316 Journal of Crystal Growth 112 (1991) 316—336 North-Holland Numerical study of the influence of reactor design on MOCVD with a comparison to experimental data Andrew N. Jansen *, Mark E. Orazem * Department of Chemical Engineering, University of Virginia, Charlottesville, Virginia 22901, USA Bradley A. Fox and William A. Jesser Department of Materials Science, University of Virginia, Charlottesville, Virginia 22901, USA Received 14 August 1990; manuscript received in final form 28 January 1991 Comparisons were made between experimental data and a two-dimensional model for the MOCVD of GaAs from trimethylgal- hum and arsine in two horizontal air cooled reactor geometries. A unique feature of this work was that comparison was made, not only on the wafer, but over the entire deposition regime. Excellent agreement was achieved for growth at low system pressures. Experimental deposition profiles under atmospheric pressure were much less uniform than those predicted by the model. This difference could not be attributed to a pressure dependence of heterogeneous reactions. Inclusion of thermal diffusion had little effect on the uniformity of the calculated deposition profile but decreased the magnitude of the growth rate by up to 12%. Through model calculations and experiments, it was determined that a channel with a tilted upper wall and a horizontal susceptor has the same growth rate profile as does a standard horizontal upper wall channel with a tilted susceptor. Substrate rotation was predicted to cause growth rate uniformity within less than 4.3% in horizontal channel flow and 3.8% in a channel with tilted walls. I. Introduction tam gallium. Typical models are based on a Levêque development in which the governing Metalorganic chemical vapor deposition equations are those for fluid mechanics coupled (MOCVD) differs from other chemical vapor de- with convective heat and mass transfer. In these position processes in the type of phenomena that models, gas phase and surface reactions are usu- limits the growth rate. In silicon deposition from ally neglected (see ref. [1] for a review). Typical silane, for example, it has been determined that boundary conditions for the convective mass the growth rate is limited by the rates of gas phase transfer are that concentrations of reactants up- reactions. Models based on this with appropriate stream of the substrate are the same as the inlet chemical reactions and rate constants are able to concentrations and that the concentration of the predict the results of various experimental condi- limiting reactant on the substrate surface is zero. tions [1,2]. In contrast, MOCVD processes are Deposition is assumed to occur only on the sub- usually assumed to be mass transport limited un- strate, and the resulting flux of the limiting re- der conventional operating conditions [3—13].In actant is infinite at the leading edge of the sub- the case of GaAs grown from arsine and trimeth- strate. ylgallium, the limiting species is assumed to con- The object of this work was to develop a mathematical model that could be used as a pre- dictive tool for the design of MOCVD reactors. * Current address: Department of Chemical Engineering, One of the first numerical models developed ex- University of Florida, Gainesville, Florida 32611, USA. plicitly for MOCVD was the three-dimensional 0022-0248/91/$03.50 © 1991 Elsevier Science Publishers B.V. (North-Holland)