1. Introduction

Molten-core fabrication of hybrid fibers emerged recently and has attracted increasing interest because of its excellent performance [1, 2]. Briefly, the selected materials are inserted into a glass tube to make a preform. Then, the preform is drawn into fibers via traditional optical fiber equipment. Although this approach is similar to rod-in-tube technique, its characteristic is that the core will melt into a liquid at a high temperature above the fusion-point. Going through a series of possible reactions and solidification after fast quenching in the drawing process, the core changes its original compositions and structure, showing some new useful properties. This method is well suited for the fabrication of very long fibers in comparison to other methods, such as high pressure chemical vapor deposition (HPCVD) and co-drawing laser-heated pedestal (CDLHPG). Thus far, the precursor core materials used to make hybrid optical fibers include semiconductors [3–7], sapphire [8], YAG [1, 9–11], spinel [12], baria [13] and Bi₁₂GeO₂₀ [14]. Having nominally unstable core compositions, these fibers exhibit a series of interesting properties. For example, YAG-derived fibers with high yttria and alumina contents show less cooperative up-conversion and photodarkening than conventional silica-based fibers. A breakthrough was achieved in stimulated Brillouin scattering via a sapphire-derived optical fiber [8]. Furthermore, some reports have demonstrated the feasibility of producing solar cells from silicon optical fibers [5]. More surprisingly, chemical reactions in the core melt at high temperatures could be ingeniously utilized, such as the production of crystalline silicon core fibers derived from an aluminum core preform [6].

However, if the cladding is in direct contact with core melt in the molten core approach, it will dissolve into the core inevitably, and thus some improvements should be made. Diffusion barriers, such as a CaO layer coated on the silica tube inner wall, were successfully used in silicon core fibers [5, 7, 15]. Nevertheless, it is limited in light guiding fibers due to the opaque CaO layer. Here, we present another approach to solve the diffusion problem of cladding. In this case, the cladding dissolution was mitigated by reducing reaction time. Then, the obtained fibers were compared with those drawn by a conventional molten core approach in terms of composition profiles, molecular structures, and optical properties.

2. Preparation and measurements

The precursor core material was a Nd:YAG transparent ceramic (Nd 1 at.%, Shanghai Institute of Ceramics), which was cut into a rod with a diameter of 3 mm and length of approximately 30 mm. In addition, several spherical granules of about 1.5 mm in size and small cylindrical particles measuring ϕ 0.5 × 3 mm were prepared. Next, the ceramic rod was sleeved into a thick-walled silica tube (with outer diameter of 25 mm, inner diameter of 3.5 mm and length of 150 mm), which was welded to a long thin-walled silica tube in advance, having the preform ready. Then, the preform was drawn into fibers at high temperatures above 2000°C on a custom-built fiber draw tower. This process is equivalent to the conventional molten-core fabrication reported elsewhere [16]. The obtained fibers with core diameters of 11 μm and 22 μm were labeled C11 and C22, respectively. After the core rod was exhausted, the spherical granules and small cylindrical particles were fed into the residual necked-down thick-walled silica tube one by one. Each particle was fed only after the previous one was exhausted. The technique of feeding the core materials during drawing hollow cladding was named post-feeding method. The fiber was thin at first, and became thick at the same drawing speed. In order to get consistent fiber diameters, the drawing velocity was increased appropriately according to the real-time measurement of diameter. Some fibers derived from spherical granules had outer diameters of about 230 μm and core diameters of 32 μm, and were labeled P32. Several fibers derived from small cylindrical particles had core diameters of ~16 μm, and were labeled P16. Generally, the lengths of fiber P32 and P16 were about 5 m and 2 m, respectively.

The element distribution profiles were determined using a scanning electron microscope (SEM, Magellan 400). The Raman spectra were collected using a laser Raman microscope (Renishaw inVia) equipped with a 488 nm laser. Each measurement was controlled at the same power, spot size (10 μm, with 50 × objective), and exposure time. The fiber transmission losses were measured by an Optical Spectrum Analyzer (OSA) equipped with a white light source in a cut-back approach. During the tests, a high refractive index liquid was applied to the fiber side faces to make the light propagating in claddings leak. Before the lasing experiments, both ends of the fibers were flat-cleaved and polished, and the side surfaces were wiped clean with acetone and alcohol to reduce the propagation losses in the claddings. Then those fibers were end-pumped by a multimode 808 nm continuous wave laser diode to judge their lasing performances [17].

3. Results and discussion

Figure 1 shows the end-face images and composition profiles of the fibers. Generally, the cores maintain circularity hardly in conventional molten core approach as reported elsewhere, and an application of opportune negative pressure can help maintaining circularity of the core [18]. However, all fibers derived from small cylindrical particles measuring ϕ 0.5 × 3 mm by post-feeding method maintain circularities well due to their short dwell periods staying in hot area and negligible gravities. Therefore, the post-feeding technique is helpful to maintain circularity of the core. On the other hand, when the diameter of core exceeds 22 μm, a huge difference of coefficient of thermal expansion (CTE) between YAG ceramic core (77~82 × 10⁻⁷/°C) and silica cladding (5.5 × 10⁻⁷/°C) resulted in pull cracks in the cladding. These long fibers were observed from the side under an optical microscope, confirming that the cracks ran through the fibers axially.

