A Tiny Quantum Network

You have seen three quantum primitives in isolation. Quantum teleportation (ch.50) moves a qubit state from Alice to Bob if they share a Bell pair. Entanglement swapping (ch.52) stitches two Bell pairs at a middle node into a single longer Bell pair. Superdense coding (ch.51) sends two classical bits down a single qubit channel when a Bell pair is pre-shared. Each primitive, on its own, has a clean protocol description and a clean proof of correctness.

What the world needs is for these primitives to compose. A useful quantum communication system is not one that moves one qubit over 10 metres in a lab; it is one that moves qubits over continental distances, across many nodes, through unreliable links and noisy hardware. Nobody has built that. But every proposal for how to build it starts with the same minimal configuration: three nodes, two links, a memory in the middle. This is the tiny quantum network — the smallest network bigger than a point-to-point line — and it is the piece you stitch end to end to make a quantum internet.

This chapter derives the full three-node protocol, step by step. You will see exactly how teleportation and swapping compose, how the resources are counted, what the hardware at each node needs to do, and why even this "tiny" configuration has not yet been run end-to-end in production.

The setup — three nodes, two links

Draw the picture before the formulas.

There are three physical nodes, and five qubits in play across them.

• Alice holds two qubits: S (the source qubit, in an unknown state |\psi\rangle = \alpha|0\rangle + \beta|1\rangle she wants to transmit), and A_1 (her half of the first Bell pair).
• Mid holds two qubits: A_2 (the other half of the first pair) and B_1 (one half of the second pair). Mid's two qubits are not entangled with each other — each is entangled with a different outside qubit.
• Bob holds one qubit: B_2 (the other half of the second pair). He will receive |\psi\rangle at the end.

The two Bell pairs are prepared in advance by local photon-pair sources — one source sits between Alice and Mid (distributing A_1 to Alice and A_2 to Mid), another sits between Mid and Bob (distributing B_1 to Mid and B_2 to Bob). Each pair is in the state |\Phi^+\rangle = \tfrac{1}{\sqrt{2}}(|00\rangle + |11\rangle). Each link has whatever photon-survival probability the hardware allows, and we assume success: once the pairs are shared and stored in the three nodes' quantum memories, the network is ready to run the protocol.

Classical channels — normal electromagnetic communication, speed-of-light constrained — connect every pair of nodes. We'll use them to send a total of four classical bits during the protocol (two from Mid to Bob for the swap, two from Alice to Bob for the teleport).

No direct quantum channel connects Alice to Bob. This is the whole point of the network: Alice's qubit |\psi\rangle will reach Bob without any physical quantum carrier ever traversing the Alice–Bob distance.

The protocol — four steps

Here is the full procedure. Walk through the diagram below while reading each step.

1. Entanglement swap at Mid. Mid performs a Bell measurement on its two qubits A_2 and B_1. The measurement yields two classical bits (m_1, m_2), and projects the outer pair (A_1, B_2) — Alice's and Bob's qubits — onto a specific Bell state determined by (m_1, m_2). This is exactly entanglement swapping (ch.52): two short Bell pairs become one long Bell pair.
2. Pauli correction from swap. Mid sends (m_1, m_2) to Bob over the classical channel. Bob applies the Pauli correction X^{m_2} Z^{m_1} to B_2. After this, Alice and Bob share a deterministic |\Phi^+\rangle_{A_1 B_2} Bell pair.
3. Teleport at Alice. Alice performs a Bell measurement on her two qubits S and A_1, yielding two more classical bits (t_1, t_2). This is the first half of the teleportation protocol (ch.50).
4. Pauli correction from teleport. Alice sends (t_1, t_2) to Bob. Bob applies X^{t_2} Z^{t_1} to B_2. After this, Bob's qubit is exactly |\psi\rangle — the state Alice started with on her source qubit S.

At no step does any qubit travel from Alice to Bob directly. The only things that cross the full Alice-to-Bob distance are classical bits — four of them, at sub-light speed. The quantum carriers (the photons that established the Bell pairs) each traversed only one link, and they did so before the protocol began.

Why it works — the state trace

Track the state of all five qubits through each step. Use qubit ordering (S, A_1, A_2, B_1, B_2).

