Deep learning-enabled atmospheric turbulence compensation for concentric perfect optical vortex beams

The concentric perfect optical vortex (CPOV) beam, composed of multiple perfect optical vortex beams, has attracted considerable interest within the domain of optical communications. Nevertheless, the propagation of the CPOV beam is susceptible to distortion induced by atmospheric turbulence (AT), which adversely affects its performance in free-space optical (FSO) communication systems. To date, effectively mitigating the impact of AT on CPOV beams remains a significant challenge. In this study, we propose an atmospheric turbulence decomposition frequency prediction network (ATDFPNet) to address this issue. Trained on an extensive dataset, the proposed model demonstrates rapid and precise prediction of turbulence phase screens and exhibits robust generalization capabilities across varying turbulence strengths and orbital angular momentum (OAM) modes. Experimental results show that, following AT compensation, the mode purity of the CPOV outer ring markedly improves from 8.00% to 38.52% under strong turbulence conditions. Furthermore, integrating ResNet18 with a Dammann vortex grating for OAM mode identification enhances accuracy from 47.77% to 98.01%. Additionally, an OAM shift keying communication link is established, achieving a Structural Similarity index of 0.98 under strong turbulence conditions after compensation. These results demonstrate that the ATDFPNet model effectively mitigates distortions in the CPOV beam, underscoring its considerable potential for deployment in OAM-based FSO communication systems.