Fluid-Dynamic Stability Filtering for Zero-Noise Extrapolation in NISQ-Era Quantum Circuits

The Zenodo description field has no hard character limit — it accepts the full abstract. However it's a good opportunity to have a slightly expanded version that includes the key results, since Zenodo descriptions are indexed by Google Scholar and other academic search engines. Use the full abstract from the paper — it's actually ideal length for Zenodo. Just paste it in plain text (no LaTeX commands). Here's a clean plain-text version ready to paste: Zero-Noise Extrapolation (ZNE) is among the most widely deployed quantum error mitigation (QEM) techniques for near-term quantum devices. However, the choice of noise scaling schedule remains largely empirical, and no standard method exists for validating whether a given ZNE extrapolation is physically reliable or has diverged into an artifact. We introduce two contributions: (i) a fluid-dynamic stability filter motivated by the structure of the Navier-Stokes energy dissipation norm, defining ε = ν∫||∇u||²dΩ and flagging a ZNE output as unreliable when ε ≥ Γ, as a post-hoc reliability criterion; (ii) a family of structured noise scaling schedules — Fibonacci (τₙ = τₙ₋₁ + τₙ₋₂), Lucas, and prime-anchored — as alternatives to conventional linear or odd-integer scaling, evaluated against all four standard extrapolants (linear, Richardson, poly-2, poly-3). We evaluate both across circuit depths d∈{2,4,6,8,10,12,16,20}, depolarizing noise levels p∈{0.001–0.2}, and qubit counts n∈{2,4,6} using shot-based density matrix simulation in Cirq, totalling 153,600 experiment runs. The stability filter achieves 0 flags in 14,400 independent trials at p≤0.01, d≤4, giving a 95% Clopper-Pearson upper bound of p_flag < 0.00022. Hardware validation on ibm_fez (Heron r2) and ibm_torino (Heron r1) confirms zero flags in 120 independent trials (24 conditions × 5 trials) across all four circuit types and d∈{2,4,6,8}, with maximum observed ε=0.00285 (5.7% of Γ=0.05). Fibonacci produces the lowest mean ε across all circuit types. Linear extrapolation produces the lowest and most consistent ZNE variance; Lucas achieves the lowest variance on random and QAOA circuits. Code and data are available at GitHub (github.com/OfficialChaos/Quantum-Vortex-QEM) and archived at Zenodo (DOI: 10.5281/zenodo.18827720).