Fiber lasers operating in the 2 $\mathrm{\mu}$m band have attracted significant interest in the past decade, since this spectral range covers the absorption lines of various chemical bonds. The strong absorption in this wavelength region by water allows for the possibility of removing water-rich human tissues in laser surgery [1], in the breakup of kidney stones [2] and the treatment of skin pigmentation disorders [3]. The 2 $\mathrm{\mu}$m wavelength can excite the vibration mode overtones of the main functional groups in numerous polymers [4], making 2 $\mathrm{\mu}$m lasers highly applicable for polymer processing [5,6]. Moreover, the overlap with the rotational-vibrational absorption lines of diverse atmospheric compounds, such as $\mathrm {NH}_3$, $\mathrm {NO}_2$, $\mathrm {CO}_2$, and $\mathrm {H}_2\mathrm {O}$, allows for their spectroscopic detection [7,8]. For the detection of several unknown species, it is beneficial to rely on broad optical spectra toward the middle infrared (mid-IR), which are typically obtained through broadening of an initial short input pulse in a highly nonlinear medium [9,10] through supercontinuum generation (SCG). Pumping at 2 $\mathrm{\mu}$m provides the means to reach further into the mid-IR. Furthermore, nonlinear effects can also be exploited to convert frequency combs from the 2 $\mathrm{\mu}$m range into the mid-IR [11]. Such frequency conversion, as well as the spectral reach and efficiency of supercontinuum generation, critically depends on various phase-matching conditions related to the wavelength of the input pulsed pump [12]. For all these practical applications, there is therefore a need for tunable, but also compact and robust, pulsed sources at 2 $\mathrm{\mu}$m.

Thulium-doped (Tm$^{3+}$ doped) silica fibers are well-established gain media in the 2 $\mathrm{\mu}$m band. They can be pumped at 790 nm or in the telecom C- and L-bands, and provide a broad emission band that can cover wavelengths of 1.6 $\mathrm{\mu}$m to 2.1 $\mathrm{\mu}$m [13,14], depending on the pumping scheme and laser cavity design. This allows widely tunable mode-locked thulium-doped fiber lasers (MLTDFLs) to be built in order to address the aforementioned demands. Using a bulk diffraction grating, Meng et al. [15] achieved a MLTDFL with 1860 nm to 2060 nm tunability. However, the incorporation of a bulk diffraction grating in the laser cavity as a wavelength-selective element adds extra complexity and loss, making the system less robust and more expensive. All-fiber tunable MLTDFLs have been realized based on various wavelength-selective elements, for instance a birefringent filter (Lyot filter) [16–19] or multimode interference filter [20]. However, the wavelength-selective elements in these demonstrations are quite sensitive to ambient disturbances. More stable demonstrations of all-fiber tunable MLTDFLs have been presented, such as one based on a chirped fiber Bragg grating (CFBG) array in combination with an intensity modulator, at the price of added complexity [21]. Another filtering approach is based on bending a single-mode fiber to suppress lasing at higher wavelengths [22], enabling tuning of the cavity on the blue side of the Tm$^{3+}$ emission spectrum up to the maximal emission wavelength at 1892 nm.

In this paper, we demonstrate a widely tunable MLTDFL operating on the red-detuned side of the Tm$^{3+}$ emission spectrum, based on a mechanically robust CFBG as a wavelength-selective element. We show stable mode-locking at 2022 nm to 2042 nm by tuning a single CFBG, limited by the translation stages in our experimental setup. This is, to the best of the authors’ knowledge, the first mode-locked Tm$^{3+}$ doped fiber laser using a tunable CFBG as wavelength-selective element.

The experimental setup of the passively mode-locked Tm$^{3+}$ doped fiber laser is presented in Fig. 1. A linear laser cavity is formed between a CFBG and a semiconductor saturable absorber mirror (SESAM). The custom-built CFBG spans a broad reflection band from 2034 nm to 2051 nm [see Fig. 3(a)] and is added to define the lasing wavelength. The SESAM is mounted 150 $\mathrm{\mu}$m away from the fiber ferrule of an FC/PC connector and has a recovery time of 10 ps (Batop SAM-2000-30-10ps), such as to initiate self-starting passive mode-locking. A wavelength-division multiplexer (WDM) is used to launch up to 1.5 W of optical pumping at 1565 nm into the core of a 1.5 m long Tm$^{3+}$ doped fiber (OFS TmDF200). With a fused fiber 2$\times$2 directional coupler, we extract 10% of the power from the laser cavity. We monitor the laser output using an optical power meter, an optical spectrum analyzer (OSA), an optical autocorrelator, and a fast photodetector with 22 GHz bandwidth in combination with an oscilloscope and a radio-frequency (RF) analyzer.

 

Fig. 1. Experimental setup of Tm$^{3+}$ doped passively mode-locked fiber laser.

