1. Introduction

The generation of few-optical-cycle light pulses in the mid-IR (3–6 μm) wavelength range is crucial for many applications ranging from time-resolved spectroscopy to high-field science. Such wavelength range overlaps several important vibrational transitions in molecules [1–3], as well as electronic transitions in highly correlated materials [4] (superconductors, colossal magnetoresistance materials) and intersubband transitions in quantum confined semiconductors [5]. In addition mid-IR pulses are expected to increase the cut-off energy for high harmonic generation [6].

Traditionally, mid-IR pulses are produced in two ways: (i) difference-frequency generation (DFG) between signal and idler of a near-IR optical parametric amplifier (OPA) [7–9], of an optical parametric oscillator [10, 11] or, more generally, between two pulses produced by different sources [12–16]; (ii) DFG between different frequency components of a single broadband pulse [17–20]. Mid-IR pulses have also been directly generated as idler pulses of an OPA, by tuning the signal as close as possible to the pump wavelength [21–25]. These OPA systems employed crystals with transparency range extended to 5–6 μm, such as KTiOPO₄, LiIO₃, LiNbO₃ and KNbO₃. Up to now, however, due to the quasi-monochromatic seeding, mid-IR OPAs have generated only relatively narrowband pulses, with transform-limited (TL) pulsewidths longer than 50 fs.

In this paper we show that several crystals with extended IR transparency, when pumped at 800 nm, display a broad phase matching bandwidth around 1 μm, allowing the direct generation of broadband idler pulses spanning the 3–5 μm wavelength range with duration down to two optical cycles. We experimentally demonstrate this configuration with LiIO₃, producing up to 2-μJ energy pulses with spectra supporting 30-fs TL duration, corresponding to approximately three optical cycles. Further advantages of our approach are energy scalability and the generation of pulses with a stable carrier-envelope phase (CEP) [26, 27].

2. Broadband mid-IR OPA

The basic idea underlying our configuration is to achieve broadband amplification of signal pulses at a carrier wavelength as close as possible to the pump wavelength, thus pushing the carrier wavelength of the resulting broadband idler as far as possible to the infrared.

In an OPA, the phase matching bandwidth is given to the first order by Δω ∝ ∣δ_si∣ where δ_si = 1/V_gs - 1/V_gi is the group velocity mismatch (GVM) between signal and idler. Broadband parametric gain is therefore achieved when the group velocities of signal and idler are matched [28]. Figure 1(a) plots δ_si as a function of signal wavelength for several nonlinear optical crystals with broad mid-IR transparency range, using type I phase matching.

 

Fig. 1. (a). Calculated signal-idler group-velocity mismatch at 800 nm pump wavelength for several nonlinear optical crystals with extended mid-IR transparency. (b) Group velocity mismatch between signal and pump for the same crystals. PPLN, periodically poled lithium niobate; PPSLT, periodically poled stoichiometric lithium tantalate.

Download Full Size | PDF

As expected, group velocity matching is achieved at the degeneracy point, where signal and idler share the same wavelength and polarization; interestingly, however, there is another group velocity matching wavelength, which occurs, depending on the crystal, between 950 and 1050 nm [29]. Qualitatively, this can be understood by recalling that these are high refractive index materials, with a zero dispersion wavelength (corresponding to a maximum of the group velocity) around 1.9–2 μm; one can therefore expect similar group velocities for the ∼1 μm signal and the ∼3.5 μm idler. In these crystals it is therefore possible to amplify a broad signal bandwidth around 1 μm and produce a correspondingly broad idler in the mid-IR. A first qualitative comparison among the crystals can be obtained by plotting (Figs. 2(a) and 2(b)) the frequency-dependent parametric gain assuming monochromatic and undepleted pump. Another important parameter in an OPA is the GVM between the signal/idler and pump, which defines the interaction length and thus the parametric gain. Figure 1(b) plots δ_sp as a function of signal wavelength for the same crystals shown in Fig. 1(a). For the crystals exploiting birefringent phase matching, GVM values are quite low for signal wavelengths around 1 μm, corresponding to the broadband phase matching condition, while the periodically poled crystals have higher GVM. This analysis indicates that LiIO₃ combines a broad phase-matching bandwidth, extended mid-IR tunability (see Fig. 2(b)) and low pump-signal GVM; this crystal has therefore been chosen for the experiments.

 

Fig. 2. (a,b). Calculated frequency dependence of the parametric gain for 2-mm-thick 800-nm pumped LiIO₃, KNbO₃ and PPSLT crystals, assuming phase-matching at zero GVM point; (c,d) complete simulation of signal and idler bandwidths for LiIO₃ according to the model described in the text.

Download Full Size | PDF

The previous qualitative predictions about the broadband mid-IR amplification have been confirmed by performing numerical simulations of a collinear OPA, pumped at 800 nm and employing a 2-mm-thick type I LiIO₃. We solved the coupled three-beam time-dependent nonlinear propagation equations in the plane wave approximation, considering only second order nonlinear interactions and including material dispersion to all orders. We used a broadband seed pulse (7-fs centred at 1 μm) to simulate the white light continuum (WLC) generated in sapphire by 800-nm 50-fs long pump pulses. Figures 2(c) and 2(d) show the calculated amplified signal and idler spectra, respectively; the idler spans the 3–5 μm wavelength range and supports a 22-fs TL pulsewidth, i.e. approximately two optical cycles.

