Scalable Network for Simultaneous Pairwise Quantum Key Distribution via Entanglement-Based Time-Bin Coding

We present a scalable star-shaped quantum-key-distribution (QKD) optical-fiber network. We use wavelength-division demultiplexing (WDM) of broadband photon pairs to establish key exchange between multiple pairs of participants simultaneously. Our QKD system is the first entanglement-based network of four participants using BBM92 time-bin coding and the first network to achieve timing synchronization solely by clock recovery based on the photon arrival times. We demonstrate simultaneous bipartite key exchange between any possible combination of participants and show that the quantum bit error rate (QBER) itself can be used to stabilize the phase in the interferometers by small temperature adjustments. The key distribution is insensitive to polarization fluctuations in the network, enabling key distribution using deployed fibers even under challenging environmental conditions. We show that our network can be readily extended to 34 participants by using a standard arrayed-waveguide grating for WDM with 100 GHz channel spacing and that reconfigurable network connections are possible with a wavelength-selective switch. In a field test, we demonstrate secure key rates of 6.3 bits/s with a QBER of 4.5% over a total fiber length of 108 km with 26.8 km of deployed fiber between two participants with high stability. Our system features a relatively simple design of the receiver modules and enables scaling QKD networks without trusted nodes to distances up to more than 100 km and to more than 100 users. With such a network, a secure communication infrastructure on a metropolitan scale can be established.

Popular Summary

The public key infrastructure on which the security of the Internet is based will be rendered insecure by quantum computers. However, physics also provides a solution, in the form of quantum key distribution (QKD): the laws of quantum physics enable provably secure cryptographic key exchange. The main focus of QKD research has been the advancement of key rates and distances for connections between two users but QKD networks are essential for a variety of applications.

So far, QKD networks have relied on daisy-chained connections requiring trusted nodes or have been very sensitive to environmental fluctuations between the network participants. In this work, we present the first robust star-shaped optical-fiber network for simultaneous QKD between any two parties based on time-bin entanglement. It is implemented with imbalanced interferometers in the photon-pair source and the receiver modules. In contrast to polarization-based protocols, it is inherently robust against environmental influences causing variations of the birefringence in the fiber link. But the interferometers must show a high degree of phase stability. We achieve superb stability and are thus able to employ the quantum bit error rate itself as the feedback signal for stabilization. As a proof of concept, we demonstrate a network with four participants and simultaneous key exchange between any two parties.

Our setup is scalable to over 100 participants, providing the first practical approach to networks in real-world applications without a trusted node. This enables the securing of communication infrastructure and the establishment of robust metropolitan-scale QKD networks.

Figure 1

(a) The scheme for time-bin entanglement quantum key distribution with the BBM92 protocol [36, 37, 38]: PLS, pulsed laser source; BS, 50:50 beam splitter; PPP, photon-pair production; DET, single-photon detector. The symbols $\varphi$, $\alpha$, and $\beta$ indicate the phase delay of the interferometers. (b) The arrival-time histogram at one of the detectors. Depending on the combination of long and short paths taken in the pump and receiver interferometers, the photons arrive in one of three time bins. Detection events in the early (short-short) and late (long-long) peak yield detections in the time basis, while events in the central peak (long-short and short-long) yield detections in the phase basis. (c) The possible configurations in a four-participant star-shaped network with pairwise mutually exclusive combinations. For one given configuration, each participant is linked to exactly one other participant, so that two connected pairs can exchange keys simultaneously and independently. The dotted green lines indicate the communication over a classical channel, which is required in addition to the unidirectional quantum channel.

Figure 2

(a) The setup for time-bin entanglement quantum key distribution with four participants: AM, amplitude modulator; EDFA, erbium-doped fiber amplifier; CIR, circulator; IF, interferometer; SHG, second-harmonic generation; SPDC, spontaneous parametric down-conversion; AWG, arrayed-waveguide grating; DET, single-photon detector. (b) The setup of the interferometers: BS, 50:50 beam splitter; FRM, Faraday rotator mirror. All interferometers are placed in temperature-stabilized boxes.

Figure 3

The histogram of photons traveling through 10.5 km of fiber before detection at one of Charlie’s detectors. The histogram shows the pulse interleaving. The structure of three peaks (early, central, and late) for each color is produced by the interferometer time delay of 3.03 ns [cf. Fig. 1]. By setting the source repetition time to 4.55 ns (corresponding to 219.78 MHz), a secondary triplet of time bins (blue) is interleaved with the primary triplet (red). The data are accumulated over 90 s. Although the peaks in the detection histogram are broader than the laser pulses of the photon source due to chromatic dispersion and detector timing jitter, they are sufficiently separated to avoid overlap, so that the QBER is kept low.

Figure 4

A simultaneous key exchange for Alice and Diana (red) and Charlie and Bob (black) without fiber spools and with a mean photon-pair number per pulse $\mu$ below 1×10^-3 demonstrating low QBERs. The temperature change $\mathrm{\Delta }T$ shows adjustments made at Alice’s and Bob’s receivers to minimize the error rate. Each data point is obtained from one run of 90 s. The average QBER is 0.24% for Alice and Diana and 0.41% for Charlie and Bob, respectively.

Figure 5

The estimated secure key rates and QBERs for simultaneous key exchange over fiber spools with a total length of 47.5 km for Alice and Diana (red) and 60.5 km for Charlie and Bob (black) with a $\mu$ in the range of 0.03. $\mathrm{\Delta }T$ denotes the temperature adjustments made at Alice’s and Bob’s receiver interferometers. Each data point represents one 90 s-long run.

Figure 6

The QBER and sifted key rate for key exchange between Charlie and Bob over 60.5 km of fiber for different AWG channel pairs symmetric around our center frequency. Each data point is averaged over eight consecutive 90-s runs. The error bars represent the standard deviation of the QBER for each channel. The error bars of the sifted key rate are so small that they are not visible.

Figure 7

The QBERs (orange) and sifted key rates (red) for all combinations of the four participants, Alice, Bob, Charlie, and Diana, over different fiber lengths. Fiber spools are assigned as shown in Fig. 2. The data are obtained during three 20-run measurements via simultaneous pairwise key exchange. The dashed lines separate the results of the three measurements. The error bars represent the standard deviation of the results for each combination.

Figure 8

The deployed standard single mode dark-fiber link of Deutsche Telekom for the field test. The link consists of a looped fiber running from the Deutsche Telekom facility in Darmstadt to Griesheim and a parallel fiber running back to Darmstadt. The fiber length is 26.8 km, with a total loss of 6.7 dB. (Google Maps 2021 ©).

Figure 9

The QBER and secure key rate acquired during the field test. A wavelength-selective switch is employed for demultiplexing of the SPDC spectrum. Alice is connected via the deployed fiber (26.8 km; see Fig. 8). The other participants are connected via spooled fiber: Bob via 81.2 km, Charlie via 9.6 km, and Diana via 20.9 km. The total distance between two users is the sum of the individual link lengths. Diana’s time tagger is synchronized to the photon source, while for the other participants synchronization is achieved via clock recovery from the photon arrival times. Black crosses indicate runs where Bob’s clock recovery fails and the delay to Alice is therefore automatically recalibrated via cross-correlation.