Uniformity improvement of etching process using principal component analysis technique

Abstract

The purpose of this study is generated principal component models in order to detect fault in Reactive Ion Etching (RIE) process. The normal data of twenty batches, eleven significant process input variables and six hundred and twenty eight interval time from one machine are selected as database. These data is used for principal component analysis calculation. The result of this study reveals that four principal component models which is fit by data and all principal component models can explain the variance up to 80 percent. The model validation is made on two additional batches; normal and abnormal batch with six hundred and twenty eight time intervals. All the principal component models can detect fault of abnormal batch, while normal batch is under the principal component control limit. The advantage of using principal component models to monitor process input variables of reactive ion, etching is etched depth uniformity improvement. Once four models are tested in production line for almost four weeks, etched depth uniformity has decrease from 1.830 to 1.795 percent.