Presets
Playback
Network and route
Fibre and link generation
Memories, swapping, purification
Time
Camera
Layers
Quality
Model check
The science
What this page models
A quantum network distributes entangled pairs between distant nodes. Light in optical fibre loses about 0.2 dB per kilometre, so the chance that one photon survives a distance L is η = η0 · 10−αL/10. Quantum states cannot be copied or amplified (no-cloning), so the network instead cuts the route into segments, makes a heralded pair in every segment, stores the halves in quantum memories and joins neighbouring pairs by entanglement swapping. Choose a source and a destination on the map; the route is the shortest surviving fibre path, and cutting a fibre makes the network re-route.
Equations used
Elementary link. A source in the middle of a segment of length Ls sends one photon to each end. The pair is heralded with probability pb = η0 · 10−αLs/10 per attempt; with m parallel modes p = 1 − (1 − pb)m. One round lasts τ = 2Lmax/v + 10 µs, because the herald must travel back (v ≈ 2×105 km/s in glass).
Waiting for a whole chain. A pair is delivered only when every segment holds a link at the same time. That is the maximum of n independent geometric waiting times: E[rounds] = ∑k≥0 [1 − ∏i (1 − (1 − pi)k)]. With n−1 swaps, each succeeding with probability q, the delivery rate is R ≈ qn−1 / (E[rounds] · τ). Splitting a link into shorter segments removes the exponential dependence on the total length from the success probability of each attempt, which is why repeaters can turn an exponential law in L into a much gentler one.
Werner states. Every link is a Werner state ρ = w|Φ+>〈Φ+| + (1 − w) I/4 with fidelity F = (3w + 1)/4. Swapping two links gives w = w1w2, equivalently F = F1F2 + (1 − F1)(1 − F2)/3. A memory with coherence time T lets w decay as e−t/T; with the cutoff on, a link older than T is thrown away and regenerated.
Purification (BBPSSW). Two pairs of fidelity F are combined; with probability ps = F2 + 2F(1 − F)/3 + 5((1 − F)/3)2 the kept pair has F′ = [F2 + ((1 − F)/3)2] / ps. It raises F for every F between 0.5 and 1 but costs at least two pairs per round, so the page multiplies the link success probability by ∏ps/2k (a mean-rate equivalent).
Secret key. For a Werner pair the error rate in either basis is Q = 2(1 − F)/3 and the asymptotic BB84-type key fraction is 1 − 2h2(Q); it drops to zero near F ≈ 0.81. For comparison the page shows direct transmission through the same fibre at a 10 MHz source clock and the repeaterless PLOB bound, −log2(1 − η) ≈ 1.44 η secret bits per use.
Model checks (run in the page)
(1) A Monte-Carlo run of independent geometric waiting times must reproduce the analytic series for E[rounds] within sampling error. (2) The swap formula is verified by an explicit four-qubit density-matrix calculation: two Werner pairs are combined, qubits 2 and 3 are projected onto |Φ+> and the fidelity of the remaining pair is read off. (3) BBPSSW must raise F on (0.5, 1) with fixed points at 0.5 and 1; 0.2 dB/km over 15 km must give 3 dB; the key fraction must vanish near F = 0.81.
What to try
Set the spacing slider to 400 km (no repeaters) and watch the rate collapse; then lower it to 40 km. Shorten the memory time and see the cutoff discard links. Raise multiplexing to 50 modes. Lower the elementary fidelity to 0.88 and turn on two purification rounds: the key fraction recovers. Cut the fibres between Alice and N3 and between N1 and N3 and watch the route change.
Honest simplifications
Single memory per node and no buffering of pairs; all segments are synchronised to the slowest segment's round time; ideal sources, perfect gates and detectors apart from the single efficiency η0 (no dark counts, no gate noise); depolarising-only memory model with a sharp cutoff; purification is represented by its mean rate and ignores the extra memory time it needs; classical communication takes the same time as light in fibre; asymptotic key rate without finite-size effects. The map, node names and distances are an invented illustration, not a real network, and parameter values are typical lab-scale numbers, not a specific device.