A quantum repeater is not a device that measures an unknown qubit and sends a fresh copy onward. Its job is to create, store, connect and verify entanglement across shorter segments so that distant nodes obtain an end-to-end quantum resource without trusting intermediate nodes with the final secret.
Long-distance quantum communication is difficult because photon transmission through fibre or free space is lossy. Direct success probability falls exponentially with distance. Classical optical networks solve loss by detecting and regenerating signals. That strategy does not work for an arbitrary unknown quantum state because measurement destroys coherence and no-cloning forbids perfect copying.
Quantum repeaters change the architecture. First create short-range entangled links. Store successful links in quantum memories. Connect neighbouring links by entanglement swapping. Purify or error-correct as necessary. Repeat this hierarchy until the endpoints share sufficiently high-quality entanglement.
Lossy channel → heralded short links → quantum memory → swapping → purification/error correction → longer entanglement → end-to-end key, teleportation or distributed computation.
1. Fibre loss is exponential in distance
If fibre attenuation is α dB per kilometre and distance is L kilometres, ideal transmission probability scales as
η(L)=10^{-αL/10}.
At telecom wavelengths, a representative low-loss fibre value is around 0.2 dB/km, though real links include connectors, coupling and detector losses too.
For α=0.2 dB/km and L=100 km,
η=10^{-2}=0.01.
Only about one photon in one hundred survives the ideal fibre attenuation before other losses.
2. Doubling distance squares transmission
Because η(L)=e−L/Latt in exponential form,
η(2L)=η(L)^2.
If 100 km gives 1% fibre transmission, 200 km gives approximately 0.01%=10−4.
This exponential distance penalty is the basic reason long-distance direct quantum communication becomes rate limited.
3. Repeaterless capacity gives a stronger benchmark than raw transmission
For a pure-loss bosonic channel with transmissivity η, the PLOB result gives the ultimate two-way-assisted secret-key/entanglement-distribution capacity without repeaters as
C=-log₂(1−η)
bits per channel use or optical mode under the theorem’s model. [4]
For η≪1, use -ln(1-η)≈η, so
C≈η/ln2≈1.4427η.
The repeaterless bound therefore inherits the same exponential distance decay. A repeater architecture is valuable if its end-to-end rate can asymptotically beat this point-to-point limit under comparable resource accounting.
4. Worked repeaterless bound
For η=0.01,
C=-log₂(0.99)≈0.01450 bits/mode.
For η=10−4,
C≈1.4427×10^-4 bits/mode.
The capacity falls by about a factor of one hundred when the distance increase makes transmissivity one hundred times smaller.
5. Divide the channel into elementary links
Suppose total length L is split into N equal elementary segments of length L/N.
Each short link has much larger transmissivity
η_seg=10^{-αL/(10N)}.
For L=200 km, α=0.2 dB/km and N=4, each segment is 50 km with ηseg=10−1=0.1 rather than the direct end-to-end η=10−4.
The challenge is then to connect successful short links without destroying the quantum information.
6. Heralded entanglement generation
An elementary link attempts to create entanglement between neighbouring repeater nodes. A detector event or another classical flag heralds whether the attempt succeeded.
If one attempt succeeds with probability p and attempts are independent, the number T of attempts until success follows a geometric distribution:
Pr(T=t)=(1-p)^{t-1}p.
Its mean is
E[T]=1/p.
If p=0.01, the average link needs one hundred attempts.
7. Attempt time includes classical heralding
A link attempt may require a photon flight and a classical heralding signal. If the segment length is L₀ and signal speed in fibre is v≈2×10⁸ m/s, a round-trip communication time scale is roughly
τ≈2L₀/v
plus source, detector and reset latency.
For L₀=50 km, 2L₀/v≈0.5 ms. If p=0.01 and attempts cannot be multiplexed, the mean elementary-link creation time is about 50 ms before local processing overhead.
