Rapid detection and discrimination of individual incoming photons are pivotal in exploring quantum phenomena at the elementary level. They enable broad-ranging applications, such as Moon–Earth laser communication [1], high-rate quantum key distribution (QKD) [2,3], Gaussian boson sampling [4–6], and linear optics quantum computation [7]. Traditional single-photon detectors (SPDs), such as avalanche photodiodes and photomultiplier tubes, offer moderate detector performance in terms of the efficiency, speed, and timing resolution [8,9]. Superconducting transition-edge sensors have intrinsic photon-number-resolving capabilities and high detection efficiency but have low detection speeds, large timing jitters, and ultralow operating temperatures. Recently developed superconducting nanowire SPDs (SSPDs) are promising candidates owing to their near-unit efficiency [10–12], ultralow dark count rate (CR) [13,14], and excellent timing resolution [15]; moreover, they enable numerous novel applications, such as quantum communication [16–18], optical quantum computation [6,19], and quantum principle verification [20,21]. Despite their excellent performance, conventional single-element SSPDs have limited detection speeds (typically tens of megahertz) owing to their readout circuitry and detector recovery times [22]. Moreover, the photon number resolution (PNR) of SSPDs is severely limited by the high nonlinearity of the superconducting-to-normal state transition. Researchers have attempted to improve the PNR of SSPDs using a tapered impedance transformer, a wideband cryogenic amplifier, and series/parallel resistors. However, efficiently distinguishing multiple photons with a high speed and a high PNR remains challenging.

Multipixel SSPD array configurations are innovative optical detector arrays comprising multiple nanowires that are each equipped with independent biasing and readout circuits. In comparison to single-pixel SSPDs, multipixel SSPDs not only have high efficiency and low dark counts due to their nanowire geometry and optical structure but also enhance the CR and PNR by operating multiple nanowires in parallel [23,24]. This feature offers a unique combination of benefits for optical detectors. Recent experimental studies demonstrate their capabilities in applications such as Moon–Earth laser communication, high-rate QKD [25], and the characterization of photon sources [26], highlighting their potential for future quantum information technologies.

In this work, we present the development of an efficient, ultrafast SSPD with high PNR. The detector consists of 64 paralleled sandwiched superconducting nanowires; these nanowires are arranged into a unique geometry and integrated on a distributed Bragg reflector (DBR), which maximizes its photon sensitivity and detection speed. A compact cryocooler based on a Gifford–McMahon (GM) refrigerator is developed for SPD reliability and convenience. This cryocooler supports 64 electrical channels and has a minimum working temperature of 2.3 K. The integrated SPD system shows a functional nanowire yield of 61/64, a maximum system detection efficiency (SDE) of 90% at a 1550 nm wavelength, a maximum CR of 5.2 GHz, and a PNR of 61. This detector system represents a significant advancement in quantum detector technology, and its excellent performance and reliability will make it a valuable tool in a range of applications, including deep-space laser communication, high-rate quantum communication, and fundamental quantum optics experiments.

The $\mathrm{NbN}/{\mathrm{SiO}}_2/\mathrm{NbN}$ sandwiched superconducting structure with a thickness of 6/3/6 nm was first prepared atop a DBR structure, which served as a mirror to enhance light absorption [27]. Then, 64 paralleled nanowires with widths and pitches of 70 and 140 nm, respectively, were achieved, encompassing a circular active region with a diameter of 28 μm. The paralleled nanowire bends were rounded to minimize current crowding, as displayed in Figs. 1(a)–1(c).

Figure 2 shows a schematic diagram of the device measurement setup. For device integration, the device was mounted on a specially designed holder with 64 channels, as shown in Fig. 2(b). It was then connected to a printed circuit board using bonding wires and linked to staggered vertical and horizontal electrical channels. The entire assembly was precision aligned to ensure connection accuracy between the device and the printed circuit board. Finally, the package was mounted on the cold head of a two-stage GM cryocooler operating at approximately 2.3 K.

A single-mode lensed fiber was used to introduce the incoming photons into the chip, which was aligned directly from the front side to the detector. The input photon flux, emitted by continuous-wave (CW) or pulsed-laser sources, was controlled using two variable attenuators and calibrated using an optical power meter to ensure accurate measurement of the photon flux. The input photons were polarized parallel to the nanowire by tuning a polarization controller. The voltage pulses generated from each nanowire of the SSPD array were individually amplified using 50 dB low-noise amplifiers (RF Bay Inc., LNA-650) operating at room temperature. The amplified voltage pulses were then transmitted to a data acquisition system for further processing and analysis. The switching current and efficiency of each channel were measured, and simultaneous measurements for all channels were conducted using the 64-channel bias and readout circuits [Photon Technology (Zhejiang) Co., Ltd.]. The self-built system can conduct SDE and PNR tests [28]. Detailed test results are presented in Figs. 3 and 4.

