Open Access
Add to BibSonomyAbstract
Glass clad semiconductor core fibers have received much attention recently for their potential utility for nonlinear optics and infrared power delivery. As these fibers have progressed, it has become evident that a greater understanding as to the dominant sources of loss is needed. This work begins that discussion by investigating intrinsic and extrinsic sources of loss in silica glass clad crystalline silicon core optical fibers. Of particular interest are, to the best of our knowledge, the first lattice-fringe images of single and poly-crystalline regions of the silicon core optical fibers as well as scattering sources. Suggested herein are methods to further reduce the presence of impurities and defects that lead to scattering and dominate optical losses.
©2012 Optical Society of America
1. Introduction
Recently, glass-clad unary and binary semiconductor core optical fibers have generated much interest due to their mid-wave infrared transparency and strong optical nonlinearities, which potentially are useful for applications in sensing, biomedicine, defense and security [1–3]. Beyond their applicability, such fibers offer unique insight into the fundamentals of materials science, namely the interplay between kinetics and thermodynamics during solidification under non-equilibrium conditions [3].
Several methods have been employed to fabricate semiconductor optical fibers including the powder-in-tube approach of Scott, et al. [4], and the high-pressure microfluidic chemical deposition approach of Sazio, et al. [5]; each having practical advantages and disadvantages. A third method for fabricating glass clad semiconductor core optical fibers is the molten core approach [6–8], in which the core composition melts at a temperature less than the draw temperature of the cladding glass. During the draw process, the semiconductor core is a fluent melt and subsequently solidifies as the fiber cools from the draw. Unlike the other approaches, the molten core approach allows for longer continuous lengths of fiber to be realized at commercially relevant draw speeds and further permits the conformation of the semiconductor core to the volume and geometry of the cladding tube [9]. This latter point will be of added value in later discussions herein relating to control of the core crystallography as a means to enhance single crystallinity and reduce grain boundary scattering.
More specifically, this work focuses on identifying and beginning to understand the principal factors that contribute to attenuation in these fibers. The sources of loss in semiconductor core fibers can arise from a variety of mechanisms that fall into two main categories: absorption and scattering. Intrinsic absorption in semiconductors mainly is caused by free electrons and resonances associated with bound electrons and ions, while extrinsic absorption mainly comes from impurities and defect centers. Extrinsic scattering is presumed to come from stress birefringence, grain boundaries, precipitates, cracks and voids due to thermal expansion mismatch, surface roughness, and diameter fluctuations. Reduction of loss is of interest, as theory suggests that ~50 dB/km is possible [3]. While initial efforts to systematically address crystallographic [9,10], impurity (specifically oxide) [7], and core/clad thermal expansion mismatch [11] have improved selected properties, including enhanced single crystallinity and reduces electrical conductivity, there has not yet been the revolutionary advance(s) that reduces measured losses to values (< 1 dB/m) where the advantages of optical fibers possessing a semiconducting core, particularly in nonlinear optics, can be most useful. Figure 1 provides a compilation of measured loss values for a variety of semiconductor optical fibers, with core size and citation specified. Regardless of the fabrication technique, furthering the fundamental understanding of loss in these novel semiconductor fibers, and subsequent work to reduce it, will help advance this growing field.
Fig. 1 Comparison of measured attenuation values and corresponding core sizes reported from various semiconductor optical fibers (silicon, Si; germanium, Ge; zinc selenide, ZnSe; amorphous hydrogenated silicon, a-H:Si. Abbreviations: Collaboration between the Pennsylvania State University and the Optoelectronics Research Center at the University of Southampton [PSU/ORC], Clemson University [Clemson]. Citations noted follow the following format (Journal name abbreviation volume, starting page number, year of publication), where “OE” is Optics Express, “JACS” is the Journal of the American Chemical Society, “APL” is Applied Physics Letters, “AM” is Advanced Materials, “JCG” is the Journal of Crystal Growth, “OME” is Optical Materials Express, and “OL” is Optics Letters.
