1. Introduction

Tunable picosecond (ps) pulses in the near- to mid-infrared spectral region, between 1µm and 4 µm, are required in many spectroscopic applications within medical, biological, and material science. Radiation in this “fingerprint” spectral region gives access to the vibrational lines and their overtones using Raman scattering spectroscopy [1, 2]. In particular, imaging techniques such as coherent anti-Stokes Raman scattering (CARS) benefits from employing relatively narrowband picosecond sources, since it reduces the background signal originating from two-photon electron resonances [3]. An additional benefit of the longer wavelengths, in terms of imaging, is the decrease of the Rayleigh-scattering efficiency. At the same time, by using infrared radiation damaging of the biological substance is avoided, and CARS-imaging has even been demonstrated on living cells using near-infrared picosecond pulse trains [2]. It has also been shown that for excitation wavelengths longer than the scattering object the backwards propagating CARS signal is strongly enhanced and can provide higher measurement sensitivity [4].

The laser source for CARS, and other spectroscopic or imaging applications, should produce relatively narrowband pulses independently tunable at two wavelengths. The repetition rate should be as high as possible in order to increase data collection rate, however, in many cases it is limited to below 500 kHz in order to reduce the thermal load on the specimen. This is especially important for in-vivo imaging [2, 3, 5]. One approach to build such a system uses two repetition rate-locked tunable picosecond Ti:Sapphire lasers with electro-optic pulse-pickers, that provides synchronized pulse trains at 250 kHz rate [3, 5]. However, reduction of the relative timing jitter to acceptable levels required rather elaborate locking electronics. It has also to be considered that operation of Kerr-lens mode-locked Ti:Sappire in the picosecond pulse regime, at wavelengths longer than 900 nm, is not as reliable as in femtosecond regime, even for commercial lasers. Another, and possibly simpler, solution is deriving tunable pulses from optical parametric devices pumped by a picosecond laser. Recently a synchronously pumped optical parametric oscillator employing periodically poled KTiOPO₄ (PP-KTP) and pumped by a mode-locked and frequency-doubled Nd:YVO₄ has been demonstrated [6]. This system provides perfectly synchronized pulses with tunable spectral separation. However, obtaining narrow pulse spectrum and tuning at the same time is not straightforward in this particular setup. The spectral and the temporal properties of synchronously pumped optical parametric oscillators are quite sensitive to the exact detuning between the pump and the parametric oscillator cavities.

Optical parametric amplifier (OPA) can be a more robust source of widely tunable picosecond jitter-free pulses. Broadband OPAs based on parametric gain in BBO are commonly used for generating tunable femtosecond pulses [7]. Due to the high peak intensity of the femtosecond pump pulses, the parametric conversion in the OPA is quite efficient even in such a low-nonlinearity material as BBO. Indeed, generation of 100 nJ level pulses at 150 kHz in two simultaneously pumped noncollinear BBO OPA systems has recently been demonstrated with a total pump-energy budget of only 7.5 µJ at 800 nm with 50 fs long pulses [8]. Rearranging regenerative amplifiers for the picosecond pulse regime unavoidably leads to a reduction of the pump peak power by at least two orders of magnitude. Materials with higher effective nonlinearities should then be used in the picosecond regime. For instance, a GaSe was used to realize a 1 MHz repetition-rate picosecond OPA in the mid-infrared region [9]. The pump energy was only 1 µJ and the tuning was achieved by rotating the GaSe crystal. Type-II phase-matching in bulk KTP has been employed in the near- to mid-infrared: (i), in a tunable picosecond optical parametric generator (OPG)-seeded OPA by pumping at 526 nm and 1053 nm at low repetition rates [10], and (ii), in a white-light continuum-seeded OPA pumped at 800 nm using a 1 kHz picosecond Ti:Sapphire regenerative amplifier [11]. The pump energy requirement can be substantially reduced by employing quasi-phase matched (QPM) structures in PPKTP. Several efficient OPAs with PPKTP from femtoseconds to picoseconds have been demonstrated in the near- and mid-infrared spectral regions [12–16]. An additional important advantage of using QPM structures is related to the possibility of spectral control of the generated signal and idler by adjusting three parameters, namely: (i), the QPM period, (ii), the pump angle with respect to the QPM grating, and (iii), the angle between pump and signal [17]. The OPA pump wavelength has a crucial importance for the parametric gain bandwidth. Recently, it was demonstrated that extremely broadband parametric gain can be realized in both collinear and noncollinear interaction by using the pump wavelengths between 820 nm and 850 nm [18].

