This preprint bounds the action required to realize a specified four-level unitary when one entry of the control Hamiltonian is forbidden. The model allows arbitrary finite ancillary levels and measurable Hermitian controls. The action is the integrated spectral semidiameter. The endpoint error is measured in operator norm on all four code columns, including leakage, up to one common phase. An affine positive semidefinite condition characterizes the compressed paths and gives a realization on at most eight levels. Reflection of the symmetric target yields half-path certificates. Rational witnesses and interval arithmetic exclude four action–error pairs and establish a strict minimum-action interval 1.22442 < Λ* < 1.235092. Exact five- and six-level constructions are certified by pulse bounds and an analytic endpoint correction. A separate two-copy readout distinguishes hidden rank for a specified intermediate frame. The archive contains source, frozen witness data, pulse words and scripts for all checks. The results concern the declared mathematical model. The proposed two-copy protocol is prospective.
This preprint bounds the action required to realize a specified four-level unitary when one entry of the control Hamiltonian is forbidden. The model allows arbitrary finite ancillary levels and measurable Hermitian controls. The action is the integrated spectral semidiameter. The endpoint error is measured in operator norm on all four code columns, including leakage, up to one common phase. An affine positive semidefinite condition characterizes the compressed paths and gives a realization on at most eight levels. Reflection of the symmetric target yields half-path certificates. Rational witnesses and interval arithmetic exclude four action–error pairs and establish a strict minimum-action interval 1.22442 < Λ* < 1.235092. Exact five- and six-level constructions are certified by pulse bounds and an analytic endpoint correction. A separate two-copy readout distinguishes hidden rank for a specified intermediate frame. The archive contains source, frozen witness data, pulse words and scripts for all checks. The results concern the declared mathematical model. The proposed two-copy protocol is prospective.
This preprint studies a cycle-reflection operation on a supplied six-qubit realization of paired trigram configurations. A four-matched-pulse identity and a minimum within its stated serial grammar are separated from an ordinary-exchange endpoint construction. The raw full-source reflection and its full domain-safe alternative are different tasks: the latter is excluded by a collective-symmetry obstruction in the admitted source-only exchange model. Strict preservation of the addressable sector throughout the path is also excluded for the full task, while finite-action lower bounds quantify peak no-address population and time-integrated exposure. The paper presents an explicit low-excursion balanced word and a conditional joint clock–amplitude error certificate. Root settings and trajectory inputs remain tied to their supplied computer-assisted certificates. The second internal review reproduced all ten advertised numerical routes and twenty-four outputs using the original programs; this is same-program reproducibility, not independent implementation or external peer review. A separate Lean supplement covers the finite raw/safe operator algebra and the obstruction for arbitrary finite words of the declared algebraic factors. It does not formalize the numerical roots, matrix exponentials, continuous-time dynamics or resource theorems. No global pulse optimum, measured hardware feasibility, general noise tolerance, or automatic connection to the companion recovery code is claimed.
Can an encoded quantum state be controlled continuously without repeated measurements to prevent leakage? This study gives a constructive answer for a four-spin-2 AKLT plaquette, where a three-dimensional singlet code sits inside a 625-dimensional physical Hilbert space. The result combines an exact restriction with a coherent control scheme: strictly leakage-free two-site Hamiltonians cannot provide universal control of the full encoded qutrit, yet two explicitly constructed interactions enable universal single-qubit control without intermediate measurements or postselection within a two-dimensional subspace. The classification begins with the complete physical leakage map for all six spin pairs and exchange polynomials through degree four. Exact rational calculations establish rank L = 6 and dim ker L = 18 for the 255 × 24 matrix. Despite this large family of leakage-cancelling combinations, every accessible logical Hamiltonian obeys [A, R] = 0, where R is a fixed geometric involution. Its two-dimensional and one-dimensional eigenspaces therefore remain separate, obstructing universal SU(3) control. The restriction extends to arbitrary anisotropic two-site Hamiltonians that preserve the code. Within the rotationally invariant pair-polynomial model, interactions confined to the four plaquette edges act only as logical scalars; noncommuting protected controls require at least three pair supports and polynomial degree three. The constructive solution uses quadratic and cubic exchange Hamiltonians acting simultaneously on three pairs, including a diagonal link. Their coefficients cancel physical leakage exactly, so QH₂P = QH₃P = 0 throughout the full system, with P projecting onto the code and Q onto its complement. Their logical generators satisfy Lie(iA, iB) ≅ u(2) ⊕ u(1), allowing arbitrary coherent qubit gates while preserving the code at every intermediate time. The manuscript supplies exact certificates and verification procedures, together with operator-norm and calibration-error bounds that make the engineering requirements explicit. The practical opportunity is to remove intermediate code-check measurements from this control scheme; demonstrating a hardware advantage requires calibrated simultaneous interactions, including the diagonal channel, and measured gate fidelity and latency.
