计算机科学
曲线坐标
过程(计算)
弹丸
一次性
工作流程
平版印刷术
管道(软件)
直线(几何图形)
忠诚
还原(数学)
高保真
GSM演进的增强数据速率
生产线
光学接近校正
投影(关系代数)
流离失所(心理学)
合成数据
电子工程
人工智能
计算机视觉
作者
Tianyi Li,Kai Ma,Yiwen Wu,Jiaqi Liu,Jingyi Yu A,Hao Geng
标识
DOI:10.1109/tcad.2026.3667094
摘要
Curvilinear masks play a crucial role in maintaining the lithographic fidelity for advanced nodes, but severe mask rule check (MRC) violations frequently emerge in conventional variable-shaped beam (VSB) mask data preparation (MDP) workflows. These conventional workflows rely on shape-based or dose-based mask process correction (MPC) prior to shot fracturing. Shape-based MPC methods rely on the movement of manually designed segmented lines for edge displacement corrections, which fail for complex curvilinear designs with dense line segments of arbitrary directions. While dose-based approaches handle curvilinear geometries, they are computationally intensive and incompatible with most VSB mask writers. Worse, subsequent fracturing often compromises the MPCed mask to enhance manufacturability, degrading mask quality. In this work, we propose Shot Splatting, an innovative MPC framework that integrates traditional MPC and shot fracturing into a unified optimization process, achieving end-to-end MDP processing. By optimizing shot fracturing strategies in conjunction with geometry, placement, size, and dose, our method achieves 22.3% EPE reduction and 7.87% fewer shots without introducing extra MRC violations, at a moderate computational cost of 5.01s compared to 3s for the state-of-the-art method. The output of our framework, i.e., optimized shots, can be directly formatted for VSB writers, enabling seamless fabrication.
科研通智能强力驱动
Strongly Powered by AbleSci AI