[Paper Review] 16-Element Superconducting Nanowire Single-Photon Detector for Gigahertz Counting at 1550-nm

Timothy M. Rambo, Amy R. Conover+1 more|arXiv (Cornell University)|Mar 25, 2021

Analytical Chemistry and Sensors7 citations

TL;DR

This paper presents a 16-element superconducting nanowire single-photon detector (SNSPD) fabricated in a scalable commercial process, achieving 83.4% system detection efficiency at 1550 nm and enabling gigahertz counting with >50% efficiency at 1 GHz. The device uses individually wired sub-elements with a 9.6-ns mean dead time and exhibits low timing jitter (86.1 ps FWHM) and minimal crosstalk (<5% per pair), demonstrating high performance and scalability for quantum and optical applications.

ABSTRACT

We present a linearly arrayed, 16-element, superconducting nanowire single-photon detector with 83.4$\%$ system detection efficiency at 1550 nm and a mean per-element dead-time of 9.6-ns, enabling counting at 1 giga-count per second with $&gt;50\%$ System Detection Efficiency. This device was designed and fabricated in an existing scalable commercial process.

Motivation & Objective

• To develop a high-count-rate, high-efficiency SNSPD array suitable for scalable commercial production.
• To overcome the trade-off between detection efficiency and recovery time in single-element SNSPDs by using a multi-element architecture.
• To achieve >50% system detection efficiency at 1 GHz photon rates, enabling applications in quantum communications and high-speed optical sensing.
• To minimize crosstalk and maintain low timing jitter across all elements in a multi-element configuration.
• To demonstrate compatibility with fiber-coupled operation using four-mode fiber (FMF), enabling practical integration in real-world systems.

Proposed method

• The SNSPD was fabricated on a Si/SiO2 substrate with a back-short mirror and a λ/4 optical cavity to enhance 1550-nm absorption.
• A 7 nm thick proprietary superconducting film (Tc ≈ 5 K) was patterned into 16 individually wired, 16 μm × 1 μm meandered nanowires with 60 nm width and 50% fill factor.
• A two-layer dielectric coating was applied on top of the cavity to maximize absorption at 1550 nm, and deep etching created a keyhole-shaped chip for self-alignment to fiber.
• A custom dc-coupled, anti-latch circuit was used to enable single-coaxial line per element, reducing cryogenic complexity.
• System detection efficiency (SDE) and dark count rates were measured using a tunable CW laser and attenuators, with bias-dependent SDE curves acquired per element.
• Timing jitter was measured using a mode-locked fiber laser and a Time Tagger Ultra, with contributions deconvolved using the relation: $ J_{meas}^2 = J_{SNSPD}^2 + J_{TT}^2 + J_{pulse}^2 $.

Experimental results

Research questions

• RQ1 Can a 16-element SNSPD array achieve >50% system detection efficiency at 1 GHz photon rates while maintaining low timing jitter and crosstalk?
• RQ2 How does the dead time and recovery dynamics of individual sub-elements affect the overall count-rate performance in a multi-element SNSPD?
• RQ3 To what extent does fiber coupling (specifically FMF) degrade dark count performance compared to single-mode fiber?
• RQ4 Can crosstalk between elements be minimized through bias control and temporal gating, and how does it vary across element pairs?
• RQ5 Does the multi-element architecture preserve the high timing resolution and detection efficiency of single-element SNSPDs?

Key findings

• The 16-element SNSPD achieved a net system detection efficiency of 83.4% at 1550 nm, with a mean per-element dead time of 9.6 ns.
• The device maintained >50% system detection efficiency at a mean incident photon rate of 1 GHz, enabling gigahertz counting.
• The average timing jitter across all elements was 86.1 ps FWHM, with a standard deviation of 10.2 ps, indicating high consistency.
• Crosstalk was minimal, with the maximum observed increase in counts on any stop channel being 5% of the source channel’s count rate at high bias.
• The device exhibited a long plateau in detection efficiency versus bias current, which contributed to elevated dark counts (71,000/s) due to blackbody photon absorption in the FMF-coupled system.
• A single defective element showed significantly worse jitter and reduced efficiency, but did not compromise the performance of the other 15 elements, highlighting the robustness of the multi-element design.