From Lithography to Nanoimprint: Physics-Based, Data-Driven, and Hybrid Frameworks for Defect Prediction in Advanced Patterning Technologies

Advanced patterning technologies have evolved from direct optical image transfer into predictive manufacturing frameworks in which defects are shaped by imaging physics, material response, process variation, and computational correction. Starting from the development of resolution enhancement techniques (RET) and optical proximity correction (OPC), this review traces how conventional optical lithography became the foundation of modern computational lithography. The discussion then extends to extreme ultraviolet lithography (EUVL) and nanoimprint lithography (NIL), where defect mechanisms no longer follow the same geometry-driven correction logic. EUVL introduces stochastic resist response, photon-limited variability, and mask 3D effects, while NIL shifts the dominant defect origins toward resist filling, residual layer variation, template interaction, viscoelastic deformation, and demolding. This review focuses on two major modeling paradigms for defect prediction from an OPC-centric view: physics-based frameworks that provide mechanism-level interpretability, and data-driven frameworks that enable rapid pattern recognition, hotspot detection, layout-to-image prediction, and uncertainty-aware screening. Recent advances also point to hybrid physics–machine learning (ML) frameworks as a promising way forward by synergistically combining physical constraints, simulation-derived knowledge, and data-driven inference. In summary, this review emphasizes the transition from contour correction to physically grounded, data-efficient, and uncertainty-aware defect prediction in optical lithography, EUVL, and NIL.