We characterized the key attributes of the device, such as speed and efficiency, by testing individual biases and performing final summing CRs or SDEs to obtain the overall results. Figure 3(a) shows the distribution of the switching currents in the device. Three channels are open circuits at low temperatures, which can be attributed to various reasons that may include unintentional damage to the nanowires during the etching or deposition processes, insufficient contact formation, or other fabrication-related issues. The remaining transition currents are mainly distributed around 16 μA, with an average value of 16.3 μA, indicating reasonable uniformity in the device. This also displays the normalized CR to total input distribution of all pixels, showing a Gaussian distribution related to the actual spot distribution. Through this figure, the center position of the spot can be confirmed, indicating the imaging capability of our device.

As shown in Fig. 3(b), the maximum overall SDE is approximately 90%. The majority of pixels reach their near-saturation platform at a bias current close to 14 μA. The uncertainty in the SDE in our experiment was approximately $\pm 1.0\%$, stemming from two sources: the count rate uncertainty from the photon response and the uncertainty in the number of photons entering the SSPD system. The uncertainty is influenced by several factors, including the uncertainty in photometric detection as measured by an optical power meter, wavelength-dependent uncertainties, nonlinearity-related uncertainties, stability concerns related to attenuators and lasers, the uncertainty in the adjustment of the polarization controller, and the uncertainty associated with fiber splicing losses [10]. Note that in Fig. 3(b), only the SDE curve under low light intensity is shown. When higher light intensities are involved, a specific correction on single-photon detection efficiency is required [29] because of the non-ignorable multiphoton events. The SDE curve in Fig. 3(c) as a function of the measured count rate is determined by calculating the ratio of the actual count rates to the total number of photons. This reflects the actual performance of the device under high count rate conditions, and no additional corrections are used.

The design used in this study helps reduce the effective photon count received by each pixel and the dynamic inductance of individual pixels. Consequently, the array can operate effectively at significantly high input photon rates and is thus a major improvement compared to single-pixel SSPDs. To illustrate the comprehensive performance of the 64-pixel array, we graph the relationship between single- and combined-pixel SDEs and CRs in Figs. 3(b) and 3(c), respectively. The overall trend of the SDE with the CR is a near steady state within 300 MHz, gradually decreasing to 50% when the CR reaches approximately ${\mathrm{CR}}_{3\mathrm{dB}}=1.7\mathrm{GHz}$. The maximum CR (MCR) exceeds 5.2 GHz at a photon flux of ${10}^{12}$ photons per second, while the total SDE remains at 0.2%. This phenomenon agrees with the findings of a previous study [22], which predicted that the counting performance of an SSPD rapidly declines once the CR exceeds a threshold of approximately ${(20\times {\tau }_d)}^{-1}$. The dead time (${\tau }_d=4.8\mathrm{ns}$) is the moment when the pulse height decreases to $1/e$ times its initial value for a single nanowire, as given by the output voltage pulse generated by a single pixel in the inset of Fig. 3(c). Notably, the time jitter of a single pixel at a current of 0.95 times the transition current ($\sim 14\mathrm{\mu A}$) is 80 ps.

For clarity, we compare several data points related to the total SDE and the total CR of the proposed array and previously reported configurations, including interleaved ones. As shown in Table 1, our detector system with 61 channels operating at 2.3 K has shown an improvement in the MCR from 3 to 5 GHz compared to similar previous works [26]. The achievement of a 5 GHz MCR may benefit the high speed quantum communication and the Moon–Earth laser communication. Note that the count rate at 3 dB efficiency ${\mathrm{CR}}_{3\mathrm{dB}}$ is limited to 1.7 GHz because the diameter of the coupled SMF is relatively small compared to the sensitive area of our detector, resulting in a low photon detection efficiency (as low as 0.001%) for pixels located far from the center of the light spot. An increased 3 dB count rate is expected when the detector coupled with a large input light spot.Table 1.

