ZnGeP WITH ITS transparency range between 2 and 12

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 10, OCTOBER 1997 1749 Parametric Generation of 1-ps Pulses Between 5 and 11 m with a ZnGeP Crystal Valentin Petrov, Yoshihito Tanaka, and Takanori Suzuki Abstract We use a novel picosecond source (a seeded optical parametric amplifier tunable near 3 m) to pump a type-ii ZnGeP 2 traveling-wave optical parametric generator. With a rather simplified two-pass arrangement, the tunability of the driving Ti:sapphire regenerative amplifier is extended continuously up to 11 m. As a result of the short pulse pumping and the pulse compression accompanying the parametric amplification, nearly bandwidth-limited pulses could be generated for the first time with this crystal. We report microjoule output energies with 20% quantum efficiency and unprecedentedly low (<100 MW/cm 2 for a crystal length of only 1 cm) parametric gain thresholds. Index Terms Frequency conversion, infrared measurements, infrared spectroscopy, nonlinear optics, optical parametric amplifiers, optical pulses. I. INTRODUCTION ZnGeP WITH ITS transparency range between 2 and 12 m and the highest second-order nonlinearity among all commercially available crystals is a very prospective material for mid-infrared (MIR) short pulse generation for nonlinear optical applications and time resolved spectroscopy of simple and complicated molecules in the condensed phase. The first single stage optical parametric generator (OPG) based on this nonlinear crystal in a traveling-wave configuration was pumped by mode-locked 100 150-ps pulses from Er :YAG (2.94 m) or Er :Cr :YSGG (2.79 m) lasers and yielded 17% of quantum efficiency in the type-ii interaction scheme [1], [2]. Later, with the same pump sources, tunability between 4 and 10 m with type-i interaction [3], [4] as well as temperature tuning [5] were demonstrated, and considerable improvement in the spatial and spectral quality of the signal and idler pulses was achieved by employing two-pass arrangements [6], [7]. In the latter case, parametric generation thresholds as low as 0.1 GW/cm for a 4-cm-long crystal were reported (the lowest value of all traveling-wave OPG s) which is an important advantage as compared to similar schemes based on GaSe [3], [4], [7], [8] or AgGaS [9], because operation far below the optical damage threshold is possible. Most recently, efficient picosecond frequency doubling in ZnGeP extended the field of its application in the ultrashort pulse technology [10]. Our interest in this material was motivated Manuscript received April 14, 1997; revised June 12, 1997.The work of V. Petrov was supported by the Research Development Corporation of Japan (JRDC) through a STA Fellowship. V. Petrov was with The Institute of Physical and Chemical Research (RIKEN), Saitama 351-01, Japan, on leave from the Max-Born-Institute for Nonlinear Optics and Ultrafast Spectroscopy, Berlin, D-12474, Germany. Y. Tanaka and T. Suzuki are with The Institute of Physical and Chemical Research (RIKEN), Hirosawa 2-1, Wako-shi, Saitama 351-01, Japan. Publisher Item Identifier S 0018-9197(97)07082-6. by its potential as a MIR frequency converter for Ti:sapphiredriven ultrafast laser amplifiers. These most widely spread high-power short-pulse sources nowadays provide superior stability and reproducibility at repetition rates ranging from 1 Hz to 250 khz and pulse durations from several picoseconds down to less than 100 fs. In addition, for a number of multiwavelength applications, the tunability of the Ti:sapphire amplifiers near 800 nm is an important advantage. Up to now, the longest MIR-wavelengths (about 5 m) achieved by direct pumping with Ti:sapphire-based short-pulse systems were produced employing MgO:LiNbO in an OPG [11] or in a seeded optical parametric amplifier (OPA) configuration [12] as well as by KNbO OPA s seeded by narrow-band radiation [13] or white light continuum [14]. KTiOPO -based OPA s are limited to somewhat shorter wavelengths [15] [18] and the same is valid for its isomorphs [19]. Although transparent up to 5.5 m, LiIO is not suitable for OPG or OPA applications because of its low damage threshold, and the only attempt to pump an AgGaS OPA near 800 nm resulted in very poor performance (extremely low conversion efficiency) because of the two-photon absorption [20], [21]. Obviously, cascaded processes remain the only alternative for frequency conversion to 5.5 m of Ti:sapphire-based high-power ultrafast systems. The difference-frequency generation (DFG) demonstrated previously with AgGaS [22], [23] and the output (signal and idler) of multistage nearinfrared OPA s pumped near 800 nm is characterized by its low efficiency and provides MIR pulse energies on the nanojoule level (of the order of 50 nj). To induce