ultrashort mid-infrared pulses rely on using widely spread amplified femtosecond Ti:sapphire driving lasers. Several configurations that involve optical parametric amplification, difference frequency generation (DFG), or both, have been demonstrated so far. The first configuration is based on either collinear or noncollinear optical parametric amplification in bulk and periodically poled nonlinear crystals, extracting the idler wave, which lies in the mid-infrared spectral range. Here the fundamental laser wavelength (800 nm) serves as a pump, while the seed signal is provided by either supercontinuum (SC) generation in bulk dielectric media or using the idler pulses produced by the near-infrared optical parametric amplifier (OPA). To this end, OPAs based on nonlinear crystals which possess transparency and phase matching in the mid-infrared, such as potassium titanyl phosphate (KTP) and potassium titanyl arsenate (KTA) [3, 4], potassium niobate (KNbO₃) [3–8], lithium niobate (LiNbO₃) [4, 7–9], lithium iodate (LiIO₃) [10] and periodically poled stoichiometric lithium tantalate [11] crystals were demonstrated to provide pulses with durations from sub-100 fs to few optical cycles in the 2–5 μm wavelength range. The second configuration makes use of DFG between the signal and idler pulses from the near-infrared OPA (e.g. that based on commonly used BBO crystal), see e.g. [12]. The third configuration employs DFG between broadband pulses from second harmonic-pumped noncollinear OPA and the fundamental laser pulses, providing few optical cycle pulses with stable carrier envelope phase, which are subsequently amplified in a broadband OPA [13–15].

Nonlinear propagation of intense mid-infrared pulses in wide bandgap solid-state dielectric materials featuring anomalous group velocity dispersion (GVD) provides an efficient pulse self-compression mechanism, resulting from the interplay between the self-phase modulation and anomalous GVD, that could be readily employed to broaden the pulse spectrum beyond the OPA gain bandwidth and further shorten the pulse duration. The numerical simulations suggest that self-focusing and filamentation of femtosecond mid-infrared pulses in dielectrics may lead to pulse self-compression down to a single optical cycle [16]. Moreover, such filamentation regime leads to formation of self-compressed quasistationary spatiotemporal light bullets, as verified experimentally in various dielectric materials, such as fused silica [17–19], sapphire [20], CaF₂ and BaF₂ [21] and BBO [22], yielding the shortest self-compressed pulsewidths of just sub-two optical cycles for the input wavelengths around 2 μm.

However, fully evolved self-compressed light bullets contain just a relatively small fraction of the input pulse energy and exhibit dispersive broadening as they leave the nonlinear medium and propagate in the free space [20], and these features may appear undesirable for a range of practical applications. Therefore a more attractive realization of anomalous GVD-induced self-compression mechanism relies upon using a shorter nonlinear medium and a larger input beam, extracting the self-compressed pulse before the filamentation regime sets in [23]. Such self-compression mechanism appears universal and is demonstrated to work equally well with mid-infrared optical parametric chirped pulse amplification systems providing tens-of-μJ pulses at high repetition rate [24] and multimilijoule pulses with multigagawatt peak power [25], by compressing the output pulses to sub-3 optical cycles in few mm thick YAG crystal.

In this work, we demonstrate a simple and compact Ti:sapphire laser-pumped mid-infrared light source, which involves DFG between the signal and idler pulses from a commercial near-infrared OPA, subsequent optical parametric amplification in LiIO₃ crystal and self-compression by the nonlinear propagation in dielectric crystals featuring anomalous GVD, and which delivers sub-3 optical cycle pulses with sub-30 μJ energy in the $3\mbox{--}4\,\mu {\rm{m}}$ spectral range.

The experimental scheme is depicted in figure 1(a). The entire setup is driven by regeneratively amplified Ti:sapphire system (Spitfire PRO, Newport-Spectra Physics), which delivers 100 fs pulses at 800 nm with an energy up to 3 mJ at a repetition rate of 1 kHz. A fraction of the laser energy (0.9 mJ) was used to pump a commercial BBO crystal-based OPA (Topas-C, Light Conversion Ltd.), which provided the near-infrared signal at 1.30 μm and idler at 2.08 μm pulses, with energies of 175 μJ and 115 μJ, and pulsewidths of 75 fs and 85 fs, respectively.

