An optimization loop for industrial applications includes several aspects
(e.g. of robustness, speed, etc.) in addition to the numerical optimization.
However, the basic structure of the optimization is the same for all
applications.
A typical optimization flow is depicted in Figure 4.1, where the
different parts of the optimization are presented. According to the different
levels where the optimization is applied, the interactions with the optimization
loop and the results of the different optimizations are different.
An optimization in a typical TCAD software environment includes several
applications as listed below:
- Device characteristics to improve performance and to reduce parasitic effects
as well as self-heating effects
- Single process steps to obtain more realistic device structures for
device simulations
- Sequences of single steps, for instance several process
steps (material deposition, mask deposition with optical proximity
correction, material etching, and mask stripping)
Typical optimization tasks in a semiconductor device fabrication environment
requires:
- Optimization of a single process step to increase yield for this
particular step or to identify critical parameters to apply them
for reliability calculations.
- Optimization of a technology is a steady process during the whole
life cycle of the particular semiconductor device. This includes
improving yield, and reliability of the device, and to port the process to a
smaller-scaled technology node to increase the number of devices per lot.
- The most complex task is to optimize the complete formula of a
technology, including the resulting device characteristics which often
require several weeks to months from fabrication to the final results of the
measurements of characteristics and reliability.
To decrease the time from the start of optimization until the final results are
obtained, certain aspects can be optimized in advance -- not on hardware but -
with appropriate simulations tools -- by using software. This offers the developer
several weeks' gain in time for each optimization run they can perform with
software tools before doing the experiments.
Although, the final measurements for the verify the device characteristics and
reliability cannot be avoided, the monetary benefits due to time savings
can be considerable.
Stefan Holzer
2007-11-19