1. Introduction

Broadband laser sources operating in the “eye-safe” 1.9 – 2.0 μm region have attracted great attention in the recent years, due to their potential applications in various fields, e.g. medicine, sensing, spectroscopy, etc [1–4]. The 2 μm wavelength is particularly desirable in environmental sensing, especially in trace-gas detection systems. The spectral range covered by Thulium-doped fiber lasers (TDFLs) overlaps with absorption lines of two harmful greenhouse gases, namely carbon dioxide (CO₂), nitrous oxide (N₂O) [3]. It is evident, that conducting research on compact, fully fiberized and broadband TDFLs (i.e. mode-locked all-fiber lasers generating dissipative solitons) might strongly contribute to the development of ultra-sensitive, robust and portable trace-gas detection systems

A significant number of the currently developed mode-locked TDFLs operates in the soliton regime, which is a consequence of the anomalous chromatic dispersion of standard single-mode fibers (SMFs) at 1.9 μm wavelength. In the literature, one can find a number of reports on anomalous dispersion lasers [5–12], mode-locked with different techniques: using semiconductor saturable absorber mirrors (SESAMs) [5], graphene [6–10], nonlinear optical/amplifying loop mirrors (NOLM/NALM) [11], or NALM in combination with carbon nanotubes (CNTs) [12]. In order to achieve different mode-locking regimes and more broadband operation of a fiber laser, it is required to compensate the anomalous dispersion of SMFs. This can be done either by introducing bulk compressors into the cavity [13–15], by using dispersion compensating fibers (DCFs) [16–21], or chirped Bragg gratings [22]. The use of bulk compressors may lead to impressive results (e.g. 173 fs achieved by F. Haxsen et al. [15]), but Martinez-type compressors require careful adjustment and introduce large losses. That is why the setups reported in [13–15] required very strong multi-watt pumping sources. Broad dissipative soliton generation with the use of DCFs was reported in [18], but the laser utilized a free-space optics set responsible for the nonlinear polarization evolution (NPE) mode-locking (a Lyot filter together with waveplates and polarizing cubes). It also required usage of 2 types of DCFs (ultra high numerical aperture fibers, UHNA) and a grating compressor to compensate the chirp outside the cavity. Also Li et al. [19] demonstrated sub-100 fs pulse generation at 2 μm from a dispersion-managed laser. However, the setup was also based on free-space optics, which requires adjustment. Fully fiberized dissipative soliton lasers were reported in the literature, but their performance differs significantly from the lasers presented in [18,19]. As an example, Q. Wang et al. demonstrated dissipative soliton generation from an all-fiber TDFL, but the bandwidth was only around 4 nm [21]. Another fully fiberized oscillator was presented by C. Huang et al., but also there the bandwidth was only 15 nm [17]. Another approach to generate ultra-broad spectra from fiber lasers is to force the oscillator to operate in so called noise-like pulse (NLP) regime. In this regime, the pulse circulating in the cavity is composed of a bunch of ultrashort pulses with varying widths and peak intensities. The autocorrelation trace of such pulse is characterized by a femtosecond peak located on the top of a broad picosecond pedestal [23], while the spectrum is very broad, comparable to or even broader than the gain bandwidth. Er-doped NLP lasers achieve bandwidths broader than 100 nm [23]. Many applications might benefit from such laser behavior, e.g. interferometry [24] or supercontinuum generation [25]. NLP lasers operating in the 1.9 μm region were reported, however, the achieved bandwidths were at the level of tens of nanometers [11,26,27].

Here, we demonstrate a fully fiberized Tm-doped laser operating in the normal dispersion regime, capable of generating broadband dissipative soliton spectra (up to 100 nm) and ultra-broadband noise-like pulses (up to 300 nm bandwidth). The mode-locking is achieved via nonlinear polarization evolution mechanism. These are, to the best of our knowledge, the broadest dissipative solitons and NLPs generated from an all-fiber Tm-doped laser so far.

2. Experimental setup

The setup of the all-fiber oscillator is depicted in Fig. 1. The resonator consists of the following components: a hybrid output coupler/isolator/wavelength division multiplexer (OC/ISO/WDM) device, a polarization controller (PC), a polarization beam splitter (PBS), a 17 cm piece of Tm-doped fiber (Nufern TSF-5/125, TDF), and a segment of dispersion compensating fiber (DCF) with group velocity dispersion at the level of 0.0246 ps²/m at 1950 nm. The dispersion was measured using white light interferometry, described in [28].

