1. Introduction

Mid-infrared sources are currently experiencing intense development due to their importance in many scientific applications, e.g. remote sensing, spectroscopy and medicine. The 1.8 – 2.0 μm spectral range, covered by Thulium-doped fiber lasers, overlaps with many absorption lines of several molecules, e.g. carbon dioxide (CO₂) or hydrogen bromide (HBr) [1], which creates the possibility of constructing cost-effective trace-gas sensing platforms. Strong water absorption in this range makes Tm-doped fiber lasers extremely desirable in biomedical applications. It has been shown, that 2 μm laser outperform 1µm and 1.55 μm sources in dermatology and surgery, serving as precise and efficient optical scalpels [2,3]. Application of mode-locking techniques for Tm-doped lasers might potentially create new possibilities and extend the range of applications of those sources.

Tm-doped fiber lasers demonstrated up to date were usually mode-locked using nonlinear polarization rotation (NPR) [4], semiconductor saturable absorbers (SESAMs) [5,6] and carbon nanotubes (CNTs) [7]. All those approaches are well-known and established over the years. Nevertheless, NPR lasers tend to be environmentally unstable and do not provide self-starting pulsed operation. SESAMs suffer from relatively narrowband operation. CNTs should have precisely selected diameters, in order to provide absorption at the desired wavelength range [8]. All of those drawbacks forced the laser community to search for new saturalbe absorber materials. Recently graphene is being considered as a promising SA for various types of lasers. Due to its ultrafast recovery time and ultra-broadband operation, graphene might also be used for lasers operating in the mid-IR spectral region. Graphene suitable for mode-locking of fiber lasers can be produced using several methods. The most common include: epitaxial growth via chemical vapor deposition (CVD), liquid phase exfoliation (LPE) and mechanical exfoliation. The CVD technique allows to obtain high-quality graphene with precisely controlled number of layers, starting from monolayer (grown on SiC [9] or Cu substrates [10]) up to multilayer (on Ni foils [11,12]). Graphene might be afterwards transferred onto mirrors [13], glass windows [14] or fiber connectors [15–17] and serve as a mode-locker for fiber lasers. LPE (chemical functionalization) offers a variety of solvents, in which the graphene flakes may be dispersed, e.g. dimethyloformamide (DMF) [18]. Such solutions can be afterwards mixed with polymers, forming stable and thin composites which also may be deposited on fiber connectors [19–21]. Although, the solutions always contain graphene flakes with different thickness. It means that the obtained layer is not always uniform. Micromechanical cleavage with the use of Scotch-tape is historically the oldest and obviously the easiest method of obtaining graphene. Also in this method the number of graphene layers is undetermined. Nonetheless, many setups utilizing mechanically exfoliated graphene have been presented recently [22–24].

Up to date, only two Tm-doped all-fiber lasers mode-locked with graphene were demonstrated in the literature. Both utilize graphene flakes obtained via LPE. In [25] the authors use a graphene-PVA composite, which allowed to achieve 3.6 ps pulses. The laser presented in [26] delivered 2.1 ps pulses and used graphene/DMF dispersion. In our work we demonstrate for the first time, to our knowledge, an all-fiber Tm-doped laser mode-locked with CVD-grown graphene layers, which are transferred onto the fiber connector with the use of poly(methyl methacrylate) (PMMA) support. Usually, after the transfer process the polymer layer is removed from the substrate [11]. Such removal is difficult and may cause damage of the graphene layer. In our experiments we have found, that PMMA exhibits nearly 100% transmission in the 1.8 – 1.9 μm spectral region. Thus, a CVD-graphene/PMMA composite can be successfully used as a SA for an all-fiber, Tm-doped fiber laser.

2. Graphene SA preparation and characterization

The graphene films were synthesized by chemical vapor deposition on the surface of 12 μm thick copper foils. To obtain graphene of top quality we use the Aixtron VP508 horizontal CVD hot wall reactor. At first, the samples were pretreated at 1000°C under an Ar gas flow and then H₂ gas flow at the pressure of 100 mbar. Afterwards, both C₃H₈ and H₂ gas were introduced into the reactor for 2 minutes. Finally, the copper substrates covering the graphene films were cooled down to room temperature in Ar atmosphere. To detach graphene from the copper foil, graphene was covered with a thin layer of PMMA by spin-coating method. Then we have removed graphene from the backside of copper foil to avoid impurities between the top and lateral graphene films formed during copper etching. The PMMA/graphene stack was subsequently introduced into an aqueous solution of ammonium persulfate. After etching copper away, the PMMA/graphene stack was cleaned with deionized water. Next, the PMMA/graphene stack was fished by the substrate to which graphene does not stick and finally dried in an air atmosphere. Figure 2(a) shows the Raman spectra of graphene on PMMA measured using a 532 nm laser beam. The spectrum contains pronounced G (1592 cm⁻¹) and 2D (2685 cm⁻¹) bands, characteristic for sp² hybridization of carbon. This confirms the presence of a graphene structure in the measured sample. A photograph of the fabricated free-standing PMMA/graphene membrane is shown inset Fig. 1(a). A small piece (about 1x1 mm) of such composite is deposited onto the tip of the angled fiber connector (FC/APC) connector and connected with another one. The fabricated graphene/PMMA SA has a flat absorption spectrum over the range of 1500 – 2000 nm. The measured transmission through the graphene/PMMA layer deposited on the connector is plotted in Fig. 1(b). The average transmission of the composite is at the level of 94-95%.