3. Experimental results

The experimental setup of the mid-IR OPA is shown in Fig. 3. It is pumped by 300 μJ, 50 fs, 800 nm, 1 kHz pulses from a regeneratively amplified Ti:sapphire laser (Coherent Legend). A 3-μJ fraction of the energy is focused in a 2-mm-thick sapphire plate to generate a single-filament WLC, used as a seed. The spectral portion of the WLC around 1 μm is first amplified in a non-collinear OPA (seed generation stage) and then fed into the mid-IR OPA, which generates the 3–5 μm idler in a collinear geometry to avoid angular dispersion. For the seed generation we used two different configurations: (i) 800-nm pumped 2-mm-thick type I LiIO₃ crystal (θ = 21.5°); (ii) 400-nm pumped 1-mm-thick type I BBO crystal (θ = 29°) [30]. The first configuration provided broader bandwidth (28 THz) but lower energy (200 nJ), while the second configuration had an higher output energy (1 μJ) with a slightly narrower bandwidth (21 THz). Since LiIO₃ has a relatively low nonlinear coefficient we preferred to use the 400-nm pumped configuration that provides higher seed energy. For different crystals with higher nonlinearity, such as PPSLT, the 800-nm pumped seed generation stage could yield better results since it gives the best matching with the gain bandwidth of the IR OPA.

 

Fig. 3. Experimental setup; BS, beam splitter; DL, delay line; SHG, second-harmonic generation; WLG, white light generation; DM, dichroic mirror; Au, gold mirror. In the first stage it is possible to switch from 800-nm pumped LiIO₃ to 400-nm pumped BBO, substituting the λ/2 plate with a SHG crystal (BBO, 1 mm, type I, θ = 29°).

Download Full Size | PDF

The output of the seed generation stage was collimated and collinearly combined with the residual pump by a dichroic mirror. The mid-IR OPA used a 2-mm-thick type I LiIO₃ crystal (θ = 21.5°) and, pumped by 200-μJ 800-nm pulses, produced more than 8-μJ signal energy around 1–1.1 μm. After the OPA the idler beam was collimated by an uncoated germanium lens, that also provided filtering of the pump and signal beams. The idler power measured by a thermal power-meter, after taking into account the Fresnel losses introduced by the lens, was 2 mW (2 μJ pulse energy), in good agreement with expectations from the Manley-Rowe relations. The mid-IR spectra were measured with a grating monochromator and a lithium tantalate pyroelectric detector together with a lock-in amplifier.

Figure 4 shows a sequence of idler spectra obtained by changing the pump-signal delay and slightly varying the phase-matching angle. We obtained pulses tunable from 3 to 4 μm with an unprecedented combination of high energy (μJ-level) and broad bandwidth (≈30 fs TL duration). We note that the dip observed around 4.2 μm is due to atmospheric CO₂ absorption, so that even broader bandwidth could be achieved by purging the system. We could tune the system further to the IR up to 5 μm but with reduced bandwidth. Our numerical simulations predict even broader idler bandwidth and we believe that we are currently limited by the chirp introduced on the 1-μm seed by the refractive optics, which prevents temporal overlap of all the spectral components with the short pump pulse. Reducing this chirp by the use of reflective optics and/or stretching the pump pulses, we should be able to improve the pump-seed temporal overlap and amplify broader bandwidths, so as to approach the predicted ≈ 20 fs limit.

 

Fig. 4. Idler spectra from the second stage.

Download Full Size | PDF

The broadband mid-IR pulses are expected to have negative dispersion as a result of the nonlinear parametric amplification process, which induces in the idler a chirp opposite to that of the signal (positive), and of the propagation in the LiIO₃ crystal, which displays negative dispersion in the mid-IR. These pulses should be compressible to the TL by propagation in a properly chosen thickness of materials such as silicon, germanium or zinc selenide, having positive dispersion in the mid-IR spectral region [31]. At the highest pump intensity levels, we observed optical damage of the LiIO₃ crystal in the second OPA stage; we expect that more robust operation and further energy scaling should be achievable by crystals with slightly less favourable phase matching characteristics but larger nonlinear optical coefficient and higher damage threshold, such as KNbO₃ and PPSLT (see Fig. 1).

In conclusion, we have proposed and experimentally demonstrated the possibility of generating broadband high-energy mid-IR pulses from an OPA. We have shown that several crystals with extended IR transparency, when pumped at 800 nm, display a broad phase-matching bandwidth around 1 μm, allowing the generation of idler pulses spanning the 3–5 μm wavelength range. In a LiIO₃ crystal we have obtained 2-μJ pulses with bandwidths supporting three-optical-cycle durations. Moreover these pulses, being generated by a DFG process between two pulses derived from the same source [26, 27], are expected to have a stable CEP, which is an important feature in high field processes which are governed by the electric field rather than the intensity of the pulse.