Pulsed lasers operating at mid-infrared (IR) are promising light sources because they can be used in many practical applications, such as eye-safe light detection and ranging (LIDAR),^[1] gas sensing,^[2] free-space optical communication,^[3] and laser surgery.^[4] Recently, the use of optical fiber laser technique in the mid-IR spectral region exhibits several advantages over solid-state counterparts in terms of beam quality, reliability, and environmental stability.^[5,6] To obtain sub-picosecond ultra-short pulses, mode-locking technique is commonly used. Mode-locking accomplished by locking relative phases of multiple longitudinal modes within the cavity is realized through two schemes: active and passive. To achieve passive mode-locking in a fiber laser, the key element is a saturable absorber (SA). Until now, commonly-used saturable absorbers have been based on III–V compound semiconductors because of their proven practical performance.^[7] However, due to their limited operating bandwidth and the need for expensive fabricating facilities, many investigations have been searching for an alternative. In the past decade, carbon-based materials such as carbon nanotubes (CNTs),^[8–15] graphene,^[16–25] graphene oxide (GO),^[26–31] and graphite^[32–34] have been intensively investigated as alternatives. Recently, topological insulators (TIs),^[35–52] transition metal dichalcogenides (TMDCs),^[53–74] gold nanoparticles,^[75–80] black phosphorus (BP),^[81–84] filled skutterudites,^[85] and MXene^[86] have also been identified as efficient saturable absorption materials.

Among these saturable absorption materials, TIs have attracted a great amount of technical attention in the field of condense matters because they possess extraordinary charge and spin properties on the edge or surface.^[87] TIs feature gapless metallic states on the surface due to band conversion by combination of strong spin–orbit coupling and time-reversal symmetry even if the bulk interior of the material exhibits insulating energy gaps. Two-dimensional (2D) topological insulator was theoretically predicted by Bernevig et al.^[88] Mercury telluride (HgTe) quantum well structures were experimentally identified as a 2D topological insulator by König et al. in 2007.^[89] After that, bulk crystal of Bi_{1 − x}Sb_x, bismuth selenide (Bi₂Se₃), bismuth telluride (Bi₂Te₃), and antimony telluride (Sb₂Te₃) were subsequently confirmed to be three-dimensional (3D) topological insulators.

The potential use of TIs in the field of photonic applications as a saturable absorber was discovered by Bernard et al. in 2012.^[35] Soon after that, intensive investigations on the use of TIs as saturable absorbers for Q-switching or mode-locking have been conducted.^[35–52] TIs have small bandgap bulk and gapless surface. Therefore, ultrabroad saturable absorption could happen under the help of Pauli-blocking phenomenon. Intensive experimental investigations on the use of Bi₂Te₃ with a small energy bandgap (∼0.15 eV) as an SA for Q-switching and mode-locking have been performed.^[35–43] Recently, our group has conducted a series of studies on the use of bulk-structured Bi₂Te₃ TI as an efficient SA for a simplified fabrication process.^[38–40] Bi₂Se₃ with an energy bandgap of ∼ 0.3 eV^[90] also has a saturable absorption wavelength range from visible to mid-infrared (4.1 μm). As a result, many studies on saturable absorption of Bi₂Se₃ in the wavelength regions of 1 μm to 1.5 μm have been reported.^[46–51] In 2014, Luo et al. demonstrated the use of a Bi₂Se₃ TI-based SA to Q-switched fiber laser in the wavelength region of 2 μm.^[52] However, to the best of our knowledge, there has been no report on the use of a Bi₂Se₃-based SA for mode-locked fiber laser in 2 μm wavelength region. In 2018, our group performed a theoretical investigation about electronic band structures and optical properties of Bi₂Te₃ and Bi₂Se₃ using density functional theory (DFT) calculations^[91] and confirmed that Bi₂Se₃ had an excellent potential as a saturable absorption material for mid-infrared laser mode-locking.^[91]

In this paper, we experimentally demonstrate the use of an SA based on a bulk-structured Bi₂Se₃ TI for the generation of mode-locked femtosecond pulses from an all-fiberized cavity in 2 μm wavelength region as an ongoing study. Our SA was prepared by depositing mechanically exfoliated Bi₂Se₃ TI layer on top of the flat side of a side-polished fiber. Using the prepared SA within a thulium–holmium (Tm–Ho) co-doped fiber ring cavity, stable mode-locked pulses with a temporal width of ∼ 865 fs could readily be generated through evanescent field interaction between oscillated beam and bulk-structured Bi₂Se₃ TI layer at a wavelength of 1912 nm. To the best of our knowledge, this is the first study that demonstrates the use of an SA based on a bulk-structured Bi₂Se₃ TI layer for femtosecond mode-locking of a 2 μm fiber laser.

