The Brillouin scattering in PFGI-POFs observed in the previous experimental setup [27,28] was not stimulated but spontaneous, because the Brillouin threshold of PFGI-POFs was estimated to be as high as 24 W, as already described in Section 3. Consequently, the power of the reflected Stokes light was so low that we had to face the following four problems: (1) a PFGI-POF longer than several meters was required, (2) the polarization state had to be optimized, (3) averaging of the spectral data had to be conducted several tens of times, and (4) signal-to-noise (S/N) ratio of the BGS was quite low, even when (1), (2), and (3) were cleared. In order to implement practical Brillouin sensors and other systems using PFGI-POFs, these problems need to be resolved.

One solution is to employ so-called pump-probe technique. As described in Section 2, when the pump power is higher than the Brillouin threshold, a transition from SpBS to SBS occurs, leading to drastic enhancement of the Stokes light. On the other hand, when probe light at the same frequency as the Stokes light is also injected into the other end of the FUT, SBS is induced even when the power of the pump light is much lower than the Brillouin threshold, because the probe light itself acts as a seed of stimulated scattering [38]. This technique, called pump-probe technique, has been used to develop Brillouin systems with high S/N ratio [17].

The FUT used in the experiment was a 1-m PFGI-POF (Asahi Glass Co., Ltd., 2009) with a core diameter of 120 μm, NA of 0.185, a core refractive index of ~1.35, and propagation loss of ~250 dB/km at 1.55 μm (manufacturer’s specifications). The experimental setup is schematically shown in Figure 1, which is similar to that of Brillouin optical correlation-domain analysis (BOCDA) [16,17]. The light beam from a 1550-nm three-electrode laser diode (LD) was divided into two. One was used as the pump light, after being chopped with an intensity modulator (IM) for lock-in detection and being amplified with an erbium-doped fiber amplifier (EDFA). The other was used as the probe light, after passing two EDFAs, a single-sideband modulator (SSBM), and a polarization scrambler (PSCR). The SSBM was employed with a microwave generator (MG) and a proper DC bias control to suppress the carrier (pump) and the anti-Stokes component of the two first-order sidebands and to maintain a stable frequency downshift from the pump light. This frequency downshift was swept from 2.5 GHz to 3.5 GHz with a period of 300 ms to obtain the BGS of the PFGI-POF, which is observed approximately at 2.8 GHz. The suppression ratio of the other frequency components was kept higher than 25 dB, as shown in Figure 2.

The PSCR, which can modulate the polarization state at 1 MHz, was inserted to suppress the polarization-dependent fluctuations of the signal. The PFGI-POF and the silica SMFs were butt-coupled [27] with the gaps filled with index matching oil (n = 1.46) to minimize the Fresnel reflection. The Stokes light was adjusted in power with a variable optical attenuator (VOA), and converted to an electrical signal with a 125-MHz photo-detector (PD). After passing a lock-in amplifier (LI-A) with a chopping frequency of 5.018 MHz and a time constant of 10 ms, the electrical signal was observed as a BGS with an oscilloscope (OSC) synchronized with the frequency sweep of the SSBM.

Figure 3 shows the measured BGS without averaging when the pump power and the probe power were 23 dBm and 22 dBm, respectively. The power was normalized so that the peak power was 1.0. Although the PFGI-POF length was only 1 m and the polarization state was scrambled, the BGS was observed with much higher S/N ratio than that previously reported [30]. The BFS was 2.86 GHz, which is slightly higher than the previously reported value of 2.83 GHz [27]. This discrepancy seems to originate from the difference in temperature and the time constant of the LI-A which is not short enough. The 3-dB bandwidth of the BGS measured in this experiment was about 160 MHz, but further research is needed on the bandwidth because it is also dependent on the time constant of the LI-A (Note that, when the time constant was shorter than 10 ms, the BGS was distorted).

Figure 4 shows the dependence of the BGS on probe power when the pump power was fixed at 23 dBm. The probe power was reduced from 22 dBm down to 6 dBm, and averaging was conducted 30 times for the observable readout when the Stokes power was very low. As the probe power decreased, the Stokes power also decreased, which proves that this BGS is caused by the interaction between the pump light and the probe light, i.e., SBS.

