Fiber-integrated quantum frequency conversion for long-distance quantum networking
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | npj Quantum Information |
| Published | 2026-03-31 |
| Authors | Zhichuan Liao, Ao Shen, Lai Xiang Zhou, Nan Jiang, Zhiliang Yuan |
| DOI | 10.1038/s41534-026-01225-y |
The fiber-integrated converter produced 1588.3 nm telecom photons from 637.2 nm input at about 9% total efficiency and 154 Hz pump-induced noise at 1.2 W pump power. At a 32.7 kHz input rate, its measured signal-to-noise ratio was 12.3, which was 76% above the setup in Ref. [34]. A model gave 52.5% expected fidelity between an NV-center spin and the converted photon after 100 km of fiber at that rate. The 100 km value is a projection from measured parameters, not a measured 100 km entanglement experiment.
Fiber-coupled conversion setup
Section titled “Fiber-coupled conversion setup”A 637.2 nm continuous-wave laser was shaped by an acousto-optic modulator into 300 ns pulses. The pulse repetition rate was 1 MHz, the duty cycle was 30%, and the measured optical power was about 1.74 mW. Two cascaded polarization-maintaining variable optical attenuators reduced the pulses to single-photon levels. The applied attenuations were 117.6, 112.0, and 107.5 dB for the three measured input rates.
The pulses were combined with a 1064.1 nm pump through dense wavelength division multiplexing. Half-wave and quarter-wave plates adjusted polarization before the light entered the PPLN waveguide. The waveguide used a lithium-niobate-on-insulator ridge structure with an 11.85 µm poling period and a length of 20 mm. A thermoelectric controller held it at 33.80 °C with ±0.01 °C stability.
The waveguide’s input and output ports connected to PM980 and PM1550 fibers. The overall transmission through the PPLN waveguide was about 36%. This figure included fiber coupling and propagation loss. The paper identifies low coupling as a principal remaining loss in the integrated design.
Efficiency and noise suppression
Section titled “Efficiency and noise suppression”The reported conversion efficiency included loss from the combined DWDM, fiber-waveguide coupling, and hybrid filtering. This distinction matters when comparing systems. At 1.2 W pump power, pulsed inputs at 32.7, 118.0, and 327.7 kHz produced total conversion efficiencies of 9.24%, 9.27%, and 8.92%. A 35.5 kHz continuous-wave input produced 8.92% at the same pump power.
The fitted pump-power curve used a simulated maximum efficiency of 10.95% after propagation losses. The authors estimated about 80% internal conversion efficiency in the waveguide. The measured total fell below that value because the DWDM, waveguide, and filter stage had propagation efficiencies of 87%, 36%, and 36%, respectively.
Pump-induced noise came mainly from spontaneous parametric down-conversion photons. The pump wavelength was shorter than the 1588.3 nm target and well separated from it. The hybrid filter combined DWDM, two 10 GHz bandwidth fiber Bragg gratings, and a 250 MHz bandwidth ultranarrowband tunable filter. The DWDM removed residual pump light, while the two Bragg gratings and tunable filter suppressed broadband down-conversion noise. The module provided more than 97 dB of noise suppression while adding about 4.4 dB loss to converted photons.
Noise grew linearly with pump power but was 154 Hz at 1.2 W. This count excluded the 54 Hz dark count from the superconducting nanowire single-photon detector. The result separates filtering performance from detector dark counts.
Signal-to-noise performance
Section titled “Signal-to-noise performance”The detection chain used a 10 ns time bin and a 300 ns gate. The SNR was defined as accumulated signal counts minus noise counts, divided by noise counts within that gate. A 10 MHz clock synchronized pulse generation and detection across the two laboratories. The SNSPD detection efficiency was 90%. The 1.2 dB fiber link between them represented about 7.5 km.
Measured pulsed performance at 1.2 W pump power:
| Input rate at 637.2 nm | Attenuation | Total efficiency | Measured SNR |
|---|---|---|---|
| 32.7 kHz | 117.6 dB | 9.24% | 12.3 |
| 118.0 kHz | 112.0 dB | 9.27% | 43.9 |
| 327.7 kHz | 107.5 dB | 8.92% | 117.8 |
At 32.7 kHz, SNR reached 12.3 at 1.2 W. Raising pump power beyond that point reduced SNR because added noise began to outweigh the conversion gain. Higher input rates produced SNRs of 43.9 and 117.8, so the 637.2 nm photon rate was a direct limit on long-link performance.
The integrated waveguide coupling of 36% was lower than the values above 70% reported for free-space implementations. The authors treated this as a trade-off between alignment-free operation and propagation efficiency. They proposed mode-matching optics or integrated spot-size converters as routes to higher coupling.
