Introduction

There has been growing interest in novel optical fibers that contain semiconducting materials in their cores. Notable successes include InP-containing fibers where InP nanoparticles were incorporated into the porous soot of a modified-chemical-vapor-deposition-derived silica preform prior to fiber draw [1]. Other recent efforts have focused on the fabrication of optical fibers where the entire core is comprised of a semiconductor material [2,3]. In that work, a molten core approach was employed where the cladding glass is chosen such that it draws at a temperature above the melting point of the core composition. While this is a variation on an older theme [4], it works well for simple systems, e.g., unary semiconductors, but not well for more complex crystallographies [5]. The purpose of this work is to determine whether the molten core approach can be extended to binary semiconductors.

There are many potentially interesting device applications for optical fibers with crystalline semiconductor cores, primarily in the area of nonlinear optics because of the large optical nonlinearities that are observed for photon energies approaching that of the bandgap. The unary semiconductors exhibit a large χ⁽³⁾ but no χ⁽²⁾ because of the symmetry of their crystal class; hence they can be used for high gain Raman amplifiers and wavelength converters, but not optical parametric amplifiers (OPA), frequency doublers or parametric oscillators (OPO). These devices, which require a large χ⁽²⁾ only available in crystal classes lacking a center of symmetry, are widely used in nonlinear optics applications. To date, such devices have exclusively used bulk single crystals or planar waveguides fabricated on single crystal substrates, which are complex and costly to produce. Crystalline core fibers have two important advantages for nonlinear optics applications: first, the optical power is confined in a waveguide to maintain high nonlinear coupling between the interacting fields over an arbitrary interaction length, and second, the interaction length is not limited by the size of nonlinear crystals that can be prepared by conventional crystal growth methods. Many binary semiconductors, such as GaAs and InP, lack centers of symmetry in the unit cell and exhibit large values for χ⁽²⁾, and hence could be attractive as potential nonlinear fiber core materials for the infrared and near infrared. Approaches for managing phase mismatch, as required for χ⁽²⁾ nonlinear processes, are beyond the scope of this work but are potentially feasible.

In this work, InSb was selected as a candidate proof-of-concept core material because it is cubic (zinc blende/sphalerite structure type; space group F-43m; #216), hence is optically isotropic. InSb also has one of the lowest melting points for any cubic compound semiconductor at 527 °C [6], which is an important consideration for molten-core-derived optical fibers. Further, InSb possesses unusually high electron mobilities and Hall constant [7] and the optical properties of pure and doped InSb have been widely investigated from the infrared to terahertz frequencies [8–11]. It has a bandgap at room temperature of about 0.17 eV (i.e., electronic absorption edge of about 7.3 μm) and is transparent to about 30 μm at room temperature [9]. InSb (and InAs) also are known for having large optical nonlinearities, particularly three-wave mixing [12], due to non-parabolicity in its band-structure [13,14]. Lastly, it is worth noting that InSb, like all of the III-V and II-VI binaries (in classes 6mm or −43m) are piezoelectric and piezo-optic. The binaries are semiconducting and so the piezo-related effects are usually short-circuited, unless the compounds are doped to make them semi-insulating.

Experimental procedures

An n-type undoped InSb wafer of unspecified purity (MTI Corporation, Richmond, CA) was diced into sections that were approximately three millimeters in width and sleeved into the center of an FU-2 phosphate glass tube (Kigre Incorporated, Hilton Head Island, SC). Fibers were drawn at Clemson University using the Heathway draw tower at approximately 700 °C and a draw speed of about 0.5 m/min. At this temperature, the InSb melts and the fluent liquid takes on the shape of the glass cladding tube. Given the small sample size afforded by the InSb wafer only approximately 2 meters of 1 mm diameter fiber was drawn yielding a core size of about 230 μm. However, the generalized molten core approach, coupled with fiber draw process, is readily scalable to long lengths [3]. The choice of a clad-to-core ratio of about 5, or greater, has proven appropriate such that a sufficient amount of the softened cladding glass is present in order to physically contain the weight of the molten core at the draw temperature. It also is useful to draw first through a section of bulk cladding glass so that the optimum draw conditions can be reached and equilibriated prior to the preform reaching the section containing the semiconducting core.

Electron microscopy was performed using a Hitachi 3400N scanning electron microscope operating at 20 kV and 10 mm working distance under variable pressure. Elemental analysis was carried out using energy dispersive x-ray spectroscopy to determine the distribution of elements across the core/clad interface. Elemental compositions were measured at several locations traversing the core in approximately 1 μm increments. Micro-Raman measurements were carried out using a Dilor XY triple-monochromator equipped with a thermoelectric cooled charged-coupled device (CCD) detector. The backscattering configuration using a 50 × objective and a Kr-laser (647.1 nm excitation) laser was used. The spectrometer had a typical resolution of 2.0 cm⁻¹ in a backscattering geometry. Fourier transform infrared spectroscopy (FTIR) was performed on the as-purchased starting InSb wafer using a Nicolet Magna IR 560. A total of 32 scans were averaged after running a equivalent background scan for subsequent subtraction. All measurements were made at room temperature.

