1. Introduction

Random fiber lasers (RFLs) have attracted extensive attention owing to underlying fundamental physics of randomly distributed feedback to offer random “cavities” and lower frequency noise. The unique spectral and noise properties make versatile applications in telecommunications, coherent light source and remote sensing, see [1] and references therein. Rayleigh scattering (RS) from refractive index inhomogeneity naturally presented in silica fiber has been employed to provide one-dimensional random distributed feedback for the improvement of lasing directionality and efficiency [2]. Once the gain overcomes the fiber loss, the photons recaptured by RS along the fibers turns to become lasing resonances. With the RS random feedback, RFLs based on different kinds of the gain mechanisms have been experimentally demonstrated such as Raman scattering [2–6], Erbium doped fiber amplification [7, 8] as well as Brillouin scattering [9–11].

Brillouin random fiber lasers (BRFLs) have been achieved with the combination of Brillouin gain and RS-based random distributed feedback in the linear cavity [9] and in ring cavity [10, 11], providing advantage of linewidth reduction for high-precision metrology [12], narrow-linewidth microwave generation [13] and truly random number generator [14], thanks to the randomly distributed feedback to get narrow laser linewidth without the phase locking loops. Although the Brillouin scattering can provide much higher gain coefficient, RS in the silica fiber is usually weak (typically ~10⁻⁵ km⁻¹) and different fibers are needed at both ends of the linear cavity [9] or embed inside the closed loop cavity [10, 11] for sufficient random feedback of the lasing generation. Coherent BRFL in the uniform fiber with the unidirectional pump injection is hard to be achieved mainly due to insufficient RS random feedback. In addition, the generation of the 2nd Stokes in long uniform fibers with low Brillouin threshold turns out to be much easier, leading to the gain saturation which is detrimental to the 1st Stokes lasing. Recently, the bi-directional Brillouin pump scheme combining with RS random distributed feedback has been proposed to efficiently build up the BRFL with the narrow linewidth as well as the low frequency noise [15]. Relative intensity noise (RIN) of the bi-pump BRFL can also be improved with the assistance of random feedback originated from random fiber gratings [16]. Thanks to the bi-directional pump, the Brillouin gain is established in both directions for the Rayleigh scattered Stokes along the fiber. However, the Stokes laser emission of the bi-pump BRFL always accompanied with strong residual pump light and a pure Stokes lasing emission can only achieved by utilizing an additional narrow bandpass filter to filter out the strong residual pump light.

Refractive index inhomogeneity in silica fiber plays a critical role in distributed Rayleigh scattering for the RFL. The refractive index in fibers can be artificially modified and shaped using fibers tapering technique [17], offering versatile applications in optical devices manufacture [18], high power laser generation [19] and nonlinearity enhancement [20], etc. Herein, tapered fiber can provide an enhancement of the inhomogeneity along the fibers for strong distributed Rayleigh scattering, indicating an alternative candidate for random fiber laser generation.

In this paper, we demonstrated a novel Brillouin random fiber laser based on a 5-km tapered fiber (TF) with unidirectional pump injection for the first time. Rayleigh scattering in the TF is significantly enhanced due to the non-uniform distribution of the effective core area as well as the refractive index. Therefore, highly accumulated Rayleigh scattered Stokes acting as sufficient random distributed coherent feedback is then amplified by the Rayleigh scattered pump light. Consequently, the efficient Stokes lasing can be built up with the combination of the effective Brillouin gain in one portion of the fibers and the distributed RS feedback along the whole fibers. The laser linewidth characterized by a delayed self-heterodyne method was measured as ~1.17 kHz. The results show that the RIN of the TF-based BRFL is suppressed comparing with the SMF-based BRFL using bi-directional pumps. Furthermore, a high-resolution (kHz) linewidth measurement method is demonstrated without long delay fiber (>100km) and extra frequency shifter thanks to the acoustic frequency shift from fiber itself.

