NbN superconducting nanowire single photon detector with efficiency over 90% at 1550 nm wavelength operational at compact cryocooler temperature

Transcription

1 Supplementary Information NbN superconducting nanowire single photon detector with efficiency over 90% at 1550 nm wavelength operational at compact cryocooler temperature W. J. Zhang, L. X. You *, H. Li, J. Huang, C. L. Lv, L. Zhang, X. Y. Liu, J. J. Wu, Z. Wang, and X. M. Xie State Key Lab of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology (SIMIT), Chinese Academy of Sciences (CAS), Shanghai, , P. R. China CAS Center for Excellence in Superconducting Electronics (CENSE) 865 Changning Rd., Shanghai, , P. R. China. * lxyou@mail.sim.ac.cn

2 Influence of filling factors for a 7-nm devices Supplementary Figure. 1 Dependence of SDE (solid dots) on the pitch between the devices. The error bars indicated the relative uncertainty of SDEs. The red dashed line is the numerically simulated absorptance including the assumed 7% absorption loss. As shown in the supplementary Fig. 1, the measured maximum SDEs as a function of the pitch for the devices with a nominal 7-nm thickness and a nominal 75-nm width were plotted using solid dots with error bar. The highest SDE of 92.1% was obtained for the 140-nm-ptich device measured in the diluted cryostat. In order to analyze the experimental result, a numerical simulation of the optical absorptance was adopted, which was plotted with a red-dashed line. The simulated curve followed the experimental pitch dependence of SDE, with the assumed 7% absorption loss as the sole fitting parameter. From the experimental results and simulation analysis, it was found that a device with a 7-nm-thick and a filling factor of 40%-60% can achieve a >85% SDE, due to the enhancement of the absorptance by the DBR structure.

3 Surface morphology of the NbN films Supplementary Figure 2 (Color online) Atomic force microscopy (AFM) images of the DBR substrate and the 8.0-nm NbN film deposited onto DBR. The roughness of the DBR substrate and NbN film were approximately 0.2 and 0.3 nm, respectively. To characterize the uniformity of the film thickness, we conducted high-resolution atomic force microscopy (AFM) measurements. The surface morphology of the DBR substrate and an 8.0-nm NbN thin film deposited onto DBR are presented in Supplementary Fig. 2. From the AFM image, RMS (root mean square) surface roughness of 0.2 and 0.3 nm were determined. The roughness of the NbN film was slightly larger than that of the DBR because of the lattice mismatch between them. The measurements revealed that thickness variation of the NbN film was less than a few percent. According to the data above, we compared our AFM results with the literature: In the paper reported by Slysz et al. Acta Physica Polonica A 120, 200 (2011), a 0.13-nm rms roughness was obtained for a 9-nm NbN film deposited on the sapphire substrate, and a 0.63-nm rms roughness was obtained for an 18-nm NbN film deposited on the SiO 2 substrate, within a 1 1 μm 2 scan size. In the paper reported by Schuck et al. IEEE Trans. Appl. Supercond. 23, (2013), a 0.16-nm rms roughness was obtained for a 3.5-nm NbN film deposited on the SOI substrate, within a nm 2 scan size. In this manuscript, a 0.3-nm rms roughness was obtained for an 8-nm NbN film deposited on the DBR substrate (SiO 2 on the top), within a 5 5 μm 2 scan size. Discussion: 1). Firstly, the rms roughness is strongly dependent on the substrate deposited. An epitaxial NbN film on sapphire usually has a smaller roughness than the polycrystalline NbN film on SOI and SiO 2 substrates. Secondly, XRD measurements indicate that the thicker films are more crystalline and have larger grains than the thinner films do. Thus, it is expected that the thicker film would have a larger roughness. Finally, our data were obtained in a relatively large scan size which was 25 (100) times larger than the case 1(2),

4 resulting a larger roughness. Notably, roughness in a large scanning area is more appropriate to evaluate the surface morphology since the actual area of the nanowire is on the scale of a hundred μm 2. 2). The thickness variations could result in inhomogeneity in nanowire (similar to the linewidth variation) which limits the switching current and the intrinsic detection efficiency, then limit the performance of the detector. In our case, the rms roughness of 0.3 nm is 3.8% in proportion for an 8-nm thick NbN film, which is smaller than the linewidth variation (± 5 nm, 6% in proportion) of the nanowire limited by our fabrication accuracy. As a result, the thickness variation of 0.3 nm can be neglected.

