Open Access
Add to BibSonomyAbstract
We employ AgGaSe2 for difference-frequency generation between signal and idler of synchronously-pumped picosecond / femtosecond OPOs at 80 / 53 MHz. Continuous tuning in the picosecond regime is achieved from 5 to 18 µm with average power of 140 mW at 6 µm. In the femtosecond regime the tunability extends from 5 to 17 µm with average power of 69 mW at 6 µm. Maximum single pulse energies of >1 nJ in both cases represent the highest values at such high repetition rates.
© 2015 Optical Society of America
1. Introduction
Picosecond and femtosecond synchronously-pumped optical parametric oscillators (SPOPOs) based on oxide nonlinear crystals pumped by mode-locked Ti:Sapphire lasers represent a unique class of ultrafast coherent sources operating at high (50-200 MHz) repetition rates, primarily in the near-IR part of the spectrum [1]. SPOPOs usually rely on uncritical phase-matching in periodically-poled ferroelectric oxide crystals such as lithium niobate or lithium tantalate setting an upper limit for the idler wavelength of 4-5 µm in the mid-IR. Their wavelength coverage can be extended into the mid-IR by difference-frequency generation (DFG) with convenient tuning possible thanks to the simultaneous variation of signal and idler wavelengths. For wavelengths exceeding 4-5 µm, non-oxide nonlinear crystals have to be employed in the DFG stage which exhibit smaller band-gap and two-photon absorption (TPA) limitations will come into play at tight focusing [2]. AgGaS2 (AGS), AgGaSe2 (AGSe), GaSe, GaS0.4Se0.6, and LiInSe2 (LISe) have been employed for DFG with such 800-nm pumped SPOPOs but operation was in all cases in the femtosecond regime, to obtain reasonable conversion efficiency [3].
The recent development of diode-pumped mode-locked solid-state and fiber lasers based on the Yb3+ ion provides new, more stable and cost-effective pump sources for picosecond and femtosecond SPOPOs. In contrast to Ti:Sapphire laser pumping such technology is scalable in average power. From optical damage considerations, it is more suitable for pumping SPOPOs built with oxide materials which then emit at longer wavelengths (degeneracy point around 2 µm) and are compatible with DFG schemes based on nonlinear crystals possessing higher nonlinearity and extended mid-IR transparency.
AGSe is such a nonlinear crystal and its application in DFG schemes based on ~800-nm Ti:Sapphire laser sources was indeed limited to long wavelengths due to birefringence and TPA restrictions: Mixing the signal and idler of a femtosecond SPOPO, tuning from 8 to 18 µm was achieved using AGSe but the maximum energy was only 12 fJ at 84 MHz [4]. In [5] a diode-pumped mode-locked at 42 MHz Yb:KGd(WO4)2 laser was used to pump a dual wavelength SPOPO and DFG between the two signal outputs was studied in GaSe and AGSe crystals. High conversion was achieved with AGSe in a narrow spectral range with maximum average power of 4.3 mW at 13.2 µm. All-fiber systems as ultrafast pump sources, however, have the advantage that they can be power scaled by Yb-fiber amplifiers, maintaining the high repetition rate. Here we study such a DFG scheme based on AGSe, mixing the signal and idler from a SPOPO pumped by picosecond / femtosecond Yb-fiber based sources emitting near 1 µm and demonstrate unprecedented average powers and single pulse energies in the mid-IR.
2. Relevant properties of AGSe
The commercially available AGSe is attractive for DFG with low energy tightly focused beams because of its high nonlinear coefficient (d36~35 pm/V for frequency doubling of 5.3 µm radiation), small spatial walk-off and broad transparency extending from ~0.78 up to ~18 µm. The band-gap of this chalcopyrite crystal is 1.83 eV which means that no TPA should be observed above 1355 nm. With an index of refraction of n~2.6 a typical figure of merit (d2/n3) is ~70 pm2/V2, roughly 6 times higher compared to its sulphide analogue AGS. The advantage of AGSe against GaSe which exhibits a similar transparency window but higher nonlinearity is the much lower spatial walk-off, the capability of directed growth and cutting at the desired orientation, as well as the availability of antireflection coatings.
