Method and apparatus for process endpoint prediction based on actual thickness measurements

• US 5,503,707 A
• Filed: 09/22/1993
• Issued: 04/02/1996
• Est. Priority Date: 09/22/1993
• Status: Expired due to Term

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Maung, Sonny, Henck, Steven A., Butler, Stephanie W.
• Primary Examiner(s): Breneman, R. Bruce
• Assistant Examiner(s): CHANG, JONI YUK YING
• Application Number:
  US08/125,954
• Time in Patent Office: 923 Days
• Field of Search: 156/626, 156/627, 156/643.1, 427/10, 356/369, 356/366, 356/367, 356/368, 356/381, 356/382
• US Class Current: 438/5
• CPC Class Codes: C23C 14/547 using optical methods
• CPC Class Codes: G01B 11/0641 with measurement of polarization
• CPC Class Codes: H01J 37/32963 End-point detection

Abstract

In accordance with one aspect of the present invention, a method is provided for predicting the endpoint time of a semiconductor process for a layer of a wafer (14). The endpoint time is the time at which a predetermined thickness of the layer occurs. A layer thickness, calculated for a first sample time, is received. It is then determined whether or not the layer thickness lies within a predetermined range. If the layer thickness does lie within the predetermined range, it is used to update a forecasted process rate. The forecasted process rate is used to predict the endpoint time. The endpoint time is used to control the semiconductor process so that the layer of wafer (14) is formed having the predetermined thickness.

Claims

1. A method for predicting an endpoint time of a semiconductor process at which a thickness of a target layer of a wafer occurs, the method comprising the steps of:

receiving an actual thickness sample of the layer calculated at a first sample time;

determining whether or not the thickness sample is valid;

updating a forecasted process rate using the thickness sample if the thickness sample is valid;

predicting the endpoint time from the forecasted process rate; and

controlling the process to generate the layer having the desired thickness.

2. The method of claim 1 wherein the step of determining comprises the steps of:

receiving a figure of merit calculated at the first sample time;

comparing the figure of merit to an acceptable figure of merit;

determining that the thickness sample is valid if the figure of merit is less than or equal to the acceptable figure of merit.

3. The method of claim 1 wherein the step of determining comprises the steps of:

calculating a predicted layer thickness at the first sample time from a prior forecasted process rate calculated at a prior sample time;

generating a difference between the predicted layer thickness and the thickness sample;

determining from the difference a moving average residual at the first sample time;

calculating from the difference a moving absolute deviation at the first sample time;

calculating a normalized residual from the difference, a prior moving average residual and a prior moving absolute deviation, wherein both the prior moving average residual and the prior moving absolute deviation are calculated for the prior sample time;

comparing the normalized residual to a predetermined cutoff; and

determining the thickness sample is within the predetermined range if the normalized residual is less than or equal to the cutoff.

4. The method of claim 1 wherein the step of determining comprises the steps of:

calculating a raw process rate;

comparing the raw process rate to minimum and maximum process rates; and

determining that the thickness sample is within the predetermined range if the raw process rate is between the minimum and maximum process rates.

5. The method of claim 1 wherein the step of predicting further comprises the steps of:

determining whether the endpoint time will occur within a predetermined interpolation time from the first sample time; and

predicting the endpoint time if the endpoint time will occur within said predetermined interpolation time.

6. A method for predicting an endpoint time of a semiconductor process at which a thickness of a target layer of a wafer occurs, the semiconductor process occurring in a chamber, the method comprising the steps of:

receiving an actual thickness sample of the thickness of the layer and a figure of merit calculated for a first sample time;

determining from the figure of merit whether the actual thickness sample is valid;

updating a forecasted process rate using the actual thickness sample if the actual thickness sample is valid; and

predicting the endpoint time from the forecasted process rate.

7. The method of claim 6 wherein the step of determining comprises the steps of:

comparing the figure of merit to an acceptable figure of merit;

determining that the actual thickness sample is valid if the figure of merit is less than or equal to the acceptable figure of merit.

