Resolving the number of photons on-chip

Photonic quantum information processing has attracted much attention over the last decade due to its potential applications in quantum key distribution (QKD), quantum simulations, and quantum computing. It offers the most secure form of cryptography as well as fast algorithms capable of solving classically intractable problems, such as the calculation of ground states for large molecules.¹ Currently, the state of the art is limited to few-quantum-bit operations due to both the complexity of the system and the coupling losses of free-space optics implementations. In such systems, sources and detectors are coupled to fibers or free-space beams, which introduce the biggest losses to the circuit. These losses can, however, be negated if all elements are integrated on the same chip. Photonic integrated circuits have been proposed to achieve this goal,² but current implementations are mostly limited to passive functionalities (e.g., couplers in waveguide circuitry). Full integration of single-photon sources, detectors, and passive circuitry in a quantum photonic integrated circuit (QPIC) is therefore desired. A dense QPIC such as this could generate tens of single photons, process them in a linear optical network, and detect the number of photons in the output channels. This functionality would enable, for example, the realization of quantum experiments that cannot be simulated classically.³

On the detection side, single-photon and photon-number-resolving detectors with high efficiency, low dark counts, high timing resolution, and short dead time (i.e., cycle time between each detection) are demanded for quantum photonic applications. Superconducting single-photon detectors (SSPDs)⁴ represent the only technology that currently combines all of these parameters at telecom wavelengths (1310 and 1550nm). Recently, we have shown that SSPDs based on niobium nitride (NbN) nanowires can be integrated on gallium arsenide (GaAs) ridge waveguides⁵ to create waveguide single-photon detectors (WSPDs). These detectors preserve the performance of SSPDs while enabling monolithic integration. WSPDs have recently been demonstrated based on both GaAs^{5, 6} and silicon,^7,8 paving the way to densely integrated QPICs that include detectors and sources.^{9, 10} Furthermore, by implementing two single-photon detectors on the same waveguide, we have demonstrated polarization-independent, integrated autocorrelators—which enable measurement of the correlation between electromagnetic fields—for on-chip measurements of the second-order correlation function. This function determines the correlation between pairs of photons (e.g., it can determine the residual two-photon emission that has come from a single-photon source).¹¹

We have developed another important building block on the road toward the development of a QPIC for quantum computing: an integrated waveguide photon-number-resolving detector (WPNRD). Unlike single-photon detectors, photon-number-resolving detectors are able to measure the number of photons incident in a single pulse. This functionality is necessary for a number of applications, including linear-optics quantum computing.¹² To date, two different photon-number-resolving detectors compatible with integration have been demonstrated. One is based on transition-edge sensors,^{13, 14} which show high detection efficiency at telecom wavelengths but are relatively slow (dead time in the μs range). The second, recently demonstrated by our group, is based on an array of superconducting nanowires.¹⁵

Our WPNRD comprises four superconducting NbN nanowires. The wires are connected in series, with each wire connected to an on-chip parallel resistance (with an overall resistance of 152Ω): see Figure 1(a) and (b). WPNRDs also preserve all the advantages of an SSPD in terms of short dead time (few ns), low jitter, high efficiency, and simple read-out (single output and room-temperature amplifiers) while allowing a photon-number-resolving capability of up to four photons per pulse. We integrated the four nanowires on a GaAs/aluminum gallium arsenide (GaAs/AlGaAs) ridge waveguide. All of the wires sense a single optical mode: see Figure 1(c). The tightly confined mode provides >90% absorptance along a waveguide a few tens of μm long. Our finite-element simulations have shown that the maximum absorptance of the central and lateral wires differ due to the difference in mode overlap: see Figure 1(c). However, this affects the ultimate fidelity of two-photon absorption by only ∼1%.¹⁵

Figure 1. (a) Schematic of our waveguide photon-number-resolving detector (WPNRD). The niobium nitride (NbN) nanowires are 5nm thick and 100nm wide, with a pitch length of 250nm. Four integrated resistances of titanium/palladium (brown) are connected in parallel to each nanowire. The ridge waveguide is 3.85μm wide and 260nm thick. (b) Scanning electron microscope image of a WPNRD. The yellow blocks are contact pads (G: ground; S: signal). (c) Calculated absorptance of the two central and the two lateral nanowires for transverse-electric (TE, black) and transverse-magnetic (TM, red) polarizations. Inset: Mode profile of a WPNRD for the TE polarization at 1310nm, showing where the mode sits and how confined it is in the gallium arsenide (GaAs) waveguide and how far it is pushed toward the NbN nanowires. SiO_x: Silicon oxide. AlGaAs: Aluminum gallium arsenide.

The device quantum efficiency (DQE), defined as the number of counts divided by the number of photons coupled into the waveguide, is measured by an attenuated continuous-wave diode laser at 1310nm. The DQE reaches 24 and 22% at bias currents close to the critical current for the transverse-electric and transverse-magnetic polarizations, respectively, as shown in Figure 2(a). The DQE does not reach 100% efficiency due to the limited length of the wires (30μm) and possibly the non-uniformity of the NbN superconducting film on GaAs. A 1/e decay time of 6.2ns is measured as a temporal response from the detectors, which corresponds to an estimated maximum count rate >50MHz. We demonstrated the photon-number-resolving capability of our WPNRDs under pulsed-illumination at 1310nm. Figure 2(b) shows that our WPNRD detects up to four photons in an optical pulse, with an average detected photon number μ_av∼1.5 photons/pulse. Distinct detection levels corresponding to the detection of 0–4 photons are clearly observed. We have also demonstrated that up to twelve distinct photon levels can be measured (in a free-space configuration) by further increasing the number of wires.¹⁶ We have proposed that a cold-stage amplifier with a high input resistance could overcome the noise in the output signal that may be a limiting factor for further scaling of the photon number.^{17, 18}

Figure 2. (a) The device quantum efficiency (QE) of a WPNRD measured for the TE (blue) and TM (black) polarizations at 1310nm. (b) Pulse-height distribution corresponding to 0–4 photon detection events. The measurements were taken with a pulse diode laser with a 2MHz repetition. a.u.: Arbitrary units.

In summary, we have developed WPNRDs that enable photon-number-resolving detection on GaAs QPICs. Our design achieves high temporal resolution, low jitter, and a DQE of >20%, and it shows distinct photon-number resolution of up to four photons. Moving forward, we intend to overcome the technical challenges standing in the way of developing fully integrated circuits for quantum information processing by designing reconfigurable photonic circuitry and single-photon sources and detectors.

This work was supported by Dutch Technology Foundation STW, applied science division of NWO, the Technology Program of the Ministry of Economic Affair, and by the European Commission through FP7 QUANTIP (Contract No. 244026).