1. Introduction

Carbon nanotubes (CNTs) have attracted researchers’ attention by their ultrafast recovery time of absorption and high nonlinearity [1]. CNT-based saturable absorbers have been realized in different device structures, such as free-space [2,3], fiber-end [4], waveguide [5], and fiber structures [6–8]. The waveguide- and fiber-type devices were first demonstrated using evanescent coupling between the guided modes and CNT layers through their evanescent fields. The evanescent coupling structure offers high optical damage threshold and high third order nonlinearity [9]. Depositing CNTs around a microfiber is one of the methods which can realize fiber-type polarization insensitive CNT-devices [7,8]. In the previous works, unfortunately, complicated processes were needed and significant number of CNTs was dissipated. The CNT-dissipation and the process complicity are due to the CNT characteristic that they tend to be entangled with each other in common solvents. A simple and efficient deposition process with area selectivity is required for functional device fabrication.

In another aspect, thermal stability of the CNTs is very attractive to realize high power passively mode-locked fiber laser. Sprayed CNTs without any coating have low thermal transfer efficiency because air is filled among the CNTs. Heat due to light absorption elevates the CNTs temperature and damage or burn the CNTs. Therefore, high thermal resistance materials should be used to support CNTs and heat transfer design should be considered to realize high damage threshold saturable absorber. To achieve high damage threshold saturable abosorber, polycarbonate was used as a high thermal resistance material for embedding CNTs [10]. Nanoporous alumina was used as a heat sink to transfer the heat from CNTs [11]. The other approach is to employ the evanescent coupling structure. The evanescent coupling structure device utilizes interaction between evanescent field of guided wave and CNTs located near the guiding structure. Effective interacting power is small, and the damage threshold of the device is high.

Recently, we proposed and demonstrated a method of CNT deposition onto optical fiber ends [12,13]. We deposited CNTs only onto a core region of an optical fiber end by injecting a light beam from the end into a CNT-dispersed dimethylformamid (DMF) solution. Independent on our proposal, another group demonstrated optically driven CNT deposition [14]. In this report, they used ethanol for a solution. Using either of the method, the CNT deposited fibers had sufficient performance as saturable absorbers at 1.55 μm wavelength range. In the report [14], they reported that the technique can be also applied at 1 μm wavelength range and reported passively mode-locked fiber laser at 1.07 μm. Additional to the system, we employed optical reflectormetry into the system, and realized in situ monitoring of CNT layer so that the layer could be more uniform and the deposition system obtained higher repeatability [15]. However, the technique should be applicable not only to the deposition onto the fiber ends, but also along the core regions without cladding or with thin cladding via evanescent light.

In this work, we propose and demonstrate optically induced deposition of CNTs around microfibers. The deposition of CNTs around a microfiber was achieved only by injection light through the microfiber, which was immersed in a CNT-dispersed DMF droplet. The number of deposited CNTs was able to be controlled by adjusting the deposition time. We also demonstrate a passively mode-locked fiber laser using a CNT-deposited microfiber fabricated by the technique.

2. Deposition of carbon nanotubes around microfibers

First, we fabricated a microfiber using an experimental setup shown in Fig. 1 . A bare standard single mode fiber (SMF) was set on two fiber holders, which were fixed on two translation stages. The fiber ends were connected to an erbium doped fiber amplifier (EDFA) and a power meter, respectively. We fabricated a microfiber by stretching the fiber with heat produced by a flame. During the fabrication process, we monitored the insertion loss of the fiber using the power meter. The fiber was tapered down so that its taper waist diameter became ~6 μm.

 

Fig. 1 Experimental setup for fabricating microfibers.

Download Full Size | PDF

Next, we deposited the CNTs around the microfiber using a setup shown in Fig. 2 . The fabricated microfiber was immersed into a CNT-dispersed DMF droplet on a slide glass. Light from a laser diode at a wavelength of 1560 nm and at −10 dBm optical power was amplified up to 13 dBm by an EDFA and consequently injected into the microfiber. The output power was monitored by a power meter to detect start of CNT deposition, and consequently to control the deposition time. We used CNTs fabricated by high pressure carbon mono-oxide (HiPco) method. Absorption spectrum of the CNTs, which was deposited onto a quartz substrate by spray method, is shown in Fig. 3 . The CNTs had three absorption peaks at around 1175, 1325, 1425 nm. The absorption band included the 1.55μm wavelength range and the saturable absorption can be expected at the wavelength for passive mode-locker.

 

Fig. 2 Experimental setup for CNT deposition around microfibers by light injection through the microfibers.

Download Full Size | PDF

 

Fig. 3 Absorption spectrum of the CNTs.

