In the first part a practical solution to the problem of inverse modeling is presented. First generalized Bernstein polynomials are introduced into TCAD in Chapter 7, since they provide lots of advantages compared to the traditional methods subsumed under the term RSM (response surface methodology). In Chapter 8 the theory of evolutionary computation and especially genetic algorithms provides the basis for applying global optimizers to optimization and inverse modeling problems in Part III.

In the last chapter of this part the design and implementation of an inverse modeling and optimization framework is presented. In the context of previous work at the Institute for Microelectronics this new implementation improves on a previous version [139] in several ways. New optimizers were introduced and successfully applied to inverse modeling problems. The functionality and extensibility of the framework was improved by utilizing the functional aspects of Common Lisp.

Relating this work to the work of predecessors and colleagues at the Institute for Microelectronics, a framework called VISTA (Viennese Integrated System for TCAD Applications) was implemented by Dr. Ch. Pichler and others [42,33,90] in order to integrate process simulations. Dr. R. Plasun [92] introduced code for DOE (design of experiments) into the framework. Dr. R. Strasser [139] worked on optimization tools, i.e., gradient based optimizers, for the framework, which was then called SIESTA (Simulation Environment for Semiconductor Technology Analysis). Dr. M. Stockinger [132] worked on applications of the SIESTA framework. Still many aspects required and justified more work in this direction regarding optimization and approximation algorithms and the general applicability of the framework.

After the theoretic aspects in the first part the second part will be dedicated to applications in the realm of semiconductor process simulation.

Clemens Heitzinger 2003-05-08