PHYSICAL REVIEW LETTERS 126, 250502 (2021)

Editors' Suggestion Featured in Physics

Field Test of Twin-Field Quantum Key Distribution through

Sending-or-Not-Sending over 428 km
Hui Liu,1,2 Cong Jiang,3 Hao-Tao Zhu ,1,2 Mi Zou,1,2 Zong-Wen Yu,4,5 Xiao-Long Hu,4 Hai Xu ,4
Shizhao Ma,3 Zhiyong Han,3 Jiu-Peng Chen,1,2 Yunqi Dai,6 Shi-Biao Tang,6 Weijun Zhang,7 Hao Li,7
Lixing You,7 Zhen Wang,7 Yong Hua,8 Hongkun Hu,8 Hongbo Zhang ,8 Fei Zhou,3 Qiang Zhang,1,2,3
Xiang-Bin Wang,2,3,4,* Teng-Yun Chen ,1,2,† and Jian-Wei Pan1,2,‡
1
Hefei National Laboratory for Physical Sciences at Microscale and Department of Modern Physics,
University of Science and Technology ofChina, Hefei, Anhui 230026, People’s Republic of China
2
CAS Center for Excellence in Quantum Information and Quantum Physics, University of Science and Technology of China,
Hefei, Anhui 230026, People’s Republic of China
3
Jinan Institute of Quantum Technology, Jinan, Shandong 250101, People’s Republic of China
4
State Key Laboratory of Low Dimensional Quantum Physics, Department of Physics, Tsinghua University,
Beijing 100084, People’s Republic of China
5
Data Communication Science and Technology Research Institute, Beijing 100191, People’s Republic of China
6
QuantumCTek Corporation Limited, Hefei, Anhui 230088, People’s Republic of China
7
State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology,
Chinese Academy of Sciences, Shanghai 200050, People’s Republic of China
8
Chongqing Optoelectronics Research Institute, Chongqing 400060, People’s Republic of China
(Received 12 January 2021; accepted 10 May 2021; published 22 June 2021)

Quantum key distribution endows people with information-theoretical security in communications.

Twin-field quantum key distribution (TF-QKD) has attracted considerable attention because of its
outstanding key rates over long distances. Recently, several demonstrations of TF-QKD have been
realized. Nevertheless, those experiments are implemented in the laboratory, and therefore a critical
question remains about whether the TF-QKD is feasible in real-world circumstances. Here, by adopting the
sending-or-not-sending twin-field QKD (SNS-TF-QKD) with the method of actively odd parity pairing
(AOPP), we demonstrate a field-test QKD over 428 km of deployed commercial fiber and two users are
physically separated by about 300 km in a straight line. To this end, we explicitly measure the relevant
properties of the deployed fiber and develop a carefully designed system with high stability. The secure key
rate we achieved breaks the absolute key rate limit of repeaterless QKD. The result provides a new distance
record for the field test of both TF-QKD and all types of fiber-based QKD systems. Our work bridges the
gap of QKD between laboratory demonstrations and practical applications and paves the way for an
intercity QKD network with measurement-device-independent security.

DOI: 10.1103/PhysRevLett.126.250502

Introduction.—Since Bennet and Brassard proposed the independent QKD (MDI-QKD) successfully [11–14].
BB84 protocol [1], quantum key distribution (QKD) has Second, the secure key rate can be further improved by
been studied extensively [2–7] towards its final goal of twin-field QKD (TF-QKD) [15]. Similar to the MDI-QKD
application in the real world. Given the fact that quantum protocol, TF-QKD is assisted by an untrusted relay,
signals cannot be amplified, the secure distance is severely Charlie, between Alice and Bob. However, instead of
limited by the channel loss. For example, considering the performing a two-photon Bell state measurement, TF-
possible photon-number-splitting attack, the key rate of a QKD is supposed to perform single-photon interference
BB84 protocol with the imperfect single-photon source is in Charlie, which allows it to provide a key rate in the
pffiffiffi
proportional to η2 , given the channel transmittance η. So square-root scale of the channel transmittance Oð ηÞ. In
far, many efforts have been made towards the more loss- principle, as a relay-assisted protocol, TF-QKD can break
tolerant QKD in practice. There are two mile-stone signs of the relayless bound which is known as the Pirandola-
progress towards this goal. First, the decoy-state method Laurenza-Ottaviani-Bianchi (PLOB) bound [16] and fun-
[8–10] can improve the key rate of coherent-state based damentally indicates that the key rate scales linearly with η
QKD from quadratic scaling η2 to linear scaling η, as a in the absence of the relay.
perfect single-photon source behaves. Importantly, the The real-world QKD aims to physically separate users on
method can be applied to the measurement-device- Earth. However, although tremendous efforts were made

