Bell Inequality Experiments

Bell inequality experiments ask whether the correlations observed between separated measurement outcomes can be reproduced by local hidden-variable models satisfying Bell-type assumptions. They are not just demonstrations that two systems are entangled. They are tests of a quantitative constraint on correlations.

The historical arc runs from the EPR argument, through Bohm’s spin reformulation and Bell’s theorem, to laboratory measurements of correlation functions. The formal theorem and the compact CHSH derivation live in Bell Theorem and CHSH Inequality. This page explains how experiments turn those inequalities into measured numbers.

Experimental Question

A Bell experiment prepares many pairs of systems, sends one member of each pair to Alice and the other to Bob, lets each side choose a measurement setting, and records two outcomes. The central question is:

Do the observed correlations obey every Bell inequality implied by the tested local hidden-variable assumptions?

In the CHSH version, Alice chooses between two settings $a$ and $a'$, Bob chooses between two settings $b$ and $b'$, and the two outcomes are encoded as $\pm1$. The experiment estimates four correlations and forms

$$S = E(a,b) + E(a,b') + E(a',b) - E(a',b').$$

Local hidden-variable models satisfying the CHSH assumptions obey

$$\lvert S\rvert\le2.$$

Quantum mechanics predicts larger values for suitable entangled states and measurement choices, up to the Tsirelson bound

$$\lvert S\rvert\le2\sqrt2.$$

The experiment is therefore not asking whether quantum theory is mathematically consistent. It is asking which class of possible explanations is compatible with observed correlations.

Experimental Requirements

A serious Bell test needs more than a source of correlated particles. It needs a controlled version of the full Bell scenario:

• a preparation expected to produce entangled pairs;
• spatially separated measurement stations;
• independently chosen settings on the two sides;
• outcomes assigned to every relevant trial;
• enough events to estimate correlations with statistical confidence;
• timing and analysis rules fixed so that the comparison is not adjusted after seeing the data.

Different experiments realize these requirements differently. Early optical experiments used atomic cascades or parametric down-conversion photon pairs. Later experiments used trapped ions, superconducting circuits, nitrogen-vacancy centers, atoms, and high-efficiency photon detection. The physical platform matters, but the logical structure is the same: estimate correlations for different setting pairs and compare them with a Bell inequality.

The clean theoretical inequality assumes ideal trials. Real experiments must specify how a trial begins, how pairs are identified, what counts as a detection, how settings are generated, and how uncertain counts become a confidence statement. These implementation details are not bookkeeping; they determine which loopholes are being addressed.

Polarization Correlations

Many historically important Bell tests used photon polarization. A polarizer or polarization analyzer has two output classes, which can be encoded as $+1$ and $-1$. For a setting pair $x,y$, the measured correlation can be estimated from same-outcome and different-outcome counts:

$$E(x,y) = \frac{ N_{\mathrm{same}}(x,y)-N_{\mathrm{diff}}(x,y) }{ N_{\mathrm{same}}(x,y)+N_{\mathrm{diff}}(x,y) }.$$

For ideal polarization-entangled photon pairs, the correlation depends on the relative analyzer angle. Up to a sign convention set by the chosen state and outcome labels, a common dependence is

$$E(\alpha,\beta) = -\cos\!\left(2(\alpha-\beta)\right).$$

The factor of $2$ reflects polarization’s angular periodicity: a linear polarization direction at angle $\alpha$ is physically the same axis as one at $\alpha+\pi$. Suitable choices of analyzer angles make the CHSH combination exceed $2$ in magnitude.

This is why Bell tests are more structured than ordinary coincidence measurements. The same source must be tested at several setting pairs, and the correlations must be combined with the sign pattern in the chosen inequality.

Early Tests

The CHSH paper made Bell’s idea experimentally concrete by formulating an inequality for real two-outcome correlation measurements. The first generation of tests then tried to measure the required correlations with optical systems.

Freedman and Clauser’s 1972 calcium atomic-cascade experiment was a landmark. It measured polarization correlations of photon pairs and reported a violation of a Bell-type inequality. The result was historically important because it moved Bell’s theorem from a theoretical challenge into laboratory practice.

Early experiments also showed why Bell tests are technically delicate. Detectors were inefficient, not every emitted pair was registered, and the experimenters had to rely on assumptions about whether the detected subset fairly represented the emitted ensemble. Such assumptions are not cosmetic. A local hidden-variable model can exploit selective detection if the experiment only samples favorable events.

The right historical lesson is therefore twofold. Early tests strongly supported the quantum prediction, but they did not close every loophole in one experiment. Later experiments were designed precisely to remove or reduce those dependencies.

Aspect Experiments

Aspect, Grangier, Roger, Dalibard, and collaborators performed a series of influential Bell tests in the early 1980s. These experiments improved the optical realization and included a famous time-varying analyzer experiment.

