Optical scatterometry of subwavelength diffraction gratings: neural-network approach

Abstract

Optical scatterometry is a method for the on-line measurement of the geometry of a diffraction grating, which is deduced from diffraction-pattern data. We demonstrate the use of a neural network as a promising method for performing an accurate quantitative characterization of the geometry. As an example, we show the deduction of the geometry of a grating with subwavelength grooves with a rms accuracy of 1.9 degrees for the slope of the groove walls, 0.7 nm for the linewidth, and 1.0 nm for the groove depth.