Chinese Optics Letters, Volume. 22, Issue 12, 121404(2024)

All-fiber and self-starting Yb-doped Mamyshev oscillator based on nonlinear polarization evolution

Zecong Li1... Kai Wang1, Zinan Huang1,2, Qianqian Huang1, Yanlü Lin1, Weixi Li1, Zhichao Luo3,** and Chengbo Mou1,* |Show fewer author(s)
Author Affiliations
  • 1Key Laboratory of Specialty Fiber Optics and Optical Access Networks, Shanghai Institute for Advanced Communication and Data Science, Joint International Research Laboratory of Specialty Fiber Optics and Advanced Communication, Shanghai University, Shanghai 200444, China
  • 2Key Laboratory of Space Active Opto-Electronics Technology, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China
  • 3Laboratory of Nanophotonic Functional Materials and Devices, School of Information and Optoelectronic Science and Engineering, South China Normal University, Guangzhou 510006, China
  • show less

    We experimentally demonstrate a Yb-doped all-fiber mode-locked laser based on the Mamyshev mechanism. The entire experimental setup operates only by injecting pump powers and adjusting polarization controllers (PCs), which realizes self-starting. Two types of pulse patterns are observed at different pump powers and polarization states, including single pulses and up to eight-pulse bound-state pulses. The operating wavelength of single-pulse mode-locking switching between 1072.3 and 1043.1 nm can be realized by increasing the pump power while keeping the PCs in a fixed state. The design can provide an attractive experimental model for all-fiber and self-starting Mamyshev oscillators.

    Keywords

    1. Introduction

    Mode-locked (ML) fiber lasers develop rapidly because of their ability to deliver ultrashort pulses[1]. Different from free-space solid-state counterparts in solid lasers, ML fiber lasers offer the advantages of small size, low cost, and good heat dissipation. Meanwhile, due to their high beam quality, high integration, and high reliability, ML fiber lasers have remained an enduring research hotspot for scholars in the optical field in recent years[25]. Because of parasitic nonlinear effects in the fiber, ML fiber lasers still struggle to obtain peak power comparable to free-space lasers. The accumulation of excessive nonlinear effects in the fiber results in pulse splitting, thereby hindering the acquisition of high-energy pulses. By utilizing a saturable absorber (SA) element into the laser cavity, passive mode-locking is an effective way to generate ultrashort high-energy pulses. Material-based SAs[68] and interference-based SAs[9,10] have been widely applied in ML fiber lasers. Nevertheless, a large number of experimental results indicate that it is hard for general SAs to provide a perfect solution. Material-based SAs such as carbon nanotubes and graphene are easily damaged at high power, and the transmission-intensity curves of interference-based SAs do not increase monotonically, thus limiting the peak power and energy of the generated pulses.

    In order to obtain high-performance pulses, Mamyshev oscillators (MOs) have attracted more attention in recent years for their excellent output performance[1113]. In 1998, Mamyshev first proposed the technique of pulse reamplification and reshaping based on self-phase modulation (SPM) and the offset spectral filtering effect[14]. An equivalent artificial SA with a modulation depth of up to 100% can be achieved by introducing two transmission curve-shifted bandpass filters separated by a section of nonlinear fiber in the laser cavity. Input pulses pass through the first filter. Due to the SPM effect of the optical fiber, pulses with higher peak power are spectrally broadened and sufficient to cover both filters. In contrast, low-intensity pulses and continuous waves (CWs) will be blocked because of insufficient spectral broadening. In this condition, the CW lasing in MOs is greatly suppressed, showing a great advantage over conventional ultrafast fiber lasers in obtaining stable high-performance pulses[15]. Moreover, the superiority of MOs can be attributed to their tolerance to nonlinear phase accumulation, which is higher than other traditional SAs previously[16]. To sum up, these characteristics indicate the great potential of MOs to be high-energy, short-pulse fiber sources, which have promising applications in biomedical imaging and material processing.