2. Principle of tapered-fiber based BRFL

The TF was designed with the mode field diameter from 5.0 μm at one end to 7.0 μm at the other end and the dispersion from 7.7 ps/nm/km to −0.3 ps/nm/km by changing the core delta. To create a distributed Brillouin frequency shift in the tapered fiber, the effective acoustic velocity was achieved by varying the SiCl₄ and GeCl₄ flows across the preform cross-section and along the preform length to control the glass composition and index. The preform was then drawn into a fiber of 5 km by using the conventional draw technology [20]. Consequently, a 5-km TF introduces strong refractive index inhomogeneity, which is intrinsically good candidate for distributed Rayleigh reflectors. Figure 1 schematically shows the principle of the TF-based BRFL. The pump light is injected from one end (Port #2) of the TF and generates the backward Stokes via stimulated Brillouin scattering (SBS). Both pump and Stokes lights can be strongly reflected by the distributed RS in the opposite directions. Similar to the bi-pump scheme [15], the transmitting and Rayleigh scattered pump light would deliver efficient Brillouin gain for the Stokes light in both directions to compensate each ‘round-trip’ losses of photons, establishing the lasing conditions.

 

Fig. 1 Principle of the TF-based BRFL.

Download Full Size | PDF

Along 5-km TF, the Brillouin gain varies with a distributed peak gain and a Brillouin frequency shift, resulting a higher Brillouin threshold. As shown in Fig. 2, the distributed Brillouin gain spectrum along TF was characterized based on the pump-probe technique [21]. The Brillouin frequency shift ν_B shows an increasing trend from ~10 GHz at Port 2 to ~10.35 GHz. The distributed Brillouin gain along the TF shows a peak value of 5.5 dB with a central frequency shift of ν_B = 10.05 GHz within the range from z = 600m to z = 1400m, which essentially gives the Brillouin gain in short fiber length (~800m) for the lasing.

 

Fig. 2 Measured Brillouin gain spectrum along the TF.

Download Full Size | PDF

Once the pump power reaches the threshold, the TF-based BRFL can be realized by efficient Brillouin gain along the one portion of the TF and coherent random feedback in both directions. Multiple reflectors from distributed RS along the fiber form many Fabry–Pérot interferometers, and their superposition finally selects random modes for lasing. The TF exhibits several superiorities for establishing the BRFL: 1) The strong inhomogeneity of the TF significantly enhances the RS with the coefficient (∼-34 dB/km) by 1-2 order higher than uniform fibers such as SMF, which is essential to recur the photon of both pump and Stokes lights for the bi-directional Brillouin gain and the lasing oscillation. 2) The smaller fiber core area of the TF confines the Stokes and pump light for a higher power density which is beneficial to a larger Brillouin gain. 3) The relatively higher Brillouin threshold in the TF postpones the 2nd Stokes generation as well as the gain saturation which is critical for a sufficient Brillouin gain of the 1st Stokes lasing. 4) Brillouin interaction between pump light would mainly locate at gain-peak fiber section (from z = 600 to z = 1400m) of the TF, alleviating the instability induced by the thermal and vibration disturbances from the external environment.

3. Experimental results and discussion

The experimental setup of the Brillouin random fiber laser (BRFL) is shown in Fig. 3. Light from the pump laser (Rock module, NP Photonics) was amplified by an Erbium doped fiber amplifier (EDFA) and then launched into the TF inside an aluminous soundproof box isolating from external disturbance. An isolator was connected at the fiber end to avoid undesired light reflection. By increasing the input pump power above the threshold, the generated Brillouin random laser outputs through an optical circulator (CIR) and is detected by a power meter and an optical spectrum analyzer (OSA).

 

Fig. 3 Experimental setup of taper-fiber (TF) based Brillouin random fiber laser (BRFL) (a) power and optical spectrum monitoring; (b) RIN measurement; (c) delayed self-heterodyne (DSH) method for linewidth measurement.

