Multi-Layer/Multi-Input/Multi-Output (MLMIMO) Models and Method for Using
First Claim
Patent Images
14. A method for using a Multi-Layer/Multi-Input/Multi-Output (MLMIMO) model to create gate structures on a plurality of substrates, the method comprising:
- a) receiving a first set of substrates and associated substrate data, the substrate data including real-time and historical data;
b) establishing a first number (I) of disturbance variables DV(I) using real-time integrated metrology (IM) data associated with a patterned photoresist layer on one or more incoming substrates, wherein the real-time IM data includes critical dimension (CD) data, sidewall angle (SWA) data, thickness data, photoresist data, BARC data, wafer substrate data, and diffraction signal data from multiple sites in the patterned photoresist layer on each incoming substrate, wherein l is a first integer greater than two;
c) establishing a second number (m) of manipulated variables MV(m), wherein m is a second integer greater than two;
d) establishing a third number (n) of control variables, wherein n is a third integer greater than two and CV(n) is defined as
CV(n)=fn{MV(1), . . . MV(m−
1), MV(m), DV(1), . . . , DV(l−
1), DV(l)}+offsetsn e) calculating optimized process settings using a quadratic objective function, and target deviations t(n) defined as;
t(n)={DV(n}−
target CV(n)};
f) calculating a plurality of manipulated variables MV(l);
g) defining an adjusted process recipe using one or more of the calculated manipulated variables MV(l) established during nonlinear programming;
h) processing one or more of the first set of substrates using the adjusted process recipe;
i) obtaining additional measurement data for one or more of the processed substrates, wherein new controlled variable (CV) data is obtained and filtered;
j) calculating one or more process errors using differences between measured control variable data and predicted control variable data;
k) calculating feedback data items, wherein errors are used to update the offsetsn using an exponentially weighted moving average (EWMA) filter;
l) updating the model offsetsn in an optimizer unit; and
m) repeating steps a)-l) using each substrate in the first set of substrates.
1 Assignment
0 Petitions
Accused Products
Abstract
The invention provides a method of processing a substrate using multilayer processing sequences and Multi-Layer/Multi-Input/Multi-Output (MLMIMO) models and libraries that can include one or more masking layer creation procedures, one or more pre-processing measurement procedures, one or more Partial-Etch (P-E) procedures, one or more Final-Etch (F-E) procedures, and one or more post-processing measurement procedures.
-
Citations
18 Claims
-
14. A method for using a Multi-Layer/Multi-Input/Multi-Output (MLMIMO) model to create gate structures on a plurality of substrates, the method comprising:
-
a) receiving a first set of substrates and associated substrate data, the substrate data including real-time and historical data; b) establishing a first number (I) of disturbance variables DV(I) using real-time integrated metrology (IM) data associated with a patterned photoresist layer on one or more incoming substrates, wherein the real-time IM data includes critical dimension (CD) data, sidewall angle (SWA) data, thickness data, photoresist data, BARC data, wafer substrate data, and diffraction signal data from multiple sites in the patterned photoresist layer on each incoming substrate, wherein l is a first integer greater than two; c) establishing a second number (m) of manipulated variables MV(m), wherein m is a second integer greater than two; d) establishing a third number (n) of control variables, wherein n is a third integer greater than two and CV(n) is defined as
CV(n)=fn{MV(1), . . . MV(m−
1), MV(m), DV(1), . . . , DV(l−
1), DV(l)}+offsetsne) calculating optimized process settings using a quadratic objective function, and target deviations t(n) defined as;
t(n)={DV(n}−
target CV(n)};f) calculating a plurality of manipulated variables MV(l); g) defining an adjusted process recipe using one or more of the calculated manipulated variables MV(l) established during nonlinear programming; h) processing one or more of the first set of substrates using the adjusted process recipe; i) obtaining additional measurement data for one or more of the processed substrates, wherein new controlled variable (CV) data is obtained and filtered; j) calculating one or more process errors using differences between measured control variable data and predicted control variable data; k) calculating feedback data items, wherein errors are used to update the offsetsn using an exponentially weighted moving average (EWMA) filter; l) updating the model offsetsn in an optimizer unit; and m) repeating steps a)-l) using each substrate in the first set of substrates. - View Dependent Claims (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 17, 18)
-
-
15-1. The method as claimed in claim 14, wherein the weightings wj are dynamically updated based on feedback error of each CV term.
Specification