Superconducting materials have experienced significant advancements in weak photon detection, and superconducting nanostrip single-photon detectors (SNSPDs) have garnered widespread interest since 2001 [1]. With their high system detection efficiency (SDE) [2], low dark count rate [3], low timing jitter [4], as well as their 100% internal detection efficiency (IDE) extending up to mid-IR wavelength [5], [6], [7], [8], SNSPDs have found applications in various fields [9], [10], [11], [12], including long-haul quantum key distribution, quantum random number generation, and quantum computing. Typically, devices with micropattern are biased by a dc current close to the critical current (I_C) to achieve high detection efficiency under half of superconducting transition temperature (T_C). While there remain numerous debates on the detection mechanism of SNSPDs [13], many experimental results [14] show that the characteristics of superconducting materials directly affect the internal detection efficiency of the devices, leading to significant research efforts in exploring new superconductors for SNSPDs to improve their detection performance [15], [16], [17], [18], [19].

Aluminum, a conventional superconductor, has found wide usage in superconducting devices such as kinetic inductance detectors [20] for terahertz detection in astronomical observation and microwave resonators [21], as well as Josephson junctions [22] for superconducting quantum circuits. In particular, the construction of superconducting quantum computation hardware relies on Al Josephson junction-based electronic resonant circuits and quantum bits. While bulk crystalline Al has a relatively low T_C of 1.2 K, it is known that Al thin-films can exhibit T_C values almost twice as high as bulk Al [23]. Additionally, by integrating high-efficiency Al-SNSPDs on-chip with the current Al-based superconducting quantum computation architecture, the prospect of constructing more intricate on-chip electronic and optical quantum information processors becomes promising in the future.

In this manuscript, we successfully demonstrated the deposition of ultra-thin Al films with significantly higher T_C than their bulk counterparts. The resistivity and inferred sheet inductance of the Al films are consistent with those of other Al films used in nano-fabrications. The Al-based SNSPDs with strip width of 178 nm were successfully fabricated using the standard nano-fabrication process, which exhibited a saturation plateau of IDE for visible wavelength photons. However, the detection performance of Al nanostrips for single infrared photons remains limited.

The Al thin films used in this study were deposited through DC magnetron sputtering in a cluster deposition system at room temperature, with a background pressure of approximately 2 × 10⁻⁶ Pa. These films were deposited onto a double-sided thermally oxidized Si substrate using pure argon (Ar). Prior to film deposition, all substrates underwent pre-cleaning with Ar plasma in a dedicated chamber. Following the deposition of the Al layer, an additional AlN capping layer was in situ deposited as a protective layer to mitigate degradation from oxidation. The deposition of both Al and AlN involved a DC power of 250 W, with deposition rates of 0.25 nm/s for Al and 0.05 nm/s for AlN. Lastly, a 3.3-nm-thick SiO₂ layer was deposited via chemical vapor deposition to safeguard the Al and AlN films against corrosion by the alkaline developer. Fig. 1 illustrates a high-resolution transmission electron microscopy (HRTEM) cross-section image of the resulting SiO_2/AlN/Al/SiO₂ films, highlighting distinct interfaces between the different layers. Despite the incident photons may generate more quasiparticles in a relatively thicker Al film with a lower T_C, an excessively thick film might lead to a significant decrease in the IDE of the SNSPD due to the reduced hospot size to nanostrip crosssection ratio [24]. Bearing this in mind, we made the deliberate decision to grow a 4.2-nm-thick film for the fabrication of the SNSPD in this experiment. Each film's thickness was accurately determined by combining TEM imaging with Energy Dispersive Spectrometer (EDS) analysis, utilizing EDS-measured depth distribution data of various elements to better define the thickness, particularly in regions where the film boundaries are less clear. Regrettably, despite our attempts at optimization, it can be observed from the TEM image that the uniformity of the films is less than satisfactory, necessitating further optimization of the deposition process to obtain higher-quality ultrathin Al-films in our future studies.