 

Fig. 1 Core composition profiles of (a) C11 and (b) C22 drawn by a conventional molten core approach; (c) P32 derived from 1.5 mm granules; and (d) P16 derived from cylindrical particles measuring ϕ 0.5 × 3 mm in a post-feeding molten core approach. The insets show the polished end faces of those fibers.

Download Full Size | PDF

The Nd concentrations are not shown in the figure, since they are below the detection limit of the apparatus. Evidently, the diffusion concentration of silica depends on the core diameter. J. Ballato, et al. [9] concluded that the relationship between the core sizes and diffusion concentrations appears to be linear at a range of 28 to 600 μm [9]. However, when the core size shrinks to 11 um, such as in fiber C11, the diffusion concentration rises to 73.46 wt.% sharply. Moreover, the composition profiles are very different from those of typical optical fibers following a diffusion-related trend [19]. Three main factors play a key role in the composition profiles of molten core-derived fibers: (i) a concentration gradient diffusion due to the dissolution of the silica at the core/cladding interface, (ii) a thermal convection in the heating melt, and (iii) an extrusion-induced flow at the neck-down region of the preform. It is a complex process, and thus composition profiles show either gradually varied or flat. Fiber C22 is drawn following C11 at a lower speed, having a core diameter of 22 μm. The composition profiles are consistent with other reports for this core size, and the silica concentration at the core center is approximately 55 wt.% [9, 20]. After the spherical granules and small cylindrical particles were fed into the tube, their dwell times in the hot zone were about 4 min and 2 min for fibers P32 and P16, respectively. In this period, a fair amount of silica still dissolved and diffused into the core melt. The diffusion trends are shown in Figs. 1(c) and 1(d). The composition profile of P32 is flat, which hardly occurs in YAG derived fibers within a 200 μm core size in conventional molten core approach [9]. It could be assumed that a strong convection flow homogenized the distribution of the core ingredients. However, its silica concentration is still obviously lower than that in other fibers drawn by the conventional approach for the same core size [9]. Though the core diameter of fiber P16 decreases to one-half of that of P32, the silica concentrations in their cores still maintain at ~45 wt.%. To the best of our knowledge, it is the least silica concentration for this core size among all YAG derived fibers having been reported. The post-feeding method is verified effective in mitigating the diffusion of silica, owing to the shorter dwell time in the hot area, comparing with the reaction time of at least 30 min in conventional method. Since the shape of the preform in the hot area is tapered, the smaller feedstock gives the shorter reaction time and better inhibition effect on silica diffusion in theory. However, the particle sizes used in this work are appropriate in consideration of the enough fiber length.

Figure 2 shows the normalized Raman spectra of a Nd:YAG ceramic and the above-mentioned fibers. The measurements of the silica glass Raman spectra were made on the fiber claddings. In the comparison of their original spectra, the signal intensities in the cores are an order of magnitude higher than that in the claddings after repeated testing. The same phenomenon can be observed in Yb:YAG derived fibers. It is therefore inferred that adding YAG to silica could strengthen Raman gain rather than reduce it [21]. Several interesting features of the spectra are apparent in Fig. 2. The broad peaks of Raman spectra indicate the amorphous nature of the cores in all obtained fibers, comparing with the sharp Raman peaks of Nd:YAG. Moreover, fibers P32 and P16 have near composition concentrations, therefore their Raman spectra are almost coincident.

 

Fig. 2 Normalized Raman spectra of Nd:YAG, fiber cores and silica cladding; the Raman spectrum of Nd:YAG is multiplied by 0.2. The area near 800 cm⁻¹ is expanded in the inset for clarity. The main bands (R, D₁, D₂, ω₁, ω₂, ω₃) of silica are marked. Qⁿ denotes the vibration frequencies of silica species.

