Adaptive extensions to a multibranch run-to-run controller for plasma etching

Publisher Website
Google Scholar
Abstract
Fuzzy logic and database learning mechanisms have been incorporated into a generic plasma etching run-to-run controller, resulting in a very dynamic, adaptable and robust system. The system features an Applied 8300 RIE controlled by a Techware II equipment controller. A TCP/IP connection links this equipment controller to the run-to-run controller residing on a SUN. The run-to-run control environment is generic in that the basic control framework and controller development results are applicable to VLSI manufacturing in general. The controller is multi-branch as it utilizes multiple algorithms in complementary fashion to achieve process optimization and control. The current implementation utilizes three branches: (1) a linear approximation control algorithm, (2) an optimization algorithm that utilizes (real-time) data collected in-situ to determine optimal run-to- run process parameter settings, and (3) a statistical optimization algorithm that utilizes run-to-run data. We have extended the controller to accommodate an automated branch selection process that utilizes fuzzy logic to incorporate process engineer as well as optimization and control algorithmic knowledge. We have also extended the controller to adapt to unforeseen events through utilization of a learning mechanism; this mechanism detects these unforeseen events, intelligently queries the process engineer, guides the engineer through the development of an event servicing scheme, and incorporates this new knowledge into its control knowledge base so that the event may be serviced automatically in the future. Implementation results of the controller (in the control of the etcher) confirm the robust control capabilities in the face of process shift and drift, and recipe change.

This publication has 0 references indexed in Scilit: