1. Introduction

Integrated quantum photonics offers great benefits for quantum information processing and communication [1–3]. With an increasing complexity of quantum photonic applications, a plurality of electro-optic components must be operated on a single chip [4–6]. These electro-optic components, such as modulators and single photon detectors, require additional ancillary electronic components for biasing, amplification, and signal transmission. Intermediate electric components will degrade the performance of the entire photonic setup by introducing noise, heatload, or bandwidth limitations [7]. To circumvent this, it may be highly beneficial to replace the intermediate electronic connections with optical links, which electrically decouple the photonic processor and the driving electronics.

Superconducting nanowire single photon detectors (SNSPDs) are a key enabling technology for quantum optical applications due to their near unity detection efficiencies [8,9], low timing jitter [10], and low dark count rates [11]. Integrating these detectors in advanced photonic circuits is non-trivial since SNSPDs require operation temperatures below 4 $\mathrm {K}$ [12]. The cryogenic environment introduces additional challenges when interfacing the detectors with other electro-optic components. The output voltage of the SNSPD is around 1 $\mathrm {mV}$ [12] in the typical configuration with a 50 $\Omega$ shunt resistor. Therefore, additional electrical components are needed to amplify and transmit the SNSPD signal to the readout electronics. In the typical configuration the readout and bias electronics are outside the cryostat, requiring electrical interconnects between the cryogenic and room temperature environments. Replacing these electrical interconnects with electro-optic components and optical fibers enables electronic decoupling of the cryogenic electro-photonics from the external environment. An all-optical interconnect for superconducting photonics must therefore deliver the operation power and transmit signals to and from the decoupled circuit.

In recent years, cryogenic electro-optic modulation has been investigated for photonic circuits across a variety of platforms [13–18]. In particular, the electro-optic readout of superconducting single photon detectors both with and without intermediate amplifiers have been investigated [14,19,20]. In these applications, the click signal of an SNSPD detection event is delivered to an electro-optic modulator, modulating an optical throughput which is subsequently read out at room temperature. The operation voltage for intensity modulators have been reported to be in a range from 100 $\mathrm {mV}$ to 10 $\mathrm {V}$ at cryogenic temperatures [13,14,16,18]. Bridging the gap between the low amplitude output of the superconducting detector to these voltages is non-trivial in a cryogenic environment.

In this paper, we present an alternative method to generate larger detection signals from an SNSPD, which drives an electro-optic modulator directly, to achieve an all-optical readout. The SNSPD works on the principle that a bias current is converted to a voltage signal through a resistive load. This resistive load is created by an impinging photon which breaks the superconductive state in the nanowire, followed by Joule heating creating a so-called hotspot. In a typical operation, a shunt resistor is introduced to redirect the bias current following a detection event. This prevents further Joule heating, allowing the nanowire to dissipate the heat from the hotspot and return to the superconductive state. In our configuration, shown in Fig. 1, we omit the shunt resistor allowing the resistive hotspot to grow to larger resistances due to self-heating. This method can generate resistances in the order of a few tens of $\mathrm {k}\Omega$ such that a signal voltage of 30 $\mathrm {mV}$ is created. This is a significant increase compared to the 1 $\mathrm {mV}$ signal from the conventional method. Nevertheless, in our configuration, the SNSPD remains “latched” in its normally resistive state, i.e. the hotspot does not have the ability to reset itself. Indeed, the latching dynamics and resulting output voltage of a nanowire is a largely unexploited effect [21,22].

 

Fig. 1. Layout of the all-optical operation of a Superconducting Nanowire Single Photon Detector (SNSPD). The cryogenic photodiode is illuminated and generates a photocurrent. When a photon impinges on the SNSPD it becomes resistive and voltage is created. To readout the detection signal, the voltage pulse is transmitted onto the cryogenic modulator. An intensity modulation is then readout at room temperature by a photodiode and timetagger.

Download Full Size | PDF

A key aspect of our method is the ability to actively reset the hotspot, ideally with a cryogenic current source. We achieve this by modulating the bias current optically with a current generated by a photodiode [23]. We have previously shown that this method shows no significant deviation from conventional biasing [23,24]. Thus the photodiode power delivery approach combines the key aspects of providing a bias current for the SNSPD and generating the electrical driving power of the modulator. This versatility in generating the SNSPD bias, sustaining the hotspot, and supplying the supply for the electro-optical modulation enables all-optical operation of the SNSPD.