Zoom In Zoom Out Reset image size

The difference frequency generation between the o-polarized signal and e-polarized idler pulses was performed in a 1 mm thick KTA crystal, cut for type II phase matching ($\theta =43^\circ$, $\phi =0^\circ$), which provides broad conversion bandwidth in the wavelength range of interest [26]. Proper timing between the input signal and idler pulses entering the DFG crystal was set by inserting a 1 mm thick BBO crystal (PD), oriented out of phase matching, hence compensating their temporal walk-off originating in BBO OPA. The DFG stage delivered o-polarized 60 fs pulse with a central wavelength of 3.5 μm and with a FWHM spectral width of 280 nm and an energy of 11 μJ, as shown by dashed curves in figures 1(b) and (c). Here, the spectra were measured using a home-built scanning prism spectrometer with PbSe detector, whereas the cross-correlation measurements were performed by the sum-frequency generation in a thin (20 μm thickness) BBO crystal using 30 fs, 720 nm reference pulse from a second harmonic-pumped noncollinear OPA (Topas-White, Light Conversion Ltd.). The difference frequency pulse was almost perfectly transform-limited, as verified by the zero-phase Fourier transform of the spectrum.

Thereafter, the difference frequency pulse at 3.5 μm was amplified in LiIO₃ OPA, which was pumped by the e-polarized fundamental laser pulse at 800 nm. As an OPA we used 1 mm thick LiIO₃ crystal cut for type I phase matching ($\theta =21^\circ$), which provides naturally broad amplification bandwidth due to intrinsically small group velocity mismatch between the o-polarized signal (at 3.5 μm) and e-polarized idler (at 1.04 μm) waves [10, 13]. The pump and injected signal beams were crossed at a very slight angle ($0.7^\circ$), so as to easily separate the pump, amplified signal and idler beams without introducing additional optical elements. With the pump pulse energy of 0.7 mJ (the pump beam FWHM diameter of 1.6 mm, the estimated pump intensity was 200 GW cm⁻²), the 3.5 μm pulse was amplified up to 35 μJ energy, simultaneously producing the idler pulse at 1.04 μm with an energy of 100 μJ, with the quantum efficiency of the OPA of 18%. Relatively low conversion efficiency of the OPA is due to relatively short crystal length, that was chosen to be shorter than the walk-off length between 800 nm and 3.5 μm pulses (the group velocity mismatch between these two is 54 fs mm⁻¹). On the other hand, the choice of the crystal length guaranteed that the OPA did not introduce any noticeable distortions as seen from a comparison of spectra and cross-correlation functions of the 3.5 μm pulse after DFG and OPA stages, presented in figures 1(b) and (c). No optical damage or degradation of the crystal was observed with the above parameters of the pump, as verified by stable and robust day-long operation of the OPA.

The amplified 3.5 μm pulse was self-compressed by the nonlinear propagation in 4 mm thick CaF₂ plate. In doing so, the OPA output beam of 1.3 mm FWHM diameter was focused by plano-convex BaF₂ lens (L) with a focal length of 100 mm onto CaF₂ plate, which was mounted on a translation stage and made movable along the beam propagation path, so allowing fine adjustment of the input beam diameter and intensity, as depicted in figure 2(a). 3 mm thick anti-reflection coated ZnSe plate was placed before the focusing lens so as to approximately compensate for the GVD introduced by BaF₂ lens of 3.5 mm thickness. The outgoing beam was then collimated by Ag-coated concave mirror with the curvature radius of 200 mm and directed to the measurement apparatus for characterization of the temporal and spatial profiles, spectrum and energy transmission. The temporal intensity profile of the self-compressed pulse was characterized by means of sum-frequency generation-based frequency-resolved optical gating (SFG-FROG), the beam profile was measured with the pyroelectric CCD camera (WinCamD, model FIR2-16-HR), and the transmitted energy was measured using the energy meter (Ophir) with a pyroelectric detector (PE9-SH).

