Physics-Informed AI and 3D Computational Lithography for Next-Generation High-NA EUV Patterning

As the semiconductor industry advances beyond the Angstrom era, High-NA and future Hyper-NA EUV lithography are becoming essential enablers for manufacturing the most advanced logic technologies. However, patterning at dimensions below 20 nm introduces fundamental challenges arising from stochastic effects, material variability, optical limitations, and complex interactions between design, mask, resist, scanner, and etch processes. These effects can lead to patterning failures, defects, reduced depth of focus, and diminished manufacturing yield.

This PhD project aims to develop a next-generation computational framework that combines rigorous 3D physics-based lithography simulations with modern machine learning techniques to understand, predict, and mitigate stochastic defect formation in advanced EUV patterning. The work will focus on metal logic structures and emerging High-NA EUV applications where resist stochasticity interacts with optical imaging conditions, mask design, focus control, and post-lithography pattern transfer processes.

The candidate will create large-scale simulation and experimental datasets describing photon absorption, resist chemistry, resist profile evolution, defect generation, metrology signatures, and etch transfer behavior. Advanced physics-informed machine learning algorithms will then be developed to model defect probabilities, optimize process conditions, and identify the most influential patterning parameters. The research will further explore innovative concepts such as 3D-engineered dry resists with depth-dependent material properties, enabling unprecedented control of resist profiles and improved pattern fidelity.

A unique aspect of this project is the integration of novel directional etch experiments with advanced computational lithography. The candidate will investigate how engineered directional etch processes can selectively modify three-dimensional resist profiles, enabling improved critical-dimension control, reduced stochastic failures, and enhanced pattern fidelity.

In addition, the project will establish links between lithographic process conditions and device-level electrical performance through integration with state-of-the-art TCAD platforms. The goal is to create predictive digital-twin methodologies capable of accelerating technology development while reducing costly experimental wafer learning cycles.

The research offers a unique opportunity to work at the intersection of computational physics, machine learning, semiconductor manufacturing, advanced materials, and nanoelectronics. The outcomes will directly contribute to enabling future technology nodes beyond A10 and support industrial roadmaps toward High-NA and Hyper-NA EUV manufacturing.