1. Introduction

Over the past decades, mid-IR tunable femtosecond laser sources have been intensively investigated to satisfy the application requirements of infrared countermeasures, strong-field physics, attosecond science, and so on [1–3]. One of the most promising methods to obtain mid-IR tunable femtosecond laser is the fiber-based SSFS effect, that has demonstrated many advantages such as continuous wide wavelength tunability, excellent beam quality, good environmental reliability, and operation flexibility. The SSFS effect was first discovered by Mitstchke and Mollenauer in 1986 [4], and is the result of Raman self-pumping of soliton pulses propagating in the anomalous dispersion region of optical fibers. Until now, SSFS-induced tunable soliton sources have been widely studied in various fibers such as tellurite fiber [5–9], fluoride fiber [10–13], chalcogenide fiber [14] and silica fibers [15–20], capable of covering a wide spectral range from the near-IR to the mid-IR. Based on the SSFS effect operating in fluoride fiber, femtosecond solitons with the widest tunable range (2.3 µm) and the longest soliton wavelength (up to 4.3 µm) have been achieved [12]. Additionally, for the mid-IR spectral region from 2 to 3 µm, silica fibers were widely used to obtain all-fiber integrated tunable Raman soliton sources, owing to their good resistance against mechanical and laser-induced damages and excellent compatibility with other silica fiber laser systems [15,20]. A recent report also indicated that a large-mode-area silica fiber, is able to realize the tunable solitons with higher pulse energy and higher output power [17].

Among the silica fibers, the Tm-doped fiber can serve as both active fiber and Raman shifting fiber, capable of generating Raman solitons beyond 2 µm with ultrashort pulse duration and high peak power. Initially, much research in recent years has focused on the SSFS process based on conventional solitons as seed pulses. In 2007, using 2-µm conventional solitons in a mode-locked Tm-Ho-doped fiber oscillator as seed pulses and a Tm-Ho fiber amplifier, a 1.97-2.15 µm tunable Raman soliton laser with an output power up to 230 mW was first achieved by Kivistö et al. [21]. However, the maximum conversion efficiency over the tuning range was only 62%. The abrupt increase of fiber absorption beyond 2.1 µm and supercontinuum generation at high pump power resulted in the efficiency decrease of Raman conversion and the limit of tuning range. After that, using the SSFS approach, Dvoyrin et al. realized a soliton shift to 2.2 µm with several watts-level average power in a Tm-doped fiber master oscillator power amplifier (MOPA) seeded by two SESAM mode-locked fiber lasers [22]. Subsequently, they reported a Raman soliton generation from 1.98 to 2.22 µm in the similar configuration to previous work, scaling the pulse energy up to 75 nJ [23]. Nevertheless, over the entire tuning range of 1.98-2.22 µm, >40% of the soliton energy was not transferred to the Raman soliton, indicating a low efficiency of Raman conversion. The Raman soliton wavelengths were limited to <2.3 µm. In addition, the seed pulses in the majority of tunable Tm-doped fiber laser systems typically used conventional solitons with Kelly sidebands [18,19,21]. As is well known, the Kelly sidebands play an important role in the interpulse interaction, which can lead to pulse time jitter and signal distortion [24–26]. Moreover, it has been proved that the Kelly sidebands can limit the soliton amplification efficiency due to their higher gain coefficient than the soliton main lobes [27]. This indicates a possible limitation on the broadband and efficient Raman soliton generation in the amplifier due to the presence of Kelly sidebands.

Recently, by optimizing the pulse properties in oscillators and amplifiers with dispersion management during the SSFS process [28], Raman solitons with the wavelengths beyond 2.3 µm and high conversion efficiency of >90% have been successfully realized [29–31]. Most impressively, Wang et al. achieved watt-level monochromatic Raman solitons with the widest wavelength-tuning range of 1.9-2.36 µm and a maximum efficiency up to 97% in a Tm-doped fiber amplifier [30]. However, long lengths of dispersion compensating fibers are needed in the oscillators and amplifiers, which results in higher loss, limited power and uncontrollable repetition rate. The suppression of the Kelly sidebands of the conventional solitons might on the other hand present a way to overcome the above shortcomings and improve the performance of the SSFS-based Tm-doped fiber laser system. Some simple sideband-suppressed soliton mode-locked fiber lasers based on the FBG-induced spectral filtering effect where the dispersion compensating fibers were unused have been reported [32–34]. However, up to now, the use of such conventional solitons in the SSFS process has never been explored.

