Skip to content

SPDX-License-Identifier: AGPL-3.0-or-later

Commercial license available

(c) Concepts 1996-2026 Miroslav Sotek. All rights reserved.

(c) Code 2020-2026 Miroslav Sotek. All rights reserved.

ORCID: 0009-0009-3560-0851

Contact: www.anulum.li | protoscience@anulum.li

scpn-quantum-control -- IBM popcount-control manifest

IBM Popcount-Control Manifest -- 2026-05-05

This manifest preregisters the next QPU run. Its purpose is to test whether the observed n=4 DLA parity leakage asymmetry is primarily a parity-sector effect, primarily an excitation-count / amplitude-damping effect, or a combination of both.

Motivation

The promoted Phase 1 and Phase 2 A+G protocols compare:

Label State Parity Popcount
Original even 0011 even 2
Original odd 0001 odd 1

This leaves an excitation-count confound. Under a simple T1 amplitude-damping model, the higher-excitation state can leak faster independently of the DLA sector structure.

Submitted scope

Run only n=4 controls on the same XY-Trotter circuit family.

Depths:

d in {4, 6, 8, 10, 14, 20}

States:

Arm State Parity Popcount Purpose
E0 0011 even 2 Original even reference.
E1 0101 even 2 Within-even state swap at fixed popcount.
O0 0001 odd 1 Original odd reference.
O1 0010 odd 1 Within-odd state swap at fixed popcount.
O3 0111 odd 3 Excitation-inversion arm against E0.

Repetitions:

12 reps per (depth, state) point.

Shots:

4096 shots per circuit.

Circuit count:

6 depths * 5 states * 12 reps = 360 parity circuits.

Readout baseline:

0011, 0101, 0001, 0010, 0111 at 8192 shots each.

Total submitted circuits:

365

Estimated QPU use:

Using the recent B-C observed rate, 305 quantum seconds for 280 circuits, this run should cost about 398 quantum seconds, or 6.64 minutes, before backend variation. Abort if live transpilation substantially exceeds the precommitted depth budget.

Abort and budget rule

Before QPU submission, transpile every circuit on the selected live backend.

Abort before submission if:

  • live maximum transpiled depth exceeds 700;
  • live maximum gate count exceeds the recent B-C maximum by more than 50%;
  • IBM backend selection changes to a non-Heron backend;
  • any generated circuit has missing measurement or zero shots;
  • the account has less than 20 minutes of remaining QPU time.

Analysis plan

For each state and depth, compute parity leakage from raw counts:

L(d, state) = shots with parity != parity(initial) / total shots

Primary comparisons:

Comparison Null / falsification use
E0 vs O0 Reproduce the original parity-sector / popcount-correlated contrast.
E0 vs E1 Estimate within-even same-popcount state spread.
O0 vs O1 Estimate within-odd same-popcount state spread.
E0 vs O3 Excitation-inversion test; pure T1 expects O3 to leak at least as much as E0.

Decision criteria:

  • If E0 vs O0 remains positive but E0 vs O3 reverses or vanishes, the original result is likely dominated by excitation count.
  • If E0 vs O0 remains positive and E0 vs O3 also shows lower leakage for the odd state despite higher popcount, the DLA-sector interpretation is strengthened.
  • If E0 vs E1 or O0 vs O1 spreads are comparable to E0 vs O0, the original effect is state-preparation/layout sensitive and should be downgraded.
  • If all contrasts vanish, treat the prior result as calibration-window dependent until multi-day replication is available.

Promotion rule

Do not promote the run unless all of the following are committed:

  • raw count dictionaries;
  • IBM job IDs;
  • live transpilation depth and gate summaries;
  • SHA256 hash of the raw-count JSON;
  • reproducer script or existing reproducer update;
  • result summary with claim boundary;
  • paper update preserving the no-broad-advantage boundary.

Paper wording before this run is complete

Use conservative wording:

parity-sector and excitation-number correlated leakage asymmetry

Do not write:

DLA parity alone causes the asymmetry