Initial state. Alice's source S is in |\psi\rangle = \alpha|0\rangle + \beta|1\rangle. The two Bell pairs are |\Phi^+\rangle_{A_1 A_2} = \tfrac{1}{\sqrt 2}(|00\rangle + |11\rangle) and |\Phi^+\rangle_{B_1 B_2} = \tfrac{1}{\sqrt 2}(|00\rangle + |11\rangle). The joint state:

Expand:

Why we leave it in this factored form for now: the first non-trivial operation is the Bell measurement on (A_2, B_1), and the most transparent way to predict its effect is to express the middle-pair subsystem in the Bell basis — just like in ch.52 on entanglement swapping. We'll expand step by step.

Step 1 — Mid's Bell measurement on (A_2, B_1). From the entanglement-swapping chapter, you know the identity:

where the sum runs over the four Bell states |\Phi^+\rangle, |\Phi^-\rangle, |\Psi^+\rangle, |\Psi^-\rangle. Applied to our state, and tensoring with Alice's source:

Mid measures (A_2, B_1) in the Bell basis, selecting one of the four Bell states with probability \tfrac{1}{4}. Say Mid reads (m_1, m_2) = (0, 0), which the standard Bell-measurement circuit maps to |\Phi^+\rangle_{A_2 B_1}. (The same analysis extends to the other three outcomes via Pauli corrections.)

After Mid's measurement and conditioning on (0, 0):

Why A_1 and B_2 are now entangled despite never having met: the two middle qubits were each entangled with an outside qubit. Measuring the middle pair in the Bell basis collapses the outer two into a correlated Bell state — this is exactly entanglement swapping. The four branches of the Bell measurement map one-to-one onto the four Bell states of the outer pair, and the two classical bits (m_1, m_2) label which outer Bell state popped out.

Step 2 — Bob applies the swap correction. Bob receives (m_1, m_2) = (0, 0). The swap table (ch.52, Example 2) says: for (0, 0), apply I (do nothing). The Alice–Bob pair is already in |\Phi^+\rangle; no correction needed. For the other three outcomes, Bob applies Z, X, or XZ respectively.

After step 2, regardless of what Mid's outcome was, Alice and Bob deterministically share |\Phi^+\rangle_{A_1 B_2}. The state is

Mid's two qubits A_2 and B_1 are now in the specific computational-basis state labelled by (m_1, m_2) — they are used up and can be discarded (or reused later after re-initialisation).

Step 3 — Alice's Bell measurement on (S, A_1). This is the first half of teleportation (ch.50). Alice applies CNOT with S as control and A_1 as target, then Hadamard on S, then measures both S and A_1 in the computational basis. The result is two classical bits (t_1, t_2) and Bob's qubit B_2 is in a Pauli-distorted version of |\psi\rangle:

Each outcome has probability 1/4.

Step 4 — Bob applies the teleport correction. Alice sends (t_1, t_2) to Bob. Bob applies X^{t_2} Z^{t_1} to B_2. The correction inverts the Pauli distortion:

• (0, 0): apply nothing. B_2 = |\psi\rangle. ✓
• (0, 1): apply X. X \cdot X|\psi\rangle = |\psi\rangle. ✓
• (1, 0): apply Z. Z \cdot Z|\psi\rangle = |\psi\rangle. ✓
• (1, 1): apply ZX. ZX \cdot XZ|\psi\rangle = |\psi\rangle (up to global phase). ✓

Final state. In all cases, Bob's qubit B_2 is in |\psi\rangle. Alice's qubits S and A_1 are in the computational-basis state |t_1 t_2\rangle — used up. Mid's qubits are already used up from step 1.

Why the protocol is deterministic despite containing two random measurements: each measurement's four possible outcomes are each paired with a specific Pauli correction that exactly inverts the measurement-dependent distortion. The randomness is absorbed by the correction, and the final result is always |\psi\rangle. The same accounting principle drove both the teleportation and swapping chapters.

The protocol is complete. Bob holds |\psi\rangle, which Alice started with. No qubit ever moved from Alice to Bob directly; all that moved between them was four classical bits (two from Mid, two from Alice).

Resource accounting

Count what the protocol used.