In this work, we demonstrate for the first time an efficient SSFS-based femtosecond laser source seeded by conventional solitons with suppressed Kelly sidebands. A 1.95 µm hybrid mode-locked Tm-doped fiber oscillator seed based on a homemade AFLF is used to obtain stable conventional solitons where the Kelly sidebands are suppressed by >21 dB. When compared with the same amplifier using sideband-unsuppressed conventional solitons from the same pumps, the maximum energy conversion and wavelength-tuning range of monochromatic Raman soliton were increased by ∼26.6% and 209.6 nm, respectively.

2. Experimental setup

The experimental setup of the all-fiber Raman soliton source is presented in Fig. 1. It mainly consists of a Tm-doped fiber oscillator seed and a typical one-stage Tm-doped fiber amplifier/shifter. The oscillator seed is a hybrid mode-locked fiber laser based on nonlinear polarization rotation (NPR) and carbon nanotube (CNT) as a saturable absorber. Employing a (2 + 1) ×1 ITF Labs’ combiner, a commercially multimode 793 nm laser diode (LD) with a maximal fiber output power of 12 W was coupled into the oscillator. The gain section was a 1.1-m Tm-doped fiber (DCF-TM-10/128, CorActive), featuring a circular fiber core with a numerical aperture (NA) of 0.22 and an octagonal inner cladding with a NA of 0.45. Behind the gain fiber, a 2-µm polarization-insensitive isolator (PI-ISO) (Advanced Photonics, USA) was used to achieve unidirectional operation. A 3-dB optical coupler (OC) was applied to extract 50% average power from the cavity. The 1% port of a 1/99 fiber coupler was used as the real-time monitoring port of the seed pulses and the 99% port was used to couple the seed pulse into the amplifier. The CNT film helps achieve the low threshold self-starting of hybrid mode-locking was sandwiched between two fiber ferrules. To obtain two various conventional solitons as seed pulses required in our experiment, a polarization-dependent isolator (PD-ISO) and a homemade AFLF were used as the polarizer respectively. As shown in the inset of Fig. 1, The PD-ISO was set at the same position with the AFLF. The PD-ISO combined with two polarization controllers (PCs) was used to obtain conventional solitons with strong Kelly sidebands. The AFLF is designed by sandwiching a piece of polarization-maintained fiber (PMF) with two 45° tilted fiber gratings (TFGs). Owing to the characteristics of the PMF and 45° TFGs, the AFLF possess high polarization extinction ratio (PER) and a thermal tuning ability for the wavelength. Thus, when the AFLF was used to replace the PD-ISO, the NPR effect and the filtering effect were introduced simultaneously. The more detailed description on preparation and features of the AFLF has been reported in [33]. The AFLF used in the experiment has a large filter depth of ∼9 dB. Using the AFLF, the Kelly sidebands of conventional solitons could be efficiently suppressed without changing the central spectrum shape and the pulse duration. The total cavity length of the oscillator seed was measured as ∼10.76 m, including 1.1 m TDF and 9.66 m SMF28e fiber. The value of net cavity dispersion at 1960 nm was estimated to be −0.865 ps². The seed signals were finally sent to the amplifier through another PI-ISO. The pump source and combiner used in the amplifier are the same as those used in the oscillator. The amplifier was composed of a 793 nm pump LD, a pump combiner and 4.5-m-long Tm-doped double cladding fiber (Nufern, SM-TDF-10P/130) that has a core/cladding NA of 0.15/0.46. A segment of SMF-28e pigtail fiber was spliced at the output port of the TDFA to strip residual 793 nm pump power.

 

Fig. 1. The experimental setup for the SSFS study in a one-stage TDFA seeded by conventional solitons with strong Kelly sidebands and the sideband-suppressed conventional solitons respectively. LD: laser diode, DC-TDF: double cladding Tm-doped fiber, PI-ISO: polarization-insensitive isolator, AFLF: all-fiber Lyot filter, OC: optical coupler, CNT-SA: carbon nanotube saturable absorber, PC: polarization controller, PD-ISO: polarization-dependent isolator.

Download Full Size | PDF

In this work, the output laser signal was detected by an InGaAs photodetector (EOT ET-5000F, USA) with a response time of approximately 28 ps, connected with a 500-MHz digital oscilloscope and an 18-GHz RF spectrum analyzer. The bandwidth of the InGaAs photodetector is larger than 12.5 GHz, as provided by the manufacturer. The spectral profiles of output pulses were monitored by an optical spectrum analyzer (Yokogawa AQ6375, Japan) with a high resolution of 0.05 nm and a wide spectrum response range covering 1.2-2.4 µm. In addition, an interference autocorrelator (APE PulseCheck, Germany) was employed to measure the pulse duration.

