Quantum repeaters: how quantum links might reach further

A quantum repeater is a device that helps quantum links stretch over distances that a single optical fibre cannot cover. Quantum links carry entanglement, a special connection between two particles that can be used for secure communication and for linking quantum computers and sensors. The idea matters because a quantum network that spans cities or countries would need repeaters, and the longest memory-to-memory links reported in the sources reviewed for this article span tens of kilometres, not hundreds.

Why quantum signals cannot simply be boosted

An ordinary internet signal can be strengthened with an amplifier, which makes a copy of the signal that is stronger than the original. A quantum signal cannot be treated this way. The no-cloning theorem says it is impossible to make an independent, identical copy of an unknown quantum state. Because the information is carried by single photons, there is no copy to boost.

The Wikipedia article on quantum networks notes that long-distance communication is hindered by signal loss and decoherence, and that quantum signals cannot be amplified because qubits cannot be copied.

How a repeater works

A repeater divides a long link into shorter segments. Entanglement is created across each segment, and a middle station then combines the two segments by a measurement. This is called entanglement swapping. The two far ends end up entangled without any single photon having travelled the whole distance. To make this work, each node needs a quantum memory that holds its entanglement until the neighbouring segments are also ready. Physics World summarises the practical requirement: the memory's storage time should exceed the time needed to generate entanglement.

What has been demonstrated

These results are peer-reviewed experiments, so each one is listed with its own conditions.

• Three nodes, 2021. Researchers at QuTech in Delft built a network of three remote nodes, each based on diamond qubits. The Science paper reports genuine multipartite entanglement across all three nodes and entanglement swapping through a middle node, both run without postselection. In plain terms, the results were not picked out from the data afterwards.
• Two nodes, 10 km apart, 2024. The same Delft group linked two diamond-based nodes through a midpoint station, using 25 km of deployed fibre (15 km and 10 km on each side). The Science Advances paper reports a fully heralded fidelity of 0.534 at about 0.022 entangled states per second, which is about 1.3 per minute, with a 15 ns window. Fidelity measures how close the result is to the ideal state, so 0.534 is well short of perfect. The authors say the signal-to-noise ratio, mainly through false heralding events, accounts for about 30% of the lost fidelity. The heralded rate is much lower than the postselected rate of 0.48 per second, which uses a 20 ns window, mainly because of the added communication delay needed for the heralding signal to reach the nodes.
• Nuclear-spin memories, Harvard, 2024. A Harvard-led team including Mikhail Lukin entangled two nodes built from silicon-vacancy centres in nanophotonic diamond cavities. Photons were converted to 1350 nm, and the nuclear-spin memories were entangled through 40 km of fibre spools and a 35 km fibre loop in the Boston area (Nature paper; arXiv version). The abstract describes second-long entanglement storage. The authors call the work an enabling step towards practical repeaters. The abstract does not report a fidelity or a rate, so this result cannot be compared directly with the Delft figures.
• Several ions per node, 2025. A Caltech team (Ruskuc, Wu, Green, Hermans, Pajak, Choi and Faraon) describes a two-node network of several ytterbium ions in nanophotonic cavities, published in Nature in February 2025 (Nature paper; arXiv version). It entangles two remote ion pairs and prepares three-ion W states. The abstract says the work aims to ease bottlenecks in the entanglement distribution rate, but it gives no numbers.

Two further results are reported in the press and are not checked against the original papers here. A February 2026 news report describes a team at USTC led by Jian-Wei Pan and Qiang Zhang linking trapped calcium-40 ion memories over 10 km of spooled fibre in a lab, with entanglement generated in an average of 450 ms and held with a coherence time of 550 ms. The report says the result is published in Nature, and describes it as crossing a threshold that makes multi-stage repeaters physically possible for the first time. The same report says a full chain still needs entanglement swapping and purification at these distances and coordination between many nodes. An earlier proof of principle reported by Physics World in June 2023 describes a repeater node of two trapped calcium ions, built by teams at Innsbruck and Paris-Saclay, that distributed entanglement over 50 km using two 25 km fibres held in the lab, in a Physical Review Letters paper. The article gives no fidelity figure for that work.

What is speculative

Several uses are described for repeater networks, but none has been shown at scale. The Delft 2024 paper lists communication, computing, sensing and fundamental science as applications that future quantum networks hold the promise of enabling (Science Advances paper). Physics World describes linking quantum computers so they can share complex calculations, and secure communication, as possible uses (Physics World). These are reasonable goals, but they are not demonstrated.

Limits and open problems

• Memory lifetime. A memory must hold entanglement for longer than it takes the neighbouring links to succeed. The USTC report gives a coherence time only about 100 ms longer than the average generation time, so there is little margin (news report). Longer chains will need memories that last longer still.
• Photon loss and rate. The Delft link produced entangled states at about 0.022 per second, and its fidelity was limited by the signal-to-noise ratio, mainly through false heralding events (Science Advances paper). Higher rates over longer distances are still needed.
• From lab link to network. The results so far involve two or three nodes, and the sources reviewed show no repeater chain that carries entanglement beyond a single fibre span. The Wikipedia article notes that today's quantum key distribution networks rely on trusted repeaters and do not yet allow end-to-end transmission of qubits, or end-to-end entanglement, between far-apart nodes (Wikipedia).
• Error correction. The same Wikipedia article says error correction schemes that protect qubits over long distances would need a very large number of qubits, so they are currently limited to very short distances (Wikipedia).
• Cost and interoperability. The sources reviewed do not quantify the cost of building repeaters. The Wikipedia article states that quantum networks have not yet been standardised, although standards bodies have pre-standardisation work under way (Wikipedia).

What to watch next

• A chain of three or more links in which entanglement is swapped across the whole route, with the rate and fidelity reported.
• Memories whose storage time comfortably exceeds the time needed to generate entanglement, shown over longer fibre.
• Independent reading of the USTC Nature paper, with its fidelities and rates checked against the press report.
• Common standards for network nodes, which would make it possible for different systems to link up.