Silicon/Germanium multilayers shells growth

Both amorphous semiconductors were synthesized by Chemical Vapour Deposition (CVD) performing slight modifications over already published methods. In the case of Si the precursor gas was disilane (Si₂H₆) and the selected decomposition temperature was 375ºC. For Ge, the precursor was germane (GeH₄) and decomposition temperature 270ºC. Lower temperatures resulted in very slow synthesis rates while higher ones allowed little degree of control. It was observed that high temperatures induce the growth of undesired particles.

To calculate the percentage of semiconductor present in the pores, reflectance spectra of the samples were performed. The spectra are then compared to the band structures. Theoretical calculations were done assuming that semiconductor growth is layered, meaning that silica spheres are surrounded by semiconductor shells. The refractive index values used for SiO₂, Si and Ge were: 1.43, 3.80 and 4.10 respectively.

Reflectance spectra of the samples loaded with different amounts of Ge for opals of 660 nm diameter spheres.

An example of control of degree of infiltration of amorphous Ge is shown in the preceding figure. In these spectra we can observe the diffraction peak caused by the first pseudogap, the Fabry-Perot fringes due to the finite thickness of the sample and other reflection peaks associated with various other photonic bands. Apart from the percentage of loaded semiconductor some other information can be deduced from the spectra. From the Fabry-Perot fringes thickness of the sample can be extracted. Our thin opals were typically made of 14-16 layers; this is a final thickness of 13-15 microns. The right panel in Fig. 2 summarizes positions and peak widths of the first pseudogap extracted from reflectance spectra as in the left panels. For low semiconductor filling fraction Full Width at Half Maximum (FWHM) decreases with increasing infiltration and then increases again for further loading. The pseudogap almost closes for a 15 % of Ge in the pores and 20 % in the case of Si (not shown). This is the result of matching the effective refractive index of the pore (air-semiconductor) and the silica spheres.

The key parameter to control the amount of semiconductor loaded is reaction time. Next Figure summarizes the percentage of pore loaded with Si or Ge as a function of synthesis time. In our experimental conditions, for a single reaction, Si reaches a maximum at approximately the 53% of the pore infilling (24 nm of layer thickness for opals made of 660 nm silica spheres). In the case of Ge, the pore can be loaded without saturation. Pressure build up by hydrogen production is a limiting factor. The only limitation for complete infillings seems to be a geometrical one related to the closure of the pores in the {111} close packed planes.

Infilling as a function of single reaction times. Si (full circles) is formed at 375ºC and Ge (open circles) at 270ºC. Silica spheres of 660 nm of diameter are used. The inset shows the layer thickness dependence of filling fraction for an fcc. Layer thickness is normalized to lattice parameter.

With this degree of control over the Si and Ge growth we can grow multilayer structures. The method allows not only growing both materials on silica but either on the other. A further degree of freedom will be provided by the selectivity of different solvents that can be used to remove some of the materials. Aqua regia  can selectively remove Ge damaging neither the silica spheres nor the Si layer. Ge can be oxidized at 500ºC without changing the properties of the rest of materials. To test this ability, a sample with a 30% of the pore loaded with Ge was re-grown with Si up to the 80%. Then, Ge was etched with aqua regia. Next Figure shows an electron microscope picture of the resulting structure after all the process. The Si network is interconnected and fixed to the substrate and remains separate from the spheres by the air layer.

Cleaved edge SEM image of a doubly connected structure. The sample shows the continuous Si layer separated from the continuous silica structure by air shell. The dark lines in the cleaved edge are the air gaps between Si shell and the silica spheres. Scale bar is 1 micron.

A more complicated structure was fabricated and optically characterized. First a sample was loaded with Si up to 20% of the pore, then a layer of Ge to complete 45% and finally Si to a 60%. At this point the composite presents the following composition: SiO₂(74%)-Si(5.2%)-Ge(6.5%)-Si(3.9%); percentages representing the total volume fraction of each material and a 10.4% air remaining. This sample was then dipped in aqua regia for 60 minutes. The outcome is two homogeneous Si shells separated by air gap and grown on silica spherical cores.

Reflectance spectra of a sample after successive semiconductors infilling and etching. a) Bare opal; b) a Si layer is grown; c) a second layer of Ge and d) a third layer of Si; e) partial etching of the Ge; f) etching is completed and regions with a pseudogap at 2.22 μm disappear.

Another tool is available as hydrofluoric acid (HF) can selectively remove both oxides. A particular case where these techniques may be applied to tailor the photonic bands is illustrated by the following four-step process. First, an opal is loaded with Ge to a 25% of the available volume, followed by its oxidation. A further Ge load completes 86% of the initially available volume. Finally the oxides (GeO₂ and SiO₂) are removed with HF. The corresponding photonic band structure is shown in the next Figure. Two PBG open now: a larger one (12.6%) between the 8^th and 9^th bands and a narrower one (1.3%) between the 5^th and the 6^th bands. The later is an interesting case that only very recently has been reported for fcc-based structures.

Photonic band diagram of interpenetrating air spheres coated with amorphous Ge (n=4.1) in a fcc lattice. Being 0.3645a the radius of the air spheres and 0.4087a the external radius of the semiconductor shell. Two gaps are developed: one between the 5th and 6th and another between the 8th and the 9th. The inset shows the corresponding real space structure