Outlier distribution detection approach to semiconductor wafer fabrication process monitoring

It is well known that most of the defect clusters found on the fabricated semiconductor wafers have an assignable cause, which if rectified quickly can improve product quality and lower the production cost. This paper proposes a statistical correlation method that extends an existing Automatic Defect Cluster Analysis System (ADCAS). The method can be implemented in real-time such that the manufacturing cost would not be negatively affected. The proposed system generates a list of equipment having a high likelihood of causing the systematic failure. This technique is fast and easy to implement, and it provides early detection and prevention of failures associated with problematic equipment/process during manufacturing.