excessive population changes by an MIR pulse requires, however, single pulse energies on the microjoule level. The intrinsic limitation of the DFG efficiency originates from the optimum condition of having an equal number of photons at two wavelengths. On the contrary, an OPG/OPA system can provide much higher quantum efficiency and in addition is attractive for its ultimate simplicity. In this paper, we investigate a ZnGeP - based OPG/OPA pumped for the first time to our knowledge by pulses as short as several picoseconds. We choose in these initial experiments picosecond instead of femtosecond Ti:sapphire driving source for several reasons: 1) better spectral resolution can be expected with picosecond pulses since femtosecond pulses have bandwidths much broader than typical vibration linewidths of condensed phase interfacial molecules in this spectral region; 2) the optical damage mechanisms and threshold values in ZnGeP are still unknown for such short pulse durations; and 3) a high-energy OPA scheme optimized in the picosecond regime, to be used for 0018 9197/97$10.00 1997 IEEE

1750 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 10, OCTOBER 1997 Fig. 1. Experimental setup. SH: second harmonic; THG: third harmonic generation; SFG: sum frequency generation; MOPO: master oscillator power amplifier; MLN: MgO:LiNbO 3 ; =2: removable waveplate; L1 L3: 20-cm CaF 2 lenses; L4: 10-cm BaF 2 lens; M1: Au-mirror; M2: removable Al-mirror for spectral measurements; DM1: dichroic mirror; 95% reflecting near 800 nm and highly transmitting near 1 m; DM2: Er 3+ -laser (2.94 m) bending mirror on a CaF 2 substrate; DM3: 800 nm bending mirror on a CaF 2 substrate; F1: a combination of an 800 nm reflecting mirror on a CaF 2 substrate and 1-mm-thick Ge-filter with antireflection coatings; F2: 4 m cut-on layered filter; F3: a combination of neutral density, color glass, and reflecting filters for separation of the SFG signal; D1: InSb or HgCdTe detector; and D2: Si-photodiode. The 25-cm monochromator is equipped with a 66.66-grooves/mm grating blazed at 11 m. the first conversion step, was developed recently in this laboratory. We address the most critical issue in all the previous work done by Vodopyanov et al. [1] [7] namely, the pulselength bandwidth product which exceeded in all cases reported [6] the value of 30, even if type-ii interaction is chosen. Using shorter pump pulses provides the most elegant possibility to reduce this product down to the limit imposed by the crystal length without applying any spectral narrowing elements which could be rather complicated at wavelengths exceeding 5 m. Our ZnGeP OPG/OPA is pumped by 2.7-ps-long pulses produced by a seeded singlestage MgO:LiNbO OPA which is used to convert the output of a Ti:sapphire regenerative amplifier to the spectral region near 3 m. We demonstrate continuous tunability of the ZnGeP output (signal and idler) between 5 and 11 m by angle tuning with type-ii phase-matching in combination with variation of the pump wavelength. The estimated pulselength bandwidth product is 1 and the achieved quantum efficiency of 20% (amplifier stage) results in MIR output pulse energies as high as 4 J. II. EXPERIMENTAL SETUP The extremely high second-order nonlinearity of ZnGeP ( 75 pm/v) [24] is a very important prerequisite for effective OPG operation. Absorption, however, precludes the use of pump wavelengths below 2.5 m and that is why the widely spread and already commercially available nearinfrared -BaB O -based OPG/OPA s pumped near 800 nm (see [25]) cannot be used as pump sources for ZnGeP. In this paper, we employed a MgO:LiNbO seeded OPA operating near 3 m to transform the Ti:sapphire amplified output to the transparency window of ZnGeP. This OPA is described in detail elsewhere [12], and here we only briefly outline the most important features for the present experiment (Fig. 1). Synchronization of the Pockels cell of the regenerative amplifier, the -switched Nd:YAG laser whose second harmonic pumps the regenerative amplifier and the MOPO (master oscillator power amplifier, Spectra-Physics model 730 optical parametric oscillator) is achieved by a sequence of pulse/delay generators which is clocked by the output of the mode-locked Ti:sapphire laser (Spectra-Physics Tsunami, picosecond version). The wavelength of the Ti:sapphire oscillator was fixed in the present experiment at 795 nm. The MOPO, which is pumped by the third harmonic of a separate Nd:YAG laser produces narrow-bandwidth ( 0.2 cm ) pulses of 3- ns duration that are tunable between 0.44 and 1.8 m (signal and idler). A small fraction of the idler output near 1 m was used in the present experiment as a seed signal for the MgO:LiNbO OPA. The repetition rate of 10 Hz for the whole system is imposed by the -switched Nd:YAG lasers.

