Composite defect classifier

• US 9,430,743 B2
• Filed: 03/03/2015
• Issued: 08/30/2016
• Est. Priority Date: 03/06/2014
• Status: Active Grant

• Current Assignee: KLA-Tencor Corporation
• Original Assignee: KLA-Tencor Corporation
• Inventors: Plihal, Martin
• Primary Examiner(s): Vu, Kim
• Assistant Examiner(s): VANCHY JR, MICHAEL J
• Application Number:
  US14/637,242
• Publication Number:
  US 20150254832A1
• Time in Patent Office: 546 Days
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/025 Extracting rules from data
• CPC Class Codes: G06T 2207/10061 from scanning electron microscope
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/001 using an image reference approach

Abstract

Methods and systems for classifying defects detected on a wafer are provided. One method includes inputting information for defects detected on a wafer into each of at least two defect classifiers included in a composite defect classifier. Such a method also includes, for at least one of the defects that is assigned to two or more bins in the composite defect classifier, determining a bin for the at least one of the defects based on a rank assigned to the two or more bins. The rank is assigned to the two or more bins based on one or more characteristics determined for the two or more bins, and the one or more characteristics are determined based on a comparison of predetermined defect classifications for defects in a training set and defect classifications determined for the defects in the training set by the at least two defect classifiers.

Claims

1. A computer-implemented method for classifying defects detected on a wafer, comprising:

acquiring a training set of data for defects detected on a training wafer and defect classifications determined for the defects in the training set;

performing two or more defect classification processes by inputting the training set of data into two or more defect classifiers;

comparing defect classification results produced by the two or more defect classifiers to the defect classifications determined for the defects in the training set;

determining one or more characteristics of one or more bins in the two or more defect classifiers based on results of said comparing;

assigning a rank to the one or more bins in the two or more defect classifiers based on the determined one or more characteristics;

selecting at least two of the two or more defect classifiers to be included in a composite defect classifier based on the rank assigned to the one or more bins;

acquiring inspection results for the wafer using an inspection system, wherein the inspection results comprise information for defects detected on the wafer; and

classifying the defects detected on the wafer by inputting the information for the defects detected on the wafer into each of the at least two defect classifiers selected to be included in the composite defect classifier and, for at least one of the defects that is assigned to two or more bins in the composite defect classifier, determining a bin for the at least one of the defects based on the rank assigned to the two or more bins, wherein said acquiring the training set, said performing, said comparing, said determining, said assigning, said selecting, said acquiring the inspection results, and said classifying are performed by a computer system.

2. The method of claim 1, wherein the defects in the training set comprise defects of interest and nuisances.

3. The method of claim 1, wherein said performing comprises tuning the two or more defect classifiers based on the defect classifications determined for the defects in the training set.

4. The method of claim 1, further comprising tuning at least one of the two or more defect classifiers prior to said assigning using a set of modified Bayesian priors to produce a plot illustrating performance of the at least one defect classifier.

5. The method of claim 1, wherein the one or more characteristics comprise purities of the defects in the training set assigned to the one or more bins in the two or more defect classifiers, and wherein the rank assigned to the one or more bins is based on the purities such that higher ranks correspond to higher purities.

6. The method of claim 1, wherein the one or more characteristics comprise stabilities of the defects in the training set assigned to the one or more bins in the two or more defect classifiers, and wherein the rank assigned to the one or more bins is based on the stabilities such that higher ranks correspond to higher stabilities.

7. The method of claim 1, wherein the one or more characteristics comprise reliability of at least one of the two or more defect classifiers to separate the defects in the training set into the one or more bins correctly, and wherein the rank assigned to the one or more bins is based on the reliability such that higher ranks correspond to higher reliabilities.

8. The method of claim 1, wherein said selecting comprises selecting a subset of the two or more defect classifiers that include the one or more bins assigned the highest rank for the largest set of defects of interest in the training set of defects.

9. The method of claim 1, wherein the rank is responsive to separation between defects of interest in the training set and nuisances in the training set in the one or more bins of the two or more defect classifiers, and wherein said selecting comprises selecting a minimum number of the two or more defect classifiers that produce maximum defect of interest separation based on the ranks assigned to the one or more bins.

10. The method of claim 1, further comprising generating the composite defect classifier by storing information for the at least two selected defect classifiers with the rank assigned to the one or more bins included in the at least two selected defect classifiers.

11. The method of claim 1, further comprising generating the composite defect classifier based on the at least two selected defect classifiers, wherein generating the composite defect classifier does not comprise altering any of the at least two selected defect classifiers.

12. The method of claim 1, further comprising generating the composite defect classifier based on the at least two selected defect classifiers, wherein generating the composite defect classifier does not comprise tuning any of the at least two selected defect classifiers.

13. The method of claim 1, further comprising generating the composite defect classifier based on the at least two selected defect classifiers, wherein generating the composite defect classifier does not comprise combining the at least two selected defect classifiers into a sequence of defect classifiers.

14. The method of claim 1, wherein, for the at least one of the defects that is assigned to the two or more bins in the composite defect classifier, determining the bin for the at least one of the defects based on the rank assigned to the two or more bins comprises comparing the rank of the two or more bins and assigning the at least one of the defects to one of the two or more bins having the highest rank.

