Multi-mode vector vortex beams generation enabled by a rotatable D²NN.

Area of Science:

• Optics and Photonics
• Machine Learning Applications
• Beam Generation and Manipulation

Background:

• Vector vortex beams (VVBs) are crucial for advanced optical applications due to their unique polarization and orbital angular momentum (OAM) properties.
• Generating diverse VVBs often requires complex setups or retraining of devices, limiting flexibility and efficiency.

Purpose of the Study:

• To introduce a novel rotatable diffractive deep neural network (R-D²NN) architecture for dynamic VVB generation.
• To demonstrate a single-element, training-free solution for producing multi-mode VVBs with controlled OAM and polarization.

Main Methods:

• A reconfigurable R-D²NN architecture utilizing diffractive layers whose rotation controls VVB output.
• Coherent superposition of orthogonally polarized optical fields with distinct OAM modes.
• Verification using Stokes parameter measurements and numerical simulations.

Main Results:

• Numerical simulations showed generation of up to 16-mode VVBs with >99% mode purity using five diffractive layers.
• Experimental realization achieved 85% average mode purity with a two-layer system on a spatial light modulator (SLM).

Conclusions:

• The R-D²NN offers a high-fidelity, dynamic VVB generation method.
• This training-free, single-element approach is promising for optical communications, sensing, and other advanced photonic applications.