Deeply Subwavelength Topological Microscopy

We demonstrate experimentally label-free far-field imaging of subwavelength objects at resolution greater than λ/20, exceeding the diffraction limit by an order of magnitude. Our imaging approach, termed Deeply Subwavelength Topological Microscopy (DSTM), is based on the combination of illumination with topological structured illumination and artificial intelligence. In DSTM, the imaging target is placed under topological illumination and multiple far-field scattering patterns are recorded for different positions of the imaging target within the illumination light field (see Fig. 1 ). The diffraction patterns are analyzed by a neural network trained on a large number of scattering events, which allows to reconstruct the object. DSTM promises unprecedented resolution approaching atomic length scales.