2. Methods

Once a photon impinges on the nanowire, a hotspot is formed and the electrical power is converted to heat by the resistive nanowire through Joule heating, as depicted schematically in Fig. 2(a). This Joule heating results in heating of adjacent superconductive regions, increasing the hotspot size. The resistive region of the nanowire grows until the electrical power provided is equal to the power dissipated due to the temperature difference between the nanowire and the substrate [25,26]. At the power equilibrium the maximum output voltage is reached, generated by the current flowing trough the resistive nanowire. The output voltage of the latched SNSPD is therefore dependent on the power dissipation of the SNSPD and the electrical power provided by the photodiode. The key to achieve the photodiode bias is to use a diode material which is still responsive at cryogenic temperature which has been investigated for our off-the-shelf photodiodes by [23,27,28]. In addition to these investigation in the conventional optical bias of an SNSPD, we characterize the power conversion capabilities and the output voltage of the photodiode in combination with the SNSPD.

 

Fig. 2. a) The optical input power is converted by the photodiode to an electrical power which is then converted to heat by Joule heating. b) Voltage and power output of the cryogenic photodiode under different illumination powers and load resistances. The photodiode is illuminated with a cw-laser at 1530 $\mathrm {nm}$ and operated at 1 $\mathrm {K}$. c) Voltage output of the combined operation of the SNSPD and bias photodiode, under different illumination powers to the bias photodiode at 1530 $\mathrm {nm}$.

Download Full Size | PDF

 

Fig. 3. Layout of the Michelson-interferometer realized in a photonic circuit. A dual core single mode fiber pigtail couples light into the waveguides and returns the reflected light for the readout. These waveguides are fabricated by titanium indiffusion in z-cut lithium niobate. An integrated beam splitter splits the input light into two paths which are then reflected at the endface to interfere again at the beamsplitter. Electrodes on the surface of one beam arm introduce a phase difference when a voltage is applied. b) Voltage sweeps of the modulators at 1530$\mathrm {nm}$. The measured intensities are normalized to the maximal out power per sweep. The $V_{\pi }$ voltage voltages are 5.9$\mathrm {V}$ at 1$\mathrm {K}$ and 6.6$\mathrm {V}$ at 300$\mathrm {K}$. The $V_{\pi }$ voltage is acquired by fitting a sin-function to the acquired data.

Download Full Size | PDF

2.1 Optical power supply

The optical to electrical conversion efficiency of the photodiode depends on the load resistance, in our case the SNSPD. The Superconducting detector is fabricated from a tungsten silicide nanowire (WSi) thin-film [29] and has been characterized in the bias dependent detection efficiency to up to 83${\% }\pm 5{\% }$ by Thiele et al. [23]. The SNSPD resistance can vary from no resistance when superconducting to a normal resistance of 5.5 $\mathrm {M}\Omega$ when the entire nanowire is normally resistive. To determine the generated voltages and dissipated powers, we cooled the photodiode down to 1 $\mathrm {K}$ and attached different load resistors at room temperature according to the inset in Fig. 2(b). The photodiode is illuminated through a single mode fiber. To do so, we aligned the fiber perpendicular to the active area of the photodiode. The fibers position is optimized by maximizing the converted current while light is transmitted through the fiber. The fibers position is then permanently fixed in place by a UV-adhesive. The photodiode is then operated at an optical input power of about 6$\mathrm{\mu} \mathrm {W}$. At this input power, a nominal bias current of 4$\mathrm{\mu} \mathrm {A}$ for the SNSPD is generated in shunted operation, given that the responsivity of the photodiode is approximately 0.65$\mathrm {A/W}$ [23].

In our characterization, we keep the input power stable and vary the load resistor in a range from 10$\Omega$ to 1$\mathrm {M}\Omega$, while measuring the current through and voltage over the resistor. As a result, we generate increasing output voltages by increasing the load resistors, until the output voltage saturates at approximately 500$\mathrm {mV}$, as can be seen in Fig. 2(b). The generated electrical power reaches a maximum power point at a resistance around 100$\mathrm {k}\Omega$, before reducing again for higher loads due to a reduction in the generated current. This maximum power point shifts to lower resistances when the optical power on the photodiode is increased. This load characterization shows that the cryogenic photodiode can provide a bias current in the shunted operation when the nanowire is superconductive and a voltage beyond 500$\mathrm {mV}$ when a resistive hotspot is created.

