Physics-informed generative AI for semiconductor manufacturing: Enforcing hard physical constraints in generative models by construction

This paper addresses the challenge of applying generative models to semiconductor manufacturing, where generated outputs must obey hard physical constraints rather than merely appearing perceptually plausible. The authors propose enforcing constraints by construction in the generative model architecture. Applications include mask generation, layout optimization, synthetic defect data creation, and process recipe derivation, all subject to lithography, transport, reaction, and device-physics constraints. This work provides a principled approach for incorporating physical laws into generative AI for semiconductor process optimization.