1. Introduction

Optical parametric amplification (OPA) is an efficient approach to generate tunable pulses from a fixed laser source. The development of femtosecond OPA techniques has been particularly rapid in the past 10 years since mode-locked Ti:sapphire lasers were demonstrated and commercialized [1]. Near infrared femtosecond OPAs were commonly based on β-barium borate (BBO) crystal, and in many cases were extended directly from the experience of long-pulse pumped OPAs, i.e., via the conventional configuration of OPG/OPA [2, 3]. By resorting to noncollinear phase-matching scheme and white-light seeding, BBO-based OPA pumped by the second-harmonic pulse of a Ti:sapphire laser is also ideal to obtain tunable femtosecond pulses in the visible [4–8]. To date, BBO-based OPA provides a routine technique for generating tunable femtosecond pulses covering the spectral range from ~0.45 μm to 2.6 μm, and is commercially available. Mid-infrared (MIR) femtosecond OPAs, on the other hand, rely mainly on the nonlinear crystals like LiNbO₃ (LN) and KTiOPO₄ (KTP), which could not be designed simply by duplicating those BBO-based OPAs [9–11]. Due to both lower damage threshold and larger nonlinear refraction index compared with BBO crystal, these MIR crystals will suffer severe problems from surface damaging and self-focusing if OPG is adopted, which necessitates a sufficiently high pump intensity.

Several approaches have been demonstrated to develop reliable MIR femtosecond OPAs. In one approach, a near infrared femtosecond OPA generates intense signal pulse for the successive MIR OPA, while tunable MIR idler pulse is obtained by difference-frequency-generation (DFG) between the pump and signal pulses in a MIR crystal [11]. The intense input of the signal pulse obtained from a BBO-based OPA makes efficient DFG realizable at moderate pump intensity. While the OPA setup is complicated and the conversion efficiency may also be limited due to an additional BBO-based OPA, this DFG approach can generate tunable MIR femtosecond pulses with Fourier-transform limit. In another important approach, a narrow-band laser source is adopted as the external seed of a femtosecond OPA to avoid the damage problems with the OPG/OPA scheme, which offers advantages of compactness and high efficiency [12, 13]. When external seeding is strong enough, efficient OPA could be accomplished even in a single-pass configuration. Unfortunately, wavelength tunability of an externally seeded OPA can only be accomplished by tuning either the pump or the seeding source and adjusting the crystal orientation accordingly, which is practically inconvenient in a daily operation. Moreover, the femtosecond signal pulse is accompanied with a background of the narrow-band seed, and the spectral purity degrades obviously indicated by a sharp spike on the broad spectrum of the femtosecond pulse. In this paper we propose and study a novel hybrid seeded MIR femtosecond OPA that introduces external CW seeding at the signal wavelength in the first stage of OPA and DFG between the idler and pump pulses in the second stage. The hybrid seeded OPA basically does not introduce additional complexity in the setup, while it generates signal pulses with clean spectra and provides direct wavelength tunability like OPG/OPA scheme. The hybrid seeded femtosecond OPA is free from damage since the external seeding lowers the pump threshold dramatically, and its conversion efficiency has been proved to be as high as that of the conventional seeded OPA in the saturation regime.

2. Experiments

The hybrid seeded OPA adopts a two-stage amplification configuration with each of the stages seeded by two different wavelengths, i.e., the signal and idler, respectively. Two or more amplification stages are commonly adopted in femtosecond OPAs, which offers the flexibility in arranging the pump intensities through the stages to optimize the conversion efficiency and minimizing the effect of group velocity mismatch (GVM) via optical delay lines. The architecture of the hybrid seeded femtosecond OPA is schematically shown in Fig. 1, which is basically similar to a conventional seeded OPA. It employs external CW seeding at the signal wavelength in the first stage. Unlike a conventional seeded OPA, the spectral component at the signal wavelength from the first stage is blocked, while the idler pulse is left entering into the second OPA stage where DFG takes place between the idler and pump pulses. Thus, temporally and spectrally clean pulses both at the signal and idler wavelengths can be obtained. Importantly, the CW seeding scheme may provide direct tunability to some extent in the hybrid seeded femtosecond OPA.

 

Fig. 1. Schematic of the hybrid seeded femtosecond OPA. BS, beam splitter; DM1–DM2, dichroic mirrors for 800 nm and 1μm; DM3, dichroic mirror for 800 nm and 3.5 μm; LN, MgO:LiNbO₃ crystal; DL, optical delay line; Ge, germanium plate; and all unmarked, silver mirrors.

Download Full Size | PDF

In our hybrid seeded femtosecond OPA, two 6-mm-thick and 8-mm-thick uncoated MgO:LiNbO₃ crystals (supplied by CASTECH) both cut at θ=42° for type I (oo ➔ e) phase matching were used in the first and second OPA stages, respectively. A 50-fs pump laser at 800 nm, 1-kHz repetition rate with pulse energy of 200 μJ was produced from a commercial Ti: sapphire regenerative amplifier. The pump pulse was divided by a beam splitter with reflectance of ~40%, and telescoped to obtain the pump intensities of ~60 GW/cm² and ~40 GW/cm² for the first and second stages, respectively. The designed pump intensities are well below the damage threshold of LiNbO₃ crystal. The first OPA stage was seeded with a 1064-nm CW diode-pumped Nd:YVO₄ laser operating in the polarized TEM₀₀ mode with output power adjustable up to 500 mW. Since a CW seed was used, synchronization between the pump and seed sources was not required. Double-pass configuration was adopted in the first stage, with the second pass deviating in the uncritical phase-matching plane of the OPA crystal. A 3-mm-thick germanium plate, AR-coated for mid-infrared range of 3μm ~ 5 μm, was inserted between the two stages to block the spectral component at the signal wavelength while pass the idler pulses from the first stage effectively. Germanium has an absorption coefficient as high as ~10⁴ cm^-1 near 1 μm, which guarantees perfect blockage of the signal component with a 3-mm-thick plate. Our measurement using the CW Nd:YVO₄ laser at its full power operation, limited by the sensitivity of the power meter used, indicated that the transmittance near 1 μm of the germanium plate should be lower than 10^-10. In the conventional seeded OPA, the narrow-band seeding is naturally accompanied with the output femtosecond signal pulses, which is regarded as a source of noises to degrade the spectral and temporal qualities of the signal pulses. In the hybrid seeded OPA with the germanium plate, on the other hand, the narrow-band seeding background of the output femtosecond signal pulses is greatly reduced by a factor of more than 10¹⁰. Thus the spectral and temporal clean signal pulses can be obtained in the hybrid seeded OPA.

