Bell's Inequalities

Why Bell matters

For thirty years after Einstein-Podolsky-Rosen (1935) accused quantum mechanics of incompleteness, the question of whether nature might secretly be governed by hidden variables looked metaphysical. John Bell, working alone at CERN in 1964, found a four-page argument that turned the question into an experiment. Hidden-variable theories of a particular kind — local ones — make quantitative predictions about correlation measurements that quantum mechanics violates by a calculable amount. Either nature follows local hidden-variable rules (and QM is wrong about something measurable) or it doesn't (and the EPR program fails). Six decades later, every loophole has been closed; nature has chosen, and her choice is not local realism.

• Foundations of physics. Bell's theorem closed the door on the most natural way of "completing" quantum mechanics. Either you keep QM as it is, or you adopt a non-local theory like Bohmian mechanics. There is no local realistic alternative.
• Quantum cryptography. E91 (Ekert 1991) and device-independent QKD use Bell-violation as a security certificate. If the observed |S| > 2, the channel cannot be classically simulated — therefore no eavesdropper has hidden in classical variables. Demonstrated for 800 km fiber (Yin 2017 with Micius satellite) and now for chip-scale photonic devices.
• Quantum random number generation. Pironio et al. (2010) showed Bell violations certify true randomness. Commercial QRNGs based on this principle (NIST randomness beacon, ID Quantique) are now in mainstream use.
• Quantum advantage demonstrations. The "quantum supremacy" claims by Google's Sycamore (2019) and Chinese Jiuzhang (2020-2024) photonic experiments use Bell-style nonlocality as part of their certification of inability to simulate classically.
• Self-testing and certification. Bell-style inequalities let you certify that a quantum device actually contains the entanglement it claims, without trusting the device — known as "device-independent" certification, central to scalable quantum networks.

From Bell to CHSH to Tsirelson

Bell's original 1964 inequality was for spin correlations measured along three angles. The CHSH form (Clauser, Horne, Shimony, Holt 1969) is more practical: two settings per side, each yielding ±1. The classical bound |S| ≤ 2 follows from a one-line algebraic argument using λ-conditioned factorization. Tsirelson's 1980 theorem proves the quantum bound 2√2 from the geometry of Hilbert space.

Why does QM violate the inequality? In classical correlation, the conditional dependence structure forces a strict bound. In QM, the entangled state |Ψ⁻⟩ = (|↑↓⟩ − |↓↑⟩)/√2 has correlation function E(a, b) = −cos(θ_a − θ_b) — sinusoidal in the angle difference. The cosine's "bend" is what allows |S| to exceed 2; the maximum 2√2 is attained when (a, a', b, b') sit at the optimal angles 0°, 45°, 22.5°, 67.5°.

The Tsirelson bound 2√2 is itself remarkable — there's no a priori reason quantum correlations should respect any bound below the algebraic maximum of 4. Why does QM stop at 2√2 rather than approaching the no-signalling bound? "Information causality" (Pawlowski et al. 2009), "macroscopic locality," and "global non-trivial communication complexity" have all been proposed as principles that single out QM from more strongly correlated theories. None has yet won universal acceptance.

A condensed experimental timeline

• 1972 — Freedman & Clauser. First Bell test, photon-cascade source from calcium atoms. Result violated the inequality but had loopholes.
• 1982 — Aspect, Grangier, Roger. Switching the polarizer settings during the photon's flight closed the locality loophole (subject to detection efficiency caveats).
• 1998 — Weihs. 400 m separation, fast random switching, in Innsbruck.
• 2001 — Rowe. First detection-loophole-free test (with trapped ions, but space-like separation small).
• 2015 (autumn) — Three independent loophole-free tests. Hensen et al. (Delft, NV centers, 1.3 km), Giustina et al. (Vienna, photons), Shalm et al. (NIST, photons). All closed locality and detection simultaneously.
• 2017 — Handsteiner et al. "cosmic Bell test." Used quasar light (settings determined by photons emitted ~7.8 Gyr ago) to address freedom-of-choice loophole.
• 2018 — BIG Bell Test. 100,000 humans worldwide pressing keys to choose measurement settings, addressing freedom-of-choice with human-scale randomness.
• 2022 — Aspect, Clauser, Zeilinger Nobel Prize. "For experiments with entangled photons, establishing the violation of Bell inequalities, and pioneering quantum information science."
• 2022 — DIQKD demonstrations. Nadlinger et al., Zhang et al., Liu et al. — first device-independent quantum key distribution using Bell violation as security certificate.

Common misconceptions

• "Bell tests prove faster-than-light signaling." No. The no-signalling theorem proves that Alice's marginal statistics are independent of Bob's settings — entanglement does not transmit usable information faster than light. Bell tests show correlations exceed local-realistic bounds, but the correlations themselves cannot be used to communicate.
• "Quantum mechanics is non-local." A specific interpretation of Bell's result. QM is non-realistic in the standard reading; non-local realism (Bohmian mechanics) is one alternative; many-worlds avoids the question by denying single outcomes. "Non-local" in colloquial use often confuses Bell-non-locality with FTL signalling — the two are different.
• "Loophole-free experiments closed everything." They closed the major loopholes (locality, detection, freedom-of-choice up to cosmological scales). Superdeterminism — a conspiracy in the initial conditions of the universe — is a logical possibility most physicists reject as ad hoc.
• "Bell's theorem is just statistics." The CHSH bound 2 is a strict consequence of any local realistic model, regardless of how clever or hidden the variables. It's not a statistical artifact but a structural constraint on the joint probability distributions.
• "Bell rules out hidden variables." Bell rules out local hidden variables. Non-local hidden variables (Bohmian) survive. Bell himself was sympathetic to Bohmian mechanics late in his life.
• "Bell only applies to spin-½ singlets." The original argument did, but generalizations cover any two-party two-setting two-outcome scenario, plus tripartite (GHZ, Mermin) and continuous-variable versions. The CHSH formulation is the most-tested but far from the only one.

Active research frontiers

• Beyond bipartite Bell. GHZ states (three-particle Mermin inequality) demonstrate "all-or-nothing" violations: certain outcomes have probability 0 in QM but probability 1 in any local realistic theory.
• Self-testing. Maximal Bell violations uniquely fix the underlying state and measurements (up to local isometries). Used to certify quantum devices without trusting them.
• Quantum networks. Multi-party Bell violations are the protocol layer for the future quantum internet — entanglement swapping, Bell-state measurements, and routing.
• Bell tests in gravity. Proposals to test whether gravitationally-mediated entanglement can violate Bell inequalities, probing whether gravity has quantum features (Bose-Marletto-Vedral 2017).
• Relativistic Bell tests. Recent proposals study how Bell violations behave under acceleration and in curved spacetime, e.g., between accelerated and inertial observers.