Image: Gavin Morley, Wikimedia Commons, CC BY-SA 4.0Entanglement cannot send messages
Two photons leave a source in an entangled state, one towards Alice in Vienna and one towards Bob in Beijing. When Alice measures the polarization of hers, quantum mechanics says that Bob's photon acquires the opposite polarization at that moment, with no delay for the distance. Readers meet this and ask the obvious question: if Alice can change Bob's photon instantly, can she send him a message that way? The answer is no, and it is no for a reason that does not depend on the details of any experiment.
What Bob sees
Bob measures his photon along some direction and records the result, plus or minus. He does this for a long sequence of photons. His record is a string of pluses and minuses, each with probability one half, in a random order. This is true if Alice has measured her photons along the same direction as Bob. It is true if she measured along a different direction. It is true if she has not measured at all, or has lost her photons, or has gone to lunch. Nothing Alice does, including the choice of whether or not to measure, changes the statistics of Bob's string by the smallest amount.
The correlation that makes entanglement remarkable exists only between the two strings. When Alice and Bob later compare their records, they find that wherever they chose the same direction, their results were opposite. To make that comparison, one of them has to send the other a copy of their record, by telephone, by radio, or by post, and that message travels no faster than light. Until it arrives, Bob's string is random noise and contains no information about anything Alice did.
Why this must be so
The mathematics of quantum mechanics guarantees the result in general, and the guarantee is called the no-signalling theorem. The predictions for any measurement Bob can make on his particle alone are computed from a reduced description of his particle, obtained by averaging over everything that could happen to Alice's. A measurement on Alice's side changes which outcome occurs over there, but the average over her possible outcomes, weighted by their probabilities, is exactly the description Bob had before. His particle's statistics are the same whether Alice's measurement has happened or not. The collapse that Alice's measurement produces on Bob's side is real in the sense that it fixes which correlated outcome Bob will get, and invisible in the sense that Bob has no way to detect that it occurred.
There is no loophole in choosing cleverer measurements. Whatever Bob does to his particle, with any instrument, the reduced description is all that enters his predictions, and Alice cannot touch it. Relativity and quantum mechanics are consistent on this point, and physicists who build relativistic quantum field theory check it as a matter of routine.
What teleportation moves
The name quantum teleportation suggests otherwise, and it was chosen by its inventors in 1993 with some mischief. The protocol, described by Charles Bennett and five colleagues in a paper in Physical Review Letters, lets Alice transfer the unknown quantum state of a particle to a particle in Bob's possession, using a shared entangled pair. She makes a joint measurement on her particle and her half of the pair, which has four possible outcomes. Bob's half of the pair is now in a state related to the original by one of four simple transformations, depending on her outcome. To finish the job, Bob needs to know which of the four it was, and that is two bits of ordinary information that Alice has to send him in the ordinary way.
Until the two bits arrive, Bob's particle is in a random one of the four states, which averaged together carry no trace of the original. The quantum state is transmitted, and it is transmitted exactly, but not one bit faster than the classical message. The original state is destroyed in Alice's measurement, which is why the process is a transfer and not a copy, in keeping with another theorem that says an unknown quantum state cannot be copied at all.
The first experimental teleportation was done in Innsbruck in 1997 with photons across a laboratory bench. In 2017 a Chinese team led by Jian-Wei Pan teleported photon states from a ground station in Tibet to the satellite Micius, 1,400 kilometres up, and the same satellite distributed entangled pairs to two ground stations 1,203 kilometres apart and verified a Bell inequality between them. In both cases the classical channel was a radio link, and in both cases the result could not have been used without it.
What faster-than-light would break
Relativity does not forbid things from happening at a distance. It forbids signals, because a signal that travels faster than light in one frame of reference arrives before it was sent in another, and from that one can build a message to one's own past. The correlations of entanglement travel faster than light in the loose sense that no slower signal could have produced them, which is what Bell's theorem shows. They are not signals, because neither end can control what the other end sees. Physicists sometimes say that entanglement is nonlocal but not signalling, and the distinction is exactly the one between seeing a correlation after the fact and causing a change that the other side can detect.
What entanglement is good for
If it cannot carry a message, what is it for? The correlations are a resource in their own right. Two parties who share entangled pairs can generate a secret key that no eavesdropper can have copied, because an eavesdropper would have disturbed the correlation in a way the Bell test reveals. Distant clocks and telescopes can be synchronized and combined more precisely than any classical link allows. Teleportation is the way quantum information will move between the modules of a large quantum computer, and between the nodes of a quantum network. None of this is communication faster than light, and all of it is impossible without entanglement.
In the Quantum Mechanics Masterclass the no-signalling argument is given in Unit 8, where the entangled state of two photons is written out and the statistics on each side are computed. The teleportation protocol, with the four outcomes and the two bits, and the key distribution schemes that use entangled pairs are worked through in Unit 14, which also presents the Micius experiments. The previous article covers Bell's theorem itself.