8. The method of claim 6 wherein the step of determining comprises the steps of:

comparing the figure of merit to an acceptable figure of merit;

determining that the actual thickness sample is valid if the figure of merit is greater than the acceptable figure of merit;

if the figure of merit is less than or equal to the acceptable figure of merit,

calculating a predicted thickness sample of the layer at the first sample time from a prior forecasted process rate calculated at a prior sample time;

calculating a difference between the predicted thickness sample and the actual thickness sample;

calculating from the difference a moving average residual at the first sample time;

calculating from the difference a moving absolute deviation at the first sample time;

calculating a normalized residual from the difference, a prior moving average residual and a prior moving absolute deviation, wherein both the prior moving average residual and the prior moving absolute deviation are calculated for the prior sample time;

comparing the normalized residual to a cutoff; and

determining the actual thickness sample is valid if the normalized residual is less than or equal to the cutoff.

9. The method of claim 6 wherein the step of determining comprises the steps of:

comparing the figure of merit to an acceptable figure of merit;

determining that the actual thickness sample is invalid if the figure of merit is greater than the acceptable figure of merit;

if the figure of merit is less than or equal to the acceptable figure of merit,

calculating a raw process rate;

comparing the raw process rate to a minimum process rate and a maximum process rate; and

determining the actual thickness sample is valid if the raw process rate is between the minimum and the maximum process rates.

10. The method of claim 6 further comprising the steps of:

receiving a prior figure of merit calculated for a prior sample time occurring before the process has started;

comparing the prior figure of merit to an acceptable figure of merit;

determining the wafer meets preprocessing tolerances if the prior figure of merit is less than or equal to the acceptable figure of merit.

11. The method of claim 6 further comprising the steps of:

receiving a prior figure of merit calculated for a prior sample time occurring before the process has started;

comparing the prior figure of merit to an acceptable figure of merit;

determining the wafer is properly positioned in the chamber if the prior figure of merit is less than or equal to the acceptable figure of merit.

12. A method for predicting an endpoint time at which a thickness of a target layer of a wafer under process occurs, comprising the steps of:

receiving an actual layer thickness sample of the layer calculated for a first sample time;

determining if the actual layer thickness sample is valid using Holt'"'"'s method;

updating a forecasted process rate with the actual layer thickness sample using a Holt filter if the actual layer thickness sample is valid;

predicting the endpoint time from the forecasted process rate; and

controlling the process, using the endpoint time, to produce the layer at the desired thickness.

13. The method of claim 12 wherein the step of determining comprises the steps of:

calculating a predicted layer thickness sample at the first sample time from a prior forecasted process rate calculated at a prior sample time;

calculating a difference between the predicted layer thickness sample and the actual layer thickness sample;

calculating a moving average residual, at the first sample time, from the difference;

calculating a moving absolute deviation, at the first sample time, from the difference;

calculating a normalized residual from the difference, a prior moving average residual and a prior moving absolute deviation, wherein both the prior moving average residual and the prior moving absolute deviation are calculated for the prior sample time;

comparing the normalized residual to a cutoff;

determining the actual layer thickness sample is valid if the normalized residual is less than or equal to the cutoff.

14. The method of claim 12 wherein the step of updating the forecasted process rate comprises the steps of:

calculating a raw process rate;

calculating a filtered process rate from the raw process rate and a prior forecasted process rate;

calculating a process rate trend from the filtered process rate; and

setting the forecasted process rate equal to a sum of the filtered process rate and the process rate trend.

15. The method of claim 12 further comprising the step of

setting the filtered process rate equal to an initial process rate before the process begins.

16. The method of claim 12 wherein the steps of updating and predicting comprise the steps of:

calculating a raw process rate;

calculating a filtered process rate from the raw process rate and a prior forecasted process rate;

calculating a process rate trend from the filtered process rate;

setting the forecasted process rate equal to a sum of the filtered process rate and the process rate trend;

calculating a predicted layer thickness sample at the first sample time;

conjugating the filtered process rate with the first sample time to generate a conjugation;

generating a first difference between the predicted layer thickness sample and the conjugation;

generating a second difference between the first difference and the predetermined thickness; and

dividing the second difference by the forecasted process rate to generate the endpoint time.

