1 Introduction

Tunable picosecond lasers in the near-infrared (NIR) region are important excitation sources for advanced microscopy and spectroscopy. Optical parametric oscillators (OPOs) are popular sources in this category. An OPO uses parametric down conversion (PDC) of a pump laser in a nonlinear medium to generate broadly tunable signal and idler beams. However, an OPO-based laser system is costly and the cavity alignment requires strong optical skills. This may be a long-term disadvantage for researchers who are not laser experts themselves.

Optical parametric generators (OPGs) may be simpler and cheaper substitutes for OPOs in some applications. An OPG also uses PDC of a pump laser to generate tunable outputs, but it is a single-pass device and, therefore, its operation is much easier. However, the challenge is to achieve high single-pass conversion efficiency. It needs a careful choice of pump laser and nonlinear medium and possibly with the help of injection seeding [1–3]. An OPG with 53 % internal conversion efficiency was reported [1] with a 55-mm long periodically poled lithium niobate (PPLN) pumped by a 7-ps Nd:YVO4 laser at 1064 nm and 20 W average power. It was seeded with a low-power CW laser at the signal wavelength, which improved the conversion efficiency. But seeding with an external laser may also bring a limitation, as the wavelength range of the seeding laser will limit the tuning range of the OPG.

Here, we demonstrate a picosecond NIR laser source based on a self-seeded OPG. In our design, the PDC of one pump laser pulse is seeded with a small portion of the signal pulse generated by the previous pump pulse. This will enhance the conversion efficiency at any signal wavelength without the need for an external seeding laser. A self-seeded OPG was reported before with a nanosecond pump at 10 kHz repetition rate [4], but to the best of our knowledge, it has never been demonstrated with picosecond lasers at 80 MHz repetition rate. Due to the large difference in laser parameters, the self-seeding scheme in our work and that in [4] are very different. Furthermore, we frequency-doubled the OPG signal output with a second harmonic generator (SHG). The second harmonic (SH) beam is the final output of our setup, and it is comparable to the output of a picosecond OPO or Ti:Sapphire oscillator in terms of pulse energy and duration. In the following, we will present the characterization of the new laser source and demonstrate its potential through imaging experiments.

2 Experimental setup

A schematic of the laser setup is depicted in Fig. 1. The pump laser is a 20-W, 7ps, 1064-nm Nd:YVO4 laser (High Q Laser, Austria) at 80 MHz repetition rate. The pump laser beam is split into two arms using a polarization beam splitter (PBS). A small portion reflected by the PBS is used as the Stokes beam for CARS imaging in Sect. 4. The major portion (~17 W maximum before the OPG crystal) is delivered to the OPG via a dichroic mirror (DM1). The polarization of the pump beam and all generated beams in the setup are vertical. The OPG consists of two plano-convex lenses (L1 and L2) of 200 mm focal length and a 40-mm long MgO:PPLN crystal (MOPO1-1.0-40, Covesion, UK). The crystal is triple-band AR coated on both ends. It contains five gratings with different periods, but only the grating with 31.02 μm period is used in the current experiments. The lenses are chosen so that the e−2 beam waist radius of the pump beam in the crystal is ~70 μm. The corresponding central peak intensity is ~468 MW cm−2, smaller than the suggested damage threshold (500 MW cm−2).

Fig. 1
figure 1

Schematic diagram of the laser setup. HWP half-wave plate, PBS polarization beam splitter, DM dichroic mirror, L plano-convex lens, W BK7 and fused silica window, BD beam dump, DL delay line, TBE telescopic beam expander, A attenuator, F 780 nm long pass filter, I Iris

Full size image

By tuning the PPLN temperature, the generated signal and idler beams are tuned from 1610 to 1652 nm and from 3138 to 2989 nm, respectively. The signal gain bandwidth is ~3 nm with fixed pump wavelength. The group velocity mismatch (GVM) between the signal and pump waves in the crystal is ~105 fs/mm [5]. For the 40-mm PPLN the temporal walk-off is 4.2 ps, less than the FWHM pump pulse duration (7 ps). Therefore, it would not cause any significant efficiency reduction or temporal broadening of the signal pulse. The idler beam is absorbed by fused silica and BK7 windows (W), and the residual pump is separated from the signal using a dichroic mirror (DM2). The signal beam is then focused by a 75-mm lens (L3) into another PPLN for second harmonic generation.

The SHG PPLN (SHG6-10, Covesion, UK) is 10 mm long and dual-band AR coated on both ends. It has nine gratings with periods from 19.00 to 21.00 μm. By tuning the temperature and gratings, it can cover a fundamental input from 1572 to 1652 nm. With the OPG signal in 1610–1652 nm as input, the phase-matching bandwidth for the SH output is ~3 nm. The GVM between the signal and SH waves in the PPLN is 273 fs/mm [5]. For the 10-mm PPLN the temporal walk-off is 2.73 ps; hence it would not cause significant efficiency reduction or temporal broadening either. A second 75 mm lens (L4) re-collimates the exiting beams, and a dichroic mirror (DM3) separates the residue OPG signal from the SH beam.

