PF r34 Qiskit Aer gallery: reproducibility and interpretation

These plots were generated on 22 September 2026 by calling the unmodified r34 quantum_channel.simulate_polarization_pair function. Every state and count shown comes from a fresh Qiskit Aer execution. The reference expressions are used only to check the output.

Reproduce

Use the PF r34 Python environment with its Qiskit/Aer dependencies:

python run_aer_gallery.py --pf-root /path/to/privileged-frame-simulation --config aer-gallery-config.json --output ./results

The configuration is an explicit function-level sweep specification for this reproduction script, not a PF GUI scenario preset. No provider credentials, hardware service or network data are required.

What was executed

• Qiskit 2.5.2 and Qiskit Aer 0.17.2, density-matrix method.
• Two source encodings: singlet and Phi-plus.
• 21 configured phase-correlation values from -1 to +1 for each encoding: 42 exact density-matrix evaluations.
• Both phase standard deviations are 0.6 rad. Source depolarization is 0.02. Deterministic phase is zero, and no storage or imported channel overrides are configured.
• Five points per encoding also execute all nine Pauli measurement settings with 8,192 simulated shots per setting: 10 tomography runs and 737,280 simulated shots.
• Seeds start at 9,222,026 and are explicitly saved for every evaluation. All original counts, real and imaginary matrix components, and the raw sampled reconstruction remain available in results/aer-gallery-raw.json.
• 210/210 checks pass: analytical Bell-family fidelity and purity references, unit trace, Hermiticity and positive semidefiniteness for each exact density matrix. Maximum fidelity-reference discrepancy is 1.9984014443252818e-15; maximum purity-reference discrepancy is 3.1086244689504383e-15.
• The finite-shot points are statistical estimates; the 210 checks concern the exact density matrices, not a claim that sampled estimates equal their exact targets.

Read the figures

qiskit-phase-sensitivity.png compares exact conditional fidelity/purity (lines) with finite-shot estimates (crosses). Under this declared Gaussian phase model, the singlet encoding is less sensitive to positively correlated common-mode phase noise; Phi-plus is less sensitive to anticorrelated noise. The result helps a researcher identify the phase-covariance information needed to choose and evaluate an encoding. It does not select a real apparatus without calibration.

qiskit-state-reconstruction.png shows the real components of the exact state and its raw finite-shot reconstruction at zero phase correlation. Small reconstructed off-diagonal components and deviations illustrate finite-count uncertainty. Both complete complex matrices are retained in the data. No physicality projection or clipping is used for this raw reconstruction.

Scope

This is a conditional two-photon polarization model. Classical PF modules separately calculate trajectories, emission/reception, optical loss and collection rates. The plotted noise covariance is a supplied apparatus model; it is not inferred from satellite location, distance, reference frame or privileged-frame selection. Simulated Born-rule counts are not measurements from quantum hardware or photon detectors. Numerical agreement with an analytical reference establishes implementation agreement for these cases, not experimental accuracy, mission qualification, a positive key budget or a PF advantage.

The results/aer-gallery-summary.json file contains versions, source-module SHA-256 hashes and execution details. The CSV contains every plotted scalar. The raw JSON preserves all state and count outputs.