Cyber-Physical Systems Framework for Predictive Metrology in Semiconductor Manufacturing Process

The process of semiconductor manufacturing is very complex, and the downtime caused by equipment degradation at any process stage reduces yield and production efficiency. Thus, semiconductor machines’ maintenance and calibration are important in maintaining stable production output and yield improvement. Traditionally, semiconductor manufacturers emphasize the importance of product quality to distinguish product defects. However, as semiconductor reaches nanometer-level precision, monitoring process quality is necessary to control process variability and achieve optimal yields. This paper presented a 5-level Cyber Physical Systems (CPS) predictive metrology architecture to investigate the process quality. The proposed architecture performs a peer-to-peer comparison to satisfy self-comparison ability, which is critical for chamber-to-chamber matching in the semiconductor manufacturing process. This concept has been demonstrated using the 2016 PHM public dataset in which the material removal rate was predicted in the chemical-mechanical planarization process. The results show that this innovative 5-level CPS predictive metrology framework is promising and feasible.