    The suppression of low-intensity pulses and CWs prevents an MO cavity from resonating, posing a technical challenge for MOs that cannot self-start because of a large modulation depth. Therefore, most high-energy MOs require a picosecond or femtosecond seed pulse from an additional oscillator to initiate pulsation[13,1618]. In addition, in order to start MOs, researchers have also proposed several other methods, including modulation of the pump power[19,20], adding an additional starting arm[2123], or using fiber Bragg grating (FBG) filters with a temperature controller[24,25]. These approaches undoubtedly increase the cost and complexity of the experimental setup. Although some designs have achieved self-starting operation[26,27], the use of free-space elements in the cavity undermines the advantages of fiber lasers in terms of high integration and reliability, and the high mode-locking threshold creates the limitation of studies in MOs. In addition, most of the studies on MOs focus on the generation of single pulses. It is also important to study the operation of muti-pulse patterns in MOs. In 2024, Zhang’s group proposed that two offset spectral filters in MOs with overlapping bandwidths could form multiple pulses as the pump power increased due to the narrow filter interval[28]. For instance, the study of bound-state pulses can help develop greater communication capacity during fiber optic transmission[2931]. Therefore, it is necessary to design an all-fiber and self-starting MO, which will achieve its performance benefiting practical applications, and study its pulse dynamic patterns.

    In this paper, we experimentally demonstrate an all-fiber and self-starting all-normal dispersion Yb-doped fiber MO. Attributed to an artificial SA based on a nonlinear polarization evolution (NPE) structure, the starting of mode-locking can be easily triggered in a relatively low mode-locking threshold by simply adjusting the pump power and polarization controllers (PCs) in the cavity without any other operation. Two dynamic patterns of pulses, including single pulses and up to eight-pulse bound-state pulses, are observed at different pump powers and polarization states. In particular, the operating wavelength of single-pulse mode-locking switching between 1072.3 and 1043.1 nm can be observed by solely increasing the pump powers without rotating the PCs. To the best of our knowledge, this is the first experimental research on self-starting, all-normal dispersion, pump-controlled wavelength switchable, and all-fiber configuration in a Yb-doped MO cavity based on an NPE structure.

    2. Experimental Setup

    The schematic of the all-fiber and self-starting MO is depicted in Fig. 1(a). It consists of two cascaded arms (ARM1, ARM2), i.e., each arm incorporating a bandpass filter (BPF1, BPF2). BPF1 and BPF2 show the central wavelengths of 1066 and 1070 nm, and the 3 dB bandwidths of 3.5 and 5.6 nm, respectively, as shown in Fig. 1(b). Two segments of 0.75 m Yb-doped fibers (YDF1 and YDF2, Yb1200-4/125, LIEKKI) with a peak core absorption of 1200 dB/m at 976 nm are used to provide optical gain. In each arm, a single-mode laser diode (LD1, LD2) with a maximum pump power of 750 mW optically pumps a segment of gain fiber via a wavelength division multiplexer (WDM1, WDM2), and a polarization-dependent isolator (PD-ISO) sandwiched between two PCs forms an artificial SA based on NPE. The pulses are extracted through a fiber optic coupler (OC), which extracts 10% laser energy from the MO. The total cavity length is 13.4m, corresponding to the fundamental repetition rate of 15.3 MHz. Except for two segments of the gain fiber, all the other fibers in the cavity are HI 1060. The MO operates in the all-normal-dispersion region, and the net group velocity dispersion of the cavity is +0.3ps2. An optical spectra analyzer with minimum resolution of 0.02 nm (OSA, Yokogawa AQ6370D), a high-speed oscilloscope (OSC, KEYSIGHT DSO90804A) with a 12.5 GHz bandwidth photodetector (PD, Newport 818-BB-51 F), a 3.5 GHz bandwidth radio frequency (RF) analyzer (AgilentN9020A), and a commercial optical autocorrelator (APE Autocorrelator) are employed to monitor the spectra, pulse trains, stability of operation, and pulse width.

    (a) Experimental setup of the all-fiber, self-starting Yb-doped MO. (b) The transmission spectra of BPF 1 and BPF 2.

    Figure 1.(a) Experimental setup of the all-fiber, self-starting Yb-doped MO. (b) The transmission spectra of BPF 1 and BPF 2.

    3. Results and Discussion

    According to the theory described in Ref. [27], when the transmission spectral interval between the two BPFs decreases, the modulation depth of the equivalent SA also decreases. By reducing filter spacing and overlapping filters, it is possible to initialize the mode-locking in MOs without an external seed. In our experiment, the filtering spectra of the two BPFs are somewhat overlapped, which makes small noise fluctuations with low peak power passing through both filters and leading to the initiation of pulse operation. An NPE-based artificial SA consisting of two PCs and a PD-ISO is incorporated in each arm, which assists in initiating the mode-locking. Besides, the self-starting mode-locking is stabilized against the typical CW-breakthrough in such mechanisms by the inherent spectral offset filtering in the MO. After the NPE structures are utilized to initiate the MO, it is the hybrid mode-locking mechanism including the NPE-based artificial SA and the SA based on the Mamyshev mechanism that is responsible for the mode-locking operation in our experiment.

