51
Part One · Quantum mechanics

Bell's theorem and the experiments

Here is a game for two players, Alice and Bob, who may plan together beforehand but cannot talk once it starts. Each receives a random bit, xx for Alice and yy for Bob, and each answers with a bit, aa and bb. They win if a⊕b=x∧ya \oplus b = x \wedge y: their answers should differ only when both received a 1. No classical plan, however clever, wins more than 75% of the time. Players who share entangled particles win 85.4%. That gap, measured in laboratories since 1972, is the most profound experimental result in physics.

On the plate: the four possible question pairs, and which of them your plan wins.

Design your plan

Without communication, the best you can do is decide each answer in advance. Alice answers a0a_0 when x=0x = 0: ; and a1a_1 when x=1x = 1: . Bob answers b0b_0: ; and b1b_1: .

Try to win all four rows. You can’t. The four win conditions, added together, demand

(a0⊕b0)⊕(a0⊕b1)⊕ (a1⊕b0)⊕(a1⊕b1)=1\begin{gathered}(a_0\oplus b_0)\oplus(a_0\oplus b_1)\\ \oplus\,(a_1\oplus b_0)\oplus(a_1\oplus b_1) = 1\end{gathered}

but the left side is always 0, since every letter appears twice. So at least one row is always lost, and the best is three in four. Random strategies are mixtures of these sixteen plans and cannot do better. Established

Bell’s inequality

John Bell’s 1964 insight, sharpened by Clauser, Horne, Shimony and Holt in 1969, turns the game into a number. Write outcomes as ±1\pm1 and let E(x,y)E(x, y) be the average product of Alice’s and Bob’s outcomes. Define

S=E(0,0)+E(0,1)+E(1,0)−E(1,1)S = E(0,0) + E(0,1) + E(1,0) - E(1,1)

If each particle carries answers fixed at the source, the EPR picture of Chapter 50, then for each pair S=a0(b0+b1)+a1(b0−b1)S = a_0(b_0 + b_1) + a_1(b_0 - b_1), and one bracket is 0 while the other is ±2\pm2. Averaging, ∣S∣≤2|S| \le 2. The winning probability is 12+S/8\tfrac12 + S/8, so S≤2S \le 2 is the 75% limit. Any theory where results are determined locally, by anything at all, obeys it.

What entanglement does

Share a singlet pair. Alice measures along angle 00 or 2θ2\theta, Bob along θ\theta or 3θ3\theta, and quantum mechanics predicts E=−cos⁡(angle between)E = -\cos(\text{angle between}), so ∣S∣=3cos⁡θ−cos⁡3θ|S| = 3\cos\theta - \cos3\theta. Set θ=\theta = 30.0°.

At 45°, S=22≈2.83S = 2\sqrt2 \approx 2.83, and the win rate is cos⁡2(π/8)=85.4%\cos^2(\pi/8) = 85.4\%. No signal passes; each player’s own results are still a fair coin (Chapter 50). Yet together they beat every possible pre-arranged plan.

DeeperWith A,A′,B,B′A, A', B, B' having eigenvalues ±1\pm1, the operator S=AB+AB′+A′B−A′B′\mathcal S = AB + AB' + A'B - A'B' obeys S2=4−[A,A′][B,B′]\mathcal S^2 = 4 - [A, A'][B, B'], whose norm is at most 8. So ∣S∣≤22|S| \le 2\sqrt2: Tsirelson’s bound (1980). Hypothetical “PR boxes” could reach 4 without signalling; why nature stops at 222\sqrt2 is an open question.

The experiments

Stuart Freedman and John Clauser made the first test in 1972; Alain Aspect’s team in 1982 switched the settings while the photons were in flight. Every test since has violated the inequality. But each left a loophole: detectors that missed photons, or settings chosen too slowly to rule out a signal. In 2015 three groups closed both at once. Ronald Hanson’s group in Delft entangled electron spins in diamonds 1.3 km apart and measured S=2.42±0.20S = 2.42 \pm 0.20 over 245 trials; groups in Vienna and at NIST did the same with photons, with far more trials.

Play rounds with an entangled pair at the best angles. Choose 250 rounds and . With few rounds the error bar is wide, as in Delft; with many, SS sits clearly above 2. Aspect, Clauser and Anton Zeilinger shared the 2022 Nobel Prize. Established

What it rules out, and what it doesn’t

Bell’s theorem rules out every theory in which outcomes are fixed by local variables and the settings are independent of those variables. It does not rule out:

Non-local hidden variables. Bohmian mechanics (Part Two) reproduces every quantum prediction with definite particle positions, at the price of instantaneous influences that cannot be used to signal. Theories without single outcomes. In the many-worlds interpretation, every result happens, and the inequality’s premises do not apply. Superdeterminism, in which the settings and the particles are correlated from the start, so the choice of measurement is not free. Most physicists find this unattractive, a few defend it, and it cannot easily be tested. Debated

What no reading escapes: the world is not locally causal in the way EPR assumed. One of their two premises is wrong.

What you now have

The CHSH game and the 75% classical limit, the inequality ∣S∣≤2|S| \le 2, the quantum value 222\sqrt2 and Tsirelson’s bound, the experiments from 1972 to the loophole-free tests of 2015, and a map of what the theorem does and does not exclude.

· · ·

Three questions to carry with you.

Why can no fixed plan win all four rounds?tap
Combining the four win conditions requires a sum in which every answer appears twice to equal 1, which is impossible.
Does Bell violation allow faster-than-light signalling?tap
No. Each side’s results stay random; only the correlations, compared later, exceed the classical limit.
Which interpretations survive Bell’s theorem?tap
Non-local ones (Bohm), ones without single outcomes (many-worlds), and superdeterministic ones; local realism does not.

And one to work out. The quantum win probability is 12+S/8\tfrac12 + S/8 with S=22S = 2\sqrt2. What is it, to four decimal places?

P =

Sources: J. S. Bell, Physics 1, 195 (1964). J. F. Clauser, M. A. Horne, A. Shimony and R. A. Holt, Phys. Rev. Lett. 23, 880 (1969). S. J. Freedman and J. F. Clauser, Phys. Rev. Lett. 28, 938 (1972). A. Aspect, J. Dalibard and G. Roger, Phys. Rev. Lett. 49, 1804 (1982). B. S. Cirel’son, Lett. Math. Phys. 4, 93 (1980). B. Hensen et al., Nature 526, 682 (2015); M. Giustina et al., Phys. Rev. Lett. 115, 250401 (2015); L. K. Shalm et al., Phys. Rev. Lett. 115, 250402 (2015). Checked by code: best deterministic plan wins 3 of 4; S(θ) = 3cos θ − cos 3θ peaks at 2√2 when θ = 45°; win probability 0.8536.

Next · Chapter 52
Decoherence and the measurement problem

You never see your coffee cup in two places at once. Quantum mechanics doesn't stop working. The air keeps measuring the cup.

Turn the page