Use of contagious distributions in the semiconductor yield models considering cluster effect

Abstract

As semiconductor chips become larger, more than one cluster can be expected to form on a single chip and consequently defect density variations may be apparent not only among chips but also within a chip. In order to reflect this situation more effectively, we propose semiconductor yield models based on the contagious distributions. These models incorporate the probability that a defect or a cluster becomes fatal. Furthermore, we consider various limiting forms of the proposed contagious distribution by taking extreme values of the related parameters. A simplified form of the proposed yield formula based on the limiting distribution is also derived and its performance is compared with others using computer simulation. The proposed three yield models are shown in a simulation study to outperform alternative models based on the existing ideas.