• Quantum memory operations: each node holds a qubit across at least one classical-communication round-trip (Mid holds A_2 and B_1 until the swap; Alice holds A_1 until the teleport; Bob holds B_2 until he receives both correction messages). The memory time is the critical hardware parameter — it must exceed the classical round-trip delay by a margin sufficient for gate operations.
• Bell pairs consumed: two. Pair 1 (Alice–Mid) and pair 2 (Mid–Bob). Both are used up; neither is reusable after the protocol.
• Classical bits transmitted: four total. Two from Mid to Bob (swap corrections), two from Alice to Bob (teleport corrections). Classical communication scales at most linearly with the number of hops.
• Quantum channel length traversed by any one qubit: one short hop. Photons from pair 1 traverse Alice–Mid; photons from pair 2 traverse Mid–Bob. The source qubit S stays in Alice's lab. Bob's qubit B_2 stays in Bob's lab.

This is the fundamental trade. You paid for entanglement in advance (distributing two Bell pairs over short hops); during the protocol, you paid only in classical bits and local gate operations. The clever accounting is what the phrase "quantum internet" really describes: an infrastructure that pre-distributes entanglement over a network of short links, so that applications later consume that entanglement to achieve long-distance quantum communication without long-distance quantum transmission.

The hardware ingredients

What does each node need?

• A quantum memory: a system that holds a qubit in a definite state long enough to wait for the classical round-trip and any additional protocol steps. Typical requirements: microseconds to milliseconds for a single tiny-network protocol; seconds to hours for multi-node chaining. Hardware candidates: atomic ensembles (DLCZ-style), trapped ions, NV centres in diamond, rare-earth-doped crystals. Current records (as of 2024): about one second in NV centres at room temperature, several hours in cryogenic rare-earth crystals.
• A Bell-pair source: a system that produces entangled photon pairs on demand. The standard technology is spontaneous parametric down-conversion (SPDC) in a nonlinear crystal. Probability of producing an entangled pair per pump pulse is typically 10^{-3} to 10^{-6}, so you pump at high repetition rate and filter out the empty pulses using coincidence detection.
• A Bell-measurement apparatus: a system that can perform the CNOT + H + measure sequence on two qubits. On photonic hardware, full Bell measurement is only 50% efficient using passive linear optics (Lütkenhaus-Calsamiglia bound); higher efficiency requires ancilla photons or hybrid photon-atom systems.
• A classical communication channel: just normal optical or radio infrastructure. Latency matters: every classical bit adds roughly L/c to the protocol time, where L is the link distance.
• A clock: coordinated timing across nodes so that quantum memories and classical signals arrive in sync. Typical precision requirement: nanoseconds to microseconds.

None of these is simple. All are active research topics. The fact that individual components exist in separate labs around the world does not mean they have been assembled into a working three-node network yet — though several groups (Delft QuTech, Innsbruck, Argonne) have demonstrated subsets of this protocol on small scales.

Worked examples

Common confusions

• "This is a quantum internet." Not yet. Three nodes with a pre-distributed pair of Bell pairs is the minimal demonstration of what a quantum internet does. A real quantum internet needs routing protocols (which path through a multi-graph of nodes should a qubit take?), admission control (which users get the limited entanglement budget first?), quantum error correction at the link and network layers, and integration with classical network infrastructure. None of this exists yet in deployed form.

• "Network means routing." In classical networking, yes — a network has routers that decide which next hop to take. In quantum networking, routing is still an open research area. The 2018 Wehner-Elkouss-Hanson roadmap [1] lays out six stages of increasing capability, and routing is explicitly in the later stages. The tiny three-node network you derived is stage 3 (entanglement generation with memory) out of 6.

• "Photonic vs matter qubits — which is better?" Different trade-offs. Photons are excellent carriers — they fly through fibre at the speed of light, don't decohere rapidly, and are easy to generate in entangled pairs. But they are terrible at being stored (you can't put a photon on a shelf; if you stop it, it's gone). Matter qubits (atoms, ions, nuclei, NV centres) are good at being stored — coherence times of seconds to hours — but are terrible at being moved long distances. Every practical quantum network is hybrid: photons for the long hops, matter qubits for the storage and processing at each node. The mapping between flying photon qubits and stationary matter qubits (called light-matter interfaces) is one of the hardest engineering challenges in the field.

