Neural network approach to rapid thin film characterization

Authors:

Abstract
A novel approach for thin film thickness and optical constant extraction from spectral reflectance data is presented here. This methodology combines the global minimization abilities of Adaptive Simulated Annealing with the high computational efficiency of Neural Networks to solve complex characterization problems in real time. The optical constants of many thin films such as Polysilicon are a function of the processing conditions and hence the real time measurement of these parameters could possibly be used in real time or run to run process control applications.