Abstract

Quantum key distribution (QKD), a technique based on quantum physics, provides unconditional secure quantum keys to be shared between two or more clients (Alice and Bob) [1]. Most QKD systems are implemented in a point-to-point link using bulky and expensive devices. Consequently a large scale deployment of this technology has not been achieved. A solution may be integrated photonic circuits, which provide excellent performances (compact, good optical phase stability, new degrees of freedom), and are particularly suitable for the manipulation of quantum states in compact chips. Some recent experiments have already demonstrated conventional binary QKD systems, using polarization and phase reference degrees of freedom [2, 3]. In this paper, we show the first silicon chip-to-chip decoy-state BB84 protocol based on spatial degrees of freedom (the cores of a multi-core fiber -MCF-). By tuning cascaded Mach-Zehnder interferometers (MZIs), it is possible to prepare the quantum states in two mutually unbiased basis (MUBs) sets: basis $\mathcal{X}\in \left\{\right|A〉,|B〉\}$, and basis $\mathcal{Z}\in \left\{\right|A+B〉,|A-B〉\}$. |A〉 and |B〉 are the quantum states related to two individual cores of the MCF, while |A + B〉 and |A − B〉 represent the superposition of the quantum state between cores, combined with a positive/negative phase relation. A train of weak coherent pulses (5 kHz repetition and 10 ns wide) are injected into the transmitter chip (Alice), where multiple variable optical attenuators (VOAs) are used to decrease the number of photons per pulse (µ < 1) [4]. Moreover, by using a combination of MZIs and VOAs, a decoy state-technique is implemented. Alice, by using an FPGA board, (Fig. 1(a)) randomly chooses one of the two bases and one of the two states to transmit to Bob. The qubits are matched to two cores of a multi-core fiber, through a highly efficient MCF grating coupler. After 3 meters link, the quantum states are coupled into Bob’s chip (Fig. 1(a)) through the MCF coupler, and randomly measured in one of the two bases. In the subsequent distillation process, counts measured in the wrong basis are discarded. In Fig. 1(b) and (c) the experimental data acquired within 11 minutes of measurement. In particular, Fig. 1(b) shows the gain of the decoy state technique (Q_µ). In Fig. 1(c) a stable bit error rate, well below the threshold limit for coherent attacks of 11%, is measured for more than 11 minutes.

© 2017 IEEE