1. Introduction

Optical fibers with GeO₂ or P₂O₅ doped cores have been widely used in optical fiber communications systems. For many applications, the refractive index difference between the core and the cladding is required to be higher than that in standard communication fibers, i.e. SMF-28 fiber. Such highly doped fibers have been found to have excess optical losses, which increase as the core index increases and the core diameter decreases [1–6]. At first this excess loss was attributed to the inhomogeneous distribution of index-raising dopant in the core [2]. But later careful experimental examinations showed that the loss increase with increasing core index and decreasing core radius is much faster than that predicted from Rayleigh scattering, and this excess loss has been attributed to the scattering losses at the core-cladding interface due to fluctuations in its geometry [1,3–6]. Therefore, careful examination and characterization of the core-cladding interface is very important for accurate fiber designs.

In the 1970s, optical microscopy (OM) and scanning electron microscopy (SEM) had already been used to study the structural characteristics of preforms and optical fibers fabricated by the modified chemical vapor deposition (MCVD) method [7,8]. Results on etched fiber ends have shown some structural inhomogeneities at the core-cladding interface (see Figs. 2–4 in [7]), but the authors did not comment on their formation mechanism nor did they discuss their effect on the total losses in the final optical fibers. Later, Biriukov et al. observed astonishing irregularities at the core-cladding interface and termed them “viscous fingers”, which were thought to be caused by the mismatched viscosities between the fiber core and cladding [9]. They also argued that the presence of such “viscous fingers” deformed the core boundary and could cause the additional loss. More recently, using SEM, McNamara et al. studied silica-based optical fiber preforms in detail and also found substantial irregularities, which they called “starburst”, at the core-cladding interface [10]. They suggested a simple model stating that the “starburst” feature was caused by the buckling of the innermost fiber core layer during the preform collapse stage, because the fiber core was not hot enough (a few hundreds °C lower than the outer starting silica tube). Whatever the mechanism is, it has been shown that structural inhomogeneities at the core-cladding interface are quite common in heavily doped fiber preforms and could be the main source for the excess losses in such optical fibers.

Re-heating is a process widely used in planar waveguide fabrication as well as in post-processing optical fibers and preforms, such as annealing, crystallization and less frequent high temperature hydrogen-loading for increasing their photosensitivity. Therefore, the study of the structural changes at the core-cladding interface in re-heated optical fiber preforms and fibers is very important for exact characterization of the final fiber devices. Besides, as the fiber-drawing process can also be regarded as a post-heating process, such examination may shed further light on the formation mechanism of the observed structural inhomogeneities. In this paper, using optical microscopy, we study structural changes at the core-cladding interface in highly Ge- or P-doped MCVD silica fiber preforms after they are re-heated at high temperatures. Its validity confirmed by corresponding SEM results, optical microscopy has proven to be a very convenient way of observing such micro-structural inhomogeneities.

2. MCVD silica fiber preforms and experimental procedure

Two MCVD fiber preforms were used in this experiment. One is Ge-doped in the core (Δn=0.024); the other is P-doped in the core (Δn=0.009). The claddings of both fiber preforms were slightly doped with P and F to facilitate the preform collapse process. Experimental samples were prepared by cutting about 1 mm thick slices from the preforms and further grinding and polishing them to an optical finish. The measured refractive-index profiles of these two fiber preforms are shown in Fig. 1.

 

Fig. 1. Refractive index profiles of the two MCVD optical fiber preforms. (a) Ge-doped preform; (b) P-doped preform.

Download Full Size | PDF

The Ge-doped samples were heated at 1150°C for 3 hours in air, while the P-doped samples were heated at 1150°C for 6 hours in air. The heating furnace was then switched off and the heated samples were allowed to cool down to room temperature naturally.

3. Results and discussions

For the heated samples, surface crystallization occurred in both types of fiber preforms. There are far more surface cracks in the cladding area due to the appearance of many large crystals, presumably because only the cladding was doped with F, which can greatly reduce viscosity and melting temperature of the doped glass [11]. The crystalline phases were analyzed with X-ray diffractometry (Shimadzu XRD-6000) and determined to be α-cristobalite.

3.1 Comparison between OM and SEM results

To test the validity of the OM result, we compared it with that from SEM. For this purpose, the heat treated Ge-doped preform samples were selected, because the atomic number of Ge (32) is much higher than that of Si (14), while the atomic number of P is just 15. Under SEM, the material with higher atomic number appears brighter than the surrounding low atomic number material, thus enhancing image contrast. Therefore, under SEM, for a Ge-doped core, structural irregularity at the core-cladding interface deviating from the perfect circle could be identified and observed.

The result is shown in Fig. 2. The optical micrographs were obtained from an Olympus optical microscope with no condenser lens. The observation was done in transmission mode. For the SEM experiment, the previous sample was cleaned and etched in 5 % HF acid for 60 seconds before being gold-coated. To further enhance image contrast, we increased the electron beam spot size at a sacrifice of resolution; for features of several microns size, this should not be a problem. By comparing Figs. 2(a) and 2(b), we can see that the so-called “starburst” features, readily seen in the SEM micrograph, can be quite clearly seen in the OM micrograph. This confirms that the features seen in OM are not just artifacts caused by strain etc., but a true representation of actual disturbance to the otherwise perfect core-cladding interface.