The dependence of the BGS on temperature was also measured as shown in Figure 5(a). The pump power and the probe power were 23 dBm and 22 dBm, respectively, and averaging was conducted 30 times. The temperature was set to 20, 40, and 60 °C. With the increasing temperature, the BGS shifted toward lower frequency. The temperature dependence of the BFS is shown in Figure 5(b). The slope of −4.05 MHz/K is in good agreement with the previous report [30], which confirms that the BGS in a PFGI-POF observed with the pump-probe technique can be applied to high-accuracy temperature sensing.

The SBS in a PFGI-POF was successfully observed at 1.55 μm with pump-probe technique. We detected the BGS with extremely high S/N ratio, even with a 1-m PFGI-POF, scrambled polarization state, and no averaging. We also investigated the BGS dependence on probe power, which proved that the measured BGS was caused not by SpBS but by SBS. Besides, we measured the BGS dependence on temperature, showing that BGS in a PFGI-POF observed with this technique can be applied to temperature sensing with high accuracy. We are sure that use of SBS with large Stokes signal will be indispensable in developing Brillouin-based distributed temperature/strain sensing systems using PFGI-POFs.

As described in Section 3, the power of the SpBS Stokes light generated in the PFGI-POFs with 120-μm core diameter was quite low, and it needs to be enhanced for detailed investigations of the BGS [39] including the linewidth narrowing effect. Since the BGS observed with the pump-probe technique described in Section 4 is easily influenced by the time constant of lock-in detection, detailed evaluation of its linewidth was not feasible. Another approach to enhance the Stokes signal is to make use of PFGI-POFs with core diameters smaller than 120 μm. Although PFGI-POFs with 62.5-μm core diameter have become commercially available very recently (Sekisui Chemical Co., Ltd., 2011), no study has been reported on their Brillouin scattering properties. Therefore, in this Section, we characterize the BGS in PFGI-POFs with 62.5-μm core diameter.

Irrespective of the length and the core diameter, PFGI-POFs used in the experiment had NA of 0.185, a core refractive index of ~1.35, and propagation loss of ~250 dB/km at 1.55 μm. The experimental setup was basically the same as that previously reported in [27], where the BGS can be observed with high resolution by heterodyne detection. A distributed-feedback laser diode (DFB-LD) at 1,547 nm was used as a light source. One end of the PFGI-POF to be measured was optically butt-coupled to the silica SMF, and the other end was kept open.

Figure 6 shows the measured BGS of a 5-m PFGI-POF with 62.5-μm core at pump power P_in of 5, 10, 15, and 20 dBm. The polarization state optimized for P_in of 20 dBm was employed for all the measurements. The center frequency of the BGS, i.e., the BFS, was approximately 2.81 GHz, which is slightly lower than previously-reported value of 2.83 GHz [27] due to the difference in room temperature [30]. Even when P_in was as low as 5 dBm, small but clear BGS was observed.

Figure 7 shows the P_in dependences of the relative Stokes power, when 5-m PFGI-POFs with core diameters of 62.5 μm and 120 μm (Asahi Glass Co., Ltd., 2009) were employed. The reference power was set to about −63 dBm, which is the Stokes power when P_in is sufficiently low. The dependence curve of the PFGI-POF with 62.5-μm core was about 10 dB lower in pump power than that with 120-μm core, which indicates that, even at the same pump power, we can largely enhance the Stokes signal by using a PFGI-POF with a smaller core diameter.

One of the reasons for the 10-dB curve shift is the difference in Brillouin threshold power P_th. By substituting into Equations (1) and (2) the values of b = 2 [33], K = 0.667 [7], g_B = 3.09 × 10⁻¹¹ m/W [27], α = 0.056 /m (= 250 dB/km), and L = 5 m, P_th of the PFGI-POF with 120-μm core (A_eff = 209 μm² [40]) was calculated to be 97.7 W. On the other hand, P_th of the PFGI-POF with 62.5-μm core (A_eff = 108.9 μm²) was calculated to be 53.3 W, which is lower than 97.7 W by 2.63 dB. Thus, the curve shift observed in Figure 7 can be partially explained by the difference in P_th, but its amount of 10 dB is much larger than the calculated value.

