1. Introduction

The mid-infrared (MIR) region of light, found between 3-25 µm wavelength [1], covers an important diagnostic tool known as the molecular fingerprint [2]. Due to the strong vibrational resonances of almost all organic species, MIR light offers a unique opportunity to study the key building blocks of biological tissues via a non-destructive spectroscopy technique [3]. This is particularly promising for several remote sensing applications, not least towards the in-vivo diagnosis of early-stage cancer. Since MIR absorption bands can be five orders of magnitude stronger than the overtones and combination vibrational absorption bands of the NIR region [4], it is easier to distinguish between the relevant and competing MIR features. Consequently, a raft of new MIR photonic components is now being developed, including passive optical fibers which will form the basis of an imaging probe.

As one of only a few materials transparent to MIR light, chalcogenide glasses have the additional benefit of being vitreous in nature i.e. making it possible for them to be fiber-drawn. Based on one or more of the chalcogen elements (S, Se or Te) [5], the ternary Ge-Sb-Se chalcogenide system is of particular interest for medical applications, as it is thought to be less toxic than its arsenic-containing counterpart [6]. Defined by 5-35 atomic % (at. %) Ge, 5-40 at. % Sb and the remainder Se [7], Ge-Sb-Se compositions exhibit robustness, relatively good thermal, mechanical and chemical properties and transparency across the 2-16 µm wavelength region [7–10]. It has already been shown that Ge-Sb-Se glasses, including those from the Ge_xSb₁₀Se_90-x at. % system, are best described by their chemically ordered network model (CONM) [11], which often exhibits an extremum for a number of properties, such as glass transition temperature (T_g), density and refractive index (n) [6]. Due to this chemical threshold, it is often difficult to identify two compositions which have appropriate thermal and optical properties, particularly if a low numerical aperture (NA) is desired. Ou et al. [12] and Zhang et al. [13] recently demonstrated successful coupling of Ge-Sb-Se/ Ge-Sb-Se and Ge-Sb-Se/Ge-Se (core/cladding) glasses, respectively, for supercontinuum generation. However, both SIFs had a high NA. For applications that require lower NA, Wang et al. [14] and Guery et al. [15] have shown that the substitution of Se for S, has little effect on the atomic glass structure and that any observed changes in physical properties are caused primarily by a difference in -S- and -Se- bond strengths. Consequently, with increasing S content in the Ge-Sb-Se/S glass system, both the linear and nonlinear refractive indices decrease, whilst the T_g and optical band gap increase [14].

For the chalcogenide optical fibers to perform at their optimum, they must be prepared with high purity. However, the MIR region of light also embraces a number of absorption bands relating to atmospheric impurities such as oxygen, hydrogen and carbon [16–18]. Therefore, to achieve low optical losses in Ge-Sb-Se/S SIFs, it is necessary to synthesize the glasses from high purity starting elements with subsequent distillation, all under high vacuum. In addition to the constituent elements, gettering chemicals such as Al and TeCl₄ are often introduced during the fabrication of high-purity chalcogenide glasses [19–21], so that [O] and [H] impurities are minimized, respectively.

This paper reports an initial investigation of two compositions from the Ge₂₀Sb₁₀Se_70-xS_x at. % system and assesses their suitability to act as core and cladding glasses in a multimode, large core SIF. A distillation procedure was investigated, specifically for the core glass composition. In section 2 the experimental methods of glass melting and annealing are set out, as well as the distillation technique and methods of characterization of the glasses in terms of XRD (X-ray diffractometry), imaging and analytical SEM (scanning electron microscopy), and thermal analysis to measure: glass transition temperatures and the temperature coefficients of viscosity. The approach to measuring refractive index dispersion of the glasses is reported, together with the fiber drawing process and method of fiber loss characterization. In section 3 the results are discussed in terms of numerical aperture and zero dispersion wavelength of the SIF and the effect of glass distillation on fiber optical loss.