8. Quantum memories change sequential waiting into parallel accumulation
Suppose two neighbouring elementary links are needed before a swap can be attempted. Without memory, both would need to succeed simultaneously. With memory, the first successful link can be stored while waiting for the second.
This is the crucial resource: a quantum memory preserves entanglement across uncertain waiting times.
Memory coherence time, retrieval efficiency and fidelity therefore directly affect repeater rate and output quality.
9. Waiting for two independent links
Let T₁ and T₂ be independent geometric waiting times with success probability p per attempt. We need max(T₁,T₂) attempts before both links exist.
The exact mean is
E[max(T₁,T₂)]=(3−2p)/(p(2−p)).
For small p this approaches
E[max]≈3/(2p).
With p=0.01, the exact mean is about 149.75 attempts, much less than the 10,000-attempt scale of demanding simultaneous success p² but longer than the 100 attempts for one link.
10. Derive the two-link waiting formula
Use
max(T₁,T₂)=T₁+T₂−min(T₁,T₂).
Each E[T]=1/p. The minimum succeeds when at least one of two links succeeds in a round, whose probability is
1−(1−p)^2=p(2−p).
Therefore E[min]=1/[p(2−p)], and subtraction gives the expression above.
11. Entanglement swapping connects the links
Suppose A–B and B–C share Bell pairs. Node B performs a Bell-state measurement on its two local qubits.
Conditioned on the measurement outcome, A and C become entangled even though they never interacted directly. A known Pauli frame depends on B’s classical result.
This is the same identity developed in Guide 30, now used recursively to extend a network link.
12. Swap success probability
Some physical Bell-state measurements are deterministic in an ideal matter-qubit gate model; linear-optical Bell measurements without extra resources can have success probability below one.
If a swap succeeds with probability ps, failed swaps may destroy the two elementary pairs and force regeneration.
A simple two-link average-time estimate is therefore
T_swap≈T_both/p_s
when each failed swap requires a complete fresh cycle and local swap time is negligible.
With p=0.01, τ=0.5 ms and ps=0.5, the earlier 149.75-attempt mean gives roughly 74.9 ms to create both links and approximately 149.8 ms per successful swapped pair under this simplified restart model.
13. Nested swapping
For four elementary segments, first create and swap segments 1–2 and 3–4 in parallel, producing two length-2 links. Store them. Then swap those two longer links to create the length-4 connection.
This binary nesting repeats over scales 1,2,4,8,… segments.
Each level introduces additional waiting, swap failure and memory exposure. The original Briegel–Dür–Cirac–Zoller repeater architecture combines this nesting with entanglement purification. [1,2]
14. Fidelity degrades under imperfect swapping
If elementary pairs are noisy, swapping usually propagates and combines their errors. If local gates and measurements are imperfect, they add new errors.
Repeated nesting can therefore produce an end-to-end pair too noisy for teleportation or key generation even when every individual short link looks acceptable.
Rate and fidelity must be propagated together. A fast repeater that outputs unusable entanglement has not solved the communication problem.
15. Memory decoherence during waiting
As a simple dephasing model, suppose stored coherence decays as
v(t)=e^{-t/T_2}.
If one link succeeds early and waits 50 ms in a memory with T₂=500 ms, its visibility factor becomes e−0.1≈0.9048.
If the same wait occurs in a 50 ms memory, the factor is e−1≈0.3679.
Memory lifetime must therefore be compared with the stochastic distribution of link-generation times, not just with one local gate duration.
16. Multiplexing raises effective success probability
If m independent spectral, temporal or spatial modes attempt an elementary link in parallel, each with success probability p, the probability that at least one succeeds is
p_eff=1−(1−p)^m.
For p=0.01 and m=100,
p_eff=1−0.99^100≈0.634.
Multiplexing can transform a very slow heralded link into a high-probability per-clock-cycle resource, provided sources, detectors and memories support the required number of modes.
17. Purification trades quantity for quality
Entanglement purification consumes multiple imperfect pairs and local operations to probabilistically produce fewer pairs of higher fidelity.
