Two-photon comb with wavelength conversion and 20-km distribution for quantum communication

The study addresses the challenge of long-distance quantum communication over optical fibers, where loss prevents direct amplification of unknown quantum states and necessitates quantum repeaters. While satellite links have demonstrated very long-distance quantum links, fiber networks offer stability and existing infrastructure. The authors aim to develop a quantum-memory-compatible, telecom-wavelength entanglement source with narrow linewidth and high fidelity to support quantum internet applications such as teleportation, entanglement swapping, and distributed/secure cloud quantum computing. They target combining a telecom-band source suitable for low-loss fiber transmission with wavelength conversion to visible wavelengths compatible with solid-state quantum memories (e.g., Pr3+:YSO), while maintaining high entanglement quality and narrow spectral features across many frequency modes for multiplexing.

Prior work includes demonstrations of quantum supremacy processors, satellite-based entanglement and QKD over thousands of kilometers, and theoretical limits on fiber QKD distance without repeaters (~550 km) with best experimental records (~421 km). Quantum repeaters with and without memories have been proposed and demonstrated in parts. Cavity-enhanced SPDC sources have been used for narrow-linewidth, bright single-mode, and multimode photon generation, including sub-MHz sources. Wavelength conversion has been demonstrated with lasers and single photons to interface telecom photons with various quantum systems (NV centers, rare-earth doped crystals, quantum dots, rubidium). However, realizing simultaneously high count rates, sub-10 MHz linewidths, and high entanglement fidelity at telecom wavelengths remains challenging. The authors build on polarization-entangled generation using orthogonal PPLN crystals and cavity enhancement, and on SFG-based wavelength conversion to 606 nm for Pr3+:YSO memories.

• Two-photon comb (TPC) source: Degenerate SPDC at 1514 nm in a bow-tie cavity (~2.5 m length, FSR ~116–120 MHz) containing two mutually orthogonal type-0 PPLN crystals (dimensions 0.5×3×10 mm and 0.5×0.5×10 mm; Jinan Institute of Quantum Technology). Crystals and holders were temperature stabilized to within ~1 mK. The SPDC pump at 757 nm was generated by SHG of a 1514-nm external cavity diode laser (Sacher TEC420-1530-1000) stabilized to acetylene. The pump was focused to ~25 µm waist using lenses and a plano-concave mirror. Cavity locking employed Pound-Drever-Hall (PDH) with an optical chopper (duty cycle 1/3) to stabilize both the cavity resonance to the QM target frequency and the entangled-state relative phase.
• Polarization entanglement: Orthogonal PPLN arrangement produced |α|HH⟩ + β|VV⟩ states with adjustable amplitudes via pump polarization (HWP). Advantages include higher brightness, reduced alignment sensitivity, and compensated birefringence-induced phase.
• Characterization: Two-photon statistics measured using a Hanbury-Brown–Twiss setup with a 50:50 beam splitter, two SSPDs (≈85% efficiency at telecom; ~40 ps jitter), and a TCSPC module (HydraHarp 400; 32 ps resolution for extended windows, 16 ps elsewhere). Tomography used a 50:50 beam splitter and polarization analyzers (1514-nm QWP, HWP, vertical polarizer) with SSPDs; maximum-likelihood reconstruction over 16 settings. Bell states formed by inserting additional HWPs after the beam splitter (one as phase shifter via yaw angle, one as bit-flip with slow axis at 45°). Typical per-basis integration 15 s, total ~10 min per set.
• Fiber distribution: After coupling to single-mode fiber, photons transmitted over 10 km fiber (telecom), yielding total pair separation of 20 km after splitting. Fiber dispersion used to estimate comb bandwidth.
• Wavelength conversion (WC): Telecom photons combined with an auxiliary laser at 1010 nm (TOPTICA TA pro stabilized via iodine) and coupled into a type-0 PPLN waveguide (NTT Electronics; 9.9×11 µm cross-section, 48 mm length). SFG to 606 nm (Pr3+:YSO transition 3H4(0)→1D2(0)) targeted. Coupling efficiencies ~60% for both beams; internal conversion efficiency ~96%, external ~60%. Spectral filtering used two dichroic mirrors and one bandpass filter to suppress SPDC, Raman, residual pump leaks, and stray light. Converted photons detected with Si APDs (~60% efficiency; ~300 ps jitter). Measurements conducted with/without 10-km fiber before WC. Normalized second-order signal-idler correlation coefficient g^(2)(0) extracted as peak-to-noise ratio.
• Additional parameters: Output coupler reflectivities of 99% and 95% were used to trade linewidth against escape/brightness. Cavity comb and envelope analyzed to extract FSR, coherence time, and linewidth. Fiber transmittance at 1514 nm measured ~62%.

