基于混合模型的光刻图形OPC校正精度提升方法

打开文本图片集

中图分类号：TN305.7；TN4 文献标志码：A 文章编码：1672-7274（2026）02-0065-03

Hybrid Model-Based Method for Improving Lithography OPC Accuracy

CHENG Ge

（Cheng Ge Beijing Yandong Microelectronics Technology Co.， Ltd.， Beijing ， China）

Abstract： As advanced integrated circuit manufacturing scales toward smaler technology nodes， lithographic imaging approaches its physical limits，and optical proximity effects increasinglydegrade patern transfer accuracy. Rule-based optical proximity correction （OPC）methods show limited efectiveness forcomplex layouts，while purely physics-based OPC methods struggle to balance correction accuracy and computational cost.To addressthese issues， this work proposes a hybrid OPC approach that combines conventional physical optical modeling with deep learningbasedresidualcorrection.Physical model predictions are incorporatedas prior information inthe initial correction stage，and neural network training samples are integrated into the conventional OPC workflow.By exploiting the ability of neural networks to compensate for local systematic deviations，stable correction of complex patterns is achieved using a staged parameter optimization strategy and an adaptive iterative scheme.Experiments conducted onrepresentative metal interconnect and contact hole paterns demonstrate that the proposed method reduces edge placement error anditerativecomputational cost while maintaining correction stability，and exhibits good adaptability across different technology nodes.These results indicate that the hybrid modeling approach is suitable for practical OPC engineering applications.

Keywords： optical proximity correction; hybrid model; deep learning; edge placement error; lithography

0 引言

光刻技术是集成电路制造的核心工艺，其图形转移精度直接决定芯片性能与良率，随着工艺节点的推进，Design的特征尺寸已接近或小于光刻机曝光波长，产生了光学邻近效应导致wafer上的图形与设计图形产生显著偏差，光学邻近校正（OPC）技术因此成为保障图形保真度的关键手段。（剩余3837字）