 

Fig. 2. Typical signal spectra from the conventional (a-c) and hybrid (d-f) seeded OPAs. The spectra were measured at three different orientations of the crystals, and the adjusting angle for each was ~1 mrad (external). The narrow peaks in (a-c) indicate the seeding wavelength at 1064 nm.

Download Full Size | PDF

In our experiments, only the signal pulses were monitored and characterized. A CW seed level of ~100 mW was sufficient to saturate the hybrid seeded OPA, with pulse-to-pulse stability of ~±4%. Figure 2 shows the output signal spectra at different crystal orientations. As a comparison, we also measured the corresponding signal spectra from a conventional seeded OPA (Fig. 2 (a-c)). Clearly, a strong spike at the seeding wavelength superposes on the output signal spectrum in the case of conventional seeded OPA. By blocking the spectral component at the signal wavelength from the first stage and mixing the pump pulse with the temporally and spectrally clean idler pulse in the second stage, the final signal spectrum is well shaped and does not contain the component of the CW seed (Fig. 2 (d-f)). Such a CW seeded OPA may provide direct tunability to some extent, as manifested by the recorded spectra. A tuning range at signal wavelength from 1.01 μm to 1.08 μm, which is comparable to the signal bandwidth, was obtained just by rotating the nonlinear crystals and slightly adjusting the optical delay lines. Since the OPA did not work with the seed off, the spectrum of the signal pulse supported by the phase-matching bandwidth should overlap more or less with the seed wavelength in order to make the seeding effective. As a result, the effective tuning range will be limited closely to the signal bandwidth centered at the seed wavelength, which is much narrower compared with the whole tuning range of OPG/OPA itself. The physical reason about the tunability in the CW seeded OPA is clear: weak enough seeding will not affect wavelength selection by rotating the OPA crystals, while the strong enough seeding will clamp the central wavelength and make the direct tuning impossible. We believe that weak seeding is crucial to wavelength tuning in our scheme, which can be regarded as a compromise between OPG/OPA and strongly seeded OPA. Although we did not characterize the idler pulse, its corresponding tuning range from 3.1 μm to 3.9 μm can be expected. This spectral range is particularly attractive for nonlinear optical spectroscopy of the vibrational dynamics in numerous systems.

Figure 3 shows the typical autocorrelation traces of the signal pulses at 1045 nm (corresponding to the maximum OPA gain and nearly zero GVMs) for the conventional seeded and the hybrid seeded schemes, respectively. Both schemes produce signal pulses with duration of ~100 fs assuming a Gaussian intensity profile. This results in a time-bandwidth product of 0.5, which corresponds to 1.2 times the transform limit. Within the whole tuning range, the signal pulse durations were less than 150 fs. The resulted relatively broad pulse duration at wavelength deviated from 1045 nm is attributed to the larger GVM.

 

Fig. 3. Typical autocorrelation traces of the signal pulses at 1045 nm: (a) conventional seeded OPA, and (b) hybrid seeded OPA. Insets: the corresponding spectra.

Download Full Size | PDF

 

Fig. 4. Pulse energy at 1045 nm vs. seeding power. Open square: conventional seeded OPA; solid circle: hybrid seeded OPA.

Download Full Size | PDF

Since pulse energy of the idler is 3–4 times lower than that of the signal from the first stage, one might expect that the conversion efficiency of the hybrid seeded OPA should be lower than that of the conventional seeded OPA. However, the measured results demonstrate similar saturated efficiencies in these two cases, as shown in Fig. 4. In the saturated amplification regime, the maximum efficiency is mainly determined by the available pump energy, though the saturation occurred at a higher seeding level for the hybrid seeded OPA. As a result, a maximum signal pulse energy of ~17 μJ was produced in both the hybrid and conventional seeded schemes, which corresponds to an overall conversion efficiency of ~11% (signal plus idler). Due to the weaker seeding hence the weaker nonlinear interaction in the second stage, the conversion efficiency of the hybrid seeded OPA is lower than that of the conventional seeded OPA before saturation is reached.

3. Conclusion

We have proposed and demonstrated a novel hybrid seeded femtosecond OPA that incorporates an external CW seeding at the signal wavelength in the first stage and difference-frequency mixing between the idler and pump in the second stage. The scheme does not introduce additional complexity both in the setup and adjustment, while it generates ~100 fs signal pulses with clean spectrum and provides direct wavelength tunability to some extent. The hybrid seeded femtosecond OPA is free from damage due to the lower pump intensity, and its conversion efficiency has been proved to be as high as that of the conventional seeded OPA. A wavelength tuning range from 1.01 μm to 1.08 μm at the signal (from 3.1 μm to 3.9 μm at the idler correspondingly) is achieved in such a MgO:LiNbO₃-based device.