PETROV et al.: PARAMETRIC GENERATION OF 1-ps PULSES WITH A ZnGeP CRYSTAL 1751 The homemade regenerative amplifier is described in more detail in [26]. It is based on a ring resonator employing a 20- mm-long Ti:sapphire crystal. For the present experiment, we increased the output energy up to 5 mj ( 1.5 mj behind the compressor) by increasing the pump level to 45 mj at 532 nm. Energies as high as 1.4 mj were applied to the MgO:LiNbO OPA in an unfocussed beam of 3 mm diameter. The amplified pulse duration at 795 nm was measured by the cross correlator depicted in Fig. 1 employing a LiB O crystal for second harmonic generation, instead of the SFG, and removing F1 and F2 from the beam path. Assuming Gaussian pulse shape, we obtained a FWHM of 2.7 ps which results in pump intensities of 3 GW/cm in the 2-cm-long MgO:LiNbO crystal (we define here this spatially averaged intensity as 50% of the peak spatial and temporal value). The seed beam from the MOPO had a larger diameter of 5 mm and the seed energy (of the order of 0.2 mj) was adjusted to operate the OPA well into the saturated regime for best pulse-to-pulse stability. The filter F1 blocks both the pump and amplified seed (signal) pulses with the remaining (useful) energy of 60 J at the idler wavelengths near 3 m. Fig. 2 shows the spectrum of the MgO:LiNbO OPA output near 3.15 m and the cross correlation with a small fraction of the regenerative amplifier output at 795 nm. The group velocity dispersion (GVM) effects in the SFG crystal used for the cross correlation (0.24 ps/mm for type SFG) are negligible in this case. Assuming again Gaussian pulse shapes, we arrive at a FWHM of 2.7 ps, i.e., the Ti:sapphire pulselength remains unchanged in this first conversion step. The spectral bandwidth of 9 cm is also very close to the spectral bandwidth measured at 795 nm and results in a pulselength bandwidth product of 0.73 at 3.15 m. The ZnGeP crystal with dimensions of 6 6 10 mm used in the present experiment was obtained from ELAN (Russia). It was cut for type-ii phase-matching at 80 45 which results in an effective nonlinearity of for the interaction. The transmission spectrum of the crystal is shown in Fig. 3. Although specified by the supplier with antireflection coatings on both faces for the 2 8- m region, one can conclude that the antireflection coatings improved the overall transmission by no more than 20% with an additional slight modulation near 3 m when the measured transmission window is compared with previously published data of uncoated samples ([3], [4], [6], [7], [24]). The ZnGeP crystal is used in a double-pass configuration where in the first pass (OPG) the spontaneous parametric fluorescence is amplified from the noise level and the second pass (OPA) serves as a power amplifier. Because of the low transmission of the crystal (Fig. 3), a collimated geometry is not suitable for two passes. A focused pump beam geometry is employed in such a way that the crystal position for the second pass is near the focal point of L1, and L2 is used as a recollimating lens. The distance between the ZnGeP crystal and the retroreflecting mirror M1 amounts to 5.5 cm and this improves the aspect ratio of the two-stage amplifier by roughly one order of magnitude as compared to the single-pass scheme with a doubled crystal length. Thus, the main function of the double-pass arrangement is to suppress the off-axial parametric generation and to improve in this way the spectral and spatial (a) (b) Fig. 2. (a) Spectrum and (b) cross-correlation function of the MgO:LiNbO 3 OPA idler pulse. The spectrum is recorded in the fourth grating order using an InSb detector as D1 and the cross correlation is recorded using a 2-mm-thick MgO:LiNbO 3 crystal for SFG (see Fig. 1). In both cases, F2 is removed. Fig. 3. Transmission of the ZnGeP 2 crystal (length L = 10 mm) measured on a Fourier transform infrared spectrometer. quality of the output pulses. The GVM is rather low when pumping in the MIR (Fig. 4) and that is why the signal and/or idler pulses are not separated from the pump pulse during the two passes in spite of the uncompensated group delays. The calculation in Fig. 4 is based on the Sellmeier expansion coefficients from [27] and is limited to crystal angles between 70 and 90. In the case of 3.5 m, the additional

