1. Introduction

The novel optical properties of single walled carbon nanotubes have recently attracted significant interest. They have been demonstrated to have a fast saturable absorption with sub-picosecond recovery times in the near infrared region [1, 2]. The saturable absorption occurs on the S ₁₁ transition, with the wavelength of this transition depending on the type of nanotube (metallic versus semiconducting) as well as the diameter of the nanotube [3]. Subsequently, single-walled carbon nanotubes have been used to passively modelock both erbium-doped fiber lasers at 1.55 µm [4–8], as well as ytterbium-doped fiber lasers at 1 µm [9].

A variety of techniques have been implemented in order to integrate carbon nanotubes into fiber devices. The first demonstration of a saturable absorber for modelocking in an erbium laser used an approach in which the nanotubes were sprayed onto a glass slide and micro-optics were used to couple out of the fiber, focus into the sample, and finally couple back into a second optical fiber [4]. In another approach, single walled carbon-nanotubes were grown directly onto optical fiber end-faces which has the advantage of direct integration with the fibers, but requires the ability to manufacture high purity nanotubes [5]. Carbon nanotubes have also been effectively incorporated into polymer thin films, which gives good performance as a saturable absorber, but requires extensive processing [6, 10].

Recently, a novel technique was demonstrated for depositing carbon nanotubes on the ends of optical fibers using optical radiation propagating in the fiber [11,12]. The carbon nanotubes are preferentially deposited in the region of the core allowing for optimal interaction with radiation propagating in the fiber, while at the same time minimizing the waste of the nanotubes during device preparation.

 

Fig. 1. Setup for depositing carbon nanotubes on the ends of cleaved optical fibers using optical radiation. Forces due to optical radiation are also shown.

Download Full Size | PDF

In this paper the process for depositing the nanotubes is presented in detail. Various mechanisms capable of depositing nanotubes on the fiber tip are discussed. By a process of elimination, thermophoresis caused by localized heating of the ethanol and nanotubes due to optical absorption is considered the likely candidate for the driving force behind the deposition. Saturable absorbers are constructed by depositing the nanotubes on fiber connectors. Using these saturable absorbers modelocking is demonstrated in an erbium-doped fiber laser with a spectral FWHM of 10.3 nm, and a pulse FWHM of 247 fs. Modelocking is also demonstrated in an ytterbium-doped fiber laser, generating 137 fs pulses from an oscillator-amplifier configuration.

2. Deposition of carbon nanotubes

The setup for depositing nanotubes on optical fiber end-faces is shown in Fig. 1. A solution of nanotubes was prepared by mixing approximately 0.5 mg of nanotubes with 12 cc of ethanol and ultrasonicating (Branson B-12) for 15 to 30 minutes. Next the optical fiber was prepared by removing the coating, cleaving and then placing the fiber into the solution. Optical radiation from a 980 nm or an amplified 1550 nm laser diode was then propagated through the fiber. The radiation was left on for a length of time ranging from a few seconds to several minutes. Then the laser diode was turned off and the fiber removed from the solution.

Figure 2(a) shows the results of the control experiment, in which the fiber was placed in solution and left for 30 minutes, but no optical radiation was propagated in the fiber during that time. In this case, the carbon nanotubes did not adhere to the fiber. In contrast, Fig. 2(b) shows the result when 10 mW of 1560 nm radiation were propagating in the fiber and the fiber was left in the solution for 1 minute. With the presence of optical radiation, carbon nanotubes have been preferentially deposited in the core region of the fiber. When the length of time was increased to 4 minutes, the area over which the nanotubes were deposited correspondingly increased, as shown in Fig. 2(c). With sufficient time and/or optical power, the entire fiber endface could be covered with nanotubes. Figure 2(d) shows the result when 30 mW of 1560 nm light was propagated in the fiber for 5 minutes. Under these conditions, the entire end face of the fiber was coated with nanotubes. Furthermore ring structures far outside the core area become visible. The minimum optical power for depositing the nanotubes was approximately 5 mW. At this time, only non-polarization maintaining fibers have been used in the experiments and no dependence of the deposition on polarization has been observed.

These ring structures were frequently observed even when the fiber end face was only partially coated with nanotubes. With optical power greater than the threshold for deposition propagating in the fiber, the nanotubes always deposited in the region of the optical core. However, frequently the nanotubes were observed to deposit in a ring shape around the core, in which the outer regions of the core were partially covered with nanotubes, as shown in Fig. 2(e).

