Deep‐Learning‐Assisted Design of Dynamically Tunable Near‐Infrared Dual‐Band Lithium Niobate Metamaterial Absorber for Optical Communication

Abstract

ABSTRACT To overcome the fixed operating wavelengths and limited tunability of conventional metamaterial absorbers, this study proposes a dynamically tunable near‐infrared dual‐band lithium niobate (LN) metamaterial absorber platform at optical communication wavelengths. Using an “end‐to‐end + gradient optimization” deep learning inverse‐design framework, we efficiently and accurately derive the structural parameters matching target absorption spectra. The designed LN metamaterial absorber (LN‐MMA) exhibits dual‐band near‐perfect absorption peaks at 1310 and 1550 nm, achieving peak efficiencies of 99.9% and 99.5%, respectively. Owing to the strong electro‐optic effect of LN, the resonance wavelengths can be dynamically tuned via an external electric field. Furthermore, by integrating the LN‐MMA with coherent technology, a dynamically tunable coherent perfect absorber (LN‐CPA) is realized, exhibiting absorption peaks at 1550 and 1704 nm with efficiencies of 98.5% and 98.1%. In this configuration, the absorption response can be actively tuned by the phase difference between two incident coherent beams, enabling all‐optical modulation. These results demonstrate a compact absorber with both electrical and optical tunability, showing strong potential for inter‐satellite laser communication, optoelectronic signal conversion, optical detection, and all‐optical signal processing.