Yield based, in-line defect sampling method

• US 6,613,590 B2
• Filed: 05/02/2001
• Issued: 09/02/2003
• Est. Priority Date: 08/21/1998
• Status: Expired due to Fees

• Current Assignee: Micron Technology, Inc.
• Original Assignee: Micron Technology, Inc.
• Inventors: Simmons, Steven J.
• Primary Examiner(s): LEBENTRITT, MICHAEL
• Application Number:
  US09/847,708
• Publication Number:
  US 20010023083A1
• Time in Patent Office: 853 Days
• Field of Search: 438/14-18, 438/701, 438/4, 438/7
• US Class Current: 438/14
• CPC Class Codes: H01L 22/20 Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps
• CPC Class Codes: Y02P 80/30 Reducing waste in manufacturing processes; Calculations of released waste quantities

Abstract

A test method provides a sample of wafer level defects most likely to cause yield loss on a semiconductor wafer subdivided into a plurality of integrated circuits (IC'"'"'s). Defect size and location data from an inspection tool is manipulated in an algorithm based on defect sizes and geometry parameters. The defects are classified by defect size to form size based populations. The contribution of each size range of defect population to yield loss is calculated and random samples for review are selected from each defect size population. The number of samples from each size defect population is proportional to the predicted yield impact of each sample. The method is rapid and permits on-line process modification to reduce yield losses.

Claims

1. A processing method for semiconductor dice on a wafer comprising:

inspecting said semiconductor dice on said wafer for determining defects thereon and classifying each of said defects by size and location, said inspecting and said classifying comprising said classifying said each of said defects into one of a plurality of size range populations of defects;

assigning a weight to said each of said defects representing an estimated effect of said each of said defects on die yield for said semiconductor dice;

determining an estimated die yield loss (DYL) for each die of said semiconductor dice based on number and weight of said defects on said each die of said semiconductor dice, determining said estimated die yield loss (DYL) including calculating an estimated die yield loss having lower and upper limits;

summing all said estimated DYL of said semiconductor dice on said wafer to obtain a wafer yield loss (WYL);

subdividing the defects into said plurality of size range populations of defects of said semiconductor dice; and

determining a relative contribution of each size range population of defects of said plurality of size range populations of said semiconductor dice to said WYL.

2. The processing method of claim 1, wherein said determining said estimated die yield loss (DYL) comprises calculating an estimated die yield loss having lower and upper limits of zero and 1.0, respectively.

3. The processing method of claim 2, wherein said lower limit comprises a representation of no yield loss attributable to said defects and said upper limit comprises a representation of fatal yield loss attributable to said defects.

4. The processing method of claim 1, wherein said determining said estimated die yield loss (DYL) comprises calculating an estimated die yield loss having a lower limit and an upper limit of zero and 1.0, respectively.

5. The processing method of claim 1, wherein said subdividing said defects into said plurality of size range populations of defects comprises subdividing said defects into a range of 0 to 10 size range populations.

7. The processing method of claim 1, further comprising:

randomly selecting defects from said each size range population of defects of said plurality of size range populations, a number selected from said each size range population of defects of said plurality of size range populations in proportion to said relative contribution thereof, said randomly selected defects being weighted to represent defects having greatest effect on the wafer yield losses.

8. The processing method of claim 7, further comprising:

reviewing said randomly selected defects and determining in-line action required to reduce wafer yield losses.

9. The processing method of claim 8, wherein said reviewing said randomly selected defects includes visual inspection by one of a scanning microscope and an optical microscope.

10. The processing method of claim 8, wherein said determining in-line action comprises determining if an individual die of said semiconductor dice on said wafer is acceptable to proceed in a manufacturing process.

11. The processing method of claim 1, wherein said inspecting said semiconductor dice is performed by an automated surface inspection tool.

6. A processing method for semiconductor dice on a wafer comprising:

inspecting said semiconductor dice on said wafer to determine defects thereon and classifying each of said defects by size and location, said inspecting and said classifying comprising classifying said each of said defects into one of a plurality of size range populations of defects;

assigning a weight to said each of said defects representing an estimated effect of said defects on die yield for said semiconductor dice;

determining an estimated die yield loss (DYL) for each die of said semiconductor dice based on number and weight of said defects on said each die of said semiconductor dice;

summing all DYL of said semiconductor dice on said wafer to obtain a wafer yield loss (WYL);

subdividing the defects into said plurality of size range populations of defects; and

determining a relative contribution of each size range population of defects of said plurality of size range populations to said WYL, said determining the relative contribution of said each size range population of defects of said plurality of size range populations to said wafer yield loss comprises;

discarding data for said each size range population of defects of said plurality of size range populations calculating, in turn, a drop in said WYL for combined size range populations excluding the discarded data;

summing the calculated wafer yield losses to obtain a drop sum;

dividing said drop sum to determine a relative drop attributable to said each size range population of defects of said plurality of size range populations; and

randomly selecting defects from said each size range population of defects.