5 Measurement of reflectance of NbN films on DBR substrates Supplementary Fig. 3. (Color online) Wavelength dependence of the reflectance of a DBR wafer (R DBR ) and 7- and 8-nm-thick NbN films deposited onto DBR wafers (R NbN ), as measured at room temperature by a spectrophotometer. At an incident photon wavelength 1550 nm, the R DBR and R NbN for 7(8) nm were 99.9% and 0.77(2.24)%, respectively. The inset shows a zoomed-in region around the R NbN minima of the 7(8)-nm films. Because the absorption A can be expressed as A = 1 Tr R and the transmission Tr is negligible (approximately 0.1%) for the high-reflectance mirror, the reflectance R can be directly measured using a spectrophotometer; A can then be deduced as A 1 R. Supplementary Fig. 3 shows the measured reflectance as a function of wavelengths from 1200 to 1900 nm for the with 7(8)-nm-thick NbN films deposited onto DBR mirrors. The reflectance minima, which appear at approximately nm, imply high absorption of the NbN films in this wavelength range. The measured reflectance data show that near-unity absorption in NbN films can be attained when the film thickness on the DBR mirror is approximately 7 nm. With increasing thickness, the reflectance increases. Notably, however, the absorption of a thin film (A film ) can differ from the A device, and the A film defines the upper bound for A device.

6 RT measurements of the devices with various thicknesses Supplementary Figure 4. (Color online) Normalized resistances of 10 devices with NbN thicknesses of 6.5, 7.0, 7.5, and 8.0 nm as functions of temperature. The resistance R was normalized to the resistance value at 10 K (R 10K ). The T c s (defined by 0.5R/R 10K ) of 6.5, 7.0, 7.5, and 8 nm devices were 7.85, 8.07, 8.26, and 8.59 K, respectively, with an error of approximately ±0.05 K. Supplementary Fig. 4 shows the RT measurements for 10 devices with different nominal thicknesses from 6.5 to 8.0 nm. The curves were normalized by their resistance values at 10 K (R 10K ). The T c s of the thin films were determined by the criterion 0.5R/R 10K. The deposition rate of the NbN films was approximately 0.8 nm/s, and the thickness was controlled via the deposition time. Because these NbN films were fabricated in the same run, clear separation of the RT curves for each thickness was obtained, indicating good control of the deposited NbN thickness.

7 Calibration of the optical attenuators Supplementary Figure 5. (Color online) Measurement of the individual attenuator linearity at a wavelength of 1550 nm as the input power as varied. Three independent attenuators supplied as much as 180 db of optical power reduction. All of the output powers were measured using the same power meter (Keysight 81624B). All the attenuators exhibited good linearity that approached unity, independent of the input optical power. Three attenuators (one 81576A, two 81570As) were calibrated separately to confirm the expected linearity, particularly at high attenuation values. Supplementary Fig. 5 shows the measured transmitted power as a function of the nominal attenuation. The attenuation was measured at input power levels of 1 5 dbm, which emitted from the tunable laser source (Keysight 81940A) employed in the SNSPD characterizations. The linear fitting revealed that the linearity of the attenuators was at the level of ± 0.001, ensuring precise attenuation.

8 Estimation of the SDE and its relative uncertainty at 1550 nm We followed the method of ref. 8 [F. Marsili et al. Nat. Photon. 7, (2013)], which is now recognized as a common calibration method in the state-of-art technology. The loss due to fiber bending and splicing has been considered in the calibration of input optical power to avoid the overestimate the SDE. The number of input photons (N ) is estimated by an expression of N = P c *α 2 *α 3 *R sp /(1-r)/(1-L b )/(1-L s )/E λ, where P c is the input optical power in the control port, α 2 (α 3 ) is the real attenuation at a certain wavelength of attenuator 2(3), R sp is the splitting ratio of the MEMS optical switch, r is the reflectivity of the ARC lens fiber, L b is the optical loss due to fiber bending, L s is the optical loss due to fiber splicing, and E λ is the photon energy at a wavelength of λ. Then, we estimated the SDE as: SDE = = / / / / (1) where PCR is the photon respond count rate of SNSPD. Assuming that all the source of uncertainty are independent in equation (1), the relative uncertainty of SDE can be expressed as: ( ) = (2) where is the uncertainty of PCR; is the uncertainty of the power incident on the control power meter (P c ); is the uncertainty of the attenuation of attenuator 2(3); is the uncertainty of the splitting ratio of the optical switch. is the uncertainty of the bent loss; and is the uncertainty of the splice loss. In equation (2), the uncertainty on r and on E λ were neglected and the uncertainty on the attenuation of attenuators 2 and 3 were assumed to be identical. In our experiment, r is less than 0.3% in the wavelength range of interest. a) Relative uncertainty on PCR ( ) Since the PCR = CR-DCR, where CR is the output pulse count rate, and DCR is the dark count rate of the SNSPD. Then, = (3) We estimated σ CR and σ DCR by calculating the standard error of the mean on 6 consecutive measurements of the CR vs I b and DCR vs I b curves. For the device 02#F9 biased at 13 μa,

