Real-time predictive control of semiconductor manufacturing processes using neural networks

As a result of consistent demands on semiconductor manufacturers to produce circuits with increased density and complexity, stringent process control has become an issue of growing importance in this industry. Earlier work has shown that neural networks offer great promise in modeling complex fabrication processes such as reactive ion etching (RIE). Motivated by these results, this paper explores the use of neural networks for real-time, model-based control of semiconductor manufacturing processes. This objective is accomplished in part by constructing a q-step ahead predictive model for the system, which can be inverted (or approximately inverted) to achieve the desired control. The efficacy of this approach is demonstrated: (1) using a process simulated by a nonlinear equation; (2) using experimental input/output data from an actual RIE process to examine run-by-run control; and (3) by performing real-time, one-step ahead predictive control of a dynamic process which reflects typical RIE behavior