Unsupervised learning-based reference selection for enhanced defect inspection sensitivity

1. An optical characterization system, comprising:

a controller configured to be communicatively coupled with one or more detectors configured to receive illumination from a sample and generate image data, including one or more processors configured to execute program instructions causing the one or more processors to:

receive the image data, wherein the image data comprises images of dies on the sample;

calculate dissimilarity values for all combinations of the images;

perform a cluster analysis to partition the combinations of the images into two or more clusters;

generate a reference image for a cluster of the two or more clusters using two or more of the combinations of the images in the cluster; and

detect one or more defects on the sample by comparing a test image in the cluster to the reference image for the cluster.

2. The system of claim 1 , wherein the one or more processors are configured to execute program instructions causing the one or more processors to:

generate a difference image by subtracting the test image from the reference image.

3. The system of claim 1 , wherein a total number of the combinations is defined by:

n

!

k

⁢

!

(

n

-

k

)

!

wherein n is a total number of the images and k is 2.

4. The system of claim 1 , wherein the sample comprises a semiconductor wafer, a reticle, or a photomask.

5. The system of claim 1 , wherein the one or more processors are configured to execute program instructions causing the one or more processors to align each of the images to others of the images.

6. The system of claim 1 , wherein the dissimilarity values are calculated using a Pearson correlation.

7. The system of claim 1 , wherein the dissimilarity values are calculated using a normalized sum squared difference (NSSD) calculation.

8. The system of claim 1 , wherein the cluster analysis comprises hierarchical clustering.

9. The system of claim 8 , wherein the hierarchical clustering comprises at least one of:

average linkage, single linkage, complete linkage, or

centroid linkage.

10. The system of claim 1 , wherein the cluster analysis comprises k-means clustering.

11. An optical characterization method, comprising:

receiving illumination from a sample using one or more detectors;

generating image data;

receiving the image data, wherein the image data comprises images of dies on the sample;

calculating dissimilarity values for all combinations of the images;

performing a cluster analysis to partition the combinations of the images into two or more clusters;

generating a reference image for a cluster of the two or more clusters using two or more of the combinations of the images in the cluster; and

detecting one or more defects on the sample by comparing a test image in the cluster to the reference image for the cluster.

12. The method of claim 11 , further comprising:

generating a difference image by subtracting the test image from the reference image.

13. The method of claim 11 , wherein a total number of the combinations is defined by:

n

!

k

⁢

!

(

n

-

k

)

!

wherein n is a total number of the images and k is 2.

14. The method of claim 11 , wherein the sample comprises a semiconductor wafer, a reticle, or a photomask.

15. The method of claim 11 , comprising aligning each of the images to others of the images.

16. The method of claim 11 , wherein the dissimilarity values are calculated using a Pearson correlation.

17. The method of claim 11 , wherein the dissimilarity values are calculated using normalized sum squared difference (NSSD) calculation.

18. The method of claim 11 , wherein the cluster analysis comprises hierarchical clustering.

19. The method of claim 18 , wherein the hierarchical clustering comprises at least one of:

average linkage, single linkage, complete linkage, or

centroid linkage.

20. The method of claim 11 , wherein the cluster analysis comprises k-means clustering.

21. An optical characterization system, comprising:

a controller configured to be communicatively coupled with one or more detectors configured to receive illumination from a sample and generate image data, including one or more processors configured to execute program instructions causing the one or more processors to:

receive the image data, wherein the image data comprises images of dies on the sample;

identify edge die images from the images of dies on the sample;

generate a reference edge die image using a first edge die image from the edge die images;

calculate dissimilarity values for all combinations of the reference edge die image and others of the edge die images;

perform a cluster analysis to partition the combinations into two or more clusters;

generate a reference image for a cluster of the two or more clusters using two or more of the combinations of the edge die images in the cluster; and

detect one or more defects on the sample by comparing a test image in the cluster to the reference image for the cluster.

22. The system of claim 21 , wherein the one or more processors are configured to execute program instructions causing the one or more processors to:

generate a difference image by subtracting the test image from the reference image.

23. The system of claim 21 , wherein the dissimilarity values are calculated using a Pearson correlation.

24. The system of claim 21 , wherein the dissimilarity values are calculated using normalized sum squared difference (NSSD) calculation.

25. The system of claim 21 , wherein the cluster analysis comprises hierarchical clustering.

26. An optical characterization system, comprising:

a controller configured to be communicatively coupled with one or more detectors configured to receive illumination from a sample and generate image data, including one or more processors configured to execute program instructions causing the one or more processors to:

receive the image data, wherein the image data comprises images of dies on the sample;

identify a test image;

calculate dissimilarity values for all combinations of the test image and others of the images;

identify two or more of the combinations having a dissimilarity value below a threshold dissimilarity value;

generate a reference image using the two or more of the combinations; and

detect one or more defects on the sample by comparing the reference image to the test image.

27. The system of claim 26 , wherein the dissimilarity values are calculated using normalized sum squared difference (NSSD) calculation.