A quantum internet cannot copy or amplify a qubit (no-cloning), so a repeater node must hold entangled halves in quantum memory until a matching partner from the other link arrives, then perform an entanglement swap (a joint Bell-state measurement) to stitch the two short links into one long-distance entangled pair.
Fidelity while stored: F(t) = F₀ · exp(−t / T₂)
Delivered fidelity: F_out ≈ F_L · F_R · η_swap
Decohered if: F(t) < 0.5 → discarded, counts as a drop
- Entanglement rate — how often each link (left, right) heralds a fresh entangled pair into an empty buffer slot. Real hardware: SPDC sources or atom-photon interfaces, typically 10²–10⁶ Hz depending on platform.
- Memory coherence T₂ — how fast a stored qubit's fidelity decays while it waits for its partner. Short T₂ (like most current quantum memories) means most pairs decohere before a swap is possible.
- Buffer size — slots per side. A full buffer forces new arrivals to be dropped (overflow) — this is the resource-contention side of quantum network architecture, distinct from the swap physics itself.
- Scheduling policy — FIFO serves the oldest waiting pair first (fair, but that pair may already be low-fidelity); Max-Fidelity first always swaps the freshest pair available (higher average output fidelity, but old pairs can starve and decohere anyway).
This scheduling problem — not the swap operation itself — is one of the open engineering challenges of a real quantum internet: memory is scarce and short-lived, so the network layer must decide who gets served before the qubits vanish.