Quantum Teleportation, Precisely

The word does most of the damage

Quantum teleportation is a fully specified protocol with a fixed price: one shared entangled pair, one joint measurement, two classical bits, and one corrective operation on the receiving qubit. Charles Bennett and his co-authors published it in 1993 in Physical Review Letters, and the name they chose has been causing confusion ever since. Nothing is transported. No particle moves between the two parties. Nothing arrives before the classical message that completes the procedure.

The imprecision is not harmless. Vendor material and briefing decks routinely describe teleportation as instantaneous transfer, as a secure channel in its own right, or as a way to send data without any signal crossing the intervening distance. Each of those descriptions is wrong in a specific way, and each one leads to a wrong architectural conclusion. What follows is the protocol, the cost accounting, and the security properties it does and does not deliver.

The resource is a pair, and the pair is consumable

A qubit is the quantum equivalent of a bit: a two-state system, such as the polarisation of a photon or the spin of an electron, that can also hold combinations of its two states. Two qubits are entangled when the pair has a joint state that cannot be written as a description of one qubit plus a description of the other. Measure one and you learn something about the other, however far apart they are, and the correlation is stronger than any arrangement of pre-agreed classical values can produce.

The four maximally entangled two-qubit states are called Bell states, after John Bell, and they form the working basis for teleportation:

|Φ⁺⟩ = (|00⟩ + |11⟩)/√2

|Φ⁻⟩ = (|00⟩ − |11⟩)/√2

|Ψ⁺⟩ = (|01⟩ + |10⟩)/√2

|Ψ⁻⟩ = (|01⟩ − |10⟩)/√2

Read |Φ⁺⟩ this way: measure both qubits in the computational (Z) basis and the two results always agree, but neither result is determined in advance. Measured in that same basis, |Ψ⁺⟩ always gives opposite results. Change the basis and the pattern of agreement changes with it, and Bell tests exploit exactly that. These four states are the only entangled resource the protocol needs, and one of them gets used up every time the protocol runs. That last point is the one we find most often missing from network diagrams. An entangled pair is not a link. It’s a consumable, closer to a one-time pad than to a fibre.

The protocol, in four steps

Take the standard cast. Alice holds a qubit Q in some unknown state |ψ⟩ that she wants Bob to end up holding. She does not know what that state is, and she cannot find out by measuring it, because measurement would collapse it to a single outcome and destroy everything else it encoded.

Step one: share the pair. Alice and Bob each take one qubit of an entangled pair, prepared in |Φ⁺⟩. Call Alice’s half A and Bob’s half B. This has to happen before any teleportation, and in a real network it’s the slow, lossy, expensive part.

Step two: the Bell state measurement. Alice measures Q and A jointly, in the Bell basis. This is a Bell state measurement, or BSM: rather than reading each qubit separately, she asks a single question with four possible answers, namely which of the four Bell states her two qubits are in. The answer is 2 bits. Her original qubit Q is gone as an information carrier, and so is her half of the entangled pair. That destruction isn’t an unfortunate side effect; it’s required. The no-cloning theorem, proved independently by Wootters and Zurek and by Dieks in 1982, says an unknown quantum state cannot be copied. A protocol that left Alice holding |ψ⟩ while Bob also held it would break that theorem, so the protocol doesn’t.

Step three: send the two bits. Alice transmits her 2-bit result to Bob over an ordinary classical channel. Email would work. The bits carry no quantum information and, on their own, say nothing about |ψ⟩.

Step four: correct. Bob’s qubit is already in a state related to |ψ⟩, but he doesn’t know which of four possible relations applies until the bits arrive. Alice’s outcome tells him which single-qubit operation to apply: nothing for |Φ⁺⟩, a bit flip (X) for |Ψ⁺⟩, a phase flip (Z) for |Φ⁻⟩, both for |Ψ⁻⟩. Once he applies it, his qubit is in the state |ψ⟩ that Alice started with.

Before step four, Bob holds something useless. Measuring it without the correction gives him random results uncorrelated with |ψ⟩. This is why the causal accounting works: the classical message travels at or below the speed of light, and nothing usable exists at Bob’s end until it arrives.

Five corrections worth making out loud

Matter stays put. The physical qubit at Bob’s end is the one that was always at Bob’s end. What moves is the description, the quantum state, and even that moves in the sense that it ceases to exist at one location and comes into existence at another.

The correlation carries no message. The entangled pair produces correlations without delay, but correlations aren’t messages. Two classical bits have to cross the distance, and they obey the same limit everything else does.

Every teleported qubit consumes a pair. One teleported qubit costs one entangled pair, plus 2 bits of classical bandwidth. Teleporting a hundred qubits means distributing a hundred pairs. Any capacity planning for a quantum network starts with the entanglement generation rate, not with the fibre bandwidth.

Fidelity is the measure, and 2/3 is the threshold. Fidelity measures how close the state Bob ends up with is to the state Alice started with, on a scale where 1 is exact. Teleportation fidelity is bounded by the quality of the entangled pair, and imperfect pairs give imperfect output. Massar and Popescu showed that any strategy based on measuring an unknown qubit and preparing a replacement can achieve an average fidelity of at most 2/3. Exceeding 2/3 is what distinguishes genuine teleportation from a classical imitation, and when a result is reported, that is the number to look for.

It is not a channel by itself. Teleportation moves one qubit, given a pair. It does not create the pair, authenticate the parties, or detect an adversary. Those are separate protocols with separate assumptions.

Teleportation inside a network

If teleportation costs more than sending the photon directly, why build networks around it? Because of where the cost falls.

