Localized Plasmonic Structured Illumination Microscopy\nUsing Hybrid Inverse Design

Super-resolution\nfluorescence imaging has offered unprecedented\ninsights and revolutionized our understanding of biology. In particular,\nlocalized plasmonic structured illumination microscopy (LPSIM) achieves\nvideo-rate super-resolution imaging with ∼50 nm spatial resolution\nby leveraging subdiffraction-limited nearfield patterns generated\nby plasmonic nanoantenna arrays. However, the conventional trial-and-error\ndesign process for LPSIM arrays is time-consuming and computationally\nintensive, limiting the exploration of optimal designs. Here, we propose\na hybrid inverse design framework combining deep learning and genetic\nalgorithms to refine LPSIM arrays. A population of designs is evaluated\nusing a trained convolutional neural network, and a multiobjective\noptimization method optimizes them through iteration and evolution.\nSimulations demonstrate that the optimized LPSIM substrate surpasses\ntraditional substrates, exhibiting higher reconstruction accuracy,\nrobustness against noise, and increased tolerance for fewer measurements.\nThis framework not only proves the efficacy of inverse design for\ntailoring LPSIM substrates but also opens avenues for exploring new\nplasmonic nanostructures in imaging applications.