1. Introduction

Synchronized multi-color mode-locked lasers, which have wide applications in ultrafast laser spectroscopy [1, 2], optical pulse synthesis [3], coherent anti-Stokes Raman scattering microscope [4], and optical atomic clockwork [5, 6], have been achieved by means of either active or passive schemes. Furthermore, synchronization between different-color lasers has opened up new avenues in generation and detection of terahertz sources [7–9]. Until now, great efforts have been devoted to improving the synchronization accuracy and robustness. By using carefully designed external phase-locked loops in the active synchronization scheme, independent mode-locked lasers were synchronized with a sub-femtosecond timing jitter [10, 11]. While synchronization robustness was dramatically improved by virtue of all-optical passive synchronization with remarkable cross phase modulation (XPM) effects in coupled-cavity lasers sharing the same Kerr-type nonlinear medium [12], or independently mode-locked lasers with master pulses injected into the slave oscillator in the master-slave configuration [13]. As a result of XPM-induced wavelength shifts and changes of the intracavity group delay that compensate for the cavity length detuning, the XPM-based passive synchronization scheme typically exhibits a large cavity mismatch tolerance. To date, a record timing jitter as low as 11 attoseconds has been achieved between a two-branch femtosecond erbium-doped fiber laser [14]. Besides the synchronization accuracy and robustness, another important issue is to amplify the synchronied lasers to high average power or high pulse energy. Note that there are growing requirements on high-repetition femtosecond lasers with hundreds of megawatt peak power to generate high-power femtosecond combs or extreme ultraviolet combs for the development of ultra-stable standard of optical frequency [15]. Nevertheless, current available synchronized lasers are mostly realized with mode-locked laser oscillators, typically a few hundred milliwatts in output power. Amplification of synchronized lasers to high pulse energy was realized by using Nd:YLF regenerative amplifiers at a low repetition rate [16], and by ytterbium-doped fiber amplifiers at 2 MHz repetition rate [17]. So far, high-power amplification of synchronized lasers has not been realized at a high repetition rate, and the influence of amplification on the timing jitter has not yet been explored experimentally.

Furthermore, recent progress in fiber lasers opens up a new way for high-power laser oscillators and amplifiers. The advent of double-clad fiber technology benefits high-power lasers and amplifiers from kHz to GHz repetition rate [18, 19]. The special fabrication of double-clad fiber not only provides an effective way to transfer the energy from diode lasers into fiber core where the signal pulses propagate, but also ensures diffraction-limited beam quality [20]. The long signal-pump laser interaction distance can afford a high optical-to-optical efficiency, while the large surface-to-volume ratio results in excellent heat dissipation [21]. Moreover, large-mode-area (LMA) photonics-crystal double-clad fiber upgrades the threshold of nonlinear effects such as stimulated Raman and Brillouin scattering, while its inner core provides a single-mode operation during amplification [22]. Up to now, pulse energy of μJ-scale at tens of MHz was realized by LMA ytterbium-doped double-clad fiber (YDCF) amplifiers [23, 24]. It is a natural idea to combine the temporal synchronization of multi-color lasers with the LMA double-clad fiber amplification technology to attain high-power synchronous lasers at high repetition rates. However, the amplified pulses suffer from the spectral and temporal distortion, nonlinear phase shifts and nonlinear polarization evolution during the amplification. Moreover, amplified spontaneous emission (ASE) noises may degrade the synchronization accuracy and thus produce detrimental influence on the timing jitter. Especially, the amplifying pulses are tightly guided in the inner cores of the double-clad fibers, wherein some undesirable optical nonlinearities may unavoidably bring about serious nonlinear phase shifts. All the factors complicate the experimental situations, which deserve careful experimental exploration.

In order to study the change of synchronization accuracy during high-power fiber amplification, a so-called spectral branch amplification method was employed in our experiments, here, a broadband few-cycle Ti:sapphire laser was spectrally filtered around the ytterbium-doped fiber gain bandwidth and used directly as a seed for multi-stage high-power amplifiers. As few-cycle Ti:sapphire lasers have ultra-broadband spectra covering the desired gain bandwidth of some sophisticated gain media, this method actually serves as an efficient way to attain synchronized multi-color lasers within wide tunable spectra. It has already been demonstrated in low-repetition rate Nd:YLF regenerative amplifiers [16], nevertheless, high-repetition rate amplification by using LMA double-clad fiber amplifier has not been realized. We note that such a spectral branch amplification method can be applicable to more general cases rather than merely limited to few-cycle lasers. Even as the signal pulses have insufficiently broadband spectra to cover the desired gain bandwidth, seed pulses could be generated at the desired wavelength by nonlinear processes with a typical synchronization accuracy of a few femtoseconds, such as by continuous-wave seeded optical parametric amplification in bulk crystals [25], or by broadening the femtosecond laser spectra with photonic-crystal fibers [17]. From the experimental point of view, it is relatively easy to control the relevant nonlinear processes and generate stable seed pulses with enough pulse energy for further amplification at a low repetition rate. But at a high repetition rate, narrowband signal pulses of Ti:sapphire laser may nevertheless have insufficient pulse energy to generate stable seeds with enough pulse energy through nonlinear processes, which may cause instability and especially difficulty in suppressing the accompanied ASE noises in further amplification. Special care should be taken to make the high-repetition rate pulse seed have sufficient pulse energy so that amplification dominates over ASE noises. This can be done by using carefully-designed multi-stage amplifiers in cascade.