 

Fig. 1 Experimental setup of the all-fiber mode-locked laser.

Download Full Size | PDF

Since the integrated hybrid device comprises three components in one package, it significantly reduces the complexity of the oscillator and makes it more simple, robust and cost-effective. The hybrid component also ensures unidirectional light propagation in the cavity (marked clockwise on the schematic in Fig. 1) and reflects 10% of the intra-cavity power to the output fiber. The used DCF was easily spliceable to standard SMFs and allowed to manage the cavity dispersion without the necessity of using free-space optics. The length of the DCF in the experiments varied between ~7-8 meters, in order to achieve normal net dispersion at 1950 nm. The oscillator was pumped at 1566 nm wavelength using a telecom laser diode, amplified in a self-made Erbium/Ytterbium-doped fiber amplifier (EYDFA). The mode-locking is provided via NPE mechanism, based on the PBS in combination with a polarization controller. In contrast to the previous reports on Tm-doped lasers, the setup is fully fiberized and does not require any free-space optics to either compensate the dispersion [13–15] or initiate the NPE mode-locking [18,19]. The setup is therefore less complicated, more robust and does not require any alignment of free space beams.

3. Experimental results

All measurements were performed using the following equipment: an optical spectrum analyzer (Yokogawa AQ6375), radio frequency (RF) spectrum analyzer with 3.6 GHz bandwidth (Keysight EXA N9010A) coupled with a 16 GHz photodiode (Discovery Semiconductors DSC2-50S), and an autocorrelator (Femtochrome FR-103XL).

Dissipative soliton mode-locking was achieved with the DCF length of 7.59 m. In this case, the estimated net GDD of the cavity is normal, and is at the level of 0.014 ps². The total length of the resonator was 10.03 m, resulting in repetition frequency of 20.75 MHz. The performance of the laser in this configuration is summarized in Fig. 2. All the parameters were recorded at pump power of 700 mW. The laser generated a very broad and smooth spectrum, with characteristic rectangular shape, typical for dissipative solitons in normal dispersion regime [29]. The spectra measured at both outputs (PBS and OC) are depicted in Fig. 2(a). The average output power measured at the PBS rejection port was 6.5 mW, while the power from the OC was lower (1.3 mW). It means, that the single pulse energy from the PBS is at the level of 0.32 nJ, several times larger than usually achieved from anomalous dispersion lasers [8–10]. Because of the net normal dispersion, the pulses are negatively chirped directly outside the cavity and might be recompressed using a standard SMF. The autocorrelation trace of the output pulse, after passing a 3.5-m long segment of compressing SMF is depicted in Fig. 2(b). Assuming a Gaussian shape, the pulse duration after deconvolution is 198 fs. The autocorrelation is characterized by a small pedestal, containing less than 22% of the total pulse energy (the calculation of the pedestal energy content was estimated by integrating the area under the measured and fitting curve), and is free of any pre- or post-pulses (which is confirmed by a wide-scan measurement shown inset Fig. 2(b)).

 

Fig. 2 Performance of the laser at 0.014 ps² estimated net dispersion: optical spectra (a), 198 fs autocorrelation trace (b), narrow span (c) and wide span RF spectrum (d).

Download Full Size | PDF

The achieved pulse duration is longer than reported e.g. by Y. Tang et al. [18], but they have used a grating pair compressor to shorten the pulses outside the cavity. Here, we decided to keep the whole setup fully integrated and fiberized, and the shortest obtained pulse after careful adjustment of the external SMF was 198 fs. We believe that the pulses might be further compressed by the use of a grating pair, prism pair or combination of highly-nonlinear fiber with SMF, analogously to the setup presented by P. Li et al. [19]. Figure 2(c) shows the RF spectrum recorded with 33 Hz resolution bandwidth (RBW), centered at the carrier frequency (20.75 MHz) with 4.5 MHz span. The spectrum recorded in full available span (3.6 GHz) is depicted in Fig. 2(d). It does not contain any modulations, which confirms stable, single-pulse operation (any multi-pulsing or harmonic mode-locking is easily observable in the RF spectrum, see e.g [30].).

