50
Part One · Quantum mechanics

Entanglement and the EPR paradox

In May 1935 Einstein, Boris Podolsky and Nathan Rosen published four pages arguing that quantum mechanics is incomplete. Their logic was careful and, step by step, sound. It rested on two assumptions that almost everyone would accept without thinking. Thirty years later John Bell showed that the two together make a prediction that can be tested, and the experiments say one of them is false. This chapter reconstructs the argument exactly, so that you can see which.

On the plate: a source sends pairs of particles to two distant detectors. Each detector flashes at random; together, they agree.

The argument

EPR used positions and momenta; David Bohm’s 1951 version uses spins, and it is cleaner. A source emits two spin-½ particles in the singlet state of Chapter 41, one to Alice and one to Bob, far apart. Whatever axis Alice measures along, Bob will find the opposite result along that axis, with certainty.

Let Alice choose: or . Either way she can predict Bob’s result for that axis without touching his particle. EPR then reasoned: (1) locality, nothing Alice does can change Bob’s particle instantly; (2) reality, if you can predict a value with certainty without disturbing a system, there is an element of reality corresponding to it. Since Alice could have chosen either axis, Bob’s particle must have had definite values for both SzS_z and SxS_x all along. Quantum mechanics says no state has both. So, they concluded, quantum mechanics is incomplete. Established, as a statement of the argument.

Bohr’s reply and Schrödinger’s word

Niels Bohr answered within months, in a paper with the same title. He accepted that Alice does not disturb Bob’s particle mechanically, but argued that her choice determines which kinds of prediction about it can meaningfully be made: the experimental arrangement is part of what defines a physical quantity. Many found this obscure, and the debate stalled.

That same year, Erwin Schrödinger gave the phenomenon its name. He called it Verschränkung, entanglement, and called it “not one but rather the characteristic trait of quantum mechanics”. He also noticed that Alice, by choosing her measurement, steers Bob’s particle into one set of states or another, a feature now called steering.

Telling entangled from not

Any pure state of two systems can be rewritten, by choosing bases well, as a single sum, the Schmidt decomposition:

∣Ψ⟩=∑iλi  ∣ai⟩∣bi⟩,λi≥0,  ∑λi=1|\Psi\rangle = \sum_i \sqrt{\lambda_i}\;|a_i\rangle|b_i\rangle, \qquad \lambda_i \ge 0,\ \ \sum\lambda_i = 1

If only one λi\lambda_i is nonzero, the state is a product and there is no entanglement; otherwise there is. The λi\lambda_i are the eigenvalues of each half’s density matrix from Chapter 49. For cos⁡θ∣00⟩+sin⁡θ∣11⟩\cos\theta|00\rangle + \sin\theta|11\rangle with θ=\theta = 30°, the plate shows the two Schmidt weights.

For mixed states the question is harder. A state is separable if it can be made by preparing product states at random; otherwise it is entangled. Mix a singlet with noise, ρ=p ∣Ψ−⟩⟨Ψ−∣+(1−p) 1/4\rho = p\,|\Psi^-\rangle\langle\Psi^-| + (1-p)\,\mathbb 1/4: it is entangled only for p>1/3p > 1/3, a result found by checking whether a partial transpose of ρ\rho has a negative eigenvalue (Peres, Horodecki, 1996).

Trying to send a message

If Alice’s choice steers Bob’s particle, can she use it to signal? Try. Alice wants to send “0” by measuring along z and “1” by measuring along x. Pick her setting, or , and watch Bob’s results pile up. He always measures along z.

His counts come out fifty-fifty either way. Bob’s particle alone is in the state 1/2\mathbb 1/2 whatever Alice does: her z measurement leaves him an equal mixture of up and down, her x measurement an equal mixture of left and right, and Chapter 49 showed those are the same density matrix. Only when Alice phones Bob her results, at light speed or slower, do the correlations appear. This is the no-signalling theorem. Entanglement creates correlations, never messages. Established

Which assumption?

EPR’s two premises together imply that every particle carries predetermined answers to all questions, fixed locally at the source. That is a local hidden-variable theory. It reproduces perfect anti-correlation along any single shared axis easily. The question Bell asked was whether it can reproduce the correlations at different axes. It cannot, and the next chapter derives the inequality and the experiments that settled it.

What you now have

The EPR argument and its two premises, Bohr’s reply, Schrödinger’s entanglement and steering, the Schmidt decomposition, separability and the noisy singlet, and the no-signalling theorem.

· · ·

Three questions to carry with you.

What two assumptions did EPR combine?tap
Locality (no instantaneous influence) and a criterion of reality (certain prediction without disturbance implies a real value).
Why can’t Alice signal to Bob?tap
Bob’s reduced state is 𝟙/2 whatever Alice measures; different ensembles, same density matrix.
When is a pure two-part state entangled?tap
When its Schmidt decomposition has more than one nonzero term.

And one to work out. For cos⁡30∘∣00⟩+sin⁡30∘∣11⟩\cos 30^\circ|00\rangle + \sin 30^\circ|11\rangle, what is the larger Schmidt weight λ1\lambda_1?

λ₁ =

Sources: A. Einstein, B. Podolsky and N. Rosen, Phys. Rev. 47, 777 (1935). N. Bohr, Phys. Rev. 48, 696 (1935). E. Schrödinger, Proc. Camb. Phil. Soc. 31, 555 (1935). D. Bohm, Quantum Theory (1951). A. Peres, Phys. Rev. Lett. 77, 1413 (1996); M., P. and R. Horodecki, Phys. Lett. A 223, 1 (1996). Checked by code: Schmidt weights cos²θ, sin²θ; noisy singlet partial-transpose eigenvalue (1 − 3p)/4, negative for p > 1/3; Bob’s marginal 𝟙/2 for either of Alice’s settings.

Next · Chapter 51
Bell's theorem and the experiments

In a two-player game, no classical strategy wins more than 75% of the time. Entangled players win 85.4% ✓.

Turn the page