1. Introduction

Discovery and investigation of perspective nonlinear crystals for nonlinear optical devices down-converting the near-infrared (IR) solid-state laser frequency (∼ 1 µm) directly into the mid-IR range (3-30 µm) present one of the most important trends in nonlinear optics [1]. Such radiation has a great potential of applications in science, technology, medicine, and biology. It is interesting to generate high-energy, narrowband, mid-IR laser pulses spectrally positioned at the narrowband absorption lines of molecules, for example H₂O, N₂O, CO₂, CO, HCl, C₂H_6, CH₂O (formaldehyde). Detection of such molecules can be used for human breath analysis, environmental monitoring, and remote sensing. There are also possible applications in medical surgery, nano- and biotechnologies, and others [2]. Various phosphide, selenide, and sulphide non-oxide crystals having high second order nonlinearity and transparency in a wide spectral range from visible up to mid-IR range can be used for frequency conversion. These materials exhibit the smallest bandgap and two-photon absorption limitations at tight focusing [1]. On the other hand, using highly nonlinear phosphide crystals is limited by a relatively narrow transmission range due to their narrow bandgap (∼ 2 eV). This results either in losses at the near-IR pump wavelength (1-2 µm) because of single and two-photon absorption in ZnGeP₂ and CdSiP₂ [3,4], or absorption losses at the mid-IR idler wavelength longer than ∼ 6.5 µm in CdSiP₂ [4,5]. Selenide and sulphide crystals have wider bandgap and transmission range. A well-known AgGaS₂ crystal is widely used because of reasonably wide bandgap (2.7 eV) and transmission range (0.5-12 µm), but its thermal conductivity (1.5 W/m·K [1]) and damage threshold (0.2 J/cm² at nanoseconds [6] and 0.06 J/cm² at picoseconds [7]) are quite low limiting the output power or complicating the down-conversion devices. For example, in the work [7] a high (50 µJ) level energy output at 5-12 µm mid-IR picosecond pulses under pumping by a 4-mJ, 25-ps, 1.064-µm Nd:YAG laser has been obtained only in the complicated down-conversion system based on two 18 mm long AgGaS₂ crystals allowing to decrease the radiation fluence to values lower than the AgGaS₂ crystal damage threshold.

Among the non-oxide nonlinear optical crystals transparent beyond 5 µm, the orthorhombic lithium ternary chalcogenides LiGaS₂, LiGaSe₂, LiInS₂, and LiInSe₂ are unique crystals having the widest bandgap (wider than 3 eV) and transmission range (0.4-12 µm). These crystals have also high damage threshold, but their nonlinearity is relatively low [8]. Their linear and nonlinear optical properties have been characterized recently when the technology for growing large-size single crystals in good optical-quality had been developed [9]. Notwithstanding the nonlinear coefficient of LiGaS₂ and LiInS₂ is lower (6-7 pm/V) than of LiGaSe₂ and LiInSe₂ (9-11 pm/V), the LiGaS₂ crystal damage threshold at nanosecond pumping is significantly higher and amounts ∼3.3 J/cm² [10] against 1 J/cm² for the LiInS₂ crystal [11] and 0.4-0.5 J/cm² for the LiGaSe₂ and LiInSe₂ crystals [12–14]. In the work [14] all these crystals with an equal length of 3 mm were compared for femtosecond difference frequency generation (DFG) at 4-11 µm. It was demonstrated that the sulphide (LiGaS₂ and LiInS₂) crystals made it possible to obtain several times higher output energies (tens of nJ) for femtosecond pulses because of higher damage threshold. Picosecond DFG was realized earlier in the LiInSe₂ and LiInS₂ [15,16]. In [15], picosecond DFG in the LiInSe₂ crystal at 5-12 µm was realized with a high repetition rate (80 MHz), but the individual pulse energy was very low (< nJ). In [16], picosecond DFG in the high-damage-threshold LiInS₂ crystal (12 mm long) at 5-6.7 µm was realized with low repetition rate (10 Hz), but with high (a 100-µJ level) pulse energy output that allowed to apply it for the surface-specific sum-frequency generation vibrational spectroscopy of peptides. However this high-energy LiInS₂ down-converter required a very high energy picosecond pump source (25-mJ, 10-Hz, 1.064-µm Nd:YAG laser) and a complicated seeding system based on a frequency doubler and an optical parametric generator/amplifier.

