1. Introduction

Ultrafast pulse lasers are of unprecedented interest for their wide applications in optical communication, military, industry, medicine, and fundamental research [1–5]. Passively mode-locked fiber lasers are among the best pulse sources available today due to their structural simplicity and their ability to generate picosecond and subpicosecond optical pulses [6, 7]. In recent years, single-wall carbon nanotubes (SWCNTs) have been mostly used as saturable absorbers (SAs) in passively mode-locked fiber lasers. Compared with traditional SAs, SWCNTs exhibit many advantages, including subpicosecond recovery time, wide operating bandwidth, and mechanical and environmental stability [8–11]. Besides, good compatibility with fiber has been realized by incorporating SWCNTs into polymers to form SAs [12]. In order to promote the practical applications of passively mode-locked fiber lasers, much research has been carried out to attain good mode-locking performance such as tunable laser operating wavelength, short pulse width, low insertion loss, and high optical-to-optical efficiency [9–13]. Particularly, low mode-locking threshold power is significant for reducing the cost of pulse fiber lasers [11]. Numerous attempts have been done to realize this purpose, for example changing polymer matrices, tuning the mixing ratio of SWCNTs and polymer, and producing polymer-free SAs [12, 14]. However, the extent of the threshold power reduction is limited. It is still a challenge to achieve pulse fiber laser with acceptable mode-locking threshold.

In the case of noble metal nanocrystals (NCs), the extraordinary physical properties of them are mainly arise from their surface plasmon (SP). The interaction between the SP and incident electromagnetic field bring about a number of interesting optical events such as surface enhanced Raman scattering [15], large third-order nonlinearity [16], and much efficient fluorescence [17]. Two possible reasons (SP resonance and SP field enhancement) have been proposed to explain these phenomenons [18–20]. Surface plasmon field enhancement (SPFE) is attributed to the collective motion of free electrons restricted to narrow regions, similar to that observed in colloidal nanoparticles exposed to an external electromagnetic field. The degree of field enhancement induced by SP is extremely sensitive to the shape and the distribution of the nanostructure. Therefore, large electromagnetic density and low excitation threshold can be induced in a proper designed material structure. For instance, ultra-low threshold high-harmonic generation by resonant plasmon field enhancement has been obtained [21]. We have also reported the highly enhanced local field density of SP leading to low pump threshold for upcoversion emissions [22, 23]. In addition, a terahertz emitter consists of gold photoconductive antenna has been studied by the finite-difference time-domain (FDTD) technique. Less pump power and higher terahertz output power were achieved by the enhancement of the local electric field [24]. The strongly confined and enhanced electromagnetic field on the surface is particularly useful for the excitation of materials with very small absorption section, such as thin films. If the noble metal NCs were introduced into the SWCNTs-based SAs, low mode-locking threshold may also be obtained by the field enhancement in pulse fiber lasers.

Herein we present a new technique to decrease the mode-locking threshold of pulse fiber laser through integrating of SWCNTs with gold nanoparticles (GNPs). Firstly, a ring cavity erbium-doped fiber laser (EDFL) was set up based on SWCNTs. Stable passively mode-locking was achieved for a threshold of ~30.5 mW, with a repetition rate of 33 MHz at 1559 nm. The pulse width was 639 fs at a pump power of ~90 mW. Secondly, water soluble GNPs were easily mixed with SWCNTs in sodium carboxymethylcellulose (NaCMC) to form new SAs. The stable mode-locked pulse outputs were achieved in all the EDFLs using these SAs. The mode-locking threshold of these lasers decreased first and then increased with the concentration of GNPs in the SAs increasing. The variation range of the mode-locking threshold was up to 21.5 mW. When the concentration was 0.006 mmol/ml, the threshold reached the lowest value as 16 mW. The mechanism associated with the mode-locking threshold reduction was discussed based on the SPFE effect.