• Comb structure and linewidth: Two-photon correlation shows periodic peaks at 8.6 ns intervals corresponding to FSR ≈116–120 MHz. The exponential envelope yields a cavity linewidth Δν ≈0.95 MHz with 99% output mirror (degenerate photon linewidth ~0.61 MHz), and 1.35 MHz with 95% mirror (higher brightness).
• Coherence and bandwidth: Coherence time up to ~1 µs observed. Fiber-dispersion analysis after 10 km indicates comb spectral range of ~1–2 THz.
• Entanglement quality: Tomography (95% mirror) gives maximal fidelity to an arbitrary pure state of 96.1% and concurrence of 93.0%; CHSH S = 2.47 (>2 indicates nonlocality). Bell-state fidelities: |Φ+⟩ 90.0%, |Φ−⟩ 90.2%, |Ψ+⟩ 89.4%, |Ψ−⟩ 88.1%.
• Wavelength conversion performance: External SFG conversion efficiency ~60% (internal ~96%). WC bandwidth limited by phase matching to ~25 GHz. Despite bandwidth mismatch to TPC (~1 THz), clear WC two-photon comb (WC-TPC) observed.
• Correlation after distribution and conversion: After 10-km fiber transmission (total separation 20 km) and WC, clear two-photon comb with normalized second-order correlation g^(2)(0) ≈3 observed with Si APDs despite larger timing jitter. Telecom-only one-photon-to-visible correlation was not clearly observed (g^(2)(0)≈1) due to uncorrelated multi-pair noise.
• SNR behavior: g^(2)(0) versus SPDC pump power indicates that g^(2)(0) after WC can exceed that before WC due to a correlation-filtering effect, selecting highly correlated spectral-temporal modes around degeneracy within WC bandwidth. Fiber transmittance at 1514 nm measured ≈62%.

The TPC source delivers a unique combination of very narrow sub-MHz linewidth with a wide multi-THz comb and high-fidelity polarization entanglement, well suited for atomic-frequency-comb quantum memories. Its long cavity length provides narrow linewidth even at moderate finesse, reducing sensitivity to alignment drift while maintaining compatibility with QM absorption windows (e.g., Pr3+:YSO ~4.6 MHz). Environmental stabilization (temperature at millikelvin levels, airflow isolation) is essential to maintain phase coherence across many modes; additional compensation crystals and dual-cavity approaches could further improve stability and fidelity. Wavelength conversion to 606 nm via SFG preserves temporal/linewidth profiles and enables interfacing with Pr3+:YSO memories. The observed increase in g^(2)(0) post-WC stems from selecting correlated photon pairs within the WC spectral acceptance (frequency-domain filtering around degeneracy) and from temporal selection that suppresses multi-pair noise within the SPDC coherence time. However, single-photon WC correlations were obscured by multi-pair background at telecom. Bandwidth mismatch (TPC ~1 THz vs WC ~25 GHz) attenuates usable modes but still yields adequate SNR. Tailored spectral filtering focused on the WC-TPC spectrum—potentially using high-dopant (~1%) Pr3+:YSO as a dynamic bandpass filter—could further suppress out-of-band WC noise. Addressing polarization sensitivity of the current WC (via polarization-insensitive WC designs) would mitigate fiber-induced polarization rotations, though at possible efficiency costs. Employing time-bin qubits is recommended for robust fiber transmission and compatibility with single-polarization WC. Overall, the results support frequency- and time-multiplexed long-distance entanglement distribution compatible with solid-state quantum memories.

The authors demonstrate a versatile entanglement source in the telecom band that generates a two-photon comb with sub-MHz linewidth, high entanglement fidelity (up to 96.1% for pure-state overlap; ~90% for Bell states), and multimode capacity. They achieve two-photon distribution over 20 km (10 km per arm) in fiber followed by wavelength conversion to 606 nm, observing clear two-photon correlations with g^(2)(0) ≈3 despite detector jitter and WC bandwidth limits. This establishes a practical path toward memory-compatible, frequency-multiplexed quantum networking. Future work includes: specialized noise filters matched to the TPC spectrum (e.g., high-dopant Pr3+:YSO dynamic filters), polarization-insensitive WC, Bell-state measurements in multimode regimes for multiplexed quantum communication, and conversion from polarization to time-bin encoding to evaluate full system performance including basis-conversion losses.

• WC bandwidth (~25 GHz) is much narrower than the TPC comb (~1–2 THz), limiting simultaneous mode utilization and contributing to loss outside the acceptance band.
• Current WC is polarization sensitive, making performance susceptible to fiber-induced polarization rotation; active stabilization or polarization-insensitive WC would be needed in practical links.
• Single-photon WC correlations were not observable (g^(2)(0)≈1) due to multi-pair noise at telecom, indicating the need for improved filtering and lower pump powers for heralded operation.
• Entanglement fidelities for explicit Bell states are ~88–90%, affected by waveplate dispersion/retardance and multi-mode effects; environmental sensitivity (temperature, air convection) can degrade phase stability.
• Fiber losses (measured ~62% transmittance over 10 km at 1514 nm including connections) and detector timing jitter in the visible limit g^(2)(0) and coincidence rates.
• No storage in a quantum memory was demonstrated; the interface was validated via wavelength conversion and correlation only.