1. Introduction

Lasers operating in the yellow-green region of the optical spectrum are a useful tool for a number of applications from atomic physics to biomedical science. The peak absorption wavelengths for various red fluorescent dyes and markers lie near 550 nm [1, 2], in addition to a number of resonances for atoms, such as Ba and Yb [3, 4], and molecules, such as hemoglobin and carbonylhemoglobin [5]. Additionally, ultrafast lasers operating within this region can be applied to two-photon fluorescence microscopy at short wavelengths. High repetition rate and average power are desirable in this application, in order to enable quick scanning, short timing resolution and a high signal to noise ratio [6].

A number of pulsed laser architectures, including frequency-doubled diode-pumped solid state lasers [7], Raman lasers [8–10] and fiber lasers [11, 12] have been developed to operate within this spectral region. Granados et al. have demonstrated a bulk-optics Raman laser able to deliver picosecond 559 nm pulses at Watt-level average power and a high repetition rate [9,10]. Sub-picosecond pulses in the yellow-green region have also been demonstrated at high average power and repetition rates using the second harmonic of fiber lasers based on Yb-doped fiber (YDF) [11, 12]. Fiber lasers are versatile and relatively inexpensive laser sources that provide an attractive alternative to their bulk optic counterparts, in part due to the inherent stability of their all-fiber construction. However, the available wavelengths for fiber-based architectures in the yellow-green region have so far been limited to < 550 nm due to confinement of the fundamental wavelength to near 1.06 μm, the center of the YDF gain band [13].

We present a fiber-based source of picosecond pulses at 554 nm with 1.4 W average power and a 300 MHz pulse repetition rate. To our knowledge, this represents the first fiber-based picosecond pulse source in the yellow-green region that is able to operate at > 550 nm, while reaching Watt-level average power. These pulses are generated through single-pass second harmonic generation (SHG) from 1.11 μm pulses in periodically-poled stoichiometric LiTaO₃ (PPSLT). The picosecond 1.11 μm pulse train was produced from re-amplified supercontinuum generation, as demonstrated in several previous studies [14–16]. Using this supercontinuum architecture, coupled with highly efficient SHG, we have been able to overcome the wavelength restrictions on ultrafast fiber lasers, while maintaining favourable properties for applications such as two-photon microscopy.

A number of previous studies have shown high SHG efficiency using ultrafast picosecond and femtosecond fiber lasers using various types of nonlinear media [17–20]. Particular attention has been paid to nonlinear crystals capable of quasi-phase matching (QPM), such as PPSLT and periodically-poled LiNbO₃ (PPLN), due to their high nonlinearities and wide range of operating wavelengths [18, 19]. In this work, PPSLT was chosen in order to avoid the effects of photorefraction present at high intensities in PPLN [21,22]. Without this inhibition, we have been able to achieve the generation of 1.4 W average power at 554 nm, with a conversion efficiency of 56% and the prospect of a further increase with greater available power at the fundamental wavelength.

2. Fundamental wavelength generation

An Er-doped fiber laser, passively mode-locked at the fifth harmonic of the fundamental cavity repetition rate, was used to generate a 1564 nm pulse train with a repetition rate of 300 MHz. The amplified and compressed 55 fs pulses were used to generate supercontinuum spanning from 1.1 to >2 μm in Sumitomo highly nonlinear fiber with a nonlinear coefficient γ = 0.02 W⁻¹m⁻¹. The resulting supercontinuum was amplified and filtered in three YDF stages, as shown in Fig. 1 [14]. In each stage, the light passed through the gain fiber, was reflected at a chirped fiber Bragg grating (CFBG), and passed through the gain fiber again before exiting through the circulator. This arrangement maximized the gain in each stage while minimising amplified spontaneous emission (ASE). The reflection band of the CFBGs was centred at 1108.6 nm with bandwidth of 1 nm. A 1.15 km coil of SMF was used between the first and second YDF preamplification stages to chirp the pulses to 20 ps. This length was selected by compromising between the nonlinearity of the pulses and the amount of dispersion that would require compensation. The preamplification stages provided a total of 27 mW of average power. As shown by Kielpinski et al. [14], there is no loss in coherence between the Er-doped seed laser and the amplified 1108.6 nm pulse train when using this technique. An IPG Photonics YAR-10K-1110-LP Yb-doped fiber power amplifier (YDFA) was used to amplify the resulting pulse train at 1108.6 nm to an average power of 5.0 W.

