טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
M.Sc Thesis
M.Sc StudentZontak Maria
SubjectDefect Detection in Patterned Silicon Wafers Using
Anisotropic Kernels
DepartmentDepartment of Electrical Engineering
Supervisor Professor Cohen Israel
Full Thesis textFull thesis text - English Version


Abstract

This work is focused on the application of anisotropic kernels to defect detection in patterned wafers using Scanning Electron Microscope (SEM) images.

Defect detection is a critical component of wafers manufacturing process. Various image processing techniques have been applied to automatic defect detection in wafers, which rely on accurate image registration of source and reference images obtained from neighboring dies. Unfortunately, perfect registration is generally impossible, due to pattern variations between the source and reference images.

In this work, we propose a defect detection procedure for a single image, which avoids image registration and is robust to pattern variations. The proposed method is based on anisotropic kernel reconstruction of the source image using the reference image. The source and reference images are mapped into a feature space, where every feature with origin in the source image is estimated by a weighted sum of neighboring features from the reference image. The set of neighboring features is determined according to the spatial neighborhood in the original image space, and the weights are calculated from exponential distance similarity function. We show that features originating from defect regions could not be reconstructed from the reference image, and hence can be identified.

We develop a kernel-based approach to multi-channel defect detection, which relies on the physical similarity relations between multi-channel images acquired by an SEM tool. We assume that the similarity between pattern-originated regions from the inspected and reference wafers is maintained across the three SEM channels and develop a multi-channel defect detection algorithm. We show that in case of pattern variations, the false detection rate can be significantly reduced by using our kernel-based approach rather than existing methods. We also demonstrate the improved performance of the multi-channel approach in case of non-periodic patterned wafers, compared to using a single-channel approach.