The starting material, Bi₂Se₃ bulk crystal (99.999%, Alfa Aesar), was commercially available. To prepare the Bi₂Se₃ TI layer, we used mechanical exfoliation method.^[92] Using an adhesive tape, a thin layer Bi₂Se₃ was repeatedly stripped off from bulk crystal to obtain a micrometer-thick layer. Since our targeted thickness of the Bi₂Se₃ TI layer was at micrometer level, no special care was taken during the exfoliation process. The thickness of the Bi₂Se₃ TI film was measured to be ∼ 15 μm. It was measured using an alpha-step profiler. It is believed that the impact of the film thickness on the saturable absorption performance of our prepared bulk-structured Bi₂Se₃ TI-deposited side-polished fiber is negligible.^[93] Figure 1(a) shows the scanning electron microscope (SEM) image of the Bi₂Se₃ TI film. We observed a clean surface without distinctive layered structures, indicating that our prepared Bi₂Se₃ TI film was bulk-structured rather than nanostructured. It should be noticed that the measured SEM image of the bulk-structured Bi₂Se₃ TI surface exhibits no distinctive layered structures unlike few-layered, nano-structured ones, as reported in Ref. [94]. Figure 1(b) shows measured Raman spectrum of the Bi₂Se₃ TI film. We observed three Raman optical phonon peaks: at ∼72 cm⁻¹, at ∼ 130 cm⁻¹, and at ∼ 173 cm⁻¹. These three peaks are typical in bulk crystalline Bi₂Se₃.^[51,94] It is well-known that peak redshifts in the nanosheets by ∼ 3 cm⁻¹,^[94] however, the sort of redshift was not observed in our prepared Bi₂Se₃ TI film. This confirms that our prepared Bi₂Se₃ layer is in a bulk crystalline state.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. Measured (a) SEM image and (b) Raman spectrum of Bi2Se3 film surface.

We then measured x-ray photoelectron spectroscopy (XPS) spectrum of Bi₂Se₃ layer. Figure 2(a) shows Bi 4f spectrum and figure 2(b) shows Se 3d spectrum. Two peaks at 157.8 eV and 163.2 eV in the Bi 4f region are consistent with reported values of binding energies of Bi 4f_7/2 and Bi 4f_5/2 while two small peaks at 53.3 eV and 54.1 eV in the Se 3d region are consistent with those of Se 3d_5/2 and Se 3d_3/2.^[95,96] The existence of those additional peaks can be attributed to the oxidation of Bi and Se atoms on the surface.^[95] Our group has recently demonstrated that the excellent potential of Bi₂Se₃ as a saturable absorption material for mid-infrared laser mode-locking cannot be negated by oxidation.^[91]

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. Measured XPS spectra of (a) Bi 4f region and (b) Se 3d region.

For implementation of an all-fiberized SA, a side-polished fiber platform was used in this experiment. Side-polished fiber was prepared by polishing on side of SM2000 single mode fiber while the fiber was fixed onto a V-grooved slide glass. The distance between the flat side and the fiber core was measured to be ∼ 7 μm as shown in Fig. 3(a). Beam propagation loss and polarization dependent loss (PDL) of the prepared side-polished fiber were measured to be ∼ 1 dB and ∼ 0.05 dB, respectively. The prepared Bi₂Se₃ layer was transferred together with scotch tape and placed onto the top flat surface of a side-polished fiber. We used a small amount of index matching oil onto the fiber flat surface to induce proper light coupling between the prepared Bi₂Se₃ film and the fiber core. The beam interaction length of the Bi₂Se₃ TI-deposited side-polished fiber was ∼ 2 mm. The insertion loss and polarization dependent loss of the Bi₂Se₃ TI-deposited side-polished fiber were significantly increased to ∼ 3.7 dB and ∼ 5.9 dB, respectively.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (color online) (a) Schematic of the prepared Bi2Se3 TI-deposited side-polished fiber. (b) Measurement setup for nonlinear transmission curves of the bulk-structured Bi2Se3 TI-based SA. (c) Measured nonlinear transmission curve for TE-mode beam.

