Short answer: Superconducting nanowire single-photon detectors, or SNSPDs, count single infrared photons, and the system detection efficiency is the fraction detected. A 2018 paper reports 90.2% for 1550 nm photons at 2.1 K with NbN, and a 2020 arXiv preprint reports 98% at 0.8 K and 95% at 2.1 K using twin-layer nanowires on a dielectric mirror. A 2019 paper studies how intrinsic efficiency depends on wavelength. These are laboratory measurements by the authors, and the record claims are theirs.

Why efficiency matters

The 2020 preprint calls an SNSPD with near-unity system efficiency a key enabling but still elusive technology for many quantum fundamental theory verifications and quantum information applications. The 2018 abstract says the rapid development of SNSPDs over the past decade has led to numerous advances in quantum information technology.

2018: NbN at cryocooler temperature

The 2018 abstract notes that the record system detection efficiency at 1550 nm then stood at 93%, reached with a detector made of amorphous WSi, which is usually operated at sub-Kelvin temperatures. The authors demonstrate detectors made of polycrystalline NbN with a system detection efficiency of 90.2% for 1550 nm photons at 2.1 K, a temperature reachable with a compact cryocooler. The efficiency saturated at 92.1% when they lowered the temperature to 1.8 K. They say they expect this to make practical, high-performance detectors available to quantum information and other high-end uses.

2020: twin-layer nanowires

The arXiv preprint, which is not peer reviewed in the version read here, says the key challenge is to have both a near-unity photon-response probability and absorption efficiency for a meandered nanowire with a finite filling ratio. This is more crucial for NbN than for materials like WSi with lower transition temperatures. The authors replace a single-layer nanowire with twin-layer nanowires on a dielectric mirror and report NbN detectors with what they call a record system efficiency. At 0.8 K the maximum system detection efficiency is 98% at 1590 nm, with over 95% across 1530 to 1630 nm. At 2.1 K, using a compact two-stage cryocooler, it is 95% at 1550 nm. The remainder of the abstract, on yield, was cut off here and is not described.

2019: wavelength dependence

A 2019 paper studies the intrinsic detection efficiency of NbN detectors versus wavelength. The authors fabricate detectors with linewidths from 30 nm to 140 nm and obtain efficiency curves against bias current for wavelengths from 510 to 1700 nm. They report clearly nonlinear energy-current relations, which they say indicates that a detection model considering only quasiparticle diffusion is unsuitable for meander-type NbN detectors. They describe the work as additional experimental data on the detection mechanism.

Limits

Efficiency values depend on wavelength, temperature and device design, and they come from different groups. The 98% result needs 0.8 K and the cryocooler result is 95%. The record statements are the authors' claims, and the preprint abstract was not fully readable. The abstracts do not report dark counts or timing jitter for the highest efficiencies, so those cannot be assumed.

FAQ

What is system detection efficiency?

The fraction of incident photons that the whole detector system registers.

How high has it been reported?

98% at 1590 nm and 0.8 K in a 2020 preprint, and 90.2% at 2.1 K in a 2018 paper.

Why does temperature matter?

Lower temperature raised efficiency from 90.2% to 92.1% in the 2018 NbN detector.

Is the detection mechanism fully understood?

The 2019 authors say a quasiparticle-diffusion-only model is unsuitable for meander NbN detectors.

Are these commercial products?

The abstracts describe laboratory devices.

All field notes