Bell's theorem in one page Image: Peter Potrowl, Wikimedia Commons, CC BY-SA 3.0

Bell's theorem in one page

Take a source that emits pairs of particles, one to the left and one to the right. At each side an experimenter has a measuring device with a dial that can be set to one of three positions, 1, 2, or 3, and the device answers with a plus or a minus. Run the experiment many times with the dials set at random. Here is the fact that needs explaining: whenever the two dials happen to be set the same, the two answers are always opposite. Left says plus, right says minus, or the other way round, without a single exception.

The obvious explanation

The natural account is that each particle leaves the source carrying a list of instructions: what to answer for setting 1, for setting 2, and for setting 3. The two particles of a pair carry opposite lists, so that same settings always give opposite answers. The lists may vary from pair to pair in any way whatever, and the source may produce them by any mechanism, deterministic or random. This is what Einstein, Podolsky, and Rosen had argued in 1935 the particles must be doing. It is also what a pair of gloves does: if one box contains the left glove, the other contains the right, and nothing mysterious passes between them when the boxes are opened.

The counting argument

Each list has three entries, each plus or minus, so there are eight possible lists. In every one of them at least two of the three entries are equal, because three signs cannot all differ from each other. Now consider pairs where the dials are set to two different positions, say 1 on the left and 2 on the right. Since the right particle's list is the opposite of the left one's, the two answers agree exactly when the left list has different entries at positions 1 and 2.

Pick a list at random from whatever mixture the source produces. Of the three pairs of positions, (1,2), (2,3), and (1,3), at least one pair has equal entries. So across all the lists, the fraction of pairs of positions with equal entries is at least one third, and the fraction with different entries is at most two thirds. Translating back to the experiment: when the two dials are set differently, the two answers agree at most two thirds of the time, whichever lists the source hands out. Equivalently, they are opposite at least one third of the time. That is a Bell inequality, in the simple form found by Eugene Wigner. It follows from counting, with no physics in it.

What quantum mechanics predicts

Now take the particles to be photons from a particular entangled source, and the three settings to be three directions of a polarizing filter, 120 degrees apart. Quantum mechanics predicts that for equal settings the answers are always opposite, as required, and that for different settings they are opposite only one quarter of the time. One quarter is less than one third. The quantum prediction violates the inequality.

If the prediction is right, then no lists exist. The particles do not carry their answers with them. Whatever determines the outcome on the left is not fixed before the measurement and is not independent of what is done on the right, even though the right may be a kilometre away and the settings may be chosen after the particles have left the source. John Bell published this in a short paper in 1964, in a new journal that folded after a few issues, and for years almost nobody read it.

The experiments

John Clauser and Stuart Freedman did the first test in Berkeley in 1972, with photons from calcium atoms, and found the quantum value. The experiments of Alain Aspect and his students in Orsay in 1981 and 1982 were the ones that convinced the field. In the last of them the setting on each side was switched every ten nanoseconds, while the photons were in flight, so that no signal travelling at the speed of light could carry the choice from one side to the other in time. The inequality was violated by many standard deviations.

Two loopholes remained. Detectors caught only a small fraction of the photons, and one could imagine, without believing it, that the detected ones were an unrepresentative sample. And in experiments with efficient detectors the two sides were too close together to exclude a signal. In 2015 three groups closed both loopholes in a single experiment: a team in Delft using electron spins in two diamonds 1.3 kilometres apart, a team in Vienna led by Anton Zeilinger using photons and detectors with efficiencies above 90 percent, and a team at NIST in Boulder doing the same. All three found the violation. Aspect, Clauser, and Zeilinger shared the Nobel Prize in Physics 2022.

What is ruled out and what is not

The theorem rules out every theory in which the outcome of each measurement is fixed by properties the particle carries from the source and the setting on the other side has no influence. The combination of those two assumptions is called local realism, and it is what Einstein wanted. One of the two must go. Quantum mechanics as usually understood gives up the first: there are no instruction lists, and the outcome is not determined until the measurement. Theories such as David Bohm's keep definite outcomes and give up the second, allowing the distant setting to matter instantly. No experiment can tell these readings apart, because they make the same predictions.

The theorem does not allow signalling. The answers on each side, taken alone, are a random sequence of pluses and minuses whatever the other side does; the correlation only appears when the two sequences are compared, and comparing them requires an ordinary message. Entanglement cannot be used to send information faster than light, and the next article in this series explains why.

Why it matters

Before Bell, the question whether the quantum description was complete was taken to be philosophy. Bell turned it into a number that an experiment could measure, and the experiments came down on the side nobody had expected in 1935. The violation is now a working resource: quantum cryptography protocols use a Bell test to certify that an eavesdropper cannot have copied the key, because copying would have required the instruction lists that do not exist.

In the Quantum Mechanics Masterclass the counting argument above is presented in Unit 8, with the quantum calculation of the one quarter and an account of the Orsay and 2015 experiments. The general form of the inequality used in the laboratory, due to Clauser, Horne, Shimony, and Holt, with its classical bound of 2 and quantum maximum of \(2\sqrt{2}\), is derived in Unit 14, together with the cryptography that rests on it.

Learn quantum mechanics properly

The QM Masterclass: weekly units, quizzes, a final exam, and a certificate.
About the program