2. Experimental section
As discussed in greater detail previously [12], a 3 mm in diameter by 30 mm in length silicon rod was sleeved into a 3.2 mm inner diameter by 30 mm outer diameter by 1000 mm in length high purity silica cladding tube and drawn into fiber with a preform feed of 4 mm/min. The cladding tube was pre-drawn to collapse and seal one end in order to contain the silicon melt during the fiber draw. A molten core approach was used, whereby the semiconductor core is a fluent melt at the draw temperature of the silica cladding glass [7]. The fibers discussed in this work were drawn (Clemson University) at a speed of about 12 meters/minute at a temperature of 1925°C, with a target fiber diameter of 1 mm and core size of approximately 100 μm.In order to better determine the effect of core/clad roughness on attenuation and crystallography, and to more easily image the crystallographic microstructure of the core, samples of the silica-clad silicon-core fiber were cleaved and acid etched in HF acid (50% in deionized water) and stirred at room temperature overnight. Several pieces of the etched silicon core were mounted in resin and polished flat to a 1-micron finish in order to examine a cross-section of the core.
Microstructural and elemental analysis was performed under high vacuum secondary electron (SE) mode using a Hitachi SU6600 analytical variable pressure field emission scanning electron microscope (FE-SEM). Electron backscatter diffraction (EBSD) analysis was carried out using Oxford Instruments HKL Channel 5 software. Lattice-fringe imaging of the silicon core was obtained using a Hitachi H-9500 transmission electron microscope (TEM). Atomic force microscopy (AFM) was carried out using a Veeco Dimension 3100 Nanoman AFM. Data was analyzed using Gwyddion freeware. The infrared transmission measurement was made using an equivalent set-up to that employed in previous work on fibers synthesized at Clemson [7]. Briefly, a single-mode tapered fiber, mounted on a three dimensional translation stage, was used to couple light from a tunable laser (operating at a wavelength about 1.55 μm) into the silicon fiber sample. The output end of the fiber is placed in front of a near-infrared objective (50 ×) to form a magnified near field image focused onto the infrared camera. A second optical path includes a white light source, a beam-splitter, a beam sampler, and a visible CCD camera and is used to image the output of this sample.
3. Results and discussion
Figure 2 provides the scanning electron microscope (SEM) images on the silicon core samples. As can be seen, etching away the cladding glass provides direct access to the silicon core for analysis. These images were made in order to examine the core surface for cracks and microstructural features.
Electron backscatter diffraction (EBSD) and what the authors believe to be the first direct lattice-fringe images of the silicon core are shown in Figs. 3 and 4 . The Fast Fourier Transform (FFT) is shown inset in Fig. 4(a). Clearly resolved are highly crystalline regions, as well as regions of polycrystallinity. Grain orientations were determined using the Kikuchi line geometry patterns, which correlate to crystallographic Miller indices [13]. While several different grain boundaries are present in the EBSD and transmission electron microscopy (TEM) images, regions of well-defined local single crystallinity are also observed. Since silicon is a cubic crystal, it is optically isotropic and scattering from grains of differing crystallographic orientation should not occur. However, differential density of gap states at grain boundaries can influence absorption [14,15] as well as the electron density which can influence scattering. The distribution of grain orientations can also yield a distribution of stresses, which can promote additional scattering via stress-optics effects [16]. Lastly, grain boundaries can serve as sinks for impurities to reside and, hence, can have locally different compositions which also could facilitate scattering. Towards this end, more highly single-crystalline fibers should contribute to a reduction in such perturbations that could help reduce loss in future fibers.
Fig. 3 EBSD image of silicon core grain boundaries and crystallographic planes. The black scale marker on the bottom of Fig. 3(a) is 500 μm.
Fig. 4 Transmission electron microscopy (TEM) images of etched silicon cores. Regions of well-defined local single crystallinity are observed though the samples are polycrystalline overall. The scale marker is 10 nm for Fig. 4(a) and 5 nm for Fig. 4(b) and the noted lattice spacing in Fig. 4(a) is 0.3326 nm (3.3326 Å).
