Implementing virtual metrology for in-line quality control in semiconductor manufacturing

Abstract

The in-line quality control measured data plays an increasingly important role in both the control and the improvement of yields in the production process of semiconductor manufacturing, especially in 300 mm semiconductor factories. Thus, a breakthrough in the measuring methodology associated with wafer level production is highly desirable. This article addresses this issue and proposes an extended concept of virtual metrology (VM) that utilises the status variables identification of process equipments to predict process quality of wafer level. In this article, we will discuss two modelling techniques, namely multiple linear regression and partial least square regression, to build prediction models for the film thickness and critical dimensions for the on-line production of semiconductor manufacturing. This article proposes to use statistical process control rules to enhance the credibility of the VM applications based on the tendency and characteristics of the residuals of the VM models. Finally, this study presents an architecture for the control and alarm mechanism of the VM approach in semiconductor manufacturing and illustrates the results on the application of this architecture with two real cases as well.