1. Introduction

Since the advent of optical quantum information applications, where a quantum bit (qubit) of information can be encoded in many different degrees of freedom of single photons, efforts have been placed to develop the fundamental parts required for photonic quantum technologies [1], named single photon sources, linear-optic components and single photon detectors. The integration of these three essential components in a monolithic platform would address the scalability requirements needed for the realization of large-scale quantum-optic experiments. The physical properties of the chosen optical material should allow the generation of non-classical state of light, fast light modulation and the integration of efficient single photon detectors. All these requirements are not trivial to fulfill and hybrid approach have been proposed [2,3].

Among different candidates, SiC has raised an increasing interest due to its outstanding optical properties combined with the presence of quantum emitters. In addition to a wide bandgap of 2.3eV and a high refractive index of 2.6, that could be exploited for the realization of small footprint photonics operating up to the near-visible spectral range, SiC has a non-centrosymmetric crystal structure that could allow second order non-linear process and fast Pockels modulators. SiC divacancy centers have proven to be a valuable resource for quantum optics as their long coherence time in isolated electron spins [4,5] that could be coherently controlled even at room temperature [6]. Even though all the most common SiC polytypes presents this quantum emitter [7], only its cubic form can be grown eteroepitaxially on top of silicon substrates with high quality, increasing the versatility of this platform. This has favored the development of photonic crystal cavities [8] that could be exploited to enhance the interaction of SiC color centers with light [9,10]. Linear components have been realized at telecom wavelengths in 3C SiC using a suspended technology [11] whose high confinement has allowed the demonstration of non-linear process [12]. Recently, the exploitation of a new platform, where 3C SiC is bonded on an insulator, has shown competitive results with common photonic platforms in terms of propagation losses in confined structures [13], increasing further the potential of this material.

In order to fully exploit the exceptional properties of SiC, the integration of an efficient single photon detector in this platform is required. The detector should ideally operate at NIR wavelengths to match the emission wavelengths of SiC divacancy centers (1100 nm) and it should be ideally efficient up to telecom wavelengths, to profit from the technology existing in this frequency range. Moreover, recently single photon emission from 3C SiC close to the telecom C-band has been demonstrated [14,15]. Superconducting nanowire single photon detectors (SNSPDs) are the only able to reach high system detection efficiency (SDE) at telecom wavelengths with low dark count rate (DCR) and jitter [16], and have been integrated in waveguide on different materials [17–20].

The similar lattice constant between SiC and NbN has shown advantages in term of superconducting properties [21], but the fabrication of superconducting nanowires on top of SiC was hampered from the generally poor surface quality of SiC layers and the difficulties in micromachining this material, capability required to achieve the self-aligning fiber-to-detector [22] or to implement photonic circuits.

Here we measure SDE and DCR of meandered superconducting nanowires realized on top of a chemical-mechanical polished 3C SiC layer, thanks to a new measurement approach that makes use of two alignment mirrors and a single-mode fiber array (FA). This alignment approach allows the testing of multiple SNSPDs fabricated on top of less fabrication-friendly materials, without the use of expensive and bulky cryogenic positioners nor the micromachining of the substrate. The 3x3 µm² active area of the fabricated SNSPDs allowed a quasi-saturated detection efficiency at telecom wavelengths of up to 6% at the operating temperature of 2.9 K, in close agreement with the expected value from numerical simulation.

2. Design

The SDE can be factorized in three independent efficiencies η_a, η_i and η_c where η_a identifies the optical absorption, η_i the internal efficiency and η_c the coupling efficiency. In order to test the compatibility of the fabrication process of narrow NbN nanowires on top of a bare SiC layer and with the aim to achieve a high fabrication yield, we reduced the nanowire length by designing SNSPDs with 3x3 µm² active area at the expenses of a lower η_c as most photons inside the fiber mode fall outside the meander area. By integrating the normalized field intensity of the fiber mode that falls within the active area, we can estimate a maximum η_c≈20% in the case of a fiber perfectly aligned to the nanowire meander. Figure 1(a) shows η_c computed as a function of the shift of the fiber in one direction.

 

Fig. 1 (a) Fiber-to-detector optical coupling as a function of the shift in one direction of the fiber position. (b) Simulated SNSPD optical absorption η_a vs the wavelength for NbN with extinction coefficient κ = 5.82 (black squares), κ = 7.57 (red circles) and in the case that the silicon layer is replaced with air (blue triangles). Inset: sketch of the layer stack composing the device made of 600 nm-thick PMMA protective layer, 6nm-thick NbN meander with filling factor of 40%, 9.5 µm-thick SiC layer and silicon substrate (not in scale).

