Optical proximity correction of hot-spot patterns with subwavelength size in DMD maskless projection lithography

Abstract

When the critical dimension (CD) of resist patterns nears the resolution limit of the digital micromirror device (DMD) maskless projection lithography (DMD-MPL), significant distortion can emerge in the silicon wafer due to the optical proximity effect (OPE). The significant distortion (breakpoints, line-end scaling, corner rounding, etc.) between resist patterns and target patterns results in reduced lithographic quality. To address this issue, we have proposed a pixel-based optical proximity correction (PB-OPC) method used for the hot-spot patterns with subwavelength sizes specifically designed for DMD-MPL. Employing an end-to-end learning neural network, the PB-OPC algorithm is both straightforward and efficient. A well-trained U-net framework facilitates the mapping from unoptimized masks to optimized masks. Experimental exposure trials have demonstrated that this method not only corrects OPC in general patterns but also effectively rectifies hot-spot patterns. The pattern error (PE) value can be reduced by about 30% in the design layouts. We believe this approach holds the potential to enhance the resolution and fidelity of resist patterns in DMD maskless lithography.