0031-9007=21=126(25)=250502(6) 250502-1 © 2021 American Physical Society

 PHYSICAL REVIEW LETTERS 126, 250502 (2021)

into fiber-based QKD field test [17–25], the maximum fiber windows. A Z window event is regarded as being effective
distance and the maximal physical separation achieved if Charlie announces that only one detector clicked. An X
between two users is around 90 km [24] to date, and window event is defined as an event that both Alice’s weak
challenges for longer distances remain. coherent source pulse is μA1 and the intensity of Bob’s
It is worth noting that experimental TF-QKD [26–31] weak coherent source pulse is μB1 and their phases satisfy
has advanced significantly up to a distance of more than an extra phase-slice condition to reduce the observed error
500 km [30,31]. However, all the experiments are imple- rate [35]. As shown in Ref. [35], we set the condition of
mented in the laboratory with either the simulated channel
loss or the optical fiber spool, leaving a vast gap between μA1 ϵA ð1 − ϵB ÞμA2 e−μA2
¼ ð1Þ
laboratory demonstrations and practical applications. Field μB1 ϵB ð1 − ϵA ÞμB2 e−μB2
trial of TF-QKD remains experimentally challenging.
In this work, for the first time, we present a field test of for the security of our asymmetric protocol.
TF-QKD on the deployed commercial fiber (428 km An error in the X window is defined as an effective event
length with 79.1 dB channel loss, buried underground, in the X window when Charlie announces a click of right
ultra-low-loss fiber, G654.E). Furthermore, it is the most (left) while the phase difference between the pulse pair from
extended fiber-based QKD field test without relying on Alice and Bob would provably cause a left (right) clicking at
trusted relays. Two users, Alice and Bob, realize the longest Charlie’s measurement setup. At a signal window, Alice
physical separation distance (about 300 km) in the terres- (Bob) puts down a bit value 1(0) when she (he) decides
trial QKD so far, to the best of our knowledge. The secure sending, Alice (Bob) puts down a bit value 0(1) when she
key rate of our work breaks the absolute key rate limit of (he) decides not sending. The values of eph
1 and n1 , the phase-
trusted-relayless QKD. The result lays the foundation for flip error rate and the number of effective single-photon
an intercity-scale QKD network in the absence of the events in the Z basis, can be calculated by the conventional
quantum repeater. decoy-state method [32,36]. Then we can calculate the
We adopt the sending-or-not-sending (SNS) protocol secure key rate by the zigzag approach proposed in Ref. [33].
[32] of TF-QKD with finite-key effects [33]. Besides, we Calculation details are shown in Ref. [37].
apply the efficient error rejection method, known as Experiment.—In our field test, Alice and Bob are located
actively odd parity pairing (AOPP) [34] with the finite- in Jinan and Qingdao, respectively. The central relay Charlie
key effects studied in Ref. [33]. Given such an asymmetric is placed in Linyi, as shown in Fig. 1(a). The distance
channel, we adopt the asymmetric protocol [35] to improve between Charlie and Alice (Bob) is 223 km with 40.5 dB
the secure key rate further. channel loss (205 km with 38.6 dB channel loss).
Protocol.—Consider the SNS-TF-QKD protocol pro- The experimental setup is comprised of the synchroni-
posed in Ref. [32]. Here, we implement an asymmetric zation system and the encoding and measurement system,
three-intensity method for decoy-state analysis. To improve as shown in Fig. 1(b). Alice and Bob are connected by two
the key rate, we take the bit error rejection by AOPP [34] in parallel field-deployed commercial fibers (in the same
the post data processing stage. In this way, the sending optical cable) with 428 km length each, named “synchro-
probability in signal windows can be far improved and nization channel” and “quantum channel,” respectively, in
hence the number of effective events is raised greatly. As a the following.
result, the final key rate is improved a lot especially in the The synchronization system includes two functions:
case of small data size with finite key effects being (1) the clock synchronization, of which the details are
considered. We use the zigzag approach to take the shown in Ref. [37]; (2) the wavelength synchronization.
finite-key effects in calculating the final key rate [33]. The first issue that makes implementation difficult is
In the protocol, Alice (Bob) randomly chooses the avoiding the rapid relative phase drift caused by Alice’s
decoy window and signal window with probabilities and Bob’s lasers’ wavelength difference. We realize the
1 − pA2 ð1 − pB2 Þ and pA2 ðpB2 Þ, respectively. In the decoy wavelength synchronization with the assistance of the laser
window, both Alice and Bob prepare and send decoy injection technique [41]. A laser with 3 kHz linewidth is
pulses. In our three-intensity protocol, there are two types placed in Charlie as the master laser. The continuous-wave
of decoy states in decoy windows for each party of Alice (cw) bright beam is produced and injected into Alice’s and
and Bob, one vacuum and one nonvacuum coherent states, Bob’s slave laser. To guarantee a 0 dBm cw bright beam
of intensity μA1 for Alice and μB1 for Bob. Private random injected into the slave laser, we add four erbium-doped
phase shifts of θA and θB are applied to each pulse. And in fiber amplifiers (EDFAs), two of which are placed in
the signal window, Alice (Bob) decides to send out a Yiyuan and Zhucheng, respectively (as shown in Fig. 1).
phase-randomized weak coherent state pulse with intensity And the other two are added in Alice’s and Bob’s
μA2 (μB2 ) or a vacuum pulse with probabilities ϵA ðϵB Þ and apparatus. A 10 GHz fiber Bragg grating (FBG) is inserted
1 − ϵA ð1 − ϵB Þ, respectively. A Z window event is defined into Alice’s (Bob’s) apparatus to filter the amplified
as an event that both Alice and Bob choose the signal spontaneous emission (ASE) noise of the EDFAs.

