Chemical vapor deposition quality prediction system based on support vector regression and fuzzy learning mechanism

In advanced semiconductor manufacturing, the in-process wafers need to be monitored periodically in order to obtain high stability and high yield rate. However, the actual measurement is usually obtained after all the work-pieces of the same lot have been processed. The parameter drift or shift of the production equipment could not be detected in real-time thereby increasing the production cost. We proposed a quality prediction system (QPS) based on support vector regression (SVR) and fuzzy learning mechanism (FLM) to overcome this problem. The SVR provided good generalization performance for prediction, and the embedded FLM implied a continuous improvement or at least non-degradation of the system performance in an ever changing environment. The effectiveness of the proposed QPS was validated by test on chemical vapor deposition (CVD) process in practical 12-inch wafer fabrication. The results show that the proposed QPS not only fulfills real-time quality measurement of each wafer, but also detects the performance degradation of the corresponding machines from the information of manufacturing process.