Accounting for stoichiometry changes on compound semiconductor surfaces

Abstract

We discuss how to self-consistently account for stoichiometry changes on surfaces with complex reconstructions, e.g. polar compound semiconductor surfaces. The key aspect of the methodology is that surface steps are allowed to act as a reservoir where atoms may be added or removed. The method is specifically applied to molecular beam epitaxy (MBE) and atomic layer epitaxy (ALE) results for GaAs(100) surfaces. The method easily demonstrates why only ∼ 12 monolayer (ML) of As is needed to convert from the Ga-rich surface to the (2 × 4) As-rich surface (as observed experimentally), even though the latter surface has an As coverage of 34 ML in the top layer. We also demonstrate how to convert from a more complex reconstruction, having two incomplete layers, to a simpler reconstruction having only one incomplete layer. The methodology also shows that ideal ALE of GaAs(100) cannot occur by cycling between the known adsorbate-free Ga-rich and As-rich surface reconstructions, because any such transition would not yield the observed 1 ML per cycle growth rate. We briefly discuss how adsorbates may stabilize ideally terminated (i.e. vacancy free) III–V surfaces. For example, methyl groups adsorbed on GaAs(100) exhibit a (1 × 2) LEED pattern, which is not seen for the clean GaAs(100) surface reconstructions. By using the electron counting model we interpret this structure as 12 ML CH3 adsorbed on a complete layer (1 ML) of dimerized Ga atoms. The ideal termination of the Ga-rich GaAs(100)-(1 × 2)-CH3 surface now allows for a plausible ALE mechanism which yields 1 ML deposition per cycle.