3. Experiment results and discussion

3.1 Conventional solitons in the hybrid mode-locked Tm-doped fiber laser seeds

In our experiment, the structure and total cavity length of the two mode-locked oscillator seeds were kept consistent, to eliminate any effect on soliton morphology due to a different cavity design. For the seed 1 (with the PD-ISO), the self-started mode-locking was achieved at a pump power of 1.25 W. The sideband amplitude was maximized by approximately adjusting pump power and PCs’ position. This way, soliton with sub-picosecond pulse duration could be obtained, owing to the pulse duration dependence on the Kelly sidebands [35]. When increasing the pump power to 1.4 W, a conventional soliton with strong Kelly sidebands was generated, shown as the blue curve in Fig. 2(a). The center wavelength and 3-dB spectral bandwidth of the soliton was 1953.2 nm and 4.27 nm, respectively. The spacing between the 1st-order Kelly sidebands and soliton main lobe was measured to be 7.04 nm, which is close to the dispersion-dependent theoretical value [26,36]. Because the Kelly sidebands are imposed on the soliton spectrum, and possess much higher amplitudes than the soliton components at the same wavelength positions, the Kelly sidebands can be suppressed using a band-pass filter technique [37]. Thus, in the seed 2, to efficiently suppress the soliton sidebands without changing spectral main shape and pulse duration, an AFLF with a free spectral range (FSR) of 20.8 nm was chosen, which is nearly 5 times of the soliton 3-dB spectral bandwidth in the seed 1. Controlling the temperature to match the AFLF transmission spectrum with the corresponding position of the soliton spectrum in the seed1 and appropriately adjusting the PCs, the sideband-suppressed conventional solitons would be obtained. Due to the slightly higher insertion loss of the AFLF than the PD-ISO, the mode-locking operation in the seed 2 was self-started at a pump power of 1.3 W. When the temperature of the AFLF was fixed at 33.1 °C, the Kelly sidebands of the soliton centered at 1953.2 nm were dramatically suppressed at a pump power of 1.52 W. The red curve presented in Fig. 2(a) shows the corresponding spectrum of the sideband-suppressed solitons. Compared to the solitons in the seed 1, the amplitude of the soliton sidebands was reduced by at least 21 dB. The inset in Fig. 2(a) shows the measured output spectra in a linear scale for the two oscillator seeds, exhibiting that the main lobes of two conventional solitons are completely consistent though the soliton sidebands have been significantly suppressed. In this case, the corresponding average powers were measured to be 16.5 mW (seed 1) and 14.9 mW (seed 2), respectively. Their difference in output powers was mainly due to the reduced amplified spontaneous emission (ASE) light and suppressed Kelly sidebands [33].

 

Fig. 2. (a) The output spectra of conventional solitons in the mode-locked Tm-doped fiber oscillator without/with an AFLF, inset: the soliton spectra in a linear scale. (b) The pulse autocorrelation trace in the seed 2, inset: the autocorrelation trace in the seed 1. (c) The oscilloscope trace captured in the seed 2. (d) The RF spectrum with a scanning range of 16 MHz corresponding to (c), inset: the RF spectrum in a scanning range of 300 MHz.

Download Full Size | PDF

Figure 2(b) shows the pulse autocorrelation traces corresponding to Fig. 2(a). When a sech²-profile fit was assumed, the same full widths at half maximum (FWHMs) and pulse durations were estimated to be 1.51 ps and 981 fs, respectively, showing that the pulse duration didn’t change with the suppression of the Kelly sidebands. The time bandwidth product (TBP) was calculated to be 0.329, close to the Fourier-transform limit. Figure 2(c) shows the captured soliton pulse sequence at a pump power of 1.52 W in seed 2. The measured repetition rate of 19.23 MHz was in good agreement with theoretical value dependent on total cavity length, implying the mode-locked fundamental frequency operation. The low amplitude fluctuation (<1.4%) of pulse sequence indicates a high temporal stability in the mode-locked seed oscillator. The radio frequency (RF) spectrum of solitons was also measured in a scanning range of 16 MHz, as shown in Fig. 2(d). A signal-to-noise ratio (SNR) higher than 66 dB was obtained. Besides, it can be seen from the inset of Fig. 2(d) that the interval of two adjacent spectrum components was stable and the peak intensity of the spectrum components monotonically decreased with the increased orders. These results exhibit the high uniformity and good stability of our seed pulse sequences.

