Method and system for analyzing wafer processing order

• US 6,336,086 B1
• Filed: 12/14/1998
• Issued: 01/01/2002
• Est. Priority Date: 08/13/1998
• Status: Expired due to Term

• Current Assignee: Bell Semiconductor, LLC (Hilco Inc. (d/b/a Hilco Global))
• Original Assignee: Agere Systems Guardian Corp. (Broadcom, Inc.)
• Inventors: Segoviano, Miguel Recio, Maria, Victorino Martin Santa, Garrido, Julian Moreno, Perez, Almudena Fernandez, Alonso, Miguel Angel Merino
• Primary Examiner(s): Teska, Kevin J.
• Assistant Examiner(s): PHAN, THAI Q
• Application Number:
  US09/211,481
• Time in Patent Office: 1,114 Days
• Field of Search: 703/2, 703/5, 703/13, 382/149, 382/141, 382/144, 382/145, 700/51, 700/50, 706/8, 706/52
• US Class Current: 703/13
• CPC Class Codes: H01L 22/20 Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps

Abstract

A computer analyzes a process for fabricating a plurality of semiconductor wafers. The process has a plurality of processing steps, performed on various fabrication machines. The program is knowledge based, and is trained using training data, which may be generated by a simulator. A decision tree is generated, based on the training data. A plurality of input data representing characteristics of the semiconductor wafers are extracted. A first order pattern in any of the processing steps is identified using a decision tree, based on the input data. A plurality of probability distribution functions are formed for each characteristic. Each distribution function identifies a probability that a particular type of order pattern is present. A threshold is based on the plurality of probability distribution functions. A second order pattern in any of the processing steps is identified by comparing the data representing characteristics to the threshold. A third order pattern in any one of the processing steps is identified based on a calculated distance from a centroid computed from the data representing characteristics. One of the first, second, and third order patterns is selected, and one of the processing steps is identified as being associated with the selected order pattern. A problem in one of the plurality of machines is identified, based on a type of problem associated with the selected order pattern.

Claims

1. A method for analyzing a process for forming a plurality of objects, the process having a plurality of processing steps, the method comprising the steps of:

(a) extracting a plurality of data representing characteristics of each object;

(b) identify a first order pattern in any one of the processing steps using a decision tree, based on the data representing characteristics;

(c) identifying a second order pattern in any one of the processing steps by comparing the data representing characteristics to a predetermined threshold;

(d) identifying a third order pattern in any one of the processing steps based on a calculated distance from a centroid computed from the data representing characteristics;

(e) selecting one of the first, second, and third order patterns, and identifying one of the processing steps as being associated with the selected order pattern.

2. The method according to claim 1, wherein each of the objects is a semiconductor wafer, and the process is a process for fabricating the semiconductor wafers.

3. The method according to claim 1, wherein:

each of the processing steps includes a step of using one of a plurality of machines, and

at least two of the processing steps use respectively different ones of the plurality of machines.

4. The method according to claim 3, further comprising the step of identifying a problem in one of the plurality of machines, based on a type of problem associated with the selected order pattern.

5. The method according to claim 1, wherein step (b) includes the steps of:

(b1) simulating fabrication of a set of objects by a set of process steps, wherein at least one of the set of process steps has a defect, and providing training data representing object characteristics as an output of the simulating step;

(b2) using the training data to form the decision tree.

6. The method according to claim 5, wherein step (b1) includes adding artificial noise to the training data.

7. The method according to claim 1, wherein step (c) includes:

computing a plurality of probability distribution functions for each characteristic, each of the plurality of probability distribution functions identifying a probability that a particular type of order pattern is present as a function of the value of the characteristic; and

determining the predetermined threshold based on the plurality of probability distribution functions.

8. The method according to claim 1, wherein step (d) includes:

computing a respective centroid vector for each type of order pattern;

forming an input vector which includes the value of each characteristic extracted in step (a);

computing a respective distance between the input vector and each centroid vector; and

identifying, as the third order pattern, any order pattern for which the distance is less than a respective predetermined threshold for that order pattern.

9. The method according to claim 8, wherein each centroid vector includes a respective mean value for each characteristic.

