Letter pubs.acs.org/NanoLett
Efficient Single Photon Detection from 500 nm to 5 μm Wavelength Francesco Marsili,† Francesco Bellei,† Faraz Najafi,† Andrew E. Dane,† Eric A. Dauler,‡ Richard J. Molnar,‡ and Karl K. Berggren*,† †
Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States ‡ Lincoln Laboratory, Massachusetts Institute of Technology, 244 Wood St., Lexington, Massachusetts 02420, United States S Supporting Information *
ABSTRACT: We report on superconducting nanowire single photon detectors (SNSPDs) based on 30 nm wide nanowires with detection efficiency η ∼ 2.6−5.5% in the wavelength range λ = 0.5−5 μm. We compared the sensitivity of 30 nm wide SNSPDs with the sensitivity of SNSPDs based on wider (85 and 50 nm wide) nanowires for λ = 0.5−5 μm. The detection efficiency of the detectors based on the wider nanowires became negligible at shorter wavelengths than the 30 nm wide SNSPDs. Our 30 nm wide SNSPDs showed 2 orders of magnitude higher detection efficiency (η ∼ 2%) up to longer wavelength (λ = 5 μm) than previously reported. On the basis of our simulations, we expect that by changing the optical coupling scheme and by integrating the detectors in an optical cavity, the detection efficiency of our detectors could be increased by a factor of ∼6. KEYWORDS: Superconducting nanowire single photon detectors, mid-infrared, niobium nitride (NbN), superconductivity, nanofabrication, quantum optics fficient single-photon detection at middle-infrared (midIR) wavelengths (2.5−25 μm wavelength) would open new development perspectives in a broad range of applications, such as (1) characterization of quantum light sources1,2 and of inter-subband transitions,3 (2) differential absorption light detection and ranging (DIAL) for remote environmental monitoring,4 (3) thermal5 and photoemission microscopy6 applied to failure analysis of scaled integrated circuits, (4) nulling interferometry7 applied to the detection of biomarkers in the thermal emission of earth-like planets,8,9 (5) characterization of the unidentified infrared emission of interstellar dust,10,11 and (6) free-space optical communication.12 Here we report on superconducting nanowire single photon detectors (SNSPDs)13 capable of detecting single mid-IR photons up to 5 μm wavelength with efficiency of ∼2%. Similar detectors also showed low reset time (below 10 ns14) and low jitter (∼30 ps full width at half maximum15). Since the introduction of SNSPDs,13 it has been proposed to extend the sensitivity of these detectors to the mid-IR wavelength range by decreasing the nanowire width.16 Prior to this work, the SNSPDs with the highest detection efficiency (η) in the mid-IR were based on 55 nm wide nanowires17 and showed η ∼ 0.01% at 3.5 μm wavelength (λ), which made the detectors unsuitable for most mid-IR applications. The results of ref 17 indicated that the nanowire width of SNSPDs needed to be drastically reduced to extend the sensitivity of the detectors to the mid-IR. Therefore, we fabricated SNSPDs
E
© 2012 American Chemical Society
based on ultranarrow (20 and 30 nm wide) nanowires.15 Here we report on the comparison of the sensitivity of 30 nm wide SNSPDs with the sensitivity of SNSPD based on wider (85 and 50 nm wide) nanowires from visible to mid-IR wavelengths (λ = 0.5−5 μm). The detection efficiency of the detectors based on the wider nanowires became negligible at shorter wavelengths than the 30 nm wide SNSPD. Our 30 nm wide SNSPDs showed 2 orders of magnitude higher detection efficiency (η ∼ 2%) up to longer wavelength (λ = 5 μm) than previously reported for this class of detector. Two other approaches to mid-IR single photon counting have been reported to date: frequency up-conversion and single photon avalanche diodes (SPADs) operating at visible wavelengths18 and blocked-impurity-band solid-state photomultipliers (BIB-SSPM).19 Frequency up-conversion exhibited orders of magnitude lower efficiency than our detectors (typical efficiencies were of the order of 10−5). Although BIB-SSPMs showed a detection efficiency comparable to our detectors,20 they have relaxation times 3 orders of magnitude longer (∼50 μs19) and jitter at least 1 order of magnitude higher21 than SNSPDs.15 Because many mid-IR single-photon counting applications require high detection efficiency, low reset time, and low jitter, these existing solutions are severely limiting. Received: June 14, 2012 Revised: July 29, 2012 Published: August 13, 2012 4799
dx.doi.org/10.1021/nl302245n | Nano Lett. 2012, 12, 4799−4804
Nano Letters
Letter