2. Experimental

2.1 Glass melting and annealing

All of the Ge-Sb-Se/S glasses investigated in this paper, were prepared via the traditional melt-quench technique [5]. Each chalcogenide glass was made within a vitreous silica ampoule (Multilab, UK), which was initially HF_aq. (aq. is aqueous) etched and annealed to remove residual stress. Carbonaceous deposits and/or physi-sorbed or chemi-sorbed water on the silica ampoule’s internal surface were also removed, by air and then vacuum baking at 10⁻⁵ Pa at 1000 °C for 6 hours each. Prior to batching, all Ge-Sb-Se/S glass melting was begun by purifying elemental Se (99.999%; Materion, ABSCO Materials), Sb (99.9999%, Cerac, ABSCO Materials) and S (99.999% Materion, ABSCO Materials) via a bake-out procedure to remove surface impurities. Although it may be possible to purify the surface of elemental Ge through techniques such as acid washing (although this may leave hydrogen impurities), it is technically difficult to purify Ge through the bake-out procedure since elemental Ge and Ge-oxides have low, and similar vapor pressures [22,23]. Therefore, as-bought Ge (99.999%, Materion, ABSCO Materials) was batched, with purified Se, Sb and S, straight into the prepared silica glass ampoule. All weighing and batching was carried out inside a dry nitrogen glovebox (MBraun 150B-G, ≤0.1 ppm O₂ and ≤0.1 ppm H₂O). For the Ge-Sb-Se glass used to investigate high-purity unstructured optical fibers, 1000 wt. ppm TeCl₄ (Alfa Aesar, UK) was also batched at this stage. Using an oxygen-propane (BOC gas) torch (Jencons’ Junior Jet 7, UK), the silica glass ampoule and its contents were sealed under a vacuum of 10⁻³ Pa. Chalcogenide glass melting was carried out in a modified rocking furnace (Instron), during which the raw materials were slowly heated to 900 °C and homogenized for 24 hours. After cooling to 700 °C, the ampoule was removed from the furnace and the Ge-Sb-Se/S melt was quenched in situ, inside the ampoule, using an external nitrogen (BOC) gas flow jetted at the exterior of the ampoule surface to cool its contents. Immediately after quenching, the ampoule, containing the Ge–Sb–Se/S product, was annealed at its onset-T_g from differential scanning calorimetry (DSC) [24] for 0.5 hours and then allowed to cool to room temperature.

2.2 Ge-Sb-Se distillation

Following the initial melting and annealing of homogenized Ge-Sb-Se + 1000 wt. ppm TeCl₄, this chalcogenide glass boule was broken into approximately 2-5 mm diameter pieces and batched into the charged end of a silica glass distillation rig (Multilab, UK), which had been prepared via the same procedure outlined in section 2.1. Once again, batching was carried out within the dry nitrogen glovebox. At the same time, 700 wt. ppm Al (Alfa Aesar, UK) was also batched with the Ge-Sb-Se + 1000 wt. ppm TeCl₄ pieces so to act as a [O] getter during subsequent heating. Through careful temperature control, the charged end of the silica glass distillation rig was heated to approximately 700 °C. Ge-Sb-Se glass distillation occurred in an open system, under a drawing vacuum.

2.3 Glass characterization

X-ray diffractometry (XRD) was used to analyze the amorphous nature of the chalcogenide glasses using a Siemens D500 diffractometer, with CuKα radiation in the 2θ range 10°- 70°, and with a step size of 0.05°; 40 seconds was spent at each step, resulting in a run time of 13.2 hours. The onset-T_g was measured using differential scanning calorimetry (DSC) on 20 mg of as-annealed Ge-Sb-Se/S chunks, with dimensions ≤2 mm, in sealed aluminum pans. Each DSC sample was run 3 times in a Perkin Elmer Pyris 1 DSC analyzer, under flowing Ar gas and with a heating and cooling rate of 10 °C/min. The mean onset-T_g value (to ± 2 °C) was calculated from both the second and third runs. Viscosity-temperature measurements were carried out using a modified parallel plate technique, in a Perkin Elmer TMA7 thermomechanical equipment with flowing He (BOC) at 20 ml/min. A constant load of 50 mN was applied to samples with dimensions of 1.6 mm height and 4.7 mm diameter, whilst 400 mN was applied to samples with dimensions of 4.1 mm height and 4.7 mm diameter. Scanning electron microscopy (SEM), conducted on a FEI Philips XL30 SEM, was used to image Ge-Sb-Se/S samples under the backscattered electron (BSE) mode. When combined with energy dispersive X-ray spectroscopy (EDX), using Oxford Instruments INCA Energy software, the compositions of carbon-coated Ge-Sb-Se/S samples were measured with an accuracy of approx. ±0.5 at. % for elements heavier than oxygen.

2.4 Fiber-drawing and characterization of Ge-Sb-Se/S glass fibers

All of the Ge-Sb-Se/S optical fibers presented in this paper, were drawn from bulk glass preforms, which had been prepared via various techniques. Unstructured Ge-Sb-Se fibers were directly drawn from as-annealed glass rods (∼70 mm long and 10 mm diameter). Unstructured Ge-Sb-Se-S fiber was drawn from a 4.7 mm diameter, ∼90 mm long glass rod which had been extruded from an as-annealed 15 mm diameter boule [25]. Ge-Sb-Se/S SIF was drawn from a co-extruded 10 mm diameter preform [25] comprising a Ge-Sb-Se core and Ge-Sb-Se-S cladding. Both the Ge-Sb-Se core and Ge-Sb-Se-S cladding glasses used for this co-extrusion, were as-annealed 29 mm diameter glass boules which had been polished to a 1 µm finish on the mating surfaces. Extrusions were carried out using an in-house built, vertical extruder, under flowing N gas. Fiber-drawing took place on a customized Heathway draw-tower (housed in a class 10,000 cleanroom), during which the draw-zone was purged using nitrogen gas (BOC) which had been freeze-dried as it traveled through a copper-coil immersed in liquid nitrogen at 77 K. The chalcogenide preforms were softened and drawn to fiber using a short graphite ring-receptor, which was heated by radio frequency from a copper coil.

Optical fiber loss measurements were conducted on fiber samples ∼7-18 m in length and 200 ± 10 µm diameter, via the cut-back method [26]. Using two set-ups, the optical fiber attenuation was first measured using an InSb cooled detector (JUDSON, UK) in the 2-5.5 µm wavelength range and then an MCT (HgCdTe) cooled detector (Kolmar Technologies, UK) in the 5-10 µm wavelength range. Both measurements were made with a Globar source and a KBr beam splitter. So that the peaks of impurity bands could be estimated, optical fiber loss measurements were also conducted on a shorter length (approx. 0.5 m) of the same fiber under investigation.

Thin film refractive index samples were prepared from 20 mm lengths of 200 ± 10 µm diameter unstructured Ge₂₀Sb₁₀Se₇₀ at. % and Ge₂₀Sb₁₀Se₆₇S₃ at. % optical fibers, which had been hot-pressed between two tungsten carbide plates, under vacuum (10⁻⁴ Pa) at T_g +40 °C (viscosity 10⁸ Pa.s) with a maximum pressure of 700 N. Thin films were annealed at the onset-T_g and allowed to cool in-situ. Using an improved Swanepoel method, with a two-term Sellmeier model, refractive index measurements were found across the 2-25 µm MIR region and had an accuracy of <0.4% and standard deviation of precision < ±0.002. This accuracy was determined using a 3.1 µm interband cascade laser (ICL) [27,28] fabricated at NRL (which emitted >200 mW cw at 25 °C) by comparison with a benchmark refractive index value obtained from prism measurements on Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses. Both the thin film and prism refractive index samples were taken from the same batch. A full description of the improved Swanepoel method can be found elsewhere [29].

3. Results and discussion

3.1 Ge-Sb-Se/S step-index fibers

Following the work by Parnell et al. [11], prospective core and cladding compositions were chosen from the Ge₂₀Sb₁₀Se_70-xS_x atomic % (at. %) chalcogenide glass system. With an overall aim of producing low NA SIFs, for greater selectivity in a future MIR imaging probe application, Ge₂₀Sb₁₀Se₇₀ at. % was selected as the core glass and Ge₂₀Sb₁₀Se₆₇S₃ at. % as the cladding. For the Ge₂₀Sb₁₀Se_70-xS_x at. % glasses specifically relevant to this work, it was predicted that a 3 at. % substitution of Se for S would result in a 3 ± 1 °C increase in T_g and a 0.02 ± 0.01 decrease in refractive index [14]. Well matched T_gs are required for the co-processing of core and cladding glasses to fabricate the SIF. Using DSC analysis, the onset-T_g of each composition was measured and is presented in Table 1. From this Table, the Ge₂₀Sb₁₀Se₇₀ at. % core glass exhibits an onset-T_g of 214 °C, while the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glass has an onset-T_g of 224 °C. This ΔT of 10 °C was larger than the predicted 3 °C, yet is particularly promising when considering the very large change in onset-T_g values when 5 at. % Se is substituted for Ge in the Ge-Sb-Se ternary glasses, viz.: Ge₁₅Sb₁₀Se₇₅ at. % (160.1 °C), Ge₂₀Sb₁₀Se₇₀ at. % (214 °C) and Ge₂₅Sb₁₀Se₆₅ at. % (316 °C) [30].

