Tunable Fano resonance with high sensitivity customizable via deep learning

A resonator structure composed of a metal–insulator–metal (MIM) waveguide is proposed, incorporating an embedded inverted triangular ring and a groove, with graphene introduced as a tuning material. Device performance characterization was conducted using the finite element method. The designed resonator shifts the Fano resonant frequency up to 105 GHz by varying the Fermi energy level. In addition, employing a deep neural network architecture, by combining the properties of graphene, a two-dimensional (2D) tunable material, with the structure-optimized design, we present a method for designing transmission spectrum that can customize the resonance frequency according to the requirements. Focusing on improving the sensitivity and adaptability of the device, enabling dynamic frequency range adjustment, and maintaining a high sensitivity, the design scheme aims to provide a highly flexible and accurately tunable Fano resonator structure. The structure can be rapidly designed based on specified target requirements. Furthermore, two sets of reverse design computational results show that the mean square error of the inverse network on the test data set are 2.8×10−5 and 6.96×10−4, respectively. It highlights the robust performance of our method in realizing the reverse design of resonators on demand.