An extensible TCAD optimization framework combining gradient based and genetic optimizers

C. Heitzinger, S. Selberherr

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Scopus citations

Abstract

Our Simulation Environment for Semiconductor Technology Analysis (Siesta) is a flexible, user programmable tool for optimization and inverse modeling of semiconductor devices. It is easily customizable through an interactive, object-oriented and functional scripting language. Dynamic load balancing enables to take advantage of a cluster of hosts with minimal requirements on the software infrastructure. Our approach combines the advantages of gradient based and evolutionary algorithm optimizers into one framework. Gradient based optimizers are well-suited for finding local extrema. Evolutionary algorithm optimizers add the capability of finding global extrema and thus make unattended optimizations without guessing starting values possible. Experiments can be interactively set up and tested. Bindings for the most common simulation tools are provided, and new bindings can easily be integrated taking advantage of the object-oriented and functional design. Results of experiments are saved in an object database and can be interactively retrieved as starting points for further computations or for visualizations. The user may impose arbitrary constraints (as functions defined on the parameter space) on the set in which solutions are searched. Evaluating the constraints before any simulation tool is called and getting rid of useless combinations of parameter values saves computation time and eliminates the risk of the simulation tools being called with input values that might lead to unforeseen behavior. The combination of gradient based and evolutionary algorithm optimizers enables many new optimization strategies and includes convenient handling of results.

Original languageEnglish (US)
Title of host publicationProceedings of SPIE - The International Society for Optical Engineering
EditorsB. Courtois, S.N. Demidenko, L.Y. Lau
Pages279-289
Number of pages11
Volume4228
DOIs
StatePublished - 2000
Externally publishedYes
EventDesign, Modeling, and Simulation in Microelectronics - Singapure, Singapore
Duration: Nov 28 2000Nov 30 2000

Other

OtherDesign, Modeling, and Simulation in Microelectronics
CountrySingapore
CitySingapure
Period11/28/0011/30/00

Keywords

  • Evolutionary computation
  • Genetic optimization
  • Gradient based optimization
  • Inverse modeling
  • Optimization
  • TCAD

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Condensed Matter Physics

Fingerprint Dive into the research topics of 'An extensible TCAD optimization framework combining gradient based and genetic optimizers'. Together they form a unique fingerprint.

  • Cite this

    Heitzinger, C., & Selberherr, S. (2000). An extensible TCAD optimization framework combining gradient based and genetic optimizers. In B. Courtois, S. N. Demidenko, & L. Y. Lau (Eds.), Proceedings of SPIE - The International Society for Optical Engineering (Vol. 4228, pp. 279-289) https://doi.org/10.1117/12.405424