Using Feedforward in on-Product Overlay Run-to-Run Control Loop for Reducing Lot-to-Lot Variation for a MEOL Layer of an Advanced Logic Node

For On-Product Overlay (OPO) control in high volume manufacturing, the process induced lot-to-lot (L2L) variation is still a persistent challenge. In this paper, we investigated a large data-set to find the root-cause and eventually the L2L variation in a critical layer of an advanced logic node. We identify the cause of large L2L variation is process change in one of the previous layers, and find out a high-order (HO) correction feedforward (FF) method which is able to reduce OPO and re-work rate by 19% and 86%, respectively. We show further that even after that process is stable, this FF is still useful for reducing OPO by 11%. Integrating the aforementioned FF control into the fab automation system has been realized in combination with ASML Litho Insight product (LIS).