Laser restructuring and photoluminescence of glass-clad GaSb/Si-core optical fibres

Semiconductor-core optical fibres have potential applications in photonics and optoelectronics due to large nonlinear optical coefficients and an extended transparency window. Laser processing can impose large temperature gradients, an ability that has been used to improve the uniformity of unary fibre cores, and to inscribe compositional variations in alloy systems. Interest in an integrated light-emitting element suggests a move from Group IV to III-V materials, or a core that contains both. This paper describes the fabrication of GaSb/Si core fibres, and a subsequent CO2 laser treatment that aggregates large regions of GaSb without suppressing room temperature photoluminescence. The ability to isolate a large III-V crystalline region within the Si core is an important step towards embedding semiconductor light sources within infrared light-transmitting silicon optical fibre.


Supplementary Note 1: Calcium Oxide Interface modifier
The effect of CaO as a protective interface layer between the silicon and the SiO2 cladding has been established in the literature 2 , with reduced oxygen incorporated in the core of fibers, and improved optical transmission. While GaSb has been reported to react with elemental Ca to form Ca5Ga2Sb6 3,4 or Ca11GaSb9 5 , these compounds have orthorhombic structures. The XRD spectrum of the fiber core was not compatible with this symmetry, suggesting little if any GaSb-Ca compound formation. In addition, Ca and O content were below the detection limits ( Fig. 1) in Electron Probe Micro-Analysis (EPMA) in the GaSb regions of the as-drawn and annealed fibers. The EPMA map for Ca showed shadows in the region of the core, but a detailed spectrum (Fig. 2) of the Sb signal from this part of the sample, compared to the spectrum of a Ca standard, showed that this is noise, not a real Ca signal The CaO layer remains well segregated at the interface in the fibers even after the laser treatment, suggesting that GaSb and CaO are not reactive, even at high temperatures.

Supplementary Note 2: XRD of as-drawn GaSb fibre
We compared the FWHM of as-drawn pure GaSb-core fibers to the composite, aligned composite fibers to assess the influence of the interfacial strain where the GaSb joins the silicon. The lattice constant of silicon is 0.54 nm, while that of GaSb is 0.61 nm. These results are shown in Fig. 3.

Supplementary note 3: Positional X-ray results for as-drawn fibre
For the longest fibre available, the rotational  angles at which the {220} and {422} Bragg reflections (the only two occurring) were observed were recorded and these are plotted in Fig. 4a. For the first 60 mm, the scan window was reduced to 5 mm, while for the separate piece from a position 150 mm away, the slit width was 15 mm. Error bars are indicative of the uncertainty of the  = 0 placement. The relative  = 0 position for the final piece was not known, but no rotation was seen among the other pieces measured. The box plot (Fig. 4b) presents the angular difference between the phi values for the two peaks, along with the expected observation for their separation based on a <111> alignment along the axis.   Fig. 6a is a secondary electron image of the pure silicon region, and Fig.6b shows that this silicon, left after the removal of GaSb, is all one orientation. The specks of a different colour are contaminant particles, visible in the SEM image. Fig. 6c is an SEM image of GaSb in the transition region between the two compositions, showing a Si crystal that penetrates into the GaSb. As seen in the EBSD of Fig. 6d, the GaSb is polycrystalline, confirming the XRD results shown below. The solidification of the GaSb region occurs rapidly, in part due to the digital laser control, and in part because there is a single temperature at which the transition occurs. The rapid cooldown leads to formation of multiple seeds and a polycrystalline structure. The upper right-hand corner of the EBSD image is noisy, due low contrast in the Kikuchli lines, hindering clear orientation determination. It is important to note that this signal is derived from the outside ~20 µm of the sample due to the large absorption of Mo X-rays by GaSb, and Figure 2(f) in the main text shows that there is still epitaxial alignment of the Si and GaSb.

Supplementary Note 7: X-ray analysis using capillary Bragg instrument
Normally, diffraction equipment with a capillary sample holder is used to identify small quantities of unknown minerals in powder form. Rotation of the capillary around its longitudinal axis assures the capture of all possible reflections for each Bragg angle, as shown in Fig. 8. In contrast, for the fibre cores studied here, there can be three-dimensional crystalline order over large regions, in which case a particular Bragg peak appears only for a small number of rotational positions () of the fibre. Measurement of the angles between those rotational positions allows determination of symmetry and orientation for a single crystal, and estimation of the number of crystallites in the excitation volume if the sample is polycrystalline.
For a single crystal, with -2 chosen for a particular d-spacing, rotation in  will result in reflections at separations that are determined by the crystal symmetry. As an example, in the cubic system, the {220} reflections appear at six values of , separated by 60° if a <111> axis coincides with that of the fibre, as shown for a pure silicon fibre in Fig. 9.
Generally, several families of planes yield reflections, for a single crystal. The angular separation of the Bragg reflections allows determination of the axial orientation. Combining the  responses for several -2 peaks on one set of axes provides an overview of the structure. For the analysis of these fibres, the  For some of the early-drawn GaSb/Si fibre, axial directions of <211> and <433> were observed. Fig. 10 shows a fibre with the <433> direction aligned with the fibre axis. 1 is the angle difference between the two {3 3 1} planes, and it was measured to be 37.69. This angle matches the calculated angle between the (331) and (313) planes of a cubic crystal. In order to calculate an angle between {2 2 0} and {3 3 1}, the angle at which the data is acquired must be considered in addition to the measured 2, as the effective zero for  is shifted by  when the detector moves to acquire a new Bragg peak. For the example here, the change in recorded  is 5.89 because the {2 2 0} peak appears at =10.65 and the {3 3 1} peak appears at =16.54. 2 was measured to be 64.78; with the correction, this gives 70.67, which coincides with one of the calculated dot products.
The  separations from the scan of the GaSb/Si fibre match (0 2 2 ̅ ) and (3 3 1), planes in a single crystal.
The axial orientation of the core is then [4 ̅ 3 3] from a vector product of [0 2 2 ̅ ] and [3 3 1]. The relative intensities of the peaks and the phi scans were the same for three samples taken from a 100 mm fibre, indicating that the crystal orientation, once established, is stable for some time.
While the angular separation of peaks is reliable for analysing these fibres, peak heights are not, as misalignments of the fibre axis of a few degrees can alter the observed amplitudes significantly. 1 For the highly oriented fibres discussed in this paper, there were only two independent families of planes observed, the {220} and the {422}, simplifying the analysis. The six-fold symmetry for both families confirmed the calculated <111> axial alignment. Fig. 10 Axial rotational X-ray scan with the detector set at 2 for the {220} and {331} planes for silicon in an early GaSb/Si-core fibre. The values of 1 and 2 were used to determine the crystalline orientation of the core material.