A wafer map yield model based on deep learning for wafer productivity enhancement

In semiconductor manufacturing, evaluating the productivity of wafer maps prior to fabrication for designing an optimal wafer map is one of the most effective solutions for enhancing productivity. However, a yield prediction model is required to accurately evaluate the productivity of wafer maps since the design of a wafer map affects yield. In this paper, we propose a novel yield prediction model based on deep learning algorithms. Our approach exploits spatial relationships among positions of dies, sizes of dies, and die-level yield variations collected from a wafer test. By modeling these spatial features, the accuracy of yield prediction significantly increased. Furthermore, experimental results showed that the proposed yield model and approach help to design a wafer map with higher productivity nearly 13%.