Theoretical model and experimental data on plastic relaxation are revi
ewed. It is concluded that neither equilibrium nor kinetic effects are
responsible for the relaxation observed under conditions of high grow
th quality in III-V epitaxy; instead dislocation multiplication mechan
isms and the space they require to operate constitute the determining
factor. Residual strain is then given by epsilon(h) = (0.8 +/- 0.1 nm)
/h for a layer of thickness h. Deviations from this rule are indicativ
e of poor crystallinity. Silicon-germanium and II-VI materials are bri
efly discussed. (C) 1996 The Institute of Materials.