 

Fig. 2. Images of SMF end-faces after nanotubes have been deposited. (a) No optical radiation propagating in the fiber. Fiber was left in solution for 30 minutes. (b) Power=10 mW; λ=1560 nm; left in solution for 1 min. (c) Power=10 mW; λ=1560 nm; left in solution for 4 min. (d) Power=30 mW; λ=1560 nm; left in solution for 5 min. (e) Ring structure observed for deposition conditions similar to (c).

Download Full Size | PDF

 

Fig. 3. Images of fiber end-faces after nanotubes have been deposited (a) Control experiment with no radiation propagating in the 980 single-mode fiber. (b) Power=6 mW; λ=980 nm; left in solution for 15 min.

Download Full Size | PDF

Deposition was also observed using 980 nm radiation. This observation is advantageous as it means the technique can be used for a variety of single-mode fibers over a range of wavelengths. In these experiments optical fibers that were single-moded at 980 nm were used. Figure 3(a) again shows the control experiment in which the fiber was left in solution for 30 min, with no optical radiation propagating in the fiber; the nanotubes did not adhere to the fiber end-face. However, when 6 mW of 980 nm light was propagated in the fiber for 15 minutes, the entire fiber end-face was observed to be coated. As in the case when 1560 nm radiation was used, a ring-structure was observed around the core of the fiber. In comparison to the size of the optical core seen in Fig. 3(a) the observed ring structure is far outside the core region.

In light of the preferential deposition of the nanotubes in the region of the fiber core in the presence of optical radiation, the forces induced by the optical radiation on the nanotubes suspended in liquids are considered. Suspensions of micron sized dielectric particles can generate strong artificial Kerr nonlinearities in which forces due to optical gradients are sufficient to move particles into regions of high optical power [13]. These Kerr nonlinearities can be sufficient to produce self focusing and self-trapped filaments, even for low power CW beams. In three dimensions, a single, tightly focused beam can have gradient forces that are strong enough to overcome the scattering forces to provide trapping of small particles in three dimensions [14]. Furthermore, optical trapping of carbon nanotubes has been demonstrated in holographic optical traps [15], so movement of the nanotubes in response to optical radiation in our setup is expected.

 

Fig. 4. Movies of the movement of nanotubes suspended in liquid due to 980 nm light. The 980 nm light is scattered by the liquid and viewed from the side through an infrared viewer. The length of the visible light region is approximately 2 cm. (a) (2.54 MB) when the fiber was vertical [Media 1], and (b) (2.81 MB) when the fiber was angled with respect to the vertical [Media 2].

Download Full Size | PDF

However, in general the divergence of the radiation from a single mode optical fiber is insufficient for the gradient force to overcome the scattering force in the forward direction. Multiple beam geometries that balance scattering forces with counter-propagating beams are frequently used when optical trapping experiments are performed using light delivery with fibers [16]. Alternatively, single beam optical traps based on lensed fibers have been demonstrated, but they do not provide trapping in three dimensions. Rather, the particle is trapped against a surface, and the normal force between the surface and the particle to be trapped helps overcome the scattering force [17]. Recently, a tapered fiber was demonstrated to be capable of single-beam trapping, however, the mode field diameter of 1 µm was considerably smaller than that available from standard single mode fibers [18]. Therefore, considering the relatively large mode-field diameter and low divergence of light from a standard single mode fiber, the optical forces on the nanotubes suspended in liquid are a combination of the gradient force, acting perpendicular to the beam to move the nanotubes from areas of low light intensity to high light intensity, and the scattering force, operating in the direction of the beam’s k vector. These forces are depicted in Fig. 1. Consequently, another process is required to overcome the scattering force and bring the nantoubes to the fiber tip.

Strong convection currents have been observed due to localized heating in suspensions of dielectric particles [19]. In the present nanotube deposition experiments, when the laser was turned on with the fiber dipped in the solution, convection currents, centered on the fiber tip, were observed. A movie of the currents caused by turning on the laser light is shown in Fig. 4. These movies were taken with an infrared viewer imaging the scattering of the 980 nm light from the side. Figure 4(a) was taken when the fiber was vertical. Prior to turning on the laser, the nanotube solution was observed to be still. The movie clearly shows the induced current moving the nanotubes upward toward the fiber tip. In contrast, Fig. 4(b) shows the current when the fiber tip was angled. In this case, it can be seen that the nanotubes do not directly follow the path of the light, but rather their motion is closer to the vertical. Therefore, gradient forces are insufficient to prevent the nanotubes from being carried out of the beam by convection currents.