The hotspot in the nanowire is not a static resistor and is expected to change in size depending on the electrical supply power. Therefore, the output voltage will depend on the optical input power of the photodiode. However, we cannot increase the input power to the photodiode indefinitely to generate a maximal optical response because the SNSPD must be operated at a nominal bias current of 4$\mathrm{\mu} \mathrm {A}$. Therefore, we need to determine the voltage at this operation point when combining the SNSPD directly with the photodiode. To characterize the devices before the all-optical demonstration, we combined both devices on a single stage in the cryostat at 1K and read out the output voltage at different input powers the photodiode, as shown in the inset of Fig. 2(c). The SNSPD cannot selfreset out of a latched state due to the omission of the shunt resistor. Both devices are connected with a coax cable to measure the voltages at room temperature. Figure 2(c) shows that voltages with the photodiode and SNSPD are generated and saturate above approximately 550$\mathrm {mV}$ when we sweep the input power from 700$\mathrm{\mu} \mathrm {W}$ to 0.1$\mathrm{\mu} \mathrm {W}$. The SNSPD returns from a resistive state to a superconducting state below 0.1$\mathrm{\mu} \mathrm {W}$ since the hotspot cannot be maintained by the input power. In the all-optical operation of the SNSPD, we provide an input power of 6$\mathrm{\mu} \mathrm {W}$ to generate the nominal bias current for the SNSPD. At this power level a output voltage of 31$\mathrm {mV}$ is reached, generated by a 40$\mathrm {k \Omega }$ nanowire resistance. This output voltage is a significant increase in relation to a click signal of below 1$\mathrm {mV}$ in the conventional method.

2.2 Electro-optic readout

The readout of the generated click signal is realized by modulating the intensity with an electro-optic modulator operated at cryogenic temperatures. We choose a titanium in-diffused lithium niobate electro-optic modulator because of its proven operation at cryogenic temperatures [15,18]. Furthermore, we can achieve fiber-to-fiber optical coupling up to 43% with single mode fibers even at 1$\mathrm {K}$ [15]. We implement an integrated Michelson interferometer consisting of a directional coupler, a reflective endface coating, and electrodes placed on one channel for the modulation, as it can be seen in Fig. 3.

To achieve high-efficiency optical access to our photonic readout circuit, we use a dual-core fiber ferrule with a core separation of 127$\mathrm{\mu}\mathrm {m}$, to match the waveguide separation at the endface. To attach the fiber ferrule to the lithium niobate sample, the fiber position is optimized by maximizing the reflected power through the fibers and waveguides. Subsequently, the fibers are permanently fixed by a UV-adhesive (NOA81) at the fiber-to-sample interface. The sample and attached fibers are then placed on a mounting block for mechanical stability, whereby the lithium niobate is fixed to the surface with vacuum grease (Apiezon N) and the dual core single mode fiber is fixed with UV-cured adhesive to the holder.

 

Fig. 4. Voltage sweeps of the modulators at 1530$\mathrm {nm}$. The measured intensities are normalized to the maximal out power per sweep. The $V_{\pi }$ voltage voltages are 5.9$\mathrm {V}$ at 1$\mathrm {K}$ and 6.6$\mathrm {V}$ at 300$\mathrm {K}$. The $V_{\pi }$ voltage is acquired by fitting a sin-function to the acquired data.

Download Full Size | PDF

We achieve a fiber-to-fiber efficiency of about 45% at 1530nm at room temperature, which reduces to 27% at cryogenic temperatures. The change in the coupling efficiency can be mainly attributed to mechanical changes in the adhesive bond during the cooling process. The supported optical mode in our waveguide have a large mode overlap with standard single mode fibers in the TM polarization of 85% and with the TE polarization of 92% [30]. We can expect an overall coupling efficiency between the waveguides and single mode fiber of 51% given a maximum mode overlap of 85%, a coating reflectivity of 96% and a fiber-to-waveguide reflectivity of 3%. The linear losses of the waveguides with the electrode structure are approximately 0.1 $\mathrm {dB/cm}$ which is characterized with a Fabry-Pérot method [31] and has been investigated for this type of modulators at cryogenic temperatures [18]. In summary, we achieve an overall device insertion loss of 3.5 $\mathrm {dB}$ at room temperature and 5.6 $\mathrm {dB}$ under cryogenic conditions.

