Adjusted EfficientNet for the diagnostic of orbital angular momentum spectrum

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Information Optics

Background:

• Orbital angular momentum (OAM) is a fundamental property of light beams, offering multiple dimensions for information encoding.
• The OAM spectrum, representing the intensity distribution of different OAM components, is crucial for understanding beam characteristics and enabling advanced applications.
• Accurate and efficient measurement of the OAM spectrum is essential for harnessing the full potential of OAM-based technologies.

Purpose of the Study:

• To develop a novel, high-precision, and high-speed method for measuring the orbital angular momentum (OAM) spectrum of light beams.
• To demonstrate the effectiveness of combining deep neural networks with phase-only diffraction optical elements for OAM spectrum analysis.
• To explore the robustness and applicability of the proposed method across various OAM channel configurations and experimental conditions.

Main Methods:

• Implementation of a phase-only diffraction optical element designed to spatially separate different OAM components of incident light beams into distinct diffraction patterns.
• Utilizing the EfficientNet deep neural network architecture, specifically adapted to analyze the unique diffraction patterns generated by the optical element.
• Training and employing the neural network to accurately calculate the OAM spectrum based on the analyzed diffraction patterns.

Main Results:

• Successful reconstruction of OAM spectra with high precision and remarkable speed, validated through experimental results.
• Demonstrated robustness of the method against common experimental challenges, including Gaussian noise and random zooming of the beam.
• Effective performance across systems with varying numbers of OAM channels, showcasing versatility.

Conclusions:

• The proposed method represents a significant advancement in OAM spectrum recognition, offering a new capability for optical metrology.
• This technique holds substantial promise for diverse advanced applications, including high-capacity optical communications, quantum information processing, and precision sensing.
• The integration of deep learning with optical elements provides a powerful framework for complex optical beam analysis.