Surface crystallization was induced in Ge-doped silica glass samples from a graded-index optical fiber preform by re-heating them at 1100°C for several hours. X-ray diffraction and second-harmonic generation (SHG) microscopy have been utilized to investigate the crystalline phases formed. Experimental results indicate that the predominant crystalline phase is α-cristobalite. The cross-sectional distribution of the crystal particles has also been measured with the SHG microscopy and the result is in good agreement with that from traditional bright field light microscopy.
©2004 Optical Society of America
Surface crystallization of glasses has been studied for many years, not only because the process of surface crystallization is very similar to that occurring in glass ceramics but also because it has important applications. The original efforts were to form a thin crystallized layer on the glass surface. This layer of crystallized material has a lower thermal expansion coefficient than the glass underside and, as a result, a high compressive stress can be generated in the glass surface to give it higher mechanical strength . Recently, new applications of surface crystallized glasses have emerged, with an emphasis on inducing second-order nonlinearity (SON) in the glass surface [2, 3]. This research is particularly important, as glass has inversion symmetry and thus, in principle, has no SON at all. Such lack of SON has confined glass materials mainly to passive applications as light transmitting media such as optical fibers in optical communications systems. For applications such as optical switching, modulation and second-harmonic generation (SHG), SON is indispensable. Silica based glass has become the backbone for modern optical fiber communications systems and integrated optics. Germanium (Ge) is the most frequently used dopant to increase the refractive index of silica to form waveguides. Taking these facts into consideration, the study of surface crystallization in Ge-doped silica glass is of particular importance to the development of nonlinear optical glasses.
Crystallization will almost always be initiated at the surface by re-heating the glass if no internal nucleation catalyst is deliberately added into the glass, because the glass surface always contains dust particles from the surrounding environment and minute mechanical scratches or flaws which can serve as nucleation sites for the crystallization process. Recently, Fujiwara et al. studied the surface crystallization phenomenon in Ge-doped silica fiber preforms and thin films [4,5]. They reported that the X-ray diffraction (XRD) peak at 2θ=22 can be deconvolved into three Gaussian components, among which the peak positioned at the highest 2θ angle has the same 2θ X-ray diffraction angle as that of the (101) diffraction of α-cristobalite. The authors also used SHG microscopy  to observe the glass film after heat-treatment and found clear SHG signals from some of the crystalline particles, but did not give conclusive assignment of crystalline phases to those SHG active crystalline particles.
In this paper, we report our results of careful re-examination of this surface crystallization phenomenon in heat-treated Ge-doped silica glass. By analyzing experimental results from both XRD and SHG microscopy, we identify the predominant crystalline phase as α-cristobalite. The β→α-cristobalite phase transformation is discussed in terms of mechanical stress. The depth distribution of the surface crystallization was also characterized by the SHG microscopy with results consistent with those of conventional light microscopy.
2. Ge-doped silica samples and experimental procedure
The Ge-doped silica glass samples used were 1 mm thick slices cut from a graded-index optical fiber preform, with a maximum Ge-doping concentration of around 10 mol% in the core center, fabricated by the vapor-phase axial deposition (VAD) method. Both sides of the cut slices were mechanically ground and polished to an optical finish with the final polishing process ending with 1 µm diamond lapping film. The refractive-index profile was measured with a preform analyzer and is shown in Fig. 1. The diameter of the fiber core is around 9.2 mm. The diameter of the whole optical fiber preform is 20 mm.
To induce surface crystallization, the samples were heated in a tube furnace at 1100°C for 3 hours in air with a heating rate of 15°C per minute in the temperature ramp-up stage. The samples were then taken out of the furnace and allowed to cool down to room temperature naturally. For heating temperatures at and below 1050°C, no sign of surface crystallization was observed even for heating periods longer than 8 hours.
The crystalline phases were analyzed with X-ray diffractometry (Siemens D5000) operating at 40 kV and 40 mA using CuKα radiation. A scanning step size of 0.02° and scanning speed of 10 seconds per step were adopted for all measurements.
SHG microscopy was performed with a 830 nm Ti:Sapphire laser with a repetition rate of 80 MHz, pulse duration of around 200 fs, and an average power of about 600 mW. From the objective lens used in the experiment, the spot size of the focused beam and the optical resolution of the SHG microscopy were calculated to be 0.83 and around 0.6 µm, respectively.