A representative inter-planar lattice spacing, obtained directly from the TEM micrograph as shown in Fig. 4(a), was 0.3326 nm (3.3326 Å). Since a = 5.43Å for silicon, the only combination of (hkl) in a cubic crystal for this range is the (111) plane. These results, in combination with the silicon core grain boundary orientations shown in the EBSD data, corroborate well with known data that silicon has octahedral cleavage along the {111} faces. SEM micrographs (not shown) exhibit fissures emanating from the core interface with the cladding at an angle of 65°, corresponding to the {111} face, suggesting that they could be a stress-induced cleavage originating at the core/clad interface due to the differential thermal expansion between the core crystalline (silicon) semiconductor and cladding (silica) glass. Most cracks do not propagate all the way across the core, so there may either be some self-healing during solidification or the stresses are sufficiently small at these core sizes to not catastrophically fail. Though not definitive, bright spots in the TEM images (and IR scattering images shown later) could be oxide precipitates. Oxygen is known to dissolve into the core as the molten silicon attacks the silica glass cladding during the elevated temperature draw process. The oxygen concentration, measured macroscopically via energy dispersive spectroscopy (EDX) and discussed previously, is believed to yield nanoscale oxide precipitates that lead to scattering of light propagating through the fiber [3,7,12].Atomic force microscopy (AFM) was performed on the etched silicon core pieces; images are shown in Fig. 5 . Additionally, a cladding glass tube was sectioned, and the inner surface analyzed in order to determine if there were any effects of the cladding glass roughness on the post-drawn core. Measured root-mean-square (rms) roughness of the silicon core was approximately 2 nm. The rms values for the silica cladding tube inner surface ranged from 0.3 to 0.6nm with a few spots of higher roughness (rms ~2 nm) possibly from incomplete cleansing of the tube after sectioning. This suggests that some of the surface roughness of the silicon core can be attributed to it conforming to the geometry of the silica tube during the molten core draw process. Previous analysis of surface roughness in silicon core optical fibers fabricated using chemical-vapor deposition processes [17] yielded rms values around 0.14nm. At these levels of roughness less than 4 nm [17], surface scattering does not contribute significantly to loss in transmission (< 0.1 dB/cm) though may be an issue worth revisiting in the future should < dB/m attenuation levels be desired.In order to measure the infrared (IR) transmission and evaluate sources of scattering and loss, a single-mode tapered fiber was used to couple light into the sample from a tunable laser (1.55 µm wavelength). Images taken using an infrared camera are shown in Fig. 6 . Specifically, Fig. 6(a) shows an image overlaid with the incident fiber taper, bright spot where the pump light exits the taper, and the illuminated front face of the silicon core. Figure 6(b) provides an infrared image of the output end of the silicon core showing a relatively uniform distribution of light across the diameter. Figure 6(c) shows light scattering from the side along the length of the silicon core. Clearly observed are bright bands which indicate a longitudinal heterogeneity. Whether these bands are associated with perturbations during the longitudinal solidification of the molten core, scattering from grain boundaries and possible oxide precipitates that would naturally segregate there, or small cracks is presently under further evaluation. Regardless, it is clear that side-directed scattering is easily observed along the length of the silicon core and likely causes much of the losses measured to date.
Fig. 6 (a) Infrared (IR) side-view image of the input end with regions noted. (b) IR image of output facet when IR light is coupled to the other end. (c) Longitudinal IR image of the silicon core. In all cases, the core-only (i.e., glass cladding etched away) samples were used.
At present, based mainly on the infrared images, the dominant loss in semiconductor core fibers appears to be scattering; either from cracks originating from the thermal expansion mismatch between core and clad, perturbations in the solidification leading to longitudinal variations in the fiber, and/or oxide precipitates that likely partition to the grain boundaries. Obviously, progressively higher purity source materials could only be of additional benefit once the sources of scattering are more definitively identified and mitigated. With respect to each scatting source, presented here are comments on what has been done to date and what additional actions might be further taken.
- (a) Cracks and stress-optic influences due to core/clad differential thermal expansion
- As the molten semiconductor cools from the draw temperature to its melting point, there is shrinkage that can lead to void formation. Once the solidification occurs, there will be either residual strain or, if this strain exceeds the fracture limit, cracking can occur. Transmission through a series of cracks can be modeled as multiple frustrated total internal reflections which can reduce transmission very quickly [18]. Recent work has focused on cladding glasses with compositions specifically tailored to match the thermal expansion of the core phase [11]. Stresses also are lessened at smaller core sizes, so methods to make smaller fibers, whether through tapering or drawing at reduced temperatures, also is worthy of further consideration.
