METHODS FOR CONSTRUCTING AN OPTIMAL ENDPOINT ALGORITHM

• US 20100330710A1
• Filed: 06/29/2010
• Published: 12/30/2010
• Est. Priority Date: 06/30/2009
• Status: Active Grant

• Current Assignee: Lam Research Corporation
• Original Assignee: Lam Research Corporation
• Inventors: Perry, Andrew James, Wang, Jiangxin, Venugopal, Vijayakumar C.
• Granted Patent: US 8,538,572 B2
• Time in Patent Office: Days
• US Class Current: 438/5
• CPC Class Codes: H01J 37/32935 Monitoring and controlling tubes by information coming from the object and/or discharge
• CPC Class Codes: H01J 37/32963 End-point detection

Abstract

A method for automatically identifying an optimal endpoint algorithm for qualifying a process endpoint during substrate processing within a plasma processing system is provided. The method includes receiving sensor data from a plurality of sensors during substrate processing of at least one substrate within the plasma processing system, wherein the sensor data includes a plurality of signal streams from a plurality of sensor channels. The method also includes identifying an endpoint domain, wherein the endpoint domain is an approximate period within which the process endpoint is expected to occur. The method further includes analyzing the sensor data to generate a set of potential endpoint signatures. The method yet also includes converting the set of potential endpoint signatures into a set of optimal endpoint algorithms. The method yet further includes importing one optimal endpoint algorithm of the set of optimal endpoint algorithms into production environment.

Claims

1. A method for automatically identifying an optimal endpoint algorithm for qualifying a process endpoint during substrate processing within a plasma processing system, comprising:

receiving sensor data from a plurality of sensors during substrate processing of at least one substrate within said plasma processing system, wherein said sensor data includes a plurality of signal streams from a plurality of sensor channels;

identifying an endpoint domain, wherein said endpoint domain is an approximate period within which said process endpoint is expected to occur;

analyzing said sensor data to generate a set of potential endpoint signatures;

converting said set of potential endpoint signatures into a set of optimal endpoint algorithms; and

importing one optimal endpoint algorithm of said set of optimal endpoint algorithms into production environment.

2. The method of claim 1 wherein said sensor data is being gathered from more than one substrate.

3. The method of claim 1 wherein said analyzing of said sensor data includes performing linear fitting on said sensor data to divide each signal stream from said plurality of signal streams into a plurality of segments based on time intervals.

4. The method of claim 3 wherein each segment of said plurality of segments is uniform.

5. The method of claim 3 wherein said analyzing of said sensor data to generate a first set of potential endpoint signatures of said set of potential endpoint signatures includes

calculating a first set of slopes and a first set of corresponding slope noise values for said sensor data, wherein a slope and a corresponding slope noise value is calculated for each segment of said plurality of segments,

calculating slope variances in said slopes to identify a set of high contrast signals from said plurality of signal streams, wherein said set of high contrast signals having high slope variance,

combining contiguous wavelengths with similar slope variance into a set of signal wavelength bands,

ranking said high contrast signals,

ranking said set of signal wavelength bands, and

identifying said first set of potential endpoint signatures by applying a set of class features to at least part of said high contrast signals and said set of signal wavelength bands, wherein said set of class features include at least one of peak, valley, and inflection.

6. The method of claim 5 wherein said analyzing of said sensor data to generate a second set of potential endpoint signatures of said set of potential endpoint signatures includes

performing a multi-variate analysis by combining slopes scaled by corresponding slope noise values of said first set of slopes and said first set of corresponding slope noise values to generate a normalized set of slopes and a normalized set of corresponding slope noise values,

calculating slope variances in said normalized set of slopes to identify normalizing signals from said plurality of signal streams, wherein said normalizing signals having high slope and low variance,

combining contiguous wavelengths with similar slope variance into a set of normalizing signal wavelength bands,

ranking said normalizing signals,

ranking said set of normalizing signal wavelength bands, and

applying a set of class features to a ratio of said high contrast signals and a set of signal wavelength bands to said normalizing signals and said set of normalizing signal wavelength bands to generate said second set of potential endpoint signatures.