9 0.10%. Following a similar procedure, to be about 0.09%. for device 04#E4 biased at 21 μa has been determined Then, the relatively large value~ 0.10% between the two devices was determined as. b) Relative uncertainty on the power incident on the control port ( ) The power incident on the control port P c may effect by the calibration on power meter ( ) and the stability of the laser ( ). Then, the relative uncertainty of P c could be expressed as: = (4) Where and were estimated by the following parameters. i. Relative uncertainty on the power meter ( ) In the measurements of the system detection efficiency of the SNSPDs presented in the main text, we used a 5-mm-diameter area InGaAs detector with a PTB traceable calibration certificate. According to the calibration report, the detector has a measurement uncertainty of 2.80% within the range of nm, in its linear region. ii. Relative uncertainty on the laser power ( )

10 Supplementary Figure 6. The optical power stability measurement. The laser power in the control port as a function of time, monitored by a high precise power meter (Keysight 81624B). During the measurements, the laser power can be monitored and calibrated at any time by switching the optical input to the control port with the MEMS optical switch. As the Supplementary Fig. 6 showed, the laser power in the control port was monitored by using the power meter (Keysight 81624B) over 3600 s, which was longer than a typical measurement time for the SDE measurements. In the stability characterization and device measurement, the integration time of power meter was 1 s. the relative uncertainty of the laser power has been determined to be 0.09% for the 1550 nm laser. Then, using equation (4), we could determine 2.80%. c) Total relative uncertainty on the Attenuator 2 and 3 The real attenuation of attenuator 2 (α 2 ) and attenuator 3 (α 3 ) were estimated from the display reading of the control power meter as: The uncertainty of the attenuator may contributed by the stability and linearity of attenuators. Therefore, the relative uncertainty of α 2 (α 3 ) could be expressed as: = (5) Where and is the relative uncertainty of the stability and linearity of attenuator. was due to the repeatability of the nominal attenuation setting of attenuator and was estimated by calculating the standard deviation on 6 consecutive measurements: 0.22%. has been determined by the relative standard error from the linear fit of the attenuator (Supplementary Fig. 5). At 1550 nm, this values has been determined to be about 0.07% for attenuator 2. Then from equation (5), 0.23%. The total relative uncertainty of the two attenuators (2, 3) was = %. d) Splitting ratio of the MEMS optical switch and its relative uncertainty ( )

11 The splitting ratio of the optical switch R sp was estimated by averaging 6 consecutive display readings of the control and detector power meters. The relative uncertainty of R sp was calculated by the standard error of mean on the measurements: = 0.20%. e) Bent loss and its relative uncertainty ( ) In the experiment, in order to suppress the blackbody radiation of the fiber in room temperature, we bent the fiber attached to the device in the low temperature with a diameter of 30 mm and 5 turns. To measure the uncertainty due to fiber bending, we connected the optical source and the high precise power meter with a SMF. The mean bent loss at 1550 nm was about db by averaging 6 consecutive measurements. The bent loss has been considered in the calibration of input optical power to avoid the overestimate the SDE. The relative uncertainty of L b was as calculated by the standard error of mean on the measurements: = 0.14%. f) Splice loss and its relative uncertainty To measure the uncertainty due to fiber splicing, we connected the optical source and the high precise power meter with a SMF. Then, we cut and spliced the fiber 6 times with power meter monitored at 1550 nm in whole process. The averaged spliced loss (L s ) due to splicing was about db, and its relative uncertainty was calculated by the standard error of mean on the measurements: = 0.15%. 2) Total relative uncertainty of SDE In Equation 2, we can calculate the relative uncertainty of SDE to be ( ) 2.84%. SOURCE RELATIVE UNCERTANTY (%) PCR 0.10 P c 2.80 Attenuators 0.33 R sp 0.20

12 L b 0.14 L s 0.15 TOTOAL 2.84 Supplementary Table 1. The relative uncertainty contributions in the measurement of SDE.