In this work we utilize the broadband parametric gain property of PPKTP to build an OPG-seeded picosecond PPKTP OPA, where the collinear interaction in the OPG provides virtually white light in the near- to mid-infrared region, spanning between 1.08 µm and about 3.8 µm in a single direction. A properly filtered seed spectrum is amplified using broadband noncollinear parametric interaction in PPKTP with properly chosen QPM period. The amplifier crystal provides signal-gain from 1.1 µm to degeneracy without changing the propagation direction of the signal. This allows for simple tuning, double-pass amplification in a single PPKTP crystal, and easy separation of the OPA signal from the remaining pump.

2. Experimental setups and results

The experimental set-up is depicted in Fig. 1. The OPG and OPA setup consists of two uncoated PPKTP crystals with QPM grating periods of 28 µm and 26.3 µm, respectively. The PPKTP crystals had dimension of 8 mm, 5mm and 1 mm along x, y, and z crystal axes, respectively. The QPM periods have been chosen different in order to realize collinear parametric interaction in the OPG crystal and noncollinear interaction in the OPA crystal [17, 18]. Both crystals were pumped by an Ti:sapphire regenerative amplifier tuned to a wavelength of 823 nm, which is generating 1 picosecond-long pulses at a repetition rate of 1 kHz. The FWHM of the pump spectrum was 1 nm. The pump beam was split in a 12:88 ratio, where the 12 % was directed to the OPG stage, while the remaining pump was propagating through a delay-line and synchronized with the seed in the OPA stage. The beam radius incident on the 12:88-beam splitter was ~3 cm (e^-2 intensity). The pump beams were collimated in the PPKTP crystals by the telescopes: f₁–f₂ and f₁–f₃ (f₁=500 mm, f₂=50 mm, and f₃=100 mm), forming the beam radii of 300 µm in the OPG crystal and 600 µm (e^-2 intensity) in the OPA crystal. We intentionally employed rather large pump beam sizes, in order to take advantage of the large average power available from the amplifier, which substantially simplifies the power measurements. In principle, the beam radii can be reduced by about an order of magnitude if higher repetition rate pump systems with lower pulse-energy budget are used. The collinear OPG, in PPKTP with QPM period of 28 µm, provided an ultrabroad bandwidth seed, which extended from 1080 nm to 3800 nm. However, here we only used the signal part of the spectrum, i.e., 1080 nm to 1650 nm. The OPG beam was collimated with the lens, f3, which was followed by a 3.2 mm circular aperture for spatial mode filtering. The OPG produces a broad-band beam without measurable angular dispersion, i.e. the spectral width is not affected by the spatial limitation of the beam. The undepleted OPG pump was blocked by a long pass filter (LPF).

 

Fig. 1. The experimental setup consists of an OPG-stage (top) and an OPA-stage (right bottom). The seed is spectrally manipulated in the “black-box” and thereafter launched into the OPA. The pump for the OPA-stage is delayed and the power output is adjusted with the λ/2-wave plate (WP) and the Glan polarizer (GP).

