Neural network approach to rapid thin film characterization

摘要

Abstract: A novel approach for thin film thickness and opticalconstant extraction from spectral reflectance data ispresented here. This methodology combines the globalminimization abilities of Adaptive Simulated Annealingwith the high computational efficiency of NeuralNetworks to solve complex characterization problems inreal time. The optical constants of many thin filmssuch as Polysilicon are a function of the processingconditions and hence the real time measurement of theseparameters could possibly be used in real time or runto run process control applications.!10