The as-grown films were subsequently patterned into nanostrips using standard fabrication processes. For electron beam lithography, the negative tone resist ma-N 2401 was chosen due to its high resolution, good thermal stability, and high resistance to dry etching. The resist was applied using a Delta 80 spinner at a speed of 2000 rpm and then baked on a 110 ℃ hotplate for 2 min, resulting in a resist thickness of approximately 140 nm. After electron beam exposure, the film was developed using ZX 238 (TMAH-based) for 10 s and thoroughly rinsed with deionized water for 1 min. Finally, the non-active area was etched using Ion Beam Etching Systems (IBE) with an ion energy of 245 eV. Considering the reported low resistivity of Al films [25], and to prevent device latching [26], we chose to prepare relatively long nanostrips. Fig. 2 presents an optical image of the fabricated Al SNSPD, which covers an active area of 50 × 50 μm², corresponding to a strip length of about 3.1 mm. The width and homogeneity of the resulting nanostrips were further examined using scanning electron microscopy (SEM). As depicted in the inset of Fig. 2, the strip width and pitch are 178 nm and 800 nm, respectively, with the strip width slightly narrower than the designed value of 180 nm. An analysis of the SEM image using ProSEM revealed that the line edge roughness was about $ L \bar{E} R=3 * \sigma=12.4 \mathrm{~nm}$, where σ is the standard deviation of the feature edge.

After the fabrication, the devices were packaged with a single-mode optical fiber with gradient-index lenses in a copper box. A homemade alignment system based on an inverted microscope was used to ensure that the focused beam spot of the incident light was aligned to the center of the active area. The packaged module was then mounted on the cold head of an adsorption cryocooler, which operated at a temperature of 0.85 K. The device was electrically-connected to a room-temperature bias-tee outside the cryocooler. A bias current was supplied to the device using a voltage source connected in series with a 100 kΩ resistor through the dc port of the bias-tee. The voltage pulses generated by the SNSPD were sent to the rf port of the bias-tee and amplified by a room-temperature 50 dB low-noise amplifier (RF Bay Inc. LNA-650). The amplifier has a 3-dB bandwidth covering the range of 30 kHz to 600 MHz, which can support the signal transition edge from 583 ps to 11 μs (0.35/BW) [27].The amplified pulse signals were either read using an oscilloscope or counted using a photon counter.

In the initial characterization of the highly disordered Al-SNSPDs, we examined the normal-to-superconducting transitions. The series resistance of the circuit was switched from 100 kΩ to 2.2 MΩ for this measurement. The normalized resistivity as a function of temperature for the fabricated Al-SNSPDs was plotted in Fig. 3 (a), covering the temperature range of 0.85 K to 6 K. The total normal resistance of the device was approximately 1.89 MΩ, with a film resistivity inferred to be 45.2 μΩ·cm which is higher than the reported value for a similar film thickness (∼30 μΩ·cm) [25]. The temperature-dependent resistivity of the Al-SNSPD can be well-described by the two-dimensional (2D) superconducting fluctuation mechanism with the 2D Aslamazov and Larkin (AL) [28] and Maki and Thompson (MT) paraconductivity [29], [30], which has been widely applied for describe the quasi-2D superconductivity in superconducting nanostrips [31], [32]. The fitted curve is incorporated by the red solid line in Fig. 3 (a), resulting in a critical temperature of 2.4 K, which is consistent with those previously reported for Al-films of comparable thickness [23]. The current-voltage (IV) characteristics of the Al-SNSPDs at 0.85 K is shown in Fig. 3 (b). The switching current (I_sw) and the re-trapping current (I_r) are 5 μA and 0.4 μA, respectively, with an inferred critical current density (J_C) of 6.7 × 10⁹A/m². The I-V curve exhibited a significant hysteresis due to local Joule heating, which generated a self-sustained resistive hot-spot in the nanostrip. As seen in Fig. 3 (b), during the down-sweep of the bias, the resistance of the device did not decrease directly with the voltage reduction but remained relatively constant. This behavior can be attributed to the poor thermal coupling between Al and Si substrate, and the relatively long hotspot relaxation process of Al-SNSPD.

