Table of Contents:
For the realization of a highly efficient light source in the optoelectronic integrated circuits on Silicon Ga(NaAsP)-quantum wells (QWs) in the multi-quantum well heterostructures grown on Si (001) by metalorganic vapour phase epitaxy were investigated using scanning transmission electron microscopy (STEM).
In order to optimize the growth conditions, a method for a quantitative comparison of the quality of the layers in the STEM images acquired at low magnifications was developed. The interface roughness and compositional fluctuation were chosen as indicators of the QW quality and mathematically defined.
The investigations of the Ga(NAsP)-QWs grown at 575 °C and having different nominal thicknesses revealed very interesting results. While the absolute roughness of the QWs increases with QW thickness, the relative roughness of the layers as well as their compositional fluctuation is independent of the thickness of the QW layer. However, if the nominal thickness is nearly the same, and the growth temperature is different, it has a significant impact on the relative interface roughness as well as on the compositional fluctuation. The relative and absolute roughness increases with increasing growth temperature, whereas the compositional fluctuation decreases, whereby the lowest interface roughness had a sample grown at 525 °C. A phase separation was not observed at all. It is unexpected for a metastable material. Ab initio calculations verified the experimental results and showed that Ga(NAsP) does not get separated into binary components up to the N content of 20%. Such stability can be explained by a good lattice match on the Si substrate.
The investigation of the annealed samples has shown that not only the growth temperature but also a temperature of the aftergrowth treatment has a strong impact on the quality of the layers. The intensity in the neighbouring GaP layers of the Ga(NAsP)-QWs increases after a rapid thermal annealing. It could be caused by diffusion of the N or As atoms out of the Ga(NAsP)-QW. Moreover, the QWs of the samples, annealed at the temperature of 925 °C and above, contain pore-like dark spots. The origin of these spots, due to their spherical form, cannot be explained by the compositional gradient. According to a quantitative analysis of the compositional fluctuation with dependence on the inner detector angle, the layer becomes more heterogeneous on average if the annealing temperature increases. Hence, the long-scale structural disorder increases due to the formation of the dark spots in the QWs, which correlates with the long-range electronic disorder. The compositional fluctuation between the dark spots correlates with the short-range electronic disorder.
The best interface quality was achieved for the growth temperature of 525 °C and the best optoelectronic properties were detected for the annealing temperature of 925 °C. A further optimization of the QWs-quality was tried by the variation of N/P content. The as grown samples become more homogeneous with increasing P content. After annealing at 925 °C the dark spots occur in the QWs for all N/P concentrations investigated in this work. Their density and size increases with increasing N-content in the annealed specimens. The long-range structural disorder is higher than in the as grown samples, whereby it decreases with increasing P content. The short-range structural disorder has a minimum at approximately 7% N. For N-concentrations above this value the regions between the dark spots in the annealed samples are more homogeneous than the QWs in according as grown samples.
The assumption of the void formation in the Ga(NAsP)-QWs was confirmed by high-resolution (HR) STEM investigations The reason for the formation of voids is most probably the high pressure in the QWs caused by the N2, which can be formed by N-diffusion in the N-rich regions.
The intensity distributions of the group V and group III atomic columns as well as of the background were quantitatively investigated for different detector angles. An excellent agreement between the simulated and experimental results was observed for high and middle detector angles. However, there is huge discrepancy of up 30% between the experimental and simulated results for low inner detector angles.