Multifunctional processor based on cascaded switchable polarization-multiplexed metasurface

Abstract

In many optical applications, vortex beam is widely used in optical communications, quantum information processing and other fields due to its unique properties. And in the field of imaging, it is crucial to obtain high quality and clear images. Current strategies mainly observe the interference fringes of the vortex beam and the Gaussian beam or use external optics to measure the topological charges (TCs), and apply machine learning in the back-end processing to denoise the image. The emerging diffractive deep neural network (D²NN) proposes a novel solution. Here, we introduce a multifunctional processor based on cascaded switchable polarization-multiplexed metasurface. It realizes the TC measurement and image denoising by exploiting the polarization-sensitive properties of anisotropic meta-atoms, which generate different phase responses under varying polarization states of incident light. Different types of noisy images, noise models, and noise ratios can be denoised by switching the metasurface. This study highlights the potential applications of integrating metasurfaces with D²NN through numerical simulation validation, expanding possibilities by transforming metasurfaces into multifunctional processors.