    During the experiment, at the beginning of pump power elevation, only the amplified spontaneous emission and the filter spectra of the BPF2 can be observed. When the other parameters are fixed and only the pump powers are adjusted to 130 mW from LD1 and 115 mW from LD2, the CW component appears at the center wavelength of 1067.3 nm. As the pump powers continue to increase, some unstable pulses suddenly appear on the oscilloscope. Then, the MO starts to operate in the multi-pulse state, accompanied by the broadening of the optical spectra. By decreasing the pump powers slowly and rotating the PCs appropriately, the original multi-pulse state cannot be maintained. Some pulses are annihilated in the cycle, causing energy to be redistributed in the cavity. Finally, periodically stable steady-state mode-locking can be observed on the oscilloscope. The MO enters single pulse operation eventually at the lowest pump power setting of 223 mW in each arm.

    The typical characteristics of single pulse operation are illustrated in Fig. 2 at pump powers of 294.7 mW from LD1 and 288.8 mW from LD2. Figure 2(a) shows the measured logarithmic spectra whose center wavelength (λc) is 1072.3 nm with a 3 dB bandwidth (Δλ) of 31.1nm. Under this condition, the output pulse trains are illustrated in Fig. 2(b) with a 0.06 µs interval between two neighboring pulses, corresponding to the fundamental repetition rate of 15.3 MHz. As shown in Fig. 2(c), the RF spectra indicate the signal-to-noise ratio (SNR) of 63.4 dB, indicating that the MO operates in a very steady state. The autocorrelation trace is depicted in Fig. 2(d). The pulse width is 7.0ps assuming a Gaussian pulse shape (orange dashed line). The time-bandwidth product is 56.76, indicating that the ML pulses are strongly chirped. The average output power is 1.89 mW.

    (a) Optical spectra with 0.02 nm resolution bandwidth (RBW). (b) Pulse trains. (c) RF spectra with 200 Hz RBW; inset: RF spectra with 10 kHz RBW. (d) Autocorrelation traces.

    Figure 2.(a) Optical spectra with 0.02 nm resolution bandwidth (RBW). (b) Pulse trains. (c) RF spectra with 200 Hz RBW; inset: RF spectra with 10 kHz RBW. (d) Autocorrelation traces.

    To further confirm the operating characteristics of the MO and understand the intracavity pulse evolution, we perform numerical simulations of the optical spectra and the autocorrelation traces by modeling MOs based on the extended Ginzburg–Landau equation, which can describe pulse transmission in fibers. Each parameter is set according to the experiment. The values of the second-order dispersion (β2), the third-order dispersion (β3), the gain saturation energy (Esat), and the Kerr nonlinear coefficient (γ) for the fibers are presented in Table 1. The spectral filters are represented by a Gaussian transmission function, which has similar effects on the super-Gaussian spectral shape of both filters used in this experiment. The MO is seeded with a Gaussian-shaped pulse with an average power of 1 W.

    • Table 1. Parameters Used in the Simulation

      Table 1. Parameters Used in the Simulation

      Fiber typeβ2 (ps2 m−1)β3 (ps3 m−1)γ (W−1 km−1)Esat (nJ)
      HI 10600.0230.0643.71
      YDF0.0250.0594.831.5

    When setting the values of the small-signal gain (g0) for the two arms to be g01=26dB and g02=25dB , after several roundtrips of evolution, a steady single pulse state is obtained eventually as shown in Fig. 3. It is shown in Fig. 3(a) that the operating wavelength and 3 dB bandwidth of the MO fit well with the linear experimental results in Fig. 2(a). Figure 3(b) shows the temporal result after convergence. The pulse width of the simulation is 5.5 ps.

    Numerical results of (a) linear spectra and (b) autocorrelation traces.

    Figure 3.Numerical results of (a) linear spectra and (b) autocorrelation traces.

    The pump powers of both LDs are continuously raised. When the pump powers of LD1 and LD2 are raised to 393.4 and 503 mW, respectively, the operating wavelength switching of single-pulse mode-locking is observed without the rotation of the PCs. The operating wavelength switches to the short wavelength, as shown in Fig. 4. Figure 4(a) shows that the center wavelength of the spectra switches to 1043.1 nm with a 3 dB bandwidth of 36.7 nm. The RF spectra in Fig. 4(b) demonstrate an SNR of over 60 dB, indicating that the MO can still operate very stably after the operating wavelength switching. Usually, in a unidirectional ring fiber laser, a polarizer such as a PD-ISO combined with intrinsic fiber birefringence can form a fiber birefringence filter[32]. In this experiment, the weak birefringence in HI 1060 can be used as a birefringence medium to form a birefringent filter with the PD-ISO in the MO cavity. Therefore, the wavelength-switchable phenomenon of our MO can be interpreted from the birefringence effect in the cavity.

