Method and apparatus for detecting faults using principal component analysis parameter groupings

• US 7,198,964 B1
• Filed: 02/03/2004
• Issued: 04/03/2007
• Est. Priority Date: 02/03/2004
• Status: Expired due to Fees

• Current Assignee: GlobalFoundries, Inc.
• Original Assignee: Advanced Micro Devices, Inc.
• Inventors: Cherry, Gregory A., Kadosh, Daniel
• Primary Examiner(s): Lindsay, Jr.; Walter
• Assistant Examiner(s): Stevenson; Andre′
• Application Number:
  US10/770,681
• Time in Patent Office: 1,155 Days
• Field of Search: 438/5, 438/7, 438 10- 11, 438/14, 438 16- 18, 438 22- 24, 438/29, 438/31, 438 34- 36, 438128-130, 438/149, 438/484, 438/538
• US Class Current: 438/14
• CPC Class Codes: G05B 19/41875 characterised by quality surveillance of production
• CPC Class Codes: G05B 2219/37224 Inspect wafer
• 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: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

A method for identifying faults in a semiconductor fabrication process includes storing measurements for a plurality of parameters of a wafer in the semiconductor fabrication process. A first subset of the parameters is selected. The subset is associated with a feature formed on the wafer. A principal component analysis model is applied to the first subset to generate a performance metric. A fault condition with the wafer is identified based on the performance metric. A system includes a data store and a fault monitor. The data store is adapted to store measurements for a plurality of parameters of a wafer in a semiconductor fabrication process. The fault monitor is adapted to select a first subset of the parameters, the subset being associated with a feature formed on the wafer, apply a principal component analysis model to the first subset to generate a performance metric, and identify a fault condition with the wafer based on the performance metric.

Claims

1. A method for identifying faults in a semiconductor fabrication process, comprising:

storing measurements for a plurality of parameters of a wafer in the semiconductor fabrication process;

selecting a first subset of the parameters, the subset being associated with a feature formed on the wafer and including at least one of a physical measurement of the feature and an electrical measurement of the feature;

applying a principal component analysis model to the first subset to generate a performance metric; and

identifying a fault condition with the wafer based on the performance metric.

2. The method of claim 1, wherein applying the principal component analysis model to the first subset to generate the performance metric further comprises applying the principal component analysis model to the first subset to generate at least one of a squared prediction error, a T² statistic, and a combined index.

3. The method of claim 1, wherein selecting the first subset of the parameters, further comprises selecting a first subset associated with a layer formed on the wafer.

4. The method of claim 1, wherein selecting the first subset of the parameters, further comprises selecting a first subset associated with a device formed on the wafer.

5. The method of claim 3, further comprising:

selecting a second subset of the parameters in the first subset, the second subset being associated with a scrap code for the wafer;

applying the principal component analysis model to the second subset to generate a performance metric for the scrap code; and

designating the wafer as suspect based on the performance metric for the scrap code.

6. The method of claim 1, wherein selecting the first subset of the parameters, further comprises selecting a first subset associated with a scrap code for the wafer.

7. The method of claim 1, further comprising applying the principal component analysis model to each parameter in the first subset to generate parameter performance metrics for each of the parameters.

8. The method of claim 1, further comprising applying the principal component analysis model to the plurality of parameters to generate an overall performance metric for the wafer.

9. The method of claim 1, further comprising:

generating a confidence limit for the first subset using the principal component analysis model; and

comparing the performance metric to the confidence limit to identify the fault condition.

10. The method of claim 1, wherein applying the principal component analysis model further comprises applying a recursive principal component analysis model.

11. The method of claim 1, wherein the plurality of parameters further comprise electrical parameters, physical parameters, and processing tool parameters.

12. The method of claim 1, wherein generating measurements for the plurality of parameters further comprises generating measurements for the plurality of parameters for a lot of wafers in the semiconductor fabrication process.

13. The method of claim 12, further comprising:

averaging the measurements for the plurality of parameters in the first subset across the lot; and

applying the principal component analysis model to the averaged measurements.

14. The method of claim 1, wherein generating measurements for the plurality of parameters further comprises generating measurements for the plurality of parameters at a plurality of sites on the wafer.

