3.
Production
Process Characterization
3.4. Data Analysis for PPC 3.4.3. Building Models


Sometimes we want to use a physical model  Sometimes, rather than approximating response behavior with polynomial models, we know and can model the physics behind the underlying process. In these cases we would want to fit physical models to our data. This kind of modeling allows for better prediction and is less subject to variation than polynomial models (as long as the underlying process doesn't change).  
We will use a CMP process to illustrate  We will illustrate this concept with an example. We have collected data on a chemical/mechanical planarization process (CMP) at a particular semiconductor processing step. In this process, wafers are polished using a combination of chemicals in a polishing slurry using polishing pads. We polished a number of wafers for differing periods of time in order to calculate material removal rates.  
CMP removal rate can be modeled with a nonlinear equation 
From first principles we know that removal rate changes with time.
Early on, removal rate is high and as the wafer becomes more planar
the removal rate declines. This is easily modeled with an exponential
function of the form:


A nonlinear regression routine was used to fit the data to the equation 
The equation was fit to the data using a nonlinear regression routine.
A plot of the original data and the fitted line are given in the image
below. The fit is quite good. This fitted equation was subsequently used
in process optimization work.