Table 1. Thermal properties of prospective Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses^a.

View Table

As well as studying the T_g of the core and cladding glasses, thermomechanical analysis was conducted to investigate the viscoelastic behavior associated with the extrusion and fiber-drawing process. Figures 1(a) and (b) present the full viscosity-temperature curves for the Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses, respectively; the average viscosity-temperature results for the extrusion and fiber-drawing processes are presented in Table 1.

 

Fig. 1. Viscosity-temperature curves of the: a) Ge₂₀Sb₁₀Se₇₀ at. % core and b) Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses. Horizontal dashed lines indicate the approximate extrusion (10^7.5 Pa.s) and fiber-drawing (10^4.5 Pa.s) viscosities, denoted by v_e and v_f respectively. The dashed tangent lines indicate the onset of deviation from the linear fall of the viscosity-temperature curves.

Download Full Size | PDF

Since the quality of the test piece can significantly impact the accuracy of the TMA viscometry results, the Ge-Sb-Se/S samples were carefully prepared from a 4.7 mm diameter, extruded rod, along which several cuts were made perpendicular to the rod axis. For both compositions, three cylindrical samples were cut with heights of 1.6 mm and another three with heights of 4.1 mm. The two sample geometries allowed the full range of the viscosity-temperature curves to be measured, considering sample spreading. Figure 1, presents the viscosity-temperature curves for all six samples. At relevant extrusion (10^7.5 Pa.s) and fiber-drawing (10^4.5 Pa.s) viscosities, the temperature was recorded for each sample curve. Therefore, the values presented in Table 1 are an average result calculated from three samples. Corresponding errors indicate the associated standard deviation. Since the errors were found to be small, they are presented only in Table 1 and not in Fig. 1.

Results showed that the temperature required to reach the estimated extrusion viscosity of 10^7.5 Pa.s was similar for both the Ge₂₀Sb₁₀Se₇₀ at. % core (274.7 ± 0.4 °C) and the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding (277.7 ± 0.4 °C) glasses. Likewise, the temperature required to achieve a fiber-drawing viscosity of approx. 10^4.5 Pa.s was also close, with a ΔT of 4 °C between Ge₂₀Sb₁₀Se₇₀ at. % core (357.8 ± 0.7 °C) and the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding (361.9 ± 0.3 °C). These results suggest that the two prospective compositions have similar thermal coefficients of viscosity above T_g and should, therefore, successfully co-extrude.

A final observation is that both viscosity-temperature curves show a slight deviation from linearity, at 339 °C for the Ge₂₀Sb₁₀Se₇₀ at. % core and at 329 °C for the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding. This change in slope indicates a deviation from Arrhenian behavior, and possibly the breaking of different bonds on the atomic level.

In Fig. 2, the refractive index measurements of the Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses are shown from 2 to 25 µm wavelength. Overall, the dispersion curves presented in Fig. 2 show that the substitution of 3 at. % Se for S in the Ge₂₀Sb₁₀Se_70-xS_x at. % system reduces the refractive index of the glass. At 3.1 µm wavelength, the Ge₂₀Sb₁₀Se₇₀ at. % core thin film had a refractive index of 2.553 whilst the cladding Ge₂₀Sb₁₀Se₆₇S₃ at. % thin film had a refractive index of 2.540. With a refractive index difference of approx. 0.01, also observed by Wang et al. [14], a SIF composed of a Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding would give an NA of 0.25 at 3.1 µm wavelength. Ideally, SIFs in a hyperspectral imaging probe might have a slightly higher NA for superior light gathering. However, tailoring the Ge-Sb-Se/S compositions for higher NA, would in turn tend to increase detrimentally the difference in thermal properties.

 

Fig. 2. Refractive index (n) measurements of Ge₂₀Sb₁₀Se₇₀ at. % core glass (dashed line) and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glass (solid line).

Download Full Size | PDF

Based on the refractive index measurements presented in Fig. 2, the calculated NA for the Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses is presented as a function of wavelength in Fig. 3 (a), along with the material dispersion in Fig. 3(b). Since the MIR long-wavelength fundamental vibrational absorption bands of the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding is at a shorter wavelength than that of the Ge₂₀Sb₁₀Se₇₀ at. % core, the refractive index dispersion of Ge₂₀Sb₁₀Se₆₇S₃ at. % decreases faster than the Ge₂₀Sb₁₀Se₇₀ at. %. For this reason, the predicted NA increases with wavelength and from Fig. 3 (a), the lowest calculated NA (0.24) is at 4.7 µm whilst the highest (0.36) is at 23 µm wavelength. From Fig. 3 (b), the Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding thin films have zero dispersion at 6.7 µm and 6.1 µm wavelengths, respectively.