Type-II eo-e phase-matching exhibits higher effective nonlinearity for DFG wavelengths exceeding 8 µm (deff~d36sinθ for type-I while deff~d36sin2θ for type-II). However, for shorter DFG wavelengths, deff approaches zero and there is no phase-matching below ~6 µm, see Fig. 1(a). For type-I eo-o phase-matching, the critical angle θ varies from 44° to 60° in the 5-18 µm DFG tuning range while deff changes from ~70% of its maximum value (~d36) near 18 µm to about 87% near 5 µm.
Fig. 1 (a) Internal phase-matching angle θ (black) and spatial walk-off angles ρ1,3 (red) for type-I (solid lines) and type-II (dashed lines) DFG in AGSe. (b) GVM parameters mixing signal and idler pulses from a 1034 nm pumped SPOPO. The indices 1,2,3 denote DFG, idler, and signal pulses. Sellmeier equations used are from [6].
In AGSe, the birefringence walk-off angle ρ3 remains between 9.8 and 11.8 mrad in type-I phase-matching, see Fig. 1(a), while ρ1 = 0. The spatial walk-off is similar in type-II phase-matching except at short DFG wavelengths (<7 µm) where both ρ1 and ρ3 decrease abruptly when the uncritical condition (θ = 90°) is approached. For comparison, the walk-off angle ρ3 in type-I eo-o GaSe is ~60 mrad in this DFG tuning range and it is even larger for type-II phase-matching.
The interaction length between the two input pulses is determined by the inverse group-velocity mismatch (GVM) parameter Δ23, see Fig. 1(b), and it is slightly longer for type-II phase-matching in AGSe. However, the temporal walk-off and consequently the pulse duration of the generated DFG pulses will be much larger in type-II phase-matching according to the calculated GVM parameter Δ12, also shown in Fig. 1(b). From this plot it can be seen that for ~2-ps pulse durations both effects can be ignored in type-I AGSe up to a crystal length of ~10 mm. For 200-250 fs pulses, AGSe thickness of 2 mm seems optimum in type-I phase-matching because the temporal walk-off will be comparable to the input pulse durations. At such sample thickness one can expect to maintain the input pulse durations in the DFG process except for the short wave part of the tuning range.
In the picosecond regime we used type-I AGSe crystals cut at φ = 45° and θ = 52° with an aperture of 3 × 5 mm2 which were anti-reflection coated for 1.7-2.7 µm and 5-15 µm on the input and output faces, respectively. Samples of 2, 5, and 10 mm length were compared but here we present only results obtained with the 10-mm long sample which delivered the highest DFG output. The type-I AGSe sample employed in the femtosecond regime was cut at φ = 45° and θ = 57°. It was uncoated, 2-mm thick, and had an aperture of 3 × 5 mm2.
3. Experimental set-up and results
In the picosecond regime, a commercial SPOPO (Levante IR, APE GmbH, Berlin, Germany) was used, pumped by an NKT Photonics (Birkerød, Denmark) Yb-fiber oscillator/amplifier operating at 1032.2 nm and 80 MHz with an average power of 7.8 W. The ~2.1 ps (Gaussian FWHM) pulses from the fiber laser system had spectral bandwidth (FWHM) of ~0.9 nm corresponding to a time-bandwidth product (TBP) of ~0.53. The SPOPO delivered up to 2.0 and 1.3 W of signal and idler average power, respectively, in the tuning range of 1380-1980 nm for the signal corresponding to tuning from ~4.1 to ~2.2 µm for the idler. The pulse duration from the SPOPO varied between 2.1 and 2.6 ps (Gaussian FWHM).