Phase purity and crystallinity were determined using powder x-ray diffraction. Samples of the fiber were ground and analyzed using a Scintag XDS 2000-2 powder diffractometer with Cu K_α radiation (λ = 1.5418 Å) and a solid-state Ge detector. Diffraction patterns were collected in 0.03° steps (1.5 seconds per step) from 5 to 65° in 2-theta. Single crystal x-ray diffraction was used to examine the single crystallinity and crystallographic orientation of a 1 cm length of fiber in 1 mm increments along the fiber length by a procedure described in detail elsewhere [15]. A Rigaku AFC8S diffractometer with MoK_α radiation collimated to a diameter of 0.5 mm and a Mercury CCD area detector was used to collect 25 screening images for subsequent unit cell determination and orientation analysis of the longitudinally oriented fiber. Axial photographs (not shown) collected by ω-scans from −30 to +30° were used to determine the crystallographic orientation with respect to the fiber axis and inspect the reflection profiles for polycrystalline contributions.

Results and discussion

An electron micrograph of the resulting fiber is shown in Fig. 1 . The core is circular in cross-section and there is a strong contrast between the semiconducting core and the phosphate glass cladding. Further, the interface is well-defined and does not show signs of bubbles or striations resulting from the melting of the core during fiber draw.

 

Fig. 1 Electron micrograph image of phosphate glass clad, InSb core optical fiber.

Download Full Size | PDF

Figure 2 provides the x-ray diffraction spectrum from the InSb starting material (wafer) and from the core of the fiber. Clearly observed are peaks well-indexed to the crystallographic reflections of InSb indicating phase purity of the core that solidifies from the melt as the fiber cools. As noted above, while this seems to work well for elemental semiconductors such as Si and Ge [2,3], it has not worked for more complex crystals such as YAG, which yield amorphous cores either when drawn [5] or grown [16]. The narrow linewidth of the x-ray reflections in Fig. 2 further verifies the high degree of crystallinity in these fibers and validates the ability to draw binary semiconductors into fiber at reasonable speeds.

 

Fig. 2 X-ray diffraction spectra for InSb starting material and fiber core. Crystallographic indices also are shown.

Download Full Size | PDF

The single crystal measurements carried out on the fiber suggested that the localized crystalline nature (given the 0.5 mm x-ray beam diameter) of the core is largely polycrystalline where a singular crystallographic orientation could not be resolved along the length. The InSb fiber studied here was found to be 90% polycrystalline over the length of the sample with the longest continuous length of polycrystalline material measuring at least 8 mm; i.e., over 8 mm length a single orientation could not be resolved. This is in contrast to previous findings on a Ge core semiconductor fiber which was ~15% polycrystalline along its length and exhibited a maximum length of 3 mm before a single crystallographic orientation could be resolved [15]. Polycrystallinity in GaSb and InSb crystals can occur from a combination of (1) a “wall effect” that promotes the formation of a greater number of crystallites in samples grown in smaller ampoules and (2) non-uniform heat conduction resulting from fast cooling rates [17,18]. Thus, in the present study it is not so surprising to see the InSb core exhibit localized polycrystalline behavior given the confinements of the cladding glass combined with the rapid rate of quenching in the fiber draw process.

Figure 3 provides the elemental profile across the core/clad interface of the fiber. Negative position values represent the cladding region whereas positive values represent the core region. The cladding is an aluminum phosphate base composition and the minor (< 1 percent) glass modifying components have been removed from the graph in order to enhance the clarity of the trends. Error bars representing a 95% confidence interval are included in the Fig. 3 data but might not be resolvable since they are smaller than the data symbols. It can be seen that there is some diffusion of In and Sb into the cladding region and, conversely, diffusion of oxygen and phosphorus into the core region.

 

Fig. 3 Elemental profile of selected core and clad constituents as a function of distance from the core/clad interface. Minor cladding glass constituents have been omitted for clarity.

Download Full Size | PDF

The level of oxygen in this InSb core fiber is less than that found in Si core fiber drawn at 1950 °C but is more than that in Ge core fiber drawn at about 1000 °C [2,3]. To first order, this would imply that the thermodynamic driving force for the oxidation of In and Sb are greater than that for Ge but less than it is for Si. However, much like the case of the Si core fiber, the presence of oxygen at what certainly is a non-equilibrium concentration does not modify the measured crystallinity implying the likelihood that oxide precipitates are amorphous. It is worth noting that such precipitates could act as defects or secondary nucleating sites that limit long range coherent crystal growth thereby contributing to the higher degree of polycrystallinity found in these fibers in comparison to the lower oxygen content Ge core fibers [3].