The voltage signal generated by the photodiode and SNSPD is transmitted to the electrodes of the lithium niobate modulator. The voltage induces an electric field in the waveguide and changes the refractive index via the electro-optic effect. The electro-optic modulator is placed in an integrated Michelson interferometer such that the refractive index change will induce a relative phase change with the second beam splitter arm. Both beams interfere again after the reflection at the reflection coated endface. As a result, the reflected power of the Michelson interferometer is modulated dependent on the applied voltage which is used for the optical readout. The Michelson-interferometer layout is chosen to reduce the required voltage to introduce a phase shift because the phase shift is accumulated in both the forward and backward propagation through the modulator. The integrated Michelson interferometer is realized by titanium indiffused waveguides in z-cut lithium niobate. The endface reflector is realized by a dielectric reflective coating matched to the waveguides refractive index with a reflectivity of 96% [32]. More details on the fabrication and waveguide characterization are given in the Supplement 1. We characterized the electro-optic modulator by acquiring the intensity of the output by sweeping the voltage, as shown in Fig. 4. The modulation voltage required to switch the intensity from maximum to minimum is extracted to be 5.9$\mathrm {V}$ at room temperature at an operation wavelength of 1530$\mathrm {nm}$. The $V_{\pi }$-voltage increases to 6.6 $\mathrm {V}$ at 1 $\mathrm {K}$ due to the temperature-dependent electro-optic coefficient which has been investigated previously [18,33].

3. Results

We combine the operation of an opto-electronic bias of the SNSPD with an electro-optic readout. To do so, the photodiode, SNSPD, and lithium niobate modulator are connected on a single cold stage of the cryostat and cooled down to a base temperature of 1$\mathrm {K}$. For the operation of the SNSPD a 6$\mathrm{\mu} \mathrm {W}$ cw-laser is externally on-off pulsed with a duty cycle of 35$\mathrm{\mu} \mathrm {s}$, as illustrated in Fig. 5(a). A detection event occurs 2.9$\mathrm{\mu} \mathrm {s}$ after the illumination of the photodiode. As a result, a hotspot starts to grow, resulting in an increasing voltage delivered to the modulator. To read out the generated detection signal, light is transmitted through the modulator and the optical response is measured at room temperature. A typical measurement trace is shown in Fig. 5(b). The resulting optical response signal has a 90% rise time of 11$\mathrm{\mu} \mathrm {s}$. Once a photon is detected the superconductor becomes resistive and no subsequent photons can be detected, because the current through the nanowire is strongly reduced. The SNSPD is reset by switching off the light to the biasing photodiode after 18.9$\mathrm{\mu} \mathrm {s}$, which results in an optical response with a fall time of 11$\mathrm{\mu} \mathrm {s}$. To maximize the optical response, we tuned the operation wavelength of the modulator to 1530$\mathrm {nm}$ and optimized the input polarization with a fiber polarization controller while a power of 3.5$\mathrm {mW}$ is transmitted through the modulator.

 

Fig. 5. a) Optical modulation of the bias power for the cryogenic photodiode to operate the SNSPD. The modulation period is 35$\mathrm{\mu} \mathrm {s}$. b) The optical response of the readout photodiode after the electro-optic modulator placed at room temperature. A photon impinges on the SNSPD at 2.9$\mathrm{\mu} \mathrm {s}$, a voltage is generated and switches the electro-optic modulator. To reset the SNSPD, the optical power to the cryogenic photodiode is switched off after 18.9$\mathrm{\mu} \mathrm {s}$. The nanowire becomes superconducting such that the generated voltage over nanowire resistance is reduced to zero. c) Histogram of the acquired countrate of the all-optical SNSPD operation. This is the difference in the countrate of the signal photons and measured dark counts. The error bars are determined by the counting errors of the count rates. The inset shows the total countrate of the measurement with and without a single photon input to the SNSPD. The mean photon number per pulse is 1.17$\pm 0.06$.

Download Full Size | PDF

The optical response is read out with a photodiode at room temperature, resulting in a signal with an amplitude of 6.9$\mathrm{\mu} \mathrm {W}$, given that the photodiode has a responsivity of 0.4 $\mathrm {mV}/\mathrm{\mu}\mathrm{W}$ and the readout amplitude is 2.76$\mathrm {mV}$, as shown in Fig. 5(b). In comparison, we expect an optical response of about 7.0$\mathrm{\mu} \mathrm {W}$, given a click signal amplitude of 31$\mathrm {mV}$, a $\mathrm {V_\pi }$ of the modulator of 6.6$\mathrm {V}$ and a fiber-to-fiber efficiency of the modulator of 27%. This shows the core principle that we can all-optically bias an SNSPD and read out the detection signal.