10. A computer implemented system for analyzing a process for forming a plurality of objects, the process having a plurality of processing steps, the system comprising:

means for extracting a plurality of data representing characteristics of each object;

means for identifying a first order pattern in any one of the processing steps using a decision tree, based on the data representing characteristics;

means for identifying a second order pattern in any one of the processing steps by comparing the data representing characteristics to a predetermined threshold;

means for identifying a third order pattern in any one of the processing steps based on a calculated distance from a centroid computed from the data representing characteristics;

means for selecting one of the first, second, and third order patterns, and identifying one of the processing steps as being associated with the selected order pattern.

11. The system according to claim 10, wherein the first order pattern identifying means include:

(b1) means for simulating fabrication of a set of objects by a set of process steps, wherein at least one of the set of process steps has a defect, and for providing training data representing object characteristics as an output;

(b2) means for forming the decision tree based on the training data.

12. The system according to claim 11, wherein the simulating means include means for adding artificial noise to the training data.

13. The system according to claim 10, wherein the second order pattern identifying means include:

means for computing a plurality of probability distribution functions for each characteristic, each of the plurality of probability distribution functions identifying a probability that a particular type of order pattern is present as a function of the value of the characteristic; and

means for determining the predetermined threshold based on the plurality of probability distribution functions.

14. The system according to claim 10, wherein the third order pattern identifying means include:

means for computing a respective centroid vector for each type of order pattern;

means for forming an input vector which includes the value of each characteristic extracted by the extracting means;

means for computing a respective distance between the input vector and each centroid vector; and

means for identifying, as the third order pattern, any order pattern for which the distance is less than a respective predetermined threshold for that order pattern.

15. The system according to claim 14, wherein each centroid vector includes a respective mean value for each characteristic.

16. A storage medium encoded with machine-readable computer program code for causing a processor to analyze a process for forming a plurality of objects, the process having a plurality of processing steps, the system comprising:

means for causing the processor to extract a plurality of data representing characteristics of each object;

means for causing the processor to identify a first order pattern in any one of the processing steps using a decision tree, based on the data representing characteristics;

means for causing the processor to identify a second order pattern in any one of the processing steps by comparing the data representing characteristics to a predetermined threshold;

means for causing the processor to identify a third order pattern in any one of the processing steps based on a calculated distance from a centroid computed from the data representing characteristics;

means for causing the processor to select one of the first, second, and third order patterns, and to identify one of the processing steps as being associated with the selected order pattern.

17. The storage medium according to claim 16, wherein the first order pattern identifying means include:

(b1) means for causing the processor to simulate fabrication of a set of objects by a set of process steps, wherein at least one of the set of process steps has a defect, and for providing training data representing object characteristics as an output;

(b2) means for causing the processor to form the decision tree based on the training data.

18. The storage medium according to claim 17, wherein the simulating means include means for causing the processor to add artificial noise to the training data.

19. The storage medium according to claim 17, wherein the second order pattern identifying means include:

means for causing the processor to compute a plurality of probability distribution functions for each characteristic, each of the plurality of probability distribution functions identifying a probability that a particular type of order pattern is present as a function of the value of the characteristic; and

means for causing the processor to determine the predetermined threshold based on the plurality of probability distribution functions.

20. The storage medium according to claim 17, wherein the third order pattern identifying means include:

means for causing the processor to compute a respective centroid vector for each type of order pattern;

means for causing the processor to form an input vector which includes the value of each characteristic extracted by the extracting means;

means for causing the processor to compute a respective distance between the input vector and each centroid vector; and

means for causing the processor to identify, as the third order pattern, any order pattern for which the distance is less than a respective predetermined threshold for that order pattern.

21. The storage medium according to claim 17, wherein each centroid vector includes a respective mean value for each characteristic.

22. A method for analyzing a process for forming a plurality of objects, the process having a plurality of processing steps, the method comprising the steps of:

(a) extracting a plurality of data representing characteristics of each object;

(b) identifying a first order pattern in any one of the processing steps based on the extracted data, using a first order pattern detection algorithm;

(c) identifying a second order pattern in any one of the processing steps based on the extracted data using a second order pattern section algorithm different from the first order pattern detection algorithm;

(d) selecting one of the first and second order patterns, and identifying one of the processing steps as being associated with the selected order pattern.