78. Quantum Networks
78.1Quantum Communication
The fundamentals, physics-first: qubits can be transmitted (photonics) but not copied (no-cloning) or amplified (no quantum repeater-amplifier — amplification is copying). Entanglement can be distributed but not created at a distance (only shipped). And none of it signals faster than light (Ch. 2.10's locality discipline — the eternal correction). What communication buys: quantum key distribution (Ch. 51.5–51.8 — the deployed application), teleportation (state transfer consuming entanglement + classical bits), and remote entanglement as a resource for distributed gates. The engineering framing that organizes everything: a quantum network is a machine for distributing entanglement as a consumable resource, metered in rate and fidelity — e-bit bandwidth. Every protocol in this chapter is plumbing for that resource.
78.2Entanglement Distribution
The network's core service: prepare Bell pairs, send half of each to distant nodes, verify/herald success, then stockpile. The physics constraint chain: photons lose coherence in fiber exponentially (dB/km — ~50% loss per 10-15 km at telecom wavelengths), so distribution rate decays with distance until, past a few hundred km, direct links die; free-space/satellite links (Micius demonstrated 1,200 km entanglement distribution, 2017) trade atmosphere for no fiber. Heralding: entanglement attempts succeed probabilistically, with detectors announcing success — the network's flow control is literally "did it work?" Protocol stack forming in the field: entanglement generation (hardware) → purification (sacrifice pairs to raise fidelity) → swapping (stitch pairs into longer ones — the repeater's core trick) → consumption (application). Read network papers as you learned to read circuits in Part VI: as layered plumbing with a resource accounting.
78.3Quantum Repeaters
The enabling technology and the field's hardest network problem. Why classical repeaters fail: amplifying a quantum signal copies it — forbidden. The quantum answer (1998, Briegel et al.): repeater nodes with quantum memories — each node entangles with neighbors, then entanglement swapping (Bell-measure its two memories, teleporting the correlation across) extends the pair; purification between swaps raises fidelity. The status ladder: first-generation repeaters (memory + swapping + purification — memory lifetimes are the bottleneck: milliseconds-to-seconds demonstrated, seconds needed) are laboratory demos; second-generation (error-corrected links) and third-generation (fully encoded) are research programs. The gap matters: without repeaters, quantum networks are point-to-point islands; with them, continental mesh. Watch memory-coherence and swap-fidelity numbers the way Ch. 77.3 told you to watch Λ — they are the quantum internet's Moore curve, currently years behind the compute side's.
78.4Quantum Memories
The repeater's heart: interface and store a photonic qubit, hold it (ms-to-s), retrieve it faithfully. The technology zoo: atomic ensembles (EIT-storage), NV-center and other defects (electron/nuclear spins — seconds demonstrated), trapped ions (natural qubits, hard to interface with light), rare-earth crystals (multi-mode storage — bandwidth play), and cavities (fast interfaces, short lives). The metrics that matter: storage time × fidelity product, retrieval efficiency, wavelength compatibility (networks want telecom; memories often don't), and multiplexing (how many qubits per node — rate multiplier). The honest 2026 score: no memory simultaneously achieves telecom compatibility, >90% efficiency, second-scale coherence, and chip-scale integration — the quadfecta the quantum internet needs; each lab has two of four. This is a physics-and-materials problem with systems consequences — the field's clearest "unsolved" signpost (Ch. 80.1 will point back here).
78.5Distributed Quantum Computing
The compute application of networking: link small processors with entanglement-assisted gates (teleport a gate's action by consuming e-bits) and run circuits too wide for any single machine. Why it might win: modular engineering beats monolithic scaling (Ch. 77.6's economics — many small cryostats, cheaper yield, replaceable modules), and heterogeneous networks (ion-trap memory + photonic interconnect + superconducting speed) could combine platforms' strengths. The costs that make it hard: every teleported gate consumes entanglement (e-bits become the new gate count — Ch. 79's resource estimation gains a currency), network latency enters circuit scheduling (Ch. 44.7's timelines, now with round-trip times), and state routing across modules is compiler-class engineering that barely exists (Ch. 55.8's interface problems, network edition). The research reading: distributed-QC papers are the field's newest systems-design literature — and notably short of practitioners who can both read a stabilizer circuit and design a network protocol. That intersection is a career.
78.6Quantum Internet
The end-state vision (Wehner et al.'s staged formulation): a network of networks where quantum devices exchange qubits and entanglement the way the classical internet exchanges bits — stages running QKD (deployed) → entanglement distribution networks (demonstrated) → device-independent protocols (lab-scale: DI-QKD, certified randomness — Ch. 51.7) → distributed computing (research) → full mesh (horizon). What it would uniquely enable: cryptography beyond QKD (position-based, blind/verified computation — Ch. 44's research seam networked), networked sensing (entangled telescopes, clock synchronization — genuinely near-term physics), and modular FTQ (77.6). What it will not be: a replacement for the classical internet (entanglement is a consumable, rate-limited resource — the quantum layer is a co-processor network riding classical control). The mature view: the quantum internet is where the classical internet was in ~1975 — protocols being invented (a quantum TCP/IP is an open problem), physics limiting, and every layer undecided. Systems people wanted.