1. INTRODUCTION

Quantum key distribution (QKD) [1] provides a promising solution for constructing a global-scale quantum network and has made significant progress in the past few decades [2–8]. Based on a low-Earth-orbit (LEO) satellite launched in 2016, satellite-to-ground QKD [9], satellite-relayed intercontinental quantum network [10], and an integrated space-to-ground quantum network over 4600 km [11] have been implemented. For a single LEO satellite, the available communication time is limited to approximately 5–10 min per day, and the average delay for key distillation based on the classic microwave communication is about several days [9]. Such performance is inefficient for worldwide connectivity and practical applications. Recently, an international space race toward the satellite-constellation-based quantum network has arisen [12], and several proposals of CubeSats and quantum constellations have been reported [13–18]. The use of high-orbit satellites can achieve greater coverage and communication time than LEO satellites [19], which also constitutes an optimal complementarity. However, despite significant advancements have been made, the existing experimental implementations of satellite-based QKD [9,20] are restricted to nighttime operation and key distillation with a delay of several days, which are far from enough to establish a practical and efficient global-scale quantum network.

 

Fig. 1. (a) All-day real-time QKD experiment in the 20-km free-space channel; (b) the QKD receiver; (c) the QKD transmitter; (d) the high-speed robust QKD light source. The transmitter and receiver first establish the terrestrial optical link using acquiring-pointing-tracking systems (see Supplement 1) based on the 815-nm downlink beacon light and the 671-nm uplink beacon laser. Then, the 1550-nm quantum photons are prepared by the high-speed robust QKD light source, output from the transmitter, and collected and detected at the receiver. Finally, real-time key distillation can be achieved by utilizing the 815-nm downlink communication laser and 1538-nm uplink communication laser. In the transmitter, DM1 reflects 1538 nm and transmits at 1550 nm, DM2 reflects 650–850 nm and transmits at 1538–1550 nm, and DM3 reflects 671 nm and transmission at 815 nm. In the receiver, the auxiliary telescope (bottom right of panel b) is used for transmitting 1538 nm. LD, laser diode; BS, (fiber) beam splitter; PM, phase modulator; CPM, customized polarization module; FP, Fabry–Perot filter; ATT, attenuator; SMF, single-mode fiber; PMF, polarization-maintaining fiber; MMF, multimode fiber. M, mirror; CAM, camera; FSM, fast steering mirror; DM, dichromatic mirror; Q, quarter-wave plate; H, half-wave plate. DWDM, 100-GHz dense wavelength division multiplexing for coarse filtering; PC, fiber polarization compensator; PBS, fiber polarization beam splitter; SNSPD, superconducting-nanowire single-photon detector.

Download Full Size | PDF

There have been several pioneering experiments and great effort has been taken [21–29], leading the way toward satellite-based daylight quantum communication. For example, Liao et al. [25] demonstrated free-space QKD over a 53-km high-loss link from 15:30 to 17:00 local time over several sunny days, and Avesani et al. [28] demonstrated successful QKD over a 145-m-long free-space link from 11:00 to 20:00 during a whole sunny day. Through these experiments, indispensable technical advances toward all-day QKD with quantum satellites have accumulated. However, these experiments can only work under certain conditions, for instance, channel losses of the magnitude of ${\sim}20\;{\rm dB}$ [21–24,27,28], or time restrictions on generating final secure keys [25,26]. Moreover, the key extraction processes in these experiments is generally processed using Ethernet or working offline, which is not suitable for future satellite-based practical applications, especially the requirement of real-time key distillation. By utilizing laser communication for QKD post-processing, real-time key distillation can be achieved, which has been verified using an emulated experiment in the laboratory [30]. Based on previous technical advances [21–28], the next crucial step is to break through these bottlenecks and complete experimental verifications toward all-day real-time QKD with quantum satellites.

