🌊 The Inequality That Split the Dirac Sea: Bell’s Inequality

From the EPR Paradox to the Experimental Challenge to Local Realism

There are moments in the history of physics when a relatively simple relation creates a deeper fracture in our understanding of the world than an entire library of philosophical debate. Bell’s inequality is precisely such a case. Its importance does not lie in telling us that quantum mechanics is strange, nor in introducing the idea of quantum entanglement for the first time. Its real power lies in showing that a very natural, classically intuitive picture of reality cannot reproduce all the predictions of quantum mechanics. To understand what Bell actually demonstrated, we have to go one step back, to the EPR paradox and to the question that stood at the center of the debate for decades: is quantum mechanics a complete theory, or merely a statistical description of some deeper structure of reality?

EPR: Is Quantum Mechanics Complete?

In 1935, Einstein, Podolsky and Rosen attempted to show that quantum mechanics might not be a complete theory. Their idea was simple in essence: if two particles are created in a common quantum state and then separated, a measurement performed on one of them can allow us to predict the result of a measurement on the other. If we also assume that a distant measurement cannot instantaneously produce a physical influence on the second particle, it seems natural to conclude that the second particle must already have possessed some physical value before the measurement itself. In that case, quantum mechanics, which does not assign such a value in advance, might merely be an incomplete description of the system.

From this point of view, it is perfectly reasonable to suppose that additional variables exist which the theory does not explicitly track, but which determine the measurement outcomes in the background. Such “hidden variables” would not necessarily be mysterious or exotic. They could simply represent a deeper level of physical description, while the wave function would provide only a statistical account of our ignorance. This was an attractive idea, especially because it matched the classical intuition of reality: a physical system possesses definite properties regardless of whether we measure them.

The Quantum-Entangled State

The simplest example for discussing Bell’s inequality is the singlet state of two spin-1/2 particles:

|ψ⁻⟩ = 1/√2 (|↑↓⟩ − |↓↑⟩)

This state does not say that the first particle “has spin up” and the second “spin down” in some predetermined classical sense. It describes the two-particle system as a single quantum whole. If we measure the spin of both particles along the same axis, the results are perfectly anticorrelated: whenever one side gives +1, the other gives −1. When the measurement axes are different, however, the correlation is no longer trivial but depends on the angle between the chosen directions.

For the singlet state, quantum mechanics predicts:

E(a,b) = −cos θ

where θ is the angle between the measurement directions a and b. This angular dependence lies at the heart of the problem, because it produces correlations stronger than local classical models can allow.

How Is the Correlation Calculated?

Suppose that each detector can produce only two results:

A = ±1, B = ±1

For a given pair of settings a and b, there are four possible outcomes:

(++), (+−), (−+), (−−)

The correlation is defined as the average value of the product of the two results:

E(a,b) = ⟨A(a)B(b)⟩

If we assign probabilities to the possible outcomes, we obtain:

E(a,b) = P₊₊ + P₋₋ − P₊₋ − P₋₊

Equal outcomes contribute +1, while opposite outcomes contribute −1. For the singlet state, this combination of probabilities gives exactly E(a,b) = −cos θ. If the axes are identical, θ = 0 and E = −1, which represents perfect anticorrelation. If the axes are perpendicular, θ = 90°, then E = 0, so that the correlation in this sense disappears. The mathematics is simple; its consequences are not.

From Probability to Bell’s Test

Bell’s key insight was to move the discussion away from interpretation and toward experimentally testable correlations. Instead of asking “what is real?”, he asked a more precise question: if local hidden variables determine the outcomes of measurements, what correlations are possible at all?

Let λ represent all additional hidden information that quantum mechanics might fail to describe. In a local model, the result on side A depends on the local setting a and on λ, but not on the distant choice b. Similarly:

A = A(a,λ)

B = B(b,λ)

At the level of probabilities, local factorization takes the form:

P(A,B | a,b,λ) = P(A | a,λ) P(B | b,λ)

This factorization is the essence of the local model. If it holds, the correlations such a model can produce are not arbitrary. They have a strict upper bound.

The CHSH Form of Bell’s Inequality

The most practical form of Bell’s test today is the CHSH inequality. We choose two possible measurement directions on each side, a and a′ for the first observer and b and b′ for the second. We then form the combination:

S = E(a,b) + E(a,b′) + E(a′,b) − E(a′,b′)

For any local hidden-variable model of this kind, the following must hold:

|S| ≤ 2

This is not a quantum assumption. It is a consequence of the local structure of the model and of the fact that the outcomes take the values ±1. Bell’s decisive step was therefore to replace a philosophical argument about whether properties are “real” with a numerical bound that could be tested experimentally.