 

Fig. 3. (a) Calculated numerical aperture (NA) values for a step-index fiber (SIF) with Ge₂₀Sb₁₀Se₇₀ at. % core glass and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glass. (b) Material dispersion (D) for a SIF with Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses. Inset shows an enlarged region between 5 to 10 µm.

Download Full Size | PDF

Following these individual investigations of the Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses, both glass melts were combined for co-processing, firstly via a co-extrusion and then fiber-drawing to a 200 ± 5 µm diameter SIF. The optical fiber-loss was measured in an ∼8 m length of this SIF (see Fig. 4), along with the individual fiber-losses of unclad Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses for comparison. Since none of the chalcogenide glasses had been purified beyond an initial bake-out procedure, relatively large impurity bands were expected and found e.g., the Se-H feature at 4.5 µm (10.51 dB/m) and Ge-O at 7.9 µm (19.05 dB/m). However, the absorption bands proved effective in demonstrating successful waveguiding in the SIF, as follows.

 

Fig. 4. Combined optical fiber loss measurements for unstructured Ge₂₀Sb₁₀Se₇₀ at. % core, unstructured Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding and a Ge₂₀Sb₁₀Se₇₀ at. % (core) Ge₂₀Sb₁₀Se₆₇S₃ at. % (cladding) step-index fiber (SIF).

Download Full Size | PDF

When light travels in a SIF, some portion of the optical mode inevitably occupies the cladding due to the evanescent wave phenomenon. Therefore, it was anticipated that traces of the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding impurity bands e.g. S-H at 4.04 µm (42.0 dB/m) and S-O at approx. 8.85 µm (14.8 dB/m), would show in the SIF fiber-loss results. However, as demonstrated in Fig. 4, there is very little difference between the SIF fiber-loss and that of the unstructured Ge₂₀Sb₁₀Se₇₀ at. % fiber. The absence of impurity vibrational absorptions associated with the Ge₂₀Sb₁₀Se₆₇S₃ at. % glass suggests that there was a successful mismatch between the refractive indices of the Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses, which promoted efficient propagation through the SIF core, accompanied by relatively weak evanescent field penetration into the cladding glass. The lowest loss for the Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding SIF was measured as 0.72 dB/m at 6.06 µm wavelength.

To confirm the presence of a core-cladding structure, SEM-EDX analysis was conducted on SIF cleaves used for the fiber-loss measurements in Fig. 4. From a backscattered electron SEM image, presented in Fig. 5 (a), two distinct regions can be identified, compromising a large circular core and a 10 µm thick cladding. With an overall SIF diameter of 200 µm, the core to cladding ratio was found to be 95%. Using SEM-EDX, each region was confirmed as a Ge_20.4Sb₁₀Se_69.6 ±0.5 at. % core and a Ge_20.4Sb_9.9Se_66.5S_3.2 ±0.5 at. % cladding i.e. close to that nominally batched. Furthermore, the elemental mapping of S, as presented in Fig. 5 (b), shows a homogenous cladding in the SIF.

 

Fig. 5. a) Scanning electron microscopy (SEM) image of a cleaved optical fiber taken from the Ge₂₀Sb₁₀Se₇₀ at. % (core) Ge₂₀Sb₁₀Se₆₇S₃ at. % (cladding) step-index fiber and b) energy dispersive X-ray spectroscopy (EDX) map for elemental S (K lines used).

Download Full Size | PDF

3.2 Low optical loss Ge-Sb-Se fibers

To minimize the optical absorption bands associated with oxygen, hydrogen and/or carbon, a distillation procedure was developed. Focusing on the Ge₂₀Sb₁₀Se₇₀ at. % core glass, 1000 ppm wt. TeCl₄ and 700 ppm wt. Al were added, to act as [H] and [O] getters, respectively. With adequate temperature control, the Ge₂₀Sb₁₀Se₇₀ at. % glass was shown to distil successfully at temperatures close to 693 °C in an open system, under a drawing vacuum. Following a re-melt at 900 °C for 10 hours the, nominally batched, distilled Ge₂₀Sb₁₀Se₇₀ at. % glass was quenched and annealed, before being drawn into 200 ± 10 µm diameter, uncladded fiber.

