Conformal Antireflective Multilayers for High‐Numerical‐Aperture Deep‐Ultraviolet Lenses

Abstract Precise surface reflectance control at specific deep‐ultraviolet (DUV) wavelengths across wide angles is crucial for semiconductor inspection and lithography tools. The inherent challenges in designing DUV antireflective multilayers stem from limited transparent materials and the resultant fabrication complexity owing to numerous interfaces. Here, wide‐angle antireflective multilayers finely tuned to 248 nm designed using an active learning scheme is presented. The active learning scheme employing factorization machines (FM) identifies the optimal configurations for binary‐material‐based multilayers (AlF 3 /LaF 3 , AlF 3 /MgF 2 , and AlF 3 /Al 2 O 3 ) with varying index contrasts, achieving minimal figure‐of‐merit (i.e., average angular reflectance) values at predetermined total thicknesses. High‐index‐contrast AlF 3 /Al 2 O 3 multilayers are fabricated via atomic layer deposition, thus enabling the conformal coating of high‐numerical‐aperture (NA) lenses with atomic precision. An optimized AlF 3 /Al 2 O 3 tri‐layer with a total thickness of 180 nm results in an average (0°–45°) reflectance of 0.4% on a CaF 2 planar substrate and 0.6% on a CaF 2 convex lens (NA = 0.47), similar to the performance of an ideal single‐layer coating requiring a practically unavailable refractive index. Phasor analysis, which considers only first‐order reflections between adjacent layers, supports the benefits of high‐index‐contrast binary materials and the use of the FM‐based active learning scheme in antireflective multilayer design.