Surface processes in GaAs epitaxial growth

There are several points to note about the GaAs (100) surface. First, the surface reconstructs into dimer rows as does silicon (100) 2x1. Flowever the arrangement of these dimers is more complex to account for the difference in chemistry of the anion and cation sites and to get the electron pairing that drives reconstruction correct. The structure removes half of the top layer of atoms and some of the second layer atoms to expose a third layer of the same chemistry as the top layer. These exposed atoms 

It has been proposed and argued based on experimental results that decomposition of As4 to deposit single As atoms on the surface requires multiple open Ga surface sites. 

Joyce [17] proposed that four open Ga sites are required and that two AS4 molecules must adsorb and migrate such that these four sites that are adjacent and in a specific orientation to one another (later has been referred to as configuration-dependent reactive incorporation by Madhukar and others). These two AS4 molecules then react to produce four adsorbed As atoms and release an AS4 molecule. This configuration requirement makes the incorporation probability for As very low on a surface containing few open sites (exposed Ga) so adsorption in the presence of excess As is self limiting. 

Diffusion across the channels is roughly four times slower. On the Ga-terminated surface the diffusivity is roughly three six times higher. [20]. Therefore Ga on an open Ga-terminated region will move quickly until it comes to a surface step and moves up or down onto an As-terminated region where diffusion is much slower. Therefore, the As-terminated regions will accumulate most of the free Ga adatoms while As will only stick to the Ga-terminated areas. 

Atomic diffusion on GaAs surfaces has been examined in the context of coarsening of surface islands on a flat surface. [21] They and others have found that coarsening does not proceed as one would expect for normal attachment/detachment and diffusion processes on surfaces. The data is consistent with a higher than usual concentration of adatoms on the surface, suggesting that the attachment rate of atoms to surface islands is relatively slow compared to the rate at which atoms leave the clusters. Likewise, it suggests that nucleation of new islands is even slower. 

---