A Tuneable Telecom Wavelength Entangled Light Emitting Diode Deployed in an Installed Fibre Network

The study addresses the challenge of integrating quantum light sources into existing telecommunication infrastructures. For effective quantum networking, sources must operate at telecom wavelengths, be wavelength-tuneable for interfacing and multiplexing, and be compatible with scalable manufacturing and low-voltage electronics for safe, reliable field operation. Entanglement distribution between remote users underpins advanced quantum-network applications beyond standard QKD, offering robustness to photon-number-splitting attacks, extended communication distances, and potential links between quantum processors. While many current sources rely on spontaneous processes with intrinsic efficiency limits, semiconductor quantum dots (QDs) offer sub-Poissonian, potentially deterministic photon-pair emission. Electrically driven QD devices (ELEDs) are promising due to compatibility with III–V semiconductor manufacturing. However, field deployment favors simple, robust operation over complex resonant optical excitation schemes. For coexistence with classical traffic, operation in the telecom O-band with sufficient spectral separation from C-band data and compliance with CWDM grid spacing (20 nm) is desirable. Post-growth tuning, particularly via static electric fields, is attractive for practical deployment. This work demonstrates a low-voltage, electrically operated, wavelength-tuneable telecom ELED with >25 nm tuning, deployed on a metropolitan fibre network, distributing sub-Poissonian entangled photons multiplexed with classical data and achieving high entanglement fidelity.

Prior works established entanglement-based quantum communication and QKD using attenuated lasers and spontaneous sources, but with efficiency and security limitations. Quantum-dot-based entangled photon sources promise deterministic generation and compatibility with advanced protocols (teleportation, swapping), though resonant excitation requires complex lasers and suppression techniques, making non-resonant schemes attractive for early field integration despite lower efficiency and purity. Telecom-wavelength QDs have enabled long-term field transmission over deployed fibres. For network coexistence with classical traffic, O-band operation and CWDM compatibility are advantageous due to reduced Raman scattering from C- to O-band. Various tuning mechanisms have been explored: magnetic fields and strain offer limited nm-scale tuning with impractical field/voltage requirements for deployment; electric-field-based Stark tuning provides larger ranges with low voltages. The present work builds on these to realize a practical, tuneable, electrically operated telecom ELED suitable for network deployment and classical-quantum multiplexing.

Device and wafer design: The device comprises a central circular tuning diode (QD region) surrounded by a ring-shaped pumping diode. A forward bias on the pumping diode generates on-chip optical excitation light; a reverse bias on the tuning diode applies a static field for Stark tuning. The wafer includes InAs/GaAs Stranski–Krastanov QDs grown in a bi-modal growth mode for O-band emission, embedded within an InGaAs/GaAs quantum well (QW) with surrounding AlGaAs barriers. A 5 nm InGaAs layer serves as both on-chip optical pump and absorption layer (~950–1000 nm emission in the pumping diode), while QD emission occurs around 1310 nm in the tuning diode. Stacked GaAs/AlGaAs DBR mirrors form a weak vertical 1/2-wavelength cavity to enhance O-band emission. The bottom DBR top repeats are n-doped and the top DBR is p-doped, forming a p-i-n diode. On-chip optical excitation and filtering: Pump emission around 980 nm from the pumping diode is absorbed in the tuning diode QW, exciting carriers that recombine via QDs to emit X and XX photons in the telecom O-band. The >300 nm spectral separation between pump and QD emission enables strong background suppression with standard filters, yielding high single-photon purity (g^(2)(0) well below 0.1 in the lab). Tuning and selection: Electric fields applied via reverse bias on the tuning diode tune the QD emission through the quantum confined Stark effect; growth of QDs within the QW increases permanent dipole moment and polarizability, enhancing tuneability. A single QD with sufficient brightness and entanglement fidelity was selected (typical yield ~0.9%). Emission was tuned by varying reverse bias from -3.8 to 0 V, achieving >25 nm tuning of both X and XX lines across the O-band center. FSS was measured via time-resolved quantum beats in polarisation-filtered X–XX correlations; an FSS ≤10 µeV ensures entanglement observability with ~100 ps detector timing resolution. The device was operated at -2.6 V, setting XX at 1310.00 nm and X at 1321.45 nm with FSS = 5.6 µeV. Field deployment and network integration: The ELED was installed at a West Cambridge site (cooled to 6 K in a closed-cycle cryostat). Emission was fibre-coupled and sent to a spectral filter (diffraction grating) to separate X and XX photons. X photons passed through a polarization analyser (HWP, QWP, LP) at the transmitter site; XX photons were sent over a separate 15 km installed network fibre to the Cambridge Research Laboratory (CRL). The XX fibre carried a bidirectional 1 Gbit/s classical data link at 1550 nm for remote control, multiplexed with quantum photons (1310 nm) using CWDM components. At CRL, XX polarisation was projected using an electronic polarisation controller and fibre PBS; arrival times of both X and XX photons were recorded with superconducting nanowire single-photon detectors. A polarisation stabilisation system compensated birefringence drifts on the XX fibre by periodically sending two orthogonal polarisation references at 1310 nm (every 15 s for 100 ms; forced every 225 s; ~98% quantum duty cycle). Calibration and measurement: Detector polarisation bases were calibrated without free-space access by measuring time-resolved X–XX correlations in three linearly independent bases, extracting the QD eigenbasis orientation and performing Müller matrix evaluation to align detection bases via injected polarisation references. Entanglement was quantified from co-/cross-polarised correlation measurements in HV, DA, and RL bases; fidelity F = (1 + C_HV + C_DA + C_RL)/4, with correlation coefficients derived from coincidence counts. Single-photon purity before and after transmission was assessed via second-order autocorrelation measurements. Losses and rates: Total system losses on the XX path (including stabilisation and multiplexing modules on both ends, receiver filter, EPC, PBS) were 3.3 dB; the XX link fibre had 8.5 dB loss at 1310 nm. The X link fibre had 12.5 dB loss; leakage of classical light at 1309/1313 nm necessitated a narrowband filter at CRL for X at 1321.45 nm, adding 2.5 dB. Extraction efficiency into single-mode fibre was ~3% (confocal collection). Detected raw rates before the network: XX ~500 kHz, X ~400 kHz. After transmission through the network: XX ~33 kHz, X ~13 kHz.