Here, to address every aspect of the above-mentioned crucial issues, we directly and comprehensively verified all-day real-time QKD over a 20-km terrestrial free-space link, which well exceeds the effective atmospheric thickness. For this purpose, we developed a satellite payload prototype serving as the QKD transmitting terminal, located at the mountain top of Silk Road Resort (Alice, ${{43}}^\circ {{25^\prime 40^{\prime \prime} {\rm N}}}$, ${{87}}^\circ {{24^\prime 44^{\prime \prime} {\rm E}}}$, altitude of 2266 m), and the Nanshan station with modified receiving optics (Bob, ${{43}}^\circ {{28^\prime 31^{\prime \prime} {\rm N}}}$, ${{87}}^\circ {{1^\prime 35^{\prime \prime} {\rm E}}}$, altitude of 2070 m) served as the QKD receiving terminal. Figure 1 shows the configuration of our experiment performed in Xinjiang Province, China.

2. EXPERIMENTAL CHALLENGES AND SOLUTIONS

All-day real-time quantum communication using satellites must meet several critical requirements: (i) daytime noise suppression that approaches the theoretical limit; (ii) a high-speed, inherently robust QKD optical source; and (iii) laser-communication-based QKD post-processing that enables real-time key distillation. The attenuation of satellite-to-ground quantum link is exceptionally high, mainly due to the limited divergence angle of the beam carrying the quantum signals and the finite aperture of the receiving telescope. To improve the signal-to-noise ratio (SNR), the background noise should be suppressed as far as possible, preferably approaching the theoretical limit of a single spatial–spectral–temporal mode. In the case of narrow linewidth spectral filtering at the ground station, the QKD optical source should also have a narrow linewidth to reduce the loss caused by filtering. Meanwhile, the QKD optical source working in the severe space environment has to be inherently robust against factors such as random vibration and temperature changes. Additionally, integrating laser communication can enable the online post-processing of quantum communication and extract the final secure keys in real time. To realize quantum-satellite-based QKD in practice, the above challenges must be resolved in comprehensive verifications.

We combined multiple techniques involving single-mode-fiber (SMF) spatial filtering, narrow linewidth spectral filtering, and gated temporal filtering for daytime noise suppression. The working wavelength was 1550 nm to reduce the background noises due to solar irradiation and Rayleigh scattering [25]. The spatial filtering was realized by collecting photons into the SMF and suppressing the corresponding field of view of the receiving telescope, similar to previous experimental demonstrations [25,26]. By employing an appropriate narrow linewidth interference filter strategy, a near single spectral-temporal mode that is close to the theoretical time-bandwidth product can be obtained. However, the time-bandwidth product limit forces a trade-off between pulse width and spectral width selection. A too narrow spectrum, such as 1 pm, brings challenges to wavelength stabilization and also limits the highest repetition frequency, whereas a too short pulse width, such as 1 ps, makes the time synchronization system difficult, especially in field experiments. Here, considering the trade-off between system performance and field experiment feasibility, the employed spectral filter is a Fabry–Perot (FP) cavity with a bandwidth of ${\sim}28\;{\rm pm}$ (full width at half-maximum, FWHM) at 1550 nm, while the pulse width after spectral filtering is tested as ${\sim}50\;{\rm ps}$ (FWHM) with the intensity autocorrelation (see Supplement 1). The calculated time-bandwidth product $\Delta \nu \Delta t$ is ${\sim}0.17$, indicating a near-single spectral-temporal mode that carries the quantum signals. When accounting for detector jitter, synchronization, and time-to-digital converter resolution, the total time precision of our system in the field experiment is approximately 240 ps. At the receiver, a temporal filter with a gate width of 800 ps is applied to further decrease the dark count noise with a factor of 0.5 while ensuring the high efficiency of quantum photons.

