Efficient and accurate lithographic imaging through physics-informed neural networks

Lithography plays a crucial role in semiconductor manufacturing, where the ongoing demand for smaller feature sizes has introduced considerable challenges, especially with the advent of Extreme Ultraviolet (EUV) lithography. This study investigates the application of Physics-Informed Neural Networks (PINNs) in lithographic imaging, demonstrating their superior accuracy and computational efficiency compared to traditional simulation methods. By incorporating the Abbe principle and Fourier transforms within the PINN framework, we achieved a marked improvement in computational speed without compromising predictive accuracy. These findings suggest that PINN-based approaches hold significant promise for advancing lithographic simulations, particularly in the realm of EUV lithography.