250502-2
 PHYSICAL REVIEW LETTERS 126, 250502 (2021)

(a)

(b)

FIG. 1. (a) Bird’s-eye view of our field test. Alice is located at the Jinan Institute of Quantum Technology (JIQT) in Jinan (36°41’0.60”
N, 117°8’10.93” E), while Bob is located at an internet data center (IDC) room in Qingdao (36°7’24.29” N, 120°27’11.88” E). The third-
party measurement is done by Charlie in a room in Linyi (36°1’39.84” N, 118°44’50.58” E), which is 223 km from Alice and 205 km
from Bob. Two yellow marks show the locations of two machine rooms at Yiyuan (36° 11’12.60” N, 118° 12’24.16” E) and Zhucheng
(36°2’59.31” N, 119°24’43.58” E), respectively. An erbium-doped fiber amplifier (EDFA) is placed in each machine room to amplify the
light for the clock and wavelength synchronization. Map data from Google, Landsat/Copernicus. (b) Illustration of the experimental
setup. A continuous-wave (cw) bright beam from a 1550.12 nm master laser is multiplexed with the pulses from two 1570 nm auxiliary
synchronization lasers (Sync Lasers) in Charlie and is transmitted along the synchronization channel. The slave laser of Alice and Bob is
seeded by the cw bright beam and generates pulses with a width of 320 ps and a repetition rate of 312.5 MHz. The optical launch power
of the slave laser is monitored in real time by a watchdog photoelectric detector PD2 of Alice (Bob). Then these pulses are sent to two
sagnac rings SR1-2, which are randomly prepared in one of the four intensities strong μr , high μA2 (μB2 ), moderate μA1 (μB1 ), and
vacuum state. Three phase modulators PM1-3 are utilized for active phase randomization. The pulses are transmitted along the quantum
channel and interfere in Charlie. The schematic of the polarization auto-alignment module is shown inside the red dashed rectangle.
EVOA: electrical variable optical attenuator, FBG: fiber Bragg grating, CIR: circulator, EDFA: erbium-doped fiber amplifier,
EPC: electric polarization controller, DWDM: dense wavelength division multiplexer, PBS: polarizing beam splitter.

