Speckle-based deep learning approach for classification of orbital angular momentum modes.

Area of Science:

• Optics and Photonics
• Machine Learning
• Quantum Information

Background:

• Orbital angular momentum (OAM) modes are crucial for advanced optical communication systems.
• Accurate and efficient OAM mode classification is essential for practical applications.
• Traditional methods for OAM mode classification can be complex and sensitive to environmental factors.

Purpose of the Study:

• To develop a robust and efficient deep learning-based approach for OAM mode classification.
• To investigate the performance of the proposed method under simulated atmospheric turbulence.
• To validate the method using experimental speckle data.

Main Methods:

• Simulated speckle fields of Laguerre-Gauss (LG), Hermite-Gauss (HG), and superposition modes were generated.
• Intensity images of speckle fields were used to train a convolutional neural network (CNN).
• The CNN model was trained and tested with simulated and experimental data, including perturbed modes.

Main Results:

• The CNN model achieved >99% accuracy for classifying simulated OAM modes.
• Classification accuracy remained high (>98%) even when models were trained with simulated atmospheric turbulence.
• A maximum accuracy of 96% was achieved with experimental speckle images, demonstrating robustness.

Conclusions:

• Speckle-based deep learning offers a highly accurate and efficient method for OAM mode classification.
• The technique is resilient to atmospheric turbulence and can utilize partial speckle field information.
• This approach simplifies OAM mode analysis, reducing the need for complex modal field capturing.