A new mathematical study says a difficult threshold calculation in a class of quantum codes can, in an important special case, be reduced to a problem about the affine geometry of a classical word set. The result gives exact witness-induced lower bounds for selected nonadditive code examples and a formula for one broader family, while the larger Rains family still lacks a scalable algorithm.
The work is an arXiv preprint, version 2, dated 29 August 2026. It examines codeword-stabilized, or CWS, codes in standard form. Each code is specified by a graph state together with a classical word set.
A test aimed at coherence
The authors construct a codeword-coherence witness for a chosen coherent superposition of codewords. In plain terms, the calculation subtracts the population on each individual word and keeps the interference between different word states.
For superpositions with unequal weights, the method computes the threshold by enumerating affine intersections and affine-quadratic phase patterns. The authors describe the result as an exact fixed-parameter algorithm whose dependence is controlled by the number of CWS words.
The sharper simplification comes when every word has the same weight. The phase optimization then collapses, and the threshold depends only on the affine-intersection profile of the classical word set. In other words, the central stabilizer calculation is governed by how that set intersects the finite geometric structures used by the method.
A formula, and five exact certificates
For a binary affine-simplex union-stabilizer family, the reported equal-weight threshold is 3 divided by 2(n + 1), where n is the block length. The corresponding robustness-of-magic result is a lower bound of 2n divided by 3 for n at least 2. The robustness figure is reported as a lower bound supplied by the witness calculation.
Five listed standard nonadditive CWS examples receive exact rational witness-induced lower bounds. They are 10/3 for ((5, 6, 2)), 44/15 for ((9, 12, 3)), 17/7 for ((10, 18, 3)), 76/33 for ((10, 20, 3)), and 14/5 for ((7, 22, 2))SSW. These certificates apply to the listed examples, not automatically to every state in the corresponding code spaces.
The analysis also reports a structural split in the equal-weight setting. Linear word sets have no equal-weight magic under the stated mode, while nontrivial equal-weight magic is linked to nonlinear affine geometry in the word set. This finding is limited to the equal-weight mode described by the paper.
Designed for correlator measurements
The witness has a finite expansion in Pauli operators, determined by the graph, word set and amplitudes. A possible certification would therefore use Pauli correlators rather than full state tomography. It assumes calibrated Pauli measurements and is not a device-independent Bell certificate.
Small-system checks compute exact robustness of magic by solving a stabilizer-decomposition linear program for systems with at most three qubits. Those checks are limited to the small cases and do not validate scalability to large systems.
A global-depolarizing benchmark uses the ((5, 6, 2)) CWS state. In that calculation, the witness remains nontrivial when the depolarizing parameter is below 7/10, and the reported lower bound is 10/3 multiplied by 1 minus that parameter. The benchmark does not settle realistic noise performance, which the paper says requires separate analysis.
The larger family remains unresolved
The clearest unresolved point concerns the Rains family. Exact subset enumeration is not scalable there. The theorem identifies upper bounds on the family's affine-intersection profile as the missing input for scalable robustness-of-magic lower bounds, but it does not provide those bounds or a scalable algorithm.
More broadly, the certificate is tailored to the selected CWS superposition rather than an arbitrary state in the code space. The study therefore offers a route for certifying particular structured states, not a claim that every state in a CWS code space is magical.
The affine-profile data used for the listed examples are recorded in Appendix C 2. Reproduction code is planned for public release upon publication and is available from the corresponding author on reasonable request.
Paper data and sources
Original title: Affine-Profile Stabilizer Thresholds for Magic in Codeword-Stabilized Quantum Codes
Authors: Li-Yi Hsu, En-Jui Kuo
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text