Comparison of CRs of Current and Previous Studies

The PNR capability of the 64-pixel SSPD array was characterized. Due to the independent biasing and readout of different channels, a maximum resolved photon number of 61 is easily obtained, corresponding to the available channels. To simplify postprocessing in practical applications, we combined the signal outputs from different channels and recorded them using an oscilloscope (Keysight Technologies, MSOV204A Infiniium V-Series). Pulse height distribution histograms were generated by sampling photon responses at various light intensities. Notably, the Gaussian peaks in these histograms are separated, allowing for accurate discrimination of the number of photons, as depicted in Fig. 4(a). In this experiment, we used a modulated pulse light source with an increasing average number of photons per pulse. As the optical intensity increases, the statistically obtained distribution of pulse amplitudes continuously increases. We note that the intensity and the resolved photon number of a single pulse are related by the detection efficiencies of the nanowires. Since the efficiency of each nanowire is different, the large light intensity (ranging from $-83.92$ to $-35.92\mathrm{dBm}$) was used to trigger the nanowire with lower efficiency. With increased optical intensity, we observe a gradual right shift in the combined pulse amplitude, indicating that an increasing number of triggered nanowires. Under weak light conditions (e.g., incident optical $\text{intensity}=-86.92\mathrm{dBm}$), the statistically measured numbers of photons corresponding to the output pulse amplitude are primarily in the range of 1–8. With an increase in the optical intensity, the maximum pulse amplitude and the distribution of photon numbers shift toward higher values. At a given optical intensity, the distribution of photon numbers exhibits a Gaussian distribution. At an incident optical intensity of $-35.92\mathrm{dBm}$, the maximum number of photons saturates at 61. The resulting persistence map displays the 61 individual pulse heights corresponding to a photon count range of 1–61. The linearity of the PNR deteriorates with the introduction of partial channel noise. In principle, a low-temperature amplifier will help improve the maximum number by increasing the signal-to-noise ratio.

Figure 4(b) depicts the linear relationship between the resolved photon number and the output pulse amplitude. The electronic readout system used in the experiment processed the signals from different channels and converted them into uniform square-wave signals. These signals were then combined, and a well-defined linear correlation was produced between the combined output signal and the resolved photon number. The linear fit shows a good agreement with the measured amplitude data as shown in the graph, where the photon number information can be easily extracted [28]. Furthermore, in Fig. 4(b), the Gaussian peaks observed in the output pulses are presented in terms of their FWHM values. A smaller FWHM indicates a higher signal-to-noise ratio and better differentiation between the output pulses at different photon counts. The graph reveals that at lower photon numbers, good differentiation is seen between different photon numbers. However, this differentiation gradually decreases as the photon number increases. This can be attributed to the activation of channels with lower signal-to-noise ratios, which reduces the overall signal-to-noise ratio of the combined signal. Additionally, the reduced accuracy of higher output pulses, caused by the oscilloscope only outputting three significant figures, may also contribute to this trend. The FWHM for each data point falls within the range of 0–42 mV.

Table 2 presents a comparative analysis of published SSPDs with PNR capabilities. Single-element SSPDs with taper impedance matching [33], principal component analysis (PCA) [34], cryogenic amplifier [35], large-inductance superconducting microstrip [36], and the PNRD-type SSPD [31] with shunted resistors have demonstrated the photon number resolution up to 24. A recently developed waveguide-integrated SSPD achieved a PNR of 100; however, it suffers from the coupling losses for vertical coupling applications [32]. It is worth mentioning that recently developed silicon photomultipliers (SiPMs) have shown the capability to resolve up to 14 photons [37] and can utilize frequency upconversion techniques for mid-infrared detection [38], but there is still potential for improvement in the efficiency and speed. In comparison, our detector array performs better than the other detectors and SSPD arrays with similar pixel counts in terms of the PNR, SDE, and detection speed. In addition, based on the efficiency distribution data of the devices, we can estimate their fidelity matrices [39]. The fidelities for one-, two-, three-, and four-photon events are 90%, 77%, 69%, and 52%, respectively, all of which exhibit significant advantages over reported devices [26].Table 2.

Comparison of PNR Capabilities of Proposed and Previous SSPDs

In summary, we present a promising SSPD that consists of 64 paralleled nanowires and combines high speed, efficiency, and good PNR capability. It is an excellent candidate for a diverse range of applications that require rapid, accurate detection of individual photons. Our fabricated SSPD yields an SDE of 90% at a 1550 nm wavelength, an impressive MCR of 5.2 GHz, and the ability to resolve up to 61 individual photons. Additionally, the detector performance can be further improved. For instance, replacing the AC coupling readout with a cryogenic DC coupling setup may lead to MCRs exceeding 10 GHz [40]. Furthermore, the PNR fidelity may be refined by enhancing the SDE of nanowires or the number of nanowires using multiplexing readout techniques. Our detector system can be used in diverse applications, including deep-space laser communication, quantum information processing, quantum metrology, and other fundamental quantum optics experiments.