Download Full Size | PDF

The laser output was monitored by a power meter as increasing the input pump power, as shown in Fig. 4(a). As the input pump power is lower than the threshold, the output power remains extremely low caused by the weak RS of the injected pump light. In the experiment, the threshold is found to be 21.8 mW. The laser output abruptly changes when the input pump power surpasses the threshold. The laser output power then gradually rises up as increasing the pump power with a slope efficiency of 11.2%. The linewidth measurement of the TF-based BRFL is conducted by a conventional delayed self-heterodyne (DSH) method. Laser output was split by a 5/95 optical coupler into two parts, as show in Fig. 3(C). The 5% branch of the laser was sent to an Acoustic-optic Modulator (AOM) with 40MHz frequency downshift while the 95% part was sent through 200-km delay fiber to remove the correlation with each other. By adjusting the variable attenuator (VA), two parts of the light with roughly equal powers are combined through a 50/50 optical coupler and then the beat signals were detected by an electrical spectrum analyzer (ESA). As shown in Fig. 4(b), the 20 dB linewidth of TF-based BRFL pumped with NP fiber laser is around 23.3 kHz which corresponds to the 3-dB laser linewidth of 1.17 kHz. The effective fiber length for Brillouin gain in TF-based BRFL can be increased significantly, as the enhanced RS enables multiple roundtrip interactions between pump and Stokes for the sufficient Brillouin gain. Enhanced distributed RS in both directions essentially increases the dwell time or the path length for narrow linewidth. The inset of the Fig. 4(b) shows the power spectrum of the beat signal between the lasing signal and the pump light. The pedestal of the Brillouin spectrum of the tapered fiber can be seen below the narrow lasing peak of ~10.05 GHz, which is in accordance with the measured Brillouin gain spectrum in Fig. 2.

 

Fig. 4 (a) Laser output power as a function of the input pump power; (b) Electrical spectrum of the BRFL pumped by NP fiber laser (Inset, the laser spike on the pedestal of ~300 MHz).

Download Full Size | PDF

In order to verify the RS enhancement along the 5km TF, single mode fibers (SMF) (Corning, SMF28) with the lengths of 25 km and 5 km with the fiber loss of 0.2 dB/km are compared with the same unidirectional pump injection. The output spectra are monitored through an OSA (AP2043B, Apex) with the resolution of 0.04 pm. The peak powers of the 1st Stokes and the RS-reflected pump is monitored at Port #2 while the transmitting pump and the RS-reflected Stokes are monitored at Port #1. The uniform 5-km and 25-km SMF exhibits lower Brillouin thresholds than the TF, as shown in Fig. 5(a). In Fig. 5(b), the transmitting pump power in 5-km TF is lower than that in 5-km SMF owing to high RS-induced fiber loss (~0.45 dB/km). Significant enhancement of the RS-reflected Stokes and pump power in TF can be seen in Figs. 5(c) and 5(d). The RS-reflected Stokes and pump power is almost 1-2 order of magnitude higher than that in 25-km and 5-km SMF, which is critical for the lasing oscillation of the random modes.

 

Fig. 5 Power monitor at both Ports of the TF as the input pump increases (a) 1st Stokes at Port #2, (b) Transmitting pump at Port #1, (c) Reflected Stokes at Port #1, (d) Reflected pump at Port #2. The power is shown in linear scale.

