Modeling semiconductor manufacturing yield by test data and partial least squares

With the shrinking of transistor sizes, semiconductor manufacturing production costs are getting higher and higher. Predicting manufacturing yield helps engineers to take decision earlier in the manufacturing process flow. This paper presents a method to model the yield from measurements obtained during the production. A common method is the use of principal component analysis (PCA) coupled to response surface modeling (RSM). In this paper, the use of partial least square (PLS) regression methods is introduced. The obtained results are compared to PCA/RSM.