Yield improvement for GaAs IC manufacturing using neural network inverse modeling

This paper describes a neural network based method of design centering for microelectronic circuits fabrication process. Process data are first evaluated for principal components and subsequently modeled using multilayer perceptron networks in a reduced and transformed input space. Perceptron network models are then inverted, and center settings of input variables are computed by using the inverse PCA transformation. The approach allows for maximizing the fabrication yield of GaAs circuits. Example of yield maximization for MMIC fabrication process is provided to demonstrate the effectiveness of the proposed technique