2. Experimental

2.1 Preparation and characterization of SWCNTs@GNPs-NaCMC and SWCNTs-NaCMC films

SWCNTs with diameters of 1-1.5 nm used in these experiments were commercially available from Carbon Nanotechnologies Inc. Stable suspensions of SWCNTs (4 mg) in 1 wt% aqueous solution (10 ml) of NaCMC (medium viscosity, Sigma) were prepared by ultrasonication. The concentration of SWCNTs was 0.04 wt%. This suspension was kept for 48 h and no precipitation was observed. GNPs were synthesized through reducing hydrochloroauric acid by trisodium citrate at 99 °C for 20 min. The GNPs were then separated via centrifugation and redispersed in aqueous solution to form ten GNPs solutions (10 ml) with increased concentration (0.002, 0.004, 0.006, 0.008, 0.01, 0.012, 0.014, 0.016, 0.018, and 0.02 mmol/ml). These GNPs solutions were mixed with ten same SWCNTs-NaCMC solutions (10 ml) by ultrasonication, separately. To avoid the impact of the decreased SWCNT concentration on the threshold power, 10 ml aqueous solution without GNPs was also mixed with 10 ml of SWCNTs-NaCMC solution to serve as the compared sample. SWCNTs@GNPs-NaCMC and SWCNTs-NaCMC films were formed by casting these solutions onto flat substrates, followed by a slow drying at room temperature.

Transmission electron microscopy (TEM) was used to characterize the SWCNTs and the GNPs. Individual SWCNT strain can be resolved from the nanotube bundles, confirming that the sample indeed contains high-quality SWCNTs as shown in Fig. 1(a). The SWCNTs are found to be attached by the aggregated GNPs in Fig. 1(b), because they were only mixed by ultrasonication. The average diameters of these GNPs are about 20 nm as can be seen from the insert of Fig. 1(b). For conciseness, only atomic force microscopy (AFM) images of the SWCNTs-NaCMC film and one SWCNTs@GNPs-NaCMC film (the concentration of GNPs is 0.006 mmol/ml) are shown in Figs. 1(c) and 1(d). The surfaces of the two thin films are both smooth and crack-free.

 

Fig. 1 TEM images of (a) the SWCNTs and (b) the SWCNTs@GNPs (Insert: TEM image of the GNPs). AFM images of (c) the SWCNTs-NaCMC film and (d) the SWCNTs@GNPs-NaCMC film (the concentration of GNPs is 0.006 mmol/ml).

Download Full Size | PDF

The absorption spectra of the SWCNTs@GNPs-NaCMC films with different GNP concentration (from 0 to 0.01 mmol/ml) at an excitation power of ~200 mW are shown in Fig. 2(a). The absorption intensities of these films increase linearly with increasing the GNP concentration as shown in the insert of Fig. 2(a). Owing to the fact that the absorption peaks of the SWCNTs cannot be clearly observed in Fig. 2(a), we also give absorption spectra of these SWCNTs@GNPs-NaCMC films at the wavelength from 1200 to 1800 nm in Fig. 2(b) and compare the absorption spectra of NaCMC, SWCNTs-NaCMC, and SWCNTs@GNPs-NaCMC films from 800 to 1800 nm as shown in Fig. 2(c). There are two main absorption bands at around 1300 and 1600 nm in the absorption spectrum of the SWCNTs-NaCMC film, the second one making them suitable for operation at 1560 nm. Moreover, the absorption band of the SWCNTs around 1600 nm indeed exists in the SWCNTs@GNPs-NaCMC film. It should be paid attention to the absorption band shifting of the film after introducing GNPs, the reason for which we have not known very exactly. The aggregation of GNPs can be induced by the adhesion and thickening effect of the polymer film. A plasmon mode coupling of the GNPs may be caused by the aggregation, which results in the red shifting and broadening of the plasmon resonance in the optical spectra. It can be seen that the absorption bandwidth becomes larger gradually in Fig. 2(a). We presume that the shifting and broadening of GNP absorption band possibly have some effect on the absorption band of SWCNTs. In our experiment, the absorption intensity of SWCNTs is much lower than that of GNPs. Although the main absorption wavelength of the GNPs locates at 520 nm, obvious effect on the absorption band of SWCNTs such as the shifting may be induced. As can be seen from Fig. 2(b), the absorption intensity of SWCNTs increases with the increase of GNPs, although the concentrations of SWCNTs maintain at the same value.

 

Fig. 2 The optical absorption spectra of (a) the SWCNTs@GNPs-NaCMC films with different GNP concentration (0 to 0.010 mmol/ml) from 400 to 1800 nm (Insert: dependence of the absorption intensity on the GNP concentration), (b) from 1200 to 1800 nm, and (c) the NaCMC film, the SWCNTs-NaCMC film, and the SWCNTs@GNPs-NaCMC film (the concentration of GNPs is 0.006 mmol/ml) from 800 to 1800 nm.

