Optical proximity correction (OPC) has become increasingly important in the lithography process as integrated circuits continue to decrease in size. One promising solution for OPC is level set-based inverse lithography technology (ILT), which offers high pattern fidelity. However, ILT is computationally intensive and is mainly used for correcting partial layers and hotspot regions. Deep learning (DL) methods have shown potential in accelerating ILT, but their lack of domain knowledge limits their effectiveness in process window (PW) enhancement. In this study, we propose an approach called inverse lithography physics-informed deep neural level set (ILDLS) for mask optimization. This approach combines level set-based ILT with DL to significantly improve printability and PW compared to pure DL and ILT. The computational time is reduced by several orders of magnitude compared to ILT alone. By incorporating knowledge of inverse lithography physics, ILDLS provides an efficient solution for mask optimization.
Mask Optimization using Deep Neural Level Set with Inverse Lithography and Physics-informed Approach. (arXiv:2308.12299v1 [eess.IV])