Fig. 3 (c) displays a single-shot trace captured by the oscilloscope, demonstrating the real-time photon response pulse of the Al-SNSPDs at 0.85 K, with a bias current of 0.9 I_sw. The unusual peak at the falling edge was due to the RF reflections which happened at the input port of the amplifier and arrived at the device ∼24 ns after the detection time.[33]. The pulse exhibits a rising time ($\tau_{\text {rise }}$) of 3 ns, while the detector recovers with a time constant of 33.2 ns ($\tau_{\text {fall }}$), calculated by fitting the exponential decay of the falling edge and defining the time width at which the amplitude reaches 1/e (0.368) of the maximum pulse amplitude.

The sheet kinetic inductance (L_ks) of our Al-SNSPD can therefore be estimated by $ L_{k s}=\tau_{\text {fall }} R_{L} w / l \approx 95.6 \mathrm{pH} /$ square, where w refers to the width of the nanostrip, l represents the total length of the nanostrip, and R_L (50 Ω) denotes the load resistance of the readout circuit [34]. This value is consistent with other Al films [35]. The relatively long rising time can be primarily attributed to the small hotspot resistance of the Al-SNSPD. By considering the relationship $ \tau_{\text {rise }}=L_{k} /\left(R_{L}+R_{n}\right)$, where L_k denotes the kinetic inductance, and R_n signifies the hotspot resistance [34], we can calculate a R_n of approximately 500 Ω.

We further investigated the detection performance of the Al-SNSPDs for different photon wavelengths, specifically 405 nm, 785 nm, and 1550 nm. Fig. 4 (a) presents the normalized photon counting rates as a function of the bias current applied to the Al-SNSPDs. As anticipated, the detector exhibits better saturated IDE at shorter wavelengths. As depicted in Fig. 4 (a), the device demonstrates a clear saturation plateau for 405-nm photons. However, no discernible saturation trend is observed for the photons at the other two wavelengths. We also characterized the SDE and dark count rate of the same devices. The input light was heavily attenuated to a single photon level by three cascaded attenuators. A polarization controller was inserted after the attenuator to manipulate the light polarization to obtain the maximum SDE. In the DCR measurement, the optical-fiber port at room temperature was shielded to eliminate any stray light from the environment. Due to the inherent limitations in absorption efficiency and fiber transmission loss, the Al-SNSPD only achieves a maximum SDE of 1.5% at 405 nm, accompanied by a dark count rate of approximately 100 Hz. The device's absorption efficiency can be further improved by introducing optical cavities in SNSPDs. This can be achieved through various approaches, such as the implementation of metallic or dielectric mirrors beneath the nanowires, or the utilization of double-sided cavities where the superconducting nanowire layer is situated between upper and lower cavity layers composed of dielectric material. Moreover, the utilization of multi-layer nanowires also holds promise in enhancing absorption efficiency.