15. The method of claim 14, further comprising:

averaging the measurements for the plurality of parameters in the first subset across the sites; and

applying the principal component analysis model to the averaged measurements.

16. The method of claim 1, further comprising:

determining a tool set history of processing tools used to process the feature on the wafer;

identifying at least one tool associated with the fault condition based on the tool set history.

17. The method of claim 16, further comprising:

identifying a plurality of fault conditions associated with the feature for a plurality of wafers;

determining a tool set history for each of the wafers with fault conditions; and

comparing the tool set histories to identify at least one tool associated with the fault condition.

18. A method for identifying faults in a semiconductor fabrication process, comprising:

generating measurements for a plurality of electrical parameters of a wafer in the semiconductor fabrication process;

selecting a first subset of the electrical parameters, the subset being associated with a feature formed on the wafer;

applying a principal component analysis model to the first subset to generate a performance metric; and

identifying a fault condition with the wafer based on the performance metric.

19. The method of claim 18, wherein applying the principal component analysis model to the first subset to generate the performance metric further comprises applying the principal component analysis model to the first subset to generate at least one of a squared prediction error, a T² statistic, and a combined index.

20. The method of claim 18, wherein selecting the first subset of the parameters, further comprises selecting a first subset associated with a layer formed on the wafer.

21. The method of claim 18, wherein selecting the first subset of the parameters, further comprises selecting a first subset associated with a device formed on the wafer.

22. The method of claim 20, further comprising:

selecting a second subset of the electrical parameters in the first subset, the second subset being associated with a scrap code for the wafer;

applying the principal component analysis model to the second subset to generate a performance metric for the scrap code; and

designating the wafer as suspect based on the performance metric for the scrap code.

23. The method of claim 18, wherein selecting the first subset of the electrical parameters, further comprises selecting a first subset associated with a scrap code for the wafer.

24. The method of claim 18, further comprising applying the principal component analysis model to each parameter in the first subset to generate parameter performance metrics for each of the electrical parameters.

25. The method of claim 18, further comprising applying the principal component analysis model to the plurality of electrical parameters to generate an overall performance metric for the wafer.

26. The method of claim 18, further comprising:

generating a confidence limit for the first subset using the principal component analysis model; and

comparing the performance metric to the confidence limit to identify the fault condition.

27. The method of claim 18, wherein applying the principal component analysis model further comprises applying a recursive principal component analysis model.

28. The method of claim 21, wherein the device comprises a flash memory cell, and selecting the first subset of the parameters further comprises selecting parameters relating to a diffusion resistivity of the flash memory cell.

29. The method of claim 21, wherein the device comprises a flash memory cell, and selecting the first subset of the parameters further comprises selecting parameters relating to a control gate of the flash memory cell.

30. The method of claim 21, wherein the device comprises a flash memory cell, and selecting the first subset of the parameters further comprises selecting parameters relating to a floating gate of the flash memory cell.

31. The method of claim 20, wherein the feature comprises a flash memory cell, and selecting the first subset of the parameters further comprises selecting parameters relating to an oxide-nitride-oxide layer of the flash memory cell.

32. The method of claim 20, wherein the feature comprises a flash memory cell, and selecting the first subset of the parameters further comprises selecting parameters relating to a tunnel oxide layer of the flash memory cell.

33. The method of claim 18, wherein generating measurements for the plurality of parameters further comprises generating measurements for the plurality of parameters for a lot of wafers in the semiconductor fabrication process.

34. The method of claim 32, further comprising:

averaging the measurements for the plurality of parameters in the first subset across the lot; and

applying the principal component analysis model to the averaged measurements.

35. The method of claim 18, wherein generating measurements for the plurality of parameters further comprises generating measurements for the plurality of parameters at a plurality of sites on the wafer.

36. The method of claim 35, further comprising:

averaging the measurements for the plurality of parameters in the first subset across the sites; and

applying the principal component analysis model to the averaged measurements.

37. The method of claim 18, further comprising:

determining a tool set history of processing tools used to process the feature on the wafer; and

identifying at least one tool associated with the fault condition based on the tool set history.

38. The method of claim 37, further comprising:

identifying a plurality of fault conditions associated with the feature for a plurality of wafers;

determining a tool set history for each of the wafers with fault conditions; and

comparing the tool set histories to identify at least one tool associated with the fault condition.