Motorola grows GaAs on silicon
...William Ooms, the Motorola Labs director of materials, device, and energy research, explained that the GaAs lattice constant (5.65 Å) is about 4% larger than silicon's (5.43 Å). This difference makes it difficult to grow high quality GaAs directly on silicon. Instead, the Motorola researchers placed a thin layer (about 50 Å) of strontium titanate (SrTiO3) between the two semiconductors. Strontium titanate's lattice constant falls roughly halfway between GaAs and Si. Moreover, an amorphous layer (10-20 Å thick) forms between Si and SrTiO3. This layer absorbs the lattice mismatch strain between the two materials, allowing the crystalline SrTiO3 to form a normal lattice without distortion from the underlying Si. Because the SrTiO3 layer is very thin, the amorphous interlayer also absorbs the mismatch strain between the SrTiO3 and GaAs layers...
Link here: http://www.thinfilmmfg.com/Noteworthy/Noteworthy01/GaAs20Sept01.htm
What a big memory you have grandma!
I remember being a little skeptical. Silicon is an amphoteric dopant of GaAs (amphoteric - fancy word meaning that it can act as either N or P type dopant depending on whether it substitutes for the galium or arsenic, which in turn depends on crystal growth conditions such as temperature gradients(?) or something like that) GaAs is "poisoned" by oxygen, so depostiting it on an oxide is problematic IMHO.
But what do I know? it's been 25 years since I built LEDs, and I've never been in the high speed GaAs end of the business.