The residual OPG signal is re-injected into the OPG via the self-seeding path. The delay line (DL1) in the seeding path is tuned to ensure that the residue signal pulse temporally overlaps with the next pump pulse in the OPG PPLN. A 2× beam expander (TBE) tunes the seed beam size to roughly match the pump beam size at L1 to improve the seeding effect. An attenuator (A) reduces the seed beam power to ~6 mW or a pulse energy of 0.075 nJ. It is enough to create significant seeding effects and further increasing it no longer helps.

The SH beam reflected by DM3 is the final output of the OPG + SHG setup. A 780-nm long pass filter (F) is used to eliminate unwanted blue or green light generated in the PPLNs. The SH beam and the low-power 1064 nm beam reflected by PBS are collinearly combined by a dichroic mirror (DM4) and sent into a confocal microscope (Nikon Eclipse Ti) for coherent anti-Stokes Raman scattering (CARS) imaging experiments. A 40× water immersion NIR objective (NA = 0.8) focuses both beams into the test sample. The forward CARS (F-CARS) signal was collected by an air condenser (NA = 0.52), filtered with a band pass filter, and then detected by a photomultiplier tube (PMT, H7422-50, Hamamatsu, Japan).

3 Experimental characterization of the laser source

3.1 The effects of self-seeding on conversion efficiency

We measured the power of the OPG signal beam and SH beam at various pump powers with and without seeding. The pump power was measured in front of the OPG PPLN. The SH power was measured after the 780-nm long-pass filter (F). The OPG signal power without seeding was measured after M3. The signal power with seeding was derived by comparing a sampled signal power after DL1 with and without the seeding. After taking into account the losses due to optics, the internal signal power and SH power inside the OPG and SHG PPLNs were derived. The available signal and SH power outside the OPG and SHG was 70.8 and 90.2 % of the internal powers, respectively.

In this measurement, the OPG PPLN was kept at 95 °C to generate 1634 nm signal. The 20.5- μm grating of the SHG PPLN was kept at 160 °C to frequency double the 1634 nm input into 817 nm output. Figure 2 shows the internal signal power and internal SH power at various pump powers for the case of no-seeding and two cases of seeding as explained below. As shown by the ‘no-seeding’ curve in Fig. 2, without seeding at 17 W pump power, the internal OPG signal power was 4.1 W (37 % efficiency including idler) and the internal SH power was about 800 mW (27 % efficiency).

Fig. 2
figure 2

a Internal signal power at 1634 nm. b Internal SH power at 817 nm. c SHG conversion efficiency at 817 nm as a function of pump power with and without seeding

Full size image

By introducing self-seeding and tuning the seeding delay DL1 carefully, the OPG and SHG output powers were increased significantly. However, interestingly, they were maximized at different positions of seeding delay. The signal power was only peaked at one delay position, and the ‘best signal’ curves in Fig. 2 refer to the results at this position. But the SH power had two peak positions on the two sides of the best signal point: one was earlier than it by about 2 ps, and the other was later than it by about 3 ps. The ‘best SH’ curves in Fig. 2 refer to the best results at them. From Fig. 2a, at 17 W pump power, the internal OPG signal power reached 6.3 W at “best signal” condition, indicating an impressive internal conversion efficiency of 57 %. This appears to be even higher than the efficiency achieved in a 55-mm PPLN with a similar pump laser and CW seeding [1]. From Fig. 2b, the internal SH power was increased to 1.2 W at “best SH” condition, which was a 50 % increase over that without seeding.

The SHG efficiency curves (Fig. 2c) revealed yet another interesting fact. Comparing with the “no-seeding” case, the SHG efficiency was increased by self-seeding at “best SH” condition. However, it was actually reduced at “best signal” condition. This may be surprising as one might expect the SHG efficiency to increase when the input power increases. But we think that it provides an intriguing clue about how self-seeding affects the bandwidth of the OPG signal, as explained below.

First, as reported in [1], without seeding the PDC process will create a broad signal bandwidth of ~10 nm in an OPG like ours at high pump powers. This is broader than the phase-matching bandwidth of the SHG PPLN (~6 nm for the input estimated from Fig. 3b). With self-seeding at “best signal” position, presumably the peak of the seed pulse is temporally overlapped with the peak of pump pulse at the focal center in the OPG PPLN. This will shorten the signal pulse because the peak of the pump pulse gets more enhancements in conversion efficiency than the rest of the pulse. This temporal shortening may result in an increase in signal bandwidth. But because the phase-matching bandwidth of the SHG PPLN is smaller, a larger signal bandwidth will actually reduce the SHG efficiency despite the increase of overall input power. This explains the efficiency drop at “best signal” condition. On the other hand, when the seed pulse has a proper delay relative to the pump pulse, the leading or trailing edge of the pump pulse will get more enhancements in conversion efficiency than the peak. This will cause temporal broadening of the signal pulse, which will likely narrow the signal bandwidth and benefit the SHG efficiency. As a result, the SH power is maximized under such conditions. Of course the above is only our hypothesis. A complete understanding of this phenomenon may need a full theoretical simulation.

