Transforming the promise of generative AI into value-added applications for physical design layout analytics, mask data synthesis and lithography simulation

Machine Learning (ML) applications for Physical Design Layout analytics, Lithography modeling, and Mask Data Synthesis have reached a considerable degree of technical maturity in the past five years, and are now being deployed in production flows, particularly at advanced process technology nodes (e.g. 5nm and below). Despite these successes, Generative AI (GenAI), heralded across virtually every technology field as the next frontier and the true ML game-changer, has not yet demonstrated cost-effective solutions for advanced lithography applications. This work presents both a lithography-specific survey of the state-of-the-art in GenAI, and a 3-year roadmap towards cost-effective, production-level flows for generating layout and mask datasets (including OPC and ILT) enabled by novel process modeling with generative mechanisms. Preliminary computational results for 3 novel GenAI prototype flows exemplify how GenAI deployment could be made feasible, and simultaneously highlight the still existing data and computational scaling gap which will need to be bridged to achieve sustainable return on investment.