Automatic modelling of reaction systems using genetic algorithms and its application to chemical vapour deposition processes: advanced utilizations of simulators for chemical systems

Abstract

The identification of appropriate reaction models is very helpful for developing chemical vapour deposition (CVD) processes. We introduced novel algorithms to analyse experimental data from CVD processes and identify reaction models automatically using genetic algorithms (GAs). The reaction models, which consist of various deposition species and gas-phase and surface reactions, were determined both quantitatively and qualitatively, based on chemical kinetics. The GA modelling algorithm consists of a process for calculating the predicted results from the reaction model candidates and a process for modifying the candidates by use of the difference between experimental and predicted results. We demonstrate the validity of this approach to successfully identify the appropriate reaction models from synthetic experimental data and real experimental data obtained during thermal CVD of tetraethylorthosilicate.