Fig. 3
figure 3

a Autocorrelation traces of the SH pulse at 817 nm. b Spectra of the SH beam at 817 nm when OPG was pumped at 15 W power

Full size image

The SH beam is the final output from our setup. In the CARS experiments (Sect. 4), about 3 W of the 1064-nm beam was reflected by the PBS in Fig. 1, and 15 W was delivered to the OPG. Therefore, the characterization of the SH beam below was carried out at 15 W pump power.

3.2 The effects of self-seeding on the temporal and spectral profile of SH output

Figure 3 shows the autocorrelation traces of the SH beam without seeding and with seeding at “best SH” condition, measured by a collinear SHG autocorrelator (Miniscan-15p, Minioptic Technology Inc, CA, USA). The trace at “best signal” condition was basically identical to the “no seeding” trace (data not shown). The traces showed a sharp peak at zero delay indicating an incoherent component in the pulse. This is a common feature of the output from PDC which was also found in the literature [1]. Furthermore, the pulse duration was increased by seeding at “best SH” condition. This is consistent with our statement that under such condition, the signal pulse is broadened and, therefore, the SH pulse will likely be broadened as well. Assuming Gaussian deconvolution factor of 1.44, the FWHM of the SH pulse was 4.2 ps without seeding, and 5.0 ps with seeding in Fig. 3a.

The spectra of the SH beam were measured by Maya2000Pro (Ocean Optics). The spectral width of the SH pulse was not affected much by the seeding, probably because it was mainly determined by the phase-matching bandwidth of the SHG PPLN. The corresponding spectral widths were 3.11 nm without seeding, and 3.22 nm with self-seeding as reflected in Fig. 3b. We also measured the pulse duration and spectral width of the SH output at different wavelengths (data not shown). In the range of 805–826 nm and at “best SH” condition, the pulse durations could vary from 5.0 to 6.6 ps. The time-bandwidth products of the SH pulse were in the range of 4.7–8.2. This relatively large time-bandwidth product was mainly due to the low temporal coherence of OPG emissions.

3.3 Wavelength tuning

The wavelength of the OPG and SHG can be tuned by tuning the temperature of the PPLNs and by switching SHG PPLN gratings. At 15 W pump power, the internal signal power and SH power as a function of wavelength without seeding and with “best SH” seeding are shown in Fig. 4. The signal power with seeding is increased by ~32 % compared with that without seeding. A slight increase of signal power at longer wavelength was observed. The SH power with seeding is increased by ~37 % compared with that without seeding. A slight decrease of SH power at longer wavelength was observed. From the seeding curves reflected in Fig. 4, it could be seen that the signal wavelengths with higher power tend to produce SH output with lower power. This may imply a positive correlation between signal power and bandwidth at “best SH” condition. A higher signal power may be linked to a larger signal bandwidth, and so it leads to lower SHG efficiency as we discussed in Sect. 3.1.

Fig. 4
figure 4

a The internal OPG signal power as a function of signal wavelength without seeding and with seeding optimized for best SH power. b The internal SH power as a function of SH wavelength without seeding and with seeding optimized for best SH power. The OPG was pumped at 15 W power

Full size image

3.4 The beam quality of SH beam

We measured the M2 factor of the SH beam at 15 W pump power and “best SH” seeding condition by a slit scanning beam profiler (M2SET-IR/M, Thorlabs). There are two irises in the setup (I1 and I2 in Fig. 1) which serve as spatial filters for the signal beam and the SH beam, respectively. When both irises were wide open, the M2 factor of the SH beam was measured to be 1.5 and 2.3 in two orthogonal axes. The beam was in fact elliptical.

By closing down I1 and/or I2 to a certain degree, the beam quality can be improved, but at the cost of losing output power. When only I1 was closed down to the point where the SH power was reduced by 50 % (e.g. from 1.2 W to 600 mW), the M2 factor of the SH beam was improved to 1.4 in both orthogonal axes, and the beam became almost round. Furthermore, when iris I2 was also closed down to the point where the SH power was reduced by another 30 % (e.g. from 600 to 400 mW), the M2 factor was improved significantly to 1.1 in both orthogonal axes. The beam became very close to Gaussian. The autocorrelation trace and spectrum of the SH beam were not affected by tuning the irises as described. In the CARS microscopy experiments, we closed down only I1 to produce a SH beam with 500 mW external power and a M2 factor of 1.4, as this condition produced relatively good CARS signal.

