Telecom-Band Quantum Interference of Frequency-Converted Photons from Remote Detuned NV Centers

Entanglement distribution over quantum networks has the promise of realizing fundamentally new technologies. Entanglement between separated quantum processing nodes has been achieved on several experimental platforms in the past decade. To move toward metropolitan-scale quantum network test beds, the creation and transmission of indistinguishable single photons over existing telecom infrastructure is key. Here, we report the interference of photons emitted by remote spectrally detuned NV-center-based network nodes, using quantum frequency conversion to the telecom $L$ band. We find a visibility of $0.79\pm 0.03$ and an indistinguishability between converted NV photons around $0.9$ over the full range of the emission duration, confirming the removal of the spectral information present. Our approach implements fully separated and independent control over the nodes, time multiplexing of control and quantum signals, and active feedback to stabilize the output frequency. Our results demonstrate a working principle that can be readily employed on other platforms and shows a clear path toward generating metropolitan-scale solid-state entanglement over deployed telecom fibers.

Popular Summary

Quantum networks hold the promise of fundamentally changing the way we share and process information. A promising candidate for such networks are optically active spins in solids, where stationary qubits can be entangled with flying photonic qubits. These states of light experience high losses in existing fiber infrastructure, due to incompatibility of the emission wavelength with the telecom standard. To bring quantum networks out of the lab, it is crucial to convert these photons to the telecom band, whilst maintaining their quantum nature and ensuring compatibility with other nodes. In this paper, we introduce a method that achieves both simultaneously. We demonstrate our technique by building two remote quantum nodes based on the nitrogen-vacancy (NV) center in diamond, each equipped with a quantum frequency conversion (QFC) module employing a periodically-poled Lithium Niobate (ppLN) crystal. We show that the resonant emission of the NV centers (637nm) can be faithfully converted to the same wavelength in the telecom L-band (1588nm) by locking to a central reference laser, for a broad range of frequencies. This technique removes any difference in emission frequency between nodes, facilitating scaling. When performing a canonical Hong-Ou-Mandel type experiment, we find a stark reduction of coincidence counts indicating the desired two-photon quantum interference. Further analysis finds a high degree of indistinguishability of the converted photons, emphasizing the effectiveness of our scheme. We believe that other spin platforms in diamond, silicon, and silicon carbide that emit in the visible spectrum can benefit from our technique. Moreover, our technique opens the door to metropolitan scale solid-state entanglement generation over deployed fibers.

Figure 1

The layout of the two independent network nodes and the midpoint. (a) A schematic of the main components of the setup. Pulses excite the NV center, which emits single photons through spontaneous decay. The photons are converted to telecom wavelength and guided toward a midpoint placed in the neighboring laboratory, where they interfere on a beam splitter (for more details on optics, see the Supplemental Material [25]). (b) The NV-center level structure, showing the spin levels in the ground state and the relevant optical transitions. The optical transitions of the two nodes have different energies due to local variations in strain (see text). (c) A detailed schematic showing the optics used for full operation of the system. The nodes are identical and no hardware is shared between the systems. Each node has a set of charge reset, spin reset, and excitation lasers, which are modulated, combined, and focused via a high-NA objective onto the NV center. The phonon-side-band (PSB) [zero-phonon-line (ZPL)] emission from the NV center is filtered using frequency (polarization) filtering and coupled into a multi- (single-) mode fiber. The PSB emission is measured locally on an Avalanche Photodiode, while the ZPL emission is sent to the quantum-frequency-conversion (QFC) module. Stabilization light is split off from the excitation path and brought into the single-photon path via a polarizing beam splitter. The QFC module contains remotely controllable optics to align the input, pump, and converted fields to the waveguide on the periodically poled lithium niobate (ppLN) crystal that converts both the single photons and stabilization light. Polarization-maintaining (PM) fibers transport the photons to a midpoint where the single photons are filtered and separated from the stabilization light using a combination of in-fiber filters. The single photons interfere on a beam splitter and are measured by single-photon superconducting nanowire detectors (SNSPDs), while the stabilization light interferes with the reference laser and is measured on a photodiode.

