Classification of mixed-type defects patterns in wafer bin maps using convolutional neural networks

Abstract

In semiconductor manufacturing, a wafer bin map (WBM) represents the results of wafer tests using a binary value according to pass or fail of the tests. Defective dies are often clustered together, forming local systematic defects. Determining the specific pattern of local systematic defects is important, because different patterns of local systematic defects are related to different root causes of failure. Recently, as the wafer size has increased and the process technology has become more complicated, the possibility of observing two or more defect patterns in a single wafer has increased. In this study, we propose the use of convolutional neural networks to classify mixed-type defect patterns in WBMs. Through simulated and real data examples, we show that CNN performs well even when there are many random defects in WBMs.