Computational i-line proximity lithography approaching sub-micrometer resolution

Proximity lithography provides a cost-effective and full-field approach for semiconductor manufacturing. However, its imaging resolution is limited by diffraction effects. This paper presents a genetic algorithm-based mask optimization method to overcome the resolution limitation in proximity lithography. Unlike traditional mask corrections focusing on line-shortening/corner-rounding corrections, our approach achieves a breakthrough in resolution by optimizing a pixel-based binary mask, which effectively encodes the finite spatial discretization of the transmitted light amplitude. The proximity lithography process is modeled using the Rayleigh–Sommerfeld diffraction theory. Numerical simulations and experimental results verify the effectiveness of the proposed methods. Key advancements are summarized as follows. (1) The algorithm is effective for different proximity distances, different critical dimensions, and different patterns. The pattern fidelity—quantified by the mean squared error (MSE)—is improved by an average of 69.8% compared to unoptimized masks. (2) Two strategies, namely edge cropping and regularization penalty, are implemented to enhance the manufacturability and processing feasibility. (3) The critical dimension of the micrometer optical device patterns explored is ≤1 µm, and the smallest line width achieved is 0.2 µm, which exceeds the resolution limits of the commercial machine (>3 µm). The proposed approach leverages the large exposure field advantage of proximity lithography and introduces optimization exclusively at the mask design stage, without modifying downstream fabrication processes, thereby offering a cost-effective solution.