Download Full Size | PDF

In between the OPG and OPA crystals a spectral filtering and conditioning can be performed (depicted as spectral management in Fig. 1). For our purposes of tunable picosecond pulse generation, we used a zero-dispersion compressor setup with a slit-aperture in the Fourier plane. The bandwidth filter is discussed below in more detail. The spectrally conditioned seed (depicted as k_{seed, out} in Fig. 1) was focused by the lens f₄=250 mm into a 600 µm beam radius inside the OPA crystal. A noncollinear OPA configuration was chosen, since it facilitates seeding arrangements and enables easy double-pass amplification, if needed.

In order to support amplification of the whole seed spectrum, the external angle between the x-axis of the PPKTP OPA crystal and the pump was set to 22 degrees, whereas the external angle between the pump and the seed was 3.6 degrees (62.8 mrad). It should be noted that there was a slight angular dispersion apparent in a separately measured parametric fluorescence spectrum; however the angular dispersion was very small and, in the seeded amplifier operation, the seed’s angle did not require additional adjustments over the entire tuning range.

For the pump pulse energy of 27 µJ (peak intensity of 9.5 GWcm^-2) incident on the OPG, the measured seed energy, after the long pass-filter and the spatial filter (k_seed, in in Fig. 1), was measured to 1.62 µJ. Without the spatial filter an overall conversion efficiency of 30 % was measured. However, it should be noted that the total efficiency was slightly higher as we employed a filter in front of the power meter in order to block the idler above 2000 nm, so that only relevant signal seed power could be monitored.

A simple way to achieve OPA tuning would be to propagate the broadband seed through a dispersive medium before amplification. Then by scanning the OPA pump-delay it would provide access to different spectral components of the chirped seed pulse. To demonstrate this we let the k_seed, in propagate through a dispersion line consisting of a total length of 30 cm glass slabs with a cross-section of 1×1 cm², where two-thirds of the length was SF6 glass and one-third was SF8 glass. The seed power after these glass slabs (k_{seed, out}) was measured to be 40 µW, corresponding to 40 nJ of pulse energy. The large power loss in the glass was not specifically investigated, but it might be caused by multi-photon absorption. Nevertheless, the remaining seed energy was large enough for efficient amplification in the OPA. The measured dependence of the OPA signal central wavelengths on the relative delay of the pump pulse is shown in Fig. 2(a). The length of the chirped seed pulse is 17 ps and by only changing the pump delay, it was possible to tune the OPA from 1616 nm to close to 1100 nm. As seen in Fig. 2 (b), the measured full-width at half maximum (FWHM) spectral bandwidths varied between 21 THz and 5 THz when going from degeneracy to 1100 nm. The reason for the variation in bandwidth is that the point of zero-group velocity dispersion (GVD) for SF6 and SF8 are at 1.9 µm and 1.6 µm, respectively, and it results in a poor temporal separation of the spectral components near degeneracy. Clearly this tuning method could be employed in ultrashort-pulse OPAs, but it gives too broad spectra for a picosecond device. Similar OPA signal power levels were observed throughout the tuning range. We have chosen the 8 ps pump-delay, corresponding to the central seed wavelength of 1280 nm, to measure the power characteristics of the single-pass OPA. For the incident pump energy of 86 µJ (peak intensity of 7.6 GW/cm²) the overall conversion efficiency in the OPA (signal power + idler power/incident pump power) was about 18 %, resulting in the estimated seed amplification factor of 36 dB. This estimate takes into account the fact that only part (1 ps-long) of the seed pulse was utilized in OPA, corresponding approximately to 2.7 nJ of the effective seed energy.

 

Fig. 2. (a). Measured amplified central wavelength as a function of the time delay of the OPA pump beam. (b). The output spectra. The grey curve is the OPG output spectrum. Dashed curve: at a delay of 1 ps. Dotted curve: at a delay of 4 ps. Dotted-Dashed: at a delay of 7.3 ps.

