Ultrafast fiber lasers exhibit high broadband gain per pass, superior thermo-optical properties, and excellent beam quality, making them very suitable for practical use. For simplicity and efficiency, advanced mode-locked oscillator designs which can compete with the amplifier systems are always favorable. Here, we demonstrate a high-peak-power Mamyshev oscillator based on single-polarization, large-mode-area photonic crystal fibers. Using properly arranged filters, the fiber oscillator directly emits pulses with 9 W average power at 8 MHz repetition rate, corresponding to a single-pulse energy exceeding 1 μJ. The pulses are dechirped to 41 fs outside the cavity, leading to a record oscillator peak power as high as 13 MW. With such unprecedented performance, the proposed single-stage oscillator should be very attractive for various applications.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
High-peak-power ultrashort laser sources play an increasing role in various scientific and industrial applications, ranging from high-field physics to materials processing. Ultrashort pulse generation from rare-earth-doped fibers is particularly widely recognized as a promising approach due to great thermo-optical characteristics, broadband gain, and beam quality. The main limitation of fibers is the accumulation of excessive nonlinearity, which hinders the directly achievable pulse energy from an oscillator. Typically, rather complex multi-stage amplifier systems are required for further power/energy scaling. To improve the simplicity and efficiency of such systems, advanced mode-locked oscillator designs with enhanced performance are always favorable. Developing compact and robust oscillators of high-energy femtosecond pulses has therefore attracted strong research interest, leading to significant advances in the field.
Over the past decades, with the maneuver of pulse dynamics in cavities [1–4] and the utilization of large-mode-area fibers , fiber laser oscillators have made great progress. The proposed all-normal-dispersion laser by Chong et al. allows for highly chirped intracavity pulse evolution, and hence the tolerance of large nonlinear phase accumulation . Meanwhile, the low-nonlinearity large-mode-area fibers further open the door to power/energy scaling [6,7]. The main issue now is to find an appropriate saturable absorber (SA) to initiate the mode-locking and stabilize the energetic ultrashort pulses. Large-mode-area photonic crystal fiber (LMA-PCF) lasers mode-locked with the nonlinear polarization evolution (NPE) method could provide sub-100 fs pulses with energies of several hundreds of nanojoules , yet the incompatibility between NPE and polarization-maintaining (PM) fibers limits the environmental stability. Furthermore, the initialization of the NPE mode-locking could be quite arduous. On the other hand, LMA-PCF lasers employing mode-lockers like semiconductor saturable absorber mirrors (SESAMs) [9,10] and nonlinear loop mirrors  have also been demonstrated. Nonetheless, they can’t quite reach the performance level of NPE-based lasers despite their compatible implementations in the PM format.
Recently, an attractive alternative to traditional passive mode-locking techniques was proposed and demonstrated by several groups [12–16]. Based on cascaded spectral broadening and offset spectral filtering [17–20], the so-called Mamyshev oscillator delivers superior results ( peak power) with just standard PM single-mode fibers (6 μm core size) . The concatenated regenerators greatly suppress the CW breakthrough and the occurrence of multi-pulse, which is a major step forward towards high-peak-power sources in environmentally stable designs. Increasing the filter separation generally allows higher-energy pulses to be stabilized [15,18]. The main drawback of this method lies in the ignition of pulsing . Although it has been reported that pulse generation could be started by directing light rejected by the filter back into the oscillator [14,22], an active modulation or external seed is often needed to generate a spectrum broad enough to sustain the initial oscillation. For a more practical use, Sidorenko et al. have also demonstrated a completely passive method for self-seeding with an auxiliary starting arm . Despite the starting issue, the unprecedented performance of the Mamyshev oscillator makes the prospects for further optimization and scaling very fruitful and fascinating. The use of large-mode-area fibers would greatly alleviate the nonlinearity issue and allow further power/energy scaling.
In this Letter, we report on the generation of femtosecond pulses from a Mamyshev oscillator featuring Yb-doped single-polarization LMA-PCF fibers. The laser reaches 9 W of average power at an 8 MHz repetition rate, corresponding to 1.1 μJ pulses. The output pulses are extracavity dechirped down to 41 fs with peak power. To the best of our knowledge, these are the highest single-pulse energy and peak power values ever reached by a single-stage, mode-locked fiber oscillator.