 

Fig. 1 (a) Schematic of experimental setup, showing mode-locked seed laser, supercontinuum generation and amplification stages in optical fiber. The three Yb-doped preamplifier stages are represented by Y1, Y2 and Y3. The pulse duration of the power YDFA output is compressed and the light is focused into an 11 mm fan-out PPSLT crystal. The fundamental and second harmonic are separated by two dichroic filters. (b) A detailed schematic of the 1108 nm preamplifier stages, consisting of a circulator, YDF and a chirped fiber Bragg grating (CFBG).

Download Full Size | PDF

The free-space output of the power YDFA was passed through a grating compressor using two LightSmyth LSFSG-1000 transmission gratings with a separation of 1.2 m, compressing the pulses to 1.9 ps in duration. ASE centered at ∼ 1090 nm accounts for ∼ 9% of the total YDFA output, as shown in Fig. 2, however this is removed through spatial filtering in the grating compressor stage. The pulse duration was measured by frequency resolved optical gating (FROG), using a MesaPhotonics FROGScan with an Ocean Optics HR4000 spectrometer, and a wavelength range of 503–577 nm. Figure 3 shows the FROG trace, pulse shape and spectrum determined from the FROG retrieval. The spectrum of the amplified IR pulse train is centred at 1108.1 nm with a FWHM bandwidth of 1.72 nm, broadened due to self-phase modulation (SPM) during amplification, giving a time-bandwidth product (TBWP) of 0.80. The relatively large TBWP for a sech² pulse is due to the remaining non-quadratic component of the spectral phase, shown in Fig. 3, which was introduced by the SPM. As it is not quadratic in form, this phase could not be compensated further using the grating compressor. The effects of SPM on the IR spectrum and spectral phase show that higher peak powers cannot be reached in the current configuration, without further spectral broadening and distortion of the spectral phase.

 

Fig. 2 Spectrum of 5 W YDFA output shown on a log scale, showing amplified spontaneous emission (ASE) centered at ∼ 1090 nm which accounts for ∼ 9% of the total power. The ASE is removed through spatial filtering in the grating compressor stage.

Download Full Size | PDF

 

Fig. 3 (a) FROG trace (inset) and time domain retrieval of output pulse from grating compressor, giving a pulse duration of 1.9 ps. (b) Measured optical spectrum of IR pulse train (solid blue), with a center wavelength of 1108.1 nm and FWHM bandwidth of 1.72 nm. In addition, the spectral phase retrieved from the FROG (dashed red) is shown on the right axis.

Download Full Size | PDF

The vertically polarized output of the grating compressor was focused down to a spot with an estimated 21.1 μm 1/e² radius, inside the PPSLT crystal used for SHG, using a 150 mm focal length lens. The maximum available power was 2.52 W, giving a maximum available peak intensity of 0.24 GW/cm².

3. Second harmonic generation

The 11 mm long fan-out style PPSLT crystal was manufactured by Oxide Corporation and anti-reflection coated for both the fundamental and second harmonic wavelengths. Fan-out poling, with a period ranging from 8.8–9.7 μm, was chosen in order to quasi-phase match at 1108.6 nm, due to the lower cost and increased versatility compared to a single-period PPSLT designed for the same wavelength. The poling period of the PPSLT varies at a rate of 0.2 nm/μm over a full width of 5 mm, with a variation of < 0.1% over the 1/e² diameter of the focused IR pump beam. As a result, we do not expect any significant effect on the phase-matching bandwidth as a result of the fan-out structure. The crystal temperature was held constant at 30.0 °C and the QPM period was tuned by translating the PPSLT orthogonally to the axis of beam propagation. The second harmonic was separated from the fundamental wavelength using two Semrock FF665-Di02 BrightLine filters, reducing the power of the fundamental light to < 0.7% of the remaining IR after the PPSLT, which is negligible compared to the SHG power.