1752 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 10, OCTOBER 1997 Fig. 4. Inverse group velocity mismatch (GVM) in ZnGeP 2 for type oe! o interaction versus signal wavelength for three pump wavelengths ( p ) indicated in the figure. The solid lines represent the (1=v i 0 1=v p) dependence and the dashed lines represent the (1=v s 0 1=v p ) dependence where v i, v s, and v p are the idler, signal, and pump pulse group velocities. restriction originating from the upper transparency limit has been taken into account. As a whole, the potential of the crystal with this cut is from 4.5 to 6 m for the signal wave and from 5.7 to 11 m for the idler wave, which means that continuous tuning over this rather broad spectral range should be possible. III. RESULTS AND DISCUSSION We first studied the parametric fluorescence of the ZnGeP crystal for a single-pass configuration at a normal incidence using the total available pump energy at 3.15 m. The threshold of detection (about 10 nj for our pyroelectric energy meter) if used as a measure for the threshold of parametric fluorescence leads to very low values for the necessary pump intensity ( 0.1 GW/cm ) achievable even without focusing the pump beam. Using a 20-cm focusing lens, we measured 10 J of total energy (signal idler) just behind the crystal at pump intensities of the order of 1 GW/cm. If corrected for the Fresnel losses, this results in 20% efficiency of the single-pass scheme. The output was, however, very divergent with a solid angle ( criterion) of 0.03 sr. This is about six times lower than the estimation based on the pump geometry (aspect ratio) in the crystal. Since the Er -laser mirror, used as a dichroic beam splitter DM2 in Fig. 2 for the double-pass scheme, was not optimized for this purpose, the pump energy available and incident on the ZnGeP crystal was slightly above 50 J. The estimated pump intensities for the first and second pass amount to 0.15 and 0.7 GW/cm, respectively (Fresnel losses taken into account). At 0.15 GW/cm, the first pass provided a seed level of 0.75 J (signal idler). This value, however, was measured again just behind the crystal and only a small part of it seeds the second stage. The second pass (power amplifier) produced a total energy output of more than 3 J. This figure represents the useful output level in our case. Using an optimized dichroic mirror for DM2 would enable the extraction of the total available energy (4 J for signal idler). Additional consideration of the Fresnel losses in the second pass for all three interacting pulses results in the estimation Fig. 5. Signal and idler angle tuning curves of the ZnGeP 2 OPG at p = 3.15 m: Experimental results (circles) and calculation (curves). of 20% total efficiency for the second pass. We checked experimentally the seeding effect by several indirect methods: 1) by simulation of the second pass in a single-pass scheme with reduced pump energy; 2) by adjusting the M1 position for maximum output power; and 3) by placing a 4-mm-thick quartz plate between the ZnGeP crystal and M1 used as a high-frequency pass filter to suppress the seeding effect above 4 m. The last test provided us an evidence that more than 80% of the output is seeded by the first pass. The spectral data were recorded with L2 and F2 removed and using mirror M2 (see Fig. 1). The angle tuning results at 3.15 m are shown in Fig. 5 together with the corresponding calculated curves for the signal and idler waves. The agreement with the predictions of the Sellmeier expansion coefficients [27] is rather good with a slightly increasing discrepancy far from degeneracy. The energy level remains constant at larger angles and starts decreasing below 75 reaching, at an idler wavelength of 7.5 m, 50% of the value corresponding to normal incidence. We filled in the gap between 6 and 6.7 m (Fig. 5) with the idler wave by changing pump wavelengths down to 2.9 m. The shortest signal wavelength reached in this case was slightly below 5 m. Wavelengths longer than 8 m could be achieved increasing the pump wavelength. At 3.5 m, we observed retracing behavior and simultaneous generation of four wavelengths. The experimentally measured wavelengths at normal incidence 80 together with the calculated signal and idler tuning curves are shown in Fig. 6. The discrepancy in this case is larger as compared to pumping at shorter wavelengths: the different wavelengths measured correspond to a critical angle deviating by less than 3 (internal angle) from the calculated one. In general, retracing behavior is predicted also at pump wavelengths shorter or longer than 3.5 m. At shorter pump wavelengths, however, the second idler wave lies outside the transparency window of the crystal, whereas at longer pump wavelengths the retracing cannot be observed at normal incidence for the 80 cut. This is the reason why in previous work with Er -laser pumping no retracing behavior has been reported [6], [7]: we note, however, that the discrepancy of the measured and calculated wavelengths found in these papers is reminiscent of the behavior we observe