Zoom In Zoom Out Reset image size

The energy of the pulse at the input face of CaF₂ plate was 29 μJ, as measured after beam steering mirrors (not shown), ZnSe plate and focusing lens. Figure 3 shows the retrieved pulsewidth and the transmission of CaF₂ plate as functions of the sample position z with respect to the geometric focus (z = 0 mm). Two distinct regimes of self-compression were uncovered, which were distinguished by the absence or presence of the nonlinear losses. A lossless self-compression regime (with energy throughput efficiency of 93.5% and the energy losses occurring just due to Fresnel reflections from the input and output faces of CaF₂ plate) was found for converging ($z\lt 0$) as well as for diverging ($z\gt 0$) input beams, when CaF₂ plate was located several mm on both sides from the geometric focus.

Zoom In Zoom Out Reset image size

Figure 4 shows the measured and retrieved SFG-FROG traces, and compares the spectra and intensity profiles of the input and self-compressed pulses at $z=-6.5\,\mathrm{mm}$, just before the onset of beam filamentation. The duration of the self-compressed pulse of 31 fs is equivalent to 2.7 optical cycles and was retrieved within grid size of 128 × 128 pixels and a reconstruction error of 0.7%. Lossless self-compression yielded a homogenous Gaussian-shaped beam profile of the self-compressed pulse, as shown in figure 2(c), attesting essentially one-dimensional dynamics of the nonlinear propagation [27, 28], where self-compression of the pulse stems from the sole interplay between the self-phase modulation and anomalous GVD, without the onset of any other nonlinear effects, such as self-focusing, nonlinear absorption and generation of free electron plasma. The measured standard deviation of short-term (over 5 min intervals) energy fluctuations of the self-compressed pulse was 1.7%. Very similar self-compression dynamics, although yielding slightly longer self-compressed pulse, was also measured for diverging input beam, as the CaF₂ plate was moved further out of the geometric focus.

Zoom In Zoom Out Reset image size

Approaching the geometric focus, the filamentation regime sets in, as indicated by rapidly decreasing transmission in figure 3(b), as due to the multiphoton absorption and absorption by the free electron plasma. As a result, the input Gaussian beam is reshaped into a filament with characteristic intensity distribution, consisting of an intense core and a ring-shaped periphery, as shown in figure 2(d), which develops due to the interplay between self-focusing and multiphoton absorption [29]. In the filamentation regime, the input pulse continues to self-compress; the maximum self-compression of the pulse down to 19 fs, that equates to 1.6 optical cycles was measured when CaF₂ plate was located 1 mm before the geometric focus, where energy transmission decreased down to 66%. As compared to the case of lossless self-compression, the self-compressed peak contained just 33% of the transmitted energy, while the rest of energy was located in well-pronounced sidelobes, which developed due to spatiotemporal reshaping of the entire wave packet. Large sidelobes and even conditional pulse splitting are the characteristic features of the self-compressed light bullet at the early stage of its formation (at short propagation lengths) in the conditions of large anomalous GVD [30].

The self-compression experiments were also performed using YAG and BaF₂ plates, which produced rather similar results, however yielding somewhat lower self-compression factors. Nevertheless, these materials were found more suitable for compressing the pulses at 3 and 4 μm, respectively, as will be shown below.

