Classification of single photons in higher-order spatial modes via convolutional neural networks

Area of Science:

• Optical communication
• Quantum information science
• Machine learning applications

Background:

• Spatial modes offer high information capacity for optical communication.
• Atmospheric turbulence distorts wavefronts, necessitating compensation for effective information transfer.
• Convolutional neural networks (CNNs) excel at image denoising and classification tasks.

Purpose of the Study:

• To improve information transfer in optical communication systems using spatial modes.
• To develop an AI-based method for correcting atmospheric turbulence effects on spatial modes.
• To classify different types of spatial modes (HG, LG, IG) with high accuracy.

Main Methods:

• Experimentally generated Hermite-Gaussian (HG), Laguerre-Gaussian (LG), and Ince-Gaussian (IG) modes using single photons.
• Applied a denoising autoencoder to correct wavefront distortions caused by turbulence.
• Utilized a convolutional neural network for classifying the mode orders based on intensity images.

Main Results:

• Achieved 99.2% classification accuracy across all spatial modes (HG, LG, IG).
• Hermite-Gaussian modes demonstrated the highest individual mode classification accuracy.
• The CNN model effectively corrected for turbulence-induced wavefront aberrations.

Conclusions:

• Convolutional neural networks provide an efficient and cost-effective solution for spatial mode compensation in optical communication.
• The proposed AI method is suitable for single-photon communication systems operating in turbulent environments.
• Intensity-based CNNs offer a robust approach for enhancing information transfer via spatial modes.