1. Introduction

Superconducting nanowires are capable of efficiently responding to single photons from visible to infrared wavelengths owing to the low gap energy of superconducting materials [1]. The first superconducting nanowire single photon detector (SNSPD) was demonstrated in 2001 by directly fabricating a nanowire on a substrate and the SNSPD has a very low absorption of incident photons [2]. Since then, different schemes have been proposed for enhancing absorption [3]. To date, backside illuminated SNSPDs with cavity structures [4,5] and front side coupled SNSPDs based on metallic [6] or dielectric mirrors [7,8] are two typical structures that are widely adopted for high detection efficiency (DE) SNSPDs. These SNSPDs have led to high DE close to unity [2–5], thereby enabling numerous applications such as long distance quantum key distribution [6,7], quantum random number generator [8], and optical quantum computation [9] in the past decade. Nonetheless, the wide band responding feature of SNSPDs is limited by the resonant effect of optical cavities. Even though superconducting nanowires are intrinsically wideband, most of the reported SNSPDs showed high DE only at around their singlet resonant wavelengths.

The recent advances have also extended the use of high DE SNSPDs from near infrared to visible light because of the critical applications in this wavelength range such as fluorescence correlation spectroscopy at 635 nm [9], satellite laser ranging/imaging at 532/1064 nm [8,10], and singlet oxygen luminescence detection at 1270 nm [11]. To meet the practical requirements of high DE SNSPDs at different wavelengths, one needs to repeat varying the detector designs and running the fabrication process accordingly [12]. This process is time-consuming and laborious. Some endeavors have been made to fabricate wideband SNSPDs [13,14]. However, it is challenging to design SNSPDs with high absorption for the whole waveband and the reported results are not as good as SNSPDs designed for single wavelengths [6,15,16]. SNSPDs designed for multiple wavelengths are thus expected to offer high absorption and high DE. Furthermore, simultaneously detecting single photons in multiple wavelengths are required and critical for multispectral applications such as dual wavelength ranging, multiple spectral imaging [17,18], and atmospheric remote sensing [19,20] that may reduce or eliminate the system complexity, associated electronics as well as the weight of the entire detection system.

In this work, we first demonstrate multispectral SNSPDs that could simultaneously detect single photons at well-separated multiple wavelengths with high DEs. The work principle relies on multiple resonant absorption wavelengths resulting from two serially coupled optical cavities formed by two quarter-wave stacks. The calculated resonant absorption reaches over 95% and the fabricated detector shows maximum DEs of >80% at around 1550 nm, 1310 nm, and 1064 nm. Our detector shows high efficiency single photon detection at multiple separated wavelengths and may also find applications in multispectral ranging or imaging.

2. Optical absorption and device design

Three typical near-infrared wavelengths (1064 nm, 1310 nm and 1550 nm) were selected as target wavelengths. Figure 1(a) shows the schematic of the proposed SNSPD, in which one quarter-wave stack 1 (S1) with central wavelength λ₁ was fabricated atop another stack 2 (S2) with central wavelength λ₂. Common materials of choice for optical film stacks are TiO₂ and SiO₂ with refractive indices of n_TiO2 = 2.30 and n_SiO2 = 1.47 at wavelengths around 1550 nm. The dispersion effect in the wavelength range under observation was neglected. An ultrathin NbN film is selected as the superconducting material because NbN SNSPD can work at 2.1 K obtained using a commercial Gifford-McMahon (G-M) cryocooler. The thickness (h) and width (w) of the superconducting NbN nanowire are chosen to be 6.5 nm and 80 nm, respectively, with a fill factor, f = 0.5 that are typical parameters for SNSPDs to respond to near infrared photons. The refractive index of NbN n_NbN was 6.50 + 5.83i at around 1550 nm and was obtained using a spectroscopic ellipsometer. The substrate parameters include a refractive index n_Si of 3.46 for Si and a thickness l of 400 μm.

 

Fig. 1 (a) Schematic of multispectral SNSPD based on dielectric mirror consisting of two film stacks with separated central wavelengths. (b) Simulated reflectivity (solid black line) and phase shift (dashed green line) on reflection at the front surface of the stacks and the absorptance of NbN SNSPD on the substrate (red line). (c) TEM image of the fabricated SNSPD sample. Pseudo colors (red, yellow and blue) are used to indicate NbN nanowires, yellow of stack 1 and blue of stack 2. Note that only four top bilayers of stack 2 are shown in the image. The layer at the very bottom of the image is the fifth SiO₂ layer which was destroyed by focused ion beam used in the TEM sample preparation. The top layer of the image is the Pt layer coated onto the sample for TEM test. (d) Measured reflectivity of the dielectric mirror without superconducting film (solid black line), absorption of NbN films (solid red line) and measured system DEs for the SNSPDs as a function of wavelength (blue scatters).