The QKD optical source is implanted in a scheme with a single laser diode (LD) and external electric-optical modulations [Fig. 1(d)] to suppress potential side channels. The 1550-nm LD is gain-switched by consecutive electrical pulses to generate short light pulses at a repetition frequency of 625 MHz. The intensity is modulated to prepare the three intensity states of decoy-state theory (signal state, decoy state, and vacuum state) [31,32]. The modulation scheme is based on the inherently stable Sagnac interferometer scheme [33], which employs a $2 \times 2$ beam splitter and a lithium niobate (${{\rm LiNbO}_3}$) phase modulator (PM). Similarly, the polarization is modulated to prepare the four polarization states of the standard BB84 protocol [1] ($| D \rangle = \frac{1}{{\sqrt 2}}({| H \rangle + | V \rangle})$, $| A \rangle = \frac{1}{{\sqrt 2}}({| H \rangle - | V \rangle})$, $| L \rangle = \frac{1}{{\sqrt 2}}({| H \rangle + i| V \rangle})$, and $| R \rangle = \frac{1}{{\sqrt 2}}({| H \rangle - i| V \rangle})$). The polarization modulation is also based on the Sagnac interferometer scheme [34] using a customized polarization module and another ${{\rm LiNbO}_3}$ PM. The Sagnac interferometric configuration is inherently stable, ensuring a highly robust optical source [34], with patterning effects mitigated [33]. The linewidth of the optical source is narrowed by another FP filter with the same bandwidth to ensure the high efficiency of spectral filtering. After spectral filtering, the optical source is attenuated to the single-photon level and then fed into the transmitting telescope.

To break the bottleneck of the classical microwave channel, our QKD system integrates bidirectional laser communication with a code rate of 156 Mbps (see Supplement 1). For downlink laser communications, specific synchronization patterns with a repetition frequency of ${\sim}9.5\;{\rm kHz}$ are embedded in the communication data, which are encoded into laser signals at the transmitter. At the receiver, the clock and communication data are recovered using the clock data recovery technique, and the target synchronization patterns can be extracted for real-time time synchronization with local-test accuracy of approximately 100 ps. After time synchronization, the bidirectional laser communication link enables the processes of basis sifting, error correction, and privacy amplification (see Supplement 1). Note that the unstable free-space channel will introduce probabilistic data loss and bit errors and may even interrupt the data transmission signal. To guard against these errors, we designed a robust key extraction process and split the key data into packets to extract the final secure keys using the decoy-state QKD analysis method (see Supplement 1). Once the data processing flow is interrupted during an optical link interruption, it can automatically resume operation when the optical link returns to normal. The block length of each packet is set to 100 kbits, and the low-density parity check (LDPC) method is adopted for error correction. This choice is mainly because the LDPC method only requires one data interaction and is particularly suitable for unstable free-space channels.

3. EXPERIMENT SETUP

At the sending terminal [Fig. 1(c)], decoy-state QKD photons are prepared in real time using a physical random number generator. The dichromatic mirrors separate the 1550-nm QKD photons, the 815-nm downlink beacon light and communication laser, the 671-nm uplink beacon light, and the 1538-nm uplink communication laser. Fast-steering mirrors are used for closed-loop fine tracking based on the 671-nm beacon laser images captured by the camera. These QKD photons and the downlink laser are sent through a telescope with a diameter of 0.18 m. After collimation by the telescope, the divergence angles of the quantum and beacon lights are 20 and 150 µrad, respectively.

At the receiving terminal [Fig. 1(b)], the QKD photons are collected into a telescope with a diameter of around 1.2 m. These QKD photons are coupled into an SMF for spatial filtering, then passed through the dense wavelength division multiplexing and another FP filter for spectral filtering. Finally, they are fed into the BB84 analysis module. By controlling the working temperatures of the FP filters at the sending and receiving terminals (one FP filter at each terminal), the respective central wavelengths can be adjusted for precise matching. Photons output from the BB84 analysis module are transferred through four SMFs (40 m) and are detected by a four-channel superconducting-nanowire single-photon detector (SNSPD) (with approximately detection efficiencies of 70% and total dark counts of less than 200 cps). The received 815 nm downlink laser signals are divided into several parts that serve different functions: the tracking beam in the acquisition, tracking and pointing system (detected by the tracking camera), and the laser-communication signal for real-time time synchronization and key extraction (coupled into a multimode fiber (MMF) with a reflective collimator).

