NEURAL-NETWORK MODELS FOR THERMALLY BASED MICROELECTRONIC MANUFACTURING PROCESSES

This paper presents artificial neural networks (ANNs) to model thermally based microelectronic manufacturing processes. The specific processes chosen are the chemically vapor deposited (CVD) epitaxial deposition of silicon in a horizontal reactor and ''pool boiling'' as applied to vapor-phase soldering. In the CVD processes, an analytic model is used to generate data under simulated production conditions. Part of the data sets are used to train the neural network models. These models, referred to hereafter as physiconeural models, are then used to predict the output as a function of input parameters for the other part of the data sets. For pool boiling, an empirical correlation is used to train the ANN model. A comparison of these predictions with the physical model's computational results for the CVD process, and the experimental data for the pool boiling shows good agreement. These results show the effectiveness of the artificial-neural-network technique for modeling complex processes. Further work is in progress to exploit fully the potential of neural models, singly or in conjunction with physical models for run-to-run and real-time process control.