Figure | Working principle of the proposal. (IMAGE)
Caption
(a) The work flow of intelligent hybrid strategy for generating the modulated DOE phase profiles. The desired phase-only profile P consists of three components: Γ1a,b, which pertains to the manipulation of wave vector and initial phase; Γ2a,b , derived from an intelligent deep neural network designed for tailoring superimposed OAM mode and corresponding amplitude; and the overall geometric phase φL3 and φR3 for SAM tailoring. (b) The architecture of the deep neural network for superposed OAM and amplitude tailoring. By training on the complex amplitude information through the strategies of feature extraction and a dynamic feature fusion, a mapping between the target superposed OAM beam and phase-only transmittance function can be established. The trained model allows direct hologram output of target OAM mode corresponding with amplitude distribution, forming different spatial mode. (c) A phase distribution instance of the finally obtained Γ, which is then encoded onto a metasurface to accomplish six-DoFs, 288-dimensional beam tailoring, where the center zooms for better view. (d) Principle of overall geometric phase modulation. By introducing left and right circular components and toward the polarization angle and of the generated beam, one can achieve the modulation of SAM.
Credit
Shiyun Zhou, Lang Li et al.
Usage Restrictions
Credit must be given to the creator.
License
CC BY