The optical fiber loss measurement presented in Fig. 6, for an 18 m section of the distilled Ge₂₀Sb₁₀Se₇₀ at. % glass (solid line), shows remarkably low loss across the 3-8 µm region: 0.62 ± 0.19 dB/m. The lowest baseline loss was measured as 0.44 dB/m at 6.42 µm wavelength. Although a small Se-H absorption can be seen at 4.6 µm, it is considered extremely difficult to remove all hydrogen impurities [31] and even the purest reported chalcogenide glasses exhibit a small [H] spectral absorption band centered at 4.6 µm [26,32,33]. To the best of the Authors’ knowledge, this result is currently the lowest loss reported to date for Ge-Sb-Se optical fibers. To highlight the improved MIR transparency, the optical loss spectrum for the undistilled Ge₂₀Sb₁₀Se₇₀ at. % optical fiber (presented in Fig. 4), is shown for comparison as the dashed line in Fig. 6. These results show that distilling the Ge₂₀Sb₁₀Se₇₀ at. % glass with [O] and [H] getters significantly reduces the Se-H and Ge-O absorption bands.

 

Fig. 6. Optical fiber loss spectra for the unstructured, distilled Ge₂₀Sb₁₀Se₇₀ at. % glass measured over an 18 m length of 200 µm diameter fiber (solid line). The optical loss in the undistilled Ge₂₀Sb₁₀Se₇₀ at. % optical fiber is included for comparison (dashed line).

Download Full Size | PDF

Using SEM-EDX, the chemical composition of the distilled Ge₂₀Sb₁₀Se₇₀ at. % glass was confirmed as Ge_21.1Sb_10.6Se_68.3 ±0.5 at. % (cf. as batched: Ge₂₀Sb₁₀Se₇₀). This measurement and standard deviation were based on several EDX points taken at regular intervals along the full 18 m length of the optical fiber. Although the final distilled glass composition did not exactly match that of the nominally-batched Ge₂₀Sb₁₀Se₇₀ at. % glass, the standard deviation of the EDX measurements indicate good homogeneity of the glass.

4. Conclusions

Two compositions from the Ge₂₀Sb₁₀Se_70-xS_x at. % glass system have been combined successfully to produce SIFs with low NA. Using DSC and TMA, it was found that the prospective Ge₂₀Sb₁₀Se₇₀ at. % core glass and the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glass, had similar thermal properties. To achieve an extrusion viscosity of 10^7.5 Pa.s there was a 3 °C temperature difference between the Ge₂₀Sb₁₀Se₇₀ at. % core glass (274.7 ± 0.4 °C) and the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glass (277.7 ± 0.4 °C). Likewise, to achieve a fiber-drawing viscosity of 10^4.5 Pa.s there was a 4 °C temperature difference between the Ge₂₀Sb₁₀Se₇₀ at. % core glass (357.8 ± 0.7 °C) and the Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glass (361.9 ± 0.3 °C). Furthermore, the measured refractive indices of 2.553 (Ge₂₀Sb₁₀Se₇₀ at. %) and 2.540 (Ge₂₀Sb₁₀Se₆₇S₃ at. %) at 3.1 µm wavelength gave a predicted SIF NA of 0.25. The zero dispersion wavelengths were also calculated to be 6.7 µm and 6.1 µm for the Ge₂₀Sb₁₀Se₇₀ at. % core and Ge₂₀Sb₁₀Se₆₇S₃ at. % cladding glasses, respectively. The two compositions were successfully drawn into a large core SIF, with the lowest optical baseline loss measured as 0.72 dB/m at 6.06 µm wavelength. To improve further the optical loss, a distillation procedure was developed specifically for high-purity Ge₂₀Sb₁₀Se₇₀ at. % optical fibers. Using Al as an [O] getter and TeCl₄ as a [H] getter, pre-homogenized Ge₂₀Sb₁₀Se₇₀ at. % glass was successfully distilled at temperatures close to 693 °C. Optical loss measurements made on an 18 m length of unstructured, nominally-batched Ge₂₀Sb₁₀Se₇₀ at. % fiber, drawn from the remelted distillate, revealed a remarkably low loss of 0.62 ± 0.19 dB/m across the 3-8 µm region. The lowest baseline loss of 0.44 dB/m, at 6.42 µm wavelength, appears to indicate that the Ge-Sb-Se fiber has the highest purity reported to date.