Skip to content

Phase 3 Entanglement/Tomography Paper Plan

Date: 2026-05-20

Working Title

Reduced-Pauli entanglement checks for DLA-sector leakage mechanisms on IBM Heron hardware

Alternative:

Mechanism-boundary tomography for parity-sector leakage in Kuramoto-XY quantum circuits

Thesis

The paper should not try to claim a new advantage result. Its strongest clean angle is mechanism separation:

  • the earlier DLA-parity programme showed a real, backend-sensitive leakage asymmetry;
  • the Phase 3 state/layout controls showed that layout and prepared state can dominate the original contrast;
  • this campaign asks whether the promoted DLA and FIM circuits also show a measurable reduced-Pauli entanglement-structure companion.

The paper remains publishable if the result is null: a bounded negative result would show that reduced-Pauli tomography at this cost does not explain the observed leakage mechanisms.

Evidence Status

Completed:

  • reduced-Pauli readiness package;
  • exact reference rows for 54 observables;
  • approval-gated IBM runner;
  • no-submit live preflight on ibm_marrakesh;
  • approved QPU submission on ibm_marrakesh;
  • raw-count artefact with job IDs;
  • generated observable-summary CSV and manifest;
  • post-run raw-count reducer;
  • paper asset generator with label summary, basis summary, top-deviation table, and heatmap figure;
  • second-backend ibm_fez replication;
  • pinned-layout ibm_fez repeat with full 16-state correlated readout calibration;
  • preregistered five-channel ZNE stress test on ibm_fez physical qubits [21,22,23,24];
  • ZNE reducer and manifest: docs/campaigns/phase3_entanglement_zne_manifest_2026-05-20_ibm_fez_zne.md.

Current first-pass result:

  • completed artefact: data/phase3_entanglement_tomography/entanglement_tomography_live_ibm_marrakesh_2026-05-20T004334Z.json;
  • jobs: d86g7h1789is738vkreg, d86ggpis46sc73f6v170;
  • 54 observable rows;
  • mean absolute deviation from exact reference: 0.12989296537986128;
  • maximum absolute deviation from exact reference: 0.5560906424788263;
  • observable rows SHA256: 3d18308d60fe32827bae7517f18fd71690240b105779287408c4749cb0e7dc72.
  • paper asset manifest: data/phase3_entanglement_tomography/entanglement_tomography_paper_assets_2026-05-20.md.

Pending:

  • final venue formatting and bibliography pass.

Completed ZNE stress-test result:

  • completed artefact: data/phase3_entanglement_tomography/entanglement_tomography_live_ibm_fez_2026-05-20T023600Z.json;
  • jobs: d86hs6qs46sc73f70h90, d86hsltg7okc73el4lg0;
  • 15 scale rows over 5 preregistered channels;
  • scale-1 mean absolute deviation: 0.4196274906488838;
  • linear-ZNE mean absolute deviation: 0.44124749932943935;
  • full-readout-mitigated linear-ZNE mean absolute deviation: 0.4468173734963394;
  • channel-summary SHA256: 775594dd3efc8f519111365179659d41903cf774bcde1d20a378d53b6ee8e7aa.

Minimum Run

Field Value
Backend ibm_marrakesh
Physical qubits [1, 2, 3, 4]
Main circuits 162
Readout circuits 4
Total circuits 166
Main shots 2048
Readout shots 8192
Estimated QPU minutes 1.5217
Budget ceiling minutes 25.0
Max transpiled depth 388
Max basis-expansion ratio 1.0718232044198894

Executed submission command:

python scripts/phase3_entanglement_tomography_ibm.py --backend ibm_marrakesh --submit --confirm-budget

Executed analysis command:

python scripts/analyse_phase3_entanglement_tomography.py \
  data/phase3_entanglement_tomography/entanglement_tomography_live_<backend>_<timestamp>.json

Claim Boundary

Safe after successful raw-count analysis:

  • reduced-Pauli correlators were measured for the preregistered DLA and FIM circuit families;
  • measured correlators can be compared with exact references under a fixed layout and shot budget;
  • the result either supports or rejects a small-system entanglement-structure companion to the leakage/retention mechanism.

Blocked even after a positive result:

  • quantum advantage;
  • scalable tomography;
  • backend-general entanglement dynamics;
  • full-state tomography;
  • claims about unmeasured qubits, depths, layouts, or backends.

Result Decision Tree

Raw-count outcome Paper framing
Correlators track exact references within uncertainty Hardware supports the reduced-Pauli reference structure for the promoted circuits; leakage mechanism is not explained by gross entanglement-observable drift.
Correlators deviate coherently by family or depth Entanglement-structure companion is visible and becomes a candidate mechanism for follow-up controls.
Uncertainty dominates deviations Measurement-cost boundary; the promoted tomography block is too small to support mechanism claims.
Readout correction flips or erases the comparison Readout-calibration boundary; no entanglement interpretation should be promoted.

Figures and Tables

Minimum paper assets:

  1. Table of preflight gates: circuits, shots, depth, basis-expansion ratio, QPU-minute estimate.
  2. Heatmap of measured minus exact Pauli expectations by circuit label and observable.
  3. Family-level aggregate table for DLA shallow/signal and FIM lambda pair.
  4. Largest-deviation table for the dominant measured correlator channels.
  5. Boundary table: supported claims, blocked claims, falsifiers.
  6. ZNE stress-test table for the four DLA transverse channels and one FIM control channel.

Generated assets:

  • figures/phase3/phase3_entanglement_deviation_heatmap_2026-05-20.png
  • figures/phase3/phase3_entanglement_deviation_heatmap_2026-05-20.pdf
  • data/phase3_entanglement_tomography/entanglement_tomography_label_summary_2026-05-20.csv
  • data/phase3_entanglement_tomography/entanglement_tomography_basis_summary_2026-05-20.csv
  • data/phase3_entanglement_tomography/entanglement_tomography_top_deviations_2026-05-20.csv

Submission Target

Best fit after raw counts:

  • workshop or short communication first, then expand if the hardware result is mechanistically sharp;
  • categories: quant-ph, with possible cs.ET cross-list if positioned as a reproducibility/control package;
  • do not submit as a quantum-advantage paper.

Current Manuscript Source

Manuscript source:

  • paper/submissions/submission_005_phase3_reduced_pauli_entanglement/phase3_entanglement_tomography.tex

Built PDF:

  • paper/submissions/submission_005_phase3_reduced_pauli_entanglement/phase3_entanglement_tomography.pdf

Historical scaffold:

  • paper/submissions/submission_005_phase3_reduced_pauli_entanglement/phase3_entanglement_tomography_short_paper.md

The canonical LaTeX manuscript now records the completed raw-count execution, readout-sensitivity extensions, second-backend replication, full correlated readout calibration, and the five-channel ZNE stress test. Next paper work is venue-specific formatting and final bibliography polishing, not additional QPU spend for this paper.