Film thickness control using spectral interferometry

• US 6,589,869 B2
• Filed: 04/23/2002
• Issued: 07/08/2003
• Est. Priority Date: 12/23/1999
• Status: Expired due to Fees

• Current Assignee: Applied Materials Incorporated
• Original Assignee: Applied Materials Incorporated
• Inventors: Lymberopoulos, Dimitris P., Davidow, Jed E., Balasubramhanya, Lalitha S., Sarfaty, Moshe
• Primary Examiner(s): HITESHEW, FELISA CARLA
• Application Number:
  US10/131,726
• Publication Number:
  US 20020119660A1
• Time in Patent Office: 441 Days
• Field of Search: 438/689, 438/690
• US Class Current: 438/689
• CPC Class Codes: H01L 22/12 for structural parameters, e.g. thickness, line width, refractive index, temperature, warp, bond strength, defects, optical inspection, electrical measurement of structural dimensions, metallurgic measurement of diffusions electrical measurement of diffusions H01L22/14
• CPC Class Codes: H01L 22/26 Acting in response to an ongoing measurement without interruption of processing, e.g. endpoint detection, in-situ thickness measurement endpoint detection arrangements in CMP apparatus B24B37/013, in discharge apparatus H01J37/32
• CPC Class Codes: H01L 2924/00 Indexing scheme for arrangements or methods for connecting or disconnecting semiconductor or solid-state bodies as covered by H01L24/00
• CPC Class Codes: H01L 2924/0002 Not covered by any one of groups H01L24/00, H01L24/00 and H01L2224/00

Abstract

A process for controlling a substrate processing operation such as a plasma etch operation. One embodiment of the method of the present invention forms a plasma within a substrate processing chamber to etch a wafer disposed within the chamber. The plasma emission is used by the process as a broadband light source. During the plasma etch process, a plurality of wavelengths of radiation reflected from the surface of the wafer being etched are measured with a spectrometer. These measurements are then compared using pattern recognition techniques to previous measurements taken during a previous plasma etch operation. Certain embodiments of the invention use principal component analysis (PCA) techniques to perform pattern recognition while other embodiment use programmed neural net pattern recognition techniques.

Claims

1. A method for controlling a substrate processing chamber, said method comprising:

transferring a substrate into said substrate processing chamber;

performing a first substrate operation on the substrate disposed within the substrate processing chamber;

measuring a plurality of wavelengths of radiation reflected from an upper surface of said substrate during the first substrate processing operation;

converting each of the plurality of measured wavelengths into digital data; and

comparing the digital data in real time with the digital data collected from a previous substrate processing operation using a principal component analysis pattern recognition technique.

2. The method of claim 1 wherein the digital data collected from a previous substrate processing operation is generated from a plurality of wavelengths of radiation reflected from an upper surface of a test substrate during a pre-production run substrate processing operation similar to the first substrate processing operation.

3. The method of claim 2 further comprising identifying a subset of the digital data associated with a time period from the pre-production substrate processing operation that corresponds to an event to be identified in the pre-production run process.

4. The method of claim 3 wherein the event to be identified is an endpoint of an etching process.

5. The method of claim 3 wherein principal component analysis is performed on the subset of the digital data.

6. The method of claim 5 wherein the subset of digital data represents a window of data collected at times before, during and after the event to be identified.

7. The method of claim 6 wherein the principal component analysis performed on the subset of digital data comprises forming a matrix from the subset of digital data, the matrix having rows comprising measured correlated attribute data and columns comprising a time attribute each correlated attribute data was measured.

8. The method of claim 7 further comprising mean centering data within the matrix.

9. The method of claim 7 further comprising performing a singular value decomposition on the matrix and generating principal component eigenvectors for the data within the matrix.

10. The method of claim 9 further comprising designating one principal component as a reference principal component that represents the event to be identified.

11. The method of claim 10 further comprising storing an evolving window of the digital data from the first substrate processing operation.

12. The method of claim 11 wherein the comparing step comprises performing principal component analysis on the window of digital data from the first substrate processing operation to generate a non-reference principal component and comparing the non-reference principal component to the reference principal component.

13. The method of claim 12 further comprising generating a signal, if the non-reference principal component is identified as matching the reference principal component.

14. The method of claim 13 wherein the first substrate processing operation is a plasma etch operation and wherein the plasma etch operation is endpointed on the generation of the signal.

15. The method of claim 12 wherein the window of digital data from the first substrate processing operation is the same size as the window of data collected during the pre-production run process.

16. A method for operating a substrate processing chamber to etch a substrate, the method comprising:

performing a calibration process within the substrate processing chamber by (i) transferring a first substrate into the chamber, (ii) etching the first substrate according to a first etch recipe, (iii) measuring a first plurality of wavelengths of radiation reflected from an upper surface of the first substrate during the etching step, (iv) converting the first plurality of wavelengths into digital data that includes a first set of digital data having data points representing times before and after an endpoint of the etch process, and (v) generating a reference principal component indicative of the endpoint from the first set of digital data;

thereafter, performing a production process within the substrate processing chamber by (i) transferring a second substrate into the substrate processing chamber, (ii) etching the second substrate according to a second etch recipe substantially similar to the first etch recipe, (iii) measuring a second plurality of wavelengths of radiation reflected from an upper surface of the second substrate during the etching step, (iv) converting the second plurality of wavelengths into a second set of digital data, (v) generating a non-reference principal component from the second set of digital data, and (vi) comparing the non reference principal component to the reference principal component and endpointing the etch process if the non reference principal component matches the reference principal component.

17. The method of claim 16 wherein, during the production process, the second set of digital data represents an evolving window of data such that once the window is full, over time, old data is dropped from the window in favor of new data.

18. The method of claim 16 wherein the reference principal component is formed from a first matrix of data having rows that comprise measured correlated attribute data from the first etch process and columns that comprise a time attribute each correlated attribute was measured and wherein the non-reference principal component is formed from a second matrix of data having rows that comprise measured correlated attribute data from the second etch process and columns that comprise a time attribute each correlated attribute was measured.

19. The method of claim 18 wherein data within the first and second matrices is mean centered prior to forming the reference and non-reference principal components, respectively.

20. The method of claim 19 wherein the reference principal component is selected from a plurality of principal components formed by performing a singular value decomposition on the first matrix and generating principal component eigenvectors for the data within the first matrix.