Orbital angular momentum-mediated optoelectronic neural network for turbulence diagnostics
Ke Cheng, Qinghan Zhang, Jianxin Lin, Xiaonan Hu, Hang Su, Baoli Li, Min Gu, Xinyuan Fang
Abstract
Atmospheric turbulence (AT) severely degrades free-space communications, imaging, and sensing systems, driving critical demand for diagnostics of turbulence strength ( C n 2 ). However, existing approaches face limited adaptability, high latency, and excessive power consumption for deployment. Here, we propose an orbital angular momentum (OAM)-mediated optoelectronic neural network (OOENN) that integrates a diffractive optical module for OAM spectrum feature extraction with a shallow electronic module for turbulence diagnostics, leveraging OAM spectrum data transformation. The optical module extracts turbulence-encoded features from distorted Laguerre–Gaussian (LG) beams and decomposes its output field into OAM spectrum data. These data are then fed into an electronic module that diagnoses turbulence strength using a minimal fully connected network with 9 input neurons and nonlinear activation. The OOENN performs feature extraction at light speed while enabling ultra-efficient electronic processing, thereby alleviating the latency and power constraints. Experimental results demonstrate diagnostics of five turbulence strengths within C n 2 =10 −16 to 10 −12 m −2/3 , achieving 82.4% accuracy at 80 ms latency per diagnosis. This fusion of structured light fields with optoelectronic intelligence establishes a technological foundation for next-generation adaptive systems in turbulence-resilient optical communications, remote sensing, and quantum information transfer.