The protocol rejects some outcomes and consumes both input pairs even when it succeeds, reducing throughput.
The gain is that swapping can proceed from a cleaner resource, preventing fidelity from collapsing over many nesting levels.
Purification is therefore another rate-versus-quality trade-off, analogous to magic-state distillation but applied to shared entanglement rather than non-Clifford computation.
18. Quantum error correction can replace two-way purification
Later-generation repeater architectures use encoded quantum error correction so logical quantum information can be forwarded or teleported through a chain with one-way classical feed-forward rather than waiting for repeated two-way purification acknowledgements.
This can improve rate over long distances when local gates and syndrome extraction are sufficiently reliable, but it greatly increases physical-qubit overhead.
Repeater generations therefore trade communication latency against local hardware complexity.
19. Trusted relay versus quantum repeater
A trusted classical relay can receive a secret key from one link and forward or combine it with another link. Security then requires trusting the relay with the key.
A true entanglement-based quantum repeater can distribute end-to-end entanglement so intermediate nodes need not learn the final teleported quantum state or secret key.
The distinction is architectural and security-critical. A chain of trusted QKD nodes is not the same thing as a quantum repeater network.
20. End-to-end rate is limited by bottlenecks
Consider a route through links with sustainable entanglement-generation capacities R₁,R₂,… and no ability to create end-to-end pairs faster than the slowest required segment.
A simple cut argument gives an end-to-end rate no larger than the minimum relevant cut capacity.
Adding many high-rate links cannot compensate for one narrow bottleneck if every end-to-end resource must cross that cut.
Quantum network routing is therefore a flow problem constrained by entanglement generation, memory occupancy, purification demand and swapping schedules.
21. Worked three-link bottleneck
Suppose three neighbouring links can supply purified Bell pairs at 1000, 600 and 900 pairs/s.
Even with perfect deterministic swapping and unlimited memories, a chain that consumes one pair from each segment per end-to-end pair cannot exceed 600 pairs/s.
If each of two swap stages succeeds with probability 0.8 and failures require new resource pairs, a crude throughput scale falls further by about 0.8²=0.64, giving at most roughly 384 pairs/s before waiting correlations and scheduling overhead.
This multiplication is a simplified independent-success estimate, not a full queueing model.
22. Network entanglement is consumable inventory
An end-to-end Bell pair can be consumed by teleportation, dense coding, remote gate protocols, clock synchronisation, distributed sensing or cryptographic key generation.
A quantum network therefore needs inventory management: which node pairs currently share entanglement, with what fidelity, age and expiration time?
Routing algorithms may choose to generate entanglement speculatively before an application asks for it, trading memory occupancy against latency.
23. Quantum memories create queueing problems
If one link repeatedly succeeds while its neighbour is slow, memory slots can fill with stored pairs waiting for partners.
Old pairs decohere while waiting. A scheduler may choose to discard stale pairs and regenerate them rather than consume low-fidelity inventory.
End-to-end rate optimisation is therefore not only a physical-layer success-probability calculation. It is also a stochastic scheduling and queueing problem.
24. Repeater spacing has an optimum
Making elementary links shorter increases photon-transmission success. But it also increases the number of repeater nodes, memories, swaps and local operations.
Making links longer reduces node count but makes each heralded entanglement attempt much less likely to succeed.
The optimal spacing depends on attenuation, detector efficiency, memory lifetime, swap success, multiplexing, node cost and the chosen repeater generation.
25. Common misconception: a repeater amplifies a qubit
A quantum repeater cannot measure and regenerate an arbitrary unknown state like a classical optical amplifier. It extends entanglement through swapping, purification and/or quantum error correction.
26. Common misconception: shorter segments automatically give higher end-to-end rate
Shorter links improve transmission but add nodes, swaps, memories and local-error exposure. The complete network has an architecture-dependent optimum.