Figure 1. (a) η vs IB/ISW curves of a 30 nm wide SNSPD for λ = 0.5−5 μm. The black arrow indicates the direction of increasing λ. The colored arrows indicate the η vs IB/ISW curves at λ = 0.5 μm (violet), 2.7 μm (cyan), and 5 μm (light gray). The switching current of the SNSPD was ISW = 7.4 μA. (b) Normalized photon flux vs λ from the tungsten halogen lamp coupled to the silica fiber. Black curve: photon flux measured with the thermal powermeter normalized by the photon flux measured at λ = 1.3 μm. Color curves: photon flux measured by using the same SNSPD as in (a) at different bias currents. The curves were normalized by the value of the photoresponse count rate at λ = 1.3 μm measured at the highest bias current (IB = 6.9 μA). The bias currents were IB = 3.3, 3.7, 4.1, 4.5, 4.9, 5.3, 5.7, 6.1, 6.5, and 6.9 μA. The black arrow shows the direction of increasing IB. (c) Normalized photon flux vs λ from the glowbar coupled to the chalcogenide fiber. Black curve: photon flux measured with the thermal powermeter normalized by the photon flux at λ = 2.4 μm. Color curves: photon flux measured with the same SNSPD as in (a) at different bias currents. The curves were normalized by the value of the count rate at λ = 2.4 μm measured at the highest bias current (IB = 6.9 μA). The bias currents used are the same as those used in (b). In (b) and (c), the error bars on the photon flux measured with the powermeter were estimated from the noise of the power meter (see Supporting Information).
= 0.5−5 μm with a dip-probe system. To simplify the comparison of the sensitivity of nanowires with different widths, in this Letter we report on the characterization of 85, 50, and 30 nm wide SNSPDs with the same fill factor (∼30%) only. Further details of the characterization of detectors based on different architectures and nanowire widths are reported in the Supporting Information. We measured the detection efficiency of our detectors at a temperature T ∼ 1.5 K by using two different broadband incoherent light sources coupled to a monochromator: a tungsten halogen lamp for λ = 0.5−1.6 μm and a glowbar heated at 1100 K for λ = 1.6−5 μm. The light was delivered to the detectors by multimode optical fibers (a silica fiber for λ = 0.5−1.6 μm and a chalcogenide fiber for λ = 1.6−5 μm). The detector chip was illuminated from the top. The optical power coupled into the fiber was measured at 300 K with a thermal powermeter (spectral range λ = 0.2−20 μm). The details of the measurement apparatus and of the experimental procedures are described in the Supporting Information.
We fabricated SNSPDs and superconducting nanowire avalanche photodetectors (SNAPs15,22) on 5.5 nm thick NbN films deposited on sapphire substrates (the details of the fabrication process are reported in refs 15, 23, and 24). The superconducting critical temperature of the NbN films was TC ∼ 10 K (measured at the midpoint of the transition). The nanowire width of the detectors was varied from 20 to 100 nm. For the detectors based on ultranarrow (20 and 30 nm wide) nanowires, the pitch of the nanowires was 100 nm and the active area (defined as in ref 15) was varied from 1.22 μm × 220 nm to 1.33 μm × 1.33 μm. For the detectors based on 50, 85, and 100 nm wide nanowires, the pitch of the nanowires was varied from 150 to 400 nm and the active area was varied from 2.97 μm × 2.49 μm to 3.04 μm × 2.94 μm. We screened the switching current (ISW, the highest bias current a device could sustain before switching to the normal state25) and the detection efficiency at λ = 1550 nm of ∼400 devices from four fabrication runs with a cryogenic probe station15 and characterized ∼30 of these detectors in the wavelength range λ 4800
dx.doi.org/10.1021/nl302245n | Nano Lett. 2012, 12, 4799−4804
Nano Letters
Letter
Figure 2. η (in color scale) vs λ and normalized bias current (IB/ISW) for SNSPDs based on (a) 85 nm wide, (b) 50 nm wide, and (c) 30 nm wide nanowires. Each pixel of the color map corresponds to an experimental data point. The switching currents and the active areas (Ad) of the devices were ISW = 20.6 μA and Ad = 2.97 μm × 2.49 μm for the 85 nm wide SNSPD; ISW = 9.3 μA and Ad = 3.04 μm × 2.94 μm for the 50 nm wide SNSPD; ISW = 7.4 μA and Ad = 1.47 μm × 0.23 μm for the 30 nm wide SNSPD. The red dashed lines are the Ico vs λ curves of the three detectors. The blue pixels for high bias current and λ < 2 μm in (b) and (c) are due to the detectors switching to the normal state. The bias and wavelength dependence of the detection efficiency of a 20 nm wide SNAP with two sections in parallel (2-SNAP) is reported in the Supporting Information.