Download Full Size | PDF

In order to estimate η_a, we performed 2D FDTD simulations (Fig. 1(b)) of the wave propagating through the different layers (schematized in the inset of Fig. 1(b)). We assumed the refractive index of the NbN layer to be n_NbN = 5.23 and an extinction coefficient κ = 5.82 at 1550 nm [23]. For SiC and Si layers, we considered instead n_SiC = 2.56 and n_Si = 3.5, respectively. Figure 1(b) shows the optical absorption as a function of the wavelength (black squares with label κ = 5.82), where maxima of 23.84% and minima of 12.10% fall at 1550nm and 1527nm, respectively. This behavior is due to the reflections occurring at both the PMMA-SiC and SiC-Si interfaces that acts as a cavity and cause the non-uniform spectral response of the SNSPD. From numerical simulation, we expect a maximum SDE = η_i·η_c·η_a≈4.8% considering a unitary internal efficiency η_i.

3. Results

Detectors are fabricated using a 9.5 µm-thick 3C SiC layer [24], heteroepitaxially grown on top of a silicon substrate and chemical-mechanical polished (RMS roughness below 1nm). A 6 nm-thick NbN film is deposited on the sample by DC magnetron sputtering at T = 550 °C in a gas mixture of N₂ + Ar (with 22% N₂). We used a first step of electron beam lithography to define via lift-off the pad area together with the two alignment mirrors (Fig. 2) using PMMA as stencil mask and a subsequent deposition of 10nm Ti and 60nm Au layers by electron gun deposition. We defined the nanowires with a second step of electron beam lithography using a 100nm-thick HSQ resist and a final RIE etch in a fluorine-based chemistry. Finally, a layer of PMMA (600nm thick) is spun on the sample to protect the surface of the device. Several devices like the one shown in Fig. 2(a) were fabricated thanks to their small footprint compared to devices based on the self-aligning fiber-to-detector coupling, whose space requirement greatly reduce the ultimate throughput [22].

 

Fig. 2 (a) SEM images of the device consisting of two alignment mirrors and four SNSPDs. The distance between detectors and mirrors is fixed to 127 µm to match the pitch of the FA.(b) A 60nm-wide superconducting nanowire folded as a 3x3 µm² active-area meander.

Download Full Size | PDF

Figure 2(a) shows the final device, composed by 4 SNSPDs and 2 alignment mirrors at both sides. The distance between the SNSPDs and the mirrors is fixed to 127µm to match fiber-to-fiber pitch of a 6-channel FA composed of standard SMF-28 telecom fibers. We performed the alignment at room temperature thanks to the first and sixth fibers of the FA where the light coming from a CW laser is reflected from the substrate surface and collected from the same fiber. We obtained a maximum of the reflected light when the fiber is placed on top of the Au mirror (13µm diameter). Once both the first and the sixth fiber were aligned on the corresponding mirrors, the FA is glued to the sample using a cyanoacrylate adhesive. The sample is successively mounted in a GM cryostat and cooled down to 2.9 K base temperature. With this procedure, fibers from 2 to 5 of the FA are aligned on top of the 4 SNSPDs that can be tested independently.

In addition to 80 nm-wide SNSPDs, we fabricated detectors with a narrower nanowire (60nm). Figure 2(b) shows a SEM image of the 60nm-width SNSPDs, whose transition temperature (T_c) is 8K (Fig. 3(a)). This is a substantial difference from previously reported results [21], where a T_c record of 11.8K was achieved in 4nm-thick NbN unpatterned films thanks to the similarity of its lattice constant with the 3C SiC one. We attribute this difference to the lower deposition temperature employed in our process and to the patterning of the film in narrow nanowires. As can be seen from the I vs V curve of Fig. 3(b), the 60nm-wide SNSPD has a critical current of 5.8µA at the temperature of 2.9K where the device is operated.

 

Fig. 3 (a) R vs T curve of the 60nm-wide SNSPD showing a transition temperature of 8 K. (b) IV curve of a 60nm-wide SNSPD with a critical current of 5.8 µA measured at 2.9K.

