Single image detection

• US 10,186,026 B2
• Filed: 11/16/2016
• Issued: 01/22/2019
• Est. Priority Date: 11/17/2015
• Status: Active Grant

• Current Assignee: KLA-Tencor Corporation (KLA Corporation)
• Original Assignee: KLA-Tencor Corporation (KLA Corporation)
• Inventors: Karsenti, Laurent, Bhaskar, Kris, Jordan, III, John Raymond, Venkataraman, Sankar, Carmon, Yair
• Primary Examiner(s): Tucker, Wesley J
• Application Number:
  US15/353,210
• Publication Number:
  US 20170140524A1
• Time in Patent Office: 797 Days
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06F 18/2411 based on the proximity to a decision surface, e.g. support vector machines
• CPC Class Codes: G06F 18/2453 non-linear, e.g. polynomial classifier
• CPC Class Codes: G06T 2207/10061 from scanning electron microscope
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/0004 Industrial image inspection
• CPC Class Codes: G06V 10/50 by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis

Abstract

Methods and systems for detecting defects on a specimen are provided. One system includes a generative model. The generative model includes a non-linear network configured for mapping blocks of pixels of an input feature map volume into labels. The labels are indicative of one or more defect-related characteristics of the blocks. The system inputs a single test image into the generative model, which determines features of blocks of pixels in the single test image and determines labels for the blocks based on the mapping. The system detects defects on the specimen based on the determined labels.

Claims

1. A system configured to detect defects on a specimen, comprising:

an imaging subsystem configured for generating images of a specimen, wherein the imaging subsystem comprises at least an energy source configured to direct energy to the specimen and at least a detector configured to detect energy from the specimen;

one or more computer subsystems coupled to the imaging subsystem, wherein the one or more computer subsystems are configured for acquiring a single test image for a portion of the specimen generated by the imaging subsystem; and

one or more components executed by the one or more computer subsystems, wherein the one or more components comprise;

a generative model, wherein the generative model comprises a non-linear network configured for mapping blocks of pixels of an input feature map volume into labels, and wherein the labels are indicative of one or more defect-related characteristics of the blocks;

wherein the one or more computer subsystems are further configured for inputting the single test image into the generative model;

wherein the generative model is configured for;

separating the single test image into multiple blocks of pixels;

for at least one of the multiple blocks of pixels, determining a feature of the at least one of the multiple blocks based on only the pixels in the at least one of the multiple blocks; and

selecting one of the labels for the at least one of the multiple blocks based on the determined feature and the mapping of the blocks of the pixels of the input feature map volume into the labels; and

wherein the one or more computer subsystems are further configured for detecting defects in the portion of the specimen based on the selected label for the at least one of the multiple blocks.

2. The system of claim 1, wherein the generative model is a deep generative model.

3. The system of claim 1, wherein the generative model is a machine learning model.

4. The system of claim 1, wherein the generative model is a convolution neural network.

5. The system of claim 1, wherein detecting the defects does not comprise aligning the single test image to any other image.

6. The system of claim 1, wherein detecting the defects does not comprise comparing the single test image to any other image.

7. The system of claim 1, wherein detecting the defects does not comprise statistical based defect detection.

8. The system of claim 1, wherein the labels indicate whether input features in the input feature map volume are associated with defects or are not associated with defects.

9. The system of claim 1, wherein the labels indicate a type of defect to which input features in the input feature map volume are associated.

10. The system of claim 1, wherein the one or more computer subsystems are further configured for determining a type of the defects in the portion of the specimen based on the selected label for the at least one of the multiple blocks.

11. The system of claim 1, wherein selecting one of the labels for the at least one of the multiple blocks comprises selecting only one of the labels for a combination of the multiple blocks.

12. The system of claim 1, wherein the one or more computer subsystems are further configured for generating a training dataset used for training the generative model, and wherein the training dataset comprises a set of pairs of a portion of design information for the specimen and an image generated for the portion of the design information.

13. The system of claim 12, wherein at least one of the pairs in the set is generated for a defect detected on the specimen or another specimen.

14. The system of claim 12, wherein at least one of the pairs in the set is generated for a synthetic defect.

15. The system of claim 12, wherein at least one of the pairs in the set is generated for a simulated defect.

16. The system of claim 12, wherein at least one of the pairs in the set is generated for a defect detected by process window qualification.

