Point defects and diffusion in thin films of GaAs

Impurity diffusivity in thin films of GaAs is affected by native defect concentrations which were grown into the film, and which enter the film from both the substrate and the external ambient. Variations in measured diffusivity are related to native defects whose concentrations gradually relax to equilibrium values via different kinetic pathways. The time required for native defects to equilibrate depends upon the experimental design which in turn determines which region of the phase diagram that an annealing experiment begins and ends in. The Gibbs phase rule is applied to several commonly used experimental designs to show why the equilibrium native defect concentrations are generally not defined solely by temperature. The necessary conditions which define an equilibrium state, and the sufficient conditions which determine whether that equilibrium state can be approximated within a short time, are explicitly discussed. Experiments begun far from equilibrium are often associated with unusually high or low time-dependent diffusivities, but the results are often irreproducible as the native defect concentrations drift for extended periods of time. Experiments begun close to equilibrium are generally associated with reproducible diffusivities because the native defect concentrations reach constant values in a time which is short compared with the anneal period. Some diffusion results can be adequately described by equilibrium models, and the so-called 'Fermi-level effect' model is shown to apply as a special case of solid-vapor equilibrium. Near-equilibrium results from several groups provide strong evidence that a Ga vacancy, with a charge of -1, controls group III element marker diffusion in n-type, intrinsic, and p-type GaAs at T> 800 degrees C. By relating an anneal ambient to a region of the ternary phase diagram, it becomes clear why the group Ii-related diffusion and other variables are extremely sensitive to the experimental design. Pinning of the Fermi energy at the surface during vapor phase epitaxy appears to explain why nonequilibrium concentrations of dopants and charged point defects can be grown into GaAs and affect diffusion in post-growth anneals. The relatively large random measurement noise often accompanying interdiffusion appears to be largely associated with the presence of Al in GaAs. We conclude that this behavior is associated with a residual contaminant, most likely oxygen. It also appears likely that small amounts of an oxidant have been present in many closed ampoule anneals and affected the reported interdiffusion results. We conclude that poorly understood metallurgical reactions are probably responsible for the enormous range of interdiffusion observed in device structures annealed under SiO2 and SI3N4 glass caps. (C) 1997 Elsevier Science S.A.