Inverse lithography source and mask optimization via Bayesian compressive sensing

Source and mask optimization (SMO) is a key technique to guarantee the lithographic fidelity for 14-5 nm technology nodes. The balance between lithography fidelity and computational efficiency is a big issue for SMO. Our earlier works of compressive sensing SMO (CS-SMO) effectively accelerated the SMO procedure by sampling monitoring pixels. However, the imaging fidelity of the results of these methods can be further improved. This paper proposes a novel Bayesian compressive sensing source and mask optimization (BCS-SMO) method, to the best of our knowledge, to achieve the goals of fast SMO and high fidelity patterns simultaneously. The SMO procedure can be achieved by solving as a series of re-weighted l₁-norm reconstruction problems, and the weights can be updated in every iteration. The results demonstrate that, with similar computational efficiency, the BCS-SMO method can significantly improve lithographic fidelity over the current CS-SMO method.