3.2 Raman soliton generation in the TDFA

Using the SSFS process, we first investigated the spectrum evolution of Raman solitons in the TDFA seeded by the conventional soliton with strong Kelly sidebands, as the blue curve shown in Fig. 3. Initially, the soliton was amplified without any obviously morphologic changes with the pump power increasing. In particular, the Kelly sidebands that took up a part of the pulse energy were amplified rapidly, much more than the soliton main lobes. This is one reason why it is difficult to increase the amplification efficiency of conventional solitons. At a pump power of 0.912 W, we observed that part of the energy of the soliton main lobe was transferred to a longer wavelength. Further increasing the pump power to 1.65 W, the shifted soliton was completely separated from the Kelly sidebands, indicating that the Raman SSFS effect had been generated in the TDFA. It should be noted that the Kelly sidebands possess a broad time domain distribution and their corresponding peak power is too low to excite the SSFS effect in the fiber, despite their spectral amplitudes look quite high. Thus, in the process, the positions of Kelly sidebands always remained unchanged. However, their amplitudes were increased sharply with the amplifier’s pump power increasing, because of the higher gain coefficient of Kelly sidebands compared with the soliton main lobe [27]. Additionally, the Kelly sidebands on the long wavelength side of the seed pulses had higher peak power compared with those on the short wavelength side, due to the gain profile and the Raman effect [38]. Consequently, the Kelly sidebands could fleetly occupy the amplification energy that was originally intended for Raman soliton generation in the TDFA. At the pump power of 1.78 W, we observed a changed soliton shape with a new component arising at the long wavelength edge, which is resulted from the soliton interaction on the perturbations of the soliton frequencies and amplitudes [39–41]. When further increasing the pump power to 2.66 W, a new Raman soliton without separating completely from the sideband was generated, due to the high-order soliton fission and the resulting SSFS effect. This is a typical spectrum evolution phenomenon in the SSFS process and has been widely observed in previous reports [42,43]. Additionally, it can be observed that the amplitudes of the Raman solitons were decreased as the pump power was beyond 2.38 W. Such behavior is caused by two factors. One is that the partial soliton energy would be taken away from the SSFS process due to the nonsolitonic radiation (NSR) [44] and some higher order nonlinear effects [45,46] when the pump power exceeds a certain value. The other is that the energy loss in the SSFS process will aggravate once the soliton wavelength moves out from the bandwidth of the Tm³⁺ absorption and the high transmission window in the fiber.

 

Fig. 3. Spectra evolution of two Tm-based fiber SSFS systems at different pump powers, using the solitons with unsuppressed and suppressed Kelly sidebands as seed pulses respectively.

Download Full Size | PDF

To further explore the impacts of Kelly sidebands on the SSFS process, we compared the output spectra of Raman soliton sources based on the seeds with unsuppressed and suppressed Kelly sidebands. As illustrated in Fig. 3, their center wavelengths were both continuously tuned when the pump power was increased from 0.912 W to 2.66 W. However, at the same pump powers, the latter always possessed a wider wavelength-tuning range and a higher spectral peak intensity. At a pump power of 1.94 W, the sideband-suppressed laser system generated monochromatic Raman solitons with a center wavelength of 2240.6 nm a FWHM of 10.4 nm. As for the sideband-unsuppressed laser system, the corresponding soliton wavelength and the FWHM at the pump power of 1.94 W were 2031nm and 7.41 nm, respectively. In this case, their wavelength difference was calculated to be 209.6 nm, implying that the wavelength-tuning range of the Raman solitons was extremely limited due to the amplified Kelly sidebands. In the second tunable laser system, the first Raman soliton was further red-shifted and a larger tuning range of the second Raman soliton was realized as a result of the SSFS effect enhanced by the high-order soliton fission, as the pump power further increased. We observed two Raman solitons with different center wavelengths (2013.2 nm and 2315 nm) coexisting in the spectrum. For the sideband-unsuppressed SSFS system, the wavelengths of monochromatic Raman soliton were only shifted to 2194.8 nm at a high pump power of 2.38 W.