Next, we measured the nonlinear transmission curve of the Bi₂Se₃ TI-deposited side-polished fiber as a function of the incident optical pulse peak power to determine the nonlinear absorption performance. For the modulation depth measurements, we used a mode-locked, 1.9-μm fiber laser with a temporal width of ∼ 660 fs at a repetition rate of ∼ 36.9 MHz and the measurement setup is shown in Fig. 3(b). Since the prepared SA had a large polarization dependent loss of ∼ 5.9 dB, a polarization controller (PC) was incorporated into the setup to change the polarization status of the mode-locked optical pulses. A variable optical attenuator (VOA) was used to adjust the optical power of the mode-locked optical pulses. A 50:50 optical coupler was used to split the optical pulses into two ports. One of the two ports was connected to the prepared SA while the other was directly connected to a power meter to monitor the input optical power into the prepared SA. Another power meter was used to monitor the output power from the SA for its comparison with the input power. Modulation depth measurements were performed for both transverse electric (TE) mode beam and transverse magnetic (TM) mode beams. Due to high PDL, no saturable absorption was observed for the TM mode. Figure 3(c) shows the measured nonlinear transmission curve for TE mode together with a fitting curve.^[37] The saturation power and modulation depth of the prepared SA were estimated to be ∼ 28.6 W and ∼ 13.4%, respectively.

Figure 4 shows the experimental schematic of our constructed passively mode-locked fiber laser using the prepared SA. The gain medium was a 1-m-long Tm–Ho co-doped fiber (CorActive, TH512) with absorption of 13 dB/m at a wavelength of 1550 nm. The laser cavity consisted of a 1550/2000 nm wavelength division multiplexing (WDM) coupler, an optical isolator, a 90:10 coupler, a polarization controller, and a Bi₂Se₃-deposited side-polished fiber. A 1550-nm laser diode with a maximum pump power of ∼ 297 mW was used as a pump source that entered the gain fiber via the 1550/2000 nm WDM coupler. The mode-locked laser output beam was extracted from the ring cavity using a 10% output port of a 90:10 coupler.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (color online) Experimental configuration of our mode-locked fiber laser.

We monitored the laser output with a combination of a 16-GHz real-time oscilloscope and a 15-GHz photodetector while the pump power was enlarged. Stable mode-locked pulses were readily obtained at a pump power of 200 mW. The PC was properly adjusted to obtain stable mode-locked pulses. Since our prepared Bi₂Se₃ TI-deposited SA has a high polarization dependent loss, it is believed that both saturable absorption and nonlinear polarization rotation (NPR) contributed to the mode-locking of this laser.^[97] The pump power range for the fundamental mode-locking was found to be between 200 mW and 297 mW. It is well known that stable mode-locking at a fundamental resonance frequency is broken by increasing the pump power due to multiple-soliton generation and soliton energy quantization effect.^[98–100] It is also known that harmonically mode-locked pulses can be produced from a cavity since pulse splitting occurs when large photon energies exist within the cavity.^[101] During our experiment, no harmonic mode-locking phenomenon was observed.

Figure 5(a) shows the measured optical spectrum of the mode-locked pulses with its sech²() fitting curve. The center wavelength and 3-dB bandwidth were measured to be ∼ 1912.12 nm and ∼ 4.87 nm, respectively. Figure 5(b) shows the measured oscilloscope trace of output pulses while the inset of Fig. 5(b) shows a magnified view of a single output pulse. The period of the output pulses was measured to be ∼ 54.5 ns, corresponding to a repetition rate of ∼ 18.37 MHz. This measured repetition rate coincided with the fundamental cavity resonance frequency.

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. (color online) Measured (a) optical spectrum and (b) oscilloscope trace of output pulses. Inset: oscilloscope trace for a narrow span.

Next, we performed the autocorrelation measurement using a second harmonic generation (SHG)-based autocorrelator. Figure 6(a) shows the measured autocorrelation trace of the mode-locked pulses with a sech²() fitting curve. The pulse width was measured to be ∼ 835 fs. Considering the 3-dB bandwidth of ∼ 4.87 nm, the estimated time-bandwidth product was ∼ 0.337, which was slightly higher than the 0.315 of transform-limited sech²() pulses, indicating that the output pulses were slightly chirped. Finally, the electrical spectrum was measured to check the phase noise of mode-locked pulses as shown in Fig. 6(b). A strong signal peak with an electrical signal-to-background ratio of ∼ 65 dB was clearly observed at a fundamental repetition rate of ∼18.37 MHz. The inset in Fig. 6(b) shows the electrical spectrum with a wide span for 1 GHz and the observed clean harmonic frequencies, confirming that the output pulses are stable mode-locked pulses. We have checked the long-term stability of the mode-locked fiber laser by measuring the output laser for several hours, but did not find any problem.