Download Full Size | PDF

In order to increase the spectral control and to reduce the seed spectral bandwidth, we removed the glass slabs and employed a zero-dispersion compressor arrangement, which was inserted after the OPG. In Fig. 3, the modified spectral management set-up is shown. The zero-dispersion setup, also called the Fourier spectral filtering setup, consists of: a λ/2-wave plate (WP), a grating with 900 grooves/mm, a lens with the focal length f₅=80 mm and a 30 mm aperture diameter, a 300 µm wide slit-aperture in the Fourier plane (AP), and an aluminium mirror (M_FP). The wave-plate rotates the polarisation by 90 degrees in order to access the highest diffraction efficiency, which was approximately 80 %. The distance between the grating and the mirror M_FP is 2×f₅. Consequently, the m=-1 diffraction order of the grating is Fourier transformed by f₅. This means that, in the Fourier plane (i.e. at the mirror M_FP) the seed’s spectrum is spatially distributed over the mirror M_FP and it is readily accessible for spectral filtering. Narrow-bandwidth tuning is achieved by translating the 300 µm wide slit-aperture along the mirror M_FP. Since the spectral bandwidth of the seed is very broad, the angular spread of the diffracted beam also becomes large and, as a consequence, the f₅-lens’ aperture was too small to collect all the frequencies at the same time. However, by rotating the grating once, in the middle of the tuning range, the whole spectrum from 1080 nm to 1650 nm could be accessed. The two grating positions correspond to the angles of incidence of 58 degrees and 69 degrees with respect to the normal of the grating. The resulting diffraction angles for the m=-1 diffraction order are 16 degrees and 24.6 degrees, which corresponds to central wavelengths of 1260 nm and 1500 nm, respectively. It is important to stress that in this spectral filtering arrangement the beam returning from the filter is not changing its pointing angle during the tuning procedure; thus, the seed’s routing mirrors do not need to be realigned. In this manner, a seed with narrow bandwidth of less than 10 cm^-1 (300 GHz) could be continuously tuned from 1080 nm to 1650 nm. In Fig. 4(a) the seed spectra are depicted as they are tuned over the parametric gain spectrum of the 26.3 µm PPKTP crystal deployed in the OPA [the grey curve in Fig. 4(a)].

 

Fig. 3. Modified experimental setup. The OPG seed is spectrally manipulated in a zero-dispersion arrangement. The seed’s polarization is rotated with the λ/2-wave plate (WP) in order to maximize the diffraction efficiency into the minus first-order of the grating.

Download Full Size | PDF

For the OPG pump energy of 27 µJ, the Fourier-filtered seed energy (k_{seed, out}) was approximated to 0.57 pJ. The seed energy was measured with a 1 GHz InGaAs-photodetector, which was carefully calibrated using amplified signal and a power meter. Example of the OPA signal spectrum with a central wavelength of 1495 nm after a single-pass through the amplifier is shown in Fig. 4(b). The spectrum was recorded at the OPA pump pulse energy of 50 µJ (4.4 GWcm^-2). At this pump intensity the amplified signal (idler) energy was 2.7 µJ (2.2 µJ). The same energies were found throughout the tuning range. This corresponds to a single-pass parametric gain of 67 dB and an overall efficiency of 10 %. As evident from Fig. 4(b), the spectral bandwidth of 2.2 nm (FWHM) was preserved after amplification. The spectrum has the typical Fraunhofer diffraction pattern of the slit in the Fourier-filter arrangement. When the pump energy was increased to 83 µJ (7.3 GWcm^-2), the single-pass signal gain increased to 69 dB producing amplified signal energy of 4.8 µJ. However, the spectrum had broadened more than five times due to the increasing of amplified parametric fluorescence (APF).

 

Fig. 4. (a). Grey curve is the OPA crystals gain spectrum. Black curves: tuning of the seed by moving the aperture in the FP. (b) Single-pass amplified spectrum. A central wavelength of 1497 nm, with a preserved bandwidth of 2.2 nm (FWHM). The spectrum is diffraction-limited, due to the slit-aperture in the Fourier plane.