27. Common misconception: high elementary-link fidelity guarantees high final fidelity
Repeated swapping, storage and local gates accumulate error. Fidelity must be tracked through every nesting level and waiting-time distribution.
28. Worked synthesis problem
A 200 km fibre link with 0.2 dB/km loss is divided into four 50 km elementary links. Each attempt has additional source/detection factors so the total elementary entanglement success probability is p=0.02. Attempts are clocked every 0.5 ms. Memories are available.
Step 1: One-link mean. E[T]=1/0.02=50 attempts, or 25 ms.
Step 2: Two parallel links. Use (3−2p)/(p(2−p)) to obtain approximately 74.75 attempts, or 37.4 ms, before both neighbouring links are ready.
Step 3: First swap. If swap success ps=0.8 and failure forces regeneration, a simple restart estimate gives about 46.7 ms per length-100 km pair.
Step 4: Build two 100 km halves in parallel. Their waiting time again exceeds one-half mean because the slower side determines readiness.
Step 5: Final swap and memory. The first completed half may wait tens of milliseconds for the second, so memory decoherence and swap fidelity must be included before quoting a usable end-to-end Bell-pair rate.
The example shows why multiplying raw probabilities alone is not enough. A repeater is a stochastic synchronisation system.
29. Practice set
- Write fibre transmissivity in terms of dB/km attenuation and distance.
- What happens to η when distance doubles in a simple exponential-loss model?
- State the pure-loss repeaterless PLOB capacity formula.
- What is the mean number of independent attempts for success probability p?
- Why are quantum memories essential in first-generation repeaters?
- What is the small-p waiting-time scale for two independent links to both succeed?
- What operation connects A–B and B–C Bell pairs into A–C entanglement?
- Why does swap success probability affect throughput?
- How can multiplexing change link success probability?
- What is the purpose of entanglement purification?
- Why is a trusted QKD relay not the same as a quantum repeater?
- Name three variables that can determine optimal repeater spacing.
Answers
η=10^{-αL/10}.- Transmission squares:
η(2L)=η(L)^2. -log₂(1-η)bits per mode under the pure-loss theorem model.1/p.- They store early successful links while neighbouring stochastic links are still being generated.
- Approximately
3/(2p)attempts. - Entanglement swapping via a Bell-state measurement at B.
- Failed swaps may consume stored pairs and force regeneration.
p_eff=1-(1-p)^mfor m independent parallel modes.- Consume several noisy pairs to produce fewer higher-fidelity pairs.
- A trusted relay learns or must be trusted with key material; a repeater distributes end-to-end entanglement without requiring intermediate trust in the final quantum data.
- Examples: fibre loss, detector efficiency, memory coherence, swap fidelity, multiplexing, node hardware cost and local gate error.
Sources and further study
[1] H.-J. Briegel, W. Dür, J. I. Cirac and P. Zoller, Quantum repeaters for communication. The foundational nested-repeater proposal.
[2] W. Dür, H.-J. Briegel, J. I. Cirac and P. Zoller, Quantum repeaters based on entanglement purification. A detailed analysis of purification, imperfect local operations and nested long-distance links.
[3] Nicolas Sangouard, Christoph Simon, Hugues de Riedmatten and Nicolas Gisin, Quantum repeaters based on atomic ensembles and linear optics. A comprehensive repeater review covering rates, memories and optical architectures.
[4] Stefano Pirandola, Riccardo Laurenza, Carlo Ottaviani and Leonardo Banchi, Fundamental Limits of Repeaterless Quantum Communications. The repeaterless rate-loss benchmark for pure-loss and other channels.
Continue through Quantum Mathematics
Guide 33: Quantum Key Distribution, BB84, Entropic Uncertainty and Security Proofs explains end-use secret-key generation. Guide 34: Quantum Authentication, Private Quantum Channels and the Quantum One-Time Pad protects quantum payloads. Guide 36: Device-Independent Randomness, Self-Testing and Bell-Certified Security uses network Bell correlations as certification evidence.