Download Full Size | PDF

2.2 Setup of EDFL

These obtained films were placed between two fiber connectors to form fiber-compatible SAs and then integrated into a laser cavity respectively as shown in Fig. 3. The fiber laser was pumped by a 980 nm laser diode (LD) through a 980/1550 nm wavelength-division multiplexer (WDM). A 20-cm-long single-mode EDF was used as the gain medium. Unidirectional light propagation was ensured by an optical isolator (ISO). The mode-locked laser pulse was output from the 10% port of the 10 dB WDM coupler. The rest of the cavity consisted of a SMF-28 single-mode fiber and the total cavity length was 6 m. The output lasers were analyzed by using an optical spectrum analyzer, a digital oscilloscope and an autocorrelator.

 

Fig. 3 Schematic illustration of the mode-locked EDFL.

Download Full Size | PDF

3. Results and discussion

When the SWCNTs-NaCMC film was integrated into the ring cavity, continuous wave (CW) laser operation started at 20 mW pump power and a narrow spectrum appeared as shown in Fig. 4(a). There was no pulse train in the digital oscilloscope (Fig. 4(b)) and no pulse profile in the autocorrelator. It is in evidence that mode-locked pulse output of the EDFL has not been obtained. Stable self-starting mode-locked pulse laser oscillation was easily achieved as soon as the pump power exceeded the threshold of ~30.5 mW. The laser operated with a single pulse in the cavity up to a maximum pump power of ~90 mW, at which point either a CW spike was observed, or the pulse broke up into multiple pulses in the cavity. The emission spectrum of the mode-locked EDFL at a pump power of ~90 mW is shown in Fig. 4(c). The spectrum becomes broad obviously. The operating center wavelength is about 1559 nm and the 3-dB width is approximately 3.9 nm. The Kelly-sidebands of the spectrum, indicating the signature of soliton mode-locking, are caused by dispersive wave generation when the soliton pulse encounters periodic variations of gain, loss, and dispersion as it propagates around the laser cavity [25]. Figure 4(d) shows the output pulse trains of the mode-locked laser. The time interval between two adjacent pulses is about 30 ns, corresponding to a repetition rate of ~33 MHz, which coincides with the length of the laser cavity (~6 m). Figure 4(e) shows a single pulse profile of the laser measured by using an autocorrelator (Alnair HAC-200). The measured full width at half maximum (FWHM) of the autocorrelation trace is 985 fs, after being well fitted by a sech² function, the real pulse duration is 639 fs. The corresponding time-bandwidth was calculated as 0.321, which was higher than the transform-limited value. The polarization state of the fiber cavity is important to achieve pulse operation because of the fact that SWCNTs and SPs of GNPs are both sensitive to polarization; however, it is not a stringent requirement. The stable mode-locking was indeed attained simply by manipulating and taping down the fiber cavity by hand, even without the use of polarization controller (PC) in our experiment, which was also reported in previous publication [26]. Of course, we have also tried to use PC to improve the mode-locking performance of the laser. However, after the mode-locked laser was self-started at the optimum polarization orientation of the PC, no further tuning of the PC was required to start the laser, which was almost the same as inexistence of the PC. The performance of the laser output was not better, but the pulse duration was increased by longing the length of the laser cavity and larger insertion loss was produced. Figure 4(f) shows the output power of the mode-locked laser as a function of the pump power of the 980 nm laser. The output power increases linearly from 0.1 to 3.2 mW with increasing the pump power from 30.5 to 90 mW, and the resulting optical-to-optical efficiency is about 5%. The mode-locked EDFL became unstable when the pump power was higher than 90 mW. The SWCNTs-NaCMC ðlm might be damaged for a large pump power (>90 mW).

 

Fig. 4 (a) Emission spectrum of the EDFL based on a SWCNTs-NaCMC film, (b) laser intensity as a function of time at a pump power of ~20 mW, (c) emission spectrum, (d) output pulse train, (e) single pulse profile of the mode-locked EDFL at a pump power of ~90 mW, and (f) the output power of the mode-locked EDFL as a function of the pump power.

