Comparison of Gaussian and vortex probe beams for deep-learning-based turbulence correction

Abstract

Optical transmission through atmospheric turbulence is a critical issue for establishing optical communication based on photonic spatial degrees of freedom. Refractive index fluctuations by a turbulent atmosphere cause structural distortion on the transmitted laser beams, resulting in crosstalk between adjacent spatial modes. We study a deep-learning technique to compensate for the turbulence effect. In particular, we mainly focus on improving the correction performance according to the selection of a probe beam. Different modes of Laguerre-Gaussian beams, including the lowest order (Gaussian), are employed as optical probes, and their abilities to compensate for turbulence effects are compared. Our results show that the Gaussian probe beam achieves a higher average correlation coefficient and a lower error rate, i.e., better correction performance, despite the spot size smaller than higher-order modes. Finally, to demonstrate the origin of the performance difference, we examine residual phases by spatial position and discuss the correction coverages.