• "Why not just send |\psi\rangle directly from Alice to Bob over a quantum channel?" Because the direct channel has limited length. Photon transmission in optical fibre drops by about 0.2 dB/km; past a few hundred kilometres, the single-photon transmission probability is below 10^{-10}, which means you'd need to wait hours to get a single surviving photon. A tiny network with intermediate nodes exchanges this exponential loss for linear memory cost. This is what makes the architecture practical in principle, even if hardware is still years from realising it.

• "The protocol is too slow because of classical communication." Yes, the classical bits take time to travel. Classical latency sets the minimum protocol time. For an Alice–Bob distance of 1000 km through Mid, the round-trip classical-light travel time is about 7 ms — very long compared to gate times (nanoseconds) but short compared to decoherence of good quantum memories. Classical latency is not the bottleneck; quantum memory lifetime and Bell-pair generation rate are. This is why every serious repeater proposal focuses on memories and not on classical networking.

• "The source qubit S travels to Bob." No. S stays in Alice's lab the entire protocol. What travels is Alice's Bell-measurement outcome (t_1, t_2) — two classical bits. After Alice measures, S is in a computational-basis state (|0\rangle or |1\rangle) with no trace of the original |\psi\rangle. Bob's qubit B_2, which started empty (as half of a Bell pair), becomes |\psi\rangle after his corrections.

Going deeper

Quantum repeaters — DLCZ and beyond

The DLCZ protocol (Duan, Lukin, Cirac, Zoller, 2001) [2] is the canonical quantum repeater architecture. It uses atomic-ensemble quantum memories — clouds of millions of atoms that collectively hold a single collective excitation — and photonic links. Key ideas:

• Each node has an atomic ensemble. A weak laser pulse has a small probability of creating one collective excitation in the ensemble and emitting one photon in a known mode.
• The emitted photon goes down the fibre to a beam splitter halfway between two nodes. It interferes with the photon from the other side. A click at one of two single-photon detectors heralds an entangled pair between the two ensembles.
• Heralded entanglement is stored in the memories until the next repeater hop is ready.
• When all hops are ready, entanglement swapping stitches them together.

Second-generation repeater proposals (2010s) add error correction at the link layer, using encoded logical qubits rather than bare physical ones. Third-generation proposals (all-photonic or memory-less) use error-correcting codes to replace memories entirely, at the cost of more complex photonic circuits.

Entanglement purification

Real Bell pairs are not perfect. Imperfect state preparation, photon loss, and gate errors produce a noisy mixed state — close to but not exactly |\Phi^+\rangle. If you tried to run teleportation or swapping on noisy pairs, the errors compound and the protocol's fidelity drops.

Entanglement purification (Bennett-DiVincenzo-Smolin-Wootters, 1996) is a protocol that takes several noisy Bell pairs and produces fewer but higher-fidelity pairs, by applying CNOT gates between pairs and measuring one of them. The measurement outcome "heralds" either a purification success (keep the pair, now higher fidelity) or failure (discard). Over many rounds, the average fidelity of the kept pairs approaches 1.

This is the key protocol that makes repeater chains fault-tolerant. Without purification, noise would compound across hops; with it, the fidelity can be kept above the required threshold at each link.

Photon loss and mitigation

Fibre loss is roughly 0.2 dB/km at the 1550 nm telecom wavelength. After 100 km, about 1% of photons survive; after 500 km, about 10^{-10}. Classical amplification cannot help: amplifying a single photon in an unknown quantum state violates no-cloning. Quantum repeaters are the only way around this.

Free-space optical links (horizontal or satellite-to-ground) have different loss profiles — atmospheric absorption at ground level, geometric beam divergence at long distances, detector efficiency. Satellite-based QKD demonstrations (China's Micius, ESA's TESAT) use free-space links to reach thousands of kilometres by going above the atmosphere.

The Wehner-Elkouss-Hanson quantum-internet roadmap

Stephanie Wehner, David Elkouss, and Ronald Hanson proposed a six-stage roadmap for quantum internet capability in 2018 [1]. The stages are:

1. Trusted-node networks (today, BB84-style QKD).
2. Prepare-and-measure networks (point-to-point quantum communication without memory).
3. Entanglement generation with memory (the tiny network of this chapter).
4. Quantum computing networks (small-scale distributed quantum computation).
5. Few-qubit fault-tolerant networks.
6. Full quantum internet (arbitrary distributed quantum computation, error-corrected, high-rate).