The inefficient single photon detection of Al-SNSPDs for infrared photons can be attributed to the relatively large diffusion coefficient (D_e) of normal state electrons in Al films. Generally, 2D Al films exhibit a D_e (∼5 cm²/s) that is one order of magnitude higher than that of disordered NbN films or amorphous WSi films (∼0.5 cm²/s) [36], [37], [38]. Consequently, the quasiparticles or hot electrons within the hotspot are not sufficiently heated, resulting in a relatively low “temperature” within the hotspot. In Al SNSPDs, since the devices are prone to latch into a resistive state, we may then estimate that the hotspot relaxation time is comparable with the kinetic induce limited recovery time, ∼ 33.2 ns. If we even estimate the hotspot relaxation as 1 ns, with a De of ∼ 5 cm²/s, the hotspot size can then be estimated to be around $ 2 \sqrt{D_{e} \tau_{\text {hotspot }}} \approx 1400$. As a result, the dissipated photon energy within the hotspot may be significantly suppressed, leading to a relatively “cold” hotspot. This may partially explain the relatively small hotspot resistance observed in our Al nanostrips. As a result, the hotspot is unable to effectively expel the bias current into the readout, thus limiting the IDE of the devices. Apart from the large D_e, another contributing factor that further diminishes the IDE of Al-SNSPDs is the relatively long coherence length ($\xi$). Due to its ultra-low upper critical field, the $\xi$ of bulk clean Al can reach up to 1.5 μm [39]. Even for our disordered Al films, the $\xi$ can still be more than 60 nm [40], which is significantly larger than the $\xi$ of other commonly used thin films in SNSPD fabrication. Finally, it is also interesting to note that as a type-I superconductor, the quantized vortices is absent in Al-SNSPD. As a result, the vortex-assisted photon detection event (either by single vortex crossing or breakup of vortex-antivortex pairs) is therefore not present in Al nanowires, which in turn further hampers the superconducting to normal transition within the Al nanowire cross section. Consequently, to completely collapse superconductivity of Al-nanostrips, it would necessitate more telecom wavelength photons to generate a detection event. Specifically, on one hand, due to the relatively long coherence length in Al-nanostrips, cooper pairs may tunnel through the small hotspot. On the other hand, due to the relatively “cold” hotspot, it would need more photon energy to “over heat” the hotspot to completely transit the nanowire from superconducting to normal state. Fig. 4 (b) and (c) illustrate the photon count rates under different incident laser power at wavelengths of 785 nm and 1550 nm, respectively, for varying bias currents. By fitting the photon count rate dependency in a log–log scale, the slopes indicate that the probability of detecting a 785-nm photon increases linearly with the mean number of photons (Fig. 4 (b)), demonstrating the operation of the detector in the single-photon region [38]. Additionally, Fig. 4 (c) indicates that two or three 1550-nm photons are required to trigger a detection event when $I_{b}<0.8 I_{S W}$. Even in the higher bias range ($I_{b}>0.8 I_{S W}$), the fitted slope remains greater than 1, suggesting the occurrence of multi-photon detection events. Typically, such cases of multi-photon detection significantly increase the timing jitter of the device, as indicated by a measured value of 644 ps for 1550-nm photons, obtained from the full width at half maximum of the time correlated photon count histogram.

In this study, we successfully demonstrated the deposition of ultra-thin Al films with a thickness of 4.2 nm. By employing a protective layer of 3.1-nm AlN and 3.3-nm-SiO₂, the fabricated Al-nanostrips exhibited a critical temperature of 2.4 K, significantly higher than that of bulk Al. We fabricated an Al-SNSPD with an area of 50 × 50 μm², a strip width of 178 nm, and a pitch of 800 nm. The device can operate in the single-photon regime at visible wavelengths, and the IDE showed a saturation plateau at the wavelength of 405 nm. However, due to the relatively large diffusion coefficient and coherence length, the IDE for 1550-nm photons remains limited and detection events might be triggered by two or more photons, resulting in large timing jitter. Nevertheless, there is potential for enhancing the detection performance of the Al-SNSPDs by further narrowing the strip width and optimizing the superconducting properties of Al nanostrips.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This work was supported by Strategic Priority Research Program (B) of the Chinese Academy of Sciences (XDB0580000), Youth Innovation Promotion Association of the Chinese Academy of Sciences (2021230), Shanghai Science and Technology Development Foundation (21YF1455500); Science and Technology Commission of Shanghai Municipality (2019SHZDZX01); National Natural Science Foundation of China (61801462, 61827823, 61971408).