In the femtosecond regime, the DFG stage was driven by a similar commercial SPOPO (femtoLevante IR, APE GmbH, Berlin, Germany), pumped by an Amplitude Systemes (Pessac, France) Yb-fiber oscillator/amplifier operating at 1034 nm and 53 MHz with an average power of 5 W. The ~260 fs (Gaussian FWHM) pulses from the fiber laser system had spectral bandwidth (FWHM) of ~11 nm corresponding to a TBP of ~0.8. The SPOPO delivered up to 1.3 and 0.8 W of signal and idler average power, respectively, in the same tuning ranges. The pulse duration from the SPOPO varied between 200 and 250 fs (Gaussian FWHM). In both regimes the measured signal / idler spectral bandwidths resulted in a TBP~0.5-0.6, comparable or smaller than the TBP of the pump pulses.
Figure 2 shows the DFG set-up. After temporally overlapping signal and idler through a delay line, the two beams were spatially recombined with a dichroic mirror (reflecting for the signal in s-polarization and transmitting for the idler in p-polarization when slightly tilted) and focused with a 150 or 100-mm lens in the picosecond or femtosecond regime, respectively. Final adjustment of the collinearity and spatial overlap was accomplished optimizing the DFG average power by the two mirrors in the signal path before the focusing lens.
Fig. 2 Schematic of the DFG set-up: DM: dichroic mirror as beam recombiner, P: periscope for polarization rotation, L: focusing lens, R: retro-reflector in delay line (also adjusting the beam height), X: AGSe crystal, C: collimation (curved Au-mirror), F: 3.6-µm cut-on Ge-filter.
Figure 3(a) shows the average mid-IR output power obtained in the picosecond regime and the external quantum conversion efficiency calculated from the incident signal and output DFG powers. The output power has been corrected only for the Ge-long-pass filter transmission and the reflection coefficient of the two Au-coated mirrors used in front of the thermophile power meter. Near 6 µm, the single pulse energy reached 1.75 nJ, more than 5 times higher than previously achieved with a 4-mm thick LISe crystal [7]. The improvement in quantum conversion efficiency compared to LISe is similar. The single pulse energy and the average power (140 mW) represent the highest values ever demonstrated from a high-repetition rate ultrafast mid-IR DFG system [3]. The improvement in comparison to earlier work is related to the more powerful pump source but also to the longer interacting wavelengths which permit the use of AGSe. In addition, the low spatial walk-off permitted to use a relatively long sample with picosecond pulse durations. The wavelength range achieved is also much broader, see Fig. 3(a), in particular the long wave limit which did not exceed 12 µm with LISe [7].
Fig. 3 (a) Average power at 80 MHz obtained in the picosecond regime behind the AGSe crystal and external quantum conversion efficiency. (b) Spectral tuning versus DFG wavelength. The dotted curve in (b) indicates the measured spectral bandwidth.
The DFG spectra shown in Fig. 3(b), recorded with a 50-cm monochromator and 75 g/mm grating, demonstrate the achieved tunability. The upper wavelength limit was determined by the N2-cooled HgCdTe detector used, however, with the power meter tuning was measured up to 18 µm, Fig. 3(a). The average spectral bandwidth from Fig. 3(b) is ~10 cm−1. Note that the reduced FWHM at 15 µm is a result of an air (water) absorption feature. The recorded by a commercial autocorrelator (pulseCheck USB MIR, APE GmbH) traces could be fitted with Gaussian shapes, see Fig. 4, and this gives DFG pulse durations of less than 2 ps in the mid-IR. The resulting TBP is <0.6 which means that the DFG pulses are nearly Fourier-limited with quality similar to that of the input signal and idler pulses.
Fig. 4 Recorded DFG autocorrelation functions in the picosecond regime using TPA: (a) at 7 µm and (b) at 11 µm. The FWHM indicated refers to the autocorrelation traces.