3. Results and discussions
After heat treatment, the samples were first observed under an optical microscope. The glass surface was found to have many cracks. This is thought to be due to stress between the crystalline phase and surrounding glassy phase. It was also found that there are more cracks in the central area than in the outer regions of the fiber preform. As the fiber preform has a graded-index profile with highest Ge-doped concentration in the core center, the concentration of≡Ge-O-Si≡bonds, which are weaker than the corresponding≡Si-O-Si ≡ bonds, should be higher in the core center. In the core region, there will also be some ≡ Ge-Si≡defect centers, the concentration of which should also be approximately proportional to the Ge-doping concentration. The glass re-crystallization process involves atomic adjustments and glass network local rearrangement, and the existence of these weaker bonds would make this process occur more readily. It was anticipated that crystallization would occur favorably in glass areas with higher Ge-doping concentration.
3.1 XRD characterization
A typical XRD measurement result is shown in Fig. 2(a). The diffraction pattern of SiO2 α-cristobalite (PDF #39-1425 from JCPDS-International Centre for Diffraction Data) is also shown as a reference. It can be seen that the predominant crystalline phase should be α-cristobalite SiO2, with the highest peak at 2θ=22° corresponding to the reflection from (101) plane. Careful examination revealed that the main peak at around 22° is asymmetric with a shoulder on the lower-angle side, which has also been observed by others [4,5]. This indicates that there are other crystalline phases present in the crystallized layer besides the α-phase.
It is known that there are two varieties of cristobalite at atmospheric pressure; the high temperature modification β-cristobalite and its room temperature counterpart α-cristobalite [7,8]. When cooling down to room temperature, β-cristobalite can transform to α-cristobalite at 220-280°C accompanied by a volume contraction of ~5%. The lattice spacings for (101) plane of α- and (111) plane of β-cristobalite are 0.4039 nm and 0.4110 nm, respectively . The corresponding XRD 2θ angles are 22.00° and 21.62°. Therefore, the shoulder on the main peak at around 2θ=22° in the measured XRD spectrum can be attributed to the presence of β-phase in the crystalline mixture, which somehow managed to remain untransformed and retain its high temperature crystal structure to room temperature. The much lower height of this shoulder indicates that the majority of the β-phase has transformed to the α-phase in the cooling process. The peak at around 2θ=36° comprises contributions from (200) and (112) planes of α-cristobalite. The reflection from the (220) plane of β-cristobalite is also at around this angle, but it cannot be distinguished in this spectrum. To make the contribution at this angle from the β-phase discernable, we annealed some samples at a higher temperature 1150°C for 2.5 hours to try to induce more crystal particles. X-ray diffraction was also conducted on these samples. It was found that the shape of the main peak at 2θ=22° is very similar to that of samples annealed at 1100°C for 3 hours, but the peak at around 36° has now clearly split into three peaks, which are shown in Fig. 2(b). The last two peaks can be assigned to the contributions from the (200) and (112) planes of α-cristobalite with corresponding lattice spacings of 0.2487 nm and 0.2467 nm, respectively. The lattice spacing of the (220) plane of β-phase is 0.2530 nm (given by JCPDS PDF #27-605 for β-cristobalite) , larger than those of (200) and (112) planes of α-cristobalite (and therefore smaller 2θ angle), so the former can be assigned to a contribution from the (220) plane of the β-phase. It also implies the co-existence of α- and β-cristobalite in the crystallized surface.
3.2 SHG microscopy results
A typical image from the SHG microscope is shown in Fig. 3, along with the image taken with phase contrast microscopy of the same sample area for comparison. From Fig. 3(a), it can be seen that crystal particles come in various sizes, ranging from several µm to more than 30 µm. The bright areas shown in Fig. 3(b) are crystal particles that can emit frequency-doubled light through SHG under the incident laser excitation. Comparing the two images, we can see that there is an almost one-to-one correspondence between particles in the two images. It should be pointed out that some crystal particles present in the phase contrast microscopy of Fig. 3(a) may be out of view in Fig. 3(b) as they may fall outside of the focus of the fundamental laser beam. Therefore, it can be safely concluded that most, if not all, of the crystal particles on the crystallized glass surface are SHG active.
The α-cristobalite is known to have a tetragonal crystal structure with P41212 symmetry, while β-cristobalite has a cubic crystal structure with Fd3m symmetry . Macroscopically, β-cristobalite is optically isotropic and thus has inversion symmetry. This property inhibits any SHG in β-cristobalite. While for α-cristobalite, it is non-centrosymmetric and thus has SON . Therefore, under the SHG microscope, while the α-cristobalite is SHG active and visible due to the frequency-doubled light emission, the β-cristobalite would be inactive and remain invisible. By comparing Fig. 3(a) with Fig. 3(b), we can conclude that most of the crystal particles are α-cristobalite, with only a very small amount of β-cristobalite particles. This conclusion is in agreement with the XRD measurement results.