Long-distance fidelity model
Section titled “Long-distance fidelity model”The fidelity calculation treated the converted photon and NV-center spin as a Bell state. For visibilities in the X, Y, and Z bases, the model used F = (1 + V_X + V_Y - V_Z) / 4. It combined conversion efficiency, input rate, conversion noise, detector dark counts, detection efficiency, and fiber transmission. Fiber attenuation was fixed at 0.16 dB/km, giving η_l = 10^(-0.16 l / 10).
One simulation used the measured parameters from this system. The other used parameters from Ref. [34]: 415 Hz conversion noise, 190 Hz dark counts, 17% conversion efficiency, and 41% detection efficiency. The reference case also used a 35.5 kHz input rate. The present system used 154 Hz conversion noise, 54 Hz dark counts, 90% detection efficiency, and about 9% conversion efficiency.
Modeled fidelity comparison at a 32.7 kHz input rate:
| Fiber length | Input rate | This work SNR | This work expected fidelity | Ref. [34] expected fidelity |
|---|---|---|---|---|
| 0 km | 32.7 kHz | 13.74 | 90.5% | 83.1% |
| 60 km | 32.7 kHz | 4.19 | 75.8% | 54.5% |
| 100 km | 32.7 kHz | 1.16 | 52.5% | 35.3% |
At 60 km, the 32.7 kHz case exceeded reference fidelity by about 21.3%. At 100 km, it retained 52.5%, compared with 35.3% for reference parameters. The model also gave 75.7% at 100 km for 118.0 kHz, against 51.0% for reference parameters. It gave 89.0% at 100 km for 327.7 kHz.
The 7.5 km laboratory link produced expected fidelities of 89.5%, 96.7%, and 98.7% for the three measured input rates. A reference point with SNR near 7 and an input rate of 35.5 kHz gave 83.3% expected fidelity. The authors attributed a 6.2% improvement in expected fidelity to the higher SNR.
Limits and engineering priorities
Section titled “Limits and engineering priorities”The 100 km results depend on a model with 0.16 dB/km attenuation. The paper did not measure NV spin-photon entanglement through 100 km of fiber. Its measured link was 1.2 dB between laboratories. The model assumes correlated measurements of the NV spin and converted photon in the X, Y, and Z bases.
Coupling efficiency remains the clearest system-level loss. The authors recommend advanced mode-matching optics or integrated spot-size converters. The filter’s 4.4 dB insertion loss remains part of the propagation budget alongside low waveguide coupling.
The NV source also limits the available signal. Zero-phonon-line photons account for only about 3% to 4% of total NV-center emission. The paper suggests a higher zero-phonon-line branching ratio through the Purcell effect and better collection through optical design. Those approaches were proposed, not demonstrated in this experiment.
The supplied manuscript is an unedited article-in-press version and warns that errors may remain before final publication. The measured conversion, noise, and SNR results stand separately from the theoretical long-distance projections.
Materials and support
Section titled “Materials and support”The paper’s quantum node uses NV centers in diamond and needs collection of 637.2 nm zero-phonon-line photons for conversion to 1588.3 nm. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal and polycrystalline diamond plates and wafers up to 125 mm. The paper does not specify the diamond grade, NV-center formation, or collection geometry, so use the engineering support page to define those custom source requirements.
View Original Abstract
Signal photons emitted by quantum nodes typically fall outside the low-loss telecom window of optical fibers, leading to severe transmission losses. Quantum frequency conversion (QFC) offers an effective optical interface that bridges quantum nodes with telecom-band channels, enabling long-distance quantum communication. In this work, we demonstrate a compact, fiber-integrated QFC system with low noise and a high signal-to-noise ratio (SNR). Using a periodically poled lithium niobate (PPLN) waveguide, input photons at 637.2 nm are down-converted to telecom photons at 1588.3 nm. Our system achieves a total conversion efficiency of approximately 9%, with pump-induced noise suppressed to 154 Hz. For input photon rates of 32.7, 118.0, and 327.7 kHz, the corresponding SNRs are 12.3, 43.9, and 117.8, respectively. We further develop a theoretical model to simulate the entanglement fidelity between nitrogen-vacancy (NV) center spins and the frequency-converted telecom photons. At the emission rate of an NV center, our QFC system maintains an expected fidelity exceeding 52% over a transmission distance of 100 km. These findings highlight the potential of our QFC system for scalable, long-distance quantum networking.
High Concentration Quantum Grade Single Crystal Diamond
View Specs →
Advanced Thermal Simulation Service Diamond Heat Spreader Performance Analysis
View Specs →Products
View Specs →References we cover
Section titled “References we cover”- Experimental demonstration of memory-enhanced quantum communication
- Quantum Frequency Conversion of Single Photons from a Nitrogen-Vacancy Center in Diamond to Telecommunication Wavelengths