Real-time Fault Detection System for Multiphase Plasma Etching Process using OES, Two-Step Division and Change Stage Alignment Method

Abstract

For plasma etching process, developing efficient real-time fault detection systems is an active research area of semiconductor manufacturing, especially the systems using Optical Emission Spectroscopy(OES) data. Besides, as a typical multiphase process, it is promising to apply multiphase methods in the OES fault detection system. However, the process drifts existing in the system will lead to severe drifts in time direction and cause model mismatch issue when applying the multiphase methods. To cope with these problems, a real-time fault detection system for multiphase plasma etching using OES, two-step division and change stage alignment method is proposed in this paper. By capturing changes in process, a two-step division method divides the process into stationary phases and transition phases. To better overcome the drift problem, change stage in transition phase is further aligned before MPCA modeling and online monitoring. Case studies have shown that the proposed system is more reliable and sensitive comparing to the reference model and can better overcome the model mismatch problem.