Another interesting possibility to realize a narrowband seeder for the mid-IR DFG is stimulated Raman scattering (SRS) of the pump laser radiation generating the signal wave for DFG at the first Stokes SRS wavelength of λ_s = (λ_p⁻¹ – ν_R)⁻¹, where λ_p is the pump wavelength, ν_R is the Raman frequency (in cm⁻¹) of the Raman medium. In this case, the idler wavenumber λ_i⁻¹ in the DFG process will be equal to the Raman frequency ν_R (λ_i⁻¹ = λ_p⁻¹ – λ_s⁻¹ = ν_R). So, it is possible to select various Raman media for seeding of the DFG at various discrete mid-IR wavelengths. Recently, in the work [17], the Raman seeder based on the single-pass SRS conversion in deuterated benzene C₆D₆ (ν_R = 944.7 cm⁻¹) was proposed for the picosecond DFG in the GaSe. Generated radiation was spectrally positioned at a wavelength of 10.6 µm with narrower spectrum in comparison with the traditional DFG system [18] allowing to obtain more energetic (up to 3 µJ) narrowband picosecond seed pulses for high-power CO₂ laser amplifiers. Using the picosecond neodymium solid-state laser as a pump source allows also developing a very simple Raman seeder based on single-pass Raman conversion in crystals with various Raman frequencies. The point is that the single-pass SRS generation threshold in crystals under picosecond laser pumping is the lowest relative to optical damage threshold in comparison with the cases of nanosecond and femtosecond pumping. This is caused by relatively low optical damage threshold at nanosecond pumping and relatively high SRS generation threshold at femtosecond pumping (when pulse duration is shorter than picosecond dephasing time of Raman-active modes in crystals). In our previous work [19], utilization of a CaCO₃ crystal (ν_R = 1086 cm⁻¹) as the single-pass Raman seeder has been proposed and realized for the DFG at various mid-IR wavelengths using not only the first Stokes, but also higher order Stokes SRS components for the signal allowing to obtain the idler wavenumber equal to not only the Raman frequency ν_R, but also to its multiple harmonics. Low-energy (∼ 0.1 µJ) picosecond DFG in LiGaS₂ at 9.2 µm (λ_i⁻¹ = ν_R) and 4.6 µm (λ_i⁻¹ = 2ν_R) under the picosecond Nd:YAG laser pumping and the single-pass CaCO₃ Raman laser seeding has been obtained.

In this work, we present a comparative study of high-energy picosecond mid-IR DFG at discrete wavelengths of 4.6, 5.4, 7.5, 9.2, and 10.8 µm in a simple down-converter based on the 8 mm long crystals of LiGaS₂ or LiGaSe₂ under the 5-mJ, 20-ps, 1.064-µm Nd:YAG laser pumping and the single-pass crystalline Raman laser seeding. In comparison with our previous work [19], we have now achieved high energy (up to 50 µJ) output at various wavelengths in the range of 4.6-10.8 µm using not only CaCO₃, but also BaWO₄ and diamond as a crystalline Raman seeder generating at the first and second Stokes SRS components. High-energy picosecond mid-infrared frequency down-conversion in the LiGaS₂ and LiGaSe₂ crystals was experimentally realized for the first time to our best knowledge.

2. Experimental setup and down-converter characterization

The experimental setup of the frequency down-converter is presented in Fig. 1. The laboratory-designed, 10-mJ (max.), 20-ps, 5 Hz, 1.064-µm Nd:YAG laser was used as an excitation source for two stages of single-pass frequency conversion: the first stage was the Raman seeder and the second was the DFG in the LiGaSe₂ or LiGaS₂ crystal. Additional optical elements controlling the light-polarization, time-delay, beam splitting-combining, and focusing conditions between the pump laser, Raman seeder, and DFG crystals were used.

 

Fig. 1. The experimental setup of the frequency down-converter: λ/2: half-wave plate; M1, M2: concave mirrors, P: total reflection prisms, L1, L2: lenses.

Download Full Size | PDF

The polarized pump laser output beam was split by a birefringent wedge into the orthogonally polarized pump and signal beams (channels) and their energy ratio was optimized using a half-wave plate (λ/2). Total reflection prisms (P), plano-convex lenses (L1, L2), and concave mirrors (M1, M2) were used to align temporal and spatial overlaps of the combined pump and signal beams in the DFG crystal. L1 in the signal channel was also used to overcome the Raman generation threshold in the crystalline Raman seeder. M1 and M2 parameters were following: high transmittance (HT) @ 1.06 µm, high reflectivity (HR) @ Raman wavelengths (∼1.2-1.4 µm); concave, various curvatures (r = 100/500/1000 mm) in order to obtain requested signal beam diameter in the DFG crystal.

