slug: bytsko-2009-positive-coupling-real-spectrum-refutation
bibkey: bytsko2011nonhermitian
doi: 10.1090/S1061-0022-2011-01148-3
url: https://arxiv.org/abs/0911.4476v3
triage: theorem
motivation_gids:
  - D5/S3/Quantum/SpinChains/QuantumGroupChainRealSpectrumRefutation.result

Positive couplings do not force a real spectrum for U_q(sl_2)-invariant spin chains

Problem

A. G. Bytsko, Non-Hermitian spin chains with inhomogeneous coupling, arXiv:0911.4476v3 (St. Petersburg Math. J. 22, 393–410, 2011), Conjecture 3:

For $|\gamma |<\frac{\pi }{2S}$, the Hamiltonian (22) of a spin chain with inhomogeneous coupling has a real spectrum if all $a_n>0$.

Here:

• The chain: $N$ sites, each carrying the spin-$S$ representation of $U_q(s{l}_2)$ with $q=e^{i\gamma }$, in the source’s normalization.
• The Hamiltonian (22): $H={\sum }_{n=1}^{N-1}{a}_n{P}_{n,n+1}^{S,s}$, where $P^{S,s}$ is the non-orthogonal projector onto the spin-$s$ summand of $V^S\otimes V^S$, given by the source’s Casimir polynomial.

The definitions are quoted in the literature note.

Issue #20040 fixes the reading: for every $S$, $s\in \{0,\dots ,2S\}$, $N\ge 2$, $0<|\gamma |<\pi /(2S)$ away from projector poles, and positive couplings, every eigenvalue of $H$ is real. This is refuted.

Motivation

For $|q|=1$ the $U_q(s{l}_2)$-invariant chains are non-Hermitian but pseudo-Hermitian, and a real spectrum is the condition under which they define a unitary quantum theory with a modified inner product. The source constructs universal metric operators for these chains and determines the ranges of $\gamma$ with real spectra for short chains; Conjecture 3 would guarantee physical reality of the spectrum throughout the standard range whenever the couplings are positive.

Gap

The literature check before any Lean is recorded in #20040:

• the five works citing the paper on INSPIRE cite it only for non-Hermitian chains, quasi-Hermiticity or projector examples;
• the author’s later papers do not revisit the spectrum of (22);
• the source’s own $N=4$ evidence covers only $s=0$ (the subsection “$N=4$ and $N=5$ for $s=0$”);
• the repository had no record.

not-found-in-searched-scope.

Route

Take $S=s=1$, $N=4$, $\gamma =\arccos (4/5)$ (so $q=(4+3i)/5$ and $2\gamma <\pi$), and $a=(1,1,1)$.

1. The projector. $14P^{1,1}$ is an explicit Gaussian-integer $9\times 9$ matrix.
2. An invariant plane. An explicit $81\times 2$ matrix $W$, supported on nine of the ten configurations of total weight $+2$, satisfies $(P_{12}+P_{34})W=WA$ and $P_{23}W=WB$ with $A=\left(\begin{array}{cc}-11/14& 1\\ -275/196& 25/14\end{array}\right)$ and $B=\left(\begin{array}{cc}1& 0\\ 0& 0\end{array}\right)$.
3. A non-real eigenvalue. $A+B$ has characteristic polynomial ${\lambda }^2-2\lambda +25/14$, with roots $1\pm i\sqrt{11/14}$. For an eigenvector $y$ of $A+B$, $Wy$ is a nonzero eigenvector of $H$ with a non-real eigenvalue.

Falsifier

The counterexample would not answer the conjecture if the module’s chain differed from the source’s. The module:

• defines $E,F,K$ and the projectors $P^{S,s}$ by the source’s formulas, and the tensor Casimir as $({\pi }_S\otimes {\pi }_S)\Delta (C)$ from Eqs. (C) and (DelEFK), the definitional left side of Eq. (dC2); the expanded right side of (dC2) does not commute with $\Delta (E)$ and is read as a misprint ([#20040](https://github.com/the-omega-institute/trureturing/issues/20040));
• defines $H$ by the source’s equation (22);
• places the witness inside the conjecture’s hypotheses: $2S|\gamma |<\pi$, $\gamma \ne 0$, nonzero projector denominators and positive couplings.

Evidence

The canonical source is D5/S3/Quantum/SpinChains/QuantumGroupChainRealSpectrumRefutation.lean (425 lines).

• Proof. It proves result : ¬ claim.
• Axioms. It uses only the standard axioms propext, Classical.choice and Quot.sound; no sorry, native_decide, or new axiom.

Frozen state: Golden/Frozen/state/D5/S3/Quantum/SpinChains/QuantumGroupChainRealSpectrumRefutation.lean.json, statement sha256:c24d3f502c28788ed74f556becefd1e53c201e8c406d76d980355f524ef56f73.

Declaration statements:

• result: sha256:3b91017ebba5ac1c974f029ba65608f932e05ad87426abfaa6909087aef0aba3;
• claim: sha256:d4e384aff69fa671d8b0d7a7d57950de08e2a1bd63e1f15a54e0c5eed4331dc8.

Triage

Tier 1 conjecture of a 2009 preprint kept in the journal versions, preregistered in issue #20040 before any Lean. theorem; resolution refuted.

The module is admitted under the open-problem-resolution basis of CLAUDE.md §3.2; the per-declaration proof shapes are listed in the delivery PR. utility: kind=certified-instance; basis=refutes. There is no digestion atom.

What the settlement shows

Proved in the module: the homogeneous spin-1 chain of four sites with the spin-1 projector at $q=(4+3i)/5$ has the non-real eigenvalue $1+i\sqrt{11/14}$, inside the range $|\gamma |<\pi /2$.

Computed, not formalized (orchestrator, numpy, recorded in #20040):

• A cone of couplings. On the invariant plane, $H(a,b)=a(P_{12}+P_{34})+b{P}_{23}$ has characteristic polynomial ${\lambda }^2-(a+b)\lambda +\frac{25}{14}ab$, so the spectrum is non-real whenever $(18-5\sqrt{11})/7<b/a<(18+5\sqrt{11})/7$; for $a=(1,2,1)$ the eigenvalue $(21+i\sqrt{259})/14$ appears.
• Other projectors. The spin-2 projector at the same $\gamma$ also gives a non-real spectrum; the spin-0 projector gives real spectra, consistent with the source’s $s=0$ analysis.
• Dependence on $\gamma$. For $s=1$ and $a=(1,1,1)$, non-real eigenvalues occur for $\gamma$ between about $0.58$ and $1.12$, and the spectrum is real near $0$ and near $\pi /2$.

Argued, not formalized.

• Mechanism. On the invariant plane, $P_{12}+P_{34}$ and $P_{23}$ are two oblique rank-one idempotents with negative overlap $\mathrm{tr}(AB)=-11/14$. For two rank-one idempotents, $\det (\lambda -aA-bB)={\lambda }^2-(a+b)\lambda +ab(1-\mathrm{tr}AB)$, whose discriminant $(a-b{)}^2+4ab\mathrm{tr}AB$ is nonnegative for all positive $a,b$ exactly when $\mathrm{tr}AB\ge 0$. The non-orthogonality of the $q$-projectors makes the overlap negative and opens a complex pair for a range of coupling ratios.
• Scope. The conjecture fails for $s=1$ and $s=2$ at $S=1$. Whether it holds for $s=0$, the case the source analyzed for $N=4,5$, is not settled here.

ASSUMED-UNVERIFIED

The bounded literature check does not establish exhaustive worldwide novelty or priority. The statements in “Computed, not formalