Abstract
The relative efficiencies of silicon dioxide, silicon nitride, chromium + silicon nitride, native gallium oxide, aluminium and native gallium oxide + aluminium as encapsulants for gallium arsenide have been investigated. The performance of each encapsulant was assessed by Rutherford backscattering and Hall measurements. The electrical properties of tin implanted GaAs as a function of both annealing temperature and time have also been studied in order to characterise the annesling kinetics. Silicon dioxide was found to be unsuitable because of the out diffusion and collection of gallium at the outer surface of the encapsulants. Although silicon nitride layers were more effective in preventing inpurity outdiffusion, they blistered and lost adhesion above 700°C. The addition of a thin (~ 200 A ) interlayer of chromium between the GaAs substrate and the silicon nitride film greatly improved the adhesion of the latter. Unfortunately, an insoluble compound was formed at the GaAs surface due to chromium indiffusion. Thermal native oxides sometimes gave good results but not reproducibly because of the difficulty in growing layers of a consistently high quality. An evaporated aluminium encapsulant was very successful at 700°C but above this temperature interdiffusion at the GaAs -A1 interface caused surface staining. The use of dual, native oxide + aluminium, layers gave similar results but, because of the oxide interlayer, staining was reduced significantly. Both these encapsulants were unreliable above about 800°C especially for annealnig times longer than 5 min. The sheet electron concentration of tin implanted GaAs increased with annealing time until, after 15 - 30 min, a saturation level was reached. The annealing process was characterised by an activation energy of about 1 eV, a value believed to be consistent with the dissociation of donor-vacancy complexes. Two distinct depth regions were observed in which different annealing kinetics occurred and a possible explanation for this is discussed.