 

Fig. 2. Experimental results of free-space QKD. (a) Efficiency; (b) ${R_0}$; (c) quantum bit error rate (QBER); (d) final key rate. The data were collected on different days from May 31 and June 13. The ${R_0}$ tests were conducted without fine-tracking in the minutes before the QKD experiments, while the data points of efficiency, QBER, and final key rate represented the average measurement results obtained from runs lasting approximately 1 h.

Download Full Size | PDF

The transmitter telescope is capable of projecting the quantum photons with a local divergence angle of approximately 20 µrad and a field tracking error of ${\sim}2\;{\unicode{x00B5} \rm rad}$ (see Supplement 1). Following the introduction of atmospheric turbulence, which leads to beam expansion, the measured beam footprint from the scanning test is approximately 1.2 m, comparable to the diameter of the receiver telescope. This result corresponds to an equivalent far-field divergence angle of around 60 µrad. The receiver telescope used was upgraded from the ground-based telescope originally designed for the Micius satellite experiment [10], in which the APT system remains unchanged (see Supplement 1). To simulate beam truncation toward future satellite-to-ground experiments and also optimize SMF coupling, we use a smaller sized coupler to intercept the received beam, reducing its diameter to 1/3 of the original beam. Therefore, the effective diameter of the receiver in this experiment is 0.4 m. The field of angles for SMF and MMF coupling at the receiver are approximately 6 and 40 µrad, respectively.

4. RESULTS AND DISCUSSION

We conducted a comprehensive QKD demonstration, spanning May 31 to June 13, 2020, encompassing both daytime and nighttime scenarios. The intermittent weather conditions impacted our measurement efficiency, leading to non-continuous data collection throughout the 24-hour period. Notably, the key extraction was performed in real time. The corresponding outcomes are illustrated in Fig. 2. Hourly, the average efficiency varied from ${-}{43.3}$ to ${-}{34.7}\;{\rm{dB}}$ [Fig. 2(a)]. The corresponding ${R_0}$ ranging from 1.6 to 13.7 cm [Fig. 2(b)]. Here, we employed the methodology of analyzing the statistics of beam pointing on the fine-tracking camera without control to calculate ${R_0}$ [35,36] and converted the results to the typical wavelength of 532 nm. It is noteworthy that the ${R_0}$ test was performed before the QKD experiment, not simultaneously. Therefore, the short-term ${R_0}$ values may not comprehensively reflect the QKD performance lasting for hours. The variation of results in each hour can be attributed to atmospheric turbulence affecting the laser beam’s wavefront, subsequently reducing SMF coupling efficiency. The quantum bit error rate (QBER) ranged from 0.87% to 2.16% hourly [Fig. 2(c)], corresponding to final key rates between 59.9 and 1529.5 bps [Fig. 2(d)]. In total, we generated 42.7 Mbits of secure keys at an average secure key rate of approximately 495 bps. See Supplement 1 for detailed experimental data.

The performance of daytime experiments strongly depends on the SNR of the QKD system, which can be enhanced by minimizing noise due to scattered solar irradiation and optimizing the efficiency of the quantum signals. In our study, we employed effective noise suppression techniques and collected sky radiation to evaluate the impact of induced dark counts. After reducing the base dark counts of the detectors, the measured dark count rate increment after spatial and spectral filtering is less than 500 cps, which can be further reduced to be less than 250 cps through the temporal filtering strategy. We also observe that as the elevation angle increases from the horizon to over 30°, the dark count gradually decreases by more than 60%, and then fluctuates within a small range. This phenomenon is mainly caused by the severe scattering effect of the ground atmosphere and various ground targets, such as mountains and vegetation. This suggests that for future satellite-to-ground quantum links with high-elevation angles, QKD system may suffer from fewer background counts at daytime. We have also compared the channel efficiency analysis of this field experiment to future satellite experiments (see Supplement 1). Considering the typical experimental parameters and the same SMF coupling efficiency as in this field experiment, the overall channel loss for LEO-satellite-based experiments is comparable to this work. Therefore, our experiment comprehensively verified the feasibility of LEO-satellite-based daytime QKD with real-time key distillation capabilities.