Download Full Size | PDF

The two serially connected stacks results in a combination of two high reflection bands and leads to two serially coupled cavities for the SNSPD, the resonant absorption peaks of which constitute the multiple high optical absorption wavelengths. The top stack S1 behaving as a mirror at a wavelength range of about λ₁, along with the superconducting nanowire comprises one cavity. The stack (S1) with more bilayer number of the cavity would result in high reflection close to unity and lead to a nearly total resonant absorption of NbN nanowires at around λ₁. On the other hand, S1 functions as a cavity spacer for another cavity composed of S2 and the nanowire at a wavelength range of about λ₂. The rapid phase change caused by the cavity spacer (S1) would lead to more absorption peaks with narrower resonant absorption bands. To ensure a high absorption efficiency around λ₁ while retaining as wide an absorption bandwidth as possible in the resonant peaks around λ₂ regarding fabrication tolerance and practical applications, 6 bilayers and another 8 bilayers were adopted in stacks S1 and S2 of our detector. In addition, the associated central wavelengths of the stacks were tuned to ${\lambda }_1=1020 nm$ and ${\lambda }_2=1440 nm$, ensuring the resulting resonant peaks were located at our target wavelengths of 1064, 1310, and 1550 nm.

To quantitatively verify the optical properties of the SNSPD, electromagnetic simulation was performed using a Matlab program based on the rigorous coupled-wave analysis method that is a semi-analytical method typically applied for solving scattering problems in periodic structures [21]. Figure 1(b) shows the calculated reflection, reflection phase shift upon the film surface, and the optical absorption of the nanowire as a function of wavelength for parallel polarization. A wide band reflection was obtained by combing two stacks with reflection bands around λ₂ and λ₁. The slight dips in the reflection zone around λ₁ are due to the limited number of bilayers (6) adopted here. Nevertheless, a high and wide absorption band was obtained around 1064 nm, as shown by the red solid line in Fig. 1(b). On the other hand, the cavity spacer (S1) resulted in a rapid phase shift (green dashed line) in the reflection zone around λ₂ and two resonant absorption peaks appeared at 1310 and 1550 nm, as indicated by the solid red line. Note that the three absorption peaks appear at around the wavelengths with phase shift of 0 so that the nanowire is located at the antinode of cavity field for enhanced interaction between nanowire and photons. The final resulting absorption peaks shown by the red line in Fig. 1(b) indicate that the resonant wavelengths at around 1550 nm, 1310 nm, and 1064 nm exhibit considerable bandwidth flexibility and high efficiency (>95%).

3. Device fabrication and measurement

We started our device fabrication with a 2-inch Si substrate. The quarter-wave optical film stacks composed of 8 periodic SiO₂/TiO₂ bilayers with a central wavelength of 1440 nm and 6 periodic SiO₂/TiO₂ bilayers with a central wavelength of 1020 nm were alternately deposited onto the Si substrate using ion beam assistant deposition, with the film thickness optically monitored to ensure adherence to the designed layer thickness. Then, a 6.5-nm-thick NbN film was deposited on the front side by reactive DC magnetron sputtering in an Ar/N₂ gas mixture at a total pressure of 0.27 Pa. The flow rates for Ar and N₂ were set to 30 sccm and 4 sccm, respectively, using mass flow controllers. The nanowire covers a circular area with a diameter of 15 µm and adopts a serially connecting avalanche architecture [22]. Next, the NbN film was coated through electron-beam photoresist PMMA, patterned into a nanowire structure via electron beam lithography, and etched via reactive-ion etching in a CF₄ plasma. Finally, a 50 Ω matched coplanar waveguide was formed using ultraviolet lithography and was etched via reactive-ion etching.

Figure 1(c) shows a transmission electron microscopy (TEM) image of the SNSPD sample fabricated in the same process with the same geometrical structure to the actual device being measured. Note that only four top bilayers of stack 2 are shown in the image with distinguished interfaces for each layer film. The layer at the very bottom of the image is the fifth SiO₂ layer which was destroyed by focused ion beam used in the TEM sample preparation. The top layer of the image is the Pt layer coated onto the sample for TEM only, which does not exist in the actual device. To verify the performance of the fabricated optical stacks, the reflection curve of the fabricated stack with/without the 6.5-nm-thick NbN film versus wavelength was measured using a spectrophotometer (Agilent Cary 7000) at room temperature. As Fig. 1(d) (black solid line) shows, the structure without NbN film shows two high reflection bands, as expected. Furthermore, by subtracting the reflectivity of the measured film with NbN film from that without NbN film, the estimated optical absorption of the NbN film was obtained, as shown by the red line in Fig. 1(d). The location of the high absorption peaks at 1550 nm and 1310 nm coincide well with theoretical calculation results. The slight deviation in the absorption peak at 1064 nm was mainly caused by the dispersion effect of the dielectric film that was neglected in our design. Note that we estimated the absorption by subtracting the reflectivity rather than measuring the transmission losses because of the Si absorption of the transmitted near infrared light (~1100nm).