    (a) Optical spectra with 0.02 nm RBW. (b) RF spectra with 200 Hz RBW.

    Figure 4.(a) Optical spectra with 0.02 nm RBW. (b) RF spectra with 200 Hz RBW.

    Keeping the pump power of 503 mW from LD2, the evolutions of the spectra, pulse duration, and average output power during the gradual increase of pump power of LD1 from 260.8 to 750 mW are shown in Fig. 5. Figure 5(a) depicts the evolution of the spectra. The spectra undergo switching of the operating wavelength twice during the process, which needs to be attributed to the birefringence effect in the MO cavity. Figure 5(b) shows the 3 dB bandwidth of the spectra monotonically increases from 28.1 to 55.8 nm when the pump power of LD1 is raised. The broadening of the spectra can be attributed to the enhancement of the SPM effect in the pulses, and the pulse spectra are broadened accordingly[33]. From Fig. 5(b), it can also be seen that the pulse width widens from 6.3 to 12.9 ps with the increasing pump power of LD1, which is mainly due to the increased gain in the cavity that enhances the broadening of pulse width caused by the chirp[34]. Figure 5(c) indicates that the average output power increases accompanied by the increasing pump power of LD1 and can reach 4.8 mW at the maximum pump power of LD1. In our experiments, the MO runs stably for more than 2 h at relatively low pump powers of the two LDs, i.e., the pump powers of LD1 and LD2 were less than 450 mW. However, at higher pump powers, because of the narrow filter interval, unavoidable CW-breakthrough may disrupt the single-pulse mode-locking operation in the MO. The MO single-pulse mode-locking may not keep stable for a long time and displays pulse splitting or even loss of the mode-locking regime, resulting in limited single-pulse energy.

    The evolution of the (a) spectra, (b) 3 dB bandwidth and pulse width, and (c) average output power during the gradual increase of the pump power of LD1 from 260.8 to 750 mW.

    Figure 5.The evolution of the (a) spectra, (b) 3 dB bandwidth and pulse width, and (c) average output power during the gradual increase of the pump power of LD1 from 260.8 to 750 mW.

    By adjusting the pump powers and the PCs appropriately, bound-state pulses can also be observed when the pump powers of two LDs exceed 310 mW. Figures 6(a)6(d) depict the characteristics of the two-pulse bound state at pump powers of 322.1 mW from LD1 and 315.6 mW from LD2. Figure 6(a) shows the spectra of bound-state pulses, which have dense periodic modulation with a period (λf) of 0.34 nm. The interval of the equal spacing of uniform pulse trains is 0.06 µs as shown in Fig. 6(b). Figure 6(c) shows the RF spectra with the SNR of 63.2 dB, indicating that the MO operates in a steady state. Figure 6(d) shows the autocorrelation trace of the two-pulse bound state. Due to the limited range of the autocorrelator, only half of the autocorrelation curve is recorded. The pulse separation is 11.5 ps, which corresponds well to λf. The pulse width of the output pulse is 4.1 ps, and the average output power is 1.42 mW.

    (a) Optical spectra with the resolution of 0.02 nm; inset: optical spectra with 1.5 nm wavelength range. (b) Pulse trains. (c) RF spectra with 200 Hz RBW; inset: RF spectra with 10 kHz RBW. (d) Autocorrelation traces.

    Figure 6.(a) Optical spectra with the resolution of 0.02 nm; inset: optical spectra with 1.5 nm wavelength range. (b) Pulse trains. (c) RF spectra with 200 Hz RBW; inset: RF spectra with 10 kHz RBW. (d) Autocorrelation traces.

    In our experiment, three-, four-, and five-pulse bound states, as well as a bound state comprising up to eight pulses, can be also achieved at higher pump powers under appropriate PC states. Figure 7 shows the corresponding autocorrelation traces. Due to the range limitation of the autocorrelator, for the seven- and eight-pulse bound-state pulses, only half of the autocorrelation curves can be observed within the range of the autocorrelator. Limited by the maximum pump powers of the two LDs, we do not observe bound states consisting of more pulses.

    The autocorrelation traces of (a) three-pulse, (b) four-pulse, (c) five-pulse, (d) six-pulse, (e) seven-pulse, and (f) eight-pulse bound states.

