Enabling Robust Inverse Lithography with Rigorous Multi-Objective Optimization

Inverse lithography technology (ILT) shows great power in optical proximity correction, which enlarges the solution space of mask optimization and generates high-quality masks in terms of various criteria, including process window. Optimizing the process window involves improving the fidelity of printed wafer patterns on various process conditions. It is non-trivial to explicitly optimize the process window during ILT optimization, which is essentially a multi-objective optimization problem. In this paper, we propose a robust inverse lithography method, RMO-ILT, to optimize the process window effectively. Instead of aggregating all the objectives into a single one, we target the multi-objective optimization directly and explicitly. Specifically, we design a rigorous multi-objective optimization algorithm that computes uniform gradients during the mask optimization process. Furthermore, we improve the algorithm efficiency from both the algorithm level and implementation level to address the intrinsic increase in the computation overhead, significantly reducing the time consumption and enhancing the scalability. Experimental results show that the proposed algorithm achieves superior performance on the process window.