Run-to-run control utilizing virtual metrology in semiconductor manufacturing

• US 9,240,360 B2
• Filed: 07/25/2012
• Issued: 01/19/2016
• Est. Priority Date: 07/25/2012
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Baseman, Robert Jeffrey, He, Jingrui, Yashchin, Emmanuel, Zhu, Yada
• Primary Examiner(s): Masinick, Michael D
• Application Number:
  US13/557,955
• Publication Number:
  US 20140031968A1
• Time in Patent Office: 1,273 Days
• Field of Search: 700/121, 702/57
• US Class Current: 1/1
• CPC Class Codes: G05B 13/048 using a predictor
• CPC Class Codes: G05B 19/41875 characterised by quality surveillance of production
• CPC Class Codes: G05B 2219/32182 If state of tool, product deviates from standard, adjust system, feedback
• CPC Class Codes: G05B 2219/32187 Correlation between controlling parameters for influence on quality parameters
• CPC Class Codes: G05B 2219/32194 Quality prediction
• CPC Class Codes: G05B 2219/45031 Manufacturing semiconductor wafers
• 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]
• CPC Class Codes: Y02P 90/80 Management or planning

Abstract

A method for run-to-run control and sampling optimization in a semiconductor manufacturing process includes the steps of: determining a process output and corresponding metrology error associated with an actual metrology for a current processing run in the semiconductor manufacturing process; determining a predicted process output and corresponding prediction error associated with a virtual metrology for the current processing run; and controlling at least one parameter corresponding to a subsequent processing run as a function of the metrology error and the prediction error.

Claims

1. A method for run-to-run control and sampling optimization in a semiconductor manufacturing process, the method comprising steps of:

determining a process output and corresponding metrology error associated with an actual metrology for a current processing run in the semiconductor manufacturing process;

determining a predicted process output and corresponding prediction error associated with a virtual metrology for the current semiconductor manufacturing processing run;

assigning a first weight to the metrology error and a second weight to the prediction error, the first and second weights being indicative of a confidence in the metrology error and prediction error, respectively; and

controlling at least one parameter corresponding to a subsequent semiconductor manufacturing processing run as a function of the metrology error, the first and second weights, and the prediction error;

wherein said semiconductor manufacturing process is executed to fabricate at least one wafer.

2. The method of claim 1, wherein the steps of determining the process output and corresponding metrology error comprises:

measuring at least one target processing parameter corresponding to a wafer fabricated in the current processing run;

comparing the measured at least one target processing parameter with an expected value for the at least one target processing parameter; and

determining the metrology error as a function of a variation between the measured at least one target processing parameter and the expected value for the at least one target processing parameter.

3. The method of claim 1, wherein the steps of determining the predicted process output and corresponding prediction error comprises:

obtaining a prediction model associated with the semiconductor manufacturing process, the prediction model adapted to predict the process output by estimating at least one processing parameter for the semiconductor manufacturing process; and

performing virtual metrology based on the prediction model, the prediction error being determined as a function of an output of the virtual metrology.

4. The method of claim 3, wherein the prediction model comprises at least one of a single processing chamber based model, a global model across a plurality of processing chambers, a metrology-based prediction model, and a metrology and error-adaptive model.

5. The method of claim 4, wherein the global model is generated by approximating coefficients corresponding to different processing chambers to a common vector given by prior information on tool capability matching, adding a positive coefficient to balance different terms in the global model, and minimizing the prediction error and optimizing the approximation.

6. The method of claim 5, wherein optimizing the approximation comprises using a block coordinate based algorithm.

7. The method of claim 4, further comprising constructing the metrology-based prediction model by:

obtaining a plurality of predictors, the predictors comprising information relating to the semiconductor manufacturing process;

decomposing the predictors into at least two types, a first type comprising variance profiles, and a second type obtained based on dynamic regression models; and

generating a combined predictor comprising at least a subset of the first type of predictors and at least a subset of the second type of predictors.

8. The method of claim 1, wherein the first and second weights are dynamically adjusted in accordance with a confidence in the metrology error and prediction error, respectively, over time.

9. The method of claim 1, wherein each of the first and second weights is a percentage, with a sum of the first and second weights being equal to one.

10. The method of claim 1, further comprising optimizing a sampling policy corresponding to the semiconductor manufacturing process as a function of the first and second weights.

11. The method of claim 10, wherein the step of optimizing the sampling policy comprises maximizing a time between actual metrology measurements obtained in connection with the semiconductor manufacturing process.

12. The method of claim 1, wherein respective values of the first and second weights are determined as a function of one or more factors related to uncertainty and chamber matching capabilities associated with the semiconductor manufacturing process.