13 Performance of 6.5-nm-thick devices Supplementary Figure 7. (Color online) Parallel polarized SDE and DCR vs. I b for device 01#G7 with an NbN thickness of 6.5 nm, as measured at 2.1 and 16 mk. Because of the reduction of thickness, a clear saturation plateau is still observed at 2.1 K. Supplementary Fig. 7 shows the parallel polarized SDE and DCR as functions of I b for device 01#G7 measured at 2.1 and 16 mk. This device features a nominal 6.5-nm-thick, 75-nm-wide, and 130-nm-pitch nanowire covering an active area of 18 μm. Because of the thinner nanowire, its I sw of 11.2 μa at 16 mk is lower than those described in Fig. 3 (15.2 and 22.5 μa). The maximum SDE for this device was 82.2%, and the PER was 3.9 at 1550 nm. A substantially lower SDE was obtained with the device with a 6.5-nm-thick film compared to its simulated absorption value of over 95%. This SDE was also approximately 10% lower than those of the 7(8)-nm-thick devices. We carefully measured other devices in the same run and observed saturated SDEs in the range from 70 to 82%, varying by device (Supplementary Tab. 2). Thus, the deviation between the experimental and simulated results might be due to the thin nanowire combined with an imperfect DBR cavity, resulting in low photon absorption.

14 Wavelength dependence of SDE(I b ) Supplementary Figure 8. (Color online) SDE vs. I b for device 04#E4, with various wavelength photons incidence, measured at 16 mk. A clear saturation plateau was obtained at 1310 nm, due to the high excitation energy of 1310-nm photons. In order to investigate the wavelength dependence of SDE, we illuminated the device using photons of different wavelength. At each wavelength, the input photon power was carefully calibrated and attenuated to a flux of photon/s. Photons with wavelengths of nm and 1310 nm were emitted from a CW-laser (Keysight 81940A) and from a ps-pulse laser (Hamamatsu, C10196), respectively. Supplementary Fig. 8 shows the SDE measured at 16 mk as a function of I b for device 04#E4, illuminated with 1310-, 1520-, 1550-, and 1630-nm photons. In the case of 1310-nm photons, a significant saturation was observed because of these photons high excitation energy. With increasing photon wavelength, the saturation plateau shrank and eventually disappeared, indicating a non-unity IDE. Besides, It was found that the SDE could be overestimated in a high bias current region when the SNSPD was illuminated, which was usually shown as an overshoot in the SDE curve at bias current close to its switching current. The reason is that an increase of the DCR under illumination was mistakenly recorded as photon respond counts. Please find the detailed discussion in the reference [S. Chen et al., "Dark counts of superconducting nanowire singlephoton detector under illumination," OE, 23, (2015)]. In the supplementary Fig. 8, the abnormal high SDEs were observed when the I b >21.7 μa (red-dashed line). For example, an overshoot (indicated with red-dashed circle) was clearly found in the saturation plateau of the SDE curve for 1310-nm wavelength. Therefore, this part of data should not be considered as the true SDE. Similar phenomena also appeared at other wavelengths, such as 1550 nm. Notably, in the main text, the overshoots of SDEs were already removed to avoid the misunderstanding. Here, to the supplementary Fig. 8, we kept these data and marked it out.

15 Estimation of the SDE relative uncertainty at the wavelengths from 1200 nm to 1700 nm 1) We measured the open circles in Fig. 5 by following a similar procedure as the measurement at 1550 nm, i.e., N = P c *α 2 *α 3 *R sp /(1-r)/(1-L b )/(1-L s )/E λ. We used the same high-precision optical power meter (Keysight 81624B) to calibrate the fibers, filters, and switch for different wavelengths. Therefore, the maximum uncertainty of the measurement was still attributed to the uncertainty of power meter, which has a measurement uncertainty of 2.80% within the range of nm, in its linear region. 2) We used an optical switch (Thorlabs Inc., OSW E, wavelength from nm). Notably, for the wavelength out of the transmission wavelengths of the switch, the switch was operated at a multimode circumstance, which could result in a nonlinear effect on R sp. Therefore, the input powers to the SNSPD at different wavelengths were corrected in each SDE measurement using the actual R sp. 3) In our measurements, the data taken with the white light source was in the range of nm and nm. The filter we used (LLTF Contrast SWIR HP8) is a continuously tunable high-resolution bandpass filter with a 4-nm FWHM bandwidth and a 60-dB suppression of out-of-band light. The high contrast filter and a long term stability of a locked output (better than ±0.5%) enabled us a reliable calibration to the number of input photons. 4) The attenuators (keysight 81570A) we used were calibrated by the manufacturer with an accuracy better than 2.3%, within the range of and nm, with a nearunity linearity. Then, by substitute α t to 2.3% and considering the stability of source of ±0.5%, the relative uncertainty of SDE at these wavelengths could estimate to be = 3.67%

