Unitho: A Unified Multi-Task Framework for Computational Lithography
Qian Jin, Yumeng Liu, Yuqi Jiang, Qi Sun, Cheng Zhuo
TL;DR
Unitho tackles the computational bottlenecks of lithography by presenting a unified multi-task Transformer framework that merges end-to-end mask/contour generation, lithography simulation, and hotspot detection into a single, process-aware pipeline. It employs a cross-attention fusion mechanism to integrate layout with process conditions and uses contrastive learning to sharpen structural and process sensitivity, with a three-stage training strategy (generation pre-training, detection pre-training, and joint fine-tuning). Evaluated on a large-scale industrial dataset, Unitho achieves state-of-the-art generation and detection performance, while delivering massive speedups (e.g., ~85× faster than a commercial tool) and close agreement with traditional simulators for LRC hotspots. The approach enables rapid PWV, improved design-for-manufacturing feedback, and scalable DTCO across diverse process conditions.
Abstract
Reliable, generalizable data foundations are critical for enabling large-scale models in computational lithography. However, essential tasks-mask generation, rule violation detection, and layout optimization-are often handled in isolation, hindered by scarce datasets and limited modeling approaches. To address these challenges, we introduce Unitho, a unified multi-task large vision model built upon the Transformer architecture. Trained on a large-scale industrial lithography simulation dataset with hundreds of thousands of cases, Unitho supports end-to-end mask generation, lithography simulation, and rule violation detection. By enabling agile and high-fidelity lithography simulation, Unitho further facilitates the construction of robust data foundations for intelligent EDA. Experimental results validate its effectiveness and generalizability, with performance substantially surpassing academic baselines.