Fast lithographic thick-mask model using overlapped patch matching

Modeling thick-mask effects is essential for lithography simulation at advanced technology nodes. This paper proposes a fast and accurate learning-based thick-mask model, dubbed the fast diffraction transfer matrix (F-DTM) model, to solve this problem in deep ultraviolet lithography. The proposed method decomposes the whole mask pattern into overlapped patches. A set of diffraction transfer matrices (DTMs) is pre-calibrated, mapping the mask patches of different geometric features to the corresponding thick-mask diffraction near-fields. The overlapping decomposition can effectively alleviate the crack effects along the decomposition boundaries, thus reducing the model errors. Additionally, an acceleration technique is proposed to greatly improve the computational efficiency of DTMs, breaking through the speed bottleneck for model calibration. The results show that the proposed methods can effectively improve the calculation accuracy and efficiency compared to the traditional thick-mask models used extensively.