Design of a high transmission illumination optics for anamorphic EUV lithography optics using deep reinforcement learning.

The design of the illumination optics for high numerical aperture (NA) anamorphic extreme ultraviolet (EUV) projection optics is a critical challenge to EUV lithography in advanced technology node. However, the EUV illumination optics design using conventional methods have flaws in illumination efficiency and illumination uniformity due to the limitations of relay configuration and matching method that can only consider one factor affecting illumination uniformity. One-mirror configuration can improve illumination efficiency by reducing the number of mirrors. Deep reinforcement learning (RL) can solve the limitations by considering multiple factors simultaneously. In this paper, a design method for a high transmission relay system and a matching method for double facets using deep RL are proposed to design NA 0.55 anamorphic EUV illumination optics. The one-mirror relay system is designed by first calculating its coaxial spherical initial configuration using matrix optics; then the one off-axis relay mirror, which is tilted and decentered to eliminate ray obscuration, is fitted into a conic surface. To satisfy the requirements of multiple factors that affect illumination uniformity, the assignment relationships between the field facets and the pupil facets are determined using deep RL under a certain illumination mode. Simulation results show that the illumination efficiency is 26.24%, and the illumination uniformity can reach 99% on the mask under different illumination modes.