The high-power Mamyshev oscillator is constructed in a ring-cavity configuration, as shown in Fig. 1. Two segments of Yb-doped single-mode polarizing LMA-PCFs (NKT Photonics A/S, Denmark) serve as gain medium for the concatenated regenerators. The fiber has a multimode pump absorption at 976 nm and a single-mode diameter of 29 μm. The dispersion of the fiber is estimated to be around 1 μm. All fiber ends are angle-polished (without endcaps installed) to eliminate the parasitic oscillation. Both fibers are cladding-pumped by laser diodes emitting at 976 nm, which provide maximum 60 W power each. Waveplates are used to optimize the laser coupling efficiency. To overcome the pump power limitations, a multi-pass cell (MPC) is incorporated to elongate the cavity to scale the pulse repetition rate. The isolator ensures the unidirectional operation, and the reflection port of polarization beam splitter 2 (PBS 2) is taken as the output. The near-infrared transmission gratings (Thorlabs, GTI25-03A), with groove densities of 300 lines/mm, form the offset bandpass filters. By tuning the grating separately, filter 1 (TG 1) is set as and filter 2 (TG 2) is . The filter bandwidth is estimated to be . The use of the transmission gratings allows for high-power/energy operations without damage issues. Since the gratings are designed for broadband uses (500 to 1800 nm range), the efficiency is only around 1000 nm. With the customized blazed transmission gratings optimized for the working range, a higher laser efficiency is plausible.
To initiate the oscillation, we introduced an external seed that spans over the filter passbands. The seed pulse can be quite weak, and a conventional single-mode mode-locked fiber laser can satisfy the requirement. Of course, one may adopt the self-seeding method by building a starting arm with a SA. With adequate pump powers, by simply rotating the waveplate before PBS 2, the mode-locking can be easily started. Once the oscillator is established, the seed can be blocked. Since the repetition rate is scaled down to 8 MHz with the MPC, the oscillator initially operates in the multi-pulse regime. We observe different multi-pulse states, including soliton bunches, soliton flow, and bound state, etc. By gradually enlarging the output coupling ratio and decreasing the pump powers, the number of supported pulses decreases. Eventually, the laser turns into single-pulse operation. Once the laser enters the single-pulse regime, the transition back to the multi-pulse regime is not observed when increasing the pump power.
We then increase the pump power from LD 2 and adjust the output coupling to obtain the highest-output pulse energy. The pump power from LD 1 is kept constant during the process. Eventually, an average output power of 5 W is measured. Corresponding to the repetition rate, a single-pulse energy of 625 nJ is achieved at the laser output. Although without the passive fibers for extra spectral broadening, the laser still exhibits a rather broad spectrum spanning from 990 to 1120 nm at the level. The output spectrum and interference autocorrelation traces of the dechirped pulses are shown in Fig. 2. After being compressed with a transmission grating pair (1000 lines/mm), the dechirped pulse duration is as short as 27 fs assuming a Gaussian pulse shape, which is close to the transform-limited duration. Rather profound side-lobes should result both from the accumulated higher-order phase in the fiber and the third order dispersion brought by the compressor. Additionally, we do notice a constant fluctuation of the spectrum (especially around 1030 nm), and eventually the CW breakthrough occurs with a further increase of the pump power. The reason for this could be the very slight overlap between the two bandpass filters, which becomes a severe issue in high single-pulse energy situations. We tried to further separate the passbands of the filters, but this would make it hard to start the laser.
To further increase the achievable single-pulse energy and stabilize the laser, a tunable longpass filter (Semrock, TLP01-1116-25x36) and shortpass filter (Semrock, TSP01-1116-25x36) were incorporated into the oscillator before gratings 1 and 2, respectively. By adjusting the angle of incidence, the edge filters offer tunability over a very wide range of wavelengths, while essentially maintaining the spectral performance in terms of efficiency (over ) and optical density (). With suitable arrangements, the combination of the gratings and the edge filters form the new offset bandpass filters, with the interval deeply blocked. The central wavelengths of the bandpass filters nearly remain the same, and the seed is still strong enough to ignite the laser. Similar to the previous experiment, we tuned the laser to the single-pulse regime after starting and then increase the pump power gradually. The laser indeed becomes stable with the assistance of the edge filters, and the pulse energy of the oscillator is boosted to above 1 μJ.
