Neural network based control strategies for improving plasma characteristics in reactive ion etching

Reactive ion etching (RIE) is a key process in VLSI circuit fabrication. It is known that this process is not very robust and requires frequent tuning to achieve acceptable yields. The research reported in this paper is motivated by the possibility of designing control techniques to significantly improve the performance of the RIE process. The research is broadly focused on the application of a real-time neural network based control technique. In this paper, we develop multivariable real-time intelligent control strategies to improve plasma characteristics in the RIE process. The objective is to demonstrate that by controlling appropriate key plasma parameters namely the fluorine concentration ions [F][F] and their energy [Vbias][Vbias] through neurocontrol strategies, it is feasible to improve the etch performance of the reactive ion etchers characterized by their selectivity, uniformity, anisotropy and etch depth. We propose robust neurocontrollers for this problem and examine their effectiveness in attenuating the effects of exogenous disturbances on etch characteristics. The nonlinear nature and robustness of the proposed intelligent control algorithm enables it to provide greater accuracy and better performance when compared to standard algorithms from the RIE literature such as the conventional PI controller.