Design of Free-Space Dual-Polarization Receiving Antenna with Deep Learning for Coherent Photonic-Integrated LiDAR

Abstract

The dual-polarization receiver antenna (DPRA) is one of the most critical components in all-solid-state coherent detection LiDAR systems. Although various photonic receiver antennas have attempted to achieve coupling for differently polarized light, significant challenges regarding free-space optical coupling efficiency, compact footprint, and design complexity persist. This work proposes a high-efficiency compact DPRA by integrating a metalens module with a grating coupler module. The metalens module employs multishaped meta-atoms, effectively enhancing focusing efficiency for dual polarization states. The optimized grating coupler, designed through a hybrid deep learning network combined with a particle swarm optimization (PSO) algorithm, efficiently couples the focused beams from the metalens into two distinct waveguides. This approach substantially accelerates the grating coupler design process, achieving nearly 1000-fold speed improvement compared to conventional electromagnetic simulation methods. Experimental results validate the effectiveness of this system, showing final coupling efficiencies of −5.31 and −5.52 dB for 0 and 90° polarized light. This integrated architecture, compatible with multilayer lithography for further efficiency enhancement, provides an innovative solution for polarization-diverse light reception in LiDAR and optical communication systems.