AI-Driven Optics Enable 1.4-km Quantum Link Under Strong Turbulence

Researchers have demonstrated a deep-learning approach to adaptive optics, successfully establishing 1.4-km and 7-km free-space quantum links despite strong atmospheric turbulence. The work addresses a critical limitation of current quantum key distribution (QKD) systems, which are largely restricted to nighttime operation due to disruptive background noise. This method directly estimates phase information from images, offering an efficient solution for turbulence correction and enabling practical daytime QKD. The team reports this approach achieves substantial improvements in the Strehl ratio under turbulent conditions, enabling higher key rates and longer distances.

Deep-Learning Adaptive Optics Corrects Strong Turbulence for Free-Space QKD

Free-space quantum key distribution (QKD) may extend beyond nighttime operation thanks to a new approach to correcting atmospheric turbulence. Researchers have developed a deep-learning-based adaptive optics system that improves signal clarity in challenging conditions, a critical step toward practical, all-day quantum communication networks. Current QKD systems are limited by daytime background noise, which overwhelms the delicate quantum signals; however, this new method offers a potential solution by maintaining high coupling efficiency despite atmospheric distortions.

The team experimentally validated the system across both 1.4-kilometer and 7-kilometer free-space channels, showcasing its functionality at varying distances. Numerical evaluations reveal the deep-learning approach achieves higher key rates and extends the potential propagation distances for QKD systems. This advancement addresses a long-standing obstacle to widespread QKD deployment.

While single-mode fiber coupling can suppress background noise, it struggles to maintain stability when faced with significant atmospheric turbulence. By bypassing the need for complex wave-front sensing, the deep-learning method offers a streamlined solution. The work, accepted July 13, 2026, and published July 30, 2026, represents a significant leap toward realizing truly global quantum communication networks.