Demonstration of the feasibility of a complete ellipsometric characterization method based on an artificial neural network - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Applied optics Année : 2009

Demonstration of the feasibility of a complete ellipsometric characterization method based on an artificial neural network

Résumé

Ellipsometry is an optical technique that is widely used for determining optical and geometrical properties of optical thin films. These properties are in general extracted from the ellipsometric measurement by solving an inverse problem. Classical methods like the Levenberg–Marquardt algorithm are generally too long, depending on direct calculation and are very sensitive to local minima. In this way, the neural network has proved to be an efficient tool for solving these kinds of problems in a very short time. Indeed, it is rapid and less sensitive to local minima than the classical inversion method. We suggest a complete neural ellipsometric characterization method for determining the index dispersion law and the thickness of a simple SiO2 or photoresist thin layer on Si, SiO2, and BK7 substrates. The influence of the training couples on the artificial neural network performance is also discussed.
Fichier non déposé

Dates et versions

hal-00419584 , version 1 (24-09-2009)

Identifiants

Citer

Y. Battie, Stéphane Robert, Issam Gereige, D. Jamon, Michel Stchakovsky. Demonstration of the feasibility of a complete ellipsometric characterization method based on an artificial neural network. Applied optics, 2009, 48 (28), pp.5318-5323. ⟨10.1364/AO.48.005318⟩. ⟨hal-00419584⟩
64 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More