In this paper, we further promote high average power synchronization technology by directly amplifying the near-infrared spectral fraction of a broadband Ti:sapphire laser at a repetition rate of 79.5 MHz with a two-stage ytterbium-doped fiber pre-amplifier followed by a two-stage LMA-YDCF power amplifier in cascade, achieving to the best of our knowledge the highest average power of 262 W. An overall power amplification up to 86.1 dB was achieved at 1030 nm with sufficient suppression of ASE noises. The jitter spectral density of the cross-correlation signal shows that the amplified high-power pulses maintained a temporal synchronization of 25 fs timing jitter with the few-cycle signal pulse.

2. Experimental setup and results

As schematically shown in Fig. 1, the high average power synchronized laser system consists of a few-cycle Ti:sapphire laser, a two-stage ytterbium-doped fiber pre-amplifier and a two-stage LMA-YDCF power amplifier. The femtosecond seed source was provided by a Kerr-lens mode-locked Ti:sapphire laser at the center wavelength of 800 nm (Rainbow, Femtocompact), producing 7-fs pulses at a repetition rate of 79.5 MHz with the average power of 170 mW, from which the pulses with the spectral fraction from 980 to 1080 nm [shown in Fig. 2(a)] were filtered out by a dichromic mirror. The filtered pulses with less than 0.1 mW average power were coupled by a microscopic objective (×60) into a guiding fiber of the first-stage pre-amplifier. The pre-amplifier was constructed with ytterbium-doped single-mode fibers (YbDF350, OFS) of 0.6 m and 1.5 m for the first and second stages, respectively, which were both pumped by fiber-pigtailed diode lasers of 300 mW at 976 nm. The power amplifier consisted of the LMA ytterbium-doped double-clad photonic-crystal fibers with the pump absorption of 10 dB/m at 976 nm (DC-200/40-PZ-Yb-01, CRYSTAL FIBER). The LMA ytterbium-doped photonic-crystal fibers have a diameter of 40 μm (NA=0.03) for the active core and a diameter of 200 μm (NA=0.55) for the inner clad. The double-clad fiber ends were sealed to protect the capillaries from environmental influences and polished at an angle of 8° in order to suppress parasitic lasing. Two high-power diode lasers emitting at 976 nm with a pigtail of 400 μm fiber core were employed as the pump sources for the power amplifiers. All the amplifiers were arranged in cascad, each of which was isolated from the following stages with optical isolators (>30 dB) so as to avoid the possible detrimental influence caused by reflection of the cascaded amplifiers. In our experiments, 80% of the pump laser was coupled into the inner clad and 60% of the seed power was coupled into the fiber core.

 

Fig. 1. Experimental setup. Dichroic mirrors: DM1 (HT@600–1000 nm & HR@1000–1120 nm), DM2 and DM3 (HT@976 nm & HR@1020–1120 nm); Wavelength division multiplexing (980/1040 nm): WDM; Optics isolator: OI.

Download Full Size | PDF

The selected seed pulses from the few-cycle Ti:sapphire laser were first amplified to the average power of 6 mW by the first-stage pre-amplifier. Because of the limited gain bandwidth of the ytterbium-doped fiber, the amplified spectra of the pulses were narrowed to 21 nm center around 1025 nm [shown in the inset of Fig. 3(a)]. Figure 2(b) compared the polarization extinction ratio under different pump powers. Polarization extinction ratio of the amplified laser was 6:1 at 180 mW pump, while it was 1:1 for ASE. Although non-polarization maintained fiber was used in the YDSF, elliptical polarization was realized for the amplified pulses. Figure 2(c) shows the radio-frequency spectrum for the seed laser (black) and amplified laser (red), indicating the optimized gain in the first-stage pre-amplifier was about 39.6 dB, much larger than the calculated amplification from the input seed power of 0.1 mW to the amplified power of 6 mW. This was mainly caused by the overestimated input seed power as we measured the integrated spectral fraction within 980–1080 nm while amplification took place only around 1025 nm with a full-width at half-maximum (FWHM) of 21 nm.

 

Fig. 2. (a) The seed spectra of the pulses from Ti:sapphire; (b) Polarization extinction ratio for the amplified laser (red) and ASE (black). (c) Radio-frequency spectrum for the seed laser (black) and amplified laser (red); (d) The output spectra of all four-stage amplifier with (red) and without (black) seeding the first-stage pre-amplifier.