below 2% beyond λ = 2.6 μm for the 85 nm wide SNSPD and beyond λ = 3.2 μm for the 50 nm wide SNSPD. The Ico vs λ curves for the three devices are plotted in red. To estimate the sensitivity of our detectors, we determined the dependence of the dark count rate (DCR) on IB. To fairly compare the sensitivity at different wavelengths for SNSPDs with different nanowire widths and active areas, we introduced the wavelength-dependent equivalent dark count rate: DCReq(λ). The DCReq(λ) of a detector was defined as the dark count rate of an equivalent detector with the same nanowire width and fill factor whose active area is as large as the diffraction-limited optical spot at the operating wavelength λ (necessary for efficient optical coupling, assuming that subwavelength light concentrators26 are not employed): DCReq(λ) = dcr·λ2, where dcr (which we called normalized dark count rate) is the intrinsic DCR of the detector divided by the total length of nanowire inside the active area (L) and where we assumed L ∝ λ2 because the length of the nanowire is proportional to the active area of the detector.27 The definition of DCReq was based on the following assumptions: (1) it is possible to fabricate detectors with an active area which would allow efficient optical coupling at the wavelength of interest, (2) the absorptance of the detector does not depend on the active area, and (3) the dark counts are nucleated by the transit of a single magnetic vortex across the nanowire.28 The third assumption implies that a dark count is equally likely to nucleate anywhere along the length of the nanowire, so the dark count attempt rate scales with the length of the nanowire. The figure of merit for the sensitivity of single photon detectors depends on the specific application. For applications in which the mean value of the DCR cannot be compensated for (e.g., quantum key distribution29), the value of η at a certain value of DCR is the relevant figure of merit. For applications in
Figure 1 a shows the detection efficiency of a 30 nm wide SNSPD as a function of the bias current (IB) for λ = 0.5−5 μm. In the range λ = 0.5−2.7 μm (see colored arrows in Figure 1a), η showed a sigmoidal dependence on IB and saturated to η ∼ 4.5−5.5% (depending on the wavelength) for IB sufficiently higher than the cutoff current (Ico, taken to be at the inflection point of the η vs IB curve as in ref 15). As the wavelength was increased, Ico shifted to higher currents, and it increased from 0.3ISW at λ = 0.5 μm to 0.6ISW at λ = 2.7 μm. Although for λ > 3 μm only the initial rise of the sigmoid was observed, the detection efficiency at λ = 5 μm was as high as η = 2.6% at IB = 0.94ISW. Figure 1b,c shows the wavelength dependence of the normalized photon flux from the tungsten halogen lamp coupled to the silica fiber (Figure 1b) and from the glowbar coupled to the chalcogenide fiber (in Figure 1c). We measured the photon flux using the powermeter (black curve) and the 30 nm wide SNSPD biased at different currents (colored curves). For λ < 3 μm and for IB higher than Ico, the shape of the photon spectra measured with the SNSPD became independent of the bias current and similar to the shape of the spectra measured with the power meter. We measured similar photon spectra using several 30 nm nanowire width detectors (see Supporting Information). The dips in the photon spectra around λ = 2.9 μm and λ = 4.1 μm were due to the absorption spectrum of the chalcogenide fiber. Figure 2 shows the bias and wavelength dependence of the detection efficiency of SNSPDs based on nanowires of different widths (w) and the same fill factor (∼30%): w = 85 nm (Figure 2a); w = 50 nm (Figure 2b); w = 30 nm (Figure 2c). The detection efficiency of the detectors based on the wider nanowires became negligible at shorter wavelengths than the 30 nm wide SNSPD. Indeed, the detection efficiency dropped 4801
dx.doi.org/10.1021/nl302245n | Nano Lett. 2012, 12, 4799−4804
Nano Letters
Letter
