System and method for performing real time data acquisition, process modeling and fault detection of wafer fabrication processes

• US 5,859,964 A
• Filed: 10/25/1996
• Issued: 01/12/1999
• Est. Priority Date: 10/25/1996
• Status: Expired due to Term

• Current Assignee: Advanced Micro Devices, Inc.
• Original Assignee: Advanced Micro Devices, Inc.
• Inventors: Cheng, Yi, Wang, Qingsu, Greig, R. Michael, Barnett, Gerald
• Primary Examiner(s): Decady, Albert
• Application Number:
  US08/736,919
• Time in Patent Office: 809 Days
• Field of Search: 395/21, 395/23, 395/183.01, 395/183.02, 395/183.09, 395/185.01, 395/185.02
• US Class Current: 714/48
• CPC Class Codes: H01L 22/20 Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps

Abstract

A system and method for detecting faults in wafer fabrication process tools by acquiring real-time process parameter signal data samples used to model the process performed by the process tool. The system includes a computer system including a DAQ device, which acquires the data samples, and a fault detector program which employs a process model program to analyze the data samples for the purpose of detecting faults. The model uses data samples in a reference database acquired from previous known good runs of the process tool. The fault detector notifies a process tool operator of any faults which occur thus potentially avoiding wafer scrap and potentially improving mean time between failures. The fault detector also receives notification of the occurrence of process events from the process tool, such as the start or end of processing a wafer, which the fault detector uses to start and stop the data acquisition, respectively. The fault detector also receives notification of the occurrence of a new process recipe and uses the recipe information to select the appropriate model for modeling the data samples. The fault detector employs a standard data exchange interface, such as DDE, between the fault detector and the model, thus facilitating modular selection of models best suited to the particular fabrication process being modeled. Embodiments are contemplated which use a UPM model, a PCA model, or a neural network model.

Claims

1. A system for detecting faults in a process tool, comprising:

means for receiving process event signals generated by the process tool;

a data acquisition device configured to acquire a sample of process parameter signals during operation of the process tool;

a model configured to receive said sample and generate a prediction error in response to said sample; and

a fault detector in communication with said model configured to receive said sample from said data acquisition device and said process event signals from said receiving means, provide said sample to said model, receive said prediction error from said model, and use said prediction error to detect a fault in the process tool.

2. The system as recited in claim 1, wherein said system further comprises a display screen configured to display notification of said fault.

3. The system as recited in claim 1, wherein said system further comprises a reference database of representative samples, wherein said model uses said reference database to generate said prediction error in response to said sample.

4. The system as recited in claim 3, wherein model uses only a range of said representative samples to generate said prediction error in response to said sample.

5. The system as recited in claim 4, wherein said fault detector provides sequence information regarding said sample to said model, wherein said model determines said range of said representative samples in response to said sequence information.

6. The system as recited in claim 1, wherein said system further comprises a report browser configured to display said prediction error.

7. The system as recited in claim 1, wherein said system further comprises a data viewer configured to display said sample.

8. The system as recited in claim 1, wherein said receiving means comprises said data acquisition device, wherein said data acquisition device is configured to receive said process event signals.

9. The system as recited in claim 1, wherein said receiving means comprises a communications port configured to receive said process event signals.

10. The system as recited in claim 1, wherein said process event signals comprise at least one event from the group comprising a wafer start, a wafer end, a new lot, a new recipe, and an alarm.

11. The system as recited in claim 1, wherein said fault detector is configured to receive said process event signals and control said data acquisition device to acquire said data sample in response to said process event signals.

12. The system as recited in claim 1, wherein said model is further configured to receive a recipe identifier and generate said prediction error in response to said recipe identifier.

13. The system as recited in claim 12, wherein said recipe identifier is specified via user input.

14. The system as recited in claim 12, wherein said recipe identifier is provided to said fault detector by said process event signals.

19. The system as recited in claim 1, wherein said model is further configured to receive a recipe identifier and generate said prediction error in response to said recipe identifier.