The pulses produced from the slave laser pass through the probability distributions are shown in Figs. 3(c) and
two sagnac rings (SRs) and three phase modulators (PMs) 3(d) and the details are shown in Ref. [37]. The dark count
for encoding and phase randomization in the encoding and of each SNSPD is about 6 cps, corresponding to a dark
measurement system. The pulses are attenuated to the count rate of 2.0 × 10−9 =pulse.
desired levels by an electrical variable optical attenuator Another challenge we have encountered is the significant
(EVOA) before being transmitted to Charlie through the changes to the relative phase drift stemming from the long
quantum channel. In Charlie, a 50∶50 BS performs a single fiber channel. A comparison of the relative phase drift in
photon interference of the incoming pulses after noise different fiber distances in previous works and our work is
filtering. The measurement results are detected by two shown in Fig. 1 of Ref. [37]. We stress that in our work, the
superconducting nanowire single photon detectors signal pulses produced by the slave laser inherit the global
(SNSPDs) with efficiencies of 73% and 76%, respectively. phase of the cw bright beam, which is influenced by the
Charlie’s overall detection efficiency is 28%, taking into 428 km synchronization channel. Then the signal pulses
account 2.4 dB insertion loss, 70% overlapping between transmit along the 428 km quantum channel before
signal pulse and detection window, and 94% polarization interference. So the relative phase drift is influenced by
alignment efficiency. The first two terms can be measured the total 856 km fiber link, which is 7.80 rad=ms in our
directly. The other two are the statistical results, of which work. Note that in our field test, the total optical fiber

250502-3
 PHYSICAL REVIEW LETTERS 126, 250502 (2021)
(a) (b)
50 1.68E-8

Measured

Crosstalk noise (/pulse)

40 Simulation 1.34E-8

Crosstalk noise (cps)

30 1.01E-8

20 6.72E-9

10 3.36E-9

0 0.00E+0
0 5 10 15

Launch power of the master laser (dBm)

FIG. 2. Characterization of the crosstalk noise. All measurements are performed under the same overall detection efficiency (28%).
(a) The crosstalk noise caused by the classical services running in some fibers in the optical cable. Without the master laser in the
synchronization channel, we test two available fiber channels, fiber 1 and fiber 2 (fiber 3 and fiber 4), from Alice (Bob) to Charlie. The
blue and red columns are the measurement results before and after filtering with two 100 GHz DWDMs, respectively. Note that we only
need one fiber channel as the quantum channel to transmits the signal pulses from Alice (Bob) to Charlie. Taken the loss and the
crosstalk noise of the fiber channel into account, we use fiber 1 (fiber 4) as the quantum channel from Alice (Bob) to Charlie. (b) The
crosstalk noise caused by the cw bright beam in the synchronization channel with different optical launch power. Each experiment lasts
5 min. The experimental results are the average and variance (1 standard deviation) calculated by 144 experiments.