Download Full Size | PDF

For the electro-optical characterization of the fabricated device, we employed a CW laser at telecom wavelength (1550 nm) with the light attenuated at the single photon level (0.1pW of input power at the fiber). We used two MiniCircuits Bias-Tee to filter and separate RF and DC signals to and from the SNSPD. The electronic pulses generated by SNSPDs are amplified by two room-temperature RF amplifiers with a 49 dB total gain in the bandwidth 0.1-500 MHz. Figure 4(a) and 4(b) show the SDE [25] and DCR for the 80nm and 60nm–wide SNSPDs, respectively, for two different devices (D1 and D2). The bias current of each meander is normalized to the value of its own switching current I_SW corresponding to the value of I_SW-D1 = 5.08µA and I_SW-D2 = 4.45µA, for 60 nm-wide nanowires, and I_SW-D1 = 8.93µA and I_SW-D2 = 7.5µA, for 80 nm-wide nanowires. The SDE and DCR of the 80nm-wide detectors have the same trend and reach a maximum of SDE of ~2.8% at a DCR of 10 kHz. This results show a good reproducibility of both the coupling method and the detector characteristics.

 

Fig. 4 Optical characterization of two SNSPDs (D1and D2) made of 80nm (a) and 60nm (b) wide nanowires. Legend: SDE for D1 (red squares) and for D2 (green squares); DCR for D1 (black squares) and D2 (blue squares). Inset: enlarged view of SDE close to I_b/I_sw ≈1 (red squares) and its smoothed curve (dot-line) showing the onset of the SDE saturation for D1.

Download Full Size | PDF

As expected, for the narrower nanowires the internal efficiency is larger (Fig. 4(b)) and the SDE is saturated at I_b/I_sw> 0.98 (SDE = 5.5%, inset of Fig. 4(b)). However, this occurs at a bias current where the DCR is not negligible (20 kHz, due to the operating temperature of 2.9K) and the result do not match the one obtained for D2 (SDE = 3.1% at DCR = 2.5 kHz). In order to produce a pronounced saturation of the SNSPD at minor bias currents, to operate the detector at lower dark counts, the detector requires to have η_i = 1. This can be reached by further decreasing the nanowire width [26] (at the expenses of the fabrication yield of the SNSPD), by lowering the operating temperature [27] and by engineering the deposition conditions. The difference from the simulated SDE (4.8%) and the measured one (5.5%) can be attributed to different parameters. The main ones that might have an impact on η_a are the refractive index of the NbN layer and the non-ideal SiC-Si interface. Ellipsometric measurements of NbN refractive index have shown a great variability depending on NbN thickness and its deposition conditions [28], consequently affecting its absorption capability. As an example, Fig. 1(b) shows the case κ = 7.57 where the absorption at 1550nm is 27.7%, resulting in this way a SDE = 5.5%, in agreement with the experimental data. On the other side, the carbonization step needed for the SiC heteroepitaxial growth on top Si causes the formation of voids in the substrate [29] and affects the optical properties of the system as depicted from the blue curve of Fig. 1(b), where the Si substrate is replaced with a void (n_void = 1). The resulting η_a spans from 38.8% (1576nm) to 5.1% (1550nm) where the position of maxima and minima can be shifted depending on SiC refracting index n_sic.

To the best of our knowledge, the rise time of SNSPDs is not greatly affected by the substrate and therefore we do not expect the timing jitter to be much different from what reported in literature for NbN SNSPDs.

4. Conclusions

We presented the fabrication and test of SNSPDs on top of 3C SiC, which is a promising candidate to host in a monolithic way all of the fundamental building blocks required for photonics quantum technologies. In spite of a reduced active area and a lack of a bottom layer with high reflectivity, the SDE showed a quasi-saturated behavior at I_b/I_SW>0.98 with a maximum of 5.5% at 2.9K for 60nm-wide nanowires. This demonstrates that high-efficiency SNSPDs can be implemented on this material providing a smooth enough SiC surface. The fabrication of SNSPDs on top of SiC represents a fundamental step toward the implementation of complex quantum optics experiments in this material, involving solid state emitters, reconfigurable linear circuits and single photon detectors in the same platform. Thanks to the low surface roughness employed in in SiC waveguides [11,13], it is possible to integrate SNSPDs on top of confined waveguides either prior to the waveguide fabrication [18] or after the waveguide fabrication [20]. To couple the light to the meandered active area we have shown a novel approach that allows the electro-optical characterization of SNSPDs integrated on to less fabrication-friendly substrates without the use of bulky cryogenic positioners and expensive low-vibration cryostats. Compared to the self-aligning fiber-to-detector coupling, our technique do not require the substrate micromachining, allowing in this way a larger filling factor that ultimately increases the device throughput.