Search papers, labs, and topics across Lattice.
This paper introduces a gradient-based inverse lithography technique for extreme ultraviolet (EUV) masks, leveraging a novel framework that integrates the differentiable waveguide method with a waveguide neural operator (WGNO) as a physics engine. By employing automatic differentiation of the full forward diffraction model, the method effectively recovers the permittivity of the mask's absorber materials. Numerical experiments demonstrate that this approach can successfully design mask structures that meet specified field requirements on the wafer, significantly advancing the precision of EUV lithography.
Achieving precise EUV mask designs through a physics-informed neural operator could revolutionize semiconductor manufacturing processes.
Gradient-based inverse lithography technology~(ILT) for extreme ultraviolet~(EUV) masks is presented. A novel framework treats the differentiable waveguide method and the recently proposed waveguide neural operator~(WGNO) as end-to-end physics engines, recovering the permittivity of the absorber of the mask through automatic differentiation of the full forward diffraction model. Numerical experiments on realistic 2D and 3D absorbers of the mask (TaBN, La, U) at $位{=}11.2$~nm show that the considered ILT methods make it possible to obtain a mask structure that achieves the desired field on the wafer.