Download Full Size | PDF

Figure 4 (a) shows a microscope image of a CNT-deposited microfiber, whose waist diameter was ~6 μm. We found that the CNTs start to be deposited around the microfiber at the incident power of 13 dBm. To ensure the existence of CNTs, we performed microscopic Raman spectroscopy. The dotted circle in Fig. 4(a) was the area where microscopic Raman spectroscopy was performed. There are three major peaks in the Raman spectra of CNTs, G-band, D-band and radial breathing mode (RBM). RBM, which appears at around 250 cm⁻¹, is the most characteristic peak of CNTs, and is a vibration mode of CNTs in their radial direction. The microscopic Raman spectrum in Fig. 4(b) confirms that the CNTs were certainly deposited around the microfiber. The optical deposition technique was confirmed to be applicable not only to deposition onto fiber ends but also to deposition around microfibers. We could detect the start of the CNT deposition by the drop of the output power due to scattering and absorption induced by the deposited CNTs. We stopped the light injection about 5 sec after the deposition started. The excess loss induced by the tapering was 0.2 dB. The CNT deposition increased insertion loss by 5.8 dB, and the total loss was 6 dB.

 

Fig. 4 (a) Microscope image of CNT-deposited microfiber (waist ~6 μm), dotted circle indicates area where microscopic Raman spectrum was measured, (b) Microscopic Raman spectrum of CNT-deposited microfiber.

Download Full Size | PDF

CNTs were deposited around the fiber in ~30 μm length as shown in Fig. 4(b), and the length was enough to achieve saturable absorption for passive mode-locking. Figure 5 shows saturable absorption characteristics of the CNT deposited microfiber. The experimental setup was almost same as Fig. 2, except the DMF droplet on the slide glass. We changed the output power of the EDFA and measured the transmitted power by the power meter. The insertion loss decreased with increase of the input power, indicating that the absorption had been saturated. The loss decrease due to the saturation was over 1.2% and the value was sufficient for passive mode-locker.

 

Fig. 5 Saturable absorption characteristics of the CNT-deposited microfiber.

Download Full Size | PDF

We also measured polarization dependent loss (PDL) of the microfiber. It had small PDL less than 0.3 dB with input power both at 10 dBm (weaker power than saturable absorption threshold) and 17 dBm (higher power than saturable absorption threshold). These results indicate the CNTs uniformly were deposited around the fiber and canceled the polarization dependence of the each CNTs which had cylinder like structure. When we deposit the CNTs longer and thicker around the fiber, we could achieve higher absorption depth, but the insertion loss of the fiber was increased.

We did not observed any damages of the CNTs or the microfiber even we injected the high power light over 17 dBm. The damage threshold depends on the microfiber waist diameter. When the waist diameter of the microfiber was shorter than ~3 μm, the heat transfer from the CNTs were not enough and CNTs were damaged or the microfiber was broken by the heat caused by the light absorption.

We presume the similar phenomenon occurred both in the case of optical CNT deposition onto fiber ends and around microfibers. In the case of the CNT deposition onto fiber ends, swirl and convection caused by light injection to a CNT-dispersed solution might deposit CNTs (Fig. 6(a) ). Under a microscope, we could see the convection when we started to incident the light into the solution and bundled CNTs were carried closer to the fiber end. On the contrary, in the present experiment, the light injection from air into the droplet thermally caused swirl and convection at the boundary (Fig. 6(b)). This convection could be seen under the microscope. The swirl and convection induced by the light injection might deposit the CNTs around the microfibers. This phenomenon occurs only at the light input side boundary, and enables us to area-selectively deposit them onto desired position.

 

Fig. 6 Schematic of optical deposition of CNTs (a) onto fiber end (b) around microfiber.

Download Full Size | PDF

Another possible mechanism is the optical tweezer effect, which can traps micro- and nano-sized objects by the optical intensity diversion in the solution. For the nano-sized objects, the objects can be treated as a point dipole. In electromagnetic field, Lorentz force is applied onto the object. Under optical power divergence or inhomogeneous electromagnetic field, the Lorentz force is stronger toward the higher optical intensity direction than the lower optical intensity diversion. The force, therefore, tends to trap the object at the region of strongest position. Evanescent field had the optical intensity diversion, and mode field mismatch between the air section and the DMF section caused the optical intensity diversion by scattering. The diversion might trap the CNTs by optical tweezer and deposited onto the microfiber.