Entanglement generation can be heralded: the network can be told that an attempt succeeded, without anybody measuring the entangled state and destroying it. So a pair of nodes can try repeatedly across a lossy link, discard the failures, and keep going until they hold a good pair. Only then does the payload qubit get committed, and it never enters the fibre at all. Compare that with sending the payload directly, where a lost photon is a lost state and there’s no copy to retry with. Teleportation doesn’t eliminate loss. It relocates the loss to a stage where retrying is cheap and failure is visible.

The second reason is reach. Entanglement swapping extends entanglement across nodes that never interact. Suppose Alice shares a pair with Bob, and Bob separately shares a pair with Charlie. Bob performs a Bell state measurement on his two qubits, one from each pair, and announces the result. Alice and Charlie now hold an entangled pair, despite never having exchanged anything. Bob is running the teleportation protocol on entanglement itself.

Chain those swaps and you get a quantum repeater: a node that extends entanglement segment by segment rather than amplifying a signal, which no-cloning forbids anyway. Two properties make repeaters interesting to anyone who has designed a trusted relay. The intermediate nodes only ever perform physical operations on their own qubits, and they never hold the teleported state, because the state doesn’t pass through them.

The demonstrations are real but early. Bouwmeester and colleagues in Innsbruck reported the first experimental teleportation in 1997, and the Rome group followed in 1998. Ren and co-authors reported ground-to-satellite teleportation over roughly 1,400 km using the Micius satellite in 2017. In 2022, Hermans and colleagues at QuTech in Delft teleported a qubit between two nodes with no direct link between them, using an entanglement swap at the node in the middle. That result runs the full pattern: distribute, swap, teleport, correct.

What this does and does not buy a security team

Quantum key distribution (QKD) is a family of protocols that produce shared random bits between two parties, with secrecy resting on the fact that measuring a quantum system disturbs it. An eavesdropper who intercepts and measures leaves statistical evidence in the results, so the parties can detect the interference and discard the compromised material. Artur Ekert’s 1991 protocol does this with entangled pairs, checking the correlations directly.

Teleportation is not a QKD protocol, and it does not generate keys. What it shares with entanglement-based QKD is the underlying resource, and that shared resource is where the practical benefit sits. Deployed fibre QKD links are distance-limited, so operators chain them using trusted nodes: relays that receive key material, decrypt it, and re-encrypt it for the next hop. Every trusted node is a place where plaintext key exists and where an operator or an intruder can read it. Repeater networks built on swapping and teleportation are the intended replacement, because a node that only performs Bell measurements never holds the material it’s helping to relay.

Three qualifications keep this honest.

The classical channel is an attack surface. Spoof Alice’s 2 bits and Bob applies the wrong correction, ending up with a corrupted state. That’s denial of service and data corruption rather than disclosure, but it still requires the classical channel to be authenticated, and that authentication is ordinary cryptography with ordinary key management. In deployment terms, a quantum link inherits a classical dependency.

Hardware is where real QKD systems have been broken, not mathematics. Lydersen and colleagues showed in 2010 that bright-light blinding could take control of the single-photon detectors in QKD hardware and extract a full key without raising the error rate. Entanglement in a system does not by itself close that class of attack. The constructions designed for it are measurement-device-independent QKD, which removes trust in the detectors, and device-independent QKD, which certifies security from an observed Bell inequality violation and demands very high detection efficiency. Those are specific protocols with specific costs, and buying an entangled photon source doesn’t grant their properties.

And teleportation protects data in transit only. Endpoints, key storage, access control, and everything at rest remain exactly as secure as the classical practices around them. A network that cannot be eavesdropped in the middle is still reachable through a compromised node at either end.

What decides the timeline

Four engineering constraints govern how far this scales, and all four are measurable today.

Coherence time. Entangled states degrade through decoherence, the loss of the delicate phase relationships that carry quantum information, caused by interaction with the environment. A repeater must hold one half of a pair in quantum memory while the next segment is established. If the memory forgets faster than the next link succeeds, the chain never completes.

Entanglement rate. One pair per teleported qubit sets a hard ceiling on throughput. Long-distance experiments have run at rates measured in pairs per minute rather than per second. Multiplexing across frequency channels, parallel sources, and memory buffers are the routes being tried.

Fidelity per hop. Errors compound along a chain of swaps. Entanglement purification consumes several low-quality pairs to produce one better pair, which buys fidelity at the cost of rate, and the trade has to close for a chain of useful length.

Field operation. Most of the results cited above came from laboratories with continuous expert attention. Deployed repeater stations have to hold alignment through temperature swings and vibration, and interface with fibre plant that was installed for classical traffic.

None of these is a physics obstacle. They are engineering problems with active programmes behind them, and progress on each is public and quantifiable.

Getting the vocabulary right is part of the job

Anyone specifying, procuring, or auditing quantum network infrastructure will be reading claims written by people who are either imprecise or hoping you are. The defence is knowing that a teleportation claim needs a fidelity figure and its benchmark, that a link claim needs an entanglement rate, that a repeater claim needs a memory coherence time, and that a security claim needs to say which protocol, with which trust assumptions, is doing the work.

Quantum Academy’s quantum networking material builds from these definitions upward: entanglement as a consumable resource, teleportation as the transfer protocol, swapping and repeaters as the reach mechanism, and QKD as one application among several that sit on top. You can see the current programs at quantumacademy.com/.

For the longer technical treatment of teleportation in network architecture, including the source material this piece was built from, see PostQuantum.com on quantum teleportation.