influencing the relative phase is longer than all the previous which is also in the same optical cable. Thus it cannot be
lab works. However, the relative phase drift is significantly filtered, whether spectrally or temporally. We found that the
lower than the 800 km result in ref [30]. It is near to most crosstalk noise becomes more ignorable as the optical
results measured in the laboratory over a shorter fiber launch power of the cw bright beam decreases, which is
channel. It makes the relative phase calculation in our field shown in Fig. 2(b). To suppress the noise, we reduce the
test much less demanding than the lab experiment in optical launch power of the master laser to about 5 dBm and
Ref. [30] over 402 km. increase the EDFA gain appropriately, resulting in a noise
We verified that we could indeed estimate and compen- level of 3.6 × 10−9 =pulse. Still, a stable and high-efficiency
sate for the relative phase drift caused by the long fiber injection can be ensured in this case.
channel. In our work, Alice and Bob sacrifice a part of the Results.—In our field test, Alice and Bob send a total of
signal pulses as bright reference pulses periodically for 5.59 × 1012 pulse pairs, and obtain 2.79 × 107 sifted key
relative phase calculation and apply a postselection method bits in the Z basis, including 27.84% error bits. According
when the signal pulse detection occurs. (for details see to the method shown in Ref. [37] and the data acquired in
Refs. [30,37]). However, the scattering of the bright the experiment, there are at least 1.29 × 107 untagged bits
reference pulses will lead to nonignorable noises in the in the sifted keys, corresponding to an 11.07% phase flip
signal pulses [31]. After being filtered by four 100 GHz error rate before AOPP. After AOPP, 5.84 × 106 keys
dense wavelength division multiplexers (DWDMs) in survive which contain 0.69% error bits. The number of
Charlie, the remaining noise is about 1.4 × 10−8 =pulse. untagged bits is 2.38 × 106 with a corresponding phase flip
Besides, we face the crosstalk noise in the field test, error rate of 20.24%. With the finite-key effect being taken
which is never met in TF-QKD lab experiments. The into consideration, we finally obtain a secure key rate of
quantum channel for transmitting signal pulses is in an 4.80 × 10−8 =pulse, which is 170% higher than the absolute
optical cable (96 fibers included). Part of the noise proceeds PLOB bound, 859% higher than the relative PLOB bound,
from the classical services running in some fibers in the and 2–4 orders of magnitude improvement than two
optical cable. Fortunately, it can be filtered by a pair of comparable experiments over 400 km which were using
cascaded DWDMs at the end of each quantum channel in a BB84 [42] and a measurement-device-independent QKD
Charlie to approximately 5.1 × 10−9 =pulse, which is configuration [14], respectively. Figure 3 shows the per-
acceptable for us, as shown in Fig. 2(a). The other part formance of our work in terms of the simulation key rates,
of the crosstalk noise is raised from the cw bright beam the achieved secure key rate, and the total efficiency of the
(same wavelength with the signal, generated from the polarization auto-alignment module and arrival time
master laser in Charlie) in the synchronization channel, synchronization.

250502-4
 PHYSICAL REVIEW LETTERS 126, 250502 (2021)
(b) 0.16

Bit error in X windows

(a)
-4 3
10 7*10
0.12
Simulation key rate
Absolute PLOB bound
Relative PLOB bound 0.08
10-5 7*102
Secure key rate (bit/pulse)

Our Work

Secure key rate (bps)

0.04
-6 1
10 7*10 0.00
0 20 40 60 80
3
Valid time (10 s)
10-7 7*100
(c) 30 (d) 40

Probability (%)

Probability (%)
-8 -1 30
10 7*10
20
20
10-9 7*10-2 10
10
200 250 300 350 400 450 500
Distance (km) 0 0
0 8 16 24 40 60 80 100
Reflectivity (%) Ratio (%)

FIG. 3. (a) Experimentally and simulated secure key rates. The secure key rate (purple pentagram point) of our work is
4.80 × 10−8 /pulse, corresponding to 3.36 bps. The red curve is the simulation results using the parameters in Table I of Ref. [37].
The green and blue dashed line are the absolute PLOB bound (assuming the overall detection efficiency of Charlie ηd ¼ 1) and the
relative PLOB bound with ηd ¼ 28%, respectively. (b) The bit error rate in X windows. Each data point represents the effective clicks
collected in 21.82 min on average. (c) The probability distribution of the reflectivity for the PBSs in Charlie, with real-time
compensation in the overall experiment. The total efficiency of the polarization auto-alignment module is about 94% (the detail see
Ref. [37]). (d) The probability distribution of the overlapping rate between the signal pulse and detection window in the overall
experiment. The total overlapping is 70% (see details in Ref. [37]).

