Skip to content

Quantum Frequency Conversion of Single Photons from a Nitrogen-Vacancy Center in Diamond to Telecommunication Wavelengths

FieldValue
JournalPhysical Review Applied
Published2018-06-19
AuthorsA. Dréau, Anna Tcheborateva, Aboubakr El Mahdaoui, Cristian Bonato, Ronald Hanson
DOI10.1103/physrevapplied.9.064031
Citations153

Individual spin-selective photons from one nitrogen-vacancy center were converted from 637 nm to 1588 nm with 17% total conversion efficiency and a signal-to-noise ratio of about 7. The converted light retained clear antibunching, with (g^{(2)}(0) = 0.32 \pm 0.08), below the 0.5 value expected for a single emitter.

The coherent zero-phonon line of an NV center emits at 637 nm but represents only about 3% of its total emission. That wavelength loses about 8 dB/km in optical fiber, compared with less than 0.2 dB/km in the telecom range. A previous kilometer-scale NV entanglement experiment had its entanglement generation rate reduced by a factor of 30 by this attenuation.

The authors used difference-frequency generation with a continuous-wave 1064 nm pump to shift spin-selective 637 nm photons into the 1588 nm L band. The nonlinear element was a Zn-doped type-0 periodically poled lithium niobate waveguide measuring 8 µm × 8 µm × 48 mm. Crystal temperature was tuned to meet the quasi-phase-matching condition. Anti-reflection coatings and aspheric lenses supported input and output coupling.

NV emission was extracted into polarization-maintaining fiber and coupled into the PPLN waveguide. The estimated signal coupling was at least about 90%, based on transmission of a 637 nm continuous-wave laser. A two-lens telescope corrected chromatic aberration and improved spatial mode matching because the waveguide was not fully single mode at both wavelengths.

Spectral cleanup used a dispersive prism, a long-pass filter, single-mode fiber, and a tunable fiber Bragg grating. The grating had a 4 pm bandwidth, about 500 MHz, and 55 dB noise rejection. The converted photons reached a superconducting single-photon detector with 41% quantum efficiency and timing jitter below 70 ps. A time-tagged card had 1 ps resolution and excitation synchronization jitter below 80 ps.

The NV-center excitation and collection setup operated at 4 K. After preparing the optical and spin state (m_s = 0), a 2 ns resonant optical (\pi) pulse drove the center to its (E_x) excited state. Spontaneous emission returned the center to the ground state and produced the spin-selective photon. The authors recorded 7.5 million excitation sequences for each comparison.

The unconverted 637 nm photons were measured with a visible avalanche photodiode with about 80% detection efficiency. Their histogram decayed with a lifetime of (12.9 \pm 1.0) ns. The 1588 nm photons appeared about 90 ns after excitation and decayed with a lifetime of (12.6 \pm 1.2) ns. The matching time constants identified the telecom counts as converted NV emission. The histograms contained raw detection events and were not corrected for detector quantum efficiency.

Signal counts were integrated in a 20 ns window after the resonant (\pi) pulse. A second window of equal duration, placed 100 ns later, measured noise. Before conversion, the detected probability was (p_{c,\mathrm{red}} = (5.7 \pm 0.1) \times 10^{-4}) counts per excitation. The inferred average photon number was ((7.1 \pm 0.1) \times 10^{-4}).

The maximum telecom probability was (p_{c,\mathrm{tel}} = (5.4 \pm 0.1) \times 10^{-5}) photons per optical excitation at about 110 mW input pump power. Detection fell at higher power because the inverse process converted telecom photons back to 637 nm. The single-photon data agreed with classical difference-frequency generation measured with a 637 nm continuous-wave laser.

After correction for the two detector efficiencies, total conversion efficiency reached about 17%, while internal efficiency inside the PPLN crystal reached about 65%. The authors said evidence pointed to imperfections in the periodicity of the poling domains as the internal limit. Total efficiency was limited by 41% FBG transmission, about 85% fiber coupling, and about 90% waveguide insertion or transmission.

