CLASSIFICATION OF SPATIAL PATTERNS ON WAFER MAPS

• US 20100057391A1
• Filed: 08/29/2008
• Published: 03/04/2010
• Est. Priority Date: 08/29/2008
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Tuv, Eugene, Borisov, Alexander, St. Pierre, Eric R.
• Granted Patent: US 7,937,234 B2
• Time in Patent Office: Days
• US Class Current: 702/81
• CPC Class Codes: H01L 22/14 for electrical parameters, e.g. resistance, deep-levels, CV, diffusions by electrical means
• CPC Class Codes: H01L 22/20 Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps
• CPC Class Codes: H01L 2924/00 Indexing scheme for arrangements or methods for connecting or disconnecting semiconductor or solid-state bodies as covered by H01L24/00
• CPC Class Codes: H01L 2924/0002 Not covered by any one of groups H01L24/00, H01L24/00 and H01L2224/00

Abstract

Classification of spatial patterns on wafer maps is generally described. In one example, a method includes applying K-means type clustering to wafer maps comprising one or more spatial patterns to group one or more clusters comprising wafer maps having similar spatial patterns and producing a dendrogram using a clustering process to display the one or more clusters.

Claims

1. A method comprising:

applying K-means type clustering to wafer maps comprising one or more spatial patterns to group one or more clusters comprising wafer maps having similar spatial patterns; and

producing a dendrogram using a clustering process to display the one or more clusters.

2. A method according to claim 1 wherein producing the dendrogram comprises using agglomerative clustering to allow a user to visually select final cluster assignment of the wafer maps.

3. A method according to claim 1 further comprising:

comparing the wafer maps with a wafer map comprising a generated random spatial pattern to filter out wafer maps that are similar to the wafer map comprising the generated random spatial pattern prior to applying the K-means type clustering.

4. A method according to claim 3 wherein comparing the wafer maps with the wafer map comprising a generated random spatial pattern comprises using a chi-square test wherein a p-value of the chi-square test is user-selectable.

5. A method according to claim 1 further comprising:

filtering out wafer maps based on factory tool and run date by applying an Apriori-type process; and

ranking the wafer maps that are not filtered out based on factory tool and run date by an ascending pareto-bound distance.

6. A method according to claim 5 wherein filtering out wafer maps based on factory tool and run date comprises using conditional support wherein nodes of the Apriori-type process comprise pairs of factory tool and run date wherein the run date is nested within the factory tool, and wherein ranking the wafer maps by the ascending pareto-bound distance is calculated on a conditional support-confidence plane.

7. A method according to claim 1 further comprising:

assigning a binary value to one or more die locations of the wafer maps to indicate whether the die location is above or below a threshold value;

filtering out wafer maps comprising spatial pattern densities that are above or below user-selectable values; and

smoothing the wafer maps using kernel-based smoothing prior to applying the K-means type clustering to the wafer maps.

8. A method according to claim 1 further comprising:

initializing a number of cluster seeds for the K-means type clustering prior to applying the K-means type clustering process to the wafer maps wherein the number of cluster seeds is proportional to a number of wafer maps in a data set and wherein the cluster seeds are initialized to have a large sum total distance between cluster seeds; and

filtering out clusters that contain less than a user-specified minimum of wafer maps prior to producing the dendrogram.

9. A method according to claim 1 further comprising:

refining the one or more clusters by comparing the wafer maps of the one or more clusters with a signature spatial pattern to remove wafer maps that are not similar to the signature spatial pattern; and

assigning a binary value to one or more lots to indicate whether the one or more lots comprise a wafer map from a user-determined cluster.

10. A method according to claim 1 wherein the method increases the speed of spatial pattern detection for semiconductor yield analysis or increases the speed of detection of factory tools that produce defective spatial patterns, or combinations thereof, and wherein the method does not require spatial pattern templates prior to applying the K-means type clustering to the wafer maps.

13. An article of manufacture according to claim 11, wherein the instructions, if executed, further result in:

comparing the wafer maps with a wafer map comprising a generated random spatial pattern to filter out wafer maps that are similar to the wafer map comprising the generated random spatial pattern prior to applying the K-means type clustering.

14. An article of manufacture according to claim 13 wherein comparing the wafer maps with the wafer map comprising a generated random spatial pattern comprises using a chi-square test wherein a p-value of the chi-square test is user-selectable.

11. An article of manufacture comprising a storage medium having instructions stored thereon that, if executed, result in:

applying K-means type clustering to wafer maps comprising one or more spatial patterns to group one or more clusters comprising wafer maps having similar spatial patterns; and

producing a dendrogram using a clustering process to display the one or more clusters.

12. An article of manufacture according to claim 11 wherein producing the dendrogram comprises using agglomerative clustering to allow a user to visually select final cluster assignment of the wafer maps.

15. An article of manufacture according to claim 11, wherein the instructions, if executed, further result in:

filtering out wafer maps based on factory tool and run date by applying an Apriori-type process; and

ranking the wafer maps that are not filtered out based on factory tool and run date by an ascending pareto-bound distance.

16. An article of manufacture according to claim 15 wherein filtering out wafer maps based on factory tool and run date comprises using conditional support wherein nodes of the Apriori-type process comprise pairs of factory tool and run date wherein the run date is nested within the factory tool, and wherein ranking the wafer maps by the ascending pareto-bound distance is calculated on a conditional support-confidence plane.

17. An article of manufacture according to claim 11, wherein the instructions, if executed, further result in:

assigning a binary value to one or more die locations of the wafer maps to indicate whether the die location is above or below a threshold value;

filtering out wafer maps comprising spatial pattern densities that are above or below user-selectable values; and

smoothing the wafer maps using kernel-based smoothing prior to applying the K-means type clustering to the wafer maps.

18. An article of manufacture according to claim 11, wherein the instructions, if executed, further result in:

initializing a number of cluster seeds for the K-means type clustering prior to applying the K-means type clustering to the wafer maps wherein the number of cluster seeds is proportional to a number of wafer maps in a data set and wherein the cluster seeds are initialized to have a large sum total distance between cluster seeds; and

filtering out clusters that contain less than a user-specified minimum of wafer maps prior to producing the dendrogram.

19. An article of manufacture according to claim 11, wherein the instructions, if executed, further result in:

refining the one or more clusters by comparing the wafer maps of the one or more clusters with a signature spatial pattern to remove wafer maps that are not similar to the signature spatial pattern;

assigning a binary value to one or more lots to indicate whether the one or more lots comprise a wafer map from a user-determined cluster.

20. An article of manufacture according to claim 11 wherein applying the K-means type clustering to the wafer maps or producing the dendrogram, or combinations thereof, increase the speed of spatial pattern detection for semiconductor yield analysis or increase the speed of detection of factory tools that produce defective spatial patterns, or combinations thereof, and wherein spatial pattern templates are not required prior to applying the K-means type clustering to the wafer maps.