Download Full Size | PDF

Ten SWCNTs@GNPs-NaCMC films were inserted into the same ring cavity respectively. The stable mode-locked pulse outputs were obtained in all these EDFLs. Figure 5(a) shows the mode-locking threshold variation of these EDFLs. With the increase of GNP concentration from 0 to 0.01 mmol/ml, the threshold decreased first and then increased. When the concentration of GNPs increased to 0.006 mmol/ml, the threshold reached the lowest value as 16 mW. Since the initial mode-locking threshold of the EDFL based on the SA without GNPs was 30.5 mW, the largest decrease ratio of the threshold was 47.5%. However, with the increase of the GNPs further, the threshold increased inversely. When the concentration of GNPs was as high as 0.01 mmol/ml, the mode-locking threshold increased to 37.5 mW. If the concentration of GNPs was higher than 0.01 mmol/ml, only Q-switched pulse outputs without sign of mode-locking were observed for a large range of the pump power. Therefore, the variation range of the mode-locking threshold was up to 21.5 mW. The CW threshold variation of these EDFLs is almost the same as that of the mode-locking threshold as shown in Fig. 5(b). With the increase of GNP concentration from 0 to 0.01 mmol/ml, the threshold also decreased first and then increased. The lowest value of the CW threshold reached 10 mW with the concentration of GNPs as 0.006 mmol/ml. The variation range of the CW threshold was 17 mW and the largest decrease ratio was 50%. Figure 5(c) shows the relationship between the output power and incident pump power of the mode-locked EDFL with a SWCNTs@GNPs-NaCMC film (the concentration of GNPs is 0.006 mmol/ml). The output power increases linearly from 0.1 to 3.1 mW with increasing the pump power from 16 to 70 mW and the resulting optical-to-optical slope efficiency is about 5%. When the incident pump power was higher than 70 W, the output power decreased slowly and the absorber was damaged.

 

Fig. 5 Dependence of (a) the mode-locking threshold and (b) the CW threshold on the GNP concentration, (c) the output power of the EDFL mode-locked with a SWCNTs@GNPs-NaCMC film (the concentration of GNPs is 0.006 mmol/ml) as a function of the pump power, and distributions of excitation power density around GNPs with mutual distance as (d) 10, (e) 5, and (f) 1 nm in the x-y plane.

Download Full Size | PDF

We consider that after adding GNPs, a strongly amplified electromagnetic field can be induced by the SPFE effect in the nanostructure film [21, 22]. The excitation power density is directly enhanced by the amplified electromagnetic field. The pump power needed for mode-locked pulse decreases due to the efficient excitation relatively. As the concentration of GNPs increases in the film, the couple between neighbouring GNPs may become more effective. As a result, the incident electromagnetic field density becomes larger, the threshold decreases further. To demonstrate the above analysis, we performed simulations by using the FDTD method. The FDTD method is one of the most popular numerical methods for solving electromagnetic problems with arbitrary geometries and inhomogeneous materials. One major advantage is that broadband results can be obtained with the FDTD algorithm run only once, taking a pulse as the excitation source. This will dramatically save calculation time compared with other steady-state calculations [27]. Gold is a dispersive material whose constitutive parameters are frequency-dependent. The electronic transitions in a solid are more directly related to the complex dielectric constant$\tilde{\epsilon }={\epsilon }_1+i{\epsilon }_2$, instead of the complex index of refraction$\tilde{n}=n+ik$. They are connected by$\tilde{\epsilon }={\tilde{n}}^2$, so that${\epsilon }_1=n^2-k^2$and${\epsilon }_2=2nk$. The optical constants n and k in a large spectral range are available from reference [28]. The dielectric constant of the NaCMC is 15.5. In the calculations, a light source peaked at 1559 nm with an amplitude of 1 was used as the pump source, and the injection direction of the pump light was set as z-axis. Three types of gold nanostructures were considered. For consistence, only four GNPs with diameters of 20 nm were applied, the mutual distance between them was set as 10, 5, and 1 nm. For these structures, the calculated power density distributions of the excitation light in the x-y plane are shown in Figs. 5(d)-5(f), respectively. The pump power densities in the areas near these GNPs were calculated to be enhanced. The largest enhancement factor of local electric field was around 3 with the mutual distance of GNPs as 10 nm. When the distance decreased to 1 nm (in other words, more GNPs were added), the enhancement factor increased from 3 to 5. The fact is evident that the shorter distance, the larger enhancement factor. It is quite probable that the threshold decrease will be recorded under such enhanced pump power density. If the doping concentration of GNPs was too large, the aggregation of GNPs became serious and the scattering loss was enhanced, therefore the mode-locking threshold increased contrarily. Too much energy loss can even make the mode-locking operation transfer to Q-switching behavior. Obviously, there was an optimal concentration of GNPs in the film at which the mode-locking threshold was the lowest. The value was 0.006 mmol/ml under our experimental conditions. Although the optimal concentration cannot be exactly confirmed by the simulation results, it can be speculated that a strongly amplified electromagnetic field is indeed induced by the SPFE effect in the nanostructure film. At the same time, the more effective couple among GNPs with nearer distance owing to the increase of concentration can also be confirmed by the simulation.