PETROV et al.: PARAMETRIC GENERATION OF 1-ps PULSES WITH A ZnGeP CRYSTAL 1753 Fig. 6. Retracing behavior of the ZnGeP 2 OPG at p = 3.5 m. Calculation (curves) and measured wavelengths at normal incidence (circles). Fig. 7. Typical spectra of the ZnGeP 2 OPG at the signal wavelength recorded in the second order of the grating. in Fig. 6 for the one pair of pulses. This indicates some decreasing accuracy of the Sellmeier expansion coefficients at idler wavelengths exceeding 8 m. At 11 m, the energy measured was less than 10% of the value corresponding to the shorter wavelength idler. We estimate this to be the longest wavelength achievable with this crystal because of the transparency cutoff edge (see Fig. 3). The drastic drop of the energy at this wavelength is caused in our case also by the substantially reduced transmission of DM2 (Fig. 1). We used pump wavelengths up to 3.5 m, this limit being set by the increasing transmission of DM2 for the pump. For reliable estimation of the spectral bandwidths, we had to rely on the InSb detector since it had much better detectivity than the HgCdTe detector available to us during these experiments. That is why we present here measurements only in the 5 5.3- m range at 2.9 3 m. Typical spectral widths (FWHM) of 35 cm were estimated at such signal wavelengths (Fig. 7). The GVM between signal and idler in this spectral region amounts to 100 fs/mm. We note that the GVM varies only weakly with wavelength for the type- II interaction (Fig. 4) so that there is no physical reason for substantial deviation of the bandwidth (e.g., by more than 50%) in the whole tuning range for the signal wave whereas the energy conservation condition leads to the same conclusion at the idler wave. Rough estimations with our HgCdTe detector at longer wavelengths provided only an upper limit of the order of 50 cm because of the insufficient spectral resolution in that case. The temporal properties of the generated signal and idler pulses were studied by cross-correlation measurements with a small fraction of the regenerative amplifier output at 795 nm (Fig. 1). For this purpose, a 1-mm-thick AgGaS crystal cut at 55 (type interaction) was applied. Even at this thickness the GVM in this crystal had to be taken into account when deconvolving the cross-correlation curves using the Gaussian pulse shape assumption. Fig. 8 shows the results of the cross-correlation measurement at one signal and two idler wavelengths. In all three cases, the FWHM of the curves is 3 ps, from which deconvolution yields a nearly constant pulse duration (FWHM) of 1 ps. Although this was a typical estimation for most of the cross-correlation data, in some cases Fig. 8. (a) (b) (c) Cross-correction functions of the output of the ZnGeP 2 OPG (signal or idler) with the Ti:sapphire pulses at 795 nm. (a) Signal wavelength s = 5.8 m achieved at p = 3.15 m, (b) idler wavelength i = 6.9 m achieved at the same pump wavelength, and (c) idler wavelength i = 7.6 m produced at p = 3.34 m. In all three cases, the FWHM of the curves is 3 s. we measured cross-correlation widths shorter than 3 ps. Note the cross-correlation method is less reliable for two pulses with distinctively different pulse duration requiring more precise autocorrelation measurements in the subpicosecond range. The 1 ps estimation for the signal/idler pulse duration leads to a pulselength-bandwidth product of which compares rather well with this product for the pump pulses. This already is an improvement by a factor of 30 as compared to previous work with ZnGeP [6]. The pulse shortening obtained at both signal and idler wavelengths can be attributed only to temporal gain narrowing in the high gain limit where the gain coefficient is defined by (, and are the corresponding indices of refraction). As shown in [28], the introduction of the spatially averaged pump intensity, which equals 1/2 of the maximum value, is a rough approximation for the overlap of beams deviating from the plane-wave assumption, allowing the use of the resulting gain coefficient for calculation of the achievable pulse shortening in the small-signal limit. Using 0.7 GW/cm, we arrive at 7cm. Following the analytical approximation in the limit of weak saturation [29], we obtain a compression factor of ( 2.6 which is in excellent correspondence with the measured value. Thus the efficient compression effect observed in the present experiments can be attributed to a

