Loïc Millet

Phase randomization is essential for the security of practical quantum key distribution (QKD) systems. Commonly, implementations rely on laser sources (either actively phase-randomized, or gain-switched). However, at high repetition rates these show correlations, which can ultimate compromise security and performance. We present a 1.25 GHz fully phase-randomized QKD source based on a super luminescent diode (SLED) operating in the C-band as a compact and cost-effective alternative. The source generates ∼ 100 ps optical pulses with 400 ps time-bin separation, compatible with high-speed time-bin encoding. Interferometric measurements demonstrate > 99% visibility between adjacent time bins, confirming strong first-order coherence within a qubit, while the spontaneous-emission-driven nature of the SLED ensures intrinsic pulse to pulse phase randomization. The broadband architecture further enables operation across multiple ITU channels, supporting wavelength-multiplexed QKD from a single emitter. This work establishes a scalable SLED-based platform for high-speed time-bin QKD systems.

Clock synchronization between the transmitter (Alice) and receiver (Bob) is essential for practical quantum key distribution (QKD) systems. In time-bin protocols, a frequency mismatch between the local clocks of Alice and Bob leads to a timing drift that broadens photon detection-time histograms, increasing the quantum bit error rate (QBER) if left uncompensated. In many practical systems this issue is mitigated by distributing a reference timing signal over a dedicated channel. However, this approach complicates the deployment of QKD in existing telecommunication networks, motivating synchronization methods that operate directly on photon detections from the quantum channel. Here, we present a lightweight synchronization algorithm for time-bin BB84 systems based on an analytical model of detection-time statistics under clock drift. The model describes how a constant frequency mismatch modifies start–stop histograms over a finite acquisition time, enabling direct estimation of the clock drift directly from detection events. This leads to a simple algorithm requiring only two consecutive histograms per update and compatible with standard hardware. We experimentally validate the method on commercial QKD systems. Laboratory tests over 100 km of fiber and under variable quantum channel attenuation demonstrate rapid convergence and low QBER, comparable to that obtained using a shared reference clock. We further demonstrate stable operation over 24 hours on a 16 km section of the Geneva Quantum Network.