Robust Gauge Reduction under Imperfect Quantum Controls

Abstract

This is a separate preprint record that develops the robustness layer of the outcome-resolved mid-circuit-measurement gauge analysis. It uses the exact complete-positivity benchmark of the preceding preprint as an input, but its main result concerns the composition of imperfect-control uncertainty, residual-gauge conditioning, and instrument-model error. The general ingredients—gate-set-tomography gauge freedom, commutant geometry, perturbation by singular-value gaps, quantum-instrument physicality, and mid-circuit-measurement learnability—are prior art. The paper does not claim a new general GST gauge principle, a dimension-free tomography theorem, or experimental validation. Related identifier ● Identifier: 10.5281/zenodo.23043202 ● Relation: Is derived from (isDerivedFrom) ● Resource type: Publication -> Preprint ● Title: Outcome-Resolved Mid-Circuit Measurements as Exact Gauge Constraints: Complete-Positivity Windows, Correlation Certificates, and Robustness Limits Description / Abstract Self-consistent quantum characterization is invariant under similarity transformations, so an experimentally fitted gate set is generally identified only up to gauge. Exact trusted controls can reduce this ambiguity, but realistic controls are imperfect; simultaneously, a mid-circuit measurement (MCM) instrument is itself only approximately described by a model. These two uncertainties act at different logical stages and should not be merged before the gauge geometry is reconstructed. This preprint gives an explicit finite-dimensional stability analysis. For a finite reference control family ({R_i}), define the commutator map (\mathcal C(B)=([B,R_i])_i) and let (\gamma) be its smallest positive singular value. If two exactly gauge-equivalent implemented gate families ({G_i}) and ({G’_i}) both lie within Frobenius errors (\delta_i) of the same references, then every trace-preserving gauge matrix (B) lies within [\operatorname{dist}{F}(B,\ker\mathcal C)\le \frac{2\Delta}{\gamma}|B|{\mathrm{op}},\qquad\Delta=\left(\sum_i\delta_i^2\right)^{1/2}.] For the two-qubit control set (H\otimes I), (S\otimes I), (I\otimes H), (I\otimes S), and CNOT, the exact commutant is (\operatorname{span}{\mathcal D,\mathrm{id}-\mathcal D}), so the residual scalar gauge is (B_c=\mathcal D+c(\mathrm{id}-\mathcal D)). In the declared unweighted Frobenius convention, direct computation gives (\gamma=0.830129650657\ldots). The main result closes two previously separate robustness budgets. Suppose an independent physical or calibration argument places the scalar projection (c) in a positive prior interval ([\ell,u]). If (\alpha=2\Delta/\gamma<1), a self-normalized estimate removes the otherwise unknown factor (|B|): [|B-B_c|_F\le \bar e=\frac{\alpha M_0}{1-\alpha},\qquad M_0=\max(1,u).] If (\bar e<m_0=\min(1,\ell)), the condition number of (B) is explicitly bounded. Composing this control-side bound with a Choi-Frobenius instrument-model error (\varepsilon) yields a closed outer complete-positivity condition for the symmetric noisy Lüders MCM benchmark, [(1-\eta)|1-2p|,|c-c^{-1}|\le \eta+2d\rho,] where (\rho) is an explicit function of (\Delta,\gamma,\ell,u), and (\varepsilon). Hence the imperfect-control problem again reduces to an explicit scalar gauge interval. The result separates transverse control uncertainty from instrument-model uncertainty and identifies the loss of robustness as the prior gauge envelope approaches singularity.