Rapid ellipsometric imaging characterization of nanocomposite films with an artificial neural network

Abstract

Imaging ellipsometry is an optical characterization tool that is widely used to investigate the spatial variations of the opto-geometrical properties of thin films. As ellipsometry is an indirect method, an ellipsometric map analysis requires a modeling step. Classical methods such as the Levenberg–Marquardt algorithm (LM) are generally too time consuming to be applied on a large data set. In this way, an artificial neural network (ANN) approach was introduced for the analysis of an ellipsometric map. As a proof of concept this method was applied for the characterization of silver nanoparticles embedded in a poly-(vinyl alcohol) film. We demonstrate that the LM and ANN give similar results. However, the time required for the ellipsometric map analysis decreases from 15 days for the LM to 1 s for the ANN. This suggests that the ANN is a powerful tool for fast spectroscopic-ellipsometric-imaging analysis.