1. Introduction

Passively mode-locked fiber lasers have attracted much attention because they can be used to construct compact, robust, and versatile ultrashort pulsed sources. Passive mode locking in fiber lasers can be achieved by nonlinear polarization evolution, nonlinear amplifying loop mirror, and real saturable absorbers (SAs). Compared with the other two techniques, incorporating a real SA into the laser cavity is expected to be a more efficient way to generate mode-locked pulses, because it is not sensitive to the polarization states of the cavity. Different SAs such as semiconductor saturable absorber mirror (SESAM) [1], single-walled carbon nanotubes (SWNT) [2], graphene [3, 4 ] and graphene derivatives [5] have been used in mode-locked fiber lasers. We have also demonstrated harmonic mode locking in fiber laser based on graphene saturable absorber [6, 7 ]. As a rising Dirac material, topological insulators (TIs) exhibit Dirac-like linear band dispersion [8], and attracted much interest in the field of passive mode-locked fiber laser [9–13 ]. In addition to the high modulation depth, it was found from Z-scan measurements that the TIs also possess large nonlinear refractive index [14]. Thus, the TIs could serve as both a high nonlinear photonic device and an ideal SA in the framework of fiber lasers. However, so far fiber lasers based on topological insulators saturable absorbers (TISAs) reported only with low output powers, about few or tens of milliwatts. The main limiting factor of the output power is the lower damage threshold of topological insulators if they are directly deposited onto the fiber end facet [15–18 ]. So it is of great importance to find effective ways to increase the optical damage power of a TISA.

There are several methods to achieve high power TISAs: TISAs based on glass plate [19], side-polished fiber [20, 21 ] and tapered fiber [22–24 ]. Obviously, glass plate based SA will destroy all-fiber structure of the fiber laser, the side-polished fiber based SA is not easy to fabricate and has smaller interaction area between light and TI. Tapered fiber based SA is a way which not only can improve the laser induced damage threshold, but also have effective absorption effect. Light propagating in the single mode fiber is guided in a nearly Gaussian mode by the fiber core. If the taper reduces the diameter of the fiber sufficiently gradually, the core-guided mode adiabatically transforms into a compressed mode strongly guided by the ultrathin fiber waist and with an evanescent part which extents largely outside the fiber clad. If there are SA nanomaterials on the surface of the tapered fiber, the light can be absorbed by the nanomaterials and a tapered fiber based SA is formed. In this paper, we demonstrate a high power Er:Yb doped double-clad fiber laser using tapered fiber based TISA. TISA based fiber laser operating above 1.6 μm is reported for the first time to our best knowledge. Moreover it is also the highest average output power yet reported in TISA based passive mode-locked fiber lasers.

2. Production and characterization of topological insulator saturable absorber

Here we use Bi₂Se₃ (99.999%, Alfa Aesar) to make the TISA. The Bi₂Se₃ nanosheets were synthesized by cost-effective hydrothermal intercalation and exfoliation method [25]. After preparing the Bi2Se3 nanosheets, they were dispersed in the acetone and ultrasonicated for 30 minutes. The concentration of TI acetone solution is ~0.1 mg/ml in this experiment. The tapered fiber which has a waist diameter of ~11 μm and a tapering length of ~8 mm was fabricated using a dedicated machine (optic fiber conic clink). The schematic for the fabrication of tapered fiber based TISA is shown in Fig. 1(a) . First, the TI acetone solution is dropped onto the taper which is previously fixed onto a piece of glass. Then an ASE light (1530–1560 nm) with a power of 16 mW is injected into the microfiber. When the light travels the waist of the microfiber, there will be a strong evanescent field outside the microfiber, and the optical deposition process starts because of the principle of optical tweezers [26]. The dimensions (length and thickness) of deposited TI can be roughly adjusted by controlling both the size of liquid drop and the deposition time. Here, the deposition time is about 10 minutes. The remaining TI acetone solution is taken out by an injector. Finally, the fabricated TISA is evaporated at room temperature and the tapered fiber based TISA is finished. Figure 1(b) shows microscopy image of fabricated microfiber-based TISA. The transmission of the TISA has been measured using a 1.61 μm home-made mode-locked pulse source [27] with an available average power of 40 mW, a pulse duration of ~880 fs and a repetition rate of ~4 GHz. We only provide the transmission curve versus the average power because the fluence could not be evaluated due to the impossibility of measurement of both the exact optical power of the evanescent field and the effective mode area in the waist of the taper. The data and the corresponding fitting curve are shown in Fig. 1(c). It can be seen that the modulation depth is ~4.3% and the non-saturable loss is ~60.7%. No polarization dependent losses were observed in our experiment.

 

Fig. 1 (a) Production of topological insulator saturable absorber. (b) Microscopy image of fabricated microfiber-based TISA. (c) Measured transmission curve and the corresponding fitting curve.

