SBIR Phase I: An Artificial Intelligence System to Accelerate Semiconductor Production using Physics-embedded Lithographic Foundation Model

Award ID: 2336079

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to expedite the wide adoption of next-generation semiconductor chips, which is a major factor in driving technological innovation across industries and societies. As technologies rapidly evolve, shifting to extreme ultraviolet lithography (EUV) systems in semiconductor manufacturing has significantly increased design and manufacturing complexities, leading to prohibitively high costs and stifling innovation. This project aims to alleviate the design and manufacturing bottlenecks by integrating leading-edge artificial intelligence into these complex processes. This innovation aims to significantly boost efficiency, reduce costs, and accelerate time-to-market for new chip designs, overcoming current limitations in next-generation process nodes. Importantly, this proposal is poised to strengthen domestic semiconductor capabilities, a crucial element for maintaining U.S. national security, global competitiveness, and technological leadership. This Small Business Innovation Research Phase I project is focused on advancing state-of-the-art artificial intelligence for simulating photolithography in rapidly emerging semiconductor technologies. As technology evolves and process precisions improve, minor design and manufacturing deviations, such as the 3D mask effect and stochastic variations, can no longer be neglected. Addressing this arising technical challenge requires a swift and precise simulation tool, essential for optimizing yield, throughput, and time-to-market, to maintain competitiveness in this market. The proposed work will create the Lithography Foundation Model (LFM), a system with physics integrated deeply into its framework that understands the intricate dynamics of extreme ultraviolet lithography processes. The technical approach of embedding physical modeling into LFM enables rigorous accuracy across any permutations of process conditions. Coupled with leading-edge hardware-software optimization, LFM promises real-time simulations with exceptional precision. The versatility and modularity of LFM enables applications for various processes, including process simulation, layout correction, and manufacturability optimization. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Program Director: Peter Atherton