Figure 3. (a) Normalized dark count rate (dcr) as a function of normalized bias current (IB/ISW) of a 30 nm wide (red), a 50 nm wide (green), and a 85 nm wide (blue) SNSPD. The nanowire lengths were L = 5.1 μm for the 30 nm wide SNSPD, L = 61.8 μm for the 50 nm wide SNSPD, and L = 26.8 μm for the 85 nm wide SNSPD. The vertical dashed lines indicate the bias currents at which the devices showed dcr ∼3 cps/μm. (b) η vs λ of a 30 nm wide SNSPD biased at IB = 0.81ISW (red), a 50 nm wide SNSPD biased at IB = 0.91ISW (green), and a 85 nm wide SNSPD biased at IB = 0.95ISW (blue). The error bars on the value of the detection efficiency are calculated propagating the uncertainty on the value of the photon flux due to the precision of the power meter (see Supporting Information).
to the standard deviation of DCReq: NEPF = λ·dcr0.5/η, where η is the detection efficiency of the device illuminated with unpolarized light. We adopted the NEPF to quantify the sensitivity of single photon detectors rather than the noise equivalent power as proposed in ref 16 because the relevant quantity in most single-photon counting applications31 is the photon f lux, rather than the optical power. Figure 4a−c shows the bias and wavelength dependence of the NEPF of the 85 nm wide (Figure 4a), the 50 nm wide (Figure 4b), and the 30 nm wide SNSPD (Figure 4c). Figure 4d shows the NEPF of the three devices biased at a current that yields dcr ∼ 3 cps/μm (see Figure 3a and the dashed lines in Figure 4a−c). Although the 85 nm wide and 50 nm wide SNSPDs were more sensitive than the 30 nm wide SNSPDs at visible and near-infrared wavelengths, the sensitivity of the 30 nm wide SNSPDs was higher than the sensitivity of wider nanowires beyond 3 μm. Our results indicate that decreasing the nanowire width below 50 nm allows extending the sensitivity of SNSPDs to mid-IR wavelengths up to λ = 5 μm. The detection efficiency of the 30 nm wide SNSPDs reported here is lower than the detection efficiency of similar devices reported in ref 15 due to the following reasons: (1) we measured the detector detection efficiency by coupling the light from the front of the sample, while in ref 15 we were illuminating from the back (through the substrate); (2) the polarization of the light was random (see Supporting Information), while in ref 15 it was controlled to
which the average value of the DCR can instead be subtracted, the sensitivity of a single photon detector is limited by the statistical variation of the DCR, which is proportional to (DCR)0.5 (assuming that the dark counts have Poissonian probability distribution). Therefore, we compared the sensitivity of our SNSPDs by using two different figures of merit. Figure 3a shows the bias dependence of the normalized dark count rate (dcr) for the 85 nm wide, 50 nm wide, and 30 nm wide SNSPD. We measured the intrinsic DCR of the detectors by blocking the optical coupling between the optical fibers and the devices with a metal shutter kept at T ∼ 1.5 K. When the silica or the chalcogenide fibers were coupled to the devices, the DCR of the detectors increased by up to 2 orders of magnitude (see Supporting Information), presumably because stray photons due to background illumination and to blackbody radiation at 300 K were coupled into the fibers.30 Figure 3b shows the η vs λ curves for the three devices biased at a current that yields dcr ∼ 3 cps/μm (see dashed lines in Figure 3a). The η vs λ curves of the three SNSPDs represent the figure of merit for the sensitivity of the detectors for the first class of applications. For the second class of applications, we quantified the sensitivity at different wavelengths of our SNSPDs by introducing the concept of noise equivalent photon flux (NEPF), which we defined as the photon flux that makes the photoresponse count rate of a detector (PCR = NEPF·η) equal 4802
dx.doi.org/10.1021/nl302245n | Nano Lett. 2012, 12, 4799−4804
Nano Letters
Letter
Figure 4. (a, b, c) NEPF (in logarithmic color scale) vs IB/ISW and λ for SNSPDs based on (a) 85 nm wide, (b) 50 nm wide, and (c) 30 nm wide nanowires. Each pixel of the color map corresponds to an experimental data point. White represents NEPF > 2 × 104 (nm·cps)0.5, while black represents η = 0. The black dashed lines correspond to a bias current that yields dcr ∼ 3 cps/μm (see Figure 3a). (d) NEPF vs λ of the 30 nm wide SNSPD biased at IB = 0.81ISW (red), the 50 nm wide SNSPD biased at IB = 0.91ISW (green), and the 85 nm wide SNSPD biased at IB = 0.95ISW (blue).