Conclusions.—Applying the SNS protocol [32], we have the Chinese Academy of Science, the National Fundamental
performed the first field test of TF-QKD over a 428 km Research Program, the National Natural Science Foundation
deployed commercial fiber. It is the longest distance of of China (Grants No. 11875173, No. 61875182, and
terrestrial real-word QKD without relying on trusted relays, No. 11674193), China State Railway Group Co., Ltd.
at present, and pushes the separation between two users Scientific and Technological Research Project under
beyond 300 km. The result demonstrated in our experiment Grant No. K2019G062, and Anhui Initiative in Quantum
exhibits the feasibility of the trusted-relayless QKD in Information Technologies and Fundamental Research
practical circumstances between cities. It motivates future Funds for the Central Universities (WK2340000083).
demonstration of an intercity-scale QKD network in the
absence of the quantum repeater.
Compared to the 2.5 GHz system repetition rate of the Note added.—We note that related experimental work has
existing BB84 protocol [42], our system’s overall repetition been reported in Ref. [44]. Both our work and Ref. [44] are
rate is not high enough. However, the TF-QKD protocol the field tests of SNS TF-QKD over almost the same
has a higher requirement in time synchronization and deployed commercial fiber. However, we realize the wave-
calibration than the BB84 protocol, making it challenging length synchronization with the assistance of the laser
to increase the TF-QKD system repetition significantly. injection technique, which is off the shelf, mature, and easy
Besides, the relative phase calculation and detection to implement; while the wavelength synchronization sys-
recovery decrease the effective repetition directly in our tem in Ref. [44] utilizes the time-frequency dissemination
experiment. Fortunately, a novel wavelength-multiplexed technology, which has less cross-talk noise and achieves a
approach for the relative phase calculation has been longer distance.
proposed and demonstrated [43]. It has the potential for
a more advantageous effective repetition rate for quantum
communication. Further extensions to higher key rates
*
include utilizing the fiber link with lower attenuation and [email protected]
†
less crosstalk noise, and enhancing the laser and detector’s [email protected]
‡
performance. [email protected]
[1] C. H. Bennett and G. Brassard, in Proceedings of the IEEE
This work was supported by the National Key R&D International Conference on Computers, Systems, and
Program of China (2017YFA0303903, 2020YFA0309800), Signal Processing (IEEE, New York, 1984), pp. 175–179.

250502-5
 PHYSICAL REVIEW LETTERS 126, 250502 (2021)

