Compensating the distorted OAM beams with near zero time delay

Vortex beams, carrying orbital angular momentum (OAM), have great potential to increase the information capacity of optical communication systems for the orthogonality and infinite mode number. For OAM beams propagating in free space, however, the atmospheric turbulence may cause mode distortions and hinder their utilization in practice. In this work, we propose a kind of diffractive deep neural network (D2NN) to compensate the distorted OAM beams. Different from those D2NNs reported before, the network reported here is dissipative, rather than unitary. In our system, the common features of various wavefront distortions are extracted and compensated, while the random distortions are filtered out by a diaphragm, which is achieved by constructing an improved loss function. The results show that multiple OAM beams with different degrees of distortions can be compensated simultaneously, and good agreement between simulations and experiments is obtained. The D2NN based OAM beam compensating reported here will greatly improve the robustness and efficiency of free space optical communication.