The fabricated SNSPD was packaged inside a copper sample mount. A single-mode fiber (Corning: SMF-28e) was aligned directly from the frontside to the SNSPD. The packaged module was then mounted on the cold head of a two-stage Gifford–McMahon cryocooler with a working temperature of 2.1 K. Outside the cryocooler, the device was connected to a room-temperature bias-tee. An isolated voltage source in series with a resistor (20 kΩ) provided a stable current bias to the detector. The bias current was fed to the device through a dc port of the bias-tee while high-frequency response pulses of the SNSPD were extracted from the ac port of the bias-tee and subsequently amplified via an ultra-wideband amplifier. The amplified pulse signals were read by an oscilloscope or counted by a photon counter.

In our measurement of the DE, the pulsed laser was prepared at a specific wavelength (1550 nm, 1310 nm, or 1064 nm) using a supercontinuum laser (NKT: EXB-3) and an acousto-optic tunable filter (NKT: SuperK SELECT). The laser source was then connected in series with two variable attenuators (Keysight: 81570A) and a polarization controller (Thorlabs: CPC900). The laser was measured with a power meter (Keysight: 81624B) and then heavily attenuated to achieve a photon flux of 10⁵ photons/s. The system DE was defined as (OPR−DCR)/PR, where OPR is the output pulse rate of the SNSPD, measured using a photon counter, DCR is the dark count rate when the laser was blocked, and PR represents the total photon rate input to the system. At each bias current, an automated shutter in a variable attenuator blocked the laser light and dark counts were collected for 10 s using a commercial counter. Then, the light was unblocked, and the output photon counts collected for another 10 s. Errors due to the calibration of the laser power were less than 3.5% as given by the power meter, and the laser power fluctuation was less than 1.2%.

Figure 2 presents the DEs and DCRs as a function of the bias current, whereby all DEs reach 80% at a dark count rate of 200 Hz at around 1550 nm, 1310 nm, and 1064 nm. The DE curves show more saturation at shorter wavelengths than at longer wavelengths due to the high intrinsic detection efficiency of high energy photons. We also notice that the three DE curves start to detect photons at around 11µA, rather than shifting along with the incident photon energy. It is owing to the existence of the avalanche current of the parallel connected nanowires, at around which the detector starts to trigger the avalanche and register photons [23]. Moreover, the DEs at around these target wavelengths were also measured, with the results plotted in Fig. 1(d), the tendencies of which show good agreement with the measured absorption curves. The slight deviation in the location of the experimental DE peaks from our simulation results were caused mainly by imperfections in the fabrication of the optical stacks and superconducting nanowires, the dispersion effect of the optical film index, as well as the influence of low temperature on the material index. It is worth noting that a lower DCR could be realized with filtering techniques to depress the DCRs caused by background irradiation such as multiple-wavelength bandpass filters and/or fiber coiling [24–26].

 

Fig. 2 DEs at 1550 nm, 1310 nm, and 1064 nm and DCRs versus bias current for the SNSPD.

Download Full Size | PDF

4. Discussion and conclusions

The structure proposed in this work allows wide-separated absorption wavelengths with features such as considerable bandwidth flexibility and high efficiency. Compared to SNSPDs designed with wideband features [13,14], the multispectral SNSPD offers higher simulated absorption as well as higher measured DEs at specific wavelengths. A series of separated wavelength combinations can be realized because the resonant wavelengths can be tuned by adjusting the cavity parameters, such as central wavelength and bilayer number of the optical stack, and/or adding cavity spacer layer atop each stack. The lossless dielectric material property and the coupled cavity structure make it capable of realizing multiple spectral SNSPDs with high efficiency at multiple wide separated wavelengths, from visible to near infrared or even middle infrared wavelengths. In addition, more separated target wavelengths can be obtained by adding more cavities in series, though the resulting thick cavity spacer may lead a narrow absorption band.

In conclusion, we realized a multispectral SNSPD that can simultaneously detect single photons at three different wavelengths. The detection principle relies on the absorption resonances of a cavity structure composed of double film stacks. The fabricated SNSPD showed maximum DEs of >80% at 1550 nm, 1310 nm, and 1064 nm. The SNSPD may replace three narrow band detectors in a system, potentially leading to a reduction in power, cooling, and cost of the overall detection system. This type of SNSPD may also find interesting applications in systems that require single photon detectors at different wavelengths.