The thicknesses of these films are about 40 μm. When the thickness was changed, the insertion loss of the film was different, but the optimal concentration of GNPs was nearly not affected. For example, the insertion loss increased with the thickness of the film as 60 μm. As a consequence, the initial mode-locked threshold also increased to 36 from 30.5 mW. The optimal concentration of GNPs was still 0.006 mmol/ml, but the lowest mode-locking threshold changed to 23 mW. If the film became further thicker, the insertion loss became higher and the pulsed laser operation was difficult to obtain. By contrast, when the film became thinner, the absorption was too low to achieve stable single pulse operation and multiple pulse operation was obtained. As a consequence, we fixed the film thickness as about 40 μm in our experiment. On the other hand, if the concentration of SWCNTs was tuned, the optimal concentration of GNPs became a little different. When the concentration of SWCNTs was reduced by half, the mode-locking of the EDFL based on SWCNTs started at ~26 mW. The optimal concentration of GNPs changed to 0.005 mmol/ml and the minimum mode-locking threshold increased to 20 mW, which is probably because that the scattering loss caused by the aggregation of GNPs became more obvious with less SWCNTs. However, a rise of the SWCNT concentration made the formation of bundles and the film can be easily damaged at low pump powers.

The spectrum, autocorrelation trace, and output pulse trains of the laser with the lowest mode-locking threshold were compared with those of the bare SWCNTs-based one as shown in Figs. 6(a)-6(f). The two mode-locked lasers both show broad spectra with 3-dB widths as approximately 3.3 and 3.4 nm at the pump power of 16 and 30.5 mW. The Kelly-sidebands of the two spectra indicate the signature of soliton mode-locking. The establishing of mode-locking can also be confirmed by the phenomena that single pulse and pulse train appeared in the autocorrelator and digital oscilloscope, respectively [26, 29]. However, the pedestals in the two autocorrelations are both distinct because of the low pump powers as shown in Figs. 6(c) and 6(d). The pedestal in the autocorrelation data is also an evidence of a little large noise in the cavity. The intercavity noise is probably caused by insertion losses and non-saturable absorbance of GNPs and polymer matrix. The FWHM of the autocorrelation traces are 1.332 and 1.185 ps, respectively. After being well fitted by a sech² function, the real pulse durations are 865 and 769 fs, resulting time-bandwidths as 0.3238 and 0.3334, which are both higher than the transform-limited value. The output pulse trains are very stable with the same repetition frequency of 33 MHz, which is determined by the length of the laser cavity.

 

Fig. 6 (a), (b) Emission spectra, (c), (d) single pulse profiles, and (e), (f) output pulse trains of EDFLs mode-locked with a SWCNTs@GNPs-NaCMC film (the concentration of GNPs is 0.006 mmol/ml) and a SWCNTs-NaCMC film.

Download Full Size | PDF

The synthesized GNPs were simply mixed with SWCNTs by ultrasonication in our experiment. The orientation of GNPs is difficult to be detected exactly due to their random distribution. However, the mode-locking threshold decrease is obvious. So far, several methods have been developed to coat metals onto SWCNTs with more regular morphology, such as impregnation, self-assembly, electrochemically deposition, and vapor deposition [30]. In the future, we will try to use these methods to synthesize SWCNTs@GNPs-NaCMC films. The influence of interaction between GNPs and SWCNTs on the mode-locking threshold power and the effect of polarization state on the final output will be further studied.

4. Conclusion

In summary, the mode-locking threshold of EDFL was successfully reduced by mixing GNPs with SWCNTs as SA. A ring cavity EDFL was firstly set up based on SWCNTs. Stable passively mode-locking was achieved for a threshold of ~30.5 mW, with a repetition rate of 33 MHz at 1559 nm. The pulse width was 639 fs for a pump power of ~90 mW. Water soluble GNPs were synthesized by using citrate as reductant. These GNPs were easily mixed with SWCNTs in sodium carboxymethylcellulose (NaCMC) aqueous solution to form new SAs. When these SAs were used in the EDFL, the mode-locking threshold firstly decreased from 30.5 to 16 mW, and then increased to 37.5 mW with the increase of GNPs. The concentration of GNPs in the SA for the lowest mode-locking threshold of EDFL is 0.006 mmol/ml. The largest decrease ratio of the initial threshold is 47.5%. The possible reason for the improved mode-locking threshold was attribute to the SPFE effect based on FDTD calculations. Our experiment results offer a promising strategy to reduce the mode-locking threshold of pulse lasers.