1. Introduction

Graphene, a two-dimensional single-layer carbon sheet arranged in a hexagonal lattice, turned out to be one of the most efficient novel saturable absorber (SA) materials for pulsed laser operation which facilitates Q-switching or mode-locking. Graphene has been successfully used in many different forms for bulk, fiber or waveguide lasers [1–5]. In addition to other SAs such as semiconductor saturable absorber mirrors (SESAMs) and ion-doped crystals, graphene exhibits superior optical characteristics including large third-order nonlinearity, ultrabroad absorption, ultrafast response to the photo-excitation and high optical damage thresholds. The relatively simple fabrication process provides more flexibility and allows for integration into diverse laser geometries and systems [6–9].

Pulsed lasers with compact cavity design are desirable light sources for a number of applications in integrated optical devices including e.g. microscopy/spectroscopy, sensing and frequency comb generation [10–12]. Planar and channel waveguide structures are ideal candidates for a further miniaturization of pulsed laser systems [3–5, 13–20]. Their monolithic designs provide very rugged cavities and the light confinement within the small waveguide diameter allows for saturation of passive SAs with low pulse energies, characteristic for high repetition rate pulsed laser sources, even at low average powers. Furthermore, due to the confinement and a good overlap between pump and laser mode, high gain can be achieved with moderate pump powers. Compared to planar waveguide structures, channel waveguides can deliver a well-defined circular fundamental transverse-mode beam and are thus even more useful in this aspect. In recent years, the femtosecond (fs)-laser inscription method turned out to be well suited for the fabrication of channel waveguides within various dielectric gain media [21–26]. Beside several other configurations the inscription of two parallel tracks is well established for the fabrication of channel waveguides in various crystalline gain media. A stress induced refractive index increase allows for waveguiding in the region between the tracks. Pulsed laser operation of fs-laser-inscribed channel waveguides by using graphene SAs has been successfully demonstrated in several gain media such as Yb-doped Bismuthate glass [17], Nd:YAG [5, 18, 19] and Tm:YAG [20]. In those reports, graphene grown by chemical vapor deposition (CVD) or liquid phase exfoliation (LPE) was transferred onto OCs or onto the top surface of the waveguides themselves. Since graphene grown by CVD allows for an easy transfer onto any substrate in a much simpler way than exfoliated or solution-processed graphene [27], CVD-grown graphene has been widely used as SA in different geometries for developing compact pulsed lasers [28].

Here we report for the first time on a fs-laser-inscribed Yb:YAG channel waveguide laser Q-switched by employing monolayer graphene. In a compact configuration without external OC, the monolayer graphene is directly placed on the one end facet of the channel waveguide acting simultaneously as the OC. For comparison, a standard configuration with external OC was investigated. In this case CVD-grown monolayer graphene is transferred onto different OCs of 10%, 20% and 30% transmission. In the following we will refer to these two different cavity configurations as Setup I for the waveguide directly coated with graphene and Setup II for the setup with external OCs. While we observed a maximum slope efficiency of 33% with the external OC of 20% transmission in Setup II, an even higher slope efficiency of 46% was achieved in the monolithic Setup I.

2. Fabrication of monolayer graphene, saturable absorbers, and waveguides

By using thermal CVD of the mixture of methane and hydrogen gases under high vacuum at a temperature of about 1000°C, high-quality monolayer graphene is synthesized on copper foil and subsequently poly(methyl methacrylate) (PMMA) of adequate thickness is spin-coated on the grown graphene as supporting layer. The copper foil substrate is etched away and after rinsing in water bath the PMMA/graphene sheet is ready for transfer onto the desired substrates. The PMMA/graphene sheet is transferred either onto OCs which are, prior to transfer, coated with an additional PMMA buffer layer or directly onto the end facet of the waveguide. The samples are finally baked at 100°C on a hot plate.

Figure 1(a) illustrates the two different approaches in which the graphene SAs were integrated into the cavity. Note that the PMMA on graphene in Setup I is required as a buffer layer to avoid the SA being placed in the node of the oscillating laser mode and enables an efficient interaction of the oscillating laser mode with graphene. In Setup II the PMMA layer between graphene and OC also acts as a buffer layer to allow for a sufficiently high mode field intensity of the laser beam at the end of the resonant cavity, whereas the upper PMMA layer serves as a supporting layer for the transfer process. Figure 1(c) shows the Raman spectrum of graphene, measured with a Renishaw inVia micro-Raman spectrometer, to verify the quality and the number of graphene layers used in the present experiments. The average intensity ratio of I_2D / I_G is approximately 4.3 and a symmetric 2D peak is centered at ~2683 cm⁻¹ with a full width at half maximum of ~26 cm⁻¹. The measurement was repeated at several positions on the sample and nearly identical results proved the homogeneity of the used graphene layer. The inset of Fig. 1(c) shows that the 2D peak is well-fitted by a single Lorentzian curve. These results indicate that our synthesized graphene (dimension: >1 × 1 cm²) is uniform and high-quality monolayered across the large area [29]. Moreover, the nonlinear optical characteristics of the used monolayer graphene including nonlinear transmission (Fig. 1(d)) and nonlinear response (inset of Fig. 1(d)) were investigated near 1.09 μm by employing a synchronously-pumped optical parametric oscillator (OPO) with an output pulse duration below 200 fs. The monolayer graphene transferred onto a quartz substrate delivers a saturation fluence of 35.7 μJ/cm², a modulation depth of 0.74%, and nonsaturable losses of 1.56%. From the time-resolved pump-probe measurement, the recorded pump-probe trace is well fitted by a bi-exponential response of the saturable absorption with a fast component of 230 fs and a slow decay of 1.64 ps.

