Optimization of variable frequency microwave curing using neural networks and genetic algorithms

Abstract

Variable frequency microwave (VFM) curing can perform the same processing steps as conventional thermal processing in minutes, without compromising the intrinsic material properties. This paper presents a central composite inscribed (CCI) experiment, which is used to characterize VFM curing of polyimide, and the subsequent modeling and optimization of the process. The statistically designed experiment is performed on samples of polyimide spin-cast on silicon wafers that are cured in a VFM fumace. During VFM processing, the temperature of the polyimide samples is ramped at an appropriate level and held for a specific amount of time. Through-plane and in-plane indices of refraction are subsequently measured using a Metricon prism coupler. The percent of imidization is measured using attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy. For process modeling, we train a neural network using back-propagation and the data from the designed experiment to model the variation of the output responses as a function of the input variables. The input variables are ramp rate, temperature, and time-at-temperature. The output variables are the in-plane and through-plane indices of refraction, the birefringence, and the percent of imidization. The neural network models are then used for process optimization via genetic algorithms. Using this approach, we determine the appropriate input conditions to achieve desirable film properties.