As a next step, we characterize the single photon response of the superconducting detector. To do so, we operated the SNSPD with our all-optical biasing method and acquired the countrate of the SNSPD with a single photon level input. To do so, we used a pulsed laser with a pulse width below 1$\mathrm {ps}$ at 1545$\mathrm {nm}$. These pulses are transmitted onto a calibrated reference detector and attenuated. The pulses are attenuated to a mean photon number of 1.17$\pm 0.06$ photons per pulse. This optical input is then pulsed synchronously with the optical bias operated with a 35$\mathrm{\mu} \mathrm {s}$ on-off period. The timing delay of the generated click signals are then recorded in relation to the on-off signal.

Based on these results, we seek to reduce the optical throughput of the modulator to reduce both heatload and scatter. This depends on the minimal acquisition power of the detectors used to acquire the click signals. Small Formfactor Pluggable modules (SFP) are ideal candidates to measure small changes in intensity. The nominal minimal input pulse power of the SFP module is about -25 $\mathrm {dBm}$ at 1550nm, which will generate an electrical output pulse of above 100$\mathrm {mV}$ (Finisar FWLF-1519-7D-59). In our experiment we reduced the input power to the modulator to about 68$\mathrm{\mu} \mathrm {W}$ ($\sim$-12 $\mathrm {dBm}$) such that the click signal has an amplitude of about 150$\mathrm {nW}$($\sim$-38 $\mathrm {dBm}$). When acquiring these signals with the SFP-module, the generated output signal of the module was reduced to about 20$\mathrm {mV}$. Additional high frequency noise in the output signal is also introduced at these low input powers, which we reduced with a 1$\mathrm {MHz}$ low-pass filter. This added noise is negligible at higher input powers and higher signal count rates.

The click signals with a single photon input are then acquired with a time tagger and displayed as a histogram in Fig. 5(c). During the measurement, signal photons as well as scattered light from the electro-optic modulator are detected. To extract the single photon events we performed a background measurement without a single photon input. By subtracting this background from the results with a single-photon input, a clear peak in the countrate at a delay of 11.3$\mathrm{\mu} \mathrm {s}$ with a full width half maximum of 1.5$\mathrm{\mu} \mathrm {s}$ can be seen, as shown in Fig. 5(c). The countrate in this histogram is negative after the main peak because the previous signal photons decrease the detection probability for subsequent background photons due to the SNSPD latching. The peak shows clearly that the SNSPD is sensitive to a single photon input in our all-optical operation method.

4. Discussion

The proof-of-principle devices presented here show promise for interfacing superconducting photonics in new performance regimes and application spaces, however there are still some non-idealities that can be improved in future work. A strong background of dark counts is present, mainly generated by scattered light introduced by the optical bias and readout of the SNSPD. Due to the lack of a self-reset, the detector clicks on the first photon that it measures, and none after that. Therefore, reducing the number of premature latching clicks from noise is necessary. In the present configuration, the components are mounted in the cryostat without intermediate shielding, the application of which would reduce the dark count rates significantly. Furthermore, one could exploit differences in the spectral sensitivity between the bias photodiode and the SNSPD to further reduce noise.

The SNSPD was operated with a repetition rate around 28$\mathrm {kHz}$, limited by the rise and fall time of the click signals. We attribute this to the charging dynamics of the capacitance of the electro-optic modulator and the growth of the resistive hotspot. Faster rise times can be achieved by introducing a parallel resistance to reduce the hotspot size. Introducing a parallel resistance also reduces the output voltage and hence the intensity modulation. Therefore, there exists a trade-off between the output voltage and the resulting rise time.

The bias current of the SNSPD is pulsed in this operation to reset the detector after a detection event. The current pulsing with cryogenically integrated current sources has been investigated before for an electronic gating of detection events [34]. The investigations show that the risetime in the detection efficiency is limited by the inductance and is upper bound by the deadtime of the SNSPD. We previously characterized this SNSPD and measured a deadtime of 100 $\mathrm {ns}$ [23].

The system jitter was acquired in the acquisition of the histogram in Fig. 5(c). The peak width in the histogram is 1.4$\mathrm{\mu}\mathrm {s}$, as shown in Fig. 5(c). A previous publication characterized and showed a timing jitter of approximately 500$\mathrm {ps}$ for the used SNSPD [23]. The increase in the jitter can be mainly attributed to the acquisition photodiode (SFP-module) which is optimized for higher input power levels and faster input signals.