The time-varying analyzer idea was aimed at the locality loophole. If the measurement settings are changed while the photons are already in flight, then a local mechanism has less opportunity to coordinate outcomes through ordinary subluminal communication between the source and the analyzers.

Aspect’s experiments were historically decisive because they made the conflict between quantum predictions and Bell inequalities vivid in a controlled optical setting. They were not, however, the final word on loopholes. The switching was periodic rather than generated by modern fast random setting choices, detector efficiencies remained limited, and later analyses clarified which locality and fair-sampling assumptions remained.

For that reason, the Aspect experiments occupy a middle historical position: stronger and more direct than the first generation, but not yet loophole-free in the modern sense.

For the focused historical treatment, see Aspect Experiments.

Loopholes

A loophole is not a vague doubt about quantum mechanics. It is a specific way in which the laboratory implementation might fail to realize the assumptions used in deriving the tested inequality.

The main loopholes are:

• The detection loophole: if only a biased subset of pairs is detected, the measured correlations may not represent all trials.
• The locality loophole: if setting choices, measurement events, or outcome registrations are not suitably spacelike separated, subluminal influences could in principle coordinate the data.
• The freedom-of-choice loophole: if hidden variables are statistically correlated with later setting choices, the usual setting-independence assumption fails.
• The coincidence-time loophole: if event pairing depends on settings or outcomes, the construction of pairs may bias the correlations.
• The memory and statistics loopholes: if trials are not independent or the statistical test is chosen after inspecting the data, significance can be overstated.

Modern experiments address these issues by using faster setting choices, better spacetime separation, event-ready heralding, high-efficiency detectors, predefined analysis windows, and statistical methods that do not assume independent identically distributed trials unless justified.

No experiment eliminates all logical assumptions. For example, extreme superdeterministic correlations between setting choices and hidden variables are not experimentally excluded in the same ordinary sense as the detection loophole. The practical achievement is narrower and stronger: major physically motivated loopholes have been closed in concrete Bell tests.

Loophole-Free Tests

In 2015, several experiments reported Bell inequality violations while closing the major detection and locality loopholes in the same implementation. Hensen and collaborators used entangled electron spins associated with nitrogen-vacancy centers separated by about $1.3$ km. Giustina and collaborators and Shalm and collaborators performed high-efficiency photonic tests.

These experiments were important because earlier tests often closed one major loophole at a time. High-efficiency ion or atom experiments could address detection but struggled with spacelike separation. Photon experiments could achieve large separation but historically suffered from detection inefficiency. The 2015 generation combined stronger spacetime control, improved detection, and more careful statistical analysis.

The phrase “loophole-free” should be read in this technical sense: the major traditional detection and locality loopholes are closed under stated assumptions. It does not mean that the experiment is assumption-free, that every philosophical escape route is impossible, or that the result licenses faster-than-light signaling.

For the focused modern treatment, see Loophole-Free Bell Tests.

What Experiments Show

Bell inequality experiments show that nature violates inequalities obeyed by local hidden-variable models satisfying the tested assumptions. Together with the theorem, the experiments rule out a broad class of local explanations of quantum correlations.

The conclusion is sharper than “quantum mechanics is weird.” A Bell violation says that observed correlations cannot be represented as

$$P(A,B\mid x,y) = \int d\lambda\, \rho(\lambda)\, P(A\mid x,\lambda)\, P(B\mid y,\lambda)$$

with the relevant independence and locality assumptions. The experimental result targets the factorized local hidden-variable structure, not merely an intuitive picture of particles carrying unknown properties.

Bell tests also support device-independent quantum-information ideas. If a violation is observed under carefully specified trust assumptions, it can certify forms of nonclassical correlation without requiring complete microscopic trust in the devices. That operational use is one reason Bell experiments became central beyond the foundations debate.

What Experiments Do Not Show

Bell experiments do not permit controllable faster-than-light signaling. The violation is seen in joint correlations after Alice and Bob compare data. The local marginal statistics remain consistent with no-signaling.

They also do not prove that measurement outcomes require consciousness, that all interpretations are ruled out, or that every entangled state violates the same inequality. Bell violation is a relation among a state, measurement choices, outcome statistics, and the tested assumptions.

Nor should “local realism” be treated as a single clean assumption. Different derivations distribute the assumptions among locality, outcome definiteness, hidden variables, setting independence, and probability theory in different ways. The responsible experimental statement is explicit: under the stated Bell-test assumptions, the observed correlations violate the corresponding local hidden-variable bound.

