Hochaufgelöste transmissionselektronenmikroskopische Untersuchungen an Galliumphosphid auf Silizium

In this work GaP/Si-heterostructures, which were grown by MOVPE, were studied utilizing different TEM techniques. With the high-resolution HAADF technique one is capable of investigating interfaces as well as possible APDs on an atomic scale. Despite the commonly praised more intuitive interpretation in comparison to conventional TEM-measurements, effects beyond pure Z-contrast are observable, making an adequate simulation indispensable. It was shown that the AP-approximation deviates by less than 5% from the more time consuming FP-approximation for TEM-sample thicknesses below 50 nm. If relative intensities are evaluated instead of absolute ones, the AP-approximation is valid for an even wider range of thicknesses. The simulations show that the intensity-ratio of Si:Ga or P:Ga, respectively, is a good measure of the thickness of a sample. This facilitates the direct determination of the local thickness from experimental HAADF images. The influence of chemical intermixing on the HAADF intensity and the optimum detector size was determined by additional simulations at a fixed sample thickness. The Si-buffers grown on exactly oriented Si-substrates show, unlike theoretical predic-tions, a clear predominance of the A-type surface reconstruction, in which the dimers are oriented perpendicular to the step edges. This is caused by the H2-rich growth conditions, which significantly differ from the UHV conditions that were assumed to calculate the surface energies. The GaP-layers grown on these substrates always exhibit P-polarity viewed along the step edges. If this polarity is projected down to the interface, it is dominated by Ga-Si bonds. This is caused by the fact that the P-precursor (TBP) is not fully decomposed at the low growth temperatures and the Si-surface is passivated by hydrogen due to the prior buffer growth. Only in the presence of the Ga-precursor, the P-precursor is decomposed catalytically. By adjusted growth conditions, which provide a pulse of the P-precursor at a higher temperature, the formation of P-Si bonds can be enforced and the polarity of the GaP-crystal can be reversed on a large scale. This results in a reduction of the APD size, while their density increases. In contrast to that, the GaP-layers on 2°-misoriented Si-subsrates show Ga-polarity viewed along the step edges, which cannot be reversed by the application of TBP at high temperatures. This can be explained by the fact that the Si-buffer is, contrary to expectations, mostly covered by single atomic steps, therefore the width of the terraces is only about 4 nm. Because of the faster growth rate of the Ga-polar GaP that was found along the [110]-direction, it always prevails. On an atomic scale intermixing of GaP and Si is present at the interface. This was quantified via HAADF-measurements. In thin TEM-samples the intermixing along the interface is directly observable. With respect to the experimental accuracy the investigated samples show the same intermixing behavior. A significant amount of Si can be found in three Ga- and three P-layers. This is in very good agreement with the results from DFT-calculations, which show seven monolayers of intermixing to be energetically favorable. This indicates that it might be the intrinsic interface roughness of the GaP/Si material system. Significant higher growth temperatures, which are used during the growth of Si on GaP, lead to a larger region of intermixing at the interface. The APDs that were observed in GaP-layers on exact substrate show an anisotropic shape viewed along [110] or [-110], respectively. Viewed along [110], the kinking APBs macroscopically lie on {112}- and {111}-planes, while they run on {110}-planes through the whole layer viewed along [-110]. This could again be caused by the different growth rates along these directions. At appropriate experimental conditions the APDs can be detected via STEM, although they should not express conventional Z-contrast. Experiments and simulations for different cameralengths between sample and detector show that the additional contrast of the APDs is due to Huang-scattering, i.e. an increase of thermal diffuse scattering into low angles. This is caused by the fact that the periodicity of the crystal is deteriorated at the APB. In high scattering angles the influence of this effect is small and the chemical analysis is possible. Because of the complex three-dimensional shape of the APDs, the found APBs appear broader than expected from the simple crystal model. Cross-sectional measurements show that the APBs are not fixed to a specific crystal plane but can jump. At positions where the APBs exhibit minimum width, Ga-Ga- and P-P-bonds could be detected. In high resolution images it becomes obvious that the kinking APBs are facetted. Because of this, they can be negatively charged, contrary to the crystal model of the {112}-APD, due to a majority of P-P-bonds.