- (b) Perturbations in the solidification leading to longitudinal variations in the fiber
- Whether from inevitable (though small) perturbations in draw speed or diameter or presently unknown flow dynamics of the molten core prior to and during solidification, Fig. 6(c) quite clearly shows longitudinal striations and side-scattering that originates from them. The most reasonable approach to lessening such striations is annealing, which has been tried on silicon optical fibers using both (photo)thermal [10] and laser annealing [19] processes. In both cases, enhancements in the quality of the crystallinity and properties were observed though deeper understanding of the annealing mechanisms that influence crystallographic reorientation is needed. Regardless, annealing – perhaps in situ during the draw – has shown promise and should be further explored and optimized.
- (c) Polycrystallinity and oxide precipitates
- As has been well characterized, the as-drawn semiconductor core fibers are polycrystalline with single crystalline grain sizes on the order of several millimeters to centimeters [20]. Since grain boundaries are, by definition, regions of disorder between crystals of differing orientations, they tend to be regions where impurities can segregate and can have differing dielectric, hence refractive, properties. While the aforementioned lengths over which single crystallinity is developed during the solidification of the fiber might be useful for selected applications, such as mid-IR light sources based on Raman shifting [21], generally, longer lengths would be preferred. Towards this end, the previously discussed annealing methods have proven effective in enhancing the degree of single crystallinity. In fact, the longest single crystalline semiconductor optical fiber (> 1 cm) was realized following photo-thermal annealing [10]. For completeness, tapering has also proven effective in controlling single crystallinity [22] as has the use of a square core rather than the more conventional round core [9]. There does not seem to be any one process that best enhances single crystallinity so future advances might further consider annealing, tapering, and fiber geometry (which might include some residual stresses to facilitate stress-induced crystallization).
- With respect to oxide precipitates, it has been conjectured since the initial molten core silicon optical fiber [12] that the measurable presence of oxygen in the core is associated with oxide precipitates that arise from dissolution of the cladding glass by the core melt during the high temperature processing of the fiber. There are two principal methods to reduce the level of cladding glass dissolution: (a) reduce the processing temperature since dissolution is a thermally-activated process or (b) reactively remove (i.e., getter) the oxide phases through chemical means. With respect to reducing the draw temperature, the aforementioned designer cladding glasses [11] can be selected such that they, in addition to being expansion-matched, draw at a temperature just slightly above the melting point of the core semiconductor phase. This would then constitute the lowest temperature that a molten-core-derived optical fiber can be fabricated and has shown some initial success [11]; though more development is necessary. With respect to the use of a reactive chemistry, the molten nature of the core phase during the fiber draw process enables as convenient and versatile route to control the core composition and crystallography. Preliminary efforts to reduce the formation of oxide precipitates have involved using in situ reactions by adding silicon carbide (SiC) to the silicon core, in sufficient quantity to react with all of the diffused oxygen. As an initial result, the amount of oxide has been decreased to nearly zero and optical quality of the fibers is improved [7].
While there are applications for multimode semiconductor core optical fibers (e.g., IR power delivery), single mode analogs would be far more interesting to the community at large, particularly as relates to these fibers’ use in nonlinear optics. Accordingly, for completeness, it is appropriate to discuss the technological limits of the fabrication processes employed. Given the extraordinarily high refractive index contrast between the semiconductor core and glass cladding, which can approach about 2.5 in the case of InSb core and borosilicate clad (larger than air/silica used in many photonic crystal fiber designs), core sizes for single mode operation are in the range of 500 nm or less [22] in the infrared spectral region. Particularly at these extremely small core sizes, issues of diffusion are exacerbated. As a case in point, Fig. 2(b) in [3] shows that a (silicon) core will completely oxidize during the molten core fiber draw process for core sizes below a few micrometers; i.e., while the fiber is still multi-moded. Such core/clad reactivity is a limitation of the molten core process provided additional measures are not taken to counter the chemical processes. However, while dissolution and diffusion are especially problematic at small core sizes, tapering results indicate that a reduced core size facilitates both a greater degree of single crystallinity and alignment of the principal crystallographic axis, relative to the fiber longitudinal axis, of the core phase [22]. There is, therefore, additional value in reduced core sizes beyond simply achieving a single mode crystalline semiconductor core fiber.
