We have fabricated and tested superconducting single-photon detectors and demonstrated detection efficiencies of 57% at 1550-nm wavelength and 67% at 1064 nm. In addition to the peak detection efficiency, a median detection efficiency of 47.7% was measured over 132 devices at 1550 nm. These measurements were made at 1.8K, with each device biased to 97.5% of its critical current. The high detection efficiencies resulted from the addition of an optical cavity and anti-reflection coating to a nanowire photodetector, creating an integrated nanoelectrophotonic device with enhanced performance relative to the original device. Here, the testing apparatus and the fabrication process are presented. The detection efficiency of devices before and after the addition of optical elements is also reported.

No abstract

We investigate the source of large variations in the observed detection efficiencies of superconducting nanowire single-photon detectors between many nominally identical devices. Through both electrical and optical measurements, we infer that these variations arise from "constrictions:" highly localized regions of the nanowires where the effective cross-sectional area for superconducting current is reduced. These constrictions limit the bias current density to well below its critical value over the remainder of the wire, and thus prevent the detection efficiency from reaching the high values that occur in these devices only when they are biased near the critical current density. PACS numbers: 74.76.Db, Superconducting nanowire single-photon detectors (SNSPDs) [1,2,3,4] provide a unique combination of high infrared detection efficiency (up to 57% at 1550nm [2] has been demonstrated) and high speed (<30 ps timing resolution [3,5], and few-ns reset times after a detection event [4]). Applications for these devices already being pursued include high data-rate interplanetary optical communications [6], spectroscopy of ultrafast quantum phenomena in biological and solid-state physics [7,8], quantum key distribution (QKD) [9], and noninvasive, high-speed digital circuit testing [10].In many of these applications, large arrays of SNSPDs would be extremely important [5]. For example, existing SNSPDs have very small active areas, making optical coupling relatively difficult and inefficient [7,11]. Their small size also limits the number of optical modes they can collect, which is critical in free-space applications where photons are distributed over many modes, such as laser communication through the atmosphere (where turbulence distorts the optical wavefront) and in fluorescence detection. Furthermore, it was shown in Ref.[4] that the maximum count rate for an individual SNSPD decreases as its active area is increased, due to its kinetic inductance, forcing a tradeoff between active area and high count rates. Count rate limitations are particularly important in optical communications and QKD, affecting the achievable receiver sensitivity or data rate [6,12]. Detector arrays could provide a solution to these problems, giving larger active areas while simultaneously increasing the maximum count rate by distributing the flux over many smaller (and therefore faster) pixels. Large arrays could also provide spatial and photon-number resolution. Although few-pixel detectors have been demonstrated [5,9,11,12], fabrication and readout methods scalable to large arrays have not yet been discussed.A first step towards producing large arrays of SNSPDs is to understand (and reduce) the large observed variation of detection efficiencies for nominally identical devices [2], which would set a crippling limit on the yield of efficient arrays of any technologically interesting size.In this Letter, we demonstrate that these detection efficiency variations can be understood in terms of what we call "constrictions:" highly localized, essential...

We measured the optical absorptance of superconducting nanowire single photon detectors. We found that 200-nm-pitch, 50%-fill-factor devices had an average absorptance of 21% for normally-incident front-illumination of 1.55-microm-wavelength light polarized parallel to the nanowires, and only 10% for perpendicularly-polarized light. We also measured devices with lower fill-factors and narrower wires that were five times more sensitive to parallel-polarized photons than perpendicular-polarized photons. We developed a numerical model that predicts the absorptance of our structures. We also used our measurements, coupled with measurements of device detection efficiencies, to determine the probability of photon detection after an absorption event. We found that, remarkably, absorbed parallel-polarized photons were more likely to result in detection events than perpendicular-polarized photons, and we present a hypothesis that qualitatively explains this result. Finally, we also determined the enhancement of device detection efficiency and absorptance due to the inclusion of an integrated optical cavity over a range of wavelengths (700-1700 nm) on a number of devices, and found good agreement with our numerical model.

No abstract