Solving Unsolved problems in IR and THz Near-field Nanoscopy with Rapid Simulation and Hybrid Machine Learning

In this work, we propose time-efficient schemes in the simulation, calculation, and understanding of tip-sample interaction in the near field, especially in the field of scattering-type scanning near-field microscope. Using rapid simulation methods, we can solve scientific problems with extended tip geometry, complex sample topography, and uniaxial or biaxial sample anisotropies, recreating tip dependent near-field spectral and images with high spatial resolution. Using machine learning, we explore the most accurate means to extract dielectric function from near-field spectra, and vice versa. Our methods shed light on the understanding of the contrast mechanism in s-SNOM imaging and spectroscopy, while also representing a valuable platform for future quantitative analysis of the experimental observations.