Loopholes in Bell Tests

Bell's theorem is airtight as mathematics, but an experiment must satisfy its assumptions for the conclusion to follow. A loophole is a way a local-realistic theory could fake a Bell violation by exploiting an imperfection in the setup. Closing them is what makes a Bell test conclusive.

The locality (communication) loophole

Bell's derivation assumes Alice's outcome cannot depend on Bob's setting. If a signal traveling at or below light speed could pass from Alice's apparatus to Bob's between the moment a setting is chosen and the moment the outcome is registered, a local model could coordinate the results and mimic the quantum correlations.

How it is closed: make the two measurement events spacelike separated. Choose the settings randomly and complete each measurement so quickly that no light-speed signal could connect Alice's choice to Bob's outcome. Aspect's 1982 experiments switched analyzer orientations fast; later experiments separated the stations by hundreds of metres to over a kilometre.

The detection (fair-sampling) loophole

Real detectors miss particles. If only a fraction of pairs are detected, one must assume the detected sample is fair — representative of all pairs. A local model could violate this: particles might "choose" whether to be detected based on the setting, so that the detected subset shows correlations the full set does not.

How it is closed: raise the detection efficiency above a threshold so that no fair-sampling assumption is needed. For a maximally entangled pair with the standard CHSH settings the threshold is about $\eta \approx 0.83$ (precisely $2(\sqrt2 - 1)$ ); cleverly using non-maximally entangled states (Eberhard, 1993) lowers it to as little as $2/3$ . This is hard for photons, which are easily lost, but achievable for matter qubits like trapped ions, atoms, and nitrogen-vacancy centres in diamond.

The freedom-of-choice loophole

Bell assumes the settings $a$ and $b$ are chosen independently of the hidden variable $\lambda$ . If some common cause in the past influenced both the source and the setting choices, a local model could again conspire to fake a violation (the extreme version is superdeterminism, which denies free choice entirely).

How it is closed: derive the random settings from sources as independent as possible — fast quantum random-number generators, or even the light of distant quasars whose photons left billions of years ago ("cosmic Bell tests").

The decisive experiments

For decades, each experiment closed one loophole while leaving another open, so a determined local realist could always retreat to the unclosed one. In 2015, several groups achieved loophole-free Bell tests — closing the locality and detection loopholes simultaneously — using electron spins in diamond, and entangled photons. The 2022 Nobel Prize in Physics recognized this body of work (Aspect, Clauser, and Zeilinger).

Bottom line

Loophole-free Bell tests are the experimental verdict on EPR's question: local realism is false. The correlations of entangled particles cannot be explained by any theory in which outcomes are predetermined and influences respect locality. Nature is nonlocal — in a precise sense we make sharp in the next lesson.