We have observed that the nature of misfit dislocations introduced near the critical thickness in GexSi1-x alloys on (001)Si changes markedly in the region 0.4 ≤ x ≤ 0.5. At or below the lower end of this compositional range, the observed microstructure is comprised almost entirely of 60° type dislocations, while at the high end, the dislocation structure is almost entirely Lomer edge type. Concurrent with this change, the dislocation density at the top of the epilayer varies by a factor of about 60X. Similarly, several other observables (e.g. dislocation length and spacing) also change appreciably.
Part of the reason for the morphological variation seems to be a change in the source for dislocation introduction, in conjunction with a change in glide behaviour of dislocations as a function of film thickness. Evidence will be presented that indicates strain, as well as thickness, has a critical value for some dislocation introduction mechanisms, and that these together determine the resulting microstructure.
Furthermore, it appears unlikely that the edge-type Lomer dislocations which appear at about x = 0.5 are either introduced directly, by climb, or grown in, as in the three-dimensional island growth and coalescence which occurs when x approaches unity. Instead, a two-step mechanism involving glissile dislocations is proposed and discussed.