Statistical methodology for yield enhancement via baseline reduction

Abstract

In this paper, we introduce a DOE-regression based methodology to identify which tools, in a segment of a fab line between two inspection stations, are defect generating. The approach estimates how much the yield could increase by repairing each of the tools. Furthermore, the tools can be ordered for repair according to this potential yield increase. The estimate of the yield increase includes an evaluation of the power of the statistical test performed to identify the defect generating tools, as well as the kill ratio. By identifying the problem-prone tools and repairing them in the order given by the estimated yield increase, the process baseline can be lowered in an effective manner and the yield increased.