Quantum Mechanics Says Something Else

For the singlet state, we have E(a,b) = −cos θ. Let us choose the angles:

a = 0°, a′ = 90°, b = 45°, b′ = −45°

Then we obtain:

E(a,b) = −cos 45° = −√2/2

E(a,b′) = −cos 45° = −√2/2

E(a′,b) = −cos 45° = −√2/2

The angle between a′ and b′ is 135°, so:

E(a′,b′) = −cos 135° = +√2/2

Substituting these values into the CHSH expression gives:

S = −2√2

and therefore:

|S| = 2√2 ≈ 2.828 > 2

Quantum mechanics therefore predicts correlations that exceed the limit allowed by local hidden-variable models of this kind.

Tsirelson’s Bound

Quantum mechanics nevertheless does not allow arbitrarily large values of the CHSH parameter. The maximum quantum value is:

|S| ≤ 2√2

This is Tsirelson’s bound. It is larger than the local Bell bound of 2, but smaller than the absolute algebraic maximum of 4. Quantum correlations are therefore stronger than all local classical correlations, but they are not unlimited. Nature chooses neither the classical limit nor the maximum mathematically possible nonlocality, but a very specific quantum bound.

What Has Experiment Actually Shown?

From the earliest experiments to modern tests in which the major experimental loopholes have been closed, the result has remained the same: quantum correlations violate Bell inequalities. This means that nature cannot be described by a theory that simultaneously satisfies all the assumptions entering Bell’s local hidden-variable framework.

Precision is essential here. Violation of a Bell inequality does not mean that no hidden variables can exist, nor does it establish that quantum mechanics must be the final theory. It does not demonstrate superluminal communication, overthrow causality, or select a single interpretation of quantum mechanics. What it rules out is a broad class of local hidden-variable models of the classical type.

This is a narrower statement, but also a much deeper one.

What Remains of Hidden Variables?

If we want a theory containing hidden variables, they can no longer function simply as local parameters that predetermine the results independently of the distant experiment. Bohmian mechanics is the best-known example. It contains additional variables, but it is explicitly nonlocal. Bell’s inequality therefore does not eliminate Bohmian mechanics. What it eliminates is the idea that we can retain a simple form of classical realism, locality and predetermined outcomes while still reproducing all quantum correlations.

In this sense, Bell did not show that a deeper level of description cannot exist. He showed that, if such a level exists, it cannot look the way classical physics would naturally lead us to expect.

Another Assumption: Independence of the Measurement Settings

There is another important assumption in the standard derivation: the hidden variables λ must not already be correlated with the choice of settings a and b. This is usually called measurement independence or statistical independence. If this assumption is also abandoned, the door opens to superdeterminism.

Formally, such a possibility exists, but the price is substantial. The choice of experimental settings is then no longer statistically independent of the hidden variables of the system; instead, the entire experiment becomes part of a pre-correlated structure. This is not logically impossible, but it changes the very meaning of what we normally regard as an independent physical test.

Quantum Entanglement Is Not the Same as Bell Nonlocality

Another common mistake is to identify quantum entanglement with Bell nonlocality. They are not the same concept. Entanglement is a property of a quantum state, whereas Bell nonlocality is a property of experimentally accessible correlations that cannot be reproduced by local hidden-variable models.

quantum entanglement ≠ Bell nonlocality

Even more importantly:

Bell nonlocality ≠ superluminal signalling

The fact that distant outcomes exhibit correlations that no local model can reproduce does not mean that one experimenter can control a random result and use it to transmit a message to the other. The no-signalling principle remains intact.

What, Then, Did Nature Reject?

The fairest conclusion is that Bell experiments have not shown that quantum mechanics is the final word in physics, nor have they shown that no deeper level of description can exist. What they have rejected is a very intuitive possibility: that behind quantum statistics lies a local classical mechanism in which physical quantities simply possess predetermined values while the distant choice of measurement plays no role in the correlations.

This is an enormous result precisely because such a picture of the world was the one closest to classical intuition. After the experimental violation of Bell inequalities, it became clear that not all assumptions entering a local-realistic description can remain simultaneously intact: locality in Bell’s sense, classical realism, statistical independence of the measurement settings, or some other assumption embedded in the construction of the theory.

It is therefore not enough to say that “nature is nonlocal” and end the discussion there. The deeper question remains: which part of our classical picture of reality must we actually abandon?

The Inequality That Split the Dirac Sea

Before Bell, we could still hope that quantum mechanics was merely an incomplete statistical description of a world that remained classically local in its deepest structure. After Bell, and especially after the experimental violation of Bell inequalities, that possibility was no longer straightforwardly available.

Bell’s inequality is therefore more than a mathematical bound. It marks the point at which classical intuition confronted an experimental result that refused to accommodate it. It does not yet tell us what the world is in its deepest structure, but it tells us with remarkable precision what the world cannot be.

We have charted routes on the maps of the Dirac Sea while sailing close to the shores of classical physics. Bell’s inequality was one of those points on the map where the familiar coastline suddenly split apart and a new sea opened before us, a sea in which correlations are stronger than a local classical world allows. From that moment onward, it was no longer possible to sail as though nothing had happened. 🌊⚛️