Figure 1 shows cross-sections of (a) PCF, (b) HAF, and (c) conventional SMF for comparison. We used a PCF with 60 air holes. Its structural parameter d/Λ was 0.54. Here, d and Λ are the hole diameter and hole pitch, respectively. Λ was 5.7 μm. We used a HAF with six air holes. The core diameter was 9 μm. The diameter of an inscribed circle linking the air holes c and the hole diameter d were 17.1 and 4.6 μm, respectively.

Figure 2 shows our experimental setup. The light sources used for the initiation and propagation of the fiber fuse were a Raman fiber laser and an amplified ASE (amplified spontaneous emission) operating at wavelengths of 1480 and 1550 nm, respectively. The amplified ASE was obtained using an ASE light source, band-pass filters (BPFs), and erbium-doped optical fiber amplifiers (EDFAs). Two cw lights were simultaneously multiplexed with a WDM coupler to obtain a high input power of more than 10 W, and then guided into a test fiber after passing through fiber 1 (dispersion-shifted fiber (DSF) or SMF). We initiated a fiber fuse by heating fiber 2 (DSF or SMF) with an arc discharge, and we examined its propagation in the test fiber. We observed the dynamics of fiber fuse termination at a splice point between fiber 2 and a test fiber using a high-speed camera.

When the test fiber was a PCF, the fiber fuse propagation terminated at the splice point between fiber 2 (DSF) and the PCF even when the coupled input power into the PCF was at its maximum value of 18.1 W (6.6 W at 1480 nm and 11.5 W at 1550 nm). Moreover, the fiber fuse did not propagate in the PCF when we heated it directly with an arc discharge but did not use fiber 2. Since the propagation threshold of a fiber fuse is proportional to a mode field diameter [4], the threshold power in a PCF at 1550 nm is higher than that at 1480 nm. Therefore, we can consider that the propagation threshold of a fiber fuse in this PCF is above 18.1 W at around 1550 nm. This threshold power is more than 10 times higher than that in conventional SMF.

Figure 3 consists of four successive pictures taken at intervals of 0.1 ms that show the dynamics of fiber fuse termination at a splice point between a DSF and a PCF. The input power was 18.1 W. The exposure time of the high-speed camera was 2.3 μs. As shown in Figure 3a, the optical discharge propagated in the DSF leaving periodic voids. When the front of the optical discharge reached the splice point, the front shape of the optical discharge changed as shown in Figure 3b. Since the optical discharge is a high temperature fluid [2,16], we consider that its front shape shown in Figure 3b indicates a jet of high temperature fluid that penetrates the air holes of the PCF [38]. After 0.1 ms the optical discharge became small as shown in Figure 3c, and then terminated as shown in Figure 3d. In Figure 3c, we can also see the distinguishing front part of the jet in the PCF. We believe that we can explain the mechanism of the fiber fuse termination in PCF as follows. The size of the optical discharge is assumed to be almost the same as the diameter of the melted area (D_melted) [18,20]. The melted area is caused by a fiber fuse. The diameter of an inscribed circle in the air holes in the first layer of the PCF, which corresponds to parameter c in HAF, was 8.3 μm, and it was much smaller than the measured D_melted value of 18.3 μm in the DSF. Therefore, the high temperature fluid reached the air holes, when the optical discharge penetrated the PCF. At the same time, the jet of fluid penetrated the air holes. That is, the high temperature fluid expanded. This resulted in a reduction in the fluid pressure and its temperature. Then, the fiber fuse terminated [38].

The propagation characteristics of the fiber fuse in HAF depend on the relationship between the diameter of an inscribed circle linking the air holes c and D_melted [18,20]. When c was much larger than D_melted, the fiber fuse propagated in the same way as in SMF. However, no fiber fuse propagated even at an input power of 15.6 W, when c was much smaller than D_melted [20]. In the present HAF, we observed no fiber fuse even at a maximum input power of 18.0 W (4.4 W at 1480 nm and 13.6 W at 1550 nm). Since the c value of the HAF was 17.1 μm, c was much smaller than the measured D_melted value of 24.6 μm in the SMF.

Figure 4 shows the dynamics of fiber fuse termination at a splice point between an SMF and a HAF. Figure 4a–d were obtained at intervals of 0.15 ms. The input power was 18.0 W. The exposure time of the high-speed camera was 2.3 μs. We used an SMF instead of a DSF as fiber 2 to reduce a splice loss between HAF and fiber 2 caused by the mode field mismatch. As shown in Figure 4a, the optical discharge propagated in the SMF leaving periodic voids. When the front of the optical discharge reached the splice point, the front shape of the optical discharge changed and we observed a jet that penetrated the air holes of the HAF, as shown in Figure 4b. After 0.15 ms, the optical discharge became small as shown in Figure 4c and then terminated as shown in Figure 4d.

