DATA ANALYTICS FOR QUALITY CONTROL IN SEMICONDUCTOR MANUFACTURING PROCESSES

Abstract

In semiconductor manufacturing, various test structures are collected on each wafer to provide information for process monitoring and improvement. The analysis of these data plays an important role in identifying potential problems and improving process yield. This thesis focuses on the analysis of both parametric and functional data collected during the manufacturing process. For parametric test results, process monitoring is implemented to detect anomalies and raise early alarms. A mixed-effects model incorporating with Gaussian process is proposed to account for the process variations. For functional test results, the identification of defect patterns in the wafer bin map (WBM) help reveal the root causes and then solve them accordingly. This thesis develops a feature engineering method as well as a convolutional neural network (CNN) based method to recognize single defect pattern on WBMs; and tensor voting is used to solve mixed-type defect pattern recognition tasks.