The present 65% internal value exceeded the 44% reported for an earlier experiment on attenuated 637 nm laser light. That earlier setup had a noise rate of about 4 million counts per second, roughly three orders of magnitude above typical ZPL count rates. The present experiment used resonant excitation and time-resolved detection to work with single NV photons despite this background.

Noise counts increased linearly with pump power. The detector contributed ((0.38 \pm 0.06) \times 10^{-5}) counts per excitation, equivalent to (190 \pm 30) counts per second. Pump-induced noise added ((0.31 \pm 0.05) \times 10^{-5}) counts per excitation for each 100 mW of input pump power.

Raman scattering and spontaneous parametric down-conversion were candidate sources of pump-induced noise. The authors favored spontaneous parametric down-conversion because the pump and converted wavelengths were well separated and poling errors were likely. They noted that slight deviations from the linear trend in some waveguides were not understood.

Resonant excitation, time-resolved detection, and spectral filtering produced a signal-to-noise ratio above 7. A 6 ns window at 75 mW produced a ratio above 10, with fewer converted photons. The FBG bandwidth was about 500 MHz, and its linewidth was 10 times the NV photon linewidth.

A Hanbury-Brown and Twiss measurement tested photon statistics before and after conversion. A polarization-maintaining fiber beam splitter sent its two outputs to separate single-photon detectors. The experiment used 15 resonant (\pi) pulses per sequence, reinitialized the spin between sequences, and checked charge state and resonance every 250 repetitions of the sequence. Pump power was 100 mW, and each histogram used a 25 ns integration window.

Before conversion, (g^{(2)}(0) = 0.003 \pm 0.002), corresponding to a signal-to-noise ratio of about 530. After conversion, (g^{(2)}(0)) rose to (0.32 \pm 0.08) because of conversion noise. The expected value from the histogram signal-to-noise ratio was about 0.35 at a ratio of 4.2. The beam splitter divided photon counts while detector dark counts remained constant.

Coincidence counts were about 13 times lower after conversion despite 43 hours of collection versus 2 hours before conversion. Both datasets showed antibunching without background subtraction. Side peaks above 1 were attributed to photon bunching from blinking caused by shelving in a long-lived metastable state. The single-photon character persisted, although conversion noise reduced its purity.

The authors calculated that the current setup could already be advantageous at a fiber separation corresponding to 2.6 km for a two-photon entangling protocol. Difference-frequency generation can also bring strain-shifted NV centers to the same telecom frequency, which removes the need for dc Stark tuning. The reported work demonstrated photon conversion and antibunching, not a remote entanglement link.

Only about 3% of NV emission lies in the coherent zero-phonon line. Better collection and Purcell-assisted emission were identified as direct routes to a proportional gain in signal-to-noise ratio. The FBG linewidth was 10 times the NV photon linewidth, and its transmission was only 41%. Detectors with lower dark counts and higher efficiency were also proposed improvements.

The authors expected a signal-to-noise ratio above 100 to become feasible with better nonlinear waveguides, optimized spectral filters, improved ZPL collection, and Purcell-assisted emission. This is a projection, not a measured result. Device design must balance pump power against inverse conversion while retaining resonant excitation and narrow temporal gates.

For NV-center quantum photonics, the paper needs a single-crystal diamond host with an optically active NV center, but it does not specify how the center was created or selected. The diamond materials catalog lists MPCVD single-crystal diamond as plates and wafers up to 125 mm, with thickness from 0.1 to 500 micrometers and substrates up to 10 mm. 6CCVD supplies the host material and offers custom fabrication and engineering support, with global shipping.

View Original Abstract

Entanglement-based quantum networks are strongly pursued worldwide, because of their potential impact on secure communication, distributed quantum computing, and timekeeping, for example. Among quantum emitters, the N-$V$ center in diamond is a leading candidate for implementing such networks, but high photon loss at the N-$V$ emission wavelength hinders long-distance entanglement (beyond ~1 km). The authors surmount this hurdle by down-converting single N-$V$ photons to a telecom wavelength, via nonlinear optics plus efficient filtering and excellent control of a lone emitter. This technological achievement is a critical step toward large-scale quantum networks.