Haobin Fu

Turbulence is a complex and chaotic fluid motion state. Atmospheric turbulence presents significant challenges for applications such as remote sensing,astronomical observations, and free-space quantum key distribution (QKD), due to its rapid evolution across temporal and spatial scales. Traditional methods for correcting atmospheric turbulence encounter difficulties, particularly under strong daytime turbulence conditions. In this study, we develop a deep learning-based adaptive method to correct strong atmospheric turbulence in field conditions, facilitating the turbulence correction over 1.4 km and 7 km free-space channels. Experimental results present better correction performance compared to wavefront sensor-based methods, yielding a 2–4 dB Strehl ratio improvement. Additionally, our approach directly estimates phase information from a defocused camera, significantly reducing the implementation cost of adaptive systems. Furthermore, we evaluate the performance of a daytime free-space QKD system incorporating our deep learning–based method, leading to higher key rates and longer propagation distances. Our method provides a practical and efficient solution for daytime QKD applications.

Satellite-based quantum key distribution (QKD) is crucial for establishing global quantum networks, while current implementations are restricted to low-Earth-orbit satellites and nighttime operations. Geosynchronous-Earth-orbit (GEO) satellites present a compelling alternative, offering continuous availability and wide-area coverage. The primary obstacles are the substantial link loss inherent to GEO distances and high daytime ambient noise, which severely degrade the signal-to-noise ratio and hinder secure key generation. To address the challenge of daytime ambient noise, we implemented a multi-faceted filtering strategy across spatial, spectral, and temporal domains. Notably, an additional background reduction of approximately 5.2 dB was achieved using the 854.45 nm Fraunhofer line for spectral noise suppression. Simultaneously, we employed deep-learning-based adaptive optics, improving single-mode fiber coupling efficiency by approximately 1.4∼5.2 dB. We then experimentally validated all-day QKD over a metropolitan 7-km free-space channel, demonstrating continuous QKD operation throughout the day with maximum tolerable channel loss exceeding 62 dB. This represents an improvement of over one order of magnitude and surpasses the projected daytime loss budget for GEO satellite links. These results signify a major advancement towards practical, globally accessible quantum communication via GEO satellites.