7. The method of claim 5 wherein said converting of said set of potential endpoint signatures into said set of optimal endpoint algorithms includes

combining a first potential endpoint signature and a second potential endpoint signature of said set of potential endpoint signatures if said first potential endpoint signature and said second potential endpoint signature having similar shapes and time period,

performing a robustness test to remove potentially non-repeatable endpoint signatures from said set of potential endpoint signatures,

performing a multi-variate correlation analysis to identify a set of optimal endpoint signatures of said set of potential endpoint signatures,

converting said set of optimal endpoint signatures into a set of real-time endpoint algorithms with minimal real-time delay, wherein said real-time delay being based on filter delay,

generating said set of optimal endpoint algorithms by performing at least one of

removing real-time endpoint algorithms with corresponding real-time delay greater than a predefined threshold, and

eliminating said real-time endpoint algorithms if said real-time endpoint algorithms fail to pass a robustness test, and

ranking each optimal endpoint algorithm of said set of optimal endpoint algorithms, wherein said ranking being based on at least one of fidelity ratio and said real-time delay.

8. The method of claim 6 wherein said converting of said set of potential endpoint signatures into said set of optimal endpoint algorithms includes

combining a first potential endpoint signature and a second potential endpoint signature of said set of potential endpoint signatures if said first potential endpoint signature and said second potential endpoint signature having similar shapes and time period,

performing a robustness test to remove potentially non-repeatable endpoint signatures from said set of potential endpoint signatures,

performing a multi-variate correlation analysis to identify a set of optimal endpoint signatures of said set of potential endpoint signatures,

converting said set of optimal endpoint signatures into a set of real-time endpoint algorithms with minimal real-time delay, wherein said real-time delay being based on filter delay,

generating said set of optimal endpoint algorithms by performing at least one of

removing real-time endpoint algorithms with corresponding real-time delay greater than a predefined threshold, and

eliminating said real-time endpoint algorithms if said real-time endpoint algorithms fail to pass a robustness test, and

ranking each optimal endpoint algorithm of said set of optimal endpoint algorithms, wherein said ranking being based on at least one of fidelity ratio and said real-time delay.

9. The method of claim 1 wherein said importing of said optimal endpoint algorithm being based on at least one of ranking and a set of user-defined conditions.

10. A method for qualifying an endpoint during substrate processing within a processing chamber, comprising:

executing a recipe on a substrate;

receiving processing data from a set of sensors during substrate processing;

analyzing said processing data by applying an optimal endpoint algorithm;

identifying a process endpoint; and

stopping said substrate processing.

11. The method of claim 10 wherein said analyzing is being performed by a computing engine, wherein said computing engine is a fast processing module configured for handling a large volume of data.

12. The method of claim 10 wherein said optimal endpoint algorithm is constructed by

receiving sensor data from a plurality of sensors during processing of at least one substrate within said plasma processing system, wherein said sensor data includes a plurality of signal streams from a plurality of sensor channels,

identifying an endpoint domain, wherein said endpoint domain is an approximate period within which said process endpoint is expected to occur,

analyzing said sensor data to generate a set of potential endpoint signatures,

converting said set of potential endpoint signatures into a set of optimal endpoint algorithms, and

importing one optimal endpoint algorithm of said set of optimal endpoint algorithms into production environment.

13. The method of claim 12 wherein said sensor data is being gathered from more than one substrate.

14. The method of claim 13 wherein said analyzing of said sensor data includes performing linear fitting on said sensor data to divide each signal stream from said plurality of signal streams into a plurality of segments based on time intervals.

