Experimental validation for high-order vector-eigenmode decomposition with polarization characteristics

【摘要】 Vector vortex beams(VVBs) have attracted considerable attention due to their unique polarization distribution and helical phase wavefront. We first attempt to retrieve the modal coefficients of hybrid VVBs measured by their multiplex polarized intensities using the deep learning(DL)-based stochastic parallel gradient descent(SPGD) algorithm. The Xception-based DL model with multi-view images can make an accurate prediction of modal coefficients that are validated by the theoretical calculations of the waveplate angles, demonstrating a high correlation of 99.65%. The universality of the algorithm to highorder vector-eigenmodes(VMs) decomposition is proved to enable the precise reconstruction of modal patterns generated by mode-selective couplers, which promotes the accurate characteristics of VVBs in laser beam characterization and fiber mode-division multiplexing.更多还原

【Abstract】 Vector vortex beams(VVBs) have attracted considerable attention due to their unique polarization distribution and helical phase wavefront. We first attempt to retrieve the modal coefficients of hybrid VVBs measured by their multiplex polarized intensities using the deep learning(DL)-based stochastic parallel gradient descent(SPGD) algorithm. The Xception-based DL model with multi-view images can make an accurate prediction of modal coefficients that are validated by the theoretical calculations of the waveplate angles, demonstrating a high correlation of 99.65%. The universality of the algorithm to highorder vector-eigenmodes(VMs) decomposition is proved to enable the precise reconstruction of modal patterns generated by mode-selective couplers, which promotes the accurate characteristics of VVBs in laser beam characterization and fiber mode-division multiplexing.更多还原