Statistical analysis of SMF coupling efficiency is crucial for assessing system performance, especially in a turbulent channel [37,38]. Detailed coupling data statistics and fitting using the probability density function are provided in the supplementary file. Furthermore, it is noteworthy that achieving high-efficiency SMF coupling is a challenging technique in free-space quantum communication. One promising approach is to integrate adaptive optics (AO) technology, although this technology was not included in this study. This choice is based on the fact that the intensity of atmospheric turbulence in the 20 km terrestrial free-space link is worse than that of the satellite-to-ground links, as well as the practical difficulty of upgrading the Nanshan station. Previous ${R_0}$ tests at the observatory’s ground stations typically yields values better than 10 cm at high-elevation angles [39]. This suggests that for future satellite-to-ground quantum links, cutting-edge AO techniques can show clear advantages, potentially achieving SMF coupling efficiency as high as 30% [40]. Furthermore, integration of artificial-intelligence-powered algorithms into the AO system also provides a promising avenue for enhancing free-space QKD performance under strong turbulence. Combining breakthroughs in AO technology and optimization of experimental parameters, we predict that daytime QKD for satellite in medium-Earth orbit of 10,000-kilometer is also feasible. For satellite-based applications, when the linewidth of the QKD light source and the linewidth of the ground filter are so narrow, the impact of the Doppler effect cannot be ignored. To address this problem, the central wavelength of the FP filter can be accurately adjusted by adjusting the temperature [41], thereby achieving high-precision compensation of Doppler frequency shift.

The performance of our QKD system can be further improved in several ways. For the LEO-satellite-based QKD link with limited communication time (about 5–10 min per orbit), the amount of received quantum photons during the satellite’s transit in an orbit is limited, and the impact of finite-key-size effect is prominent [42,43]. In the current field experiment, the block length of the QKD data post-processing is set to 100 kbits, and we imposed Chernoff bounds [44,45] on the observed values to estimate the lower and upper bounds and the failure probability was set to ${10^{- 7}}$. The block length can be effectively increased (for example, 1–10 Mbits), while the failure probability can be reset to be a smaller value, to mitigate the impact of the finite-key-size effects and extract secure keys more stringently. By utilizing the high-precision time synchronization method [46] and low-jitter detectors, the temporal filtering gate width can be further reduced to improve the overall SNR and help generate positive keys in a more lossy environment, at the cost of additional quantum photon loss. Changing the wavelength from 1550 nm to the visible light band can reduce the diffraction-limited divergence angle and improve the link efficiency, while selecting an appropriate Fraunhofer window [24,47] can also reduce the solar background. In free-space and satellite-based QKD with SMF coupling, atmospheric turbulence causes strong fluctuations in transmittance, and we can perform post-selection in real time and improve the key rate by employing the prefixed-threshold real-time selection method [48]. Integrated photonics technique [28,49,50] can significantly reduce the size and weight of the QKD system, and wavelength division multiplexing technique can be adopted to improve the key rate [51].

5. CONCLUSION

In conclusion, we employed a wide range of leading technologies, including daytime noise suppression approaching a single spatial–spectral–temporal mode, a high-speed robust QKD optical source, and laser-communication-based real-time key extraction, and demonstrated a 20-km QKD experiment with 24-hour coverage and real-time key distillation capacity between a satellite payload prototype and the Nanshan ground station. The background photons in our field experiment with a near-horizon link are higher than those of satellite-based applications with high-elevation angles. The total efficiency in our field experiment is comparable with future quantum-satellite application scenarios. Our experimental results will contribute to the follow-up realization of all-day real-time QKD using quantum satellites. In future work, we will attempt quantum light emission near the diffraction limit with large-aperture telescopes and high-efficiency SMF coupling based on adaptive optics, and we expect that a global quantum network combining multiple LEO satellites and high-orbit satellites will be gradually realized after several years of devoted efforts.