For leading edge technology node, many proximity effects during mask manufacturing process will change the mask details. Model-based Mask error correction (MEC) is needed for ensuring the mask fidelity. With the development of multi beam mask writers (MBMW), curvilinear mask offers many quality and performance advantages over Manhattan mask. It offers superior process window comparing to Manhattan mask for EUV process. In this paper, we discuss the results of model based curvilinear MEC based on Proteus platform. The quality and performance were compared between conventional compact model and Machine-Learning (ML) models. ML-based model can be accurately predicting mask printing signatures otherwise could not be predicted by convection compact model. Integrating MEC into Proteus platform offers seamless flow between different applications, like OPC, ILT and RET while preserve the device hierarchy.
|