Intelligent Process Control of Silicon Nitride Chemical Vapor Deposition

Abstract

AbstractThe chemical vapor deposition of silicon nitride can be used to protect advanced materials and composites from high temperature, corrosive, and oxidative environments. Desired coating characteristics, such as uniformity and morphology, cannot be measured in-situ by traditional sensors due to the adverse conditions within the high-temperature reactor. A control strategy has been developed which utilizes a process model and an advanced laser-based sensor to measure the deposition rate of the silicon nitride coating in real-time. The control system is based on a three level hierarchical architecture which functionally separates the process control into PID, supervisory and advanced sensor-based control. Optimal setpoint schedules for the supervisory level are derived from a quasi-fuzzy logic inverse mapping of the process model. An advanced sensor utilizing laser ultrasonics provides real-time coating thickness estimates. Model bias is characterized for each reactor and is correlated on-line with the sensor's deposit thickness estimate. Deviations from model predictions may result in parametric changes to the process model. New setpoint schedules are then created as input to the supervisory control level by regenerating the inverse map of the updated process model.