15. A computer-implemented method for classifying defects detected on a wafer, comprising:

acquiring inspection results for the wafer using an inspection system, wherein the inspection results comprise information for defects detected on the wafer; and

classifying the defects detected on the wafer by;

inputting the information for the defects detected on the wafer into each of at least two defect classifiers included in a composite defect classifier; and

for at least one of the defects that is assigned to two or more bins in the composite defect classifier, determining a bin for the at least one of the defects based on a rank assigned to the two or more bins;

wherein the rank is assigned to the two or more bins based on one or more characteristics determined for the two or more bins,

wherein the one or more characteristics are determined based on a comparison of predetermined defect classifications for defects in a training set and defect classifications determined for the defects in the training set by the at least two defect classifiers, and

wherein said acquiring and said classifying are performed by a computer system.

16. A non-transitory computer-readable medium, storing program instructions executable on a computer system for performing a computer-implemented method for classifying defects detected on a wafer, wherein the computer-implemented method comprises:

acquiring a training set of data for defects detected on a training wafer and defect classifications determined for the defects in the training set;

performing two or more defect classification processes by inputting the training set of data into two or more defect classifiers;

comparing defect classification results produced by the two or more defect classifiers to the defect classifications determined for the defects in the training set;

determining one or more characteristics of one or more bins in the two or more defect classifiers based on results of said comparing;

assigning a rank to the one or more bins in the two or more defect classifiers based on the determined one or more characteristics;

selecting at least two of the two or more defect classifiers to be included in a composite defect classifier based on the rank assigned to the one or more bins;

acquiring inspection results for the wafer using an inspection system, wherein the inspection results comprise information for defects detected on the wafer; and

classifying the defects detected on the wafer by inputting the information for the defects detected on the wafer into each of the at least two defect classifiers selected to be included in the composite defect classifier and, for at least one of the defects that is assigned to two or more bins in the composite defect classifier, determining a bin for the at least one of the defects based on the rank assigned to the two or more bins.

17. A system configured to classify defects detected on a wafer, comprising:

an inspection subsystem configured to acquire inspection results for the wafer, wherein the inspection results comprise information for defects detected on the wafer; and

a computer subsystem configured for;

acquiring a training set of data for defects detected on a training wafer and defect classifications determined for the defects in the training set;

performing two or more defect classification processes by inputting the training set of data into two or more defect classifiers;

comparing defect classification results produced by the two or more defect classifiers to the defect classifications determined for the defects in the training set;

determining one or more characteristics of one or more bins in the two or more defect classifiers based on results of said comparing;

assigning a rank to the one or more bins in the two or more defect classifiers based on the determined one or more characteristics;

selecting at least two of the two or more defect classifiers to be included in a composite defect classifier based on the rank assigned to the one or more bins; and

classifying the defects detected on the wafer by inputting the information for the defects detected on the wafer into each of the at least two defect classifiers selected to be included in the composite defect classifier and, for at least one of the defects that is assigned to two or more bins in the composite defect classifier, determining a bin for the at least one of the defects based on the rank assigned to the two or more bins.

18. The system of claim 17, wherein the defects in the training set comprise defects of interest and nuisances.

19. The system of claim 17, wherein said performing comprises tuning the two or more defect classifiers based on the defect classifications determined for the defects in the training set.

20. The system of claim 17, wherein the computer subs stem is further configured for tuning at least one of the two or more defect classifiers prior to said assigning using a set of modified Bayesian priors to produce a plot illustrating performance of the at least one defect classifier.

21. The system of claim 17, wherein the one or more characteristics comprise purities of the defects in the training set assigned to the one or more bins in the two or more defect classifiers, and wherein the rank assigned to the one or more bins is based on the purities such that higher ranks correspond to higher purities.

22. The system of claim 17, wherein the one or more characteristics comprise stabilities of the defects in the training set assigned to the one or more bins in the two or more defect classifiers, and wherein the rank assigned to the one or more bins is based on the stabilities such that higher ranks correspond to higher stabilities.

23. The system of claim 17, wherein the one or more characteristics comprise is reliability of at least one of the two or more defect classifiers to separate the defects in the training set into the one or more bins correctly, and wherein the rank assigned to the one or more bins is based on the reliability such that higher ranks correspond to higher reliabilities.

24. The system of claim 17, wherein said selecting comprises selecting a subset of the two or more defect classifiers that include the one or more bins assigned the highest rank for the largest set of defects of interest in the training set of defects.

25. The system of claim 17, wherein the rank is responsive to separation between defects of interest in the training set and nuisances in the training set in the one or more bins of the two or more defect classifiers, and wherein said selecting comprises selecting a minimum number of the two or more defect classifiers that produce maximum defect of interest separation based on the ranks assigned to the one or more bins.

26. The system of claim 17, wherein the computer subsystem is further configured fir generating the composite defect classifier by storing information for the at least two selected defect classifiers with the rank assigned to the one or more bins included in the at least two selected defect classifiers.

27. The system of claim 17, wherein the computer subsystem is further configured for generating the composite defect classifier based on the at least two selected defect classifiers, and wherein generating the composite defect classifier does not comprise altering any of the at least two selected defect classifiers.

28. The system of claim 17, wherein the computer subsystem is further configured for generating the composite defect classifier based on the at least two selected defect classifiers, and wherein generating the composite defect classifier does not comprise tuning any of the at least two selected defect classifiers.

29. The system of claim 17, wherein the computer subsystem is further configured for generating the composite defect classifier based on the at least two selected defect classifiers, and wherein generating the composite defect classifier does not comprise combining the at least two selected defect classifiers into a sequence of defect classifiers.

30. The system of claim 17, wherein, for the at least one of the defects that is assigned to the two or more bins in the composite defect classifier, determining the bin for the at least one of the defects based on the rank assigned to the two or more bins comprises comparing the rank of the two or more bins and assigning the at least one of the defects to one of the two or more bins having the highest rank.