Controlled Hardware Tests on ibm_fez (September 2026): Null Result for the Tetrahedral RZ Correction, and ZZ-Aware Staggered Dynamical Decoupling with Genetic-Algorithm Evolution

Read full text
Authors: Devin Phillip Davis
Cited by: 2

Abstract

Controlled hardware tests on ibm_fez, 20 September 2026 Two experiments, four IBM Quantum jobs, 227 s of QPU time, all raw counts included. 1. Tetrahedral RZ "correction" — null result (`tetrahedral_correction_test/`) Tests the claim (10.5281/zenodo.18450507, README of `quantum-advantage/tetrahedral-correction`) that inserting RZ(+δ) on the target and RZ(−δ·χ) on the control after every CX, with δ = 0.050493 rad and χ = 0.946, raises GHZ fidelity by +3.1 % (N=8) and +7.9 % (N=12). Design: same backend as the claim, one calibration-selected linear chain per N, identical 2q-gate counts, five conditions (standard, +δ, −δ, two random δ'), 3 interleaved repeats, proper GHZ fidelity F = ½(P_pop + parity-oscillation amplitude over 2N+2 phases). Result (job `dannuhgpqrnc73991ntg`): ΔF(+δ) = +0.004 ± 0.002 at N=8 and +0.002 ± 0.006 at N=12, indistinguishable from the sign-flipped and random-angle controls; 8σ below the claimed effect at N=12. The claim is not supported. Baseline GHZ fidelities on the chain: F₈ = 0.795, F₁₂ = 0.667. 2. ZZ-aware staggered dynamical decoupling, with GA evolution (`dd_evolution/`) Preserve |+⟩^⊗8 on chain [89,90,91,98,111,110,109,118] through idle windows of 8/16/32 µs. Jobs `dano3qtr85ps73feb46g`, `dano4gopqrnc73991uk0`, `dano5dtr85ps73feb5og` (gen 0/1/2; 4 interleaved replicates × 1024 shots per sequence and window). Result: simultaneous CPMG/XY4 collapse to ≈0.59 at 32 µs; XY4×2 with the odd sublattice offset by half a slot holds 0.946 / 0.888 / 0.795 (8/16/32 µs), reproduced within one point across the three jobs. A genetic algorithm evolved on a quasi-static surrogate and then bred from hardware rankings reproduced the stagger but did not beat the textbook baseline (best 0.933 / 0.875 / 0.785). Gen 1 includes four sequences that fail by construction (odd Y-parity) — a pipeline error, kept for completeness. Both experiments are reproducible with the included scripts and an IBM Quantum account. Author: Devin Phillip Davis, Agile Defense Systems LLC. License: CC-BY-4.0.

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
20 September 2026
Document type
Dataset