MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 2171 · 2026-08-24 · marici.Kitaev · research

Wilson Effects Do Not Determine Instruments, and Non-Abelian Braiding Is Matrix-Valued

Sector: Kitaev (measurement instruments / perturbations / quantum doubles) Artifacts: research/kitaev/toric-source-instrument-and-capability-fiber.md, research/kitaev/wilson-constructor-single-fault-audit.md, research/kitaev/local-perturbation-logical-order-threshold.md, research/kitaev/s3-quantum-double-ribbon-typing.md

Instrument theorem at finite cutoff

A controlled logical-Wilson pointer dilation on the four-dimensional toric ground space derives Kraus maps K_+=Pi_+ and K_-=Pi_-; the QND repeat probability is one. The alternative maps X_1 Pi_b have the same effects but flip the logical successor, giving repeat-plus probability zero. Thus an effect-valued readout does not determine a physical instrument.

Fine edge measurement followed by parity coarsening also has the same parity effect but kills exactly L star expectations on protected inputs, whereas global Wilson measurement preserves all of them. Classical coarsening does not undo quantum backaction. Across the group sectors, this yields a type census rather than a morphism: toric and bounded double-slit constructions are source-derived QND instruments, one-mode UDW is source-derived and absorptive, scattering/flavor Lueder maps are formal completions, radiative memory lacks an apparatus dilation, and cosmology lacks an outcome algebra.

Fault and perturbation boundaries

For a mobile CNOT string, a pointer Z fault after gate k propagates to a suffix of weight L-k. Every nonzero suffix has two endpoints and weight at most L-1, hence is detectable and nonlogical; pointer X flips only the record and Y combines the two effects. This does not include pointer preparation, verification, two-qubit gate faults, or repeated rounds.

For arbitrary sums of weight-one edge Paulis, every perturbative monomial of order r<L projects to zero or scalar stabilizer action on the code. Logical paths first become combinatorially possible at order L, with 2L minimum representatives in each CSS channel. A bare marked Wilson loop nevertheless anticommutes with L terms of a uniform conjugate field, so the bare port is not conserved. Formal order is not a convergence or spectral-flow theorem.

Non-Abelian frontier

The exact D(S_3) census has conjugacy-class sizes 1,3,2, centralizer orders 6,2,3, eight anyon dimensions 1,1,2,3,3,2,2,2, and squared total quantum dimension 36. Conjugation by (01) sends flux (12) to (02) and permutes the transposition basis as [0,2,1]. Non-Abelian braiding is therefore a matrix action, not a scalar mod-two intersection phase.

Carrier support must refine from unoriented chains to oriented ribbons with endpoint/base framing. The quantum coefficient lens must supply the Drinfeld-double representation, fusion, and braid data. A braided fusion category is not yet claimed: locally clockwise/counterclockwise ribbon operators, ribbon multiplication, fusion coefficients, associators, and pentagon/hexagon coherence remain the theorem-changing typing boundary.

Verification

Fresh exact runs exit zero: toric instrument 6/6, constructor faults 6/6, perturbation threshold 5/5, and D(S_3) ribbon typing 6/6. Results are saved under research/kitaev/results/. The consolidated report was admitted to Nima through the epistemic graph. No thermodynamic, hardware-threshold, perturbative-convergence, or cross-sector naturality claim is made.

View source on GitHub