[2] N. Gisin, G. Ribordy, W. Tittel, and H. Zbinden, Rev. Mod. [24] Y.-A. Chen et al., Nature (London) 589, 214 (2021).
Phys. 74, 145 (2002). [25] M. Sasaki et al., Opt. Express 19, 10387 (2011).
[3] V. Scarani, H. Bechmann-Pasquinucci, N. J. Cerf, M. [26] M. Minder, M. Pittaluga, G. Roberts, M. Lucamarini, J.
Dušek, N. Lütkenhaus, and M. Peev, Rev. Mod. Phys. Dynes, Z. Yuan, and A. Shields, Nat. Photonics 13, 334
81, 1301 (2009). (2019).
[4] S.-K. Liao, W.-Q. Cai, W.-Y. Liu, L. Zhang, Y. Li, J.-G. Ren, [27] Y. Liu, Z.-W. Yu, W. Zhang, J.-Y. Guan, J.-P. Chen, C.
J. Yin, Q. Shen, Y. Cao, Z.-P. Li et al., Nature (London) 549, Zhang, X.-L. Hu, H. Li, C. Jiang, J. Lin, T.-Y. Chen, L. You,
43 (2017). Z. Wang, X.-B. Wang, Q. Zhang, and J.-W. Pan, Phys. Rev.
[5] F. Xu, X. Ma, Q. Zhang, H.-K. Lo, and J.-W. Pan, Rev. Mod. Lett. 123, 100505 (2019).
Phys. 92, 025002 (2020). [28] S. Wang, D.-Y. He, Z.-Q. Yin, F.-Y. Lu, C.-H. Cui, W. Chen,
[6] S. Pirandola, U. L. Andersen, L. Banchi, M. Berta, D. Z. Zhou, G.-C. Guo, and Z.-F. Han, Phys. Rev. X 9, 021046
Bunandar, R. Colbeck, D. Englund, T. Gehring, C. Lupo, C. (2019).
Ottaviani, J. L. Pereira, M. Razavi, J. S. Shaari, M. [29] X. Zhong, J. Hu, M. Curty, L. Qian, and H.-K. Lo, Phys.
Tomamichel, V. C. Usenko, G. Vallone, P. Villoresi, and Rev. Lett. 123, 100506 (2019).
P. Wallden, Adv. Opt. Photonics 12, 1012 (2020). [30] X.-T. Fang, P. Zeng, H. Liu, M. Zou, W. Wu,
[7] C. Portmann and R. Renner, arXiv:2102.00021. Y.-L. Y.-J. Sheng, Y. Xiang, W. Zhang, H. Li et al., Nat.
[8] W.-Y. Hwang, Phys. Rev. Lett. 91, 057901 (2003). Photonics 14, 422 (2020).
[9] X.-B. Wang, Phys. Rev. Lett. 94, 230503 (2005). [31] J.-P. Chen, C. Zhang, Y. Liu, C. Jiang, W. Zhang, X.-L. Hu,
[10] H.-K. Lo, X. Ma, and K. Chen, Phys. Rev. Lett. 94, 230504 J.-Y. Guan, Z.-W. Yu, H. Xu, J. Lin, M.-J. Li, H. Chen, H.
(2005). Li, L. You, Z. Wang, X.-B. Wang, Q. Zhang, and J.-W. Pan,
[11] S. L. Braunstein and S. Pirandola, Phys. Rev. Lett. 108, Phys. Rev. Lett. 124, 070501 (2020).
130502 (2012). [32] X.-B. Wang, Z.-W. Yu, and X.-L. Hu, Phys. Rev. A 98,
[12] H.-K. Lo, M. Curty, and B. Qi, Phys. Rev. Lett. 108, 130503 062323 (2018).
(2012). [33] C. Jiang, Z.-W. Yu, X.-L. Hu, and X.-B. Wang, Phys. Rev.
[13] Y.-H. Zhou, Z.-W. Yu, and X.-B. Wang, Phys. Rev. A 93, Applied 12, 024061 (2019).
042324 (2016). [34] H. Xu, Z.-W. Yu, C. Jiang, X.-L. Hu, and X.-B. Wang, Phys.
[14] H.-L. Yin, T.-Y. Chen, Z.-W. Yu, H. Liu, L.-X. You, Y.-H. Rev. A 101, 042330 (2020).
Zhou, S.-J. Chen, Y. Mao, M.-Q. Huang, W.-J. Zhang, H. [35] X.-L. Hu, C. Jiang, Z.-W. Yu, and X.-B. Wang, Phys. Rev. A
Chen, M. J. Li, D. Nolan, F. Zhou, X. Jiang, Z. Wang, Q. 100, 062337 (2019).
