The molecular beam epitaxial growth of GaAs on Si(211) has been investigated. Theoretical considerations had suggested the (211) orientation to be particularly suitable for the nucleation and growth of a zincblende‐type compound semiconductor on a diamond‐type elemental one. The experimental results support the theoretical prediction. Morphologies of thin (≤0.1 &mgr;m) (211) layers are substantially better than for (100) layers, which nucleate poorly and require large layer thicknesses (&bartil;1&mgr;m) to yield good morphologies. When the (211) layer growth is initiated with a thin (GaAs/Al, Ga)As superlattice buffer (0.1 &mgr;m), consisting of 10 periods of 5+5 nm, the (211) morphology rivals that of GaAs(100) homoepitaxial growth. Chemical etching studies as well as transmission electron microscope investigations show the layers to have the (211)Borientation and to be free of antiphase domains, both as predicted. The (211) layers show strong photoluminescence at 4 K. Not intentionally doped layers arentype, with electron concentrations on the order of 2×1016cm−3, except in the immediate vicinity of the Si interface. Some of the conduction parallel to the layer plane appears to take place on the Si side of the interface, where a two‐dimensional electron gas seems to be present. Then‐GaAs/p‐Si heterojunctions themselves show excellent rectification characteristics and photosensitivity. It is concluded that the (211) orientation is almost certainly preferable to the (100) orientation for the growth of GaAs on Si.