 

Fig. 1 Raman spectrum of the PMMA/graphene composite with photograph of an exemplary fabricated free-standing membrane (a). Transmission measurement of the PMMA/graphene composite (b).

Download Full Size | PDF

3. Experimental setup

The schematic of the all-fiber ring laser is shown in Fig. 2. The cavity consists of a 2 m long Tm-doped fiber (Nufern SM-TSF-9/125), a 1570/2000 nm wavelength division multiplexer (WDM), 20% output coupler (OC), a fiber isolator, a three-paddle fiber-based polarization controller (PC) and the graphene-based saturable absorber. The cavity contains only single-mode fibers with anomalous dispersion at the operating wavelength. The laser is pumped by a 1570 nm laser source, which is based on a laser diode amplified by an erbium-ytterbium doped-fiber amplifier (EYDFA).

 

Fig. 2 Experimental setup of the Tm-doped fiber laser.

Download Full Size | PDF

4. Experimental results

The performance of the laser was observed using an optical spectrum analyzer (Yokogawa AQ6375), 12 GHz digital oscilloscope (Hameg DSO9254A), 7 GHz RF spectrum analyzer (Agilent EXA N9010A) coupled with a 12 GHz photodetector (Discovery Semiconductors DSC2-50S), and an interferometric autocorrelator.

Stable mode-locking of the laser was observed at pump powers from the range of 140 – 170 mW. In order to initiate the mode-locking, slight adjustment of the PC is required. All measurements were done at 150 mW of pumping and the laser output power was 1.35 mW. We estimate the intra-cavity fluence on the saturable absorber to be 230 µJ/cm². The measurements performed by I. Baek et al. and W.B. Cho et al. show, that the saturation fluence of single-layer CVD-graphene is at the level of 14 µJ/cm² and 66 µJ/cm², measured at 1250 nm and 800 nm wavelengths respectively [27,28]. This suggests, that the graphene in our experiment is well saturated. The modulation depth of monolayer CVD-graphene is estimated to be lower than 0.4% at wavelengths above 2 µm [29]. In our case the transmission measurement indicates a bi-layer graphene structure. Hence, we may assume the modulation depth to be twice that for monolayer graphene. In order to confirm that the mode-locking originates from the graphene SA, the laser was firstly tested with clean connectors (without deposited graphene). No mode-locking was observed in such configuration at any position of the PC. The optical spectrum of the mode-locked emission measured with 0.05 nm resolution is depicted in Fig. 3(a). The graph inset Fig. 3(a) shows the spectrum recorded with 50 nm span (black line) and output spectrum of the laser operating in the CW regime (blue line). The mode-locked spectrum is centered at 1884 nm and has a typical soliton-like shape, which results from the all-anomalous cavity design. The full width at half maximum (FWHM) bandwidth of the emission is 4 nm. Additionally, characteristic dips can be seen in the high-resolution spectrum. We have found that they result from the water absorption lines in air, which are densely located in the 1.9 μm region. In order to confirm it, we have simulated the absorption of light over 1 m path in air with 1% water content using HITRAN database [28]. The simulated absorption lines are plotted (with red line) in Fig. 3(a). It can be seen, that the location and amplitude of the absorption peaks on the spectrum ideally match the simulated water lines.

 

Fig. 3 Measured output spectrum of the laser together with the water absorption lines taken from HITRAN database (a) and the 1.2 ps pulse autocorrelation trace (b).

Download Full Size | PDF

Figure 2(b) shows the autocorrelation of the laser output pulse together with a sech² fitting. The pulse duration after deconvolution is 1.2 ps. The sensitivity of our autocorrelator is low at 1.9 μm wavelength, which causes the relatively large noise level. Nevertheless the signal power was enough to perform a reliable measurement. With 4 nm bandwidth (337 GHz), the time-bandwidth product (TBP) is equal to 0.42. It means, that the pulses have an about 25% longer pulse duration than resulting from the transform limit (TBP equal to 0.315).

The radio frequency (RF) spectrum measured with 2 MHz frequency span and 1 kHz resolution bandwidth (RBW) is depicted in Fig. 4(a). The repetition rate of the laser was 20.5 MHz, which corresponds to an approx. 10 m long cavity. A graph inset Fig. 4(a) illustrates the RF spectrum recorded in the 5 GHz frequency span, showing a broad spectrum of harmonics. The oscilloscope trace is depicted in Fig. 4(b). The pulses are equally spaced by approx. 49 ns, corresponding to 20.5 MHz repetition frequency. No signs of dual-pulsing were observed. The graph inset Fig. 4(b) illustrates the pulse train in 250 ns wide span.

 

Fig. 4 Measured RF spectrum (a) and the oscilloscope trace (b) of the laser output.

Download Full Size | PDF

5. Summary

Summarizing, we have demonstrated a Tm-doped fiber laser operating in the mid-infrared, mode-locked with the use of a CVD-graphene/PMMA composite saturable absorber. The laser was capable of delivering 1.2 ps pulses at 20.5 MHz repetition rate, centered at 1884 nm wavelength with 4 nm FWHM bandwidth. Thanks to the flexibility of the developed graphene/PMMA composite, it can be easily transferred onto fiber connectors, which allows to build fully fiberized, environmentally stable lasers, without the use of bulk free-space components. The presented laser is the first, up to date, Tm-doped all-fiber laser mode-locked with epitaxially grown graphene.