12. A processing method for semiconductor dice on a wafer comprising:

inspecting said wafer to determine defects thereon and classifying each of said defects by size and location;

assigning a weight to said each of said defects representing an estimated effect of said each of said defects on die yield;

determining an estimated die yield loss (DYL) for each die based on number and weight of said defect(s) on said each said die of said semiconductor dice;

summing all DYL of said semiconductor dice on said wafer to obtain a wafer yield loss (WYL);

subdividing the defects into a plurality of size range populations of defects;

determining a relative contribution of each size range population of defects of said plurality of size range populations to said wafer yield loss WYL;

randomly selecting defects from said each size range population of defects of the plurality of size range populations, a number selected from said each size range population of defects of the plurality of size range populations in proportion to said relative contribution thereof, said randomly selected defects weighted to represent defects having greatest effect on yield losses;

reviewing said randomly selected defects.

13. The processing method of claim 12, further comprising:

reviewing said randomly selected defects and determining in-line action required to reduce said wafer yield losses.

14. The processing method of claim 12, wherein said inspecting said wafer said dice and said classifying each of said defects comprises classifying each of said defects into one of said plurality of size range populations of defects.

15. The processing method of claim 12, wherein said determining said estimated die yield loss (DYL) comprises calculating an estimated die yield loss having lower and upper limits of zero and 1.0, respectively.

16. The processing method of claim 15, wherein said lower limit comprises a representation of no yield loss attributable to said defects and said upper limit comprises a representation of fatal yield loss attributable to said defects.

17. The processing method of claim 12, wherein said subdividing said defects into said plurality of size range populations of defects comprises subdividing said defects into a range of 0 to 10 size range populations.

18. The processing method of claim 12, wherein said determining the relative contribution of said each size range population of defects of said plurality of size range populations to said wafer yield loss comprises:

discarding data for said each size range population of defects of said plurality of size range populations and calculating, in turn, a drop in said wafer yield loss for combined size range populations excluding the discarded data;

summing the calculated drop in wafer yield losses to obtain a drop sum; and

dividing said drop sum to determine a relative drop attributable to said each size range population of defects of said plurality of size range populations.

19. The processing method of claim 13, wherein said reviewing said randomly selected defects includes visual inspection by one of a scanning microscope and an optical microscope.

20. The processing method of claim 13, wherein said determining in-line action required to reduce said wafer yield losses comprises determining if an individual die of said semiconductor dice on said wafer is acceptable to proceed in a manufacturing process.

21. The processing method of claim 12, wherein said inspecting wafer is performed by an automated surface inspection tool.

22. A processing method for semiconductor dice on a wafer comprising:

inspecting said semiconductor dice on said wafer to determine defects thereon and classifying each of said defects by size and location, said inspecting and classifying comprising classifying said each of said defects into one of a plurality of size range populations of defects;

assigning a weight to said each of said defects representing an estimated effect of said each of said defects on die yield;

determining an estimated die yield loss (DYL) for each semiconductor die based on number and weight of said defects on said each semiconductor die of said semiconductor dice, said determining said estimated die yield loss including calculating an estimated die yield loss having lower and upper limits;

summing all DYL of said semiconductor dice on said wafer to obtain a wafer yield loss (WYL);

subdividing the defects into said plurality of size range populations of defects;

determining a relative contribution of each size range population of defects of said plurality of size range populations to said WYL;

randomly selecting defects from said each size range population of defects of the plurality, a number selected from said each size range population of defects of the plurality in proportion to said relative contribution thereof, said randomly selected defects weighted to represent defects having greatest effect on yield losses; and

reviewing said randomly selected defects and determining in-line action required to reduce said wafer yield losses.

23. The processing method of claim 22, wherein said determining said estimated die yield loss comprises calculating an estimated die yield loss having lower and upper limits of zero and 1.0, respectively.

24. The processing method of claim 23, wherein said lower limit comprises a representation of no yield loss attributable to said defects and said upper limit comprises a representation of fatal yield loss attributable to said defects.

25. The processing method of claim 22, wherein said subdividing said defects into said plurality of size range populations of defects comprises subdividing said defects into a range of 0 to 10 size range populations.

26. The processing method of claim 22, wherein said determining the relative contribution of said each size range population of defects of said plurality of size range populations to said wafer yield loss comprises:

discarding data for said each size range population of defects of said plurality of size range populations; and

calculating, in turn, a drop in wafer yield loss for combined size range populations excluding the discarded data;

summing the calculated drop in wafer yield losses to obtain a drop sum; and

dividing said drop sum to determine a relative drop sum attributable to said each size range population of defects of said plurality of size range populations.

27. The processing method of claim 22, wherein said reviewing said randomly selected defects includes visual inspection by one of a scanning microscope and an optical microscope.

28. The processing method of claim 22, wherein said determining in-line action comprises determining if an individual die of said semiconductor dice on said wafer is acceptable to proceed in a manufacturing process.

29. The processing method of claim 22, wherein said inspecting said semiconductor dice is performed by an automated surface inspection tool.