Figure 5 shows the average mid-IR output power obtained in the femtosecond regime and the external quantum conversion efficiency calculated from the incident signal and output DFG powers. The average DFG power is decreasing with wavelength, a consequence of the wavelength dependence of the coupling coefficient for the nonlinear process although the effective nonlinearity is increasing. The decrease of the DFG power at 5 µm is related to idler absorption near 2.6 µm due to air humidity and the long beam path of ~2 m. Near 6 µm, the single pulse energy reached 1.3 nJ, an unprecedented level in the mid-IR for such high repetition rate femtosecond coherent sources which exceeds the previous best result with LISe [7]. Note that we were able to cover a very broad tuning range (5-17 µm) with a single AGSe sample, again much broader compared to LISe [7]. The quantum conversion efficiency from signal to DFG photons reached ~24% near 6 µm in Fig. 5.
Fig. 5 DFG average power (black line and circles) and calculated DFG external quantum efficiency (red line and squares) in the femtosecond regime.
The DFG spectra shown in Fig. 6(a) indicate bandwidths corresponding to DFG pulse durations of 200-250 fs if a time-bandwidth product of ~0.65 is assumed, similar to the one measured for the input signal and idler pulses. The upper wavelength limit for the recorded spectra in the femtosecond regime was determined by the detector noise level. From Fig. 6(b) one can conclude that the DFG pulses at 7.2 µm have a Gaussian FWHM of 305 fs (or ~280 fs assuming sech2 shape). With the corresponding spectral width of ~520 nm from Fig. 6(a) (red solid line) this gives a TBP of ~0.9 for Gaussian pulses which is roughly two times the Fourier limit. Thus, it can be concluded that the spectral bandwidth supports ~150 fs Gaussian pulses. It is also possible that some reshaping of the DFG pulses takes place in the nonlinear process in the presence of GVM, see Fig. 1(b), towards more square-shaped pulses which have larger TBP.
Fig. 6 (a) Recorded DFG spectra demonstrating the achieved spectral tunability in the femtosecond regime and (b) typical autocorrelation function (ACF) recorded at 7.2 µm. The red dashed line in (a) shows the spectral bandwidth in wavenumbers (right axis). The autocorrelation trace in (b) is fitted with a Gaussian function (red solid line) corresponding to a DFG pulse duration of 305 fs (FWHM).
4. Conclusion
The nonlinear crystal AGSe showed excellent performance in DFG of ultrashort pulses at high repetition rates by mixing the signal and idler outputs of Yb-fiber laser pumped SPOPOs. Tuning from 5 to 18 µm in the picosecond and from 5 to 17 µm in the femtosecond regime was achieved using a single crystal cut, with record high single pulse energies and average power for high repetition rate mid-IR ultrashort pulses. Near 6 µm, the average power of the picosecond version at 80 MHz reached 140 mW equivalent to 1.75 nJ single pulse energy while that of the femtosecond version at 53 MHz reached 69 mW corresponding to single pulse energy of 1.3 nJ. Both duration and spectral bandwidth of the input pulses are largely maintained in the DFG process. The nearly Fourier-limited DFG pulses will be useful for high spectral resolution time-resolved spectroscopy.
No damage to the AGSe samples was observed for the average intensities applied in the present experiment. Higher conversion efficiency can be expected with type-II phase-matching in AGSe albeit only above:8 µm. Further power scaling, with more powerful Yb-fiber pump sources, seems straightforward by increasing the beam sizes in the DFG nonlinear crystal. The upper wavelength limits of 17-18 µm in our experiment were determined by the sensitivity of the thermophile power meter (~0.5 mW). Detectors operating at longer wavelengths, however, already exist and suitable thickness of the AGSe sample will in principle enable DFG tuning up to at least 20 µm without serious absorption loss in the nonlinear crystal, which will allow one to fully utilize the tuning potential (signal and idler) of the present commercial SPOPOs.