Common Misconceptions

• A Bell test is not just an entanglement witness.
• A violation of CHSH does not reach the algebraic maximum $4$ in quantum mechanics.
• Closing the detection loophole is not the same as closing the locality loophole.
• Aspect’s experiments were landmark tests, but not modern loophole-free tests.
• “Loophole-free” does not mean assumption-free.
• Bell violation does not imply a usable superluminal communication channel.
• The experiment does not by itself choose between all interpretations of quantum mechanics.

Cross-Links

• Foundations Experiments and Quantum Reality
• What These Experiments Do and Do Not Prove
• EPR Argument
• Bohm’s Spin Version of EPR
• Bell’s Theorem as Historical Turning Point
• Aspect Experiments
• Loophole-Free Bell Tests
• Bell Theorem
• CHSH Inequality
• Bell Tests
• Bell Test Experiments
• AMO Experiment Index
• Entanglement in Foundations
• Bell States
• Local Measurement Statistics
• Classic Papers
• Certification of Entanglement gives the operational trust hierarchy from calibrated witnesses through steering to Bell tests and robust self-testing.

References

• J. S. Bell, “On the Einstein Podolsky Rosen Paradox,” Physics 1, 195-200, 1964, DOI: 10.1103/PhysicsPhysiqueFizika.1.195.
• J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, “Proposed Experiment to Test Local Hidden-Variable Theories,” Physical Review Letters 23, 880-884, 1969, DOI: 10.1103/PhysRevLett.23.880.
• S. J. Freedman and J. F. Clauser, “Experimental Test of Local Hidden-Variable Theories,” Physical Review Letters 28, 938-941, 1972, DOI: 10.1103/PhysRevLett.28.938.
• A. Aspect, P. Grangier, and G. Roger, “Experimental Realization of Einstein–Podolsky–Rosen–Bohm Gedankenexperiment: A New Violation of Bell’s Inequalities,” Physical Review Letters 49, 91-94, 1982, DOI: 10.1103/PhysRevLett.49.91.
• A. Aspect, J. Dalibard, and G. Roger, “Experimental Test of Bell’s Inequalities Using Time-Varying Analyzers,” Physical Review Letters 49, 1804-1807, 1982, DOI: 10.1103/PhysRevLett.49.1804.
• B. Hensen et al., “Loophole-free Bell inequality violation using electron spins separated by 1.3 kilometres,” Nature 526, 682-686, 2015, DOI: 10.1038/nature15759.
• M. Giustina et al., “Significant-Loophole-Free Test of Bell’s Theorem with Entangled Photons,” Physical Review Letters 115, 250401, 2015, DOI: 10.1103/PhysRevLett.115.250401.
• L. K. Shalm et al., “Strong Loophole-Free Test of Local Realism,” Physical Review Letters 115, 250402, 2015, DOI: 10.1103/PhysRevLett.115.250402.
• N. Brunner, D. Cavalcanti, S. Pironio, V. Scarani, and S. Wehner, “Bell nonlocality,” Reviews of Modern Physics 86, 419-478, 2014, DOI: 10.1103/RevModPhys.86.419.
• J. S. Bell, Speakable and Unspeakable in Quantum Mechanics, 2nd ed., Cambridge University Press, 2004.

Exercises

1. Suppose a setting pair gives $N_{\mathrm{same}}=860$ and $N_{\mathrm{diff}}=140$. Compute the estimated correlation $E$.

Solution

Use

$$E = \frac{N_{\mathrm{same}}-N_{\mathrm{diff}}} {N_{\mathrm{same}}+N_{\mathrm{diff}}}.$$

Thus

$$E = \frac{860-140}{860+140} = 0.72.$$

2. Four measured correlations are $E(a,b)=0.70$, $E(a,b')=0.68$, $E(a',b)=0.69$, and $E(a',b')=-0.71$ using the sign convention on this page. Does the CHSH value violate the local bound?

Solution

With

$$S = E(a,b)+E(a,b')+E(a',b)-E(a',b'),$$

one gets

$$S = 0.70+0.68+0.69-(-0.71) = 2.78.$$

Since $\lvert S\rvert=2.78>2$, this violates the CHSH local hidden-variable bound.

3. Why can inefficient detection create a loophole in a Bell test?

Solution

If only some emitted pairs are detected, the recorded sample may not represent the full ensemble. A local hidden-variable model could, in principle, make detection depend on hidden variables and settings so that the detected subset shows stronger correlations than the full set. High-efficiency detection or an inequality designed for the actual detection structure is needed to avoid relying on a fair-sampling assumption.

4. Explain why Bell violation and no-signaling are compatible.

Solution

Bell violation concerns correlations between Alice’s and Bob’s outcomes after data from both sides are compared. No-signaling concerns each side’s local marginal distribution before comparison. Quantum mechanics can make the joint correlations violate a Bell inequality while keeping each local marginal independent of the remote setting.