This bias and readout method has no effect on the internal detection efficiency of the SNSPD. We have previously shown that an optical bias and a conventional bias method achieve the same detection efficiencies [23]. Nevertheless, the lack of a self reset means that "true" counts may be lost if the detector has previously been triggered by a dark count, therefore reducing the noise also plays an important role in increasing the system efficiency. This could be further optimized by careful synchronization of bias and readout light sources with respect to the expected arrival time of the single photons.

The all-optical operation promises a lower power dissipation in the cryostat in comparison to the typical operation of an SNSPD. The heat conduction of the transmission line can be computed from the dimension and material properties [35]. The used cryostat (Photonspot) in this work has two thermal stages at 45$\mathrm {K}$ and 4$\mathrm {K}$ with transmission lines with the length of 20$\mathrm {cm}$ and 30$\mathrm {cm}$, respectively. The thermal conduction of a typical coax-line and single mode optical fiber can be then computed [36,37]. A coaxial cable conducts 170$\mathrm {mW}$ between the 300$\mathrm {K}$ and 45$\mathrm {K}$ as well as 51$\mathrm {mW}$ between 45$\mathrm {K}$ and 4$\mathrm {K}$. As a direct comparison, a standard single mode fiber conducts 24$\mathrm{\mu} \mathrm {W}$ between the 300$\mathrm {K}$ and 45$\mathrm {K}$ as well as 1.2$\mathrm{\mu} \mathrm {W}$ between 45$\mathrm {K}$ and 4$\mathrm {K}$ [37]. The total dissipated thermal energy is significantly lower in the all-optical operation and is dominated by the optical operation power.

The optical readout is realized with a titanium in-diffused lithium niobate modulator which achieves a modulation strength of 13.2$\mathrm {Vcm}$ with a 2$\mathrm {cm}$ long electrodes. This platform has the benefit of a large coupling efficiency with single mode fibers. Cryogenic modulators with a higher waveguide confinement and smaller footprint could be used in the future such a thin-film lithium niobate modulators [16] or ring resonators [14].

Independent of optical biasing and readout, exploiting the latched state of an SNSPD is an important method to step up the click signal. For example, amplitudes of 10$\mathrm {mV}$ could be used to drive a Schmitt-trigger, which can enable feedforward modulation with electronic amplifiers. With this technique, the initial click signal is significantly increased, such that cryogenic low noise amplifiers can be avoided. The key to this method is to integrate a current source which can reset the latched SNSPD after the detection.

By connecting the output of the detector to a modulator, this method is an important step towards all-optical feedforward modulation [38] at cryogenic temperatures, which is integral to quantum photonic one-way computing [39]. This can be achieved by a direct matching of the generated output voltage of the SNSPD signal and the switching voltage of the modulator. In this technique, a photon can be detected by the single photon detector switching the electro-optic modulator from one state to the other. Future work is needed to match the detector output and modulation voltages at cryogenic temperatures. In addition, the rise time of the click signal must be improved to minimize optical delays when using feedforward modulation with low latency processes.

Since optical fibers inherently have a lower thermal conductivity than coaxial cables, this can significantly reduce the heatload on the cryostat [14,24]. The passive heatload of coaxial cables can be mitigated by thermal anchoring in the cryostat but are a major contributor to the thermal load on a cryostat [7]. In our operation of the SNSPD, three independent optical fibers are directly connected from room temperature to the photodiode and electro-optical modulator without the need of intermediate thermal anchoring. The active heatload of these electro-optic devices are 6$\mathrm{\mu} \mathrm {W}$ for the photodiode and 68$\mathrm{\mu} \mathrm {W}$ for the electro-optic modulator. The active heatload of the modulator can be reduced further to a few $\mathrm{\mu} \mathrm {W}$ by reducing the modulation voltage $V_{\pi }$ and increasing the detection signal voltage, such that the full throughput intensity is modulated. In addition, wavelength-division multiplexing can also be used for the bias and readout to operate multiple SNSPDs in parallel to reduce the passive heatload even further [24].

5. Conclusion

In summary, we have realized an all-optical interface for quantum photonic applications. The interface provides the operation power for a superconducting single photon detector via a cryogenic photodiode. In addition, the detection signals are readout optically via an electro-optic modulator at cryogenic temperatures. The all-optical operation shows promising techniques for the combined operation of superconducting electronics and photonics circuits, which are electrically isolated from their driving circuitry. Increasing the signal amplitudes and modulation capabilities of the opto-electronic components enables further applications such as feed-forward. Furthermore, the all-optical operation of the SNSPD achieves a low power operation of the SNSPD by providing only a total operation power of 75$\mathrm{\mu} \mathrm {W}$.