Novel Methods for Identification and Analysis of Various Yield Problems in Semiconductor Manufacturing

Abstract

Overwhelming data is produced during semiconductor processing and it becomes more important to classify a large number of wafers into various types of failures for the root cause analysis of the yield excursion as quickly as possible. In this paper, feature vector based methods have been suggested for the classification of wafers and their application to the root cause analysis. Local bin profile has been calculated to generate a feature vector for a wafer. K-means clustering method has been used to cluster these vectors for the classification of wafers. ANOVA or Kruscal-Wallis test has been applied to one of the components of a feature vector for the yield analysis, depending on its normality. Our yield analysis examples have proven that these analysis methods are very effective and quick in pinpointing the root cause for the various types of failures, especially the equipment-originated ones, including those otherwise would be impossible with the conventional methods