By lengthening the DCF from 7.59 to 7.87 m and slightly increasing the pump power (to 900 mW), we were able to achieve even broader optical spectra at the laser output. In this case, the estimated net GDD of the cavity was 0.021 ps². The achieved experimental results are summarized in Fig. 3. The bandwidth of the generated dissipative solitons from both output ports exceeded 100 nm (see Fig. 3(a)). Unfortunately, with such broad bandwidth simple pulse compression in SMF is challenging, mostly because of the steep dispersion slope of SMF. However, after careful optimization of the external SMF length, we were able to compress the pulse down to 371 fs. The autocorrelation is depicted in Fig. 3(b). The pulse contains a quite large pedestal, most likely originating from uncompressed third order dispersion (TOD). Based on integration of the autocorrelation curve, we estimate that approx. 35% of the total pulse energy is stored in the main peak. We believe, that also in this case usage of a bulk compressor could lead to better performance in terms of pulse duration, similarly to [18]. The repetition rate in this case was equal to 20.2 MHz (see RF spectrum shown in Fig. 3(c)). Again, the RF measurement in the wide span shows a broad spectrum of stable harmonics, indicating stable, single-pulse operation. The average output power from the PBS port was at the level of 11.3 mW, while the power from the OC was again lower (2 mW). The calculated pulse energy from the PBS is at the level of 0.56 nJ.

 

Fig. 3 Performance of the laser at 0.021 ps² estimated net dispersion: optical spectra (a), 371 fs autocorrelation trace (b), narrow span (c) and wide span RF spectrum (d).

Download Full Size | PDF

At the same net dispersion (0.021 ps²) as in the previous case, further increase of the pump power to 1.1 W results in noise-like operation of the laser. The emission bandwidth extremely increases even up to 300 nm considering 10 dB width (optical spectrum is shown in Fig. 4(a), which is few times broader than previously reported in the literature [11, 26,27]. To our knowledge, the broadest spectrum generated from a NLP laser operating in the 1.9 – 2.0 μm band was 60 nm [26]. Figure 4(b) shows the measured autocorrelation trace of the output pulses from the NLP mode-locked laser. Typically for NLP lasers, it contains a narrow spike on a broad pedestal. The width of the pedestal is at the level of 5-6 ps, which is far shorter than in previously reported NLP lasers (e.g. in [11] the pulse was longer than 150 ps). As a consequence of a noisy behavior of the laser, the signal-to-noise ratio (SNR) in the RF spectrum is lower than in case of dissipative soliton mode-locking (approx. 40 dB indicated in Fig. 4(c)). Also the wide RF scan (Fig. 4(d)) indicates a quite strong noise level in the signal. The average output powers from the PBS and OC were 26.3 mW and 4.5 mW, respectively. The pulse energy in the NLP regime cannot be exactly determined, since the noise-like emission is a bunch of pulses with varying amplitudes and width. However, one can estimate the energy of the one NLP bunch by dividing the average output power by the repetition frequency (as shown by other authors [21,26,27]). By doing such calculation, the estimated energy is 1.3 nJ, comparable to the Tm-doped NLP laser reported by Q. Wang et al. [21].

 

Fig. 4 Performance of the laser operating in the noise-like pulse regime at 0.021 ps² estimated net dispersion: optical spectrum (a), autocorrelation trace (b), RF spectra (c) and d)).

Download Full Size | PDF

4. Summary

Summarizing, we have experimentally shown the possibility of generating ultra-broadband dissipative solitons from a simple, fully fiberized mode-locked Tm-doped fiber laser. Thanks to the use of a dispersion compensating fiber, we were able to achieve normal net dispersion of the cavity and force the laser to operate in the dissipative soliton regime, without the need of using free-space optics. Depending on the cavity dispersion, the laser was capable of generating 60 nm broad dissipative solitons with pulse duration of 198 fs, or ultra-broad spectra exceeding 100 nm, with pulse duration at the level of 371 fs. This are the broadest dissipative solitons generated from a mode-locked Tm-doped fiber laser so far. By increasing the pump power, the laser could operate in the noise-like pulse regime, which resulted in significant increase of the emission bandwidth even up to 300 nm (considering 10 dB width). This result also outperforms previous reports on Tm-doped NLP lasers in terms of bandwidth. It is worth mentioning, that the presented laser is fully fiberized, thus it does not require any free-space optics to either compensate the dispersion or initiate the NPE mode-locking.