Universal spatial pattern recognition system

• US 7,106,897 B1
• Filed: 04/29/2002
• Issued: 09/12/2006
• Est. Priority Date: 04/29/2002
• Status: Active Grant

• Current Assignee: Advanced Micro Devices, Inc.
• Original Assignee: Advanced Micro Devices, Inc.
• Inventors: McIntyre, Michael G., Morris, James E.
• Primary Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US10/134,718
• Time in Patent Office: 1,597 Days
• Field of Search: 382/149, 382/141, 382/145, 382/156, 382/168, 382/173, 382/199, 382/224, 382/225, 382/254, 382/278, 382/299, 382/305, 716/4, 716/19, 257/E21.525, 702/83, 356/237, 437/8, 371/22.1, 364/578, 438/16, 438/14, 700/121
• US Class Current: 382/145
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/0006 using a design-rule based approach
• CPC Class Codes: G06V 10/46 Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features colour feature extraction G06V10/56

Abstract

A method and apparatus for analyzing patterns in semiconductor wafers wherein the patterns are compared to a plurality of patterns stored in a common pattern library. A spatial pattern recognition engine is operable to receive a first set of data corresponding to a pattern on a semiconductor wafer and to generate a normalized contour representation of said first data set. A pattern analyzer compares the normalized data set to a plurality of reference contour data sets stored in a common pattern data reference library and generates a correlation label associating the first data set with one of the plurality of reference contour data sets. The label associated with the first data set is stored in a label storage database that can be accessed to perform subsequent analysis on the data associated with a specific wafer. The system can be used to analyze multiple types of patterns, including defect data, bin data, positional parameter data and in-line site data.

Claims

1. A system for analyzing patterns in semiconductor wafers, comprising:

a spatial pattern recognition engine operable to receive a first set of data corresponding to a pattern on a semiconductor wafer and to generate a normalized contour representation of said first data set;

a common pattern data reference library containing a plurality of reference contour data sets;

a pattern analyzer operable to compare said normalized data set to said plurality of reference contour data sets and to generate a correlation label associating said first data set with one of said plurality of reference contour data sets; and

a pixel grid generator for converting said first data set into a plurality of virtual pixels.

2. The system according to claim 1, further comprising a label storage database operable to store said normalized first data set and said label associated with said first data set.

3. The system according to claim 1, said pattern analyzer being operable to detect new patterns not contained in said common pattern reference library and to initiate an alarm indicating detection of said new pattern.

4. The system according to claim 1, further comprising a pattern library editor operable to edit the reference contour data sets in said common pattern data reference library.

5. The system according to claim 4, said pattern library editor comprising a graphical user interface.

6. The system according to claim 1, said first data set comprising defect data.

7. The system according to claim 1, said first data set comprising bin data.

8. The system according to claim 1, said first data set comprising positional parameter data.

9. The system according to claim 1, said first data set comprising in-line site data.

10. A system for analyzing patterns in semiconductor wafers, comprising:

a plurality of process sensors for providing status data relating to the status of a semiconductor wafer process;

a fault detector for receiving said status data and for generating a fault signal and for associating said fault signal with a first data set representing a wafer pattern;

a spatial pattern recognition engine for generating a normalized contour representation of said data set;

a common pattern data reference library containing a plurality of reference contour data sets; and

a pattern analyzer operable to compare said normalized data set to said plurality of reference contour data sets and to generate a correlation label associating said first data set with one of said plurality of reference contour data sets; and

a pixel grid generator operable to convert said first data set into a plurality of virtual pixels.

11. The system according to claim 10, further comprising a label storage database operable to store said normalized first data set and said label associated with said first data set.

12. The system according to claim 10, said pattern analyzer being operable to detect new patterns not contained in said common pattern reference library and to initiate an alarm indicating detection of said new pattern.

13. The system according to claim 10, further comprising an editor operable to edit the reference contour data sets in said common pattern data reference library.

14. The system according to claim 13, said editor comprising a graphical user interface.

15. The system according to claim 10, said first data set comprising defect data.

16. The system according to claim 10, said first data set comprising bin data.

17. The system according to claim 10, said first data set comprising positional parameter data.

18. The system according to claim 10, said first data set comprising in-line site data.

19. A method for analyzing patterns in semiconductor wafers, comprising:

obtaining a first data set corresponding to a pattern on a wafer, said first data set being in a first format;

converting said first data set in said first format to a second format comprising a normalized contour data representation of said first data set;

comparing said normalized contour data representation to a plurality of reference contour patterns in a common pattern reference library; and

generating a correlation label associating said normalized data set with one of said plurality of reference contour data sets; and

a using a pixel grid generator to convert said first data set into a plurality of virtual pixels.

20. The method according to claim 19, further comprising the step of storing said normalized first data set and said label associated with said first data set in a label storage database.

21. The method according to claim 19, further comprising the step of detecting new patterns not contained in said common pattern reference library and initiating an alarm indicating detection of said new pattern.

22. The method according to claim 19, said first data set comprising defect data.

23. The method according to claim 19, said first data set comprising bin data.

24. The method according to claim 19, said first data set comprising positional parameter data.

25. The method according to claim 19, said first data set comprising in-line site data.