As the crystalline Raman seeder, various Raman crystals were used: BaWO₄, CaCO₃, and diamond having characteristics presented in Table 1. The pump laser pulse duration of τ_p = 20 ps is comparable with the vibrational dephasing time τ_R of the crystals (see Table 1), and therefore the SRS process is transient and determined by not only the Raman gain g, but also the ratio τ_p/τ_R. In this case, SRS generation threshold can be estimated [20] as

Table 1. Raman seeder characteristics [22–27] together with signal and idler wavelengths summary (g is the Raman gain, ν_R is the Raman frequency, Δν_R is the Raman mode linewidth, τ_R = 1/(π·c·Δν_R) is the Raman mode dephasing time).

View Table

Figures 2(a) and 2(b) shows dependences of the phase-matching φ-angle on the idler wavelength for eo-e type-II phase matching in the XY plane (Θ = 90°) at 1.064-µm pumping calculated from the Sellmeier equations data for the biaxial LiGaSe₂ [8] and LiGaS₂ [21] crystals. The points show the phase-matching angles at the idler wavelengths of λ_i = 4.6, 5.4, 7.5, 9.2, and 10.8 µm derived from the signal waves listed in Table 1. It can be seen that the phase matching angles at the idler wavelengths higher than 4.6 µm are quite close (a difference isn’t higher than 10°), and therefore only slight rotation of the LiGaSe₂ or LiGaS₂crystal can be used for DFG in the whole idler wavelength range of 4.6-10.8 µm.

 

Fig. 2. Calculated dependences of the phase-matching angle on the idler wavelength at the 1.064-µm pumping for (a) LiGaSe₂ and (b) LiGaS₂. The points correspond to various Raman laser seeding wavelengths.

Download Full Size | PDF

In order to present a comparative study, the LiGaSe₂ (Θ = 90°, φ = 34°) and LiGaS₂ (Θ = 90°, φ = 39°) crystals with the equal length of 8 mm and cross-section of 4×4 mm² were used. These samples were cut from the crystal boules grown on oriented seeds by the Bridgman-Stockbarger technique in a vertical setup (Institute of Geology and Mineralogy, SB RAS, and Novosibirsk State University). Precursors were Li (99.9%) and Ga, S, Se elementary components preliminarily purified to 99.999% [8,9,28]. The crystals were grown in a glassy carbon crucible, located inside a silica ampoule. The ampule was filled with Ar (1-1.5 atm). The low axial temperature gradient furnace (2.5 K/cm) was used for the crystal growth at a rate of 1 mm/day. As a result, 30 mm long single crystal boules with a diameter of 15 mm and high optical quality were grown and the experimental samples were polished. The samples were anti-reflection (AR) coated at the pump (1.064 µm) as well as signal (1.2-1.4 µm) spectral ranges.

Transmission spectra of both AR-coated crystals and their photos are presented in Fig. 3. The spectra were measured along the long side of 8 mm using a spectrophotometer (Shimadzu UV-3600) and FT-IR spectrometer (Nicolet). Slight noise at 5.5–7 µm is caused by water vapor. It can be seen that because of AR coating, the maximum transmittance was about 95% at 1.2-1.5 µm. The transmission spectrum of the LiGaSe₂ is generally wider then of the LiGaS₂ in agreement with previous works [e.g. 8]. We can see that transmission at 9.2 µm is only 9% in the LiGaS₂ and 47% in the LiGaSe₂. Generation at 10.8 µm is possible in the LiGaSe₂ only because radiation at this wavelength is almost fully absorbed in LiGaS₂ and transmittance in the LiGaSe₂ is 23% only.

 

Fig. 3. Transmission spectra of 8 mm long, AR-coated LiGaSe₂ and LiGaS₂ nonlinear crystals. Slight noise at 5.5–7 µm is caused by water vapor. Photos of the AR-coated crystals are presented in insets.

