CHARACTERIZATION AND OPTIMIZATION OF THE LPCVD SILICON OXYNITRIDE PROCESS, USING THE DESIGN OF EXPERIMENTS METHOD

Abstract
The LPCVD of silicon oxynitride has been characterized and optimized using the Design of Experiments method. The data of a half factorial matrix were converted with multiple linear regression analysis into an adequate empirical model : it appeared to be possible to predict the process output parameters (e.g. deposition rate), when applying more or less random input parameters (e.g. process pressure). The method supplies a tool to optimize a process in an efficient way : the optimum settings, found for the silicon oxynitride process are : temperature 800°C, ramp ±8°C, pressure 400 mTorr, DCS flow 70 sccm, (N₂O + NH₃) / DCS = 3, NH₃ / N₂O= 0.18. Corresponding predicted output parameters are : deposition rate 14.2 Å/min., uniformity over the wafer ±2 % and the uniformity over the boat 0 %. The refractive index becomes 1.70 ± 0.01.