Modeling, optimization, monitoring, and control of polymer dielectric curing by variable frequency microwave processing

Several statistically designed experiments were performed to verify that VFM curing results in comparable material properties to conventionally cured films. These experiments were on samples of the polymer benzocyclobutene (BCB) and the polyimide PI 2611 cured on silicon wafers. Curing was performed in the Lambda Technologies MircoCure(TM) 2100 system, as well as a conventional thermal furnace. All samples were heated to an appropriate temperature and held at temperature for a specific amount of time for both processing methods. After the BCB samples cooled, through-plane and in-plane indices of refraction were measured via ellipsometry. Upon cooling of the PI 2611 samples, the indices of refraction were measured using a Metricon prism coupler. The percent imidization was measured using attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy.; Neural networks were then trained using the data from the designed experiments to model the variation of the responses as a function of the process conditions. The back-propagation algorithm was utilized to train the neural networks. The network inputs were temperature and time-at-temperature. The outputs for cured BCB were the in-plane and through-plane indices of refraction. The indices of refraction were then used as metrics to determine the extent of cure of the BCB. The output variables of the cured polyimide were the in-plane and through-plane indices of refraction, the birefringence, and percent of imidization. To validate the neural network models, the root-mean-square (RMS) error was used as a performance metric. The neural network models were then used for process optimization via genetic algorithms. Using this approach, the appropriate input conditions to achieve desirable film properties were determined.; To enhance process monitoring for the VFM system, an acoustic temperature sensor was developed by depositing a ZnO transducer onto a sapphire buffer rod or sapphire wafer. The acoustic temperature sensor operates on the principle that the velocity of an acoustic wave in silicon is a function of temperature. Thus, an estimate of the temperature of a silicon wafer can be determined by measuring the time it takes for an acoustic wave to travel through silicon. For this research, the time-of-flight of a longitudinal wave is measured by generating a high voltage pulse, which produces the acoustic wave in the transducer. The acoustic wave travels through the sapphire into the silicon wafer and reflects off of either the silicon/air interface (or polymer/air interface if the silicon is coated with a polymer) and returns back to the transducer which then converts the mechanical signal back to an electrical signal, where it is read on an oscilloscope.; A sensor holder was developed to house the acoustic sensor for use in the VFM furnace. Once the sensor was mounted in the holder and placed inside the VFM furnace, a GUI developed in MATLAB was used to monitor both the time-of-flight of the returned pulses and a thermocouple attached to a polymer-coated wafer. The time-of-flight data had a strong correspondence to the temperature profile from the thermocouple that was attached to the polymer-coated wafer.; Based on the capability of neural networks to model nonlinear processes, a neural network indirect adaptive control scheme was developed for VFM curing of polymer dielectrics. This adaptive control scheme was used along with the ATS to monitor and control the temperature during curing of a polymer on a silicon wafer in the VFM furnace.