1754 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 10, OCTOBER 1997 greater extent to the second stage of amplification. We note that more precise simulation of the experimental results is complicated in our case by the unknown pulse parameters after the first pass. The approximate analytical theory [29] predicts a minimum achievable pulse duration of where is the final pulse duration (signal or idler) and. Thus, if we assume the initial pulse duration before the second pass to be equal to the pump pulse duration, the minimum achievable pulse duration after the second pass can be estimated to be 0.63 ps from the GVM of 100 fs/mm. Pulse shortening can be easily seen also in the frequency domain in terms of gain-induced spectral broadening effect of the bandwidth of the parametric process. For a monochromatic pump, a first-order consideration (linear dependence of the phase mismatch on the spectral bandwidth assumed) leads to 46 cm. Note that in the limit of DFG where the spectral bandwidth is determined solely by temporal walk-off effects a smaller value of 29 cm is predicted. Thus we conclude that the experimentally measured bandwidths are determined by the gain under the phasemismatching condition (essentially by the GVM, the crystal length, and the pump level). The contribution of the pump bandwidth through the frequency-dependent gain is estimated to be negligible because of the exponential dependence of the amplification factor and the narrow pump spectrum. The off-axial parametric generation and especially its effect on the bandwidths can also be neglected because of the twostage geometry. The pump beam divergence in the second pass has a contribution to the bandwidths which is about five times smaller than that of the GVM mainly due to the small birefringence. From the estimated gain coefficient, we calculate a small-signal amplification factor of 2.3 10 at the signal (idler) wavelength for the second pass. Simple geometrical consideration taking into account the measured divergence of the first pass indicates that only several percent of the signal and idler photons generated in the first pass survive. This leads to substantially lower estimation for the amplification factor than in the small-signal case. In the simplified geometry used (see Fig. 1) saturation effects could not be studied directly. However, indirect measurements with a single pass revealed that the second pass is well into the saturation regime where an increase of the pump energy by a factor of 2 results in a similar increase of the output (in the small-signal limit the latter should be of the order of 100). Saturation analysis is additionally complicated by the spatial variations due to the exponential dependence of the parametric gain. At the conversion efficiencies reported for the second pass, however, smearing of the complex beam profiles can be expected. It is thus clear from the above estimations that the actual amplification factor should be smaller than the smallsignal value which ignores pump depletion. Thus a smaller value for the saturated gain should be used for the calculation of the bandwidths, which will lead to estimations lying between the unsaturated value of 46 cm and the DFG limit of 29 cm, in reasonable agreement with the experimentally measured bandwidths in Fig. 7. The estimated parametric generation threshold (for better accuracy with a single-pass scheme) leads to values of the order of 0.2 GW where is the maximum (spatial and temporal) of the threshold pump intensity. Such values of are substantially lower ( 7 times if compared with results for type-ii phase-matching and 2 times if compared with results for type-i phase-matching) than previous estimations with 100-ps pump pulses at 2.8 m [6]. This discrepancy cannot be explained by absorption losses at 2.8 m nor Fresnel losses, since our estimation leads to unrealistically low amplification factors at the threshold for parametric generation in this experiment. We note that our measurement of the threshold pump intensity should be regarded as an upper limit on one hand because of the method of detection and on the other hand because of the not well-defined spatial profile of the pump pulses. Nevertheless, we conclude on the basis of this upper limit that some revisions of the exact value of might be necessary. Note that previous estimations based on the saturation behavior of a picosecond OPG [1] [4] yielded a larger value for (88 pm/v) than the one we adopted in the present work. Also, in a recent frequency-doubling experiment [10], the assumption of 126 pm/v was necessary in order to fit the measured conversion efficiency. Obviously further work with better defined spatial beam profiles and wavelengths well inside the transparency range of the crystal is necessary