    Figure 7.The autocorrelation traces of (a) three-pulse, (b) four-pulse, (c) five-pulse, (d) six-pulse, (e) seven-pulse, and (f) eight-pulse bound states.

    In order to verify the universality of the experimental results and explore the role of the filter interval in generating single- or multi-pulse oscillations during the self-starting process, another pair of BPFs with a smaller filter interval is used in the experiment. BPF1 and BPF2 have the central wavelengths of 1064 and 1066 nm, and the 3 dB bandwidths of 3.6 and 3.5 nm, respectively. When LD1 and LD2 are set to 182 and 163 mW, respectively, i.e., lower mode-locking threshold, the MO starts to stabilize mode-locking. Nevertheless, no matter how the PCs are tuned, the MO never obtains single-pulse mode-locking but operates directly in a double-pulse bound state, as shown in Fig. 8. Figure 8(a) shows the spectra with dense period modulation with a period of 0.32 nm. Figure 8(b) shows the pulse trains with an interval of 0.07 µs. As shown in Fig. 8(c), as a result of the longer cavity length after replacing the filters, the repetition frequency becomes 14.54 MHz. The RF spectra demonstrate an SNR of about 55 dB, indicating that the MO is in a stable operating condition. The pulse width of a double-pulse bound state is measured to be 5.7 ps assuming a Gaussian fitting profile as shown in Fig. 8(d). The pulse separation is 12.3 ps, which also corresponds well to λf.

    (a) Optical spectra with the resolution of 0.02 nm; inset: optical spectra with 1.5 nm wavelength range. (b) Pulse trains. (c) RF spectra with 200 Hz RBW; inset: RF spectra with 10 kHz RBW. (d) Autocorrelation traces.

    Figure 8.(a) Optical spectra with the resolution of 0.02 nm; inset: optical spectra with 1.5 nm wavelength range. (b) Pulse trains. (c) RF spectra with 200 Hz RBW; inset: RF spectra with 10 kHz RBW. (d) Autocorrelation traces.

    According to the experimental results above, the center-wavelength interval between the bias filters is a crucial parameter that controls the state and starting dynamics of MOs. This parameter controls the effective saturable absorption of MOs. A larger filter separation between two BPFs corresponds to an SA with greater saturation power. Thus, large filter separation produces single-pulse states with high power, while smaller separation leads to multi-pulsing[35].

    Accordingly, for the purpose of avoiding the average output power limitation imposed by the CW-lasing and observing more multi-pulse phenomena, we assume reasonably that by varying the center wavelength separation between the two BPFs or using a tunable filter in the cavity, better and more in-depth studies of the pulse performance and the pulse dynamics patterns in all-fiber and self-starting MOs can be performed. Furthermore, the utilization of high-power optical elements and large mode area fibers may be also helpful to improve pulse performance[17].

    4. Conclusion

    In short, we have proposed and demonstrated an all-fiber and self-starting ML Yb-doped fiber MO with good simplicity and ease of starting. By slowly adjusting the pump powers and appropriately rotating the PCs, two dynamic patterns of pulses including single pulses and up to eight-pulse bound-state pulses are observed. Attributed to the weak birefringence effect formed by the PD-ISOs and HI 1060 in the MO cavity, the operating wavelength switching of single-pulse mode-locking is also observed. Furthermore, the filter interval plays an important role in generating single- or multi-pulse oscillations during the self-starting process. By optimizing the MO cavity parameters, we think that we can further improve the pulse performance and research the pulse dynamics in more depth.

    [14] P. Mamyshev. All-optical data regeneration based on self-phase modulation effect. 24th European Conference on Optical Communication. ECOC’98 (IEEE Cat. No. 98TH8398), 475(1998).

    Tools

    Get Citation

    Copy Citation Text

    Zecong Li, Kai Wang, Zinan Huang, Qianqian Huang, Yanlü Lin, Weixi Li, Zhichao Luo, Chengbo Mou, "All-fiber and self-starting Yb-doped Mamyshev oscillator based on nonlinear polarization evolution," Chin. Opt. Lett. 22, 121404 (2024)

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Lasers, Optical Amplifiers, and Laser Optics

    Received: Apr. 22, 2024

    Accepted: Jun. 26, 2024

    Posted: Jun. 27, 2024

    Published Online: Jan. 2, 2025

    The Author Email: Zhichao Luo (zcluo@scnu.edu.cn), Chengbo Mou (mouc1@shu.edu.cn)

    DOI:10.3788/COL202422.121404

    CSTR:32184.14.COL202422.121404

    Topics