Computational fluid dynamics (CFD) is used in the semiconductor industry for the analysis and design of chemical vapor deposition reactors. One critical input needed for prediction of epitaxial thin film growth rates and uniformity is the chemistry occurring in the gas phase and at the surface. Traditionally, simplified chemistry derived from experimental observations has been used for this purpose. With the advent of modern high-speed computational techniques, it is now possible to formulate detailed reaction mechanisms using ab initio methods. Such detailed reaction mechanisms, comprising mostly of elementary reactions, have the advantage that they require little or no calibration. This paper presents a methodology in which the density functional theory, in combination with rate theories, was used to determine the reaction pathways and rates in the gas phase as well as at the surface for gallium nitride growth from tri-methyl gallium and ammonia. The reaction mechanisms were then used as input to a multi-dimensional CFD code enabling accurate prediction of growth rates. Validation studies were performed for four different laboratory-scale reactors, and one commercial reactor. In each case, the predictions agreed reasonably well with the experimental data indicating the universality of the reaction mechanisms. (c) 2005 Elsevier B.V. All rights reserved.