 

Fig. 1 (a) Schematic illustration of monolayer graphene SAs for Setup I (left) and Setup II (right). (b) Cross-sectional images of the fs-laser-inscribed two tracks separated by 16 μm serving as the channel waveguide. (c) Raman spectrum measured for the synthesized monolayer graphene, showing the 2D peak well-matched by a single Lorentzian curve (inset). (d) Measured nonlinear transmission and pump-probe traces (inset) near 1.09 μm in monolayer graphene with fit curves (red lines).

Download Full Size | PDF

A YAG crystal with 7% Yb-doping with respect to the yttrium sites is chosen as gain medium to demonstrate pulsed operation in a channel waveguide. The channel waveguide is fabricated by fs-laser inscription employing 140-fs pulses at 775 nm from a 1-kHz Ti:sapphire regenerative amplifier system (Clark-MXR CPA-2010). By moving the sample on a motorized translation stage in the vertical direction to the incident beam, tracks were inscribed into the crystal. The pulse energy was varied between 1.3 and 1.8 μJ depending on the depth of the tracks below the crystal surface (19 - 390 μm). The refractive index of the tracks is decreased in the order of 10⁻³ in comparison to unmodified YAG. Additionally, a stress induced refractive index increase in the surrounding of the tracks allows for waveguiding between two tracks. In our experiments we utilized a waveguide with a distance of 16 μm between the tracks, as shown in Fig. 1(b). A detailed description of the fabrication process of channel waveguides can be found elsewhere (e.g. Type II, scheme B in [24]).

3. Experimental results

Figure 2 shows the experimental setup of the monolayer graphene Q-switched Yb:YAG channel waveguide lasers in the two different configurations. A continuous-wave (CW) single-mode laser diode (9-mm TO Can LD, Axcel photonics Inc.) operating at 940 nm is used as pump source. The collimated pump beam passes through a Faraday isolator which suppresses back-reflections into the laser diode, and subsequently a half-wave plate and a Glen-Taylor polarizer for attenuation of the p-polarized pump beam. An aspheric lens (L1) of f = 18.4 mm allows for efficient coupling into the channel waveguide. A dichroic mirror with high transmission at the pump wavelength and high reflection at the laser wavelength serves as the incoupling mirror. It is closely attached to the input facet of the 9.3-mm-long Yb:YAG channel waveguide where the maximum available incident pump power is measured to be 285 mW. Note that, to minimize Fresnel losses, appropriate index matching gel (Thorlabs G608N3) is used for attaching the mirrors to the waveguide facets effectively.

 

Fig. 2 Experimental setup of Yb:YAG waveguide lasers Q-switched by monolayer graphene on an end facet of the waveguide without external output coupler in Setup I (upper) and on an output coupler in Setup II (lower). L1: aspheric lens with f = 18.4 mm, IC: incoupling mirror, OC: output coupler with T = 10%, 20% or 30%, L2: convex lens with f = 11 mm.

Download Full Size | PDF

Prior to Q-switching experiments we investigated the CW laser performance of the Yb:YAG channel waveguide in three different laser configurations without graphene: The bare waveguide itself as a monolithic cavity, the waveguide with attached incoupling mirror (IC) (Setup I without graphene) and the waveguide with IC and 20% OC (Setup II without graphene) (see Fig. 2). Due to the Fresnel reflection at both plan-parallel end facets, the channel waveguide itself works as a Fabry-Perot oscillator and hence laser oscillation is also possible without any mirrors. In this configuration the laser exhibits a very high output coupling rate of about 99% [21]. Although the laser is emitting roughly the same output power through the end facets in both directions, only the laser beam propagating along with the pump in the forward direction is measured in the present work. Thus, the output powers for the bare waveguide shown in Fig. 3(a) have to be multiplied by roughly a factor of two to obtain the total output power. In this case the slope efficiency was 25% (≙50% for both directions). Utilizing only an incoupling mirror (Setup I), the slope efficiency amounted to 47%. By adding the 20% OC (Setup II) the slope efficiency decreased to 32% due to a decreased resonator extraction efficiency in this case. A similar CW lasing behavior was observed for the OCs with T = 10% and 30%. The laser thresholds were reached at incident pump powers of 135 mW, 85 mW and 46 mW in these three setups, respectively (Fig. 3(a)). The corresponding maximum average output powers were measured to be 38 mW (≙76 mW for both directions), 91 mW and 73 mW, respectively.

 

Fig. 3 Laser operation of Yb:YAG channel waveguide lasers in Setup I and Setup II: (a) CW output power, (b) Q-switched average output power (upper) and corresponding pulse energy characteristics (lower).