13. A method for run-to-run control and sampling optimization in a semiconductor manufacturing process, the method comprising steps of:

determining a process output and corresponding metrology error associated with an actual metrology for a current processing run in the semiconductor manufacturing process;

determining a predicted process output and corresponding prediction error associated with a virtual metrology for the current semiconductor manufacturing processing run; and

controlling at least one parameter corresponding to a subsequent semiconductor manufacturing processing run as a function of the metrology error and the prediction error;

wherein the steps of determining the predicted process output and corresponding prediction error comprises;

obtaining a prediction model associated with the semiconductor manufacturing process, the prediction model adapted to predict the process output by estimating at least one processing parameter for the semiconductor manufacturing process; and

performing virtual metrology based on the prediction model, the prediction error being determined as a function of an output of the virtual metrology;

wherein the prediction model comprises a wafer quality prediction model for metrology variable prediction which utilizes tensor input process variables by directly operating on the tensor input process variables.

14. A method for run-to-run control and sampling optimization in a semiconductor manufacturing process, the method comprising steps of:

determining a process output and corresponding metrology error associated with an actual metrology for a current processing run in the semiconductor manufacturing process;

determining a predicted process output and corresponding prediction error associated with a virtual metrology for the current semiconductor manufacturing processing run;

controlling at least one parameter corresponding to a subsequent semiconductor manufacturing processing run as a function of the metrology error and the prediction error;

assigning a first weight to the metrology error and a second weight to the prediction error, the first and second weights being indicative of a confidence in the metrology error and prediction error, respectively;

controlling the at least one parameter corresponding to a subsequent processing run as a function of the first and second weights; and

optimizing a sampling policy corresponding to the semiconductor manufacturing process as a function of the first and second weights;

wherein the step of optimizing the sampling policy comprises at least one of;

minimizing a sampling frequency of the semiconductor manufacturing process, the sampling frequency being indicative of at least one of a number of processing runs and a number of wafers between consecutive metrology measurements obtained during actual metrology; and

maximizing a time between actual metrology measurements obtained in connection with the semiconductor manufacturing process.

15. A method for run-to-run control and sampling optimization in a semiconductor manufacturing process, the method comprising steps of:

determining a process output and corresponding metrology error associated with an actual metrology for a current processing run in the semiconductor manufacturing process;

determining a predicted process output and corresponding prediction error associated with a virtual metrology for the current semiconductor manufacturing processing run;

controlling at least one parameter corresponding to a subsequent semiconductor manufacturing processing run as a function of the metrology error and the prediction error;

assigning a first weight to the metrology error and a second weight to the prediction error, the first and second weights being indicative of a confidence in the metrology error and prediction error, respectively;

controlling the at least one parameter corresponding to a subsequent processing run as a function of the first and second weights; and

calculating a given one of the first and second weights according to an expression

16. A method for run-to-run control and sampling optimization in a semiconductor manufacturing process, the method comprising steps of:

determining a process output and corresponding metrology error associated with an actual metrology for a current processing run in the semiconductor manufacturing process;

determining a predicted process output and corresponding prediction error associated with a virtual metrology for the current processing run; and

controlling at least one parameter corresponding to a subsequent processing run as a function of the metrology error and the prediction error;

wherein determining the predicted process output and corresponding prediction error comprises;

obtaining a prediction model associated with the semiconductor manufacturing process, the prediction model adapted to predict the process output by estimating at least one processing parameter for the semiconductor manufacturing process, the prediction model comprising a wafer quality prediction model for metrology variable prediction which utilizes tensor input process variables; and

performing virtual metrology based on the prediction model, the prediction error being determined as a function of an output of the virtual metrology.

17. A method for run-to-run control and sampling optimization in a semiconductor manufacturing process, the method comprising steps of:

determining a process output and corresponding metrology error associated with an actual metrology for a current processing run in the semiconductor manufacturing process;

determining a predicted process output and corresponding prediction error associated with a virtual metrology for the current processing run;

controlling at least one parameter corresponding to a subsequent processing run as a function of the metrology error and the prediction error;

assigning a first weight to the metrology error and a second weight to the prediction error, the first and second weights being indicative of a confidence in the metrology error and prediction error, respectively;

controlling the at least one parameter corresponding to a subsequent processing run as a function of the first and second weights; and

optimizing a sampling policy corresponding to the semiconductor manufacturing process as a function of the first and second weights;

wherein optimizing the sampling policy comprises at least one of;

minimizing a sampling frequency of the semiconductor manufacturing process, the sampling frequency being indicative of at least one of a number of processing runs and a number of wafers between consecutive metrology measurements obtained during actual metrology; and

maximizing a time between actual metrology measurements obtained in connection with the semiconductor manufacturing process.