Download Full Size | PDF

In order to circumvent the onset of APF, to preserve the narrow bandwidth, and to increase the power level, a double-pass configuration was constructed by re-using the undepleted OPA-pump. Figure 5 illustrates this arrangement. The undepleted pump and the signal are imaged off-axis with a lens, f₃=100 mm, in order to displace the constituent beams when sending them back for a second pass amplification in the PPKTP crystal. The distance between the PPKTP and the mirrors, M1 and M2, are 2×f₃, where M1 is mounted on a translation stage for temporal synchronization between the pump and the signal. The signal and the pump on their return path are displaced by the same distance so the interaction angle in the PPKTP crystal stays approximately the same as in the first pass. Moreover, because of the signal displacement after the second pass through the amplifier, a pick-out mirror can be inserted for the signal output. It should be noted that the idler, which is generated in a conjugate direction on the opposite side from the pump wave vector, is not used in the second pass through the amplifier.

 

Fig. 5. The second-pass experimental setup. The signal and the pump propagate off-axis through the lens f_s and the mirrors M1 and M2 displaces the beams when returning for a second-pass through the PPKTP crystal.

Download Full Size | PDF

The second-pass amplification increased the overall OPA efficiency to 20 %, generating signal (idler) pulse energy of 6.5 µJ (5.3 µJ) for the pump energy of 60 µJ (5.3 GWcm^-2) as shown in Fig. 6(a). At this pump intensity the parametric gain increased to about 70.5 dB. The stability of the output signal was ±3%, which is close to that of the pump. The maximum pump intensity was determined by monitoring the amplified signal spectrum. The two solid curves in Fig. 6(b) are the signal spectra at two different wavelengths of the seeded double-pass OPA measured at 60 µJ of pump energy, whereas the dotted line corresponds to the unseeded one. It is clearly seen that almost all the APF are completely suppressed when the OPA is seeded, and by tuning away from degeneracy it decreases even more. At pump intensity of 5.3 GWcm^-2 the APF constitutes only -30 dB in the OPA when tuned close to degeneracy. Increasing the pump energy further would increase the APF background, which might not be acceptable for most applications.

 

Fig. 6. (a). overall OPA efficiency and signal energy as a function of incident pump energy. (b): Spectra at 60 µJ of pump energy. Dotted curve: - unseeded double-pass OPA. Solid lines - seeded double-pass OPA.

Download Full Size | PDF

3. Conclusion

In conclusion, a narrow bandwidth picosecond noncolinear PPKTP OPA that is continuously tunable between 1100 nm and 1650 nm have been demonstrated. The seed source is an ultrabroad bandwidth-generating collinear PPKTP OPG. The broadband seed was spectrally filtered by using a Fourier spectral filtering arrangement, which allows simple tuning of the seed wavelength. The generated signal bandwidth is slightly smaller than 300 GHz, which is maintained throughout the tuning range. The bandwidth is determined by the width of the slit in the Fourier plane. In principle, the spectral width can be reduced even further in this setup, however, at the cost of increasing slit-diffraction ripples appearing in the spectrum, and decreasing available seed energy. The latter can be boosted to a certain degree by increasing the pump power in the OPG stage. In order to increase the conversion efficiency of the OPG-OPA system, we re-used the undepleted pump in a second-pass through the OPA PPKTP crystal. In doing this, a parametric gain exceeding 70 dB was obtained. For a total pump energy of 100 µJ (both the OPG and the OPA) a signal (idler) output energy of 6.5 µJ (5.3 µJ) was achieved, corresponding to a total optical-to-optical efficiency of 12 %. It should be noted that with this PPKTP OPG-OPA arrangement the same amplification factors should be achieved in higher-repetition rate picosecond systems. For instance, a typical Ti:Sapphire regenerative amplifier operating at 150 kHz repetition rate provides pulse energies of about 8 µJ. In order to achieve the same amplification ratio the pump beam radius in the OPG and OPA PPKTP crystals have to be reduced by approximately a factor of 3.5 down to 86 µm and 171 µm, respectively.