Improved model-based infrared reflectrometry for measuring deep trench structures

Abstract

Model-based infrared reflectrometry (MBIR) has been introduced recently for characterization of high-aspect-ratio deep trench structures in microelectronics. The success of this technique relies heavily on accurate modeling of trench structures and fast extraction of trench parameters. In this paper, we propose a modeling method named corrected effective medium approximation (CEMA) for accurate and fast reflectivity calculation of deep trench structures. We also develop a method combining an artificial neural network (ANN) and a Levenberg-Marquardt (LM) algorithm for robust and fast extraction of geometric parameters from the measured reflectance spectrum. The simulation and experimental work conducted on typical deep trench structures has verified the proposed methods and demonstrated that the improved MBIR metrology achieves highly accurate measurement results as well as fast computation speed.