CN-121616605-B - Wafer defect detection method and wafer defect detection system

CN121616605BCN 121616605 BCN121616605 BCN 121616605BCN-121616605-B

Abstract

The application provides a wafer defect detection method and a wafer defect detection system based on similarity matching among multiple Die images, the wafer defect detection method comprises the steps of establishing a Die position index table, scanning to obtain an original optical image of Dies, defining an image to be detected and a reference image, defining the size of a neighborhood range, constructing a candidate reference image set, constructing a similarity measurement set, defining the number of reference images in Job, obtaining a final reference object, outputting a defect detection result of Job, traversing all Dies on a wafer and outputting a complete defect detection result. According to the application, by dynamically evaluating a plurality of Dies scanned in the same straight line, the optimal reference image is selected in a self-adaptive manner to construct and detect Job, a reference image with higher quality is provided for executing differential calculation, noise signals caused by non-defect differences on the differential image are reduced, the signal-to-noise ratio of real defects in a scanner station, the detection accuracy and stability are improved, and the method is convenient to popularize and apply in the field of semiconductor optical detection.

Inventors

CAI XIONGFEI
CHEN YI
Ke Kexin

Assignees

苏州矽行半导体技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260203

Claims (10)

1. The wafer defect detection method based on the similarity matching among the multiple Die images is characterized by comprising the following steps of: s1, acquiring wafer design layout data, defining the position distribution of Dies on a wafer, and establishing an aligned Die position index table; S2, based on a Die position index table, executing a scanning step, and acquiring original optical images of all Die in the same straight line according to the Die position index table; s3, defining an original optical image to be detected in the current detection flow as an image to be detected based on the original optical image, and taking the rest original optical images with the same straight line as reference images; s4, obtaining the size of a neighborhood range defined by a user; s5, constructing a corresponding candidate reference image set based on the image to be detected, the reference image and the neighborhood range size; s6, respectively calculating similarity measurement values between the image to be detected and the reference image based on the candidate reference image set, and constructing a similarity measurement set; s7, acquiring the number of reference images in the Job defined by the user; s8, based on the similarity measurement value set, the candidate reference image set and the number of reference images in Job, acquiring one or more reference images with highest similarity as a final reference object; S9, constructing and detecting Job based on the final reference object, executing image processing and differential calculation, and outputting a defect detection result of Job; and S10, traversing all Dies on the wafer, repeating the wafer scanning and Job construction flow until the defect detection results of all Dies are output, and finally outputting the complete defect detection result of the wafer.
2. The method of claim 1, wherein creating a Die position index table based on wafer design layout data comprises: s11, defining the position distribution of Die on the wafer to be detected in a software program of a scanner station based on wafer design layout data; s12, acquiring a wafer coordinate system based on wafer design layout data; s13, establishing a Die position index table with numbers based on Die position distribution and a wafer coordinate system.
3. The method for detecting wafer defects according to claim 1, wherein the user defines a size of a neighborhood range, and wherein the size of any neighborhood range is not less than 2.
4. The method of claim 1, wherein the user defines a neighborhood range size such that any neighborhood range size does not exceed half the maximum number of Die in a row in which the current Die is located.
5. The method of claim 1, wherein the calculating the set of similarity metrics comprises: S61, respectively calculating similarity measurement values of the image to be detected and each candidate reference image based on the image to be detected and the candidate reference image set, wherein the similarity measurement values adopt one or more of gray histogram correlation coefficients, structural similarity indexes or normalized cross correlation coefficients; S62, constructing a similarity measurement value set based on the calculated one or more similarity measurement values.
6. The method for detecting wafer defects according to claim 1, wherein the number of reference images in Job is not more than 2 when the user defines the number of reference images in Job.
7. The method for detecting a wafer defect according to claim 1, wherein the step of calculating a defect detection result for detecting Job comprises: S91, based on detection Job, sub-pixel level registration is carried out on the image to be detected and the reference image, and the registered image to be detected and the reference image are obtained; S92, performing differential operation based on the registered image to be detected and the reference image to generate a differential image; S93, executing a user-defined threshold strategy based on the difference graph, and outputting a defect detection result of Job.
8. The method for wafer defect inspection as set forth in claim 7, wherein the user-defined threshold strategy defines a threshold value T and the final defect determination is defined as : In the formula (I), in the formula (II), Indicating the process of the defect arbitration, To mark the potential defect on the differential map, The definition is as follows: Wherein the difference operation , Representing the image to be measured corresponding to Die with index number k, Representing the reference image set after sub-pixel alignment is performed.
9. The method for detecting wafer defects according to claim 1, wherein the method supports real-time online detection, i.e., acquisition-while-matching during scanning, and the similarity calculation and differential analysis can be started without waiting for the completion of the whole line scanning.
10. A wafer defect detection system comprises a wafer motion platform, an image acquisition module and a processor, and is characterized in that the wafer defect detection system controls the movement of the wafer motion platform, the scanning acquisition of the image acquisition module, the image processing of the processor and the defect result output based on the wafer defect detection method according to any one of claims 1-9.

