Monitoring Defects in IC Fabrication Using a Hotelling<tex>$T^2$</tex>Control Chart

Abstract

Monitoring the wafer defects in integrated circuit (IC) fabrication is essential for enhancing wafer yield. However, significant defect clustering occurs when the wafer is large, so the conventional defect control chart, based on the Poisson distribution, is inappropriate. Defect clustering must also be analyzed to monitor effectively defects in IC fabrication process control. This study developed a novel procedure using the multivariate Hotelling T/sup 2/ control chart, based on the number of defects and the defect clustering index (CI) to monitor simultaneously the number of defects and the defect clusters. The CI does not require any statistical assumptions concerning the distribution of defects and can accurately evaluate the clustering phenomena. A case study of a Taiwanese IC manufacturer demonstrates the effectiveness of the proposed procedure.