in order to derive more reliable results about the second-order nonlinearity of ZnGeP. The output powers in the present paper exceed 1 MW at both signal and idler wavelengths. The output beams are, however, far from diffraction-limited, mainly because of the poor pump beam spatial quality. Nevertheless, we estimated that focusing with optics should provide MIR intensities as high as 5 GW/cm. Substantial improvement of the spatial properties is expected if another seed source is used instead of the MOPO (Fig. 1), which is expected to produce a diffraction-limited output of the MgO:LiNbO OPA. The pulse stability of the ZnGeP OPG is determined to a great extent by the Ti:sapphire regenerative amplifier since the parametric amplification takes place under the condition of saturation. The long-term (minutes) fluctuations are limited to 20%. IV. CONCLUSION We studied the performance of ZnGeP as a type-ii OPG with a novel pump source which enabled for the first time to our knowledge the generation of 1-ps pulses with this crystal that exhibits an extremely high figure of merit. At a crystal length of only 1 cm, we report higher efficiency and lower threshold than in previous work whereas the reduction of the pulselength/bandwidth product down to the limit set by the crystal dispersion revealed the potential of this material for time-resolved spectroscopy in the MIR. Nearly bandwidth-limited signal and idler pulses were generated with a very broad tunability ranging between 5 and 11 m which is complementary to the tuning range of the pump pulses. The obtained microjoule energy levels make the MIR output suitable as pump pulses for surface sum-frequency generation. Improvement of the useful output energy by at least a factor of

PETROV et al.: PARAMETRIC GENERATION OF 1-ps PULSES WITH A ZnGeP CRYSTAL 1755 2 at the same pump level should be possible using an optimal beam splitter (see Fig. 1) and better antireflection coatings on the nonlinear crystal. The higher efficiency and the ultimate simplicity are important advantages of the presented scheme as compared to previous arrangements based on DFG and different pump sources. At the given pump pulsewidth, a crystal length of about 1.5 2 cm in a double pass is expected to be optimum since such a crystal would allow further reduction of the bandwidth without affecting the pulse-shortening effect and in addition may further reduce the threshold. This estimation is based on the GVM properties of the crystal when pumped at slightly above 3 m, where the GVM was identified as the main factor limiting the bandwidth. The applied pump intensities were far below the damage threshold of ZnGeP. We conclude that there are no principle limitations for extension of the present results to the femtosecond regime where crystal lengths of the order of 2 mm can be used. With typically 100 200-fs pulses produced by currently available Ti:sapphire amplifiers, safe operation with the same efficiency as in the present work can be expected. REFERENCES [1] K. L. Vodopyanov, V. G. Voevodin, A. I. Gribenyukov, and L. A. Kulevskii, Picosecond parametric superluminescence in the ZnGeP 2 crystal, Bull. Acad. Sci. USSR, Phys. Ser., vol. 49, pp. 146 149, 1985. [transl. from Izv. Akad. Nauk SSSR, Ser. Fiz., vol. 49, pp. 569 572, 1985]. [2], High efficiency picosecond parametric superradiance emitted by a ZnGeP 2 crystal in the 5 6.3 m range, Sov. J. Quantum Electron., vol. 17, pp. 1159 1161, 1988 [transl. from Kvantovaya Elektron., Moscow, vol. 14, pp. 1815 1819, 1987]. [3] K. L. Vodopyanov, L. A. Kulevskii, V. G. Voevodin, A. I. Gribenyukov, K. R. Allakhverdiev, and T. A. Kerimov, High efficiency middle IR parametric superradiance in ZnGeP 2 and GaSe crystals pumped by an erbium laser, Opt. Commun., vol. 83, pp. 322 326, 1991. [4] K. L. Vodopyanov, Parametric generation of tunable infrared radiation in ZnGeP 2 and GaSe pumped at 3 m, J. Opt. Soc. Amer. B, vol. 10, pp. 1723 1729, 1993. [5] K. L. Vodopyanov, Yu. A. Andreev, and G. C. Bhar, Parametric superluminescence in a ZnGeP 2 crystal with temperature tuning and pumping by an erbium laser, Quantum Electron., vol. 23, pp. 763 765, 1993. [transl. from Kvantovaya Elektron., Moscow, vol. 20, pp. 879 882, 1993]. [6] K. L. Vodopyanov and V. G. Voevodin, Type I and II ZnGeP 2 travelling-wave optical parametric generator tunable between 3.9 and 10 m, Opt. Commun., vol. 117, pp. 277 282, 1995. [7] K. L. Vodopyanov and V. Chazapis, Extra-wide tuning range optical parametric generator, Opt. Commun., vol. 135, pp. 98 102, 1997. [8] K. L. Vodopyanov and V. G. Voevodin, 2.8 m laser pumped type I and type II travelling-wave optical parametric