Figure 2

The removal of spectral offset using conversion. (a) The fluorescence measurement sequence used in (b). A charge-reset pulse is followed by repeated resonant excitation, during which fluorescence in the PSB is monitored. (b) Resonant excitation spectra at node 1 (node 2) shown in blue (orange) diamonds, revealing the frequencies of the optical transitions (vertical lines) used for photon generation. The horizontal axis shows the optical excitation frequency with respect to a $470.477$-THz offset. The gray dotted lines are Lorentzian fits. The right vertical axis shows the frequency of the QFC pump with respect to a $281.635$-THz offset. The blue and orange dots show the QFC pump-laser frequency when the stabilization is active. The solid gray line is a linear fit. (c) A pulse schematic for measuring the transmission through the ultranarrow filter (UNF) shown in (d). Reflections from the excitation light off the sample surface, which follow the same path as the resonant NV emission, are converted to the target telecom frequency and measured on the SNSPDs. The excitation frequency is swept by modulating the AOM frequency [see Fig. 1]. The transmission data are corrected for the frequency dependence of the losses in the AOMs. (d) The layout of relevant laser frequencies and the transmission data of the UNFs. The UNF transmissions (blue and orange dots) are actively stabilized via their temperature to half the transmission of the reference laser (gray vertical line). The stabilization light (red vertical line) is detuned from the target (green vertical line) and therefore reflected off the UNF (see main text).

Figure 3

The generation of single photons using a fixed heartbeat. (a) The measurement sequence for synchronized generation of single photons using a distributed heartbeat. For detailed timing, see the Supplemental Material [25]. (b) A histogram of the SNSPD counts in a single excitation round, averaged over approximately $2.8\times {10}^{11}$ repetitions and analyzed in 80-ps bins. The measured count rate (green) is a combination of background (gray) and NV fluorescence (red), which is calculated by subtracting the background from the total counts. The dotted line depicts exponential decay with a 12.5-ns lifetime and serves as a guide to the eye. The NV signal before time $0$, the peak of the excitation pulse, is due to the nonperfect extinction ratio of the devices generating the optical excitation pulse. The background data are taken continuously over 24 h to include any drifts that occur over the same time scale as the signal data. (c) The signal-to-background ratio (SBR) for both detectors as calculated from data in (b) (solid lines). The solid vertical gray line shows the start of our chosen detection window, whereas the dashed (dash-dotted) line shows the end for the data shown in Fig. 4 [Fig. 4]. The difference between the two curves is due to nonequal detector performance.

Figure 4

Two-photon quantum interference at telecom wavelength. (a) The histogram of the measured coincidences (blue) for the analyzed time bins, overlaid with a model assuming distinguishable photons (orange), based on independently determined parameters. The histogram bin size is 2.4 ns. The vertical lines depict the time origin of each detection-bin difference. The horizontal scale bars show the relevant time scales in and between the detection-bin difference. The inset shows the extraction of the interference visibility, corrected for the imbalance of measured photons per excitation of the two nodes, using a linear fit to the total counts per bin difference (see text). (b) A histogram showing the temporal shape of the nonzero (zero) bin-difference coincidences shown in the top (bottom) panel. We overlay the data with the same model (see the Supplemental Material [25]), taking into account brightness and background rates and the indistinguishability of converted NV photons of 0.9 [see (c)]. The bin size for the top (bottom) panel is 0.8 ns (2 ns). (c) The measured visibility and extracted indistinguishability of converted NV photons for varying window length. The error bars for visibility and indistinguishability are $1\sigma$ and a $68\mathrm{\%}$ confidence interval, respectively. The green (blue) circled points indicate data corresponding to the window length shown in Fig. 4 [Fig. 4].