The oscillator generates output pulses with an average power of 9 W at a repetition rate of 8 MHz, corresponding to a record pulse energy of 1.12 μJ. Further attempts to boost the laser pulse energy would again result in the CW breakthrough. The output pulse profiles are shown in Fig. 3. The spectral width is around 127 nm, which is similar than the previous result. However, the spectral flatness is not as good as before. Using the transmission grating pair with a distance of , the pulses are compressed externally to a minimum duration of 41 fs. The compressed pulse width is approximately 1.6 times the transform-limited duration, which is slightly broader than the previous result. This indicates the nonlinear chirp component in the pulse which cannot be compensated for by our compressor. The reason for this might be the deviation of the new bandpass filter profiles from the Gaussian shape, since the filter shape plays an important role in terms of the pulse quality and laser peak performance . To check the pulse quality, the intensity autocorrelation trace is also shown in the inset of Fig. 3(b). The exhibited pedestal results both from the spike in the spectrum and the previously mentioned uncompensated higher-order phase. Nonetheless, by reconstructing the pulse profile using the phase and intensity from cross-correlation and spectrum only (PICASO) method , we estimate the eventual peak power would still be 13 MW considering the compressor efficiency () which is, to our knowledge, the highest peak power generated by a mode-locked fiber oscillator.
We systematically investigated the laser performance as a function of the pump powers of LD 2. Figure 4(a) depicts the evolution of the output spectra. As the single-pulse energy increases with the pump power, the pulse spectrum broadens monotonically. The comparison between the measured compressed pulse duration (orange circles) and the calculated transform-limited (TL) pulse duration (blue squares) is shown in Fig. 4(b). Generally, the deviation between the measured and TL duration increases with the pulse energy, which can be attributed to the accumulated uncompensated nonlinear phases.
To confirm the single-pulse operation and stability of the laser, a 400 MHz analog oscilloscope, a 20 GHz sampling oscilloscope (with a fast photodiode of rise time), and an RF spectrum analyzer were used to monitor the status of the mode-locking. No satellite pulses were visible within the two consecutive pulse peaks on the sampling oscillator [Fig. 5(a)]. This verifies that only one pulse was circulating in the cavity. The pulse train monitored with the analog oscilloscope exhibited a very low amplitude noise level. This was confirmed by the RF spectrum analyzer measurement. The RF signal was recorded for a frequency window of 100 kHz span with 100 Hz resolution, as depicted in Fig. 5(b). Despite the introduced long optical path of the MPC, the laser still exhibited an absence of sidebands and harmonic frequencies to at least 60 dB below the fundamental frequency. The fundamental repetition rate of the pulse train was measured to be 8.1 MHz.
To gain more insight into the laser mechanism, we operated the oscillator at the high repetition rate without the MPC for cavity lengthening. The laser still initiated with similar pump power as before. However, the laser now directly operated at the single-pulse regime without the multi-pulse occurrence. Consequently, the oscillator became simpler and more reliable to start when the laser was turned off and on again. By enlarging the output ratio and increasing the pump power of LD 2 to maximum, the laser can generate an average power of 22 W at a repetition rate of 45 MHz. Notice that we do not observe any CW breakthrough during the whole tuning process. This implies that the occurrence of CW breakthrough results from the excessive single-pulse energy of the oscillator rather than the average power. This ensures the feasibility of further power scaling at high repetition rate. Also, the shortened light paths in free space would favor the stability of the laser. Hence, a high repetition rate, high average power/energy, environmentally stable oscillator is promising with just enough pump power. With the passive nonlinear spectral broadening inside the cavity, which can extend the pulse spectrum beyond the gain bandwidth, an octave-spanning laser may even be possible in the future [24,25].
In conclusion, we have presented a Mamyshev oscillator using single-polarization LMA-PCF. The laser exhibits excellent performance with high pulse energy (μJ-class), extremely short pulse duration (sub-50 fs), and environmental stability that can directly compete with the state-of-the-art fiber laser amplifying systems. With a simple configuration, reliable starting mechanism, and unprecedented performance, the proposed fiber oscillator should be very attractive for various applications in ultrafast science and technology.
National Natural Science Foundation of China (NSFC) (61535009, 61827821, 11527808, 61605142, 61735007); Tianjin Research Program of Application Foundation and Advanced Technology (17JCJQJC43500).