4 CARS imaging experiments with the new laser source

To demonstrate that our new laser source may serve as a substitute for picosecond OPOs in certain applications, we carried out CARS microscopy experiments with it. CARS microscopy uses a pump beam (ω p) and a Stokes beam (ω s < ω p) to interact with a sample via χ (3) to generate an anti-Stokes output at (2ω p−ω s). This process is enhanced by Raman resonance when the beat frequency between the two excitation fields (ω p−ω s) matches the frequency of a Raman band [6]. Synchronized Ti:Sapphire lasers [7] or OPO sources [8–10] are commonly used for CARS imaging. Recently a novel OPA source for CARS microscopy was reported [11]. This OPA source also used a MgO:PPLN as conversion medium and had good tunability, but its current drawback seems to be the relatively low output power (~100 mW at maximum).

In our CARS imaging experiments, the SH beam (805–826 nm) served as the pump beam for CARS. The 1064-nm beam reflected by the PBS (~3 W) in Fig. 1 served as the Stokes beam. These two beams can provide Raman shifts from 2709 to 3024 cm−1, covering the aliphatic CH vibration region. The power of the pump and Stokes beams before the microscope objective was 100 and 300 mW, respectively. Note that in some other CARS imaging setups the required excitation powers were only tens of mW. We had to use higher average power to achieve a good signal level because we did not use a pulse picker to reduce the repetition rate and the pulse duration of our excitations were also longer than most.

The test sample was a single layer of 1 μm polystyrene beads on a cover glass. The F-CARS images at CH2 symmetric stretching band (2856 cm−1) and an off-resonance frequency (2722 cm−1) were acquired and shown in Fig. 5. The corresponding central wavelength of the SH beam was 816 and 825 nm, respectively. The images have been normalized with the non-resonant CARS signal from a cover glass at the corresponding frequencies. From Fig. 5 it is clear that at the resonant frequency, the beads produced apparently higher signal than at off-resonance, which demonstrates the resonant enhancement in CARS. The bandwidth of SH beam (50 cm−1) was larger than the bandwidth of a single CH2 band (~10 cm−1), which may lead to less than ideal resonant contrast. But for biological lipid structures with a broad CH2 resonant peak [12], the bandwidth of our source could be a good match.

Fig. 5
figure 5

F-CARS images of 1 μm polystyrene beads on a cover glass and the corresponding line profiles at a 2856 cm−1, b 2722 cm−1, respectively

Full size image

5 Discussion on advantages, limitations, and possible improvements

We have shown that our laser source can provide an output similar to that from a picosecond OPO or Ti:Sapphire laser. Comparing with a synchronously pumped OPO, the key advantage of our setup is that there is no resonant cavity, and hence the alignment and operation of the source are much easier. An OPO, on the other hand, requires lots of training to really be able to align it and maintain it efficiently. The overall cost of the source is also much cheaper than an OPO. Comparing with a double-stage OPG/OPA setup [2], our setup is still relatively simpler and has better OPG conversion efficiency (57 %). These features of our laser source may help ultrafast lasers to be more attractive to general users and stimulate the widespread of advanced optical techniques.

The current setup also has a few limitations. First, the wavelength tunability is limited by the properties of PPLN. A significant reduction in conversion efficiency was observed when we tried to tune the OPG signal wavelength to ~1800 nm. Similar observation was also reported in [13]. To extend the wavelength range, it may be necessary to replace the PPLNs with other types of crystals such as CdSiP2 or PPKTP [14–16]. However, more study is needed to determine the details of the desired crystal.

The relatively large time-bandwidth product of the current source (5–8) may also be a limitation for some applications. It can be improved by reducing the bandwidth of the seed beam with narrow band pass filters, thus making it closer to a CW seed. A narrowband seed would reduce the signal bandwidth and make the pulses closer to transform-limited [1].

6 Conclusion

We have presented a new type of tunable NIR picosecond laser source based on a self-seeded OPG and an SHG unit. By seeding the OPG with its own signal and adjusting the seeding delay, the internal conversion efficiency of OPG can reach 57 % and the efficiency of SHG can reach 31 %. The final output beam from the SHG contains 5–7 ps pulses tunable from 805 to 826 nm, with an average power of over 1 W at 80 MHz repetition rate. The temporal profile, spectral profile, and beam quality of the output beam were characterized in the study. This output is similar to that from a picosecond OPO, while our laser setup has the advantage of being simpler and less costly. With the new laser source, we have successfully demonstrated CARS imaging experiments on polystyrene beads, which shows that it can serve as an effective excitation source for microscopy and spectroscopy applications.