Silicon has become the most important material for the semiconductor industry, due to several advantages like good heat conductance or the high quality of its oxide. Nevertheless, for opto-electronic devices, the limitation of its indirect band-gap has anticipated a breakthrough. To increase the probability for radiative recombination, one way is to use Heisenberg's uncertainty principle. If the carriers are confined in a quantum well, this leads to a broadening of the wave functions in reciprocal space, and a higher probability for recombination processes. For Si, this can be achieved by the deposition of Ge, as this material grows on a silicon surface in Stranski-Krastanov mode. After the formation of a 3-5 monolayer (ML) thick wetting-layer, Ge dots form which can act as quantum wells within a Si matrix. While holes of the valance band are confined in these dots, the silicon surrounding the Ge dots is under tensile strain and therefore acts as a slight quantum well for the electrons of the conductance band. Crucial parameters for the confinement of the carriers and therefore of the probability for radiative recombination are the density, size, and composition of the dots. One way to influence the density and size of Ge dots is to modify the Si surface by the predeposition of carbon. The deposition of submonolayers of carbon leads to a c(4x4) reconstruction of those parts of the surface, the carbon is incorporated in. If Ge is deposited on such modified surfaces, it starts to grow on the c(4x4) free areas, due to the strain induced by the carbon. As a result Ge grows directly in a three-dimensional way, and smaller dot sizes and higher densities can be achieved. As these physical values depend on the size and density of the c(4x4) reconstructed areas, the modification of the Si surface by the pre-deposition of carbon was studied by Scanning Tunnelling Microscopy (STM). It was found, that the deposition between 0.11 ML and 0.2 ML of carbon leads to the best compromise between density and size of c(4x4) reconstructed areas. In addition we studied the carbon induced Ge dots by photoluminescence spectroscopy. The intensity of the photoluminescence signal indicates an increase for the probability of no-phonon-assisted recombination. Besides the size and density, the composition is, as already mentioned, of importance. We found by STM that due to capping of Ge-dots with Si, as it is necessary to embed them in a Si matrix, at high temperatures unwanted intermixing occurs. That can even lead to a shape transformation from dome to hut clusters. This observation was proven by Energy Filtered Transmission Electron Microscopy giving the information, which parts of the dots intermix strongest. For a quantitative analyses, reciprocal space maps were measured with x-ray diffraction measurement. The simulation of these space maps gave a quantitative insight into the composition of the dots under its restrictions. To prevent the intermixing, the growth temperature for the silicon cap was lowered. Afterwards no shape transformation was found by STM, but during the initial steps of overgrowing (3 ML of Si), another type of cluster appeared, whose origin has not fully been understood yet. To proof the concept of growth temperature reduction for whole devices, two stacks of Ge dots, one overgrown at high temperature and one at low temperature, were investigated by photoluminescence investigations. They gave a hint, that the lowering of the overgrowth not only prevents the intermixing during the initial stages of overgrowth, as investigated before, but also when the dots are completely overgrown.