Another reason for the curve shift is the improvement of optical coupling efficiency at the butt coupling when the Stokes light generated in the PFGI-POF propagates back and is injected into the SMF. Although, due to the unstable core alignment and the rough surface of the PFGI-POF, it is difficult to measure the coupling loss accurately, we confirmed that the loss with the PFGI-POF with 62.5-μm core was several dB lower than that with 120-μm core. This fact, along with the difference in internal structure designed by different manufacturers, moderately explains the 10-dB curve shift.

According to Equations (1) and (2), to employ long PFGI-POFs is another way to reduce P_th and to enhance the Stokes signal. Figure 8 shows the measured BGS of 80-m and 200-m PFGI-POFs with 62.5-μm core at P_in of 5, 10, 15, and 20 dBm. Much larger Stokes signals (~7 dB higher at P_in of 20 dBm) than those of the 5-m PFGI-POF shown in Figure 6 were observed.

There was, however, almost no difference between the BGS of the 80-m PFGI-POF and that of the 200-m PFGI-POF, only with a slight discrepancy of the BFS caused by the room-temperature difference. This means that the incident light is considerably attenuated after propagation for 80 m in the PFGI-POF. In order to estimate this effect quantitatively, the effective PFGI-POF length L_eff was plotted as a function of actual length L as shown in Figure 9, where L_eff gradually approaches 18 m (P_th ~ 13 W) with the increasing L. Thus, we proved that employing a PFGI-POF longer than ~50 m is not an effective way to enhance the Stokes signal at 1.55 μm.

According to Equations (1) and (2), as the core diameter decreases, the Brillouin threshold P_th also becomes lower. When P_in is higher than P_th, SBS is induced and consequently the Stokes signal is exponentially enhanced [7]. Here, under the rough assumption that the multimode nature, NA, refractive index, and loss do not change with core diameters, we calculated P_th of a 50-m PFGI-POF with 10-μm core diameter (hypothetical; A_eff = 17.4 μm²) to be 2.22 W (= 33.5 dBm). This value is more than one order magnitude higher than the pump power of several tens to hundreds of mW commonly used in BGS characterization in silica SMFs [39]. Even when the PFGI-POF is treated as an SMF (i.e., b = 1 [7] but A_eff becomes larger [35] in Equation (1)), this difference cannot be compensated. Thus, it seems to be difficult to reduce P_th of PFGI-POFs down to the same level of that of long silica SMFs by decreasing the core diameter, which is due to the limited effective length of 18 m associated with the high propagation loss at 1.55 μm.

Although the Brillouin linewidth of a PFGI-POF has been reported to be 105 MHz at P_in of 20 dBm in [27], detailed investigations were difficult because the Stokes power was extremely low. Here, by making use of the enhanced Stokes power, we investigated the Brillouin linewidth dependence on P_in.

Figure 10(a) shows the measured BGS of the 200-m PFGI-POF with 62.5-μm core at P_in of 14.5, 17.5, and 23.5 dBm, where the Stokes power is normalized so that the maximum power is 0 dB. Figure 10(b) shows the Brillouin linewidth dependence on P_in. From these figures, we can see that, with the increasing P_in, the 3-dB linewidth of the BGS decreases, but that its slope gradually becomes small. This behavior agrees well with the experiment and theory of the linewidth narrowing effect in silica-based SMFs [39].

The BGS properties of PFGI-POFs with 62.5-μm core diameter were investigated. The Stokes power was extremely high compared to that of a PFGI-POF with 120-μm core, and the Brillouin threshold power for 5-m PFGI-POF was estimated to be 53.3 W. It was also shown that employing a PFGI-POF longer than ~50 m is not an effective way to enhance the Stokes signal. In addition, it was theoretically found that it is difficult to reduce the Brillouin threshold power of PFGI-POFs at 1.55-μm wavelength below that of long silica SMFs even if their core diameter is sufficiently decreased to satisfy the single-mode condition. Finally, the Brillouin linewidth narrowing effect was confirmed. These results will be a good guideline for developing practical Brillouin systems using PFGI-POFs as well as for designing new PFGI-POF structures for Brillouin applications in future.