Jpn. J. Appl. Phys. Vol. 39 (2000) pp. 1642–1649 Part 1, No. 4A, April 2000 c 2000 The Japan Society of Applied Physics Kinetics of GaAs Metalorganic Chemical Vapor Deposition Studied by Numerical Analysis Based on Experimental Reaction Data Masakazu SUGIYAMA, Olivier FERON 1 , Sinya SUDO 2 , Yoshiaki NAKANO 2 , Kunio TADA 2 , Hiroshi KOMIYAMA and Yukihiro SHIMOGAKI 1 Department of Chemical System Engineering, School of Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan 1 Department of Materials Science and Metallurgy, School of Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan 2 Department of Electronic Engineering, School of Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan (Received October 18, 1999; accepted for publication January 19, 2000) In order to develop a computer-assisted process optimization of In 1x Ga x As y P 1y metalorganic chemical vapor deposition (MOCVD), the kinetics of GaAs growth was studied as the first step. For the accumulation of reaction data of source materials, the decomposition rates of trimethylgallium (TMGa) and tertiarybutylarsine (TBAs) were studied using a flow tube reactor and a Fourier transform infrared spectrometer (FT-IR). Special attention was paid to the effect of TBAs concentration on the decomposition rates of TMGa. The GaAs growth rate profile in a commercial MOCVD reactor was analyzed through both experiment and simulation. The profile was dependent on the gas velocity and total pressure. This dependency was explained by a reaction model which was deduced from the experimental observations: TMGa decomposes to a gas-phase intermediate which subsequently forms the GaAs film. The fluid dynamic calculations combined with this reaction model led to growth rate distributions which agreed well with the experimental data. The analysis revealed that the GaAs growth rate is limited by the gas-phase reactions of TMGa as well as the mass-transport of the intermediates, and that precise measurement of the reaction between TMGa and TBAs is essential for an accurate simulation. KEYWORDS: metalorganic chemical vapor deposition (MOCVD), InGaAsP, trimethylgallium, trimethylindium, tertiarybutylarsine, tertiarybutylphosphine, decomposition rate, growth simulation 1642 1. Introduction Process optimization of Metalorganic chemical vapor de- position (MOCVD) of In 1x Ga x As y P 1y is important for the fabrication of optoelectronic devices such as laser diodes and optical modulators. MOCVD is a complicated process that in- cludes surface- and gas-phase decomposition of source gases, transport of both source gases and decomposition products, and finally surface reactions of film precursors. This com- plexity has obstructed the understanding of the film-growth processes, and thus, the trial-and-error optimization of growth conditions is usually necessary. Some previous works dealt with the elementary reac- tions of source gases such as trimethylgallium (TMGa), trimethylindium (TMIn), arsine (AsH 3 ) and phosphine (PH 3 ). 1–5) Other works were concerned with numerical analy- sis of the mass-transfer and heat-transfer in MOCVD reactors and attempted to combine such analyses with elementary re- action models. 6–10) Those works aimed to simulate the growth rate and the composition of obtained films. Those simula- tions could be ultimately used for the computer-assisted opti- mization of process conditions. Among those attempts for the crystal growth simulation, the case of GaAs has been most ex- tensively studied, partially because it is a simple binary sys- tem. One of the most complete works was carried out by Mountziaris and Jensen 9) and Jensen et al. 10) They listed the possible elementary reactions in the TMGa–AsH 3 system and the corresponding reaction-rate constants from the literature. They incorporated these reaction data in a two dimensional calculation of fluid dynamics in a horizontal MOCVD reactor. Their simulation results showed excellent agreement with the experimental growth rate profile and carbon incorporation. Indeed, the latter could not be explained without considering such a variety of reactions. Nevertheless, since they consid- nisms and decomposition rates of MOCVD source gases such as TMGa, TMIn, TBAs and TBP using a flow tube reactor and a gas monitoring system using a Fourier transform infrared spectrometer (FT-IR). We further investigated the effect of group V species on the reaction rate of group III species, since this data seems indispensable for accurate numerical simula- ered more than 40 gas-phase and surface reactions, a three dimensional calculation based on their framework would take too much time using conventional computers. Therefore, a simpler model is preferable as long as it can simulate the de- sired features such as growth rate, composition, and eventu- ally, impurity level. Currently, commercial MOCVD reactors are becoming larger to meet the needs of mass production. The industry requires large-scale reactors that can grow several wafers si- multanuously with excellent uniformity in growth rate and film composition. At the present stage, we cannot avoid mak- ing numerous trial runs to optimize the growth conditions in such a large reactor. Growth simulations that can deal with complex flow and mass- and heat-transport are indispensable, but we do not have sufficient reaction data or experience in growth simulations at this moment. Therefore, much more work is necessary to understand the reactions in MOCVD processes, particularly in combining the reaction data with the fluid dynamics calculations in a suitable way. Furthermore, organic group V source gases such as ter- tiarybutylarsine (TBAs) and tertiarybutylphosphine (TBP) have come in use because they are less toxic than conventional group V sources such as AsH 3 and PH 3 . The decomposition rate data of these new precursors are necessary for the sim- ulation of processes using such gases. Previous studies were concerned with the decomposition mechanism of TBAs 4) and TBP. 5) In a previous paper, 11) we examined the reaction mecha-