Data processing: generic control systems or specific application – Generic control system – apparatus or process – Optimization or adaptive control
Reexamination Certificate
2000-10-12
2003-04-29
Follansbee, John (Department: 2121)
Data processing: generic control systems or specific application
Generic control system, apparatus or process
Optimization or adaptive control
C700S045000, C700S052000, C706S002000
Reexamination Certificate
active
06556876
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method of controlling a semiconductor fabrication process and, in particular, to a method of controlling variation in a semiconductor fabrication process utilizing a fuzzy-controlled learning feedback system.
2. Description of the Related Art
Fabrication of integrated circuits is an extremely complex process requiring the performance of carefully coordinated photolithography, etching and doping steps. Because these steps are performed in succession, process deviation from desired targets has a cumulative effect and, thus, must be carefully monitored to ensure that the final product operates within specified tolerances.
FIG. 1
depicts a conventional method for controlling process variation utilizing statistical process control (SPC). Statistical process control is a methodology for characterizing, interpreting and minimizing variation in the results of a process. Statistical process control is described extensively by Grant et al.,
Statistical Quality Control
, 6th Ed. (1988), hereby incorporated by reference. A well-established set of rules for determining process variation requiring correction is set forth in the statistical Quality Control Handbook of the Western Electric Company (1956), also hereby incorporated by reference.
As shown in
FIG. 1
, results
100
of a given semiconductor fabrication process are received and plotted on chart
102
. Chart
102
includes a target mean
104
for the plotted parameter. Chart
102
also includes a series of dashed lines indicating positive incremental deviations from target mean
104
, designated as +&PHgr;
1
, +&PHgr;
2
, and +&PHgr;
3
, and a series of dashed lines indicating negative incremental deviations from target mean
104
, designated as −&PHgr;
1
, −&PHgr;
2
, and −&PHgr;
3
.
Once the process results
100
have been plotted on chart
102
, the quantity and quality of this process variation is characterized according to rules dictated by statistical process control (SPC) analysis.
Based upon the process variation characterized by chart
102
, the engineer supervising the fabrication process must interpret this variation and then manually adjust process parameters in such a way as to minimize future variation.
While satisfactory for some applications, the conventional method of process control suffers from a number of disadvantages.
One disadvantage of the conventional method shown in
FIG. 1
is its labor intensity. Human intervention is required to compile and characterize the process variation, as well as to interpret this process variation and to adjust process parameters to reduce or eliminate variation. Therefore, there is a need in the art for a method of controlling a semiconductor fabrication process requiring as little human intervention as possible.
Another disadvantage of the conventional process is inherent delay in responding to process variation. Because a human operator must first characterize and interpret process variation before an adjustment is made, the inherent delay can result in the loss of product in the intervening time which has experienced unacceptable variation. Therefore, there is a need in the art for a method of controlling a semiconductor fabrication process which provides rapid recognition of process variation and corresponding adjustment of process parameters.
Yet another disadvantage of the conventional process is the inherent variability in process control that is introduced by manual interpretation of process variation. While the SPC rules described above aid somewhat in recognizing process variation, the decision of whether or not to adjust process parameters is influenced by such non-easily quantifiable factors as the operator's experience and discretion. Thus, the ultimate decision regarding whether or not to modify the process parameters based upon a given type of variation is not automatically reproducible. This manual form of decision-making can, in turn, lead to non-uniform process control. Therefore, there is a need in the art for a method of controlling a semiconductor fabrication process which provides reproducible modification of process parameters given a specific process variation.
In summary, there is a need in the art for a method of process control which responds quickly to deviation to ensure a minimum amount of wasted material.
SUMMARY OF THE INVENTION
The present invention relates to a method of controlling variation in a semiconductor fabrication process utilizing a fuzzy-controlled learning system. The fuzzy-controlled learning system receives as inputs variation in the results of a semiconductor process flow. Utilizing a fuzzy rule base, an inference engine of the fuzzy system generates fuzzy output sets which dictate the parameters of the semiconductor fabrication process. Automatic feedback of the output of the fuzzy-controlled learning system keeps process variation low with a minimum of manual human intervention.
REFERENCES:
patent: 5283575 (1994-02-01), Koa et al.
patent: 5497331 (1996-03-01), Iriki et al.
patent: 5544256 (1996-08-01), Breecher et al.
patent: 5767785 (1998-06-01), Goldberg
patent: 5822740 (1998-10-01), Haissig et al.
patent: 6101444 (2000-08-01), Stoner
patent: 6125976 (2001-07-01), Lemelson et al.
patent: 6282526 (2002-08-01), Ganesh
1942 -G. J. Meyers, Jr. Why the Control Chart Works; Some Examples. Statistical Quality Control, Sixth Edition.
1956—Wester Electric Company, Introduction to Statistical Quality Control and Introduction to Control Charts. Statistical Quality Control Handbook.
1997—J.-S.R. Jang et al., Introduction to Neuro-Fuzzy and Soft Computing Chapters 1-4. Neuro-Fuzzy and Soft Computing. A Computational Approach to Learning and Machine Intellegence.
Ferland Patrick H.
Prosack Hank
Swan Kenneth
Follansbee John
Hartman Jr. Ronald D
National Semiconductor Corporation
Stallman & Pollock LLP
LandOfFree
Hybrid fuzzy closed-loop sub-micron critical dimension... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Hybrid fuzzy closed-loop sub-micron critical dimension..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Hybrid fuzzy closed-loop sub-micron critical dimension... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3027200