Figure 4 shows the measured SHG power and conversion efficiency, plotted against fundamental power, where a total of 1.41 W of SHG is produced from 2.52 W at the fundamental wavelength. We have found this average power level to be stable to within ∼ 5% over several hours. The fundamental power was tuned continuously using a half waveplate and Glan laser polarizer. The total conversion efficiency increases to 56% as the fundamental power increases to 2.52 W, where the efficiency is limited by the maximum available input power. No limitations due to photorefraction were observed using PPSLT, as compared to the presence of this problem at similar SHG power levels in PPLN.

 

Fig. 4 SHG power (circles) and efficiency (squares, inset) plotted against fundamental power. A maximum SHG output of 1.41 W, at 56% efficiency is reached at an input power of 2.52 W. Using the maximum measured nonlinear efficiency of 51.6%/W, a quadratic estimate of SHG power levels and efficiencies (solid blue) and an estimate accounting for pump depletion (dashed green) are shown on both axes.

Download Full Size | PDF

Due to depletion of the pump beam the SHG power does not follow an $\eta P_0^2$ relationship beyond small signal levels, as shown in Fig. 4. SHG under the effect of pump depletion can be estimated by

Figure 5 shows the autocorrelation trace and optical spectrum of the SHG pulses. We measured an autocorrelation duration of 4.10±0.02 ps for the sech² pulses using an APE PulseCheck autocorrelator, ultimately giving a pulse duration of 2.66 ± 0.01 ps. The maximum peak power of the SHG pulses is therefore 1.36 ± 0.01 kW. We measured a center wavelength of 554.09 nm with a FWHM bandwidth of 0.19 nm, using the Ocean Optics HR4000 spectrometer used for the IR FROG measurements. The larger than expected TBWP of 0.49 is once again due to the higher-order spectral phase of the fundamental pulse, but we are unable to verify the spectral phase of the SHG pulse for comparison. The large reduction in the SHG spectral bandwidth, and resulting increase in pulse duration, is accounted for by the expected 0.27 nm phase matching bandwidth of the 11 mm long PPSLT crystal, as calculated from the group velocity mismatch.

 

Fig. 5 (a) Autocorrelation trace of the second harmonic pulse. The solid line represents a sech² fit to the data, which gives a 2.66 ± 0.01 ps pulse duration. (b) Measured SHG spectrum, with 554.09 nm center wavelength and 0.19 nm FWHM bandwidth.

Download Full Size | PDF

A time domain trace of the SHG pulse train is also shown in Fig. 6. The time domain signal was measured using a ThorLabs SV2-FC 2 GHz Si photodiode, coupled via multimode fiber, and recorded using an Agilent DSO8104A Oscilloscope with a bandwidth of 1 GHz. We observe a pulse to pulse amplitude variation of < 5%.

 

Fig. 6 Time domain photodiode trace of SHG pulse train, showing pulse to pulse amplitude variation of < 5%.

Download Full Size | PDF

4. Conclusion

We have demonstrated a source of 2.7 ps, 554 nm pulses at an average power of 1.41 W. The 554 nm pulse train was generated in PPSLT from the second harmonic of a 1.9 ps, 1.11 μm fiber laser source. 2.52 W of fundamental power was used, resulting in a total conversion efficiency of 56%, with a maximum nonlinear efficiency of 51.6%/W.

The design of the yellow-green pulse source is scalable to higher average power levels, as commercially available YDFAs can be made at considerably higher power levels, although consideration has to be taken for the distance from the YDF gain peak. In order to make the system more compact, the 1.2 m grating compressor could be replaced by hollow-core photonic crystal fiber to provide the anomalous dispersion required for compression of the amplified pulses. Additionally, gain in YDF is still present at wavelengths longer than those presented here, allowing for the center wavelength of the IR pulse train to be shifted up to 5 nm farther from the YDF gain peak, or indefinitely closer, without altering the design beyond the wavelength of the CFBG reflection band. This would allow for some versatility in the presented design to address applications of interest, or the fundamental flexibility of the supercontinuum-based design could be exploited to reach an even broader range of wavelengths.