Recognition of semiconductor defect patterns using spatial filtering and spectral clustering

Diverse defect patterns shown on the wafer map usually contain important information for quality engineers to find their root causes of abnormalities. Today, even with highly automated and precisely monitored facilities used in a near dust-free clean room and operated with well-trained process engineers, the occurrence of spatial defects still cannot be avoided. This research presents a spatial defect diagnosis system and attempts to solve two challenging problems for semiconductor manufacturing: (1) to estimate the number of clusters in advance, and (2) to separate both convex and non-convex defect clusters at the same time. In this paper, a spatial filter is used to denoise the noisy wafer map and to extract meaningful defect clusters. To isolate various types of defect patterns, a hybrid scheme combining entropy fuzzy c means (EFCM) with spectral clustering is applied to the denoised output. Furthermore, a decision tree based on two cluster features (convexity and eigenvalue ratio) is constructed to identify the specific defect type and to provide decision support for quality engineers. The proposed approach is validated with an empirical wafer bin maps obtained in a DRAM company in Taiwan. Experimental results show that four kinds of mixed-type defect patterns are successfully extracted and classified. More importantly, the proposed method is very promising to be further applied to other industries, such as liquid crystal or plasma display.