Though not yet attempted at single-mode core sizes, the aforementioned successes in employing melt-phase chemistry to reactively getter oxide phases that dissolve into the molten core from the cladding glass during fiber draw [7] is the most obvious initial path forward. One disadvantage of the molten core approach is that the process would seem unable to permit a multi-semiconductor layer structure, as has recently been done using high pressure CVD approaches [23,24]. Conceivably, semiconductors of reasonably similar refractive index could be deposited to create a core and inner cladding with sufficiently small numerical aperture so as to yield single mode operation at reasonable core sizes. Both molten core and CVD methods would have the previously discussed issues relating to optimized loss, crystallinity, and crystallography in common. It is, therefore, the authors’ hope that this work is more broadly applicable to the field than just for those fibers fabricated using the molten core approach.
4. Conclusions
The principal sources of loss in present silicon core fibers were discussed and selected measurements reported to further identify whether cracks, grain boundaries (polycrystallinity), surface roughness, longitudinal perturbations, or oxide precipitates dominate. TEM and EBSD results showed that while regions of localized single crystallinity are present in the core, overall polycrystallinity could possibly lead to scattering due to stress-optics effects and absorption due to differential densities of gap states. Surface roughness was shown, via AFM measurements, to be relatively low, commensurate with previous measurements, and is not likely to contribute significantly to scattering. Potential evidence for precipitates was shown in infrared transmission images. It is likely that future progress towards the reduction in loss of these novel fibers will require a combination of (a) tailored cladding glasses to negate expansion mismatches and reduce cladding dissolution during fiber fabrication, (b) in situ melt-phase chemical reactions to reactively getter oxide and other impurities/contaminants, (c) greater single crystallinity in order to reduce grain boundary scattering though annealing, tapering, or core geometric control. Methods for reducing loss in semiconductor core optical fibers, as discussed herein, should lead to the development of higher quality optical fibers to expand the growing field of silicon photonic waveguides.
References and links
1. D. Graham-Rowe, “Fibres get functional,” Nat. Photonics 5(2), 66–67 (2011). [CrossRef]
2. B. Jalali, V. Raghunathan, D. Dimitropoulos, and O. Boyraz, “Raman-based silicon photonics,” IEEE J. Sel. Top. Quantum Electron. 12(3), 412–421 (2006). [CrossRef]
3. J. Ballato, T. Hawkins, P. Foy, B. Yazgan-Kokuoz, C. McMillen, L. Burka, S. Morris, R. Stolen, and R. Rice, “Advancements in semiconductor core optical fiber,” Opt. Fiber Technol. 16(6), 399–408 (2010). [CrossRef]
4. B. Scott, K. Wang, V. Caluori, and G. Pickrell, “Fabrication of silicon optical fiber,” Opt. Eng. 48(10), 100501 (2009). [CrossRef]
5. P. J. A. Sazio, A. Amezcua-Correa, C. E. Finlayson, J. R. Hayes, T. J. Scheidemantel, N. F. Baril, B. R. Jackson, D. J. Won, F. Zhang, E. R. Margine, V. Gopalan, V. H. Crespi, and J. V. Badding, “Microstructured optical fibers as high-pressure microfluidic reactors,” Science 311(5767), 1583–1586 (2006). [CrossRef] [PubMed]
6. J. Ballato, T. Hawkins, P. Foy, S. Morris, N. K. Hon, B. Jalali, and R. Rice, “Silica-clad crystalline germanium core optical fibers,” Opt. Lett. 36(5), 687–688 (2011). [CrossRef] [PubMed]
7. S. Morris, T. Hawkins, P. Foy, C. McMillen, J. Fan, L. Zhu, R. Stolen, R. Rice, and J. Ballato, “Reactive molten core fabrication of silicon optical fiber,” Opt. Mater. Express 1(6), 1141–1149 (2011). [CrossRef]
