Low-Loss Parowax-Imprinted Diffractive Neural Network for Orbital Angular Momentum Terahertz Holographic Imaging

Abstract

The helical phase front of orbital angular momentum (OAM) waves offer additional multiplexing degree-of-freedom to increase the capacity of communication systems in the terahertz domain, which in turn can significantly benefit forthcoming high-speed wireless sixth-generation communication networks. This work introduces a diffractive neural network approach for recognizing the topological charge of OAM waves and their superposition. Moreover, it is shown that the diffractive network can further enable mathematical operations through the topological charges (TCs) of the superposed OAM waves. The diffractive neural networks (DNN) are fabricated through an imprinting technique with low-loss parowax material. To validate the feasibility of this general approach, experimental demonstrations are conducted, which show that the low-loss parowax DNN effectively detects the TCs of the OAM waves and display them in a numerical format.