Download Full Size | PDF

3. Experimental setup

The schematic of the fiber ring laser is shown in Fig. 2 . We use a C-band double-clad Er:Yb doped fiber amplifier manufactured by Keopsys (model KPS-BT2-C-30-BO-FA). Two identical laser diodes operating at 980 nm and emitting about 2.5 W each are used in a counter propagating geometry. The 8 m-long double-clad fiber has a second-order dispersion of −15 ps²/km. A polarization controller (PC) was used to adjust the state of polarization in the cavity. A polarization-independent isolator (PI-ISO) was employed to force unidirectional operation of the cavity. A 20% output coupler is used to extract the power from the cavity. It is a good compromise allowing to maximize the output power as well as to get L-band mode-locked operation [28–30 ]. The cavity length is about 29.2 m, including 21.2 m standard single mode fiber (SMF) with GVD of −22 ps²/km. The net cavity dispersion was about −0.586 ps². The output beam is detected with a high-speed photodiode (Newport TIA 1200 13 GHz) and analyzed with either a high-speed oscilloscope (Tektronix TDS 6124C 12 GHz, 40 GS/s) or an electronic spectrum analyzer (Rohde & Schwarz FSP Spectrum Analyzer 9 kHz–13.6 GHz). Pulse duration is measured with an optical autocorrelator (Femtochrome FR-103 XL) with a scanning range scalable up to 170 ps, and an optical spectrum analyzer (Anritsu MS 9710C) is also used.

 

Fig. 2 Experimental setup. OC: output coupler, PC: polarization controller, TISA: topological insulators saturable absorber, PI-ISO: polarization-independent isolator.

Download Full Size | PDF

4. Experimental results and discussion

Self-starting mode locking appears when the pump power is increased to ~1.2 W. Because of the anomalous dispersion regime and the high gain of the amplifier, once the mode locking state is achieved, the laser operates in multiple-soliton regime with dynamical patterns depending on the exact cavity parameters [31]. The fundamental repetition rate of the cavity is 7.04 MHz. Figure 3(a) shows the temporal trace of multiple-solitons with the pump power of ~2 W. The corresponding spectrum and autocorrelation trace are shown in Fig. 3 (b). The spectrum has a central wavelength of 1610 nm with a 3 dB bandwidth of 0.91 nm. Such long wavelength emission is the consequence of the low-loss cavity [30]. The symmetric Kelly sidebands on the spectrum indicate that the laser operates in the soliton regime. The pulse duration is 3.1 ps as it is shown in the autocorrelation trace. The time-bandwidth product of the pulses is ~0.32, the small deviation from the 0.315 transform limit for a typical hyperbolic-secant soliton shape indicates that the pulses are almost transform-limited. In the experiment, the spectral width and pulse width of the solitons did not change significantly with different pump powers. By adjusting the PC, high order harmonic mode locking can be obtained in which the solitons distribute along the cavity with equal spacing.

 

Fig. 3 (a) Temporal trace of multiple-solitons. (b) The corresponding spectrum and autocorrelation trace (inset).

Download Full Size | PDF

The number of solitons in the cavity increases when the pump power increases. While the pump power increases and the polarization controller is adjusted, the solitons in the cavity tend to form a soliton bunch, in which many soliton pulses group themselves in a tight packet whose duration is much larger than the individual pulse width. Figure 4(a) shows the temporal trace of soliton bunch, it has very high intensity and few nanoseconds width. The corresponding spectrum and autocorrelation trace are shown in Fig. 4(b), the spectrum is centered at 1610 nm with a 3 dB bandwidth of 1.06 nm. The soliton pulses in the bunch have the pulse duration of ~2.76 ps. The large pedestal in the autocorrelation indicates that solitons are in relative motion inside the soliton bunch.

 

Fig. 4 (a) Temporal trace of soliton bunch. (b) The corresponding spectrum and autocorrelation trace (inset).

Download Full Size | PDF

By finely adjusting the PC, the soliton bunch can be split into many small soliton bunches distributed along the cavity with nearly equal spacing. We finally obtain high-order harmonic mode locking of soliton packets. In this state, there are many partial phase-locked solitons in every bunch as it is suggested by the optical spectrum of Fig. 5 . Figure 5 (a) shows the temporal trace of the 91st harmonic mode locking of soliton bunches [31, 32 ] with a pump power of 5 W, the width of the bunch is ~0.7 ns. The corresponding spectrum and autocorrelation trace are shown in Fig. 5(b). Figure 5(c) shows the corresponding radio frequency (RF) spectrum, the repetition frequency of the 91st harmonic mode locking is 640.9 MHz, the supermode suppression ratio is 35 dB. Evolution of the output power as a function of the pumping power is shown in Fig. 5(d), the output power increases almost linearly versus the pump power. While the pump power increases to 5 W, the average output power reaches 308 mW which is actually a record to our best knowledge. For higher pumping powers the absorption of the TISA saturates and the laser becomes continuous.

 

Fig. 5 (a) Temporal trace of 91st harmonic mode locking of soliton liquid with a pump power of 5 W. (b) The corresponding spectrum and autocorrelation trace (inset). (c) The corresponding RF spectrum. (d) Evolution of the output power as a function of the pumping powers.

Download Full Size | PDF

5. Conclusion

We have experimentally demonstrated a passive mode-locked Er:Yb doped double-clad fiber laser using a microfiber-based topological insulator (Bi₂Se₃) saturable absorber. The mode-locked fiber laser can operate at 1610 nm with a maximum output power of 308 mW. Our results show that we can get high average output power mode-locked fiber laser using microfiber-based topological insulator saturable absorber. This is the first report that the TISA based erbium-doped fiber laser operates above 1.6 μm and is also the highest output power yet reported in TISA based passive mode-locked fiber laser.