fibers; (5) estimation of the uncertainty on the value of the detection efficiency; (6) polarization of the light; (7) photon spectra measured with 30 nm wide SNSPDs and SNAPs; (8) detection efficiency of a 20 nm wide 2-SNAP; (9) dark count rate measurements. This material is available free of charge via the Internet at http://pubs.acs.org.
maximize the detection efficiency (which depends on polarization32). On the basis of the results reported in ref 15 and our optical simulations,32 it should be possible to achieve an increase in the detection efficiency by a factor of ∼3 by controlling the polarization of the light and by an additional factor of ∼2 by illuminating from the back of the sample. Finally, although the active area of our detectors does not allow efficient optical coupling at mid-infrared wavelengths, it should be possible to increase the active area while keeping the reset time below 10 ns by combining three approaches: (1) increasing the value of the load resistor in the read out of the detector;33 (2) integrating nanowires with metallic nanoantennas,26 which would allow the pitch of the nanowires to be increased without loss in absorptance; and (3) adopting a modified N-SNAP (i.e., a SNAP with N sections in parallel) architecture34 which, unlike conventional SNAPs,15,22 does not require series inductors, whose reset time would be a factor of N2 lower than SNSPDs with the same active area.
■
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors thank H. Byun, J. Daley, M. Mondol Prof. R. Ram, and V. Manfrinato for technical support and Dr. V. Anant, Dr. S. W. Nam, Dr. R. Mirin, Dr. M. Csete, and Prof. D. E. Prober for scientific discussions. The initial phases of the work at the MIT campus were sponsored by the National Science Foundation under NSF Award ECCS-0823778. The later phases were supported by the Center for Excitonics under Award DE-SC0001088. This work was completed while Prof. K. K. Berggren was on sabbatical at Delft University of Technology and supported by The Netherlands Organization
ASSOCIATED CONTENT
S Supporting Information *
(1) Complete list of characterized devices; (2) experimental setup; (3) estimation of the device detection efficiency; (4) wavelength dependence of the numerical aperture of the optical 4803
dx.doi.org/10.1021/nl302245n | Nano Lett. 2012, 12, 4799−4804
Nano Letters
Letter
(30) Miller, A. J.; Lita, A. E.; Rosenberg, D.; et al. Presented at the Proceedings of the 8th International Conference on Quantum Communication, Measurement and Computing, 2007 (unpublished). (31) Eisaman, M. D.; Fan, J.; Migdall, A.; et al. Rev. Sci. Instrum. 2011, 82 (7), 071101−071125. (32) Anant, V.; Kerman, A. J.; Dauler, E. A.; et al. Opt. Express 2008, 16 (14), 10750−10761. (33) Yang, J. K. W.; Kerman, A. J.; Dauler, E. A.; et al. IEEE Trans. Appl. Supercond. 2007, 17 (2), 581−585. (34) Ejrnaes, M.; Casaburi, A.; Quaranta, O.; et al. Supercond. Sci. Technol. 2009, 22 (5), 055006.
for Scientific Research. Additional support from DARPA is acknowledged. The work at MIT Lincoln Laboratory was sponsored by the Assistant Secretary of Defense for Research and Engineering under Air Force Contract #FA8721-05-C0002. Opinions, interpretations, recommendations, and conclusions are those of the authors and are not necessarily endorsed by the United States Government.