Since we observed no fiber fuse propagation in this HAF for an input power below 18.0 W, we can consider that the propagation threshold of a fiber fuse in the HAF is above 18.0 W. Therefore, we found that the propagation threshold of a fiber fuse in PCF and HAF can exceed 18 W, which is more than 10 times that in conventional SMF. Since the fuse threshold power in conventional single-mode fibers is proportional to the mode field diameter (MFD) [9], we consider that such a high threshold could not be obtained with a method that uses the MFD expansion of conventional single-mode fibers. High threshold powers in PCF and HAF are considered to be obtained because a fiber fuse in PCF and HAF is terminated by a jet of high temperature fluid penetrating the air holes. A photonic bandgap fiber can also be expected to have a high propagation threshold for the fiber fuse phenomenon. However, we consider that a further loss reduction in the photonic bandgap fiber is desirable if we are to employ it as an optical transmission line [39].

Failures at bends in an optical fiber can be caused by light leaking from the core when the fiber is accidentally bent tightly with a high power input. The absorption of the leaked light in the fiber coating will heat it and result in failures at bends. Failures in a tightly bent fiber were examined by 180° bend of conventional SMF with a transparent UV coating in the similar way described in reference 32. We used a 1480 nm CW Raman fiber laser as a high power optical source. The power coupled into the sample fiber was 9 W. We observed two failure regimes, “R1 failure” and “R2 failure” [29]. Figure 5 shows an example of an R1 failure. The bend diameter was 11 mm. After 24 min’s exposure, the color of the coating started to change as a result of oxidization. As shown in Figure 5a, combustion occurred after 1.5 h’s exposure. The coating around the combustion point became black from further oxidization and the coating carbonized. Simultaneously the fiber deformed and the coating was burned, as shown in Figure 5b. It is considered that the fiber temperature exceeds 1100 °C in an R1 failure, which causes the fiber to deform under stress [29]. An R1 failure is serious because it causes deformation of the glass leading to complete attenuation of the light and it can also cause the coating to burn, as shown in Figure 5b.

When the bend diameter was 17 mm, we observed an R2 failure after ten hours’ exposure, as shown in Figure 6. The coating changed color as a result of oxidization and carbonization. Furthermore, part of the coating vaporized, as shown in Figure 6. Although we observed no deformation of the glass as observed with the R1 failure shown in Figure 5b, handling or local disturbance after an R2 failure will cause the fiber to fracture because the coating is vaporized or burned off [29]. The bend diameter corresponding to the R1/R2 boundary in conventional SMF was 16–17 mm when the input power was 9 W at a wavelength of 1480 nm.

It is important to note that we observed the generation of a fiber fuse when the bend diameter was 13 mm, which corresponded to an R1 failure. After two hours’ exposure, the coating burned and there was simultaneous deformation of the glass. About 1/30 second after the deformation of the glass, the fiber fuse appeared and propagated toward the light source. Therefore, we found that an R1 failure is dangerous because a fiber fuse can be generated under certain conditions in the R1 failure regime and this will seriously damage the optical transmission system. We will examine the details of the conditions under which a fiber fuse is generated in the R1 failure regime in the near future.

Two kinds of solutions have been reported for the problem of failures at bends in optical fibers. One involves improving the coating to avoid failures. The other is to use optical fibers with low bending losses to reduce the light that leaks from the core as a result of the bending.

First, there are two methods for improving the coating. One is to use a polymer coating with a lower refractive index than the cladding [31]. By using this method, we can restrict the leaked light to the glass portion and avoid heating the coating. The leaked light can also be restricted to the glass portion by employing a second glass cladding with a lower refractive index [32]. The other method involves the use of a special coating with a higher damage threshold. A special coating has been reported that realized (1) a reduction in the coating absorption and stress, and (2) an increase in the homogeneity to increase the threshold of laser induced damage at a wavelength of 2140 nm for medical applications [34]. This coating had a lower refractive index than that of silica and it had a high resistance to failures at bends at 2140 nm.

Several kinds of optical fibers with low bending losses have already been developed, namely, PCFs [25,40], HAFs [27], trench-index-profile fibers [41], and nano-engineered fibers [42]. Recent progress on fiber to the home (FTTH) has accelerated the development and standardization of bend-insensitive fibers. ITU-T has produced recommendations on bend-insensitive fibers for use in access networks [43]. Therefore, with a view to avoiding failures at bends, we should use the low bending-loss optical fibers described above with the improved coating that has high resistance to failures at bends.