Fig. 6.

	Figure Option ViewDownloadNew Window
	Fig. 6. (color online) Measured (a) autocorrelation trace and (b) electrical spectrum of output pulses. Inset: electrical spectrum for a span of 1 GHz.

Table 1.

Table 1.

Table 1. Output performance comparison between the present work and previous mode-locked fiber lasers using other saturable absorption materials at 2 μm wavelength region. . Saturable absorption Fiber platform Operating wavelength/nm Modulation depth/% Saturation level Pulse width/ps Ref. CNTs fiber ferrule 1930 NAa NAa 2.05 [13] CNTs tapered fiber 1885 NAa NAa 0.75 [14] CNTs fiber ferrule 1870 NAa NAa 0.45 [15] graphene fiber ferrule 1940 NAa NAa 3.6 [22] graphene fiber ferrule 1884 NAa NAa 1.2 [23] graphene fiber ferrule 1935 4 NAa 2.1 [24] graphene fiber ferrule 1876 3.7 110 μJ/cm2,b 0.603 [25] GO side-polished fiber 1910 NAa NAa NAa [28] GO side-polished fiber 1950 7.1 ∼ 14.83 Wc 0.59 [29] Bi2Te3 side-polished fiber 1935 20.6 ∼ 29 Wc 0.795 [39] Bi2Te3 tapered fiber 1909.5 9.8 NAa 1.26 [41] Sb2Te3 side-polished fiber 1945 NAa NAa 0.89 [45] MoS2 HR gold mirror 1905 13.6 23.1 MW/cm2,d 843 [58] WS2 side-polished fiber 1941 10.9 ∼ 1.9 Wc 1.3 [60] MoSe2 side-polished fiber 1912.6 4.4 ∼ 25.7 Wc 0.92 [68] WSe2 microfiber 1863.96 1.83 3.7 MW/cm2,d 1.16 [71] WTe2 microfiber 1915.5 31 7.6 MW/cm2,d 1.2 [74] gold nanorod fiber ferrule 1982 4.1 35.5 MW/cm2,d 4.02 [78] BP fiber ferrule 1910 4.1 NAa 0.739 [83] Bi2Se3 side-polished fiber 1912.12 13.4 ∼ 28.6 Wc 0.835 this work a Not available b saturation fluence c saturation power d saturation intensity.

Table 1. Output performance comparison between the present work and previous mode-locked fiber lasers using other saturable absorption materials at 2 μm wavelength region. .

Finally, the output performance of our mode-locked fiber laser was compared to that of mode-locked fiber lasers using other saturable absorption materials at 2 μm wavelength region. Results are summarized in Table 1. The modulation depth of our Bi₂Se₃-based SA is almost comparable to the modulation depth (ca. 13.6%) of the MoS₂-based SA,^[68] exhibiting a smaller modulation depth than Bi₂Te₃-based SA.^[39] In terms of the pulse width, our laser exhibits a reasonable level of the output pulse width (835 fs) for femtosecond applications.

In summary, we have experimentally demonstrated the use of a bulk-structured Bi₂Se₃ TI-based saturable absorber for the generation of femtosecond mode-locked pulses in the 2 μm wavelength region. The SA was constructed on a side-polished fiber platform by depositing a bulk-structured Bi₂Se₃ TI layer onto the flat side of the platform. Using the prepared SA within a Tm–Ho co-doped fiber ring cavity, stable optical pulses with a temporal width of ∼ 853 fs could readily be generated at ∼ 1912 nm.

Compared to nanostructured TI films that demand complicated high precision fabrication facilities, the fabrication process of bulk-structured TI films with non-negligible oxidation is practical and straightforward. This is a huge advantage of a bulk-structured TI-based SA. As a further study after our group’s theoretical prediction in Ref. [91], this experimental demonstration reaffirms that a Bi₂Se₃ is a superb base material for mid-infrared passive mode-locking even under oxidation.

We believe that bulk-structured TI films should be able to provide a low-cost material platform for the implementation of practical saturable absorber, which are essential for industrial ultrafast lasers.