Spontaneous Raman scattering spectra are presented in Fig. 4 . The characteristic zincblende longitudinal optical phonon (LO) mode is observed at ~190 cm⁻¹ in both the fiber core and the InSb starting material. The transverse optical phonon (TO) mode is forbidden in [100] InSb due to the selection rules for Raman scattering [19]. However, the selections rules are relaxed for polycrystals and, in such cases, the TO mode has been previously observed at ~180 cm⁻¹ [19,20]. Consistent with the x-ray diffraction analysis showing the fiber core to be polycrystalline, a peak at ~180 cm⁻¹ is observed from the fiber and is assigned to this TO mode. Further, a weak mode is found at ~185 cm⁻¹ in the Raman spectrum of the InSb starting material and is attributed to a weakly-coupled LO phonon-plasmon L_- mode (or the screened LO phonon) [12]. The LO phonons in InSb-like polar semiconductors can produce a net macroscopic electric field. Such a field has long range interactions with electrons from different bands, which are known as Frohlich interactions. These electron-plasmon interactions are equivalent to three-wave mixing of the electric fields of the incident photons, plasmons from the LO phonons, and the scattered photons and obeys the dipole selection rules [21,22]. Finally, an overtone of the LO phonon (2LO) is also observed at ~380 cm⁻¹.

 

Fig. 4 Raman scattering spectra from InSb starting material and the fiber core of the drawn optical fiber. The black curves are the de-convoluted best-fit to the LO and TO modes. The dashed trace is shown magnified and represents the L_– mode discussed in the text.

Download Full Size | PDF

Figure 5 shows the Fourier transform infrared spectra taken through the as-purchased InSb wafer. Observed is a very gradual and monotonic increase in transmission from zero at a wavelength of about 7.5 microns to 28% at a wavelength of 30 microns, which is the expected limit, based on Fresnel reflections. Initial measurements on the drawn fiber (about 2 cm in length) were unsuccessful. It is postulated that the losses are due to several factors including free carrier absorption, the fact that the precursor material is of unknown purity, and, in the fiber, the likely significant scattering from grain boundaries and oxide precipitates.

 

Fig. 5 Fourier transform infrared (FTIR) spectra through the as-purchased InSb wafer.

Download Full Size | PDF

The free carrier absorption cross-section for InSb due to electrons in the long-wave infrared (LWIR) range is about 0.25 × 10⁻¹⁶ cm² [23]. This value, coupled with the vendor-specified n-doping level of about 5 × 10¹⁶ cm⁻³, would result in an extinction of about 1 cm⁻¹ at 9 μm, representing the base level of possible transmission from this particular fiber. Additionally, the elemental and single crystal x-ray diffraction analyses indicated a significant oxygen content and degree of polycrystallinity, respectively. As noted above, since the powder x-ray diffraction showed phase-pure InSb the oxygen in the core is most likely present as amorphous oxide precipitates. Such precipitates would certainly yield significant scattering given the very large refractive index difference with respect to the InSb (refractive index, n ~4). The presence of phosphorus in the core, certainly diffusing in from the cladding phosphate glass, could also create defects and traps that lead to added absorptions. Further, while InSb is cubic and grains of differing orientation would exhibit the same refractive index, the grain boundaries are often sites for impurities, defects, and possibly a preponderance of the aforementioned precipitates. Hence the grain boundaries are likely significant contributors to scattering and the diminution of optical transparency.

One last point, added for completeness, is that the phosphate glass chosen for the cladding might not seem a particularly appropriate choice of materials given its lack of transparency at the infrared wavelengths where the InSb core could function as a waveguide. Principally, the choice was a convenient expedient for this proof-of-concept work given its relative ease of fabrication and fiberization. That said there do exist several applications for multimode fibers, including mid-wave infrared (MWIR) power transmission in ladar and chemical sensor systems or Raman amplifiers for wavelength shifting and beam clean-up. In such cases, given the large refractive index difference between the InSb core and the phosphate cladding, very little of the optical power would propagate in the cladding and so the transparency of the cladding is less influential to the overall attenuation. Secondly, there are some refractory chalcogenide glasses in the Ga₂S₃ – La₂S₃ family that possess sufficiently high draw temperatures to make infrared transparent glass-clad InSb core fibers. Such glasses are the present focus of continuing efforts.

Future efforts will focus on InSb samples of known (and high) purity to reduce the free electron absorption as well as methods to reduce the oxygen diffusion into the melt during the draw process. Absorption levels of about 1 – 10 m⁻¹ likely would need to be realized to truly enable practical device applications. This would require a negligible hole concentration and an electron concentration in the range of 10¹⁶ cm⁻³. The important aspect reported here is the realization that glass-clad optical fibers containing binary semiconducting core can be fabricated using conventional fiber-draw techniques.

Conclusions

Phosphate glass-clad optical fibers containing an InSb core were successfully fabricated using a molten core approach. The cores were found to be phase pure and highly crystalline with clear indications of a dominantly granular (polycrystalline) microstructure. This work verifies that optical fiber can be drawn using commercially-accepted and scalable production methods that contain semiconductor cores of higher crystallographic complexity than previously achieved. Optical fibers comprising binary III-V semiconductors could enable a myriad of intriguing electro-optic and nonlinear optical applications not attainable in glass, dielectric crystal or unary semiconducting systems.