Download Full Size | PDF

According to N. K. Nasikas, et al. [22], a broad and asymmetric Boson peak is located in the range of 50 to 90 cm⁻¹. Though the peak is out of measuring range, the rises of the Raman spectra intensities below 200 cm⁻¹ imply its existence. The main peaks near 440 cm⁻¹, attributable to the Si-O-Si bending vibration mode [23, 24], become broader with increasing Al and Y concentrations. This could be partly attributable to a wider dispersion of Si-O-Si bond angles in more highly compositionally modified glasses. More likely, a new broad band arising from Y-O-Si vibrations is superimposed on the spectra. Moreover, possibly residual Y-O stretching vibrations are located at 400 cm⁻¹ [22]. The peaks at 495 cm⁻¹ (D₁) and 606 cm⁻¹ (D₂) are attributable to ‘breathing’ modes of regular puckered 4-ring and planar 3-fold ring [25], respectively, which are useful in identifying the defect structures of silica. They appear to decrease considerably when the silica concentration diminishes in the core. A conceivable reason may be that the Al and Y break considerable silica networks, and impede the formation of corresponding ring structures. For the same reason, the peaks near 800 cm⁻¹, which are assigned to the Si-O-Si symmetric stretch mode, diminish accordingly [24]. Besides, the stretching mode of Al-O terminal bonds in Q²AlO₄ species appearing at 780 cm⁻¹ results in a slight red shift of this band [22].

Pure silica has two Si-O-Si asymmetric stretch modes, and their Raman peaks are located at 1068 cm⁻¹ (TO mode) and 1180 cm⁻¹ (LO mode). Following the addition of YAG, a new peak near 950 cm⁻¹ appears. It is assigned to the silica Qⁿ species participating in the glass network, with corresponding vibration frequencies of Q¹ ~890 cm⁻¹, Q² ~950 cm⁻¹ and Q³ ~1050 cm⁻¹. These Qⁿ species carry various numbers of non-bridging oxygens (NBOs) [26]. Among the obtained fibers, the increasing intensities of the peaks near 950 cm⁻¹ verify that NBOs increase as the Al and Y contents increase. YAG glass consists of AlOs (s = 4-6) and YOz (z = 6-9) polyhedra, and AlO₄ tetrahedra with two terminal oxygens are the predominant type of AlOs polyhedra. In the Y₃Al₅O₁₂-SiO₂ glass, overall the proportion of Q³ is low and about constant, and Q² dominates the high when silica contents exceed 40 mol% [22]. Therefore, in the fibers C22, P32 and P16, the formation of Q² silica species is easier than that of Q¹ and Q³, which maintain an environment similar to YAG glass with Q²-AlO₄ tetrahedra.

Figure 3(a) shows the excitation and emission spectra of the fiber P16 core, identifying three excitation bands centered at around 744 nm, 808 nm and 878 nm [27]. Under the excitation of 808 nm lasers, the emission peaks occur at 892 nm and 1062 nm. The broad peak widths indicate the amorphization of the core. The fluorescence lifetime is 292 μs, a value between that of the Nd:YAG crystal (260 μs) and Nd:Al-doped silica (490 μs) [11]. The output laser spectra of fiber P16 with a length of 1.2 m is shown in Fig. 4(a). When the pump power reaches 0.5 W, the laser starts oscillating. As expected, the laser peaks are not unique, but two peaks at 1058 nm and 1062 nm appear simultaneously. Other modes will oscillate as the out power increases, which should be attributed to a high numerical aperture of the fiber. The lasing characteristics of the four fibers are compiled in Table 1. Fiber C11 with the maximum silica content has the lowest loss of ~0.6 dB/m. However, both excessively diluted Nd concentration and small core diameter result in extremely low absorption efficiency, impairing lasing performance severely. The output power reaches only 0.45 W, with a dissatisfactory slope efficiency of about 5%. Fiber C22 with a Nd₂O₃ concentration of 0.091 mol% and a core diameter of 22.49 μm gives a maximum output laser power of about 2.19 W with a slope efficiency of 28%, which is lower than other report [10]. Fibers P32 and P16 fabricated by the post-feeding molten core approach give output powers of above 4 W with slope efficiencies of above 35%. The cracks-free cladding of fiber P16 supplies a better optical propagation than fiber P32. Although their core diameters differ greatly, the lasing performances are similar. Without doubt, the cracks and lack of polyester coatings impair the lasing performance and repeatability.

 

Fig. 3 (a) Excitation and emission spectra, (b) fluorescence decay of the fiber P16 core.

Download Full Size | PDF

 

Fig. 4 (a) Output laser spectra of fiber P16, (b) slope efficiencies of the fibers; the symbols are measure values and the lines are fitted lines.

Download Full Size | PDF

Table 1. Lasing Characteristics of Nd:YAG-Derived Fibers.

View Table

4. Conclusions

Nd:YAG ceramic derived silica fibers were made by a post-feeding molten core approach. Benefiting from the reduction of reaction time in the hot zone, the diffusion of silica into the molten core is mitigated. The lowest silica diffusion is controlled to 45.33 wt.% in fiber P16, comparing with other reports for similar core diameters. As the silica decreases, not only the Nd concentration increases greatly but also the Q² silica species form in the glass maintaining an environment similar to YAG glass. The lasing performance obtains a great enhancement. The maximum output power reaches 4.02 W with a slope efficiency of about 37% in fiber P16. The post-feeding molten core approach first presented here exhibits a considerable advantage in this aspect. Moreover, this method will show considerable potential for other materials in the fabrication of hybrid fibers.