Hybrid-Integrated InGaAs/InP SPAD Arrays for Quantum Communications

TL;DR

The paper addresses the need for compact, non-cryogenic QKD receivers by developing GHz-gated InGaAs/InP SPAD arrays integrated with low-loss silica waveguides. The authors implement a scalable hybrid approach using SACGM SPADs, sub-ns gate durations to suppress inter-pixel crosstalk, and quasi-planar coupling to deliver efficient, densely packed detectors for BB84 QKD. Key contributions include achieving record-low crosstalk (<0.1%), demonstrated BB84 operation with secure key rates up to ~1 Mbps (and 2.1 Mbps at room temperature for 0 dB) and 100 km fibre viability, and outlining practical integration pathways for metropolitan networks. The work advances scalable, miniaturized quantum receivers with potential impact beyond QKD to quantum sensing and related photonic technologies, while also highlighting packaging and passive-receiver design considerations for real-world deployment.

Abstract

Photonic integration is a promising route to miniaturise the hardware of quantum key distribution (QKD), yet the monolithic integration of single photon detectors remains a significant challenge. QKD receiver chips integrating superconducting detectors have been demonstrated, but their requirement for cryogenic cooling restricts their practical applications. High-frequency gated single-photon avalanche diodes (SPADs) provide a mature non-cryogenic alternative and their fabrication into compact arrays would enable scalable hybrid integration. However, this faces several challenges related to efficient GHz array gating, inter-pixel crosstalk, and scalable waveguide coupling, which to date remain unaddressed. Here, we overcome the key challenges and develop GHz-gated InGaAs/InP SPAD arrays with performance viable for QKD and negligible inter-pixel crosstalk. We combine the arrays with low-loss silica waveguide chips to produce compact hybrid QKD receivers and perform BB84 protocol experiments, achieving secure key rates over 2 Mbps at short distances and 15 kbps over 100 km of fibre. Our work provides a method for flexible and scalable integration of waveguide-coupled SPADs for quantum information processing applications.

Figures (3)

• Figure 1: a The quasi-planar coupling technique. A diagonal cut at one end of the silica chip causes light propagating within the waveguides to be reflected downwards and into the aperture of a SPAD. b An illustration of a GHz-gated linear SPAD array with a silica waveguide chip. The entire array is collectively gated at 1 GHz and biased (Vu), whilst the circuit allows for the DC bias (Vi) of each pixel to be individually adjusted to compensate for inherent variations in waveguide loss ($\kappa$i) and detector efficiency ($\eta$i). The pitch of the waveguide array is matched to that of the SPAD array to achieve simultaneous coupling to each pixel. c The photonic circuit for the BB84 QKD experiment. A chip-based quantum transmitter (Alice), is adapted from previous work and is capable of producing three decoy-state time-bin encoded qubits at 1 GHz repetition rate and with low error Dolphin2023. An example of the optical signal is shown after the laser PIC and again after the encoding PIC, showing the three-decoy intensity levels and encoding phase $\phi$ respectively. Our quantum receiver (Bob) includes the integrated-SPAD receiver assembly with an asymmetric Mach-Zehnder interferometer. Polarisation controllers (Pol. C.) ensure polarisation alignment at each stage, a variable optical attenuator (VOA) is used to bring the flux down to secure single photon levels, and a discrete phase modulator provides random basis switching.
• Figure 2: Four-device SPAD array performance during 1 GHz gated-mode operation. a Dark count rate and afterpulsing probability (after 100 ns deadtime) against single photon detection efficiency (SPDE) as the universal array bias is varied. Inset: SPDE against universal array bias across the most relevant voltage range. b Synchronous crosstalk probability across the array for each of the three victim pixels (at successively increasing distance) as the universal array bias was increased. c Array specificity: Detected count rates for the array as each device is illuminated, with dark count rates subtracted. All four devices have their DC bias set to achieve 15% SPDE. We see specificity of at least three orders of magnitude for all pixel combinations.
• Figure 3: Plot of raw bit rate, QBER and secure bit rate against channel attenuation with detectors at -30°C. The loss of a variable optical attenuator in the quantum channel was gradually increased to emulate increasing fibre channel distance. Equivalent fibre channel distance is calculated at 0.18 dB/km. Data was also taken over a real 100 km fibre spool (stars), with a loss of 19.2 dB (elevated due to fibre connector losses). The dotted lines show the results of simulations based on measured system parameters.