• Advanced Photonics Nexus
  • Vol. 2, Issue 4, 046008 (2023)
Ling Hong, Haoxu Guo, Xiaodong Qiu, Fei Lin, Wuhong Zhang*, and Lixiang Chen*
Author Affiliations
  • Xiamen University, Department of Physics, Xiamen, China
  • show less
    DOI: 10.1117/1.APN.2.4.046008 Cite this Article Set citation alerts
    Ling Hong, Haoxu Guo, Xiaodong Qiu, Fei Lin, Wuhong Zhang, Lixiang Chen. Experimental optical computing of complex vector convolution with twisted light[J]. Advanced Photonics Nexus, 2023, 2(4): 046008 Copy Citation Text show less
    The schematic diagram of the optical complex vector convolution C=A*B, based on five-dimensional orbital-angular-momentum state vector.
    Fig. 1. The schematic diagram of the optical complex vector convolution C=A*B, based on five-dimensional orbital-angular-momentum state vector.
    Schematic of experimental layout for optical computing of complex vector convolution based on OAM eigenstates. HWP, half wave plate; SLM, spatial light modulator; L1, L2, L3, L4, L5, lenses; Mask, a phase triangular object; CCD, camera. Inset, holographic example of encoded complex vector with four OAM states in SLM1 and SLM2.
    Fig. 2. Schematic of experimental layout for optical computing of complex vector convolution based on OAM eigenstates. HWP, half wave plate; SLM, spatial light modulator; L1, L2, L3, L4, L5, lenses; Mask, a phase triangular object; CCD, camera. Inset, holographic example of encoded complex vector with four OAM states in SLM1 and SLM2.
    Architecture of the residual neural network.
    Fig. 3. Architecture of the residual neural network.
    Complex vector convolution process and experimental results of 7-dimensional OAM state vectors. The input vectors (a) A and (b) B convolved to get the output (c) vector C, where the vector elements are an=al exp(iφla), bn=bl exp(iφlb), and cn=cl exp(iφlc), respectively.
    Fig. 4. Complex vector convolution process and experimental results of 7-dimensional OAM state vectors. The input vectors (a) A and (b) B convolved to get the output (c) vector C, where the vector elements are an=alexp(iφla), bn=blexp(iφlb), and cn=clexp(iφlc), respectively.
    The distributions of (a) proximity and (b) relative errors obtained by comparing each experimental predicted output vector with the theoretical output vector for 7-dimensional OAM state vector convolution.
    Fig. 5. The distributions of (a) proximity and (b) relative errors obtained by comparing each experimental predicted output vector with the theoretical output vector for 7-dimensional OAM state vector convolution.
    Dimension of vector C(d)Average proximity (Savg)Mean relative error (Errmean)
    130.9778±0.01880.0427±0.0341
    170.9633±0.01920.0498±0.0403
    210.9463±0.02260.0571±0.0477
    Table 1. Average proximity and mean relative error of the output vector from 7-, 9-, and 11-dimensional OAM state vector convolution.
    Ling Hong, Haoxu Guo, Xiaodong Qiu, Fei Lin, Wuhong Zhang, Lixiang Chen. Experimental optical computing of complex vector convolution with twisted light[J]. Advanced Photonics Nexus, 2023, 2(4): 046008
    Download Citation