15. A system for detecting faults in a process tool, comprising:

a data acquisition device configured to acquire a sample of process parameter signals during operation of the process tool;

a model configured to receive said sample and generate a prediction error in response to said sample, wherein said model is one of a plurality of possible models; and

a fault detector in communication with said model configured to receive said sample from said data acquisition device, provide said sample to said model, receive said prediction error from said model, and use said prediction error to detect a fault in the process tool;

wherein said fault detector is configured to use said prediction errors to detect a fault independent of which one of said plurality of possible models is used to generate said prediction error.

16. The system as recited in claim 15, wherein said model is a model chosen from the group consisting of a Universal Process Model, a Principle Component Analysis model, and a neural network.

17. The system as recited in claim 15, wherein said system further comprises a display screen configured to display notification of said fault.

18. The system as recited in claim 15, wherein said system further comprises a reference database of representative samples, wherein said model uses said reference database to generate said prediction error in response to said sample.

20. A system for detecting faults in a process tool, comprising:

means for receiving process event signals generated by the process tool;

a data acquisition device configured to acquire a sample of process parameter signals during operation of the process tool;

a model configured to receive said sample and generate a prediction error in response to said sample, wherein said model is one of a plurality of possible models; and

a fault detector in communication with said model configured to receive said sample from said data acquisition device and said process event signals from said receiving means, provide said sample to said model, receive said prediction error from said model, and use said prediction error to detect a fault in the process tool;

wherein said fault detector is configured to use said prediction errors to detect a fault independent of which one of said plurality of possible models is used to generate said prediction error.

21. A method for detecting faults in a process tool, comprising:

receiving process event signals generated by the process tool;

acquiring a sample of process parameter signals during operation of the process tool;

generating a prediction error in response to said sample; and

detecting a fault in the process tool using said prediction error.

22. The method as recited in claim 21, wherein said detecting a fault in the process tool using said prediction error comprises comparing said prediction error with a prediction error threshold value.

23. The method as recited in claim 21, further comprising displaying notification of said fault on a display screen after said detecting said fault.

24. The method as recited in claim 21, wherein said generating a prediction error in response to said sample comprises using a reference database of representative samples to generate said prediction error.

25. The method as recited in claim 24, wherein said using a reference database of representative samples to generate said prediction error comprises using only a range of said representative samples to generate said prediction error.

26. The method as recited in claim 25, further comprising receiving sequence information regarding said sample, and determining said range of said representative samples in response to said sequence information.

27. The method as recited in claim 21, further comprising displaying said prediction error.

28. The method as recited in claim 21, further comprising displaying said sample.

29. The method as recited in claim 21, further comprising acquiring said data sample in response to said receiving process event signals generated by the process tool.

30. The method as recited in claim 21, further comprising receiving a recipe identifier, wherein said generating a prediction error in response to said sample is in response to said receiving said recipe identifier.

31. A method for detecting faults in a process tool, comprising:

selecting a model for generating a prediction error in response to a sample of process parameter signals from a plurality of possible models;

acquiring said sample during operation of the process tool;

said model generating a prediction error in response to said sample; and

detecting a fault in the process tool using said prediction error;

wherein said detecting a fault in the process tool using said prediction error is performed independent of which one of said plurality of possible models is used to generate said prediction error.

32. The method as recited in claim 31, wherein said selecting a model comprises selecting a model from the group consisting of a Universal Process Model, a Principle Component Analysis model, and a neural network.

33. The method as recited in claim 31, further comprising displaying notification of said fault on a display screen.

34. The method as recited in claim 31, wherein said generating a prediction error in response to said sample comprises using a reference database of representative samples to generate said prediction error.

35. The method as recited in claim 31, further comprising receiving a recipe identifier, wherein said generating a prediction error in response to said sample is in response to said receiving said recipe identifier.

36. A method for detecting faults in a process tool, comprising:

selecting a model for generating a prediction error in response to a sample of process parameter signals from a plurality of possible models;

acquiring said sample during operation of the process tool;

receiving process event signals generated by the process tool;

said model generating a prediction error in response to said sample; and

detecting a fault in the process tool using said prediction error;

wherein said detecting a fault in the process tool using said prediction error is performed independent of which one of said plurality of possible models is used to generate said prediction error.