On-demand orbital angular momentum modes through hollow-core multimode fiber via a neural network model

Area of Science:

• Photonics
• Optical Communications
• Machine Learning

Background:

• Orbital Angular Momentum (OAM) beams offer unique properties for optical communications.
• Controlling OAM modes in multimode fibers is challenging.
• Inhibited-coupling hollow-core photonic crystal fibers (IC-HCPCFs) support a large number of modes.

Purpose of the Study:

• To present a machine-learning approach for generating and reconfiguring OAM beams.
• To demonstrate the capability of IC-HCPCFs for structured light transport.
• To enable on-demand beam shaping using artificial intelligence.

Main Methods:

• Training a neural network-based digital twin for a 139 µm core IC-HCPCF.
• Guiding over 60 LP-like modes per polarization in the green spectral range.
• Validating the neural network's accuracy through experimental intensity pattern comparison.

Main Results:

• High fidelity between experimental and predicted output intensity patterns was achieved.
• Median Pearson correlation coefficients exceeded 98%, confirming model accuracy.
• Demonstrated dynamic reconfiguration of OAM beams within the IC-HCPCF.

Conclusions:

• Multimode IC-HCPCFs are a versatile platform for structured light.
• Machine learning provides an effective method for OAM beam control.
• This approach enables on-demand, dynamic beam shaping capabilities.