Download Full Size | PDF

In this experimental setup, the CFBG serves as a wavelength-selective element to enforce lasing on the long-wavelength tail of the Tm$^{3+}$ emission spectrum, where the gain cross section is low. When the CFBG is replaced by a retroreflector, lasing occurs at 1930 nm, toward the peak of the gain. The custom-designed CFBG is directly inscribed into the core of a stripped Ge-doped silica fiber using a cylindrical lens (25 mm focal length), a phase mask, and 100 fs long laser pulses at 800 nm central wavelength with 460 $\mathrm{\mu}$J pulse energy and a repetition rate of 250 Hz. To achieve a broad bandwidth, several scans across the fiber core along the fiber length were performed, resulting in a total grating length of 50 mm. After the inscription of the CFBG, the fiber was recoated with a low-index polymer coating (AngstromBond DSM DF-0016). A detailed description of the inscription setup can be found in the first section of Ref. [23]. The fabricated CFBG has a center grating period of 709 nm with a linear chirp rate of 1.26 nm/cm. It provides a high reflectivity ($>90$%) from 2034 nm to 2051 nm [see Fig. 3(a)], and adds a negative dispersion of $-61.8\,\mathrm {ps}^2$ to the laser cavity, which dominates the lower dispersion of all other fiber components, which we estimate to be $+0.9\,\mathrm {ps}^2$.

Without applying any tension or compression to the CFBG, the laser was passively mode-locked and the laser spectrum was initially centered at 2034.6 nm. We achieve wavelength tuning using the beam bending technique, where we shift the Bragg wavelength of the CFBG by applying tension or compression, as described in detail in Ref. [24]. To do so, we glued the fiber section containing the CFBG onto the surface of a 153 mm long and 0.5 mm thick flexible steel sheet. This steel sheet was mounted between a fixed stage and a linear translation stage, as shown in Fig. 2(a). By moving the translation stage inward, controlled by a micrometer screw, we bent the steel sheet either up or down, thereby stretching or compressing the CFBG, as illustrated in Figs. 2(b) and 2(c), respectively. This leads to a shift in the CFBG reflection band, as shown for different bending radii in Fig. 2(d). This shift occurs as a result of two effects. First, the actual grating period is elongated or shortened. Second, the applied mechanical strain induces a change in the refractive index due to the elasto-optic effect [24].

 

Fig. 2. (a) Setup to tune the CFBG when glued on a flexible steel sheet. Moving the translation stage inward, the steel is bent up or down, thereby (b) stretching or (c) compressing the CFBG. (d) Resulting red- and blueshifted CFBG reflection bands at different bending radii.

Download Full Size | PDF

The shift in the CFBG reflection band results in tuning of the lasing wavelength in the mode-locked laser cavity. Figure 3(a) shows the optical output spectrum, together with the CFBG reflection band when no strain is applied to the CFBG. The optical spectrum has a full width at half maximum (FWHM) of 0.45 nm and is centered around 2034.6 nm. The emission occurs on the short-pitch side of the CFBG reflection band; we attribute this to the higher Tm$^{3+}$ gain toward shorter wavelengths. This agrees with the forward and backward amplified spontaneous emission (ASE) spectra of the pumped Tm$^{3+}$ doped fiber in Fig. 3(b), which show a decrease within the CFBG spectral band indicated in gray from the blue to the red side.

 

Fig. 3. (a) Optical spectrum (red) within the CFBG reflection band (blue) for the laser cavity without strain on the CFBG. (b) Forward and backward ASE spectrum of the Tm$^{3+}$ doped fiber. The spectral reflection band of the CFBG without strain is indicated in gray. (c) RF spectrum and (d) pulse train of the laser output.

Download Full Size | PDF

We characterized the laser output with a fast photodetector connected to an RF analyzer and an oscilloscope. The RF trace in Fig. 3(c) indicates a fundamental repetition rate of 9.4 MHz with a signal-to-noise ratio of 60 dB, consistent with the estimated cavity length of 11 m. The oscilloscope trace in Fig. 3(d) shows a stable pulse train. From this, we conclude that the laser operates in the CW mode-locked regime [25]. Figure 4(c) shows the measured intensity autocorrelation. Assuming a $\mathrm {sech}^2$ pulse form, we fit a pulse duration of 10.9 ps, which corresponds to a time-bandwidth product (TBP) of 0.36, which is 14% above the Fourier limit. The average output power remained at $12.6\,\mathrm {dBm}\pm 0.1\,\mathrm {dBm}$ during 10 h of continuous operation, without any signs of $Q$ switching. The mode-locking threshold was found at 0.6 W of pump power and a slope efficiency of 1.6%.

 

Fig. 4. (a) Tuned optical output spectra (blue to red from left to right). Corresponding intensity autocorrelation traces for (b) the most blue-detuned emission, at 2022.1 nm, (c) the original emission, at 2034.6 nm, and (d) the most red-detuned emission, at 2042.2 nm.