8. J. Ballato, T. Hawkins, P. Foy, C. McMillen, L. Burka, J. Reppert, R. Podila, A. M. Rao, and R. R. Rice, “Binary III-V semiconductor core optical fiber,” Opt. Express 18(5), 4972–4979 (2010). [CrossRef] [PubMed]
9. S. Morris, C. McMillen, T. Hawkins, P. Foy, R. Stolen, J. Ballato, and R. Rice, “The influence of core geometry on the crystallography of silicon optical fiber,” J. Cryst. Growth 352(1), 53–58 (2012). [CrossRef]
10. N. Gupta, C. McMillen, R. Singh, R. Podila, A. M. Rao, T. Hawkins, P. Foy, S. Morris, R. Rice, K. F. Poole, L. Zhu, and J. Ballato, “Annealing of silicon optical fibers,” J. Appl. Phys. 110(9), 093107 (2011). [CrossRef]
11. S. Morris, T. Hawkins, P. Foy, J. Ballato, S. W. Martin, and R. Rice, “Cladding glass development for semiconductor core optical fibers,” Int. J. Appl. Glass. Sci. 3(2), 144–153 (2012). [CrossRef]
12. J. Ballato, T. Hawkins, P. Foy, R. Stolen, B. Kokuoz, M. Ellison, C. McMillen, J. Reppert, A. M. Rao, M. Daw, S. R. Sharma, R. Shori, O. Stafsudd, R. R. Rice, and D. R. Powers, “Silicon optical fiber,” Opt. Express 16(23), 18675–18683 (2008). [CrossRef] [PubMed]
13. R. A. Schwarzer, “Automated crystal lattice orientation mapping using a computer-controlled SEM,” Micron 28(3), 249–265 (1997). [CrossRef]
14. W. B. Jackson, N. M. Johnson, and D. K. Biegelsen, “Density of gap states of silicon grain boundaries determined by optical absorption,” Appl. Phys. Lett. 43(2), 195–197 (1983). [CrossRef]
15. P. Nagpal and V. I. Klimov, “Role of mid-gap states in charge transport and photoconductivity in semiconductor nanocrystal films,” Nat Commun 2, 486 (2011). [CrossRef] [PubMed]
16. W. N. Ye, D. X. Xu, S. Janz, P. Cheben, M. J. Picard, B. Lamontagne, and N. G. Tarr, “Birefringence control using stress engineering in silicon-on-insulator (SOI) waveguides,” J. Lightwave Technol. 23(3), 1308–1318 (2005). [CrossRef]
17. N. Healy, L. Lagonigro, J. R. Sparks, S. Boden, P. J. A. Sazio, J. V. Badding, and A. C. Peacock, “Polycrystalline silicon optical fibers with atomically smooth surfaces,” Opt. Lett. 36(13), 2480–2482 (2011). [CrossRef] [PubMed]
18. S. Zhu, A. W. Yu, D. Hawley, and R. Roy, “Frustrated total internal reflection: a demonstration and review,” Am. J. Phys. 54(7), 601–606 (1986). [CrossRef]
19. N. Healy, S. Mailis, T. D. Day, P. J. Sazio, J. V. Badding, and A. C. Peacock, “Laser annealing of amorphous silicon core optical fibers,” in Specialty Optical Fibers, OSA Technical Digest (Optical Society of America, 2012), paper STu1D.1.
20. C. McMillen, T. Hawkins, P. Foy, D. Mulwee, J. Kolis, R. Stolen, R. Rice, and J. Ballato, “On crystallographic orientation in crystal core optical fibers,” Opt. Mater. 32(9), 862–867 (2010). [CrossRef]
21. P. Wang, C. Charlton, T. Lee, R. Ismaeel, T. Hawkins, Y. Semenova, L. Bo, Q. Wu, C. McDonagh, G. Farrell, J. Ballato, and G. Brambilla, “Mid-infrared Raman sources using spontaneous Raman scattering in germanium core optical fibers” (submitted to Appl. Phys. Lett. ).
22. C. McMillen, G. Brambilla, S. Morris, T. Hawkins, P. Foy, N. Broderick, R. Rice, and J. Ballato, “On crystallographic orientation in crystal core optical fibers II: effects of tapering,” Opt. Mater. (to be published), doi:. [CrossRef]
23. R. He, P. Sazio, A. Peacock, N. Healy, J. Sparks, M. Krishnamurthi, V. Gopalan, and J. Badding, “Integration of gigahertz-bandwidth semiconductor devices inside microstructured optical fibres,” Nat. Photonics 6(3), 174–179 (2012). [CrossRef]
24. M. Krishnamurthi, E. Barnes, J. Sparks, R. He, N. Baril, P. Sazio, J. Badding, and V. Gopalan, “A magnifying fiber element with an array of sub-wavelength Ge/ZnSe pixel waveguides for infrared imaging,” Appl. Phys. Lett. 101(2), 021108 (2012). [CrossRef]