 

Fig. 5. (a) Photo of the fiber-integrated carbon-nanotube saturable absorber. (b) Schematic of the modelocked erbium-doped fiber ring laser.

Download Full Size | PDF

However, nanotubes were still observed to deposit on the fiber core when the fiber was angled. In fact, deposition occurred even when the fiber was bent to point upwards, and convection was observed to carry nanotube particles away from the fiber tip. Consequently, another process is required to explain the deposition process. Given that convection currents show the occurrence of heating in the nanotube suspension, another potential process to consider is thermophoresis, which is the motion of particles suspended in a temperature gradient from hot areas to cold in a liquid or a gas. Thermophoresis in liquids has been shown to be capable of causing particles to deposit out of a hot liquid onto a cold wall [20].

According to simple estimates, the thermophoretic motion of the nanotubes suspended in ethanol should be very small as the nanotubes have a thermal conductivity many times higher than that of ethanol, effectively eliminating the temperature gradient of the ethanol in the vicinity of a nanotube [21, 22]. However, the study of thermophoresis in nanofluids is a topic of significant current interest, and recent studies have shown that metal particles dispersed in aqueous solution can, in contrast to simple considerations, have very large thermophoretic velocities, potentially caused by surface potential distributions [23, 24]. Given these considerations, thermophoresis occurring at the fiber/ethanol interface seems likely to be playing a role in the deposition process, and the interplay between the steady state temperature gradients caused by the optical absorption of the nanotubes and ethanol, as well as the additional influence of the gradient force directing the nanotubes towards the core of the fiber could be responsible for the observed ring structures. However, more study is required to fully understand the role of all the forces involved.

3. Modelocked fiber lasers with optically deposited carbon nanotubes

In order to make saturable absorbers that could be easily integrated with fiber lasers, the experiments were repeated using connectorized fibers, rather than cleaved fibers. Angled FC/APC connectors were used, as flat connectors tended to destabilize the laser modelocking.

The process was the same as described above. The optical power out of the connector was measured. Then the connector was dipped in the nanotube solution for approximately 30 seconds. The connector was then removed, the ethanol was allowed to evaporate, and the optical power re-measured. As the nanotubes deposited on the core of the fiber, the transmitted optical power dropped. This procedure was repeated until the linear loss of the connector was between 1 and 3 dB. Finally, the connector with nanotubes deposited on the end face was mated to a clean connector to create a saturable absorber that could be spliced into our fiber laser. A photo of a typical device is shown in Fig. 5(a).

Commercially available single walled carbon nanotubes manufactured by a range of techniques can be obtained from a variety of sources. Nanotubes from Carbon Nanotechnologies, Inc. (CNI) [25] and Southwest Nanotechnologies [26] were tested. The nanotubes from CNI have a mean diameter 1.0 nm, while those from Southwest Nanotechnologies have a mean diameter of 0.8 nm. Those nanotubes with larger diameters are more suited to operation at 1.55 µm whereas the smaller tubes are for operation at 1 µm [3].

 

Fig. 6. (a) Spectrum, (b) correlation, and (c) pulse train measured from the an erbium fiber laser modelocked with the carbon nanotube saturable absorber. The pulses had a time-bandwidth product of 0.312.

Download Full Size | PDF

A unidirectional ring laser was constructed as shown in Fig. 5(b). Depending on the particular device, the threshold for modelocking was typically between 20 and 30 mW of 975 power, and the maximum pump power before the pulse broke up into multiple pulses was around 45 to 50 mW. The laser was self starting and modelocked independent of the polarization state in the cavity, although the spectrum was somewhat polarization dependent and a polarization controller was included to optimize the spectral width.