15. The method of claim 14 wherein said analyzing of said sensor data to generate a first set of potential endpoint signatures of said set of potential endpoint signatures includes

calculating a first set of slopes and a first set of corresponding slope noise values for said sensor data, wherein a slope and a corresponding slope noise value is calculated for each segment of said plurality of segments,

calculating slope variances in said slopes to identify a set of high contrast signals from said plurality of signal streams, wherein said set of high contrast signals having high slope variance,

combining contiguous wavelengths with similar slope variance into a set of signal wavelength bands,

ranking said high contrast signals,

ranking said set of signal wavelength bands, and

identifying said first set of potential endpoint signatures by applying a set of class features to at least part of said high contrast signals and said set of signal wavelength bands, wherein said set of class features include at least one of peak, valley, and inflection.

16. The method of claim 15 wherein said analyzing of said sensor data to generate a second set of potential endpoint signatures of said set of potential endpoint signatures includes

performing a multi-variate analysis by combining slopes scaled by corresponding slope noise values of said first set of slopes and said first set of corresponding slope noise values to generate a normalized set of slopes and a normalized set of corresponding slope noise values,

calculating slope variances in said normalized set of slopes to identify normalizing signals from said plurality of signal streams, wherein said normalizing signals having high slope and low variance,

combining contiguous wavelengths with similar slope variance into a set of normalizing signal wavelength bands,

ranking said normalizing signals,

ranking said set of normalizing signal wavelength bands, and

applying a set of class features to a ratio of said high contrast signals and a set of signal wavelength bands to said normalizing signals and said set of normalizing signal wavelength bands to generate said second set of potential endpoint signatures.

17. The method of claim 15 wherein said converting of said set of potential endpoint signatures into said set of optimal endpoint algorithms includes

combining a first potential endpoint signature and a second potential endpoint signature of said set of potential endpoint signatures if said first potential endpoint signature and said second potential endpoint signature having similar shapes and time period,

performing a robustness test to remove potentially non-repeatable endpoint signatures from said set of potential endpoint signatures,

performing a multi-variate correlation analysis to identify a set of optimal endpoint signatures of said set of potential endpoint signatures,

converting said set of optimal endpoint signatures into a set of real-time endpoint algorithms with minimal real-time delay, wherein said real-time delay being based on filter delay,

generating said set of optimal endpoint algorithms by performing at least one of

removing real-time endpoint algorithms with corresponding real-time delay greater than a predefined threshold, and

eliminating said real-time endpoint algorithms if said real-time endpoint algorithms fail to pass a robustness test, and

ranking each optimal endpoint algorithm of said set of optimal endpoint algorithms, wherein said ranking being based on at least one of fidelity ratio and said real-time delay.

18. The method of claim 16 wherein said converting of said set of potential endpoint signatures into said set of optimal endpoint algorithms includes

combining a first potential endpoint signature and a second potential endpoint signature of said set of potential endpoint signatures if said first potential endpoint signature and said second potential endpoint signature having similar shapes and time period,

performing a robustness test to remove potentially non-repeatable endpoint signatures from said set of potential endpoint signatures,

performing a multi-variate correlation analysis to identify a set of optimal endpoint signatures of said set of potential endpoint signatures,

converting said set of optimal endpoint signatures into a set of real-time endpoint algorithms with minimal real-time delay, wherein said real-time delay being based on filter delay,

generating said set of optimal endpoint algorithms by performing at least one of

removing real-time endpoint algorithms with corresponding real-time delay greater than a predefined threshold, and

eliminating said real-time endpoint algorithms if said real-time endpoint algorithms fail to pass a robustness test, and

ranking each optimal endpoint algorithm of said set of optimal endpoint algorithms, wherein said ranking being based on at least one of fidelity ratio and said real-time delay.

19. The method of claim 12 wherein said importing of said optimal endpoint algorithm being based on at least one of ranking and a set of user-defined conditions.

20. The method of claim 14 wherein each segment of said plurality of segments is uniform.