It was recalled that the critical layer thickness in the hetero-epitaxial growth of zincblende semiconductor compounds, as defined by a minimum energy configuration between coherent deformation of the layer and a full misfit dislocation system, clearly varied only as [lattice mismatch]-3/2. New results were reported here regarding situations which were defined mainly by the fact that the misfit dislocation array was, as often observed experimentally, not complete in length or density. In the simplest case, the interfacial misfit dislocations had a maximum length (but along only one direction) and the layer was partially relaxed; with a strain energy that was an approximately linear function of the thickness. Pseudo-critical layer thicknesses were found which were smaller than that of a completely relaxed layer. In the case of an incomplete interfacial cruciform misfit dislocation array, dislocation loops with threading dislocation lines were assumed to exist and the system involved strain energy levels which were always higher than that of a full misfit dislocation system. It was shown that the setting up of the full misfit dislocation system was not progressive but was rather a sudden one, like a phase transition. It was also observed that partial misfit dislocation systems led to apparent critical layer thicknesses which were different to that for a complete misfit dislocation network. In the case of low-density misfit dislocation arrays (with misfit dislocations that were infinite in length and were located at various distances from each other), stepped stages in the setting up of the total system were also found upon varying the array density. The apparent critical layer thicknesses varied from 0.7t to 1.4t, where t was the equilibrium critical layer thickness in an ideal unit cell. Phenomenological models were proposed that furnished results which were in good agreement with computational data.

F.Bailly, M.Barbé, G.Cohen-Solal: Journal of Crystal Growth, 1995, 153[3-4], 115-22