16 Response pulses of photon response Supplementary Figure 9. (Color online) Response pulses of devices 02#F9 and 04#F9 recorded by an oscilloscope at I b = 14.5 and 21 μa, respectively, when the devices were operated at 16 mk. The fast decay time of device 04#E4 was due to its small active area of 15 μm and thick thickness. We characterized the decay time of the response pulse by directly monitoring the amplified output electronic pulse with an oscilloscope. Response pulses of two devices 02#F9 and 04#F9 are shown in Supplementary Fig. 9; the devices were biased at I b = 14.5 and 21 μa, respectively, and operated at 16 mk. We fitted the decay time using an exponential decay function expressed as a + b(e t/τ ), where τ is the time at which the height of the pulse is reduced to 1/e = of its initial value and a and b are fitting parameters. Thus, by fitting the decay part of the pulse, we obtained the decay time τ = 48.5 (27.3) ns for device 02#F9 (04#F9).

17 List of characterized devices Supplementary Table 2. (Color online) List of characterized devices. No. 1 Operating Thick. W-P Φ Temp. (nm) (nm) (μm) (μa) (%) (K) Isw SDEmax PER SDEARC (%) PERARC From left to right: sequence numbers (No.), nanowire width (w) and pitch (p); diameter of active area ( ), switching current (I sw ), maximal SDE (SDE max ), polarization extinction rate (PER). SDE ARC : maximal SDE after antireflection coating (ARC). PER ARC : PER measured after ARC. For the devices (No. 8 11), the influence of the ARC was studied. The ARC layer was deposited by electron-beam evaporation, with a SiO thickness of approximately 408 nm (i.e., λ/2/n SiO, where n SiO = 1.89 and λ is the target wavelength). The value of PER substantially decreased after the ARC, whereas the SDE did not show a notable change. The relatively low SDEs around 70% were measured by using fiber connectors causing an optical loss.

18 Comparison of the designs and performances between NbN-, MoSi-, and WSi-SNSPDs and W-TES Supplementary Table 3 (Color online) Performance of the NbN, WSi, MoSi-SNSPDs, and W-TES at 1550 nm wavelength. SNSPD TES Material NbN MoSi 1 WSi 2 W 3 Cavity half-cavity double-side full-cavity with backside Au (Al for TES) design with DBR cavity mirror SDE (%) approxim ately 1550 nm 92.1@1.8K 90.2@2.1K 80@2.1K 4 76@2.5K @0.7 K 82@2.3K 93@0.12K 90@2K 95@0.1K DCR (c/s) ~0 Jitter (ps) 79@2.1K @0.12K Isw (μa) approximat 14.5@1.8K 12@2.1K 4 9.5@0.7K 4@0.12K ely 13.8@2.1K 6.5@2.5K 5 4.3@2.3K 1.8@2k 17@0.1K 120 Decay 5 5, (rest time time 48.5 approximately 35 approximately 40 (ns) 30 4 ns) 800 Package fiber frontside package fiber back-side package 4 ;nano -positioner 5 front-side fiber self-alignment package Supplementary References 1 Verma, V. B. et al. High-efficiency superconducting nanowire single-photon detectors fabricated from MoSi thin-films. Opt. Express 23, , doi: (2015). 2 Marsili, F. et al. Detecting single infrared photons with 93% system efficiency. Nat. Photon. 7, , doi: (2013). 3 Lita, A. E., Miller, A. J. & Nam, S. W. Counting near-infrared single-photons with 95% efficiency. Opt. Express 16, , doi: (2008). 4 Yamashita, T., Miki, S., Terai, H. & Wang, Z. Low-filling-factor superconducting single photon detector with high system detection efficiency. Opt. Express 21, , doi: (2013). 5 Rosenberg, D., Kerman, A. J., Molnar, R. J. & Dauler, E. A. High-speed and high-efficiency superconducting nanowire single photon detector array. Opt. Express 21, 1440, doi: (2013).