Figure 4(a) shows the shifted wavelengths of the first Raman solitons in the TDFAs with respect to pump power. In the suppressed-sideband Raman soliton source, the wavelength of the Raman soliton could be continuously tuned from 1.95 to 2.34 µm, as the pump power was increased from 1.22 to 2.95 W. While the first Raman soliton source was tunable over a spectral range of 1.95-2.27 µm with the same pump power range. It is worth noting that the wavelength shifting rate of two Raman soliton sources both increased at first and then gradually saturated with increased pump power. This can be attributed to the change in the fiber parameters (such as fiber loss, group velocity dispersion, nonlinear coefficient) and the increased infrared absorption of silica glass along at longer wavelengths. The conversion efficiency of Raman soliton in the amplifier is defined as the ratio of the shifted soliton output power to the overall output power from the amplifier, including the unconverted seed and amplified spontaneous emission [21]. Thus, using a band-pass filter to exact the shifted solitons from the output pulses, the corresponding Raman conversion efficiencies were obtained, as seen in Fig. 4(b). When the Kelly sidebands of seed pulse were efficiently suppressed, a high soliton conversion efficiency of more than 93% was achieved over the range of 1.95-2.24 µm, suggesting that almost all of the seed pulse energy has been shifted to the Raman solitons. The efficiency was comparable to previously reported values in the Tm-based fiber SSFS systems with dispersion compensation [29–31]. The soliton conversion efficiency dropped rapidly once the soliton wavelength was shifted beyond 2.24 µm, due to the decreased gain in the Tm-doped fiber and the increased background loss of silica fiber. In addition, Raman solitons with a center wavelength beyond 2.2 µm would not be amplified, due to the limitation of the Tm³⁺ emission spectrum [47]. At a pump power of 1.94 W, the monochromatic Raman soliton efficiency reached 93.1%. For the Raman soliton system with strong Kelly sidebands, the corresponding efficiency was calculated to be 66.5%, suggesting that approximately 26.6% of the average power was contained in the amplified Kelly sidebands. The residual energy is mainly from the unconverted seed pulse and ASE light. Consequently, it can be concluded that using the sideband-suppressed seed pulses allows the generation of the Raman solitons with wider wavelength-tuning range and higher efficiency in an amplifier.

 

Fig. 4. (a) The shift wavelengths of Raman solittons. (b) The conversion efficiency to a Raman soliton at different wavelengths.

Download Full Size | PDF

Figure 5(a) shows the measured pulse duration and spectral bandwidth of the Raman solitons in the SSFS-based system seeded by the sideband-suppressed solitons. The pulse duration varied between 319 fs and 490 fs, and the corresponding spectral FWHM varied between 10 nm and 21.5 nm. The evolution of the pulse durations and the spectral FWHMs for the Raman solitons at different wavelengths can be divided into two main stages: (1) the pulse broadening induced by the larger dispersion at the longer wavelengths and the slight spectrum broadening caused by the self-phase modulation (SPM) effect; (2) the pulse narrowing and spectrum broadening due to the enhanced SPM effect. The calculated TBP was less than 0.38 in the whole spectral range of 1.95-2.34 µm. The maximum output power of the first-order Raman soliton was measured to be 91.5 mW at a wavelength of 2.27 µm. Figure 5(b) shows its autocorrelation trace with the FWHM of 717 fs when a sech² profile was assumed to fit. The corresponding pulse duration was around 465 fs. The pulse energy and peak power were calculated to be 4.8 nJ and 10.3 kW, respectively. It is possible to realize the Raman soliton source with better performance, by further optimizing our homemade seed laser and the amplifier. Besides, it is also expected to overcome the SSFS saturation effect and obtain a widely efficient mid-infrared Raman soliton source, by cascading our laser system with some excellent nonlinear fibers and special rare earth ion doped fiber amplifiers.

 

Fig. 5. (a) The pulse duration and the spectral bandwidth of Raman solitons at different wavlengths. (b) The autocorrelation trace of Raman soliton at the wavelength of 2.27 µm.

Download Full Size | PDF

4. Conclusion

In summary, we built and investigated a mid-IR tunable Tm-based fiber Raman soliton system using a sideband-suppressed conventional soliton oscillator as seed source. By efficiently suppressing the soliton Kelly sidebands, Raman solitons with wider wavelength-tuning range of 1.95-2.34 µm were successfully achieved in the TDFA. The conversion efficiency of Raman solitons was more than 93% over the tuned wavelength range of 1.95-2.24 µm. These results are comparable to the previously reported values using the tunable Tm-based fiber laser sources with dispersion management. In addition, we have shown that in the same laser system with sideband-unsuppressed soliton seed, the amplified Kelly sidebands caused low efficiency conversion and narrow wavelength-tuning range of the Raman soliton. Our results demonstrate that using the sideband-suppressed conventional solitons as a good seed for a SSFS-based Tm fiber system leads to improved performance.