Download Full Size | PDF

3. Experimental results

Part of pump laser radiation was directed into the signal channel (Fig. 1) and converted via single-pass SRS in the diamond, BaWO₄, or CaCO₃ Raman seeder into the signal wave whose frequency corresponds to the first or second Stokes SRS component. Some part of the SRS radiation was generated in conical waves because of phase-matched parametric Raman interaction [24–26], but for further conversion, we used only the axial part of SRS radiation transmitted through an iris. Increasing the input pump pulse energy into the Raman seeder from the above mentioned SRS threshold value ($W_p^{th}$) by approx. 2 times resulted in increasing the SRS conversion efficiency into the axial component of the first Stokes Raman radiation up to 6% (90 µJ), 13% (52 µJ), and 14% (280 µJ) for the diamond, BaWO₄, and CaCO₃ crystal, respectively. Lower conversion in the diamond was probably caused by larger part of SRS radiation in conical waves that can be again explained by essentially higher divergence angle of spontaneous Stokes seeding in the short crystal. This crystal was also partially damaged. Further increasing the pump pulse energy in the Raman seeder led to the SRS conversion into the second Stokes also used as a signal wave for DFG. In the CaCO₃ crystal, the second Stokes SRS threshold was at 2.3 mJ, and the 100-µJ pulse energy in the axial second Stokes component was generated at the input pump energy of 2.4 mJ. In the BaWO₄ crystal, the maximum energy in the axial second Stokes component reached 44 µJ. The second Stokes from the diamond was not used in these experiments. Focusing conditions were not fully optimized here because our main interest was to generate signal wave for following DFG stage. Generally, the optimal focusing conditions are very different in crystals with different lengths, but it is necessary to note that SRS conversion efficiency can be increased up to 85% [24].

Various focusing conditions of the input pump beam into the DFG crystal were used because of different nonlinear coefficients and damage thresholds of LiGaS₂ and LiGaSe₂. “Small” and “large” pump spots of 350 and 780 µm in diameter (FW1/e²) were used (controlled by various lenses L2 f = 400/500 mm at various places in the pump channel). The input signal beam was kept at approximately fixed diameter of 650 µm by concave mirrors M1 and M2.

The results of comparative study of DFG at 4.6 µm in the LiGaS₂ and LiGaSe₂ crystals under the 1.064-µm picosecond Nd:YAG laser pumping and the 1.384-µm second Stokes CaCO₃ Raman laser seeding for the small (350 µm) pump spot and the optimized ratio between the signal and pump energy inputs are presented in Fig. 4. Pump and signal energies were measured by an energy probe (Coherent J25 MB-LE), the idler and low-energy signal by a high-sensitive energy probe (Coherent J10 MB-LE). Idler energy was measured behind a set of Germanium filters blocking pump and signal radiation.

 

Fig. 4. Conversion characteristics into the 4.6 µm idler in the LiGaS₂ and LiGaSe₂ at the small (350 µm) pump spot and the optimized ratio between the signal and pump energy inputs.

Download Full Size | PDF

It can be seen that in the LiGaSe₂ crystal with higher (∼ 10 pm/V [14]) nonlinearity, as the input (pump + signal) energy increases, the output energy growth was saturated at 12 µJ, conversion has its maximum of about 5% and even decreases down to 3.3% at further input pulse energy increase up to 320 µJ. Saturation of the output energy growth can be caused by back-conversion (sum-frequency generation) because of high energy in the idler. Increasing the input energy higher than 320 µJ was limited by an optical damage threshold of the LiGaSe₂ crystal at the input radiation fluence of 0.35 J/cm² causing darkening of the LiGaSe₂ crystal output face after several hours of operation and emergence of grey tracks at the output face after that. The appearance of the grey tracks was partially reversible as in the case of femtosecond excitation described in [14]. This optical damage effect is similar (but at 6 times higher threshold) to that observed earlier for AgGaS₂ at picosecond excitation [7] explained by some photochemical process induced by the laser beam. The output face damage can be also caused by self-focusing of the laser beam in LiGaSe₂.

Using the LiGaS₂ crystal having lower (∼ 6 pm/V [14]) nonlinearity at the same small pump spot made it possible to obtain unsaturated output energy growth at increasing input energy up to 560 µJ resulting in higher energy output up to 17 µJ and the conversion efficiency of 3%. The input energy was increased up to 560 µJ for the LiGaS₂ (against up to 320 µJ for the LiGaSe₂) because of higher measured optical damage threshold of 0.65 J/cm². The character of the damage was similar.

Similar characteristics were obtained for all output wavelengths (4.6, 5.4, 7.5, and 9.2 µm) at the small and large pump spots in the LiGaS₂ and LiGaSe₂ crystals. Figure 5 demonstrates the overview of the results for the small (350 µm) pump spot at the 4.6 and 5.4 µm idler wavelengths and for the large (780 µm) pump spot at the 7.5 and 9.2 µm idler wavelengths too.

