English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Bell Inequality: The Experiment Where Einstein Lost

Forum topic · 小凯 · 2026-11-07

Summary

This article explains the Bell inequality (CHSH game) as a 'lie detector' for local realism. It translates the hidden-variables ('pre-written answer sheet') worldview into a game where two parties measure entangled particles and compute S from average products of ±1 outcomes. Pure algebra shows any local hidden-variable theory obeys S ≤ 2, while quantum mechanics predicts S = 2√2 ≈ 2.83 for suitable measurement angles—exceeding the bound by 40%, the Tsirelson limit. The article reviews three generations of experiments: closing the communication (locality) loophole starting in 1982 and with satellite-based tests over ~1000 km after 2017, and the detection (selection) loophole closed in 2015 by three teams with high-efficiency detectors, culminating in the 2022 Nobel Prize in Physics. Crucially, the no-communication theorem survives: the correlations cannot transmit messages, so special relativity is unharmed—what falls is 'pre-existing answers,' not locality itself. This foundation underpins no-cloning and true quantum randomness discussed in later installments.

Bell Inequality: The Experiment Where Einstein Lost

> Prerequisites: Episodes 003 (the glove camp and the entanglement mystery), 016 (cross terms), 017 (changing measurement axes). > Next up: 028, "No-Cloning: The Common Starting Point of Quantum Money and Quantum Cryptography."

Episode 003 settled one mystery: why entanglement can't send packages. The conclusion there was "the correlations are real, but no information gets through." But there is a more fundamental lawsuit that was only opened then—

The glove camp claims: two entangled particles are like a left and right glove packed in two boxes. Before you open the boxes, the answers were already set at the factory—which glove is in which box is written on each particle's "factory sheet"; we just haven't read it. No magic, no spooky action at a distance, only "information not yet available." This claim is eminently reasonable, and even Einstein stood with it ("spooky action at a distance" was his mockery of the rival view).

How does a reasonable claim lose? The answer: in 1964 the mathematician-physicist Bell found a lie detector—a numerical red line that any "answers pre-written at the factory" theory must obey—and experimentalists, over three generations, measured the quantum world crossing the line. This post disassembles that lie detector. The entire argument needs only addition, multiplication, and ±1.

The lie-detection game: the limit of factory answer sheets

Translate the glove camp's claim into a game. Alice and Bob are far apart, each holding one particle of a common (entangled) origin. According to the glove camp, each particle carries a factory answer sheet: a pre-written answer to every possible question, and each answer is either +1 or −1.

Simplified setup: Alice's question box has only two questions (call them a and a′), Bob's has two (b and b′). So Alice's sheet pre-writes ±1 for a and for a′; Bob's sheet pre-writes ±1 for b and b′.

Rules: each round, Alice and Bob independently and randomly pick one question each, read the pre-written answers, then—after the fact—compute the product of the two numbers. After thousands of rounds, compute the "average product." Note: reconciling scores afterwards via classical communication is fine; the only foul is secretly coordinating on the spot.

Now the lie detector appears. Combine the four average products as:

S = (avg product of a & b) + (avg product of a & b′) + (avg product of a′ & b) − (avg product of a′ & b′)

How large can S be under the glove camp's theory? Look at any single round: with sheet values a, a′, b, b′ all ±1, the per-round score is a·b + a·b′ + a′·b − a′·b′ = a·(b+b′) + a′·(b−b′). Since b and b′ are ±1, one of (b+b′) and (b−b′) must be 0 and the other ±2—so the per-round score is always ±2, no matter how the sheet is drawn. Averaging over all rounds:

> S ≤ 2. This is the glove camp's fatal spot: however the factory sheets are written, however complex the mechanism, as long as answers are "pre-written before departure and read locally," S can never exceed 2.

This is pure algebra, nothing to do with quantum theory. It is a red line for worldviews: any theory that is both "local and real" (answers fixed in advance, information travels only through local channels) lies below 2.

Quantum mechanics: blatantly over the line at 2.83

What does quantum mechanics predict for the same game? Play it with the entangled dice pairs from Episode 003: both sides change measurement axes (Episode 017: a question = choosing a measurement axis), with the four questions set at four specific angles. Each average product is then given by the cross term (Episode 016: angle enters the ledger), and the math gives—

> S = 2√2 ≈ 2.83.

Quantum mechanics predicts: the red line can't hold—exceeded by 40%. And this is not a theoretical flourish: 2√2 is nature's bounty for this game (mathematically proven: no correlation can exceed it), and quantum mechanics takes it in full.

Two worldviews, for the first time, give different numbers on the same measurable quantity: one says "at most 2," the other "up to 2.83." At this point only one thing remains: run the experiment and count.

Three generations: plugging the loopholes one by one

The experiments were harder than expected, because the lie detector has two cracks defense lawyers will exploit:

Crack one: secret collusion. If the two sides are close enough to signal each other during measurement, the "±2 per round" derivation leaks—answer sheets that coordinated aren't factory sheets. From 1982 on, experiments ensured both choices and measurements completed in less time than light takes to travel between them; after 2017, the two sides were separated to the thousand-kilometer scale (ground and satellite), sealing this crack completely.

Crack two: selective response. If detectors respond to only some particles, one could in principle "pick which rounds to answer"—counting only responding rounds could flatten the books. In 2015, three teams almost simultaneously sealed this with high-efficiency detectors: nearly every particle answers; nothing to pick from.

Three generations of experiments (from the first real game in 1972, to 2015 when both major loopholes were closed the same year, to thousand-kilometer-scale retests) agree: S ≈ 2.8, the red line trampled underfoot. The 2022 Nobel Prize in Physics went to the three generations of main forces who turned this lawsuit from a thought experiment into an ironclad case—the closest trophy to today in this column's storyline.

What was lost, what was won

At final judgment, the wording must be precise:

What the glove camp lost is the "factory answer sheet"—before measurement, particles carry no pre-written complete set of answers. The world, before a measurement occurs, is not a finished script. This is not a philosophical nicety; it is a physical assumption excluded by numbers.

But Einstein's other treasure is intact: relativity. The no-communication result from Episode 003 stands unchanged in this lawsuit—the S ≈ 2.8 correlations cannot send any message (Alice's own score sequence alone is perfectly uniform random; 2.8 hides in the joint ledger, not in either side's separate book). So the "distance action" has no transmitter, no telegraph, no usable information channel—what was overthrown is the "realism" half of local realism; the "locality" half survived in the form of no-communication.

For quantum computing, the lawsuit's spoils include something even more practical: since there are no pre-written answers before measurement, the correlations carried by entangled bits are something classical bits cannot in principle contain—Episode 028's no-cloning and Episode 029's true randomness both start from this cornerstone: "the answers are not written at the factory."

Taking the analogy apart

What the "lie detector" analogy borrows—and where it fails:

  • Disanalogy 1: A lie detector tests whether someone is lying *now*; this tests whether answers were *pre-written in the past*—it excludes not deception but a worldview, and its subject (nature itself) can hardly "intend to hide."
  • Disanalogy 2: A court's evidence chain collapses with one false testimony; here thousands of statistical rounds mean no single "accident" can overturn the verdict—the judge is the law of large numbers, not a judge.

Luggage to take home

1. The Bell/CHSH game is a lie detector tailored for worldviews: factory answer sheets (±1) force the per-round score to ±2, hence S ≤ 2; quantum entanglement at the right angles gives S = 2√2 ≈ 2.83—two worldviews give different numbers for the first time. 2. Three generations of experiments closed the "secret collusion" and "selective response" loopholes, capped by the 2022 Nobel Prize: no pre-written answers before measurement—local realism is out. 3. No-communication leaves relativity unharmed—what acts at a distance is correlation, not information; the lawsuit's spoils ("answers not pre-written") are the foundation for 028 (no-cloning) and 029 (true randomness).

Hook for the next stop: if answers aren't written at the factory, then "copying an unknown state" isn't merely technically hard—it is directly forbidden by physics. And that "forbidden" is precisely the fail-safe of quantum money and quantum cryptography. Episode 028: "No-Cloning: The Common Starting Point of Quantum Money and Quantum Cryptography."

Tags

#quantum-computing#bell-inequality#chsh-game#local-realism#entanglement#quantum-foundations#nobel-prize-2022#no-communication-theorem

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178636861