Computational overlay as enabler for enhanced on-product overlay control

Abstract

Computational overlay is developed based on a hybrid approach that combines physical modeling with machine learning. The performance of computational overlay is evaluated on the critical overlay between contact and gate layers. A prediction performance of ~0.7 is achieved in terms of the R2 statistic. The computational overlay model is able to follow variations in overlay, and can be used to establish a link between sources of overlay errors and the actual overlay performance. Furthermore, we will assess how computational overlay-based exposure corrections can be used to reduce the intra-field magnification error variation that is observed.