Current technology is at stage 2 (with some stage-3 lab demonstrations at Delft and Innsbruck). India's 2022 ISRO QKD demo is a stage-1 implementation (trusted-node). Stage 3 — what you learned this chapter — is an active frontier; stage 6 is a 30-40 year aspiration.

India's satellite QKD — Bengaluru and Mt. Abu

In March 2022, ISRO (with the Raman Research Institute and the Physical Research Laboratory) demonstrated free-space quantum key distribution between Bengaluru's RRI campus and Mt. Abu's IIA telescope — a ground distance of about 300 km through open atmosphere [5]. Each key bit was encoded in the polarisation of single photons from a table-top BB84 transmitter.

This is a stage-1 network: two nodes, no entanglement, no quantum memory, trusted intermediate nodes. But the infrastructure it demonstrated — single-photon sources, single-photon detectors, timing synchronisation across hundreds of kilometres, free-space transmission — is the same infrastructure that would be needed for a stage-3 network with entanglement generation.

India's National Quantum Mission (₹6003 crore over 8 years, launched 2023) explicitly funds the progression from stage 1 to stage 3, with named goals including (i) metropolitan-scale fibre-based QKD networks in Bengaluru, Mumbai, Delhi, and Hyderabad, (ii) satellite-based entanglement distribution via a dedicated quantum communication satellite, and (iii) ground-to-ground entanglement swapping demonstrations. If the mission's targets are met, India will have a tiny-to-medium quantum network by roughly 2030.

Active Indian research groups

• TIFR Mumbai — superconducting qubits, photonic entanglement, quantum algorithms.
• Raman Research Institute, Bengaluru — atomic ensembles, single-photon sources, entanglement-distribution experiments.
• IIT Madras — trapped-ion quantum computing, atomic clocks, quantum optics.
• IIT Delhi — NV-centre diamond memories, quantum networking theory.
• IISc Bengaluru — photonic quantum computing, photon-atom interfaces.
• Physical Research Laboratory, Ahmedabad — satellite-QKD hardware.
• Indian Institute of Astrophysics, Bengaluru/Mt. Abu — free-space quantum optics, used as ground station for the 2022 ISRO demo.

These groups are linked through the National Quantum Mission's coordination centre and through collaborations with the international quantum-internet community (Quantum Internet Alliance in Europe, Q-NEXT in the US, China's Micius programme).

Where this leads next

• Quantum teleportation — the single-hop primitive this protocol composes.
• Entanglement swapping — the link-stitching primitive this protocol composes.
• Bell states — the maximally-entangled resource consumed on every hop.
• Classical control — the circuit notation for Pauli corrections conditioned on classical measurement bits.
• Quantum repeaters — how chains of tiny networks scale to continental distances with error correction.
• DLCZ protocol — the specific atomic-ensemble + photonic link architecture that turned tiny-network theory into near-term hardware.

References

1. Stephanie Wehner, David Elkouss, Ronald Hanson, Quantum internet: A vision for the road ahead (2018) — arXiv:1803.00608. The six-stage roadmap for quantum-internet development and the canonical reference on what a quantum network actually is.
2. Duan, Lukin, Cirac, Zoller, Long-distance quantum communication with atomic ensembles and linear optics (2001) — arXiv:quant-ph/0105105. The DLCZ quantum-repeater protocol.
3. Wikipedia, Quantum internet — concise overview of architectures, stages, and experimental status.
4. John Preskill, Lecture Notes on Quantum Computation, Ch. 4 — theory.caltech.edu/~preskill/ph229. Pedagogical treatment of teleportation, swapping, and the composition into networks.
5. ISRO, India's Satellite Based Quantum Communications (2022 demo, RRI/PRL/IIA) — ISRO press release. The Bengaluru–Mt. Abu free-space QKD experiment. See also RRI's programme pages.
6. Pirandola, Laurenza, Ottaviani, Banchi, Fundamental limits of repeaterless quantum communications (2017) — arXiv:1510.08863. The PLOB bound on repeaterless quantum-channel capacity, which quantifies exactly why repeaters are needed past 300 km.