1. L. F. Mollenauer and R. H. Stolen, Opt. Lett. 9, 13 (1984). [CrossRef]
2. K. Tamura, E. P. Ippen, H. A. Haus, and L. E. Nelson, Opt. Lett. 18, 1080 (1993). [CrossRef]
3. A. Chong, J. Buckley, W. Renninger, and F. Wise, Opt. Express 14, 10095 (2006). [CrossRef]
4. W. H. Renninger, A. Chong, and F. W. Wise, Phys. Rev. A 82, 021805 (2010). [CrossRef]
5. T. Schreiber, F. Röser, O. Schmidt, J. Limpert, R. Iliew, F. Lederer, A. Petersson, C. Jacobsen, K. P. Hansen, J. Broeng, and A. Tünnermann, Opt. Express 13, 7621 (2005). [CrossRef]
6. F. Röser, J. Rothhard, B. Ortac, A. Liem, O. Schmidt, T. Schreiber, J. Limpert, and A. Tünnermann, Opt. Lett. 30, 2754 (2005). [CrossRef]
7. S. Lefrançois, K. Kieu, Y. Deng, J. D. Kafka, and F. W. Wise, Opt. Lett. 35, 1569 (2010). [CrossRef]
8. M. Baumgartl, C. Lecaplain, A. Hideur, J. Limpert, and A. Tünnermann, Opt. Lett. 37, 1640 (2012). [CrossRef]
9. B. Ortaç, C. Lecaplain, A. Hideur, T. Schreiber, J. Limpert, and A. Tünnermann, Opt. Express 16, 2122 (2008). [CrossRef]
10. B. Ortaç, M. Baumgartl, J. Limpert, and A. Tünnermann, Opt. Lett. 34, 1585 (2009). [CrossRef]
11. W. Liu, H. Shi, J. Cui, C. Xie, Y. Song, C. Wang, and M. Hu, Opt. Lett. 43, 2848 (2018). [CrossRef]
12. K. Regelskis, J. Želudevičius, K. Viskontas, and G. Račiukaitis, Opt. Lett. 40, 5255 (2015). [CrossRef]
13. I. Samartsev, A. Bordenyuk, and V. Gapontsev, Proc. SPIE 10085, 100850S (2017). [CrossRef]
14. Z. Liu, Z. M. Ziegler, L. G. Wright, and F. W. Wise, Optica 4, 649 (2017). [CrossRef]
15. P. Sidorenko, W. Fu, L. G. Wright, M. Olivier, and F. W. Wise, Opt. Lett. 43, 2672 (2018). [CrossRef]
16. M. Olivier, V. Boulanger, F. Guilbert-Savary, P. Sidorenko, F. Wise, and M. Piché, Laser Congress Advanced Solid State Lasers (ASSL) (Optical Society of America, 2018), paper ATu1A.4.
17. P. V. Mamyshev, 24th European Conference on Optical Communication. ECOC ‘98 (1998), Vol. 1, pp. 475–476.
18. S. Pitois, C. Finot, L. Provost, and D. J. Richardson, J. Opt. Soc. Am. B 25, 1537 (2008). [CrossRef]
19. K. Sun, M. Rochette, and L. R. Chen, Opt. Express 17, 10419 (2009). [CrossRef]
20. T. North and M. Rochette, Opt. Lett. 39, 174 (2014). [CrossRef]
21. J. Želudevičius, M. Mickus, and K. Regelskis, Opt. Express 26, 27247 (2018). [CrossRef]
22. K. Regelskis and G. Raciukaitis, “Method and generator for generating ultra-short light pulses,” International patent application WO2016020188 A1 (July 21, 2015).
23. J. W. Nicholson, J. Jasapara, W. Rudolph, F. G. Omenetto, and A. J. Taylor, Opt. Lett. 24, 1774 (1999). [CrossRef]
24. A. Chong, H. Liu, B. Nie, B. G. Bale, S. Wabnitz, W. H. Renninger, M. Dantus, and F. W. Wise, Opt. Express 20, 14213 (2012). [CrossRef]
25. A. Khanolkar, C. Ma, and A. Chong, Conference on Lasers and Electro-Optics (Optical Society of America, 2018), paper JTh2A.111.