Although models used in TCAD simulators are very sophisticated, there remain
uncertainties in some of the parameters of a model. This is due to the
abstraction that takes place whenever a physical effect is approximated by a
mathematical model. Such models usually contain parameters that serve as
physical constants but the exact values are either not very well known or not
measurable respectively. Other parameters (*fitting parameters*) might not
even be physically motivated but serve as a help to best characterize a certain
effect. Calibration is used to fit such parameters of a simulator model to
measured data. An uncalibrated simulator can not be expected to deliver accurate
results.

From the point of view of the optimization this task is quite
similar to the *inverse modeling* task as described above. The only difference is that there
is no need for a parameterizable `Wafer` model. Instead, a `Wafer` that best
characterizes the process family should be used.

2003-03-27