Multidimensional optimisation of process parameters by experimental design for the deposition of aluminium and silicon oxynitride films with predictable composition

In this study the deposition of aluminium and silicon oxynitride films by reactive d.c. magnetron sputtering is systematically planned by design of experiments, carried out and evaluated with the application of specialised statistics software. With these sample applications it is shown that statistical process modelling is a modern tool for process description and optimisation with the outstanding opportunity to gain fast and precise control of a technological process. The influence of the deposition parameters, such as working and reactive gas flow and sputtering power, on the concentration of oxygen, nitrogen and aluminium or silicon in the resulting films is evaluated. With the obtained data a model for the quantitative prediction of the deposition parameters necessary to obtain films with desired composition was built. This is demonstrated by a confirmatory experiment. This is also of technological importance, since the physical properties of the films depend strongly on their composition. With the help of this procedure the sample position, which was not expected to be of any relevance, could be discovered as an influential process parameter. Consequently, design of experiments can be a valuable problem solving technique. Furthermore. the flexibility of the statistical model is demonstrated by implementing the degree of process long-term repeatability into the model and, in addition, by excluding two deposition parameters with a lesser degree of significance and calculating the prediction accuracy of a process operating under this less complicated general condition. (C) 1999 Elsevier Science S.A. All rights reserved.