With the introduction of the dual damascene metal technology, control of the chemical-mechanical polishing (CMP) process becomes more sophisticated. Not only the removal rate and uniformity are concerned in the polishing process, but the control of the dishing and erosion problems occurring in the CMP process are also important issues. In this study, we present an architecture based on a system point of view, in which we combined the innovative in-situ Kalman filter based endpoint detection technology proposed by Yueh [1] and the advanced artificial neural network algorithm [2]. Details of the system modeling and calibration for the endpoint are presented, and then the approach to the multi-step artificial neural network polishing control for reducing the dishing and erosion in the copper CMP will be introduced.
展开▼