Relative wavefront error correction over a 2.4-km free-space optical link via machine learning.

Area of Science:

• Optical communication
• Atmospheric physics
• Machine learning

Background:

• Coherent optical communication systems often use reference beacons multiplexed with signals.
• It is commonly assumed that beacons and signals experience equivalent wavefront distortion in turbulent atmospheric channels.
• This study investigates the discrepancy in wavefront distortion between polarization-multiplexed beacons and signals.

Purpose of the Study:

• To experimentally demonstrate and quantify relative wavefront errors (WFEs) between reference beacons and signals after atmospheric transmission.
• To develop and apply machine learning (ML)-based algorithms for wavefront correction to mitigate observed WFEs.
• To analyze the impact of relative WFEs on continuous-variable quantum key distribution (CV-QKD) and assess the potential for improved secure key rates.

Main Methods:

• Experimental setup involving a 2.4-km atmospheric link.
• Polarization multiplexing of reference beacons and information-encoded signals.
• Development and implementation of ML-based phase retrieval algorithms for wavefront correction.
• Analysis of relative phase error variance and excess noise contributions.

Main Results:

• Experimental evidence of significant relative wavefront errors (WFEs) between polarization-multiplexed beacons and signals.
• ML-based wavefront correction algorithms achieved up to a 2/3 reduction in relative phase error variance.
• Quantification of excess noise contributions stemming from relative WFEs in CV-QKD.

Conclusions:

• The assumption of equivalent wavefront distortion between beacons and signals in atmospheric optical links is challenged.
• ML-based wavefront correction is effective in mitigating relative WFEs, improving signal fidelity.
• Implementing these correction algorithms in CV-QKD could lead to substantial increases in secure key rates.