3.3 Distribution depth of the crystallization layer
Z-series scanning SHG microscopy with a step size of 1 µm was performed to obtain information on the cross-sectional distribution of crystals. The result is shown in Fig. 4, with the first image (#1) focused slightly above sample surface.
The z-direction resolution, depending on the numerical aperture of the objective lens used, was estimated to be around 2 µm. Thus the depth of the crystallized surface layer is estimated to be 15±1 µm. For comparison, we also cut the sample across the crystallized layer and used transmission bright-field light microscopy to observe the sample cross-section, shown in Fig. 5. A silicon wafer was glued to the crystallized surface before cutting to minimize possible mechanical damage to the sample surface in the cutting process. A 50× objective lens was used. A thickness of about 14 µm was estimated for the crystallized layer. We can see the result from SHG microscopy is in very good agreement with that from traditional light microscopy. The SHG microscopy has the further advantage that it is a non-destructive method, unlike the surface polishing or chemical etching [4,5].
Surface crystallization behavior in graded-index Ge-doped silica glass has been investigated. Using X-ray diffractometry and SHG microscopy, we have examined the crystalline phases of induced surface crystallization and found that most of the crystal particles are α-cristobalite, mixed with a very small amount of β-cristobalite. The cross-sectional distribution of the crystal particles was also characterized with the non-invasive SHG microscopy. The result is in good agreement with that from traditional light microscopy.
The authors would like to thank Ron Bailey of OFTC for providing the VAD optical fiber preform used in this experiment and Robert Mair, Ellie Kable and Associate Professor Guy Cox of the Electron Microscope Unit at the University of Sydney for their help in light microscopy and SHG microscopy.
References and links
1. P. M. McMillan, Glass-ceramics (Academic, 1964), Chap. 4.
2. A. Narazaki, K. Tanaka, and K. Hirao, “Optical second-order nonlinearity of transparent glass-ceramics containing BaTiO3 precipitated via surface crystallization,” J. Mater. Res. 14, 3640–3646 (1999). [CrossRef]
3. Y. Takahashi, Y. Benino, T. Fujiwara, and T. Komatsu, “Second harmonic generation in transparent surface crystallized glasses with stillwellite-type LaBGeO5,” J. Appl. Phys. 89, 5282–5287 (2001). [CrossRef]
4. T. Fujiwara, H. Nagata, Y. Benino, and T. Komatsu, “Crystallization behaviors with pre-treatments of thermal annealing and ultraviolet laser irradiation in Ge-doped SiO2 glass fiber preforms,” J. Ceram. Soc. Jpn. 111, 8–10 (2003). [CrossRef]
5. T. Fujiwara, T. Sawada, Y. Benino, T. Komatsu, M. Takahashi, T. Yoko, and J. Nishii, “Direct observation of second-harmonic generation from crystalline particles in Ge-doped SiO2 glass films,” Opt. Express 11, 1598 (2003), http://www.opticsexpress.org/abstract.cfm?URI=OPEX-11-14-1598. [CrossRef] [PubMed]
6. Y. Uesu, S. Kurimura, and Y. Yamamoto, “Optical second harmonic images of 90° domain structure in BaTiO3 and periodically inverted antiparallel domains in LiTaO3,” Appl. Phys. Lett. 66, 2165–2167 (1995). [CrossRef]
7. D. M. Hatch and S. Ghose, “The alpha-beta phase-transition in cristobalite, SiO2-symmetry analysis, domain-structure, and the dynamic nature of the beta-phase,” Phys. Chem. Miner. 17, 554–562 (1991). [CrossRef]
8. C. H. Chao and H. Y. Lu, “Stress-induced β→α-cristobalite phase transformation in (Na2O+Al2O3)-codoped silica,” Mater. Sci. Eng. A328, 267–276 (2002).
9. C. Cabrillo, F.J. Bermejo, J.M. Gibson, J.A. Johnson, D. Faccio, V. Pruneri, and P.G. Kazansky, “Thermally poled silica samples are structurally heterogeneous: Electron diffraction evidence of partial crystallization,” Appl. Phys. Lett. 78, 1991–1993 (2001). [CrossRef]