Inverse design of tunable high-sensitivity Fano resonators with black phosphorus utilizing deep learning

Abstract A deep learning-based inverse design approach is proposed to design tunable high-sensitivity Fano resonators efficiently. The approach parameterizes a planar micro-hole slab (PMS) covered with monolayer black phosphorus (BP) via a topologic strategy, then builds and trains a linear neural network (LNN) and customizes a differential evolution (DE) algorithm. By combining LNN and DE algorithm, the approach enables a fast and high-fidelity inverse design of the proposed Fano resonators. This work provides a comprehensive and efficient inverse design pipeline. Results show that the proposed approach can export the corresponding design from a target spectrum in less than a minute with a mean square error (MSE) value lower than 0.001. The designed tunable Fano resonator presents a maximum frequency shift of 100 GHz, by adjusting BP’s doping concentration within the range of 0.01 × 10 13 cm −2 to 1 × 10 13 cm −2 . Furthermore, a figure of merit (FOM) of sensitivity as high as 9,000 underscores the resonator’s exceptional sensing performance.