■
REFERENCES
(1) Elvira, D.; Michon, A.; Fain, B.; et al. Appl. Phys. Lett. 2010, 97 (13), 131907−131903. (2) Kaspi, R.; Ongstad, A. P.; Dente, G. C.; et al. Appl. Phys. Lett. 2006, 88 (4), 041122−041123. (3) Schneider, H.; Maier, T.; Walther, M.; et al. Appl. Phys. Lett. 2007, 91 (19), 191116−191113. (4) Weibring, P.; Edner, H.; Svanberg, S. Appl. Opt. 2003, 42 (18), 3583−3594. (5) Ippolito, S. B.; Thorne, S. A.; Eraslan, M. G.; et al. Appl. Phys. Lett. 2004, 84 (22), 4529−4531. (6) Tosi, A.; Stellari, F.; Pigozzi, A.; et al. Presented at the Proceedings of the 44th Annual IEEE International Reliability Physics Symposium, 2006 (unpublished). (7) Wallace, K.; Hardy, G.; Serabyn, E. Nature 2000, 406 (6797), 700−702. (8) Labadie, L.; Kern, P.; Labeye, P.; et al. Adv. Space Res. 2008, 41 (12), 1975−1982. (9) Perryman, A. C. Detection and Characterization of Extra-Solar Planets: Future Space Missions; Springer: New York, 2007. (10) Kim, H. S.; Wagner, D. R.; Saykally, R. J. Phys. Rev. Lett. 2001, 86 (25), 5691−5694. (11) Cook, D. J.; Schlemmer, S.; Balucani, N.; et al. Nature 1996, 380, 227−229. (12) Martini, R.; Bethea, C.; Capasso, F.; et al. Electron. Lett. 2002, 38 (4), 181−183. (13) Gol’tsman, G. N.; Okunev, O.; Chulkova, G.; et al. Appl. Phys. Lett. 2001, 79 (6), 705−707. (14) Marsili, F.; Najafi, F.; Dauler, E.; et al. Appl. Phys. Lett. 2012, 100 (11), 112601−112605. (15) Marsili, F.; Najafi, F.; Dauler, E.; et al. Nano Lett. 2011, 11 (5), 2048−2053. (16) Semenov, A.; Engel, A.; Il’in, K.; et al. Eur. Phys. J. 2003, 21 (3), 171−178. (17) Korneeva, Y.; Florya, I.; Semenov, A.; et al. IEEE Trans. Appl. Supercond. 2010, 99, 1−1. (18) Temporão, G.; Tanzilli, S.; Zbinden, H.; et al. . Opt. Lett. 2006, 31 (8), 1094−1096. (19) Petroff, M. D.; Stapelbroek, M. G.; Kleinhans, W. A. Appl. Phys. Lett. 1987, 51 (6), 406−408. (20) Kim, H.-S.; Saykally, R. J. Rev. Sci. Instrum. 2003, 74 (4), 2488− 2494. (21) Baek, B.; McKay, K. S.; Stevens, M. J.; et al. IEEE J. Quantum Electron. 2010, 46 (6), 991−995. (22) Ejrnaes, M.; Cristiano, R.; Quaranta, O.; et al. Appl. Phys. Lett. 2007, 91 (26), 262509. (23) Yang, J. K. W.; Dauler, E.; Ferri, A.; et al., IEEE Trans. Appl. Supercond. 15 (2 PART I), 626−630 (2005). (24) Kerman, A. J.; Dauler, E. A.; Yang, J. K. W.; et al. Appl. Phys. Lett. 2007, 90 (10), 101110. (25) The switching current of a detector is usually lower than the critical current due to constrictions; see ref 24. (26) Xiaolong, H.; Dauler, E. A.; Molnar, R. J.; et al. Opt. Express 2011, 19 (1), 17−31. (27) We considered the case of front illumination for a single-pass structure immersed in liquid He. (28) Bulaevskii, L. N.; Graf, M. J.; Batista, C. D.; et al. Phys. Rev. B 2011, 83 (14), 144526. (29) Gisin, N.; Thew, R. Nat. Photonics 2007, 1 (3), 165−171. 4804
dx.doi.org/10.1021/nl302245n | Nano Lett. 2012, 12, 4799−4804
