CNN-based few-mode fiber modal decomposition method using digital holography.

Area of Science:

• Optics and Photonics
• Computational Physics
• Machine Learning Applications

Background:

• Modal decomposition (MD) is crucial for analyzing optical fiber modal characteristics.
• Existing MD methods face challenges in precisely determining eigenmode superposition and phase information.
• Few-mode fibers (FMFs) require advanced techniques for accurate modal analysis.

Purpose of the Study:

• To introduce a novel MD approach for few-mode fibers utilizing convolutional neural networks (CNNs).
• To accurately retrieve the superposition of eigenmodes, including amplitude and phase differences.
• To validate the reliability and feasibility of the proposed CNN-based MD method.

Main Methods:

• Development of a new MD approach based on CNNs.
• Utilizing near-field beam intensity and phase patterns obtained from digital holography.
• Numerical simulations to assess the performance and accuracy of the method.

Main Results:

• The CNN-based MD approach accurately predicts eigenmode amplitudes and phase differences.
• High similarity scores achieved: 97.0% for intensity and 85.6% for phase patterns when considering ten modes.
• Demonstrated reliability and feasibility through numerical simulations.

Conclusions:

• The proposed CNN-based MD method offers a powerful tool for analyzing few-mode fibers.
• It enables precise prediction of eigenmode superposition, crucial for advanced optical communication systems.
• This technique advances the characterization of complex light propagation in optical fibers.