Download Full Size | PDF

In the Q-switched regime a maximum average output power of 85 mW with a corresponding maximum pulse energy of 64 nJ and a slope efficiency of 46% were achieved with Setup I (Fig. 3(b)). As previously observed in the CW experiments, also in the Q-switching experiments the slope efficiency reduces to 33% in Setup II utilizing the OC with T = 20%. At comparable laser threshold pump powers of Setup I and II, a lower maximum average output power of 65 mW and a pulse energy of 49 nJ are achieved in the latter case. Compared to our previous results using carbon nanotubes as SA in a configuration similar to Setup II [14], the data presented here – utilizing a waveguide with very similar parameters and CW laser performance – are superior in terms of output power, pulse energy and slope efficiency. In particular, the newly proposed scheme (Setup I) where the graphene SA is directly deposited onto the end facet of the waveguide without OC makes the laser cavity much more compact and allows for a significantly improved performance.

The slope efficiencies obtained in the Q-switching experiments are very similar to those obtained in CW operation. The slightly lower maximum average output power is mainly caused by the higher threshold due to the saturable and nonsaturable losses of the graphene layer and the losses in the PMMA. In Fig. 4(a), repetition rates and pulse durations of Q-switched pulses vs. incident pump power are shown for both setups. In Setup I, the repetition rate increases from 0.60 MHz to 1.33 MHz with increasing pump power, while the pulse duration decreases from 520 ns to 96 ns. On the other hand, Setup II operates at repetition rates from 0.90 MHz to 1.33 MHz at pulse durations from 250 ns to 79 ns during increasing the pump power. Repetition rate and pulse duration strongly depend on the effective gain cross section and the intra-cavity intensity, which both increase with pump power in waveguide geometry. Such behavior is typically observed in lasers passively Q-switched by fast SAs [30, 31]. Figure 4(b) shows the pulse traces measured at maximum average output powers in both setups.

 

Fig. 4 Performance of monolayer graphene Q-switched Yb:YAG channel waveguide lasers in Setup I and Setup II: (a) Repetition rate (black dots) and pulse duration (blue squares) characteristics vs. incident pump power, and (b) pulse traces measured at maximum average output powers.

Download Full Size | PDF

Additionally, the Q-switching characteristics for OCs of T = 10% and 30% in Setup II were investigated. Longer pulse durations of 96 ns and 132 ns as well as lower average output powers of 30 mW and 54 mW and slope efficiencies of 16% and 28% were obtained with T = 10% and 30%, respectively. Although a higher slope efficiency would be expected for higher output coupling rate from the CW results, the efficiency drops at T = 30% compared to the case of T = 20%. Since such behavior is generally not expected, we currently do not have a suitable explanation. However, alignment issues or slightly different qualities of the coated graphene layers might be the reason. The pulse durations generated with Setup II are slightly shorter than those achieved in Setup I. Nevertheless, the overall better Q-switching characteristics regarding slope efficiency, output power and pulse energy as well as the compactness favor Setup I in further experiments.

Figures 5(a) and 5(b) show the laser spectra and the pulse train in the Q-switched regime. The lasers operate near 1029 nm in both cases with spectral bandwidths of about 0.5 nm and 0.7 nm for Setup I and II, respectively (Fig. 5(a)). The output spectra were measured by an optical spectrum analyzer (Anritsu MS9710A) providing a spectral resolution of 0.05 nm. After the collimation by a convex lens L2 (f = 11 mm) the spatial beam profiles (right insets) were recorded by a beam profiler (LaserCam-HR, Coherent Inc.). The nearly circular shapes indicate fundamental transversal mode operation. The near field beam diameter at the waveguide facet was calculated to be about 14 µm. As shown in Fig. 5(b) the Q-switched pulses are fully modulated and exhibit a good pulse-to-pulse stability of ≤ 10%.

 

Fig. 5 (a) Laser spectra and beam profiles of monolayer graphene Q-switched Yb:YAG channel waveguide lasers in Setup I and Setup II measured at maximum average output powers and (b) corresponding pulse trains.

Download Full Size | PDF

4. Conclusion

We demonstrate diode-pumped monolayer graphene Q-switched Yb:YAG channel waveguide lasers with and without OC. For cavity configurations with monolayer graphene coated OCs (Setup II), we obtained a maximum average output power of 65 mW and a pulse energy of 49 nJ at a repetition rate of 1.33 MHz. At the highest pulse energy, the pulse duration was as short as 79 ns. For a more compact configuration which was demonstrated for the first time in the present work, the monolayer graphene was deposited directly onto the end facet of the waveguide which simultaneously serves as the output coupling mirror (Setup I). In this case, the average output power increased to 85 mW under 285 mW of incident pump power. At a pulse energy of 64 nJ, a somewhat longer pulse duration of 96 ns was observed.

By direct coating of the incoupling mirror onto the waveguide, our new approach allows for a monolithic setup and hence enables ultra-compact and rugged Q-switched lasers. Such lasers can be further applied for diverse applications and also pave the way towards laser mode-locking in fs-laser-inscribed waveguide structures at gigahertz repetition rates.