Description

Wafer defect detection method and wafer defect detection system Technical Field The invention belongs to the field of semiconductor detection, relates to a wafer defect detection method and system based on similarity matching among multiple Die images, and particularly relates to a method for adaptively constructing a reference image detection group (Job) in the wafer defect detection process, which is used for improving the accuracy and stability of a differential detection result. Background In the wafer defect detection process, the existing mainstream scanner adopts a progressive scanning strategy, and generally adopts an inter-chip (Die-to-Die) comparison strategy. The Die-to-Die method constructs Job by using the target Die of the same line and the reference Die of the adjacent position during scanning, and identifies potential defects in Job through strategies such as image processing, differential calculation, threshold screening and the like. The method is suitable for ideal conditions, the brightness and the structure between the target Die and the adjacent reference Die on the wafer are highly consistent, and the same area on the adjacent reference Die is free of defects, so that the reference Die can be used as an effective reference, and the difference calculation result can accurately reflect the pattern difference between the contrast Die, so that the True Defect (True Defect) can be efficiently detected. In the actual production or detection process, due to the influence of factors such as the wafer manufacturing process or the equipment state, non-defect differences such as uneven brightness, illumination deviation, focal length difference or pattern distortion often exist between the same line Die on the wafer. This results in a certain difference in gray distribution and texture features etc. between adjacent Die. Using conventional Die-to-Die comparison strategies, the non-defective differences described above can severely interfere with the quality of the difference map (DIFFERNCE MAP), thereby introducing a significant amount of noise signals. Noise signals can seriously influence the signal-to-noise ratio calculation of real defects, so that a large number of noise signals are misjudged to be defects when the real defects are detected, a large number of interference defects (Nuisance) are generated, and the accuracy and the reliability of detection equipment are seriously influenced. Furthermore, although signal noise can be processed in the subsequent algorithm level using means such as image processing and threshold screening, effective true defect distribution is obtained. However, when the noise signal is strong or the real defect signal is weak, the real defect signal is covered by the distribution of the noise signal by using the image processing method, and at this time, the purpose of filtering noise and retaining the real defect is difficult to achieve truly by using the threshold value screening method. The Die-to-Die method builds jobs based on neighboring locations, which are not optimal reference objects in the case of actual production inspection. It is therefore difficult to adaptively cope with the above-mentioned abnormal situation by a strategy of fixedly selecting Die to construct Job. Disclosure of Invention In order to overcome the defects in the prior art, the invention aims to provide a wafer defect detection method and a wafer defect detection system based on similarity matching among multiple Die images, so as to solve the problems in the prior art. The design principle is that based on a wafer defect detection method of similarity matching among multiple Die images, an optimal reference object is selected among the multiple Die to form Job, the difference between the reference image and the image to be detected in Job is reduced, and the accuracy and the stability of a detection algorithm are improved. The overall design is as follows. A wafer defect detection method based on similarity matching among multiple Die images comprises the following steps. S1, acquiring wafer design layout data, defining the position distribution of Die on a wafer, and establishing an aligned Die position index table. S2, based on the Die position index table, executing a scanning step, and acquiring original optical images of all the Die under the same straight line according to the Die position index table. And S3, defining the original optical image to be detected in the current detection flow as an image to be detected based on the original optical image, and taking the rest original optical images with the same straight line as reference images. S4, obtaining the size of the neighborhood range defined by the user. S5, constructing a corresponding candidate reference image set based on the image to be detected, the reference image and the neighborhood range size. S6, based on the candidate reference image set, similarity measurement values between the image to be detected and