Entanglement distribution and quantum storage of more than 8000 modes over a metropolitan network

AI summary

Overview

Research area: Quantum information science, specifically fibre-based quantum repeaters, light–matter entanglement, and multimode (temporal-mode) quantum memories based on rare-earth-ion-doped crystals.

Technical level: Intermediate. The core ideas are accessible with undergraduate physics, but the Schmidt-decomposition analysis of the joint temporal amplitude and the CHSH Bell-test analysis assume familiarity with quantum optics.

Scope (one sentence): The paper demonstrates a broadband, highly multimode ¹⁷¹Yb³⁺:Y₂SiO₅ quantum memory that stores one photon of an entangled pair while the partner travels through up to 25.3 km of optical fibre, including a 5.66 km link through the Geneva metropolitan fibre network.

What This Paper Is About

A quantum repeater links remote locations by entangling telecom photons with matter memories, but the photon round-trip time in fibre (about 5 μs/km in silica) severely limits the entanglement rate if a memory can hold only one mode at a time. The fix is multiplexing: storing many temporal modes at once so that many entanglement attempts run in parallel. This paper builds a repeater node from a rare-earth quantum memory and a bandwidth-matched entangled-photon source, then quantifies exactly how many effective temporal modes the memory really stores — and shows that entanglement survives storage while the partner photon travels through a real city fibre link.

Key Contributions

1. A working multimode repeater node. A ¹⁷¹Yb³⁺:Y₂SiO₅ atomic-frequency-comb (AFC) memory with 250 MHz bandwidth and a (76.6 ± 11.0) μs memory time, paired with a bandwidth-matched spontaneous parametric downconversion (SPDC) source producing 979 nm signal and 1553 nm telecom idler photons.
2. A new quantitative measure of temporal mode capacity. The authors introduce a Schmidt decomposition of the joint temporal amplitude, and derive a lower bound N̄ on the Schmidt mode number that can be computed directly from the measured cross-correlation function g⁽²⁾(τ) without phase information.
3. Entanglement storage over lab fibre spools. Bell tests at fibre lengths from 0.5 km to 25.3 km, using a Franson interferometer and a double-AFC scheme, with up to 16340 temporal modes stored for 125 μs.
4. A metropolitan field deployment. Entanglement distributed over a 5.66 km standard dark fibre link in the Geneva metropolitan network between two university buildings separated by 2.15 km, while storing 8235 modes for 63 μs.

Main Findings

• Long memory lifetime in the quantum regime: The AFC efficiency decays as η_AFC = η₀ exp(−1/(ΔT_M)) with T_M = (76.6 ± 11.0) μs, corresponding to an effective AFC coherence lifetime T₂^AFC = 4T_M = (307 ± 44) μs — stated as the longest achieved in the quantum regime in any AFC memory to the authors' knowledge, though still shorter than the 1 ms optical coherence time.
• Baseline efficiency: At short storage times the measured efficiency is η₀ = (19 ± 1)%, in good agreement with the expected value for an optimal square-shaped AFC, essentially limited by the finite optical depth (d = 3.2 after spin polarization) of the sample.
• Large mode counts: Using the conservative lower bound N̄ and an effective mode duration T̄_mode = 7.65 ns (containing ∼94% of the photons in the correlation peak), the mode capacity ranges from 327 modes (2.5 μs storage, 0.5 km) to 16340 modes (125 μs storage, 25.3 km) — described as an improvement of one to two orders of magnitude over other works.
• Entanglement preserved for every fibre length: The CHSH parameter S is 2.72(3) at 0.5 km, 2.73(4) at 2.005 km, 2.73(3) at 6.521 km, 2.68(4) at 12.9 km and 2.70(4) at 25.3 km, all violating the classical bound S ≤ 2.
• High interference visibility: A phase scan at 1/Δ = 2.5 μs gives a visibility of V = (97 ± 1)% from raw data without subtracting dark counts or accidental coincidences. The measured S at 0.5 km matches the value S = 2.73 ± 0.03 expected from this visibility via S = 2√2 V, indicating no unexpected correlated noise.
• Consistency of the correlation level: The average cross-correlation ḡ_si⁽²⁾ stays roughly constant across all spool measurements (83, 85, 99, 94, 99), implying it is limited by the SPDC pair-creation probability; these values imply an expected CHSH parameter between S = 2.76 and 2.77, in good agreement with the measured values.
• Field results: Over the 5.66 km Geneva link, S = 2.65 ± 0.04 at 32 μs storage (4183 modes) and S = 2.67 ± 0.04 at 63 μs storage (8235 modes), demonstrating light–matter entanglement over a 2.15 km straight-line distance.
• Rate–mode trade-off: Memory efficiency decays exponentially with storage time, so coincidence counts fall as mode number grows (e.g. 49 coincidence counts per second at 125 μs and 25.3 km, versus 450 at 2.5 μs and 0.5 km). At 10 μs, 1307 modes are stored while efficiency still reaches 88% of its optical-depth-limited maximum.
• Validation of the metric: The experimentally extracted Schmidt mode number approaches the numerically computed theoretical lower bound N̄ for small data binning and stays below the full theoretical Schmidt mode number N, as expected.

Methodology in Plain English

The team used a crystal of yttrium orthosilicate doped with 2 ppm of ¹⁷¹Yb³⁺ (95% isotopic purity, dimensions 15.8 × 3.9 × 3.1 mm), cooled to about 3 K in a closed-cycle cryostat. The photons pass through the crystal four times to boost the optical depth. Using the atomic frequency comb technique, they burn a comb-like modulation into the absorption profile at the 978.87 nm transition with periodicity Δ; the comb stores light for a pre-set time of 1/Δ.

