IQM Garnet Parity-Asymmetry Depth-Profile Preregistration

Date: 2026-07-22

This preregistration commits the design, decision rule, and budget for the depth-profile follow-up to the executed powered backend-sensitivity block (iqm_dla_backend_sensitivity_powered_prereg_2026-07-21.md, executed 2026-07-21, primary rejected H0 at p = 1.1 × 10⁻⁵). It is committed BEFORE any of its own hardware data exist. Execution requires a separate explicit per-submit owner GO.

Motivating (already-collected) evidence

The 2026-07-21 powered block found the pooled even-over-odd parity-leakage asymmetry replicates the IBM direction, but with a depth profile unlike IBM's flat reference: +0.230 at depth 4, +0.083 at depth 6, −0.049 at depth 10 (sign reversed), with both sectors approaching ~0.46–0.48 total leakage at depth 10 (near sector equilibration). Two open questions follow:

Q1: Where between depths 6 and 10 does the Garnet asymmetry cross zero? Q2: Is the decay consistent with sector equilibration (total leakage saturating toward ~0.5 compresses the asymmetry), or does the odd sector genuinely overtake the even sector at depth?

Those depths (8, 12) were NOT part of the executed matrix; this campaign collects them. The 2026-07-21 data are design input here and are never re-used as confirmatory evidence for this campaign's endpoints.

Circuit Matrix

Field	Value
Device	IQM Garnet (20 qubits, square lattice) via Resonance
n	4
Circuit family	committed iqm_dla_pinned_n4_d{8,12}_{even,odd} builders (same code path as d4/6/10)
States	0011 even, 0001 odd
Depths	8, 12
Repetitions	4 per state/depth
Main shots	1024 per repetition → 4096 per state/depth
Readout states	0011, 0001, 0000, 1111 at 2048 shots (fresh block — new calibration window)
Layout	[2, 7, 12, 13] pinned (same as the powered block); fallback [9, 4, 3, 8] recorded before submission if unavailable

Circuit count: 2 × 2 × 4 = 16 main + 4 readout = 20 circuits. Shot count: 16,384 main + 8,192 readout = 24,576 shots.

Endpoints and Decision Rules (frozen)

Metric: relative asymmetry (leak_even − leak_odd) / leak_odd per depth, with per-depth one-sided two-proportion z-tests (4,096 shots per arm per depth; minimum detectable relative asymmetry ≈ ±0.076 at 90 % power for a baseline odd-arm leak near 0.42 — the committed Lane 1 secondary power).

• Primary (decay ordering): one-sided two-proportion comparison that the even-minus-odd leakage difference at depth 8 exceeds that at depth 12 (Δ₈ > Δ₁₂, α = 0.05, z-test on the difference of differences with binomial variances). This tests the monotone-decay reading of the executed profile on entirely new data.
• Secondary S1 (zero-crossing localisation): per-depth signs with Wilson 95 % intervals; the crossing is reported as the interval between the last significantly positive and first significantly non-positive depth across the JOINT profile {4, 6, 8, 10, 12}, with the cross-calibration-window caveat stated wherever the joint profile is shown (d4/6/10 are from 2026-07-21; d8/12 from this campaign).
• Secondary S2 (equilibration check): total leakage (leak_even + leak_odd)/2 per depth with Wilson intervals — reported descriptively against the saturation-toward-0.5 explanation.
• Repetition drift table as in the powered block.
• Readout handling: four-state exact-state parity-confusion correction, no full-matrix claim (identical to the powered block).

Either primary outcome is publishable: Δ₈ > Δ₁₂ supports monotone decay; failure to reject reports the bounded null at the committed power. No coherent-dynamics claim is available in any branch (the exact statevector baseline pins noiseless leakage at zero for every depth; committed module analysis/dla_parity_exact_baseline.py).

Live Readiness Gates (block submission until all pass)

• IQMFakeGarnet dry run of all 20 circuits from committed code only;
• transpiled depth envelope: linear interpolation/extrapolation of the executed ladder (d4→69, d6→99, d10→159, i.e. ≈ 15 layers per Trotter step) predicts ≈ 129 at d8 and ≈ 189 at d12; the frozen gate is the same +25 % rule applied to the prediction, so d8 ≤ 161 and d12 ≤ 236;
• pinned-layout calibration check on the day's live snapshot (all three chain edges calibrated), fallback substitution recorded before submission if required;
• explicit owner per-submit GO immediately before each block.

Budget and Stop Rules

• Expected cost by the observed 2026-07-21 batching pattern (~1 credit per job; 4 mains batch to one job per repetition, readout to one job): ~5 jobs ≈ 5 credits. Balance before submission read from the Resonance dashboard by the owner.
• Submit repetition 1 (4 mains + 4 readout) alone first; abort the campaign if that block consumes more than one quarter of the visible allowance.
• No submissions beyond this 20-circuit matrix without a fresh preregistration.

Claim Boundary

Blocked regardless of outcome: quantum advantage; coherence protection; coherent-dynamics parity claims; extrapolation beyond the sampled device, calibration window, layout, and depths; any modification of frozen submissions under paper/submissions/ (results extend the NEW manuscript submission_007 or its successor only). IQM and IQM Resonance are credited in every resulting output.

Submission Boundary

This preregistration is complete once committed. QPU execution stays blocked until the readiness gates pass and the owner grants a separate explicit per-submit GO.