17. The system of claim 12, wherein the design information comprises design data.

18. The system of claim 12, wherein the design information comprises simulated images generated from design data.

19. The system of claim 12, wherein at least one of the images in the pairs comprises an actual image generated of the specimen or another specimen by the imaging subsystem.

20. The system of claim 12, wherein at least one of the images in the pairs comprises a synthetic image generated based on 1) the design information for the at least one image and 2) other images generated by the imaging subsystem for other specimens.

21. The system of claim 12, wherein at least one of the images in the pairs comprises a simulated image generated based on 1) the design information for the at least one image and 2) one or more characteristics of the imaging subsystem.

22. The system of claim 12, wherein the one or more computer subsystems are further configured for detecting other defects in the images in the pairs in the set and associating labels with pixels in the images corresponding to results of said detecting the other defects.

23. The system of claim 22, wherein the one or more computer subsystems are further configured for performing supervised training of the generative model by progressively modifying parameters of the generative model until features determined by the generative model for the images in the pairs are mapped to the labels associated with the pixels in the images.

24. The system of claim 23, wherein the one or more computer subsystems are further configured for modifying the training dataset by adding a new pair to the set, detecting one or more additional defects in an image in the new pair, associating one or more additional labels with pixels in the image corresponding to results of said detecting the one or more additional defects, and re-performing the supervised training by progressively modifying the parameters until the features determined for the images in the pairs and the new pair by the generative model are mapped to the labels associated with the pixels in the images in the pair and the new pair.

25. The system of claim 1, wherein the one or more computer subsystems are further configured for generating multiple perspectives of the single test image, and wherein the single test image input by the one or more computer subsystems into the generative model comprises the multiple perspectives of the single test image.

26. The system of claim 1, wherein the one or more computer subsystems are further configured for detecting the defects based on the selected label in combination with design information for the specimen.

27. The system of claim 1, wherein the one or more computer subsystems are further configured for detecting the defects based on the selected label and without design information for the specimen.

29. The system of claim 1, wherein the imaging subsystem is an optical based imaging subsystem.

30. The system of claim 1, wherein the imaging subsystem is an inspection subsystem.

31. The system of claim 1, wherein the imaging subsystem is a defect review subsystem.

32. The system of claim 1, wherein the imaging subsystem is a metrology subsystem.

33. The system of claim 1, wherein the specimen is a wafer.

34. The system of claim 1, wherein the specimen is a reticle.

28. The system of claim I, wherein the imaging subsystem is an electron beam based imaging subsystem.

35. A non-transitory computer-readable medium, storing program instructions executable on a computer system for performing a computer-implemented method for detecting defects on a specimen, wherein the computer-implemented method comprises:

separating a single test image generated for a portion of a specimen by an imaging subsystem into multiple blocks of pixels;

for at least one of the multiple blocks of pixels, determining a feature of the at least one of the multiple blocks based on only the pixels in the at least one of the multiple blocks, wherein said separating and said determining are performed by a generative model included in one or more components executed by one or more computer subsystems, wherein the generative model comprises a non-linear network configured for mapping blocks of pixels of an input feature map volume into labels, and wherein the labels are indicative of one or more defect-related characteristics of the blocks;

selecting one of the labels for the at least one of the multiple blocks based on the determined feature and the mapping of the blocks of the pixels of the input feature map volume into the labels; and

detecting defects in the portion of the specimen based on the selected label for the at least one of the multiple blocks, wherein said detecting is performed by the one or more computer subsystems.

36. A computer-implemented method for detecting defects on a specimen, comprising:

separating a single test image generated for a portion of a specimen by an imaging subsystem into multiple blocks of pixels;

for at least one of the multiple blocks of pixels, determining a feature of the at least one of the multiple blocks based on only the pixels in the at least one of the multiple blocks, wherein said separating and said determining are performed by a generative model included in one or more components executed by one or more computer subsystems, wherein the generative model comprises a non-linear network configured for mapping blocks of pixels of an input feature map volume into labels, and wherein the labels are indicative of one or more defect-related characteristics of the blocks;

selecting one of the labels for the at least one of the multiple blocks based on the determined feature and the mapping of the blocks of the pixels of the input feature map volume into the labels; and

detecting defects in the portion of the specimen based on the selected label for the at least one of the multiple blocks, wherein said detecting is performed by the one or more computer subsystems.