The spectral measure of a Hamiltonian and the POVM implemented by an apparatus need not coincide. We formulate observable attribution relative to an apparatus class defined by its preparations, controls, ancillary couplings, memory, and readable records. For a nondegenerate qutrit coupled to a binary threshold pointer and an amplifying environment, the explicit dynamics induces a classical observer channel that preserves only ground-versus-excited weight. We prove complete factorization of every allowed primitive through this channel and lift it to all finite adaptive protocols. Thus repetition, coherent controls within the excited sector, sector-only ancillas, calibration, and feed-forward cannot distinguish the two excited eigenstates. The full energy PVM is not attributable to the base class, and its optimal worst-case total-variation error is exactly one half. Adding a selective sector-changing control realizes the full PVM. Within the specified finite-protocol resource model, its exact minimum integrated operator-norm action is ; the optimal one-pulse approximate cost is also obtained in closed form. For imperfect devices we derive an exact minimax decoder for arbitrary calibrated binary responses, explicit error floors from record overlap, leakage, detuning, and dephasing, and a diamond-norm stability theorem for finite adaptive protocols. Observable attribution is therefore determined by physically available channels and resources, although the outcome statistics of every fixed experiment remain objective. This is a preprint version of a manuscript currently under journal review.
Consider exact scalar gate-sequence data consisting of one preparation, one binary effect, and the probability of that effect after every word over a fixed two-gate alphabet. This preprint constructs a single word law of exact noncommutative Hankel rank four that admits a control-irreducible unitary quantum realization on C^d for every finite d >= 2. In each realization the state is full rank, the binary effect is strict, and the two controls generate the full matrix algebra M_d(C), excluding direct-sum padding and invariant spectator sectors. The associated four-dimensional angular-momentum operator systems have the same first-level Lorentz cone in every dimension, while a level-2 matrix pencil is positive exactly when t >= (d+1)/(d-1). Thus matrix-level-two order distinguishes all members of the constructed family. The result is a scoped non-identifiability theorem for the Hilbert-space dimension of an arbitrary compatible irreducible realization. It does not claim that the minimum compatible dimension is unidentifiable, does not contradict informationally complete gate set tomography within a declared model dimension, and does not reconstruct a tensor-product structure. The record includes the preprint PDF, LaTeX source, bibliography, numerical verification script and output, citation metadata, and a targeted priority audit.
A symmetry specifies which interactions are compatible with a system; it does not specify which interactions its controls can produce. We resolve this distinction for a finite recognition apparatus with a main signal, a retained record and an independent clock. An explicit unitary change of coordinates turns the two signal ports into four equivalent copies of a four-dimensional symplectic representation. The invariant source Hamiltonians are therefore exactly the Hermitian 4×4 coefficient matrices acting between those copies. The original controls preserve one decoded record component, excluding even a simple exchange allowed by the symmetry. Adding forward waits under one fixed positive-strength recorded-exchange Hamiltonian removes this obstruction: every invariant source unitary, with identity on the clock, lies in the operator-norm closure of the resulting finite programs. The proof derives the required operations from the stated inventory: a clock echo removes unwanted clock motion, and a record-restoring subroutine supplies control of the main register. We also carry the original energy readout through the same programs, retaining its reference and positivity conditions. The deposit includes the paper and its LaTeX source with six vector schematics. AI-use disclosure: OpenAI Codex assisted with mathematical formalization, manuscript drafting and editing, schematic code, and source and layout checks. The results are supported by explicit mathematical arguments and machine-checked proofs in the author's research project.
We study what can be inferred from a coherent two-copy return on a qubit code. An antisymmetric input reduces the return operator to its determinant. A fixed measurement recovers the mean complex amplitude and the joint code-return probability while retaining loss. For diagonal masks whose single-pass ensemble is a complete dephasing, we derive a dimension-dependent lower bound on incorrect-pair return. A four-port tetrahedral construction attains the bound 1/3 on every incorrect pair and every input. We construct exact counterexamples to recovery of coherence from excitation counts and to calibration without a loss standard. An independent one-copy measurement, implemented either with a spectator qubit or four unentangled inputs, quantifies the deviation of the successful operation from a target isometry. The accompanying matrix calculations, circuit replays and prospective statistical tests specify how to distinguish the modeled alternatives. The results concern declared finite-dimensional models. No laboratory data are used.