The light source is a periodically poled lithium niobate waveguide pumped at 600.35 nm (stabilized to 300 μW), generating pairs at 978.87 nm (signal) and 1552.52 nm (idler). The idler is filtered by a Fabry–Pérot cavity (90.3 MHz FWHM, 55 GHz free spectral range) and a dense wavelength-division multiplexer (channel 31, 75 GHz), while the signal is coarsely filtered by a 20 GHz volume Bragg grating before the memory's own 250 MHz acceptance acts as the final filter.

The experimental sequence runs on a 700 ms cycle: 300 ms of spin polarization, 64 ms of AFC burning, and 40 ms of storage trials (a 364 ms preparation phase in total). The SPDC pump is held on for exactly the storage time 1/Δ so that as many temporal modes as possible are populated, then the memory reads out for another 1/Δ. Single photons are detected with superconducting nanowire single-photon detectors and time-tagged for post-processing.

To probe entanglement, the telecom photon passes through a fibre Michelson interferometer and the 979 nm photon is analysed with a double-AFC scheme, giving 19.4 ns time bins and controllable phases on both sides. The authors then compute coincidence fringes, visibilities and CHSH S values. For the field test, the detector, DWDM filter and interferometer were moved to a second building, with a clock signal sent over a second fibre strand to synchronize the time tagger and phase stabilization.

Why This Matters

Impact on research: The paper supplies the first half of an elementary quantum repeater link and, more importantly, a standardized way to count how many modes a memory actually delivers. Since multiplexing is widely seen as the key to practical repeater rates, having a measurable quantity (the Schmidt mode number lower bound) that can be compared across platforms is a useful benchmarking tool. The demonstrated T₂^AFC of (307 ± 44) μs in the single-photon regime is also a record for AFC memories, and the authors note the AFC coherence time is only about one third of the optical coherence time, leaving room to improve by up to a factor of 3.

Real-world applications:

• Fibre-based quantum repeaters and long-distance quantum key distribution using standard telecom C-band hardware.
• Metropolitan-scale quantum networks, demonstrated here on deployed dark fibre rather than controlled lab spools.
• Distributed quantum computation and quantum sensing, which the paper cites as motivations for future quantum networks.
• Frequency- and time-multiplexed quantum hardware that reuses off-the-shelf telecom components such as DWDM filters and SNSPDs.

Industry relevance: The node uses 1553 nm telecom-wavelength photons, standard dual-strand single-mode fibre, a DWDM channel, and a classical service channel for clock synchronization — all compatible with existing fibre infrastructure. The authors note the telecom detection rates are in the kcps regime, representative of heralding rates expected for elementary links in DLCZ-type repeaters.

Future Directions

• Millisecond storage with on-demand readout: The memory duration must reach at least the millisecond regime and be tunable during storage. The spin-wave AFC scheme has been shown in ¹⁷¹Yb³⁺:Y₂SiO₅ in the classical regime for bandwidths up to 100 MHz and storage times up to a few milliseconds; the challenge is doing it at the single-photon level by suppressing noise from the intense control fields.
• Higher efficiency: Exploiting a low-finesse cavity in the impedance-matching regime is proposed to raise efficiency beyond the (19 ± 1)% set by the current optical depth.
• Benchmarking the metric: Whether the Schmidt mode number corresponds to the actual speed-up over a single-mode repeater is left as an open question. The authors show it holds exactly for a simplified model with perfect conditional-mode filtering, but in practice a rectangular detection window introduces excess noise. They propose comparing multimode (continuous-pump) repeater rates to ideal single-mode rates.
• Cross-platform comparison: The authors hypothesize that the Schmidt mode number could enable simple quantitative benchmarking of different quantum repeater experiments in terms of multimode capacity and potential speed-up.

Target Audience

Quantum-photonics experimentalists and theorists working on quantum repeaters, quantum memories and entanglement distribution; engineers building metropolitan quantum communication infrastructure; and graduate students entering the field who need a clear worked example of how multimode capacity is defined, measured and validated against a Bell test. Readers interested in rare-earth-ion-doped materials will find the specific Y₂SiO₅ parameters (doping, optical depth, coherence lifetimes) particularly useful.

Authors’ abstract

Entanglement generation between telecommunication photons and matter is central to fibre-based quantum repeaters. Achieving practical communication rates requires multiplexing, which multimode quantum memories can provide. Rare-earth-ion ensembles offer large temporal multimode storage by exploiting the numerous spectral channels within their absorption spectrum. Here, we report on a quantum repeater node comprised of a $^{171}$Yb$^{3+}$:Y$_2$SiO$_5$ multimode quantum memory, featuring a 250 MHz bandwidth and a $76.6~μ\mathrm{s}$ lifetime, and a bandwidth-matched entangled photon-pair source. We introduce and validate a quantitative measure of the effective temporal mode capacity using a Schmidt decomposition. With this platform, we demonstrate entanglement between a telecom photon propagating through a 25.3 km fiber spool and a 979 nm photon stored for $125~μ\mathrm{s}$ across 16340 temporal modes. Finally, we report a field deployment distributing entanglement over 5.66 km through the Geneva metropolitan fibre network while storing 8235 modes for $63~μ\mathrm{s}$.