High accurate process proximity correction based on empirical model for 0.18 /spl mu/m generation and beyond

A pattern correction engine with high accuracy was developed for a full chip level process proximity correction (PPC) system. The performance was demonstrated by core circuit patterns with 0.18/spl mu/m-ground rule in a test site of 1 giga-bit dynamic random access memory (DRAM). There are two main important points for high accurate PPC. One is the prediction accuracy of the pattern feature after etching. The conventional single gaussian model is not sufficient but the empirical model based on multiple gaussian functions is necessary for predicting the feature after etching. Another is the fitting accuracy of the empirical model to the experimental data without process fluctuation and measurement noise. This paper focus on the empirical model fitting and the prediction accuracy.