Finally, it shall be mentioned that high repetition rate ultrashort pulses in the mid-IR have been also generated with AGSe employed directly in a SPOPO [3,8–10]. These AGSe SPOPOs were pumped between 1.54 and 1.577 µm by the signal pulses from oxide-based near-IR SPOPOs. Such tandem SPOPO schemes were also shown to be capable of generating single pulse energies on the nJ level both in the picosecond [8] and in the femtosecond [10] regimes. However, apart from the increased complexity requiring synchronization of three cavities, their main limitation is the restricted tunability which extended in the best case only up to 7.9 µm for the idler [9]. Thus, one of the most attractive features of AGSe, its extended transparency into the mid-IR, remains unexploited. The main reason for this limitation, which holds for any down-conversion parametric device besides DFG, is the reduction of the parametric gain away from degeneracy.
Acknowledgment
The research on AGSe development has received funding from the RFFI (Russia) under grant n° 13-02-96500 ”Growth and investigations of new nonlinear crystals for the near- and far-IR spectral ranges”.
References and links
1. M. H. Dunn and M. Ebrahimzadeh, “Parametric generation of tunable light from continuous-wave to femtosecond pulses,” Science 286(5444), 1513–1517 (1999). [CrossRef] [PubMed]
2. V. Petrov, “Parametric down-conversion devices: The coverage of the mid-infrared spectral range by solid-state laser sources,” Opt. Mater. 34(3), 536–554 (2012). [CrossRef]
3. V. Petrov, “Frequency down-conversion of solid-state laser sources to the mid-infrared spectral range using non-oxide nonlinear crystals,” Prog. Quantum Electron.in press.
4. J. M. Fraser, D. Wang, A. Haché, G. R. Allan, and H. M. van Driel, “Generation of high-repetition-rate femtosecond pulses from 8 to 18 µm,” Appl. Opt. 36(21), 5044–5047 (1997). [CrossRef] [PubMed]
5. R. Hegenbarth, A. Steinmann, S. Sarkisov, and H. Giessen, “Milliwatt-level mid-infrared (10.5-16.5 μm) difference frequency generation with a femtosecond dual-signal-wavelength optical parametric oscillator,” Opt. Lett. 37(17), 3513–3515 (2012). [CrossRef] [PubMed]
6. A. Harasaki and K. Kato, “New data on the nonlinear optical constant, phase-matching, and optical damage of AgGaS2,” Jpn. J. Appl. Phys. 36(2), 700–703 (1997). [CrossRef]
7. M. Beutler, I. Rimke, E. Büttner, V. Petrov, and L. Isaenko, “Difference-frequency generation of fs and ps mid-IR pulses in LiInSe2 based on Yb-fiber laser pump sources,” Opt. Lett. 39(15), 4353–4355 (2014). [CrossRef] [PubMed]
8. Ch. Grässer, S. Marzenell, J. Dörring, R. Beigang, and R. Wallenstein, “Continuous-wave mode-locked operation of a picosecond AgGaSe2 optical parametric oscillator in the mid infrared,” OSA TOPS on Advanced Solid-State Lasers (1996), Vol. 1, S. A. Payne and C. Pollock (eds), OSA, 1996, paper OP6, pp. 158–163.
9. S. Marzenell, R. Beigang, and R. Wallenstein, “Synchronously pumped femtosecond optical parametric oscillator based on AgGaSe2 tunable from 2 µm to 8 µm,” Appl. Phys. B 69(5-6), 423–428 (1999). [CrossRef]
10. R. Hegenbarth, A. Steinmann, S. Mastel, S. Amarie, A. J. Huber, R. Hillenbrand, S. Y. Sarkisov, and H. Giessen, “High-power femtosecond mid-IR sources for s-SNOM applications,” J. Opt. 16(9), 094003 (2014). [CrossRef]
