-
公开(公告)号:US20190155164A1
公开(公告)日:2019-05-23
申请号:US15939311
申请日:2018-03-29
Applicant: Taiwan Semiconductor Manufacturing Co., Ltd.
Inventor: Chien-Huei Chen , Hung-Yi Chung , Chao-Ting Hong , Cheng-Kuang Lee , Xiaomeng Chen , Teng-Cheng Hsu
IPC: G03F7/20 , G06T7/00 , H01J37/28 , G01N21/956 , G01N21/95
Abstract: A defect inspection method and a defect inspection system are provided. In the method, a plurality of candidate defect images are retrieved from inspection images obtained by at least one optical inspection tool performing hot scans on at least one wafer and a plurality of attributes are extracted from the inspection images. A random forest classifier including a plurality of decision trees for classifying the candidate defect images is created, wherein the decision trees are built with different subset of the attributes and the candidate defect images. A plurality of candidate defect images are retrieved from the optical inspection tool in runtime and applied to the decision trees, and classified into nuisance images and real defect images according to votes of the decision trees in which the nuisance images are filtered out. The real defect images with the votes over a confidence value are sampled for microscopic review.
-
公开(公告)号:US10809635B2
公开(公告)日:2020-10-20
申请号:US15939311
申请日:2018-03-29
Applicant: Taiwan Semiconductor Manufacturing Co., Ltd.
Inventor: Chien-Huei Chen , Hung-Yi Chung , Chao-Ting Hong , Cheng-Kuang Lee , Xiaomeng Chen , Teng-Cheng Hsu
Abstract: A defect inspection method and a defect inspection system are provided. In the method, a plurality of candidate defect images are retrieved from inspection images obtained by at least one optical inspection tool performing hot scans on at least one wafer and a plurality of attributes are extracted from the inspection images. A random forest classifier including a plurality of decision trees for classifying the candidate defect images is created, wherein the decision trees are built with different subset of the attributes and the candidate defect images. A plurality of candidate defect images are retrieved from the optical inspection tool in runtime and applied to the decision trees, and classified into nuisance images and real defect images according to votes of the decision trees in which the nuisance images are filtered out. The real defect images with the votes over a confidence value are sampled for microscopic review.
-