Automated malfunction diagnosis of a plasma etcher

The authors present a prototype tool for real-time diagnosis of equipment malfunctions in IC fabrication processes. The approach taken focuses on integrating quantitative empirical models with qualitative knowledge-based methods. The diagnostic system uses evidential reasoning techniques to identify malfunctions by combining various sources of noisy information which originates chronologically from three primary sources: before processing (maintenance diagnosis), during processing (on-line diagnosis), and after processing (in-line diagnosis). The system has been implemented on the Lam Research Autoetch 490 automated plasma etcher located in the Berkeley Microfabrication Laboratory