17. A method for predicting an endpoint time at which a desired thickness of a target layer of a wafer under process occurs, comprising the steps of:

receiving an actual layer thickness calculated by a processor for a first sample time;

calculating a raw process rate;

comparing the raw process rate to a minimum process rate and a maximum process rate;

determining the actual layer thickness is valid if the raw process rate is between the minimum and the maximum process rates;

updating a forecasted process rate with the actual layer thickness if the actual layer thickness is valid;

predicting the endpoint time from the forecasted process rate; and

controlling the process, using the endpoint time, to produce the layer at the desired thickness.

18. The method of claim 17 wherein the step of updating comprises the steps of:

updating a shifting window with the raw process rate;

computing the forecasted process rate from the shifting window.

19. The method of claim 17 wherein the step of predicting comprises the steps of:

generating a conjugation between the forecasted process rate and the first sample time;

computing a first difference between the actual layer thickness and the conjugation;

generating a second difference between the first difference and the desired thickness; and

dividing the second difference by the forecasted process rate to obtain the endpoint time.

20. A system for controlling a semiconductor process for a wafer within a process chamber and having at least one layer, the system comprising:

a generator for generating and transmitting a source signal, having a first set of parameters, toward the wafer and for receiving a reflected signal, having a second set of parameters, from the wafer;

a processor in communication with the generator for calculating for a first sample time an actual layer thickness and a figure of merit from a set of differences between the first and second sets of parameters;

an endpoint detector in communication with the processor for using the figure of merit to determine if the actual layer thickness is within a predetermined range, for updating a forecasted process rate from the actual layer thickness if the actual layer thickness is within the predetermined range and for predicting an endpoint time from the forecasted process rate; and

a controller in communication with the process chamber and the endpoint detector, the controller responsive to the endpoint time for generating control signals for controlling the semiconductor process.

21. The system of claim 20 further comprising the endpoint detector for:

comparing the figure of merit to an acceptable figure of merit;

determining that the actual layer thickness is outside of the predetermined range if the figure of merit is greater than the acceptable figure of merit;

if the figure of merit is less than or equal to the acceptable figure of merit,

calculating a predicted layer thickness at the first sample time from a prior forecasted process rate calculated at a prior sample time;

calculating a difference between the predicted layer thickness and the actual layer thickness;

calculating from the difference a moving average residual at the first sample time;

calculating from the difference a moving absolute deviation at the first sample time;

calculating a normalized residual from the difference, a prior moving average residual and a prior moving absolute deviation, wherein both the prior moving average residual and the prior moving absolute deviation are calculated for the prior sample time;

comparing the normalized residual to a predetermined cutoff; and

determining the actual layer thickness is within the predetermined range if the normalized residual is less than or equal to the predetermined cutoff.

22. The system of claim 20 comprising the endpoint detector for

comparing the figure of merit to an acceptable figure of merit;

determining that the actual layer thickness is outside of the predetermined range if the figure of merit is greater than the acceptable figure of merit;

if the figure of merit is less than or equal to the acceptable figure of merit,

calculating a raw process rate;

comparing the raw process rate to a minimum process rate and a maximum process rate; and

determining the actual layer thickness is within the predetermined range if the raw process rate is between the minimum and the maximum process rates.

23. The system of claim 20 comprising the endpoint detector for

determining whether the endpoint time will occur within a predetermined interpolation time from the first sample time;

and

predicting the endpoint time if the endpoint time will occur within the predetermined interpolation time.

24. The system of claim 20 further comprising the endpoint detector for

comparing the figure of merit to an acceptable figure of merit;

and

determining that the actual layer thickness is outside of the predetermined range of the figure of merit is greater than the acceptable figure of merit.