Skip to content

SCPN/FIM repeated IBM follow-up analysis

Date: 2026-05-05

Backend: ibm_kingston

Public run label: ibm-run-cf4835290f607387

Artefacts

Pending submission:

  • data/scpn_fim_hamiltonian/fim_ibm_repeated_followup_pending_2026-05-05_ibm-run-cf4835290f607387.json
  • SHA256: b2c183631e1ead2b41120a69eca311d784a067b3ad95489570061d9eac2858f9

Raw counts:

  • data/scpn_fim_hamiltonian/fim_ibm_repeated_followup_raw_counts_2026-05-05_ibm-run-cf4835290f607387.json
  • SHA256: 6e4df78f1c679cd29b9c503bf1fecf39be76e707b1e6c4df99bbcc87b8e50d44

Analysis:

  • scripts/analyse_fim_ibm_repeated_followup.py
  • data/scpn_fim_hamiltonian/fim_ibm_repeated_followup_analysis_2026-05-05_ibm-run-cf4835290f607387.json
  • SHA256: 8bd361b360248a0f00880ad23e9d4898d6e4b2c1ac12a60739e977890b0b18db

Row metrics:

  • data/scpn_fim_hamiltonian/fim_ibm_repeated_followup_row_metrics_2026-05-05_ibm-run-cf4835290f607387.csv
  • SHA256: de26e6d9e9332e0fcc72e80948362a03326ab4b20add437f496fc6a1b28fa8c9

Comparisons:

  • data/scpn_fim_hamiltonian/fim_ibm_repeated_followup_comparisons_2026-05-05_ibm-run-cf4835290f607387.csv
  • SHA256: e9f78765d988224c95a40af9435446c12202079eb80e594b0fa315ae8c5569fe

Reproduction command:

env PYTHONPATH=src /home/anulum/.local/bin/python scripts/analyse_fim_ibm_repeated_followup.py --verify-integrity

Run size

  • Circuits: 166
  • Main repeated FIM rows: 150
  • Readout baseline rows: 16
  • Shots: 339,968
  • Shots per circuit: 2048
  • Wait wall time: 446.0912413597107 seconds
  • Maximum live transpiled depth in returned metadata: 540
  • Maximum live two-qubit gates in returned metadata: 158

Readout baseline

The readout-only baseline over all 16 computational basis states gives:

  • Mean exact-state retention: 0.98272705078125
  • Mean magnetisation leakage: 0.01727294921875
  • Mean parity leakage: 0.017181396484375

This supports state-specific sanity checks and parity-flip correction. It is not a full confusion-matrix inversion.

Primary result

The repeated follow-up falsifies the simple hardware-protection interpretation for this backend/circuit family.

For lambda = 4 relative to lambda = 0, across 15 matched state/depth comparisons with five replicates per condition:

  • State retention mean delta: -0.0796875
  • State retention Fisher p: 1.1055758219655449e-48

  • State retention deltas: 0 / 15 positive, 15 / 15 negative

  • Magnetisation leakage mean delta: +0.08955729166666666

  • Magnetisation leakage Fisher p: 7.488374675730006e-55

  • Magnetisation leakage deltas: 14 / 15 positive, 1 / 15 negative

  • Parity leakage mean delta: +0.06151692708333333

  • Parity leakage Fisher p: 6.267550425892641e-48

  • Parity leakage deltas: 13 / 15 positive, 2 / 15 negative

  • Readout-corrected parity leakage mean delta: +0.06439517298321257

  • Readout-corrected parity leakage Fisher p: 6.267550425892548e-48
  • Readout-corrected parity leakage deltas: 13 / 15 positive, 2 / 15 negative

Interpretation: increasing lambda from 0 to 4 decreased exact-state retention and increased leakage for nearly all matched conditions. The sign is opposite to a simple FIM hardware-protection claim.

Claim boundary

Allowed claims:

  • The repeated IBM run completed and produced complete raw count dictionaries.
  • On ibm_kingston, for this n=4 Trotter circuit family, lambda = 4 increased measured leakage relative to lambda = 0 in the repeated design.
  • This falsifies the simple claim that the tested FIM term improves hardware coherence under this circuit/backend configuration.

Blocked claims:

  • No backend-general FIM protection claim.
  • No hardware many-body-localisation claim.
  • No full confusion-matrix mitigation claim.
  • No claim that the negative result invalidates all possible FIM Hamiltonian designs; it applies to this backend, circuit construction, depth set, and observable set.

Scientific consequence

The SCPN/FIM paper should be framed as:

  • A theoretically and numerically defined self-referential Hamiltonian family.
  • An offline spectral/entanglement/VQE methods result.
  • A hardware falsification of the simple coherence-protection hypothesis for the tested digital Trotter implementation on ibm_kingston.

This is still scientifically useful because it prevents overclaiming and identifies that the FIM term, as implemented here, adds circuit complexity/noise faster than any protection effect can be observed on this hardware configuration.