• Demonstrated a fully electrically operated, on-chip optically excited, telecom O-band ELED with wavelength tuneability exceeding 25 nm (≈30 nm observed for X and XX) via low-voltage Stark tuning (-3.8 to 0 V), compatible with ITU CWDM 20 nm channel spacing.
• Set operating point at XX = 1310.00 nm (O-band center), X = 1321.45 nm; FSS remained below 10 µeV across the entire tuning range (at setpoint: 5.6 µeV), ensuring entangled photon-pair emission with typical detector timing.
• High single-photon purity: laboratory g^(2)(0) ≈ 0.08; after 15 km installed fibre transmission multiplexed with a 1 Gbit/s classical C-band link, g^(2)(0) = 0.26 (still strongly antibunched and below the 0.5 classical limit).
• Entanglement distribution over 15 km installed fibre with coexisting classical data: maximum Bell-state fidelity after the link of 94.4 ± 0.3% using a 48 ps post-selection window; with a 200 ps window, fidelity 91.9 ± 0.2% (>190σ above classical limit).
• Measured high-contrast polarisation correlations: HV basis maximum contrast 94.5%; DA and RL bases showed high-contrast quantum beats due to FSS.
• Stable long-term operation: over 40 h continuous run, fidelity remained high with corresponding QBER ~3.8%, classical data link active throughout.
• Practical deployment metrics: extraction efficiency into fibre ~3%; detected rates before link XX ~500 kHz, X ~400 kHz; after link XX ~33 kHz, X ~13 kHz; losses: XX path system loss 3.3 dB, XX fibre 8.5 dB, X fibre 12.5 dB (+2.5 dB narrowband filter to suppress leaked classical light).

The work demonstrates that an electrically operated, Stark-tuneable telecom ELED can meet key requirements for integration into existing optical networks: operation in the O-band, low-voltage tuning across CWDM channels, and coexistence with classical C-band data on the same fibre. The device maintains sufficiently low FSS across its tuning range, enabling high-fidelity entanglement distribution despite use of non-resonant, on-chip optical excitation. Field tests over 15 km of installed fibre validate robust entanglement transmission with high fidelity and stable QBER in the presence of real-world impairments and classical traffic. The results address the core research goal of practical network integration of sub-Poissonian entangled photon sources, showing that electrical operation and wavelength tuneability enable flexible deployment and multiplexing while preserving quantum performance. The slight degradation in single-photon purity and fidelity compared to laboratory measurements is attributed primarily to link and system losses rather than classical co-propagating data, indicating that improved filtering and lower-loss components could further enhance performance.

This work reports the first electrically operated, wavelength-tuneable telecom ELED suitable for multiplexing with classical communication signals and deployment on real network infrastructure. The device achieves low-voltage operation and >25 nm wavelength tuneability, meeting ITU CWDM requirements, and delivers high single-photon purity and entanglement fidelities exceeding 94% after transmission over 15 km of installed fibre with simultaneous classical data traffic. The approach is compatible with standard low-voltage supplies and laser-safety constraints, making it attractive for practical end-user applications. The large tuning range enables flexible wavelength-division multiplexing of multiple ELEDs and integration with classical channels, facilitating low-cost quantum network deployment. Future work should focus on enabling pulsed GHz operation via reduced device capacitance, improving photon collection with nanophotonic structures (circular Bragg gratings, optical antennas, microlenses), enhancing generation efficiency via charge control, and exploring droplet epitaxy QDs for improved coherence under non-resonant excitation, as well as advancing indistinguishability for higher-level network protocols (relays, repeaters).

• Current implementation uses non-resonant on-chip optical excitation, which typically yields lower efficiency and photon purity than resonant schemes; photon indistinguishability (coherence times ~50 ps vs ~1 ns lifetimes) is insufficient for some scalable protocols requiring high indistinguishability.
• Field deployment showed increased g^(2)(0) and reduced entanglement fidelity compared to lab measurements due mainly to link and component losses, reducing signal-to-background ratio.
• The X-photon fibre exhibited higher loss (12.5 dB) and suffered leakage from neighbouring classical transceiver light (1309/1313 nm), necessitating additional narrowband filtering (+2.5 dB loss).
• Yield of suitable QDs is low (~0.9%), requiring selection; further improvements in growth or post-fabrication tuning could enhance device yield.
• Operation demonstrated under DC drive; future high-speed, pulsed operation requires device capacitance reduction and system optimization.