 

Fig. 5. Energy characteristics of idler waves generated by DFG in the LiGaS₂ and LiGaSe₂.

Download Full Size | PDF

It can be seen that the small pump spot was better for the LiGaS₂ crystal with lower nonlinearity and allows obtaining 20-µJ output pulse energy of the 4.6 and 5.4 µm idler. The maximum input radiation fluence was close to the measured LiGaS₂ damage threshold of 0.65 J/cm² (at 20 ps) corresponding to the input energy of 0.62 mJ and the conversion efficiency of 3%. Using the small pump spot for the LiGaSe₂ crystal with higher nonlinearity resulted in saturation of conversion, probably because of the back-conversion effect (for the 4.6 and 5.4 µm idler). The large pump spot (780 µm) was optimal for the LiGaSe₂ crystal allowing to achieve the highest 50-µJ output pulse energy of the 7.5 µm idler and 20-µJ output pulse energy of the 9.2 µm idler (the lower energy is probably caused by the operation at the edge of the transmission range). The input radiation fluence was close to the measured LiGaSe₂ damage threshold of 0.35 J/cm² (at 20 ps) corresponding to the input energy of 1.7 mJ (conversion efficiency up to 3%). Higher output energy for the LiGaSe₂ can be explained by better mode matching between the large pump spot (780 µm) and the signal spot (650 µm) in the DFG crystal.

Figure 6 demonstrates the mid-IR DFG spectra and spatial profiles for the obtained idler waves in the LiGaSe₂ (similarly in the LiGaS₂). The spectra were measured using a monochromator (Oriel 77250) together with a liquid nitrogen cooled mercury-cadmium-telluride (MCT) detector (Judson-Teledyne J15). The spatial profiles were measured using a beam profiling camera (Spiricon Pyrocam III). Narrowband idler waves at the discrete wavelengths of 4.6, 5.4, 7.5, and 9.2 µm with a good beam quality at high pulse energies (10-50 µJ) were obtained. The generated waves linewidths were measured to be close to the monochromator resolution limit of < 2 cm⁻¹ (∼10 nm @ 7.5 µm) and they can be even narrower. It can be concluded that the idler wave linewidth should be comparable with the Raman mode linewidth (Δν_R = 1.2-1.6 cm⁻¹, see Table 1) of the crystalline Raman seeder.

 

Fig. 6. Spectra of pump, signal and idler waves. Insets: idler beam profiles at respective wavelengths. Intensity of the presented spectra were normalized.

Download Full Size | PDF

Generation at 10.8 µm was achieved in the LiGaSe₂ only and the output energy was at in the order of ∼100 nJ (close to the measuring probe resolution limit).

4. Conclusions

A comparative study of high-energy picosecond mid-IR difference frequency generation in the wavelength range of 4.6-9.2 µm in the LiGaS₂ and LiGaSe₂ crystals was presented. The developed frequency down-conversion system was pumped by the 10-mJ (max), 20-ps, 1.064-µm Nd:YAG laser. The narrowband signal wave for the DFG was generated by single-pass stimulated Raman scattering in Raman-active crystals: CVD-diamond, BaWO₄, and CaCO₃. The SRS conversion efficiencies and energies in the axial component of the first Stokes Raman radiation were up to 6% (90 µJ), 13% (52 µJ), and 14% (280 µJ) for the diamond, BaWO₄, and CaCO₃ crystal, respectively. Further increase of the pump pulse energy into the Raman seeder led to the SRS conversion into the second Stokes and the energies were 44 µJ and 100 µJ for the BaWO₄ and CaCO₃ crystal, respectively. Narrowband (<2 cm⁻¹) generation of ∼20 ps pulses with energy higher than 10 µJ at discrete wavelengths of 4.6, 5.4, 7.5, and 9.2 µm with the optical frequencies equal to the vibrational Raman frequency or its second harmonic of various crystalline Raman seeders was demonstrated. The LiGaSe₂ crystal with wider transmission band allowed to obtain also 10.8-µm generation, but with lower energy because of a transmission edge. Optimization of the pump spot and mode-matching between the pump and signal spots allowed to increase the output pulse energy at 7.5 µm up to 50 µJ in the case of LiGaSe₂. The LiGaS₂ crystal having higher damage threshold has potential for further increasing the output energy via the same optimization provided here for LiGaSe₂.