Experimental Passive Round-Robin Differential Phase-Shift Quantum Key Distribution

In quantum key distribution (QKD), the bit error rate is used to estimate the information leakage and hence determines the amount of privacy amplification—making the final key private by shortening the key. In general, there exists a threshold of the error rate for each scheme, above which no secure key can be generated. This threshold puts a restriction on the environment noises. For example, a widely used QKD protocol, the Bennett-Brassard protocol, cannot tolerate error rates beyond 25%. A new protocol, round-robin differential phase-shifted (RRDPS) QKD, essentially removes this restriction and can in principle tolerate more environment disturbance. Here, we propose and experimentally demonstrate a passive RRDPS QKD scheme. In particular, our 500 MHz passive RRDPS QKD system is able to generate a secure key over 50 km with a bit error rate as high as 29%. This scheme should find its applications in noisy environment conditions.

Figure 1

(a) Original RRDPS scheme [4]. VDL stands for variable delay line. Bob splits the received signals into two paths and applies a variable delay $r$ to one of the paths. (b) Equivalent model. Bob obtains a click at position $i$, generates two random numbers $r\in \{1,\dots ,L-1\}$ and $b\in \{0,1\}$ to obtain $j=i+(-1{)}^{b}r$, and publicly announces $i$ and $j$ to Alice. (c) Passive RRDPS scheme. Bob uses a local laser to generate an $L$-pulse reference, which interferes with Alice’s $L$-pulse signal. Bob then records the coincidence clicks.

Figure 2

Experiment setup. ATT, attenuator; IM, intensity modulator; BS, polarization maintaining beam splitter; PM, phase modulator; Delay, optical adjustable delay line; PC, polarization controller; PBS, polarization beam splitter (single mode to polarization maintaining). The attenuation on Bob’s side is realized by a polarization controller and a polarization beam splitter.

Figure 3

The dependence of the key rate on the block size $L$ at various distances. For each distance, we repeat the experiment 10 times and take the standard deviations in 10 trials as the error bar.