Particle swarm optimization for ellipsometric data inversion of samples having an arbitrary number of layers

摘要

A method is presented for performing the inversion of ellipsometric data using a hybrid approach involving a particle swarm optimization algorithm and a Levenberg-Marquardt algorithm. A sample may be composed of any number of layers of transparent or absorbing materials on a substrate. The method described is applicable to single- or multiple-angle, single-wavelength ellipsometry. The results of the particle swarm optimization algorithm agree well with previously published data calculated using different ellipsometric inversion algorithms, and converges for wide ranges of initial parameter estimates.