Recovery gates as a resource in repetition-code process tomography: exhaustive gate-optimal designs for three, five and seven qubits
Abstract
Repetition-code process tomography compares the logical channel of a code with a reference channel through a decoder almost always the standard minimum-weight one. We ask how many physical single-qubit Pauli gates that decoder needs when the measurement basis is fixed---the constraint that makes the count a resource. Exhaustive enumeration at $n=3$, $5$ and $7$, each rank decided in exact arithmetic, gives minima of $0$, $1$ and $5$ gates, against $6$, $10$ and $14$ for first-order decoding and $6$, $50$ and $308$ for full decoding. The count depends only on how many sites are corrected, not which, so the problem splits into a budget and a placement; the placement decides identifiability, and more sites is not monotonically better: at $n=7$ full decoding has rank $24$ of $27$, and under a calibration-residual floor the $n=3$ precision curve is U-shaped. The three-gate design is $9.0$ times more precise than the standard six-gate decoder at the median of $40$ working points and better at every one of them, by $1.8$ to $88$ times; with the placement fixed in advance the factor falls to $6.9$, or to $3.1$ for the pre-commitment never worse, both medians over the same $40$ points. Because the full advantage needs per-point placement, we also certify two placements in parallel: they lose no direction either sees and, wherever either is identifiable and the calibration floor permits, reach a $10$\,mrad target within twice the shots. The optimal designs are not covariant under chain reversal.