DLA State/Layout Randomisation Preregistration¶
Date: 2026-05-06
This preregistration prepares a systematic state/layout randomisation control for the DLA parity hardware programme. It does not submit an IBM job, reserve backend time, or authorise QPU spend.
Scientific Question¶
How much of the promoted n=4 leakage asymmetry is explained by parity sector,
excitation count, physical qubit layout, readout properties, and coupling-map
placement?
Claim Boundary¶
Supported after successful execution and analysis:
- mechanism-separation evidence for the existing small-system DLA parity observation;
- state-level and layout-level leakage summaries;
- identification of whether excitation count or layout explains a substantial share of the original contrast.
Blocked even after a positive result:
- DLA-parity-only causality;
- backend-universal protection;
- monotone scaling;
- quantum advantage;
- full readout-matrix mitigation, unless all 16 basis states are calibrated for every selected layout.
Backend and Layout Rule¶
Use one Heron-class backend selected immediately before live readiness checks.
The backend may be ibm_kingston only if the explicit purpose is same-device
mechanism separation rather than multi-device transfer.
Select three connected four-qubit windows from the live coupling map:
- each window must support the generated circuit topology after transpilation;
- windows should minimise recent readout error and two-qubit error where the provider exposes calibration metadata;
- no window may be selected manually after seeing outcome counts.
Circuit Matrix¶
| Field | Value |
|---|---|
n |
4 |
| Coupling model | same heterogeneous Kuramoto-XY matrix as Phase 2 A+G |
| States | 0011, 0101, 0001, 0010, 0111 |
| Depths | 6, 8, 10, 14 |
| Layouts | 3 connected four-qubit windows |
| Repetitions | 8 per state/depth/layout |
| Main shots | 4096 per circuit |
| Readout states | all five prepared states per layout |
| Readout shots | 8192 per circuit |
Circuit count:
- main circuits:
5 states x 4 depths x 3 layouts x 8 reps = 480; - readout circuits:
5 states x 3 layouts = 15; - total circuits:
495.
QPU-Time Estimate¶
Expected IBM-reported QPU time: 8-15 minutes if live-transpiled depths remain
within the popcount-control envelope.
Budget ceiling for this preregistered block: 20 IBM-reported QPU minutes.
Abort before submission if the live estimate exceeds the ceiling or if the remaining allocation cannot cover the block plus a 25 % safety margin.
Live Readiness Gates¶
Before submission:
- confirm backend is Heron-class, account-visible, and operational;
- select three connected four-qubit windows before outcome data exists;
- record window qubits, calibration timestamp, readout errors, and two-qubit error summaries where available;
- generate circuits from committed code only;
- live-transpile every circuit on the selected backend and layout;
- reject if max depth exceeds the completed popcount-control depth envelope by more than 25 %;
- reject if max two-qubit gate count exceeds the completed popcount-control envelope by more than 25 %;
- record shot count, circuit count, expected QPU minutes, depth summary, and two-qubit gate summary;
- get explicit approval immediately before submission.
Analysis Plan¶
Primary observables:
- parity leakage;
- exact-state retention;
- excitation-count leakage where the observable is well defined;
- readout-only exact-state retention for the prepared calibration states.
Primary model:
- compare leakage by state, parity sector, excitation count, depth, and layout;
- report layout-stratified and layout-pooled summaries;
- treat layout as a grouping factor, not as a nuisance to average away without reporting.
Promoted summaries:
- per-state/depth/layout leakage table;
- parity-sector contrast at matched excitation controls;
- same-popcount within-sector swap contrasts;
- excitation-inversion contrast;
- layout variance and worst/best layout spread;
- sign agreement with the original Phase 2 A+G contrast where applicable.
Readout handling:
- use exact-state readout calibrations for the five prepared states;
- do not claim full
2^n x 2^nconfusion-matrix mitigation from five-state calibration; - record whether a future complete 16-state basis calibration is justified.
Falsification Rules¶
The clean parity-sector explanation is weakened if:
- same-popcount within-sector swaps are comparable to or larger than the original parity contrast;
- layout variance dominates the parity/excitation effect;
- the sign changes across selected layouts;
- readout-only correction removes the promoted sign;
- excitation count explains the observed ordering better than parity sector.
If the result is mixed, report it as mechanism-separation evidence rather than as confirmation or failure of the original paper.
Output Artefacts¶
Expected paths after approved execution:
data/phase3_state_layout_dla/phase3_state_layout_<backend>_<timestamp>.json;data/phase3_state_layout_dla/phase3_state_layout_summary_<date>.json;data/phase3_state_layout_dla/phase3_state_layout_row_metrics_<date>.csv;data/phase3_state_layout_dla/phase3_state_layout_layout_metrics_<date>.csv;docs/campaigns/phase3_state_layout_dla_manifest_<date>.md.
Each artefact must include job ID, backend, layout mapping, calibration metadata, raw counts, SHA256 hashes, depth/gate summaries, and reproduction commands.
Submission Boundary¶
This preregistration is complete. QPU execution remains blocked until backend selection, live layout selection, transpilation readiness artefacts, budget confirmation, and explicit approval are completed in a separate task.