Wafer map pattern detection based on supervised machine learning
Various aspects of the disclosed technology relate to training and applying a machine learning model for defect pattern detection. Defect pattern variants of one or more defect patterns are generated. The one or more defect patterns are extracted from wafer maps of wafers having at least systematic defects. Each of the generated defect pattern variants is superimposed on wafer maps of wafers having no systematic defects to generate positive training data of wafer maps, which are included in a training dataset. Based on the training dataset, a trained machine-learning model for recognizing known defect patterns on wafer maps is derived.
1. One or more non-transitory computer-readable media storing computer-executable instructions, the computer-executable instructions, when executed, causing one or more processors to perform a method, the method comprising:
generating defect pattern variants of one or more defect patterns, the one or more defect patterns being extracted from wafer maps of wafers having at least systematic defects, a wafer map of a wafer being a map generated based on results of one or more measurements performed on the wafer;
superimposing each of the defect pattern variants on wafer maps of wafers having no systematic defects to generate positive training data of wafer maps to be included in a training dataset; and
deriving a trained machine-learning model for recognizing known defect patterns on wafer maps based on the training dataset, the known defect patterns comprising the one or more defect patterns.
2. The one or more non-transitory computer-readable media recited in claim 1 , wherein the method further comprises:
analyzing wafer maps for newly-fabricated wafers based on the trained machine-learning model.
3. The one or more non-transitory computer-readable media recited in claim 2 , wherein the method further comprises:
adjusting manufacturing processes, manufacturing equipment or both based on root causes identified from the analyzing.
4. The one or more non-transitory computer-readable media recited in claim 2 , wherein the method further comprises:
identifying dies that pass measurements but are at higher risk of field failure based on failure patterns identified from the analyzing, the measurements comprising the one or more measurements.
5. The one or more non-transitory computer-readable media recited in claim 1 , wherein the machine learning model is constructed based on an artificial neural network.
6. The one or more non-transitory computer-readable media recited in claim 1 , wherein the one or more defect patterns are extracted using an unsupervised machine learning model or a statistical technique.
7. The one or more non-transitory computer-readable media recited in claim 1 , wherein the defect pattern variants comprise rotational variants, location variants, size variant, or any combination thereof.
8. The one or more non-transitory computer-readable media recited in claim 1 , wherein the one or more measurements comprise circuit probe testing.
9. A method, executed by at least one processor of a computer, comprising:
generating defect pattern variants of one or more defect patterns, the one or more defect patterns being extracted from wafer maps of wafers having at least systematic defects, a wafer map of a wafer being a map generated based on results of one or more measurements performed on the wafer;
superimposing each of the defect pattern variants on wafer maps of wafers having no systematic defects to generate positive training data of wafer maps to be included in a training dataset; and
deriving a trained machine-learning model for recognizing known defect patterns on wafer maps based on the training dataset, the known defect patterns comprising the one or more defect patterns.
10. The method recited in claim 9 , wherein the method further comprises:
analyzing wafer maps for newly-fabricated wafers based on the trained machine-learning model.
11. The method recited in claim 10 , wherein the method further comprises:
adjusting manufacturing processes, manufacturing equipment or both based on root causes identified from the analyzing.
12. The method recited in claim 10 , wherein the method further comprises:
identifying dies that pass measurements but are at higher risk of field failure based on failure patterns identified from the analyzing, the measurements comprising the one or more measurements.
13. The method recited in claim 9 , wherein the machine learning model is constructed based on an artificial neural network.
14. The method recited in claim 9 , wherein the one or more defect patterns are extracted using an unsupervised machine learning model or a statistical technique.
15. The method recited in claim 9 , wherein the defect pattern variants comprise rotational variants, location variants, size variant, or any combination thereof.
16. The method recited in claim 9 , wherein the one or more measurements comprise circuit probe testing.
17. A system comprising:
one or more processors, the one or more processors programmed to perform a method, the method comprising:
generating defect pattern variants of one or more defect patterns, the one or more defect patterns being extracted from wafer maps of wafers having at least systematic defects, a wafer map of a wafer being a map generated based on results of one or more measurements performed on the wafer;
superimposing each of the defect pattern variants on wafer maps of wafers having no systematic defects to generate positive training data of wafer maps to be included in a training dataset; and
deriving a trained machine-learning model for recognizing known defect patterns on wafer maps based on the training dataset, the known defect patterns comprising the one or more defect patterns.
18. The system recited in claim 17 , wherein the method further comprises:
analyzing wafer maps for newly-fabricated wafers based on the trained machine-learning model.
19. The system recited in claim 18 , wherein the method further comprises:
adjusting manufacturing processes, manufacturing equipment or both based on root causes identified from the analyzing.
20. The system recited in claim 18 , wherein the method further comprises:
identifying dies that pass measurements but are at higher risk of field failure based on failure patterns identified from the analyzing, the measurements comprising the one or more measurements.
21. The system recited in claim 17 , wherein the machine learning model is constructed based on an artificial neural network.
22. The system recited in claim 17 , wherein the one or more defect patterns are extracted using an unsupervised machine learning model or a statistical technique.
23. The system recited in claim 17 , wherein the defect pattern variants comprise rotational variants, location variants, size variant, or any combination thereof.
24. The system recited in claim 17 , wherein the one or more measurements comprise circuit probe testing.