Multiobjective Statistical Learning Optimization of RGB Metalens

Modelling of multi-wavelength metasurfaces relies on adjusting the phase indi-vidual nanoresonators at several wavelengths.The traditional procedure neglects thenear-field coupling between nanoresonators, which dramatically reduces over-all diffraction efficiency, bandwidth, numerical aperture and device diameter.Anotheralternative design strategy is to combine a optimization technique with full-wave simulations mitigate this problem optimize entire metasurface once.Here, we present global multiobjective that utilizes statisticallearning method RGB spherical metalenses visible wavelengths. Theoptimization procedure, coupled high-order solver, accounts for nearfield resonators. High lenses(NA= 0.47and NA= 0.56) 8μm 10μm diameters are optimized average1 focusing efficiencies 55% 45%, an average focusing error smaller than 6%for colors. The fabricated experimentally characterized devices present44.16% 31.5% respective efficiencies. reported performances represent thehighest highNA >0.5 obtained so far. Theintegration in portable wearable electronic de-vices requires high offer variety applications ranging from classicalimaging virtual augmented reality.