The laser modelocked over a wide range of cavity lengths, erbium fiber lengths, and repetition frequencies, from less than 10 MHz to greater than 80 MHz. The length of anomalous dispersion SMF was used to balance the normal dispersion of the erbium-doped fiber to provide the broadest possible spectrum. The broadest measured modelocked spectrum, plotted in Fig. 6(a), had a FWHMof 10.3 nm. The pulse repetition frequency for this laser was 81.7MHz, the output power was 2.6 mW, and pulse energy was 32 pJ. The measured correlation width was 381.1 fs corresponding to a pulse width of 247.5 fs (assuming sech² pulses) and a time-bandwidth product of 0.312, compared to an expected value of 0.315 for sech² pulses. The nanotubes used in this laser were from CNI. Although the nanotubes from Southwest Nanotechnologies had smaller diameters and were intended for operation at 1 µm, they were also tested in an erbium laser. While the laser modelocked, the observed spectrum was usually no broader than 3.5 to 4 nm.

Note that the performance of this laser is very good in comparison to other techniques for carbon nanotube saturable absorber preparation. Pulse widths of 318 fs were observed using a saturable absorber made by the spray-on technique [4] and 473 fs were obtained using a D-shaped fiber for enhanced interaction with nanotubes [7]. Pulse widths of 171 fs have been obtained from a polymer-nanotube thin film [6] by having careful control of the nanotube concentration in the polymer film. With the current approach, the transmitted optical power could be monitored during the deposition process in order to gain fine control over the number of nanotubes deposited on the core of the optical fiber and optimize the saturable absorber properties.

An Yb-doped fiber ring laser was also constructed that operated at 1.07 µm. For this laser the nanotubes from Southwest Nanotechnologies were used. The ring laser was similar to that shown in Fig. 5(b), however, at 1 µm, anomalous dispersion in an optical fiber is more difficult to achieve. A specially designed fiber with anomalous dispersion in a higher-order mode (HOM) was used to balance the normal dispersion of the other fibers in the cavity [27,28]. The spectrum from the modelocked laser is shown in Fig. 7(a). The spectral FWHM for this laser was 7.2 nm. At a repetition frequency of 20.1 MHz, the laser output power was 0.1 mW and pulse energy was 5 pJ. The power was amplified to 20 mW in an external Yb amplifier as the correlator did not have sufficient sensitivity to measure pulses directly from the oscillator. After the amplifier the pulses were re-compressed in a second HOM module. The correlation, giving a pulse FWHM of 137 fs, is shown in Fig. 7(b). While the amplification and subsequent pulse compression distort the pulses as is evident in Fig. 7(b), the peak to background ratio for the correlation was as expected, and there was no indication of a coherence spike in the correlation, indicating the nanotube saturable absorption was sufficient to produce a high-quality pulse train. The Yb laser would not modelock without the HOM fiber in the cavity. Furthermore, neither the Yb or Er laser would modelock without the nanotube saturable absorber in the cavity. Again, performance of this laser compared very well to previous demonstrations of Yb lasers where the saturable absorbers were made by other techniques [9].

 

Fig. 7. (a) Spectrum from the Yb oscillator modelocked by carbon nanotubes as well as after an Yb amplifier. (b) Correlation measured after the amplifier.

Download Full Size | PDF

To date, long term performance degradation in the erbium-doped fiber lasers has not been observed when the lasers were operated at power levels required for a single pulse in the cavity. At 1.07 µm, the ytterbium laser remained modelocked while the pump power was on. However, the performance of the saturable absorber degraded when the laser was cycled off and on again. Ultimately, the laser ceased modelocking. In both the erbium and ytterbium lasers the cw-modelocking state is preceded by a brief period of Q-switching. Potentially, this Q-switching state was sufficient to damage the nanotubes in the Yb laser. In the erbium laser, nanotube damage was also observed, but it occurred at much higher pump powers (>100 mW) than was required for single pulse modelocked operation.

4. Conclusions

In summary, a very simple, inexpensive technique for making fiber integrated, carbon-nanotube saturable absorbers using commercially available single walled carbon nanotubes has been presented. The nanotubes were dispersed in liquid and deposited on the end face of cleaved optical fibers and optical fiber connectors using optical radiation propagating in the fiber. Thermophoresis is considered a likely mechanism for the nanotube deposition on the endface of the fiber. Using these saturable absorbers, 247.5 fs pulses at 1.56 µm in an erbium-doped fiber laser, and 137 fs pulses at 1.07 µm in an amplified ytterbium-doped fiber were obtained.