Zhang, X.-B. Wang, and J.-W. Pan, Phys. Rev. Lett. 117, [36] Z.-W. Yu, X.-L. Hu, C. Jiang, H. Xu, and X.-B. Wang, Sci.
190501 (2016). Rep. 9, 3080 (2019).
[15] M. Lucamarini, Z. L. Yuan, J. F. Dynes, and A. J. Shields, [37] See Supplemental Material at http://link.aps.org/
Nature (London) 557, 400 (2018). supplemental/10.1103/PhysRevLett.126.250502 for de-
[16] S. Pirandola, R. Laurenza, C. Ottaviani, and L. Banchi, Nat. tailed theoretical scheme, detailed experimental technolo-
Commun. 8, 15043 (2017). gies, and full experimental results, which includes
[17] M. Peev et al., New J. Phys. 11, 075001 (2009). Refs. [38–40].
[18] D. Stucki, M. Legre, F. Buntschu, B. F. Clausen, N. Felber, [38] H. Chernoff, Ann. Math. Stat. 23, 493 (1952).
N. Gisin, L. Henzen, P. Junod, G. Litzistorf, P. Monbaron [39] A. Vitanov, F. Dupuis, M. Tomamichel, and R. Renner,
et al., New J. Phys. 13, 123001 (2011). IEEE Trans. Inf. Theory 59, 2603 (2013).
[19] T.-Y. Chen, H. Liang, Y. Liu, W.-Q. Cai, L. Ju, W.-Y. Liu, J. [40] C. Jiang, X.-L. Hu, H. Xu, Z.-W. Yu, and X.-B. Wang, New
Wang, H. Yin, K. Chen, Z.-B. Chen, C.-Z. Peng, and J.-W. J. Phys. 22, 053048 (2020).
Pan, Opt. Express 17, 6540 (2009). [41] Z. L. Yuan, B. Fröhlich, M. Lucamarini, G. L. Roberts,
[20] J. F. Dynes, A. Wonfor, W. Tam, A. W. Sharpe, R. J. F. Dynes, and A. J. Shields, Phys. Rev. X 6, 031044
Takahashi, M. Lucamarini, A. Plews, Z. L. Yuan, A. R. (2016).
Dixon, J. Cho et al., npj Quantum Inf. 5, 101 (2019). [42] A. Boaron, G. Boso, D. Rusca, C. Vulliez, C. Autebert, M.
[21] T.-Y. Chen, J. Wang, H. Liang, W.-Y. Liu, Y. Liu, X. Jiang, Caloz, M. Perrenoud, G. Gras, F. Bussières, M.-J. Li, D.
Y. Wang, X. Wan, W.-Q. Cai, L. Ju, L.-K. Chen, L.-J. Wang, Nolan, A. Martin, and H. Zbinden, Phys. Rev. Lett. 121,
Y. Gao, K. Chen, C.-Z. Peng, Z.-B. Chen, and J.-W. Pan, 190502 (2018).
Opt. Express 18, 27217 (2010). [43] C. Clivati, A. Meda, S. Donadello, S. Virzì, M. Genovese, F.
[22] S. Wang, W. Chen, Z.-Q. Yin, Y. Zhang, T. Zhang, H.-W. Li, Levi, A. Mura, M. Pittaluga, Z. L. Yuan, A. J. Shields, M.
F.-X. Xu, Z. Zhou, Y. Yang, D.-J. Huang, L.-J. Zhang, F.-Y. Lucamarini, I. P. Degiovanni, and D. Calonico,
Li, D. Liu, Y.-G. Wang, G.-C. Guo, and Z.-F. Han, Opt. Lett. arXiv:2012.15199.
35, 2454 (2010). [44] J.-P. Chen, C. Zhang, Y. Liu, C. Jiang, W.-J. Zhang, Z.-Y.
[23] Y.-L. Tang, H.-L. Yin, Q. Zhao, H. Liu, X.-X. Sun, M.-Q. Han, S.-Z. Ma, X.-L. Hu, Y.-H. Li, H. Liu, F. Zhou, H.-F.
Huang, W.-J. Zhang, S.-J. Chen, L. Zhang, L.-X. You, Jiang, T.-Y. Chen, H. Li, L.-X. You, Z. Wang, X.-B. Wang,
Z. Wang, Y. Liu, C.-Y. Lu, X. Jiang, X. Ma, Q. Zhang, T.-Y. Q. Zhang, and J.-W. Pan, arXiv:2102.00433.
Chen, and J.-W. Pan, Phys. Rev. X 6, 011024 (2016).

250502-6