Silicon contamination of III–V compounds is a potentially serious problem because of silicon's amphoteric behavior and its high affinity for oxygen. A potential source of silicon is fused quartz, which is widely used in systems for the growth of normalGaAs and normalGaP . Most open growth systems use flowing hydrogen because it is easily purified and it can remove solid gallium sesquioxide from the system. In this paper, calculations are presented for the rates of contamination of liquid gallium with silicon and of flowing H2‐normalHCl mixtures with volatile silicon compounds. Ideal gas behavior and local thermodynamic equilibrium are assumed. It is shown that the contact of Ga or normalHCl with quartz at normal growth temperatures can lead to significant silicon contamination in very dry systems. In addition, it is found that at 800°C, the pressure of water cannot be controlled below 10−8 atm for H2 in contact with SiO2 and 5×10−7 normalatm for normalHCl in contact with SiO2 , regardless of the initial purity of the H2. These pressures rise appreciably with increasing temperature. In addition, models are presented for the formation of high concentrations of Si at the substrate‐epitaxial interface and for the formation of SiO2 precipitates in vapor growth systems. The former has previously been associated with the formation of “ I ” or insulating layers. The calculations show that normalHCl synthesistype systems should give rise to much less silicon contamination than transport‐type systems, and vapor growth systems using H2O as a transporting agent give rise to nearly negligible Si contamination below 1000°C. Other possible ways to minimize silicon contamination are also discussed.

No abstract

A GaP decomposition source for producing a dimer phosphorus molecular beam free of gallium and tetramer phosphorus J.Thermodynamic calculations are presented which provide explicit expressions for the combined temperature and phosphorus partial pressure dependences of the Ga-and P-vacancy concentrations in GaP. This work extends the applicability of earlier results for the vacancy concentrations, obtained by ~n analysIs of the solidus data, from solid-liquid (i.e., LPE and LEe growth) to solid-vapor eqUlhbna (vapor growth, annealing, and diffusion). The calculations are represented in the form of Ga-and P-vacancy concentration and solid-composition isobars. In addition, the previously established correlation between the reciprocal of the nonradiative shunt-path lifetime (liT,,) and the relative Ga-vacancy concentration for GaP crystals prepared by a variety of techniques is transformed into a correlation with absolute vacancy concentrations. The electron-capture cross section of a Ga-vacancy behaving as a shunt path in p -type material is estimated to be ~4.S X 10-17 em'. Finally, the Ga-vacancy concentration isobars and the liT" correlation are discussed in the context of LPE, vapor growth, and annealing processes for GaP.

A model based on the formation of silicon-oxygen pairs is proposed to explain a variety of anomalous phenomena associated with GaAs grown in the presence of silicon, SiO2, and/or oxygen. It is suggested that silicon atoms on gallium sites pair with interstitial oxygen atoms, forming a complex which behaves as an acceptor with energy levels near 0.2 and 0.4 eV below the conduction band. It is assumed that the complex can dissociate upon annealing below 850°C by the reaction 2(SiGaOi)−=(SiGaO2)0+SiGa++3e−. This reaction may be reversed at higher temperatures. The electrical and thermochemical properties of such a complex can explain annealing behavior, the occurence of a mobility maximum between 150 and 200°K, site distribution of silicon in p-type silicon-doped GaAs, change from n- to p-type conductivity as a function of Si concentration in GaAs grown from Ga solutions, the formation of I (insulating) layers in GaAs grown by vapor-phase epitaxy, and some of the anomalous behavior observed in GaAs devices. A variety of growth experiments and characterization techniques which could aid in verification of the SiGaOi complex model are suggested.

No abstract