Structural characterisation of silicon-germanium virtual substrate- based heterostructures grown by low pressure chemical vapour deposition

Silicon-germanium heterostructures incorporating compositionally graded virtual substrates are important for the fabrication of a variety of advanced electronic devices. Their successful application depends critically on their surface morphology and defect content. The aim of this research project is to characterise the way in which these structural properties are influenced by the growth parameters used in low pressure chemical vapour deposition (LPCVD) at the Southampton University Microelectronics Centre (SUMC). To this end, a comparative study of the surface quality and the distribution and density of misfit strain relaxation induced defects in SiGe virtual substrate-based heterostructures grown under varying conditions, was carried out. The growth parameters varied have been: growth temperature, initial and final Ge content, Ge concentration gradient, type of Ge grading profile (linear and stepwise) in the virtual substrate, and thickness and presence of a device structure in the capping layer of constant composition. Characterisation was performed using Nomarski differential interference contrast microscopy, atomic force microscopy (AFM) and transmission electron microscopy (TEM). Growth conditions combining a temperature of 800 deg C and a Ge concentration variation in the virtual substrate between 13% and 42% were found to activate an inefficient misfit strain relaxation mechanism in linear-graded heterostructures, whereby dislocation nucleation prevails over dislocation motion to relieve the misfit strain. Results showed rough surfaces with deep trenches and deep faceted pits, a high density of short misfit dislocation segments extending well into the capping layer of constant composition and a high density of threading dislocations reaching the surface. Additionally, threading dislocation pileups were observed at faceted pits. At lower growth temperature (750 deg C) and for a Ge concentration variation in the virtual substrate between 12% and 42%, bot