However, the force of the optical tweezer might be too weak to trap bundled or entangled CNTs, since the optical power gradient of the evanescent field is not steep. We assume combination of the two mechanism described above deposited the CNTs around the core. The swirl and the convection carried the CNTs to the microfiber, and optical tweezer effect, consequently, trapped part of the CNTs which was closely flowed to the microfiber by the swirl and the convection.

3. Passive mode-locking of fiber laser incorporating CNT-deposited microfiber

In order to confirm the performance of the fabricated device as a passive mode-locker, we applied it to a fiber laser for passive mode-locking. We inserted the fiber, mentioned in the previous section, in a fiber laser cavity to achieve a passive mode-locking. The experimental setup is shown in Fig. 7 . We used a high power EDFA as gain of the laser. Polarization state inside the laser cavity was controlled by a polarization controller (PC). An isolator eliminated back-reflection inside the laser cavity, mainly occurred at the tapering region and the CNT-deposited part. The laser output came out from a 10% port of a 10:90 coupler. We measured optical spectrum and second harmonic generation (SHG) autocorrelation trace of the laser output using an optical spectrum analyzer and a SHG autocorrelator. We measured pulse repetition rate using a photodiode and an oscilloscope.

 

Fig. 7 Schematic of a passively mode-locked fiber laser using a CNT-deposited microfiber

Download Full Size | PDF

We controlled the polarization state inside the laser cavity and gain of the EDFA. With the increase of injection current of the pump laser diode (LD) of the EDFA, the laser started to oscillate in passive mode-locking regime with the LD current at the 1307mA. Further increase of the LD current made the laser stable, and the shortest pulse width was achieved at 1419 mA. The average power output of the EDFA was 18.8 dBm, and the average power output of the laser was −1.8 dBm. Since the EDFA contained long fiber, laser cavity was long and the laser output was slightly unstable. When we inserted 10 to 30-m-long single mode fiber to manage the round trip dispersion of the laser cavity or raised the EDFA output, the laser started to harmonically oscillate, and soliton pulse output could not be achieved.

Figure 8 shows the optical spectrum, the autocorrelation trace and the pulse train of the fiber laser output. The data was measured when we achieved the shortest pulse width and the highest pulse power. The laser had the center wavelength at 1565 nm and the 3-dB bandwidth of 3.7 nm. The full width half maximum (FWHM) of the autocorrelation trace was 1.61 ps, which correspond to inferred pulse duration of 1.14 ps (assuming Gaussian pulse profile). The time bandwidth product (TBP) of the pulse was 0.528. Comparing to an unchirped transform-limited value of Gaussian pulse (0.441), the result indicates the pulse had chirp. The chirp was due to the residual dispersion in the laser cavity. The repetition rate of the fiber laser was 1.54 MHz according to the time interval of the output pulse train. After the experiment, we measured the insertion loss of the microfiber and confirmed the loss did not change before and after the experiment. We expect that we can achieve much shorter and higher intensity pulse output with high stability by employing highly Er³⁺-doped EDF and designing the laser cavity parameters, such as length and round trip dispersion.

 

Fig. 8 Passively mode-locked fiber laser output (a) Optical spectrum of the fiber laser (resolution: 0.1 nm) (b) Autocorrelation trace of laser output (resolution: 50 fs,) corresponding pulse width: 1.14 ps (assuming Gaussian pulse) (c) Output pulse train of the fiber laser (repetition rate: 1.54 MHz)

Download Full Size | PDF

4. Conclusions

In this study, we proposed and demonstrated optically manipulated CNT-deposition around a microfiber. First, we fabricated a microfiber by stretching a bare SMF with heat. Deposition of CNTs around microfiber was achieved only by injection light into the microfiber. During the deposition, we performed in situ monitoring of the transmitted power to control the number of deposited CNTs. We achieved the CNT deposited microfiber with total insertion loss of 6 dB. The existence of CNTs was ensured by microscopic Raman spectroscopy. The microfiber had over 1.2% of saturable absorption depth at 17 dBm input power. The microfiber was used as a saturable absorber in a fiber ring laser, and provided passive mode-locking. The laser output had the bandwidth of 3.7 nm and the pulse duration of 1.14 ps at 1.54 MHz repetition rate.

In the whole experiment, we required only few drops of the CNT-dispersed DMF and CNT dissipation was small. Although the number of used CNTs was small, the CNTs were on the surface of the microfiber, and the CNT-layer and the evanescent field was effectively coupled with each other. Therefore, we were able to realize sufficient saturable absorption depth and passive mode-locking of the fiber laser. Our proposed technique possesses large advantages of the efficiency of CNT use and the simplicity of the device fabrication process compared to the alternative fabrication techniques. We believe our technique contribute to development of ultrafast photonic devices based on CNTs.