Download Full Size | PDF

The laser output spectrum exhibits several side maxima on its long-wavelength tail [see Figure 3(a)], identified as Kelly sidebands, which occur in fiber lasers with negative dispersion as a result of the interaction of a cavity soliton with its dispersive wave [26–28]. Kelly sidebands have been reported for various laser cavity designs similar to ours, i.e., laser cavities that comprise a CFBG in combination with a saturable absorber. However, when the laser output spectrum is located completely within the CFBG reflection band, Kelly sidebands usually occur symmetrically on both tails of the laser emission spectrum [29,30]. In other demonstrations, when the laser emission occurs at the edge of the CFBG reflection band, such as in our case, Kelly sidebands appear only on one side of the laser spectrum [31,32]. Simulating the soliton formation based on a nonlinear Schrödinger equation pulse propagating model, it was shown that a CFBG can suppress the dispersive wave on one side of the laser spectrum, thereby allowing the formation of Kelly sidebands only on the opposite side of the spectrum [32].

To confirm that the maxima are Kelly sidebands, we varied the pump power between 0.6 W and 1.5 W and observed that the positions of the sidebands remained the same. This agrees with the expected behavior of Kelly sidebands, whose position is independent of the peak power of the pulse circulating inside the laser cavity [26]. In addition, the sideband separation from the carrier wavelength ($\Delta \lambda$) scales with the sideband order ($N$) as $\Delta \lambda \propto \sqrt {N+\mathrm {const.}}$, which also confirms the presence of Kelly sidebands [26].

Finally, we measured almost transform-limited output pulses with a TBP that is 14% above the transform limit. For similar laser cavities that comprise a CFBG in combination with a saturable absorber, similar values, ranging from 17% to 27%, were reported [29–32].

Figure 4(a) shows the obtained optical output spectra over our accessible tuning range. Without applying any tension or compression to the CFBG, the laser spectrum was initially centered around 2034.6 nm (gray trace). We could obtain a blueshift of up to 12.5 nm by compression tuning (blue traces), and a maximum redshift of 7.6 nm by tensile tuning (red traces). Thus, an overall tunability of 20.1 nm, spanning from 2022.1 nm to 2042.2 nm, was achieved. We are able to tune the CFBG without damaging it and can repeatedly set an emission wavelength within the tuning range. The output spectrum is not sensitive to ambient disturbances, such as vibrations or touched single-mode fibers. While tuning the output wavelength, we are only limited by the force that the translation stage can apply in order to bend the steel sheet. With our stage, we can inflect the 153 mm long and 0.5 mm thick steel down to a curvature radius of 67 mm.

The oscilloscope trace and RF spectrum of the pulse train confirmed that the fiber laser remained in the CW mode-locked regime throughout the whole tuning measurement. Intensity autocorrelation traces were measured at 2022.1 nm, 2034.6 nm, and 2042.2 nm and are shown in Figs. 4(b)–4(d). Assuming a $\mathrm {sech}^2$ pulse shape, we obtain pulse durations of 12.8 ps, 10.9 ps, and 9.0 ps, respectively. The decreasing pulse duration agrees with an increase in the spectral width, from 0.45 nm at 2022.1 nm to 0.50 nm at 2042.2 nm [see Fig. 4(a) and Table 1]. From this we obtain TBPs between 0.36 and 0.42, i.e., 14% and 34% over the Fourier limit of 0.315 for a $\mathrm {sech}^2$-shaped pulse.

Table 1. Spectral Width, Pulse Duration, and TBP at Different Emission Wavelengths

View Table

When tuning from 2022.1 nm to 2042.1 nm, the insertion losses of the WDM and the directional coupler each increase by 1 dB. This yields an additional loss of approximately 4 dB per round trip, along with the subsiding Tm$^{3+}$ gain toward longer wavelengths. This leads to a 3.8 dB lower output power at 2042.1 nm, compared with the output power at 2022 nm [Fig. 4(a)]. With the same difference in pulse energy, the SESAM remains saturated for a longer time span at shorter wavelengths, resulting in the measured increased pulse duration. Since the Kelly sidebands persist and the TBP remains close to the transform limit, we conclude that the fiber laser operates in the soliton regime for all emission wavelengths. We believe that the operating wavelength of our MLTDFL could be further shifted up to 2100 nm, since CW operation of Tm-doped fiber lasers has been demonstrated up to 2200 nm using uniform fiber Bragg gratings to obtain wavelength-selective emission [33]. However, owing to the subsiding Tm$^{3+}$ gain toward higher wavelengths, this would require a longer thulium-doped fiber in conjunction with a suitable CFBG.

In conclusion, to the best of our knowledge, this is the first demonstration of a wavelength-stabilized tunable mode-locked Tm$^{3+}$ doped fiber laser emitting transform-limited pulses at wavelengths between 2022.1 nm and 2042.2 nm. The tunability was demonstrated by applying strain to a CFBG. This approach allows the operating wavelength to be set precisely, and it stays remarkably stable over a long time. Our demonstration provides a mean for the development of stable, broadband laser sources in the 2 $\mathrm{\mu}$m band for practical applications outside the laboratory.