 

Fig. 2. Ge-doped preform core showing “starburst” feature. (a) optical micrograph (transmission); (b) scanning electron micrograph (secondary electron).

Download Full Size | PDF

3.2 Structural changes at the core-cladding interface after heat-treatment

The results reported here are only for the P-doped preform, since changes in the Ge-doped preform were much less dramatic. This is possibly due to the difference in the doping materials between the two fiber preforms. Sample slices were cut from the same 2cm long preform section to ensure the structure homogeneity over such short preform length. Due to the stress at the core-cladding interface, the core usually cracked during the cutting process. First, both surfaces of one sample (the un-cracked one) were polished and observed under the optical microscope before heating. After heating, the same sample was observed again under the microscope. Reflection mode was used for both observations. The comparison result is shown in Fig. 3 where the core essentially fills the picture. In an attempt to further reduce glare, the other (cracked) sample was heated without further surface polishing to try to minimize back reflection from the bottom surface. A dark background was provided by attaching a piece of black adhesive tape on the microscope slide on which the preform sample was placed to further minimize back reflection from the surface of the microscopy glass slide. With such measures taken, the image contrast was much better and the features at the core-cladding interface were enhanced greatly, as shown in Fig. 4.

From Fig. 3, we can see that, before heating, the “starburst” features could hardly be seen under the current observation conditions; but, after heating, the “starburst” features became quite distinct. In Fig. 4, the effect is quite striking, with many anemone tentacle-like features, resembling the traces left by a shrinking viscous material. The crack, nearly across the whole core region, was left in the cutting process. From these experimental results, we can see that the structural irregularity at the core-cladding interface has been intensified by post-heating.

 

Fig. 3. Surface micrographs from optical microscope of a P-doped preform sample (a) before heating and (b) after heating.

Download Full Size | PDF

 

Fig. 4. Feature-enhanced micrograph of a heated un-polished P-doped sample.

Download Full Size | PDF

3.3 Discussions

Such an effect might be explained by considering the thermodynamic instability of core and cladding glasses at the core-cladding interface during the post-heating process. As the core and cladding are differently doped, their viscosities and thermal expansion coefficients (TEC) are all different. During the preform collapse process, the silica substrate tube already deposited with cladding and core forming layers is made hot enough and then collapsed to form a solid rod. During this process, when the temperature is still high enough, any internal stress caused by the difference in the TEC between core and cladding glasses can be released in a very short time through viscous relaxation (a few seconds above the annealing temperature, which is about 1100°C for pure silica) [12]. When the temperature of the collapsed preform rod drops below a certain value, the viscosities of the core and cladding are so high that any further stress release is no longer possible, and the residual stress will be frozen in the preform. Such residual stress is expected to be located at the core-cladding interface, and this actually has been confirmed by experimental results by different measurement techniques [13,14]. When the fiber preform is later re-heated to a high temperature for a long enough time (1150°C and several hours in our experiment), the high residual stress gradients at the core-cladding interface tend to be released through viscoelastic relaxation. This process may be further helped by bond breaking in the glass network at the core-cladding interface. In Ge-doped optical fiber, there are many oxygen-deficient Ge-Ge, Ge-Si, and Si-Si wrong bonds in the core region. Acted on by the high stress, these relatively weaker bonds could be ruptured, leading to local glass network displacement to relieve the tension. GeE’ centers, which are the supposed products of this bond breaking process, have been profiled and found to be concentrated at the core-cladding interface where the stress gradient is highest [15,16]. Such bond rupture will definitely lower the glass viscosity at the interface region, making the viscoelastic relaxation easier. P doping has a similar effect on the glass viscosity by breaking the glass network through joining one of the four surrounding O atoms by a double bond. We believe that it is this stress-releasing process through viscoelastic relaxation that has intensified the irregularity at the core-cladding interface. In our experiment, surface crystallization occurred in both heated preform samples, indicating that the glass viscosities have been low enough for structural relaxation and re-arrangement to happen.

It is interesting to note that the core-cladding interface features in Ge- and P-doped fiber preforms are different. The fingers in Ge-doped preform end in points, very similar to those reported by McNamara et al. [10]; the fingers in P-doped preform split into a multiplicity of tiny branches, very similar to those reported by Biriukov et al. [9]. The exact reason for this is unclear at the moment. Presumably it is attributed to the TEC/viscosity difference between these preforms and possible variation in the preform fabrication process.

From the above discussion, even if the wrinkling of the core layer does not happen during the preform collapse at a higher temperature, the difference in the TEC of core and cladding materials can still induce structural inhomogeneities at the core-cladding interface in the preform cooling and subsequent fiber-drawing process. The design of viscosity-matched optical fibers may be able to eliminate the instability of the core-cladding interface caused by the different viscosities between the liquid core and cladding glasses [17], the TEC mismatch between core and cladding glasses could still lead to some inhomogeneities at the core-cladding interface. Therefore, it is desirable to seek better fiber designs with both matched viscosities and TECs between the core and cladding.

4. Conclusion

We have used optical microscopy to study the structural changes at the core-cladding interfaces of highly Ge- and P-doped optical fiber preforms after post-heating treatment. Results show that the structural inhomogeneities at the core-cladding interface have been intensified by such post-heating treatment. The mechanism was discussed in terms of bonds rupture and viscous relaxation. Possible better fiber design for achieving low-loss highly doped optical fibers is also discussed.