generator in GaSe, Opt. Commun., vol. 114, pp. 333 335, 1995. [9] T. Elsaesser, A. Seilmeier, W. Kaiser, P. Koidl, and G. Brandt, Parametric generation of tunable picosecond pulses in the medium infrared using AgGaS 2 crystals, Appl. Phys. Lett., vol. 44, pp. 383 385, 1984. [10] J. M. Auerhammer, A. F. G. van der Meer, P. W. van Amersfoort, Q. H. F. Vrehen, and E. R. Eliel, Efficient frequency doubling of ps-pulses from a free-electron laser in ZnGeP 2, Opt. Commun., vol. 118, pp. 85 89, 1995. [11] S. Lin and T. Suzuki, Tunable picosecond mid-infrared pulses generated by optical parametric generation/amplification in MgO:LiNbO 3 crystals, Opt. Lett., vol. 21, pp. 579 581, 1996. [12] S. Lin, Y. Tanaka, M. Aono, and T. Suzuki, Optical parametric amplification using the phase-matching retracing behavior in MgO:LiNbO 3 for the generation of intense widely tunable mid-infrared pulse, Jpn. J. Appl. Phys., vol. 36, pp. 3510 3514. 1997. [13] V. Petrov and F. Noack, Mid-infrared femtosecond optical parametric amplification in potassium niobate, Opt. Lett., vol. 21, pp. 1576 1578, 1996. [14] J. D. Kafka and M. L. Watts, A potassium niobate OPA pumped by an amplified Ti:sapphire laser, in Springer Series in Chemical Physics, P. F. Barbara, J. G. Fujimoto, W. H. Knox, and W. Zinth, Eds. Berlin, Germany: Springer-Verlag, 1996, Ultrafast Phenomena X, vol. 62, pp. 38 39. [15] V. Petrov and F. Noack, Tunable femtosecond optical parametric amplifier in the mid-infrared with narrow-band seeding, J. Opt. Soc. Amer. B, vol. 12, pp. 2214 2221, 1995. [16] G. R. Holtom, R. A. Crowell, and X. S. Xie, High-repetition-rate femtosecond optical parametric oscillator-amplifier system near 3 m, J. Opt. Soc. Amer. B, vol. 12, pp. 1723 1731, 1995. [17] D. E. Gragson, D. S. Alavi, and G. L. Richmond, Tunable picosecond infrared laser system based on parametric amplification in KTP with a Ti:sapphire amplifier, Opt. Lett., vol. 20, pp. 1991 1993, 1995. [18] D. E. Gragson, B. M. McCarty, G. L. Richmond, and D. S. Alavi, High-power broadly tunable picosecond IR laser system for use in nonlinear spectroscopic applications, J. Opt. Soc. Amer. B, vol. 13, pp. 2075 2083, 1996. [19] V. Petrov, F. Noack, and R. Stolzenberger, Seeded femtosecond optical parametric amplification in the mid-infrared spectral region above 3 m, Appl. Opt., vol. 36, pp. 1164 1172, 1997. [20] P. Hamm, C. Lauterwasser, and W. Zinth, Generation of tunable subpicosecond light pulses in the midinfrared between 4.5 and 11.5 m, Opt. Lett., vol. 18, pp. 1943 1945, 1993. [21] P. Hamm, S. Wiemann, M. Zurek, and W. Zinth, Highly sensitive multichannel spectrometer for subpicosecond spectroscopy in the midinfrared, Opt. Lett., vol. 19, pp. 1642 1644, 1994. [22] F. Seifert, V. Petrov, and M. Woerner, Solid-state laser system for the generation of midinfrared femtosecond pulses tunable from 3.3 to 10 m, Opt. Lett., vol. 19, pp. 2009 2011, 1994. [23] M. K. Reed and M. K. Steiner Shepard, Tunable infrared generation using a femtosecond 250 khz Ti:sapphire regenerative amplifier, IEEE J. Quantum Electron., vol. 32, pp. 1273 1277, 1996. [24] P. D. Mason, D. J. Jackson, and E. K. Gorton, CO 2 laser frequency doubling in ZnGeP 2, Opt. Commun., vol. 110, pp. 163 166, 1994. [25] V. Petrov, F. Seifert, O. Kittelmann, J. Ringling, and F. Noack, Extension of the tuning range of a femtosecond Ti:sapphire laser amplifier through second-order nonlinear frequency conversion processes, J. Appl. Phys., vol. 76, pp. 7704 7712, 1994 and references therein. [26] H. Ohde, S. Lin, A. Minoh, F. O. Shimizu, M. Aono, and T. Suzuki, Tunable picosecond infrared pulses generated by stimulated electronic Raman scattering of a mode-locked Ti:sapphire laser in potassium vapor, Appl. Phys. B, vol. 62, pp. 15 19, 1996. [27] G. C. Bhar, L. K. Samanta, D. K. Ghosh, and S. Das, A tunable parametric ZnGeP 2 crystal oscillator, Sov. J. Quantum Electron., vol. 17, pp. 860 861, 1987 [transl. from Kvantovaya Elektron., Moscow, vol. 14, pp. 1361 1363, 1987]. [28] V. Petrov and F. Noack, Narrow-band seeding of an optical parametric amplifier in the femtosecond regime, Opt. Quantum Electron., vol. 28, pp. 1105 1110, 1996. [29] S. A. Akhmanov, A. S. Chirkin, K. N. Drabovich, A. I. Kovrigin, R. V. Khokhlov, and A. P. Sukhorukov, Nonstationary nonlinear optical effects and ultrashort light pulse formation, IEEE J. Quantum Electron., vol. QE-4, pp. 598 605, 1968. Valentin Petrov, photograph and biography not available at the time of publication. Yoshihito Tanaka, photograph and biography not available at the time of publication. Takanori Suzuki, photograph and biography not available at the time of publication.