Both experiments on Si(100) and Si(111) resulted in the observation of a polycrystalline aluminum mediator film after the deposition of Si diffusors. This result confirms the successful diffusion of Si into the respective Al(111) and Al(100) films. However, the formation of p-Al after the process proves that there is far greater damage done to the Al mediator film during Si diffusion than on unpatterned samples. This may have been due, in part, to Al island formation as well as damage possibly caused by rapid lateral diffusion of Si as proposed by the diffusor growth model.
Despite the appearance of p-Al on both samples, selected area transmission electron diffraction (TED) measurements indicate crystalline Si over oxide regions. This result demonstrates that polycrystalline Al mediates the diffusion of Si to a buried interface. Thus, the appearance of p-Al over oxide regions should not prove to be an obstacle to this fabrication technology.
For oxide patterned wafers, the cleaning method used[1], a 0.5% HF solution was used for 20 minutes followed by the UV-ozone cleaning used here. Thus, over-exposure to hydrofluoric acid may have resulted in the formation of aluminum islands both on the oxide and in seed regions.
In the case of SMM-SOI on Si(100), where Al islands dominate the oxide regions, the film thickness of Al compared to the oxide height may be the most likely cause of Al film separation from the oxide wall. In seed regions, the Al film remains intact aside from the separation of the film from the oxide walls. The film is crystalline as provided by the Si seed beneath. Over oxide, the Al film is separated from the oxide wall entirely along the length of the strip as shown in every SEM-observed oxide region. However, the Al film is also separated throughout the oxide surface into chains of islands. In the 50/500 $\mu$m oxide region, it is clearly noted that oxide contaminants are present that appear to be the mechanism for film separation over oxide. Again, these contaminants are most likely the result of improper wafer cleaning and handling methods[bib]240[/bib].
Nonetheless, the appearance of crystalline ELO Si regions by TED may be explained in terms of the successful vertical movement of the SMME growth front from unaffected edges of the seed area that then provides a suitable seed for crystalline ELO of Si. Over oxide regions, where it has been observed by SEM particularly on the 10/10 $\mu$m region, the Al island chains manage to collect into one continuous chain in the center of the oxide pattern during the Si deposition stage. It is suggested that the lateral SMME growth front has pushed the Al islands away from their original location near the oxide edges, as well as the possibility that diffused Si from above has allowed the Al islands to spread out across the oxide regions. In comparision, the 5µm oxide pattern of the 2/5 $\mu$m region (Fig. 5.18) shows very tightly packed Al islands over oxide while the 10 $\mu$m oxide pattern of the 5/10 $\mu$m region (Fig. 5.19) shows Al islands spread out moreso across the width of the oxide pattern. The diffusive coalescence mechanism may account for this spreading out of the islands better than the suggestion of the SMME growth front mechanism. Furthermore, a careful inspection of Figures 5.15 and 5.14 reveals that the Al islands are originally smooth compared to their texture after the diffusion of Si. Therefore, it remains inconclusive as to which mechanism dominates over oxide regions.
The TED investigations conducted were of the 2/2 $\mu$m region. Investigations of other regions were inconclusive. This inconclusivity may be the result of the fact that there is no crystalline Si overgrowth on the larger oxide regions. In all oxide patterned regions investigated by SEM, the Al film over oxide has coalesced into Al chains. Moreover, the Al film has been shown to be fully removed from the oxide edge. The 10/10 $\mu$m region of Figure 5.15 clearly indicates that the Al film has already peeled away from the oxide edge before the deposition of Si. The region between oxide edge to Al island chain is about 1-2 $\mu$m. In this region, once Si is deposited, an amorphous Si film forms, with the potential for some polycrystallinity due to the underlying alumina domains. This predominantly a-Si film would act as an impenetrable barrier to an advancing ELO Si SMME growth front. Therefore, the inconclusivity of the TED on most oxide patterns can be explained.
In the 2/2 $\mu$m region, the Al film on oxide has no space to retreat, and so there may have initially been adequate Al coverage near the oxide edges that would mediate the ELO of Si from seed regions. This could not be explored directly since the 2/2 $\mu$m oxide pattern is at the center of the oxide-patterned wafer (Fig. 4.2) and has no "Al-only" coverage. Still, the means by which results of crystalline overgrowth occurred warrants further investigation.