Filamentation in CaF₂ produced a broadband SC emission, with the broadest SC spectrum obtained at the point of maximum self-compression. The emergence of SC is indicated by the occurrence of a distinct blue spot in the far field, which is easily detected by visual means, while the entire SC spectrum extends up to $5\,\mu {\rm{m}}$ and is composed of two separate bands, as measured with a home-built scanning prism spectrometer with Si, Ge and PbSe detectors and illustrated in figure 5(a). A particularly strong SC signal is detected around the carrier wavelength in the $2\mbox{--}5\,\mu {\rm{m}}$ range. In the visible range, SC spectrum features intense the so-called blue peak centered at 410 nm, with an ultraviolet cut-off at 385 nm, as estimated at the 10⁻⁴ intensity level. The characteristic feature of the SC spectrum in CaF₂ is the absence of the spectral components in between, in line with the observations reported in [31]. The SC spectrum also features a prominent peak centered at $1.75\,\mu {\rm{m}}$ that is the second harmonic generated in ZnSe, and a deep double dip around 4.25 μm as due to strong absorption band of atmospheric CO₂. A more spectrally homogenous SC, as shown in figure 5(b), was generated by replacing CaF₂ with BaF₂ plate of the same length and keeping the same position of the nonlinear medium with respect to the geometric focus. The SC spectrum in BaF₂ features a broad blue peak centered at 630 nm and provides a continuous spectral coverage from 510 nm in the visible to beyond $5\,\mu {\rm{m}}$ in the mid-infrared, which is the long-wave detection limit of our spectrometer.

Zoom In Zoom Out Reset image size

Finally, we verified the potential of wavelength tunability of the developed source and the possibility to perform lossless self-compression of the pulses at other wavelengths. 70 fs pulses with an energy of 31 μJ were obtained after DFG and OPA stages as the output wavelength was set at $3\,\mu {\rm{m}}$. Three-fold self-compression of these pulses down to 23 fs (2.3 optical cycles) was achieved in the lossless self-compression regime in 3 mm thick YAG plate, whose input face was located 8 mm before the geometric focus ($z=-8$ mm). Figure 6 presents the measured and retrieved SFG-FROG traces, spectrum and retrieved temporal profile of the self-compressed pulse with the center wavelength of $3\,\mu {\rm{m}}$. The superimposed dashed curves in figures 6(c) and (d) illustrate the spectrum and retrieved temporal profile of the input 70 fs pulse. Somewhat longer, 97 fs pulses with an energy of 14 μJ were generated as the DFG and OPA output wavelength was tuned to $4\,\mu {\rm{m}}$. The best self-compression was achieved in 4 mm thick BaF₂ plate located at $z=-3\,\mathrm{mm}$, yielding the compressed pulsewidth of 42 fs (not shown), which is equivalent to 3.2 optical cycles. In the latter case, a generally lower self-compression factor was in part attributed to CO₂ absorption, which considerably distorts the spectrum of the self-compressed pulse before it reaches the measurement setup.

Zoom In Zoom Out Reset image size

In conclusion, we developed a simple and compact, Ti:sapphire laser-pumped mid-infrared source, which provides sub-3 optical cycle pulses with sub-$30\,\mu {\rm{J}}$ energy in the $3\mbox{--}4\,\mu {\rm{m}}$ spectral range. The developed source employs difference frequency generation in KTA crystal between the signal and idler waves from a commercial near-infrared BBO OPA and subsequent parametric amplification in LiIO₃ crystal, which provides naturally broad amplification bandwidth. Thereafter the parametrically amplified pulses are self-compressed by the nonlinear propagation in wide bandgap dielectric media featuring anomalous GVD. In particular, we demonstrated lossless (with the energy throughput of above 90%) self-compression of sub-100 fs pulses down to sub-3 optical cycles in YAG, CaF₂ and BaF₂ plates of few mm thickness due to the interplay between the self-phase modulation and anomalous GVD and without the onset of self-focusing effects, as verified by the measurements of the output beam profile. Pulse self-compression down to sub-2 optical cycles was demonstrated at 3.5 μm in the filamentation regime in CaF₂, with the energy throughput of 66%. The developed source perfectly fits for diverse studies of ultrafast light–matter interactions in the mid-infrared spectral range in dielectric (e.g. as demonstrated by ultrabroadband supercontinuum generation in CaF₂ and BaF₂ in the present study), as well as in semiconductor media.

This research was funded by a grant No. APP-8/2016 from the Research Council of Lithuania.