Method and system employing optical emission spectroscopy for monitoring and controlling semiconductor fabrication

• US 5,347,460 A
• Filed: 08/25/1992
• Issued: 09/13/1994
• Est. Priority Date: 08/25/1992
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Gifford, George G., Osborn, Brock E.
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): TRAMMELL, JAMES P
• Application Number:
  US07/935,076
• Time in Patent Office: 749 Days
• Field of Search: 364/468, 364/498, 364/552, 364/550, 156/626, 156/643, 204/192.33, 204/298.32, 437/129
• US Class Current: 700/121
• CPC Class Codes: G01J 3/443 Emission spectrometry
• CPC Class Codes: H01J 2237/332 Coating
• CPC Class Codes: H01J 2237/334 Etching
• 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
• CPC Class Codes: H01J 37/32972 Spectral analysis
• CPC Class Codes: H01J 37/3299 Feedback systems
• CPC Class Codes: H01L 21/67253 Process monitoring, e.g. flow or thickness monitoring

Abstract

Automated, closed loop method and system for monitor and control of semiconductor fabrication processing are described. Optical emissions spectrometer (OES) data readings from a fabrication plasma chamber are statistically analyzed and a novel pattern model (based on Markov random fields) is used in combination with a selective stochastic relaxation technique to identify gaseous species within the chamber from the OES readings. Wavelength and intensity information is also employed to accurately estimate relative concentration levels of identified gases within the chamber. The unique statistical analysis approach described allows real-time monitor and control of physical processing within the fabrication chamber. Several practical algorithms are set forth, including techniques for OES peak identification, peak sharpening, gas identification, and physical processing control.

Claims

1. An automated method for operating a semiconductor manufacturing apparatus within predefined or derived limits, said apparatus having a fabrication chamber with at least one of a flow rate controllable gaseous species inlet, an adjustable radio frequency powering means and a pressure adjustment mechanism, said automated method comprising the steps of:

(a) producing a plasma process within said fabrication chamber of said semiconductor manufacturing apparatus;

(b) simultaneous with said step (a), collecting from within said chamber wavelength and relative intensity data on a plurality of different electromagnetic wavelengths simultaneously occurring therein;

(c) determining the location of at least some intensity peaks for the simultaneously occurring plurality of different electromagnetic wavelengths from said wavelength and relative intensity data;

(d) utilizing said wavelength data collected in step (b) and said intensity peak location information determined in step (c) to identify using a Markov random field model at least one gaseous species contained within said chamber during said plasma processing; and

(e) regulating in an automated manner said plasma process of step (a) to ensure operation of semiconductor manufacturing apparatus within said predefined or derived limits, said regulating being based on information determined in said steps (b)-(d), said regulating step including automatically regulating said plasma process of step (a) through control of one of said at least one flow rate controllable gaseous species inlet, said adjustable radio frequency powering means and said pressure adjustment mechanism associated with said fabrication chamber.

2. The automated method of claim 1, wherein said steps (b)-(e) are accomplished in real-time.

3. The automated method of claim 1, further comprising the step of utilizing said wavelength and relative intensity data collected in said step (b) and said peak location information determined in said step (c) to determine in an automated manner the relative concentration level of each gaseous species identified in said step (d), and wherein said automated regulating step (e) includes regulating processing of said semiconductor manufacturing apparatus using said information determined in said steps (b)-(d) and said determined, relative gas concentration levels.

4. The automated method of claim 1, wherein said wavelength and relative intensity data collecting of step (b) includes collecting said wavelength and relative intensity data occurring within the chamber utilizing optical emission spectroscopy.

5. The automated method of claim 4, wherein said wavelength and relative intensity data comprises optical emission spectroscopic data, and wherein said spectroscopic data collecting of step (b) includes sensing at a plurality of optical sensors said wavelength and relative intensity data occurring within said chamber, each of said plurality of optical sensors being dedicated to a particular wavelength.

6. The automated method of claim 1, wherein said identifying step (d) includes referencing a predefined look-up table containing known wavelength location information for various gaseous species to compare said wavelength and relative intensity data with said known wavelength location information and thereby identify said at least one gaseous species contained within said chamber.

7. The automated method of claim 1, wherein said Markov random field model is pre-established based on Markov random fields for a plurality of preselected possible gaseous species within said fabrication chamber and wherein said identifying step (d) includes solving said Markov random field based model for said at least one gaseous species contained within said chamber using a maximization algorithm.

8. The automated method of claim 7, wherein said maximization algorithm comprises selective stochastic relaxation.

9. The automated method of claim 1, wherein said plasma processing of step (a) continues for a predefined interval, and wherein said method further comprises the step of periodically repeating steps (b)-(e) during said predefined interval.

10. The automated method of claim 1, wherein said peak locating step (c) further comprises sharpening at least some of said intensity peak location information determined in said step (c) prior to said gaseous species identifying step (d).

11. The automated method of claim 10, wherein said intensity peak sharpening includes employing an extrapolation procedure to refine the wavelength and relative intensity data collected in said step (b).

12. A closed loop method for monitoring and controlling a plasma process, said method comprising the steps of:

(a) providing a plasma processing system having a fabrication chamber with at least one electronically adjustable control for regulating plasma processing within said chamber;

(b) establishing a plasma process within said fabrication chamber;

(c) simultaneous with said step (b), collecting an electromagnetic wavelength spectra from said plasma process occurring within said fabrication chamber, said collected spectra comprising wavelength and relative intensity data;

(d) determining an intensity peak location for each of at least some of said different electromagnetic wavelengths in said wavelength data;

(e) utilizing said wavelength and relative intensity data of step (c) and said peak location information of step (d) to identify using a Markov random field model at least one vaporized species contained within said chamber; and

(f) regulating in an automated manner said plasma processing of step (b) based on said wavelength and relative intensity data, peak location information and vaporized species information of said steps (c)-(e), respectively, said plasma process regulating including producing an electronic signal to automatically regulate said at least one adjustable control associated with said fabrication chamber, and thereby regulate said plasma processing occurring therein.

13. The monitoring and control method of claim 12, wherein said plasma process of step (b) is established for a predefined fabrication interval, and wherein said method further comprises periodically repeating said steps (c)-(f) during said predefined fabrication interval.

14. The monitoring and control method of claim 13, wherein predefined or derived limits for said plasma processing within said fabrication chamber are established, and wherein said method further comprises terminating plasma processing within said chamber whenever said regulating step (f) fails to bring said plasma processing within said predefined or derived limits through said automatic regulating of said at least one adjustable process control.

15. The monitoring and control method of claim 12, wherein said steps (c)-(f) are accomplished in real time.

16. The monitoring and control method of claim 12, wherein said step (c) and said step (d) include utilizing optical emission spectroscopy to collect said electromagnetic wavelength spectra and to determine said peak wavelength location information, respectively.

17. The monitoring and control method of claim 12, wherein said Markov random field model is pre-established for the collected wavelength spectra based on a predetermined table of vaporized species, and wherein said identifying step (e) further includes solving said Markov random field model for said predefined plurality of vaporized species using selective stochastic relaxation analysis.

18. The monitoring and control method of claim 12, wherein said providing step (a) includes providing a plasma processing system including a fabrication chamber having multiple electronically adjustable controls for regulating plasma processing within said chamber, and wherein said regulating step (f) includes producing an electronic signal to regulate each of said multiple adjustable controls pursuant to at least some of said wavelength and relative intensity data, peak location information, and vaporized species information of said steps (c)-(e), respectively.

19. The monitoring and control method of claim 12, further comprising the step of sharpening at least some of said intensity peak location information determined in said step (d) prior to said vaporized species identifying step (e).

20. The monitoring and control method of claim 19, wherein said peak sharpening includes employing an extrapolation procedure to refine wavelength and intensity data collected in said step (c).

21. A closed loop method for operating a semiconductor fabrication chamber within predefined or derived constraints, said fabrication chamber having at least one electronically adjustable mechanism for controlling processing therein, said operating method comprising the steps of:

(a) establishing a plasma process within said fabrication chamber;

(b) simultaneous with said step (a), collecting multivariate data indicative of gaseous species occurring in said plasma process established in step (a) within said chamber, said multivariate data being collected through at least one sensor associated with said fabrication chamber;

(c) using said collected multivariate data of step (b) to identify in real-time using a Markov random field model a gaseous species contained within said chamber, and its relative concentration level; and

(d) automatically regulating said plasma process within said fabrication chamber in real-time based on said identified gaseous species and its determined relative concentration level of said step (c), said regulating step including producing an electronic signal to adjust said at least one adjustable mechanism based upon the multivariate data collected and the relative concentration information determined in said steps (b) and (c), respectively, whenever plasma processing exceeds said predefined or derived constraints, said adjusting being such as to bring said plasma process within said predefined or derived constraints.

22. The closed loop method of claim 21, wherein said plasma process of step (a) is for a predefined interval, and wherein said method further comprises the step of periodically repeating steps (b)-(d) during said predefined interval.

23. The closed loop method of claim 21, wherein said collecting step (b) includes utilizing a spectrometer in association with said fabrication chamber to collect said multivariate data.

24. The closed loop method of claim 21, wherein said Markov random field model is pre-established for a plurality of preselected potential gaseous species, and said identifying step (c) includes solving said Markov random field based model using a maximization algorithm to determine the identity of at least some of a plurality of gaseous species within said chamber.

25. The closed loop method of claim 24, wherein said maximization algorithm utilized in said step (c) comprises a selective stochastic relaxation algorithm.

26. The closed loop method of claim 21, wherein said multivariate data collecting step (b) includes:

collecting from within the fabrication chamber wavelength and relative intensity data on a plurality of different electromagnetic wavelengths simultaneously occurring therein; and

wherein said gaseous species identifying step (c) includes;

determining the location of at least some intensity peaks for the simultaneously occurring plurality of different electromagnetic wavelengths from said wavelengths and relative intensity data;

sharpening at least some of said located intensity peaks; and

identifying in real time using said Markov random field model a gaseous species contained within said chamber employing the sharpened intensity peaks.

27. The closed loop method of claim 26, wherein said sharpening of at least some of said located intensity peaks includes employing an extrapolation procedure to refine the wavelength and relative intensity data of each located intensity peak.

28. An automated method for operating and controlling within predefined or derived limits a plasma processing system having a fabrication chamber with at least one electronically adjustable, primary plasma process control point, said automated method comprising the steps of:

(a) producing a plasma process within said fabrication chamber;

(b) collecting optical emission spectra data on said plasma process occurring within said fabrication chamber;

(c) locating intensity peaks in said optical emission spectra data;

(d) electronically identifying using a Markov random field model gaseous species contained within said chamber during said plasma processing based on said optical emission spectra data of step (b);

(e) determining whether any incorrect gaseous species is identified in said step (d) for said plasma process of step (a); and

(f) whenever said step (e) identifies an incorrect gaseous species, regulating in an automated manner said plasma processing by producing an electronic control signal to adjust said primary process control point to move said plasma processing of step (a) within said predefined or derived limits.

29. The automated method of claim 28, further comprising the step of sharpening at least some of said located intensity peaks in said optical emission spectroscopy data prior to said gaseous species identifying step (d).

30. The automated method of claim 28, further comprising the steps of:

determining relative concentration levels for selected ones of said identified gaseous species of step (d);

evaluating whether said determined concentration levels are within predetermined acceptable ranges; and

regulating in an automated manner said plasma processing of step (a) whenever said concentration levels are outside said predetermined acceptable ranges by producing an electronic control signal to adjust said primary process control point to move the gases occurring within said plasma processing of step (a) within said predetermined acceptable ranges.

31. A closed loop system for manufacturing a semiconductor, said closed loop system comprising:

semiconductor manufacturing apparatus including a fabrication chamber and apparatus for generating a plasma process therein;

at least one electronically adjustable mechanism for controlling plasma processing within said fabrication chamber;

means, associated with said fabrication chamber, for collecting an electromagnetic wavelength spectra from within said fabrication chamber while plasma processing is occurring therein, said collected spectra comprising wavelength and relative intensity data;

means for determining an intensity peak location for at least some of the different electromagnetic wavelengths represented by said wavelength data;

processor means coupled to said collecting means and said determining means for automatically identifying using a Markov random field model at least one vaporized species contained within said chamber during plasma processing using said wavelength and relative intensity data and said intensity peak location information; and

real-time feedback control means responsive to said processor means for regulating within predetermined limits said plasma processing occurring within said fabrication chamber based on said wavelength and relative intensity data, located peak information and identified vaporized species, said feedback control means being coupled to said at least one electronically adjustable mechanism and including means for electronically adjusting said at least one adjustable mechanism to maintain said plasma processing occurring within said fabrication chamber within said predetermined limits.

32. The closed loop method of claim 31, wherein said means for collecting an electromagnetic wavelength spectra includes an optical emission spectrometer coupled to said fabrication chamber.

33. The closed loop method of claim 31, further comprising means for sharpening the electromagnetic wavelength of each located intensity peak, said sharpening means being coupled to provide sharpened peak location information to said processor means.

34. An automated method for operating within predefined or derived limits a plasma processing system having a fabrication chamber with at least one electronically adjustable, primary plasma process control point, comprising the steps of:

(a) producing a plasma process within the fabrication chamber;

(b) simultaneous with said step (a), collecting optical emission spectra data on said plasma process produced within the fabrication chamber;

(c) locating intensity peaks in said optical emission spectra data collected in said step (b);

(d) sharpening at least some of said located intensity peaks of said step (c); and

(e) electronically identifying at least one gaseous species contained within the fabrication chamber during said plasma processing using said intensity peaks sharpened in said step (d).

35. The automated method of claim 34, wherein said intensity peak sharpening of said step (d) includes employing for each located intensity peak an extrapolation procedure to refine the corresponding optical emission spectra data collected in said step (b).

36. The automated method of claim 35, wherein said employing of said extrapolation procedure includes for each located intensity peak fitting a regression line to measured data on each side of the located intensity peak and substituting wavelength and intensity data of a point where the two regression lines intersect for corresponding optical emission spectra data collected in said step (b).

37. The automated method of claim 34, wherein said locating of intensity peaks of said step (c) includes reordering optical emission spectra data collected in said step (b) so that the data is arranged in descending order of intensity and wherein said step (c) further includes grouping the collected optical emission spectra data such that each intensity peak is defined by a group of collected optical emission spectra data.

38. The automated method of claim 34, wherein said electronically identifying step (e) includes employing a Markov random field model to identify the at least one gaseous species contained within the fabrication chamber during the plasma processing.

39. The automated method of claim 34, further comprising the step of regulating in an automated manner said plasma process of said step (a) by producing an electronic control signal to adjust said primary process control point to move said plasma process of said step (a) to within the predefined or derived limits.

40. A method for identifying at least one gaseous species present in a process, said method comprising the steps of:

(a) generating wavelength and relative intensity data from the process using an optical emissions spectrometer;

(b) providing a Markov random field model correlating wavelength and relative intensity data with particular gaseous species; and

(c) electronically identifying at least one gaseous species contained within the process by employing the Markov random field model of step (b) and the generated wavelength and relative intensity data of step (a).