Single-photon and photon-number-resolving detectors based on superconducting nanowires

Nanowire superconducting single photon detectors (SSPDs) [1] are characterized by very high sensitivity in the near infrared (detection efficiency η up to 30%, for a dark count rate DK of few Hz), speed (up to ∼1 GHz repetition rate) and time resolution (jitter of 20 ps full width at half maximum, FWHM). They can be operated at temperatures near 4 K, so they can be packaged in cryogenic dipsticks or cryogen-free refrigerators. These features make SSPDs the most promising detectors for telecom-wavelength single-photon counting applications. The basic structure of an SSPD is a narrow (w=50 to 120 nm), thin (th∼4-10 nm) NbN superconducting nanowire folded in a meander pattern. The typical detector active area (i.e. the size of the pixel) is Ad=10 × 10 µm2 (which allows an efficient coupling with the core of optical fibers at telecom wavelengths) with filling factor (f, the ratio of the area occupied by the superconducting meander to the device total area) ranging from 40% to 60%. The meanders are embedded in a 50 Ω coplanar transmission line. At present, the SSPD detection efficiency is limited by its absorbance (α, the ratio of the number of photons absorbed in the nanowire to the number of incident photons on the device active area). Indeed, it has been shown that in the classic front-illumination configuration α cannot exceed 30%. Our approach to increase α consists in integrating SSPDs with advanced optical structures such as distributed Bragg reflectors (DBRs) and optical waveguides. This requires to transfer the challenging SSPD technology (i.e. the deposition of high-quality few-nm thick NbN films and the nano-patterning by electron beam lithography, EBL) from the usual comfortable substrates, i.e. sapphire and MgO, which are known to allow the deposition of few-nm thick NbN films of excellent quality, to an optical substrate like GaAs, on which DBRs and waveguides can be easily obtained. Our first task was then to optimize a process for the deposition of high-quality few-nm thick NbN films on GaAs and AlAs/GaAs-based DBRs. Because of the requirement of compatibility with GaAs, the substrate temperature used for the depositions is 400°C, in order to prevent As evaporation. As GaAs and DBRs are highly mismatched substrates, the deposition parameters were first optimized with respect to the superconducting properties of NbN films on MgO substrates, which allow the growth of high crystal quality NbN films at low temperature. This made easier to separate the influence of stoichiometry from that of microstructure. The optimized deposition parameters were then used to grow NbN films on GaAs and DBRs, under the reasonable assumption (later checked and confirmed) that changing the substrate would not produce a change in film stoichiometry, but only in its microstructure. NbN films ranging from 150nm to 3nm in thickness were then deposited on epitaxial-quality single crystal MgO, GaAs and DBRs structures. The deposition technique is the current controlled DC magnetron sputtering (planar, circular, balanced configuration) of Nb in an Ar + N2 plasma. NbN films deposited on MgO exhibit superconducting critical temperature ΤC=10 Κ, superconducting transition width ΔΤC=0.8 Κ and residual resistivity ratio RRR=R(20K)/R(300K)=0.8 for th=4 nm, which are state of the art values, proof of the excellent quality of our low-temperature deposition process. The quality of films deposited on GaAs and on DBRs is lower than that of NbN deposited on MgO, as for any thickness they systematically exhibit higher ΔΤC and lower ΤC and RRR. However, 5.5 nm-thick NbN films on GaAs still exhibit ΤC=10.7 Κ, ΔΤC=1.1 Κ and RRR=0.7, which compares with 4.5 nm thick films on MgO, making them suitable for device fabrication. To our knowledge, the growth of such high quality thin NbN films on GaAs and DBRs, has never been reported in literature. The degradation of the superconducting properties exhibited by NbN films on GaAs and DBRs was attributed to a highly defected microstructure, due both to a higher lattice misfit between NbN and GaAs and to a poorer quality of the substrate surface. Encouraging preliminary results show that the quality of these films can be improved either cleaning the GaAs/DBR substrate surface more effectively or adding an MgO buffer layer. SSPDs were fabricated on thin NbN films (th=3-7 nm) deposited under optimal conditions on MgO and GaAs by EBL and reactive ion etching. The geometrical parameters of our detectors are: Ad=5×5 µm2, w=60-200 nm, f=40%-60%. The devices were then characterized both electrically and optically. I-V curves of test structures were measured, from which it was possible to deduce important physical parameters used as figures of merit to estimate the superconducting properties of the nanowires, or for the design and the simulation of the devices. The quality of the devices fabricated on GaAs is poorer than those on MgO, most likely due to the lower quality of NbN films deposited on GaAs and to issues related to the EBL nano-patterning step. Measurements of η and of DK as a function of the bias current were performed on SSPDs fabricated on MgO and GaAs. The best performance was exhibited by a w=100 nm, f=40%, th=4 nm meander, showing η=20% and noise equivalent power NEP=10-16 W/Hz1/2 (at λ=1.3 µm and T=4.2 K), which are state of the art values. This result showed for the first time that high performance NbN SSPDs can be realized on a different substrate and from a deposition process at lower temperature than previously reported. High detection efficiencies could not be measured with SSPDs fabricated on GaAs, but it should be noted that at present only first-generation devices (fabricated on GaAs substrates of poor surface quality) have been tested. Better results are expected from devices fabricated on the improved NbN films grown on clean or MgO-buffered GaAs substrates. Although SSPDs on MgO have shown high detection efficiency, the fabrication yield of high performance detectors has to be improved. Variations of the critical current along a nanowire are responsible for the wide distribution in efficiency values of nominally identical SSPDs. In order to understand the physical origin of the nanowire constrictions (i.e. regions of suppressed superconductivity) we performed a spatially-resolved characterization of η of a long straight nanowire, followed by a high resolution SEM (scanning electron microscope) scan on its whole length. Two types of inhomogeneities were evidenced, corresponding to localized efficiency dips and peaks. The peaks likely correspond to constrictions. SEM observations did not evidence any width narrowing at the position of the efficiency peaks, which suggests that constrictions might be due to thickness or quality inhomogeneities of the film occurring during the film deposition or later in the process. On the other hand, the efficiency dips have been correlated with lithography problems discovered on SEM images. Finally, a new photon number resolving detector, the Parallel Nanowire Detector (PND), has been demonstrated, which significantly outperforms existing approaches in terms of sensitivity, speed and multiplication noise in the telecommunication wavelength range. In particular, it provides a repetition rate (80 MHz) three orders of magnitude larger than any existing detector at telecom wavelength, and a sensitivity (NEP=4.2×10-18 W/Hz1/2) one-two orders of magnitude better, with the exception of transition-edge sensors (which require a much lower operating temperature). An electrical equivalent model of the device was developed in order to study its operation. The modeling predicts a physical limit to the reset time of the PND, which is lower than initially estimated. Furthermore, the figures of merit of the device performance in terms of efficiency, speed and sensitivity were defined and their dependency on the design parameters analyzed. Additionally, we developed modeling tools to fully characterize the device and an algorithm to estimate the photon number statistics of an unknown light using the PND. The reconstruction proved to be successful only for low photon fluxes, most likely due to the limited counting capability and the poor calibration of the detector. The PND, with its high repetition rate and high sensitivity, is then suitable for measuring an unknown photon number probability distribution assuming accurate calibration and sufficient counting capability. ______________________________[1] G. N. Gol'tsman, O. Okunev, G. Chulkova, A. Lipatov, A. Semenov, K. Smirnov, B. Voronov, A. Dzardanov, C. Williams, and R. Sobolewski, Appl. Phys. Lett. 79, 705 (2001).