Machine Learning-Assisted Light Management and Electromagnetic Field Modulation of Large-Area Plasmonic Coaxial Cylindrical Pillar/Ring Nanoarray Patterns

Hexagonal coaxial cylindrical gold pillar/ring nanoarray patterns can be fabricated with an anodic aluminum oxide (AAO) template or nanosphere lithography. It is time-consuming and expensive for experimental work solely to tune and optimize geometrical parameters for achieving desirable optical properties. Herein, finite-difference time-domain (FDTD) simulation has been performed to investigate how the key geometrical parameters govern optical properties such as plasmonic resonance band, local electric field enhancement, and quality factor (Q-factor). FDTD simulation results reveal that these three important optical properties can be modulated by coupling localized surface plasmon resonance (LSPR) and charge distributions on the metal–dielectric interface to suppress its radiative damping, concentrate the electric field, and tune a spectral resonance. The impact of specific geometric parameters on optical properties was further analyzed via machine learning for visualization. For the gold pillar/ring nanoarrays, the local electric field enhancement can occur at the gap between two adjacent nanostructures or at the gap between pillar and ring. Adjusting the height and gap width proves to be the most effective to optimize both the Q-factor and electric field enhancement. These machine learning-assisted studies will provide a theoretical framework for tailoring the geometrical parameters of the coaxial cylindrical pillar/ring nanoarray patterns toward desirable optical properties.