Design of multifunctional tunable dual-layer metalens based on deep learning

To address the challenges of complex control in individually tunable meta-atoms and the limited modulation performance of single-layer holistic tunable metalenses, we propose a dual-layer tunable multifunctional metalens that utilizes holistic modulation based on the tunable BaTiO₃ (BTO) material. This design enables rapid switching between different functionalities to be achieved by varying the voltage across the top and bottom layers. To accelerate the device design process, we employ a deep learning algorithm based on the transformer structure and incorporate a band embedding method that aids in predicting the high-frequency response. This approach enables accurate and rapid prediction of the transmittance and phase of meta-atoms in different voltage states, and can generate 100,000 arbitrary geometric meta-atoms in less than 1 s. Detailed simulation results demonstrate that the designed metalens can achieve near-perfect switching between dual-functionality and quadruple-functionality within the visible light range. The switchable functionalities include focal length, focal position, beam deflection angle, beam splitting, and the transitions between these different functionalities. Furthermore, to enhance the manufacturability of the designed metalens, we replace arbitrary geometric meta-atoms with simple geometric meta-atoms to design the structure of the overall metalens, which can also achieve the dual-function switchable effect with minimal performance degradation.