Download Full Size | PDF

With the cross-spectrum method [22], the relative intensity noise (RIN) power spectral density of the TF-based BRFL were characterized using two independent photodetectors (PD450C-AC, Thorlabs) connected to an oscilloscope (DS081204B, Agilent), as shown in Fig. 3(B). The in-built RIN suppressed NP Photonics fiber laser was measured as the benchmark for the RIN measurement while the RIN of BRFL in 10-km SMF with the bi-pump scheme [15] was measured for the comparison. In Fig. 6, both the BRFLs based on TF and bi-pumped SMF show a higher RIN than the commercial NP fiber laser due to intensity fluctuation by gain competition and mode hopping among dense random modes from distributed Rayleigh feedback. However, random modes originated from the distributed RS along 5-km TF and 5-km SMF would be fewer than that in 10-km SMF owing to fewer RS-based random Fabry–Pérot resonators. Moreover, the short Brillouin-gain fiber length (~800 m) in the TF exhibits a higher selectivity of random distributed feedback modes in terms of the strong phase and polarization matching between the pump and Stokes waves. Compared to the bi-pump BRFL with 10-km SMF, both the TF-based BRFL and bi-pump BRFL with 5-km SMF alleviate the instability induced by the thermal and vibration disturbances from the external environment, providing lower RIN in low frequency domain less than 0.2 kHz. Different with over-km long uniform fibers, the effective Brillouin gain located along 800-m fiber length in the TF introduces larger mode frequency spacing and thus less mode density, which significantly alleviates the thermal-induced Brillouin gain spectra shift as well as the mode hopping-induced intensity instability. Thus, oscillations peaks in tens kHz of the RIN in bi-pump BRFL with 5 km and 10 km SMFs can be suppressed in TF-based BRFL.

 

Fig. 6 RIN comparison of the TF-based BRFL, bi-pump BRFL with 5km/10km SMF and NP fiber laser.

Download Full Size | PDF

In typical linewidth measurement based on the DSH method, hundreds of kilometer delay fiber and the frequency shifter are required to eliminate the light coherence for the improvement of the spectral resolution and the accuracy. For instance, 200-km delay fiber has to be utilized for the 1-kHz resolution linewidth measurement. The BRFL has been demonstrated for the linewidth characterization, exhibiting a robustness in improved spectral resolution [12]. Here, we proposed a simple linewidth measurement without the need of long delay line and external frequency shifter. As shown in Fig. 7, the laser under test is amplified by an EDFA and then injected into the 50/50 coupler. The 50% of the light as the Brillouin pump is launched into the 5km TF for the generation of random fiber laser. Another 50% of the pump light combines with the random laser output through another 50/50 coupler. The beat signal is then detected by a PD and an ESA.

 

Fig. 7 TF-BRFL based laser linewidth measurement setup for (a) NP fiber laser; (b) External Cavity Laser (ECL).

Download Full Size | PDF

A commercial NP fiber laser and an External Cavity Laser (ECL) are tested by utilizing the proposed setup. As shown in Fig. 8, results show that the detected signal is the beating spectrum between pump light and the Stokes lasing light centered at ~10.0355 GHz and ~10.0575 GHz, respectively. By averaging over 50 beating spectra to remove undesired noises, the 20-dB linewidths of the beat signals are measured as 55.80 kHz and 308.00 kHz, respectively, as shown in Figs. 8(a) and 8(b). Correspondingly, the linewidth of the NP photonics fiber laser can be calculated as 2.79 kHz while the ECL laser linewidth is 15.40 kHz, which agree well with the values that acquired from the conventional DSH method, as shown in the inset of Fig. 8.

 

Fig. 8 TF-BRFL based laser linewidth measurement results for (a) NP fiber laser; (b) External Cavity Laser (ECL). Insets are the corresponding DSH-based linewidth measurements.

Download Full Size | PDF

4. Conclusion

In summary, a TF-based BRFL with the unidirectional pump injection is proposed and demonstrated. Random lasing oscillation was achieved via the combination of Brillouin gain and random distributed feedback of enhanced RS along the 5-km TF. Experimental results show the linewidth as narrow as 1.17 kHz and a low relative intensity noise. The proposed scheme provides superior simplicity in the laser structure and robustness in the high-resolution linewidth measurement, which could be applicable in fields of high-resolution spectrometers, narrow linewidth laser sources and fiber sensing.