Download Full Size | PDF

Figure 3 shows the slope efficiency for the four amplifiers, with their spectra shown in the insets. In the second-stage pre-amplifier, the output spectrum was narrowed to exhibit an FWHM of 14.9 nm, and up to 80 mW was obtained under a diode pump power of 250 mW, corresponding to 11.3 dB of laser gain. During the pre-amplification process, no spectral splitting was observed, indicating that no pulse distortion occurred within such a range of pulse energy even though the amplified pulses were tightly confined in the single-mode fibers. The second-stage pre-amplifier was necessarily required for the efficient suppression of ASE in the following power amplifier as only under appropriate net gains in each stage should the amplification dominate over ASE through gain competition in the ytterbium-doped fibers. The first-stage power amplifier generated 6 W of average power at 1030 nm and an FWHM of 9.4 nm with 22 W pump power, corresponding to 18.8 dB net gain. The second-stage power amplifier generated 262 W of average power at 1030 nm and a FWHM of 10.5 nm under 380 W pump power, corresponding to 16.4 dB net gain. The slope efficiency was 70 % for the second-stage power amplifier [shown in Fig. 3(d)]. This corresponds to an extracted power of 130 W/m. At the extracted power level greater than 100 W/m, a water-chiller was required to remove the thermal-optical heat and protect thermal damage of the polymer coating. For each stage of the cascaded fiber amplifiers, the pump power was optimized with the corresponding input seed power to attain a proper gain so that ASE could be suppressed efficiently. We further checked the amplified laser spectra of all four-stage amplifier without seeding the first-stage pre-amplifier. Figure 2(d) clearly shows the distinct difference between the output spectra with seeding and without seeding the first-stage pre-amplifier. Two arbitrary peaks with narrow bandwidth represent much of the ASE spectra, which could be totally suppressed by injecting the seed laser. The total gain for the whole amplifier system approached to 86.1 dB with no evident gain saturation at the second-stage power amplifier, implying promising further amplification of the average power to much higher level with a more robust pump.

 

Fig. 3. The slope efficiency for the first-stage (a) and second-stage (b) pre-amplifier, the first-stage (c) and second-stage (d) power amplifier, with their corresponding spectra shown in the insets.

Download Full Size | PDF

Amplified laser pulses were further compressed by a diffraction-grating compressor based on transmission gratings of 1250 lines/mm, which were designed to have a maximum diffraction efficiency at 1064 nm with 41.7° Littrow degree. Figure 4(a) shows the measured autocorrelation trace at 50-W output power of the fiber amplifiers. Assuming sech² pulse shapes, the uncompressed pulse duration was measured to be 2.28 ps, while compressed pulse duration was 240 fs. The overall compressor throughput efficiency is about 60%, resulting in 30-W average power of the compressed femtosecond laser pulses. An excellent quality of amplified pulses with small pulse pedestals and very low nonlinear phase shifts was observed under such a power level.

 

Fig. 4. (a) Experimental autocorrelation trace of the uncompressed high-power laser (red-triangle) with sech² fit (blue-curve), and autocorrelation trace of the compressed pulse (green-square) with sech² fit (orange-curve); (b) Noise floor of detection (gray), relative jitter spectrum density (black) and integrated RMS timing jitter (green) versus frequency between compressed 1030 nm fiber laser and 7-fs 800 nm Ti:sapphire laser.

Download Full Size | PDF

To characterize the timing jitter, we focused the compressed laser pulses and the femtosecond pulses from Ti: sapphire laser so that they crossed in a 0.5-mm β-barium borate (BBO) crystal. The sum-frequency generated from the BBO was detected by a photomultiplier tube (PMT). The relative delay between two pulses were positioned at half-maximum of the cross-correlation trace. The Fourier-transformed spectrum of fluctuation of the correlation intensity was recorded by a FFT spectrum analyzer (SRS, SR760). We adopted the same method as reported in Ref.14 in calculating the jitter spectral density S_Δt(f) and the RMS jitter. For the reason of jitter measurement between different optical carrier frequencies, the parameter 1/v₀ was assumed as the double width of the compressed pulse. Figure 4(b) illustrates the power spectral density and its integration of the timing jitter. As compared to the noise floor of the detection (gray curve), the jitter spectrum density (black) reveals that most of the relative timing jitter was originated from the frequency range from 1 to 100 Hz, resulting an integrated jitter of 24.7 fs, which was mainly caused by mechanical vibrations.

3. Conclusions

In summary, we synchronized 262-W average power ytterbium-fiber laser to a few-cycle Ti:sapphire laser. Seeded with the pulses with the near infrared spectral sideband of the few-cycle Ti:sapphire laser, ytterbium-doped fiber amplifiers generated 262-W average power, corresponding to 86.1 dB total gain. No evident gain saturation was observed in the amplification, implying even higher average power could be realized by using more robust pump. As far as we know, this is the first realization of such a high-power laser synchronized with a few-cycle Ti:sapphire laser at a high repetition rate, which would promote interesting studies of XUV comb generation, atomic and molecular spectral measurements, development of the ultra-stable standard of optical frequency.