ProductBlogAlgorithmsAboutSign inSign up

Research Replication

Bell/CHSH: Correlations and the Classical Bound

Study a four-setting Bell experiment, compare simulator correlations with the CHSH bound, and understand what the recorded evidence establishes.

Hosted simulator evidence · Public package pending

Can the correlations of an entangled two-qubit state exceed the CHSH bound satisfied by local hidden-variable models?

Source paper: Clauser, Horne, Shimony and Holt, Proposed Experiment to Test Local Hidden-Variable Theories (1969).

Read the source paper →

What this replication investigates

This reference prepares an entangled pair and samples four measurement settings. It reconstructs correlations and combines them into the CHSH statistic S, comparing the finite-shot estimate with the classical bound 2 and the ideal quantum value 2√2.

This is a hosted-simulator demonstration of the CHSH calculation. It is not a physical Bell test and does not implement separated laboratories or detection/locality loophole closure.

Method and parameters

  • Prepare the two-qubit Bell state Φ+ and apply the fixed measurement rotations for each of the four settings.
  • Sample 10,000 shots per setting: 40,000 shots in each four-circuit dataset.
  • For each setting, assign +1 to equal outcomes and −1 to unequal outcomes to estimate E.
  • Use the reference convention S = E(a,b) − E(a,b′) + E(a′,b) + E(a′,b′).
  • Compare the original hosted dataset with a separately executed restored-source rerun; retain each dataset independently.

Results and comparison

Scroll the table horizontally to see every results column.

Scoped CHSH reference: theory and two recorded hosted-simulator datasets
ReferenceShots per settingS95% simultaneous S interval
Local hidden-variable boundNot a sampled dataset2Not applicable
Ideal quantum predictionExpectation value2√2 ≈ 2.8284Not applicable
Original hosted simulator10,0002.8230[2.7638, 2.8822]
Restored-source hosted rerun10,0002.8404[2.7812, 2.8996]

The two sampled estimates lie near the ideal quantum prediction. A finite-shot estimate can exceed 2√2; that does not establish a quantum expectation above the theoretical bound.

The retained comparator detected no discrepancy using a two-sided Hoeffding screen with a union bound across 21 discrepancies at α = 0.05. The table’s simultaneous intervals use a separate family covering the two S estimates. Non-rejection is not proof that the two distributions are equivalent. Sampling uncertainty does not account for drift, systematic error, or correlated shots.

What the evidence establishes

The current Bell evidence includes eight completed hosted-simulator circuit jobs: four original settings and four restored-source rerun settings. Preserved counts and source identities support this particular demonstration.

A source archive and an authored Report are distinct from a publicly reusable platform Capsule. Public Capsule access and a separate receiving team’s reproduction have not been verified.

Limitations

  • Simulator correlations do not establish a physical Bell violation or a loophole-free experiment.
  • Computational-basis counts from the editor’s simple Bell preparation alone are insufficient: the CHSH reference uses four measurement settings.
  • Historical hardware records are separate evidence and are not used here as proof of a new hardware run or statistically established hardware violation.

Access and reproduction

The public paper and this scientific overview are available now. Publication of the experiment package, Report and Capsule is still being qualified by the Replications project.

The experiment's source package and runnable Capsule are not yet public. This page describes the scoped reference and its evidence; it does not provide a one-click reproduction.

Learning example or research replication?

The learning example prepares a Bell state and shows its computational-basis histogram. This research reference adds four measurement settings, a CHSH statistic and a scientific comparison. Running the learning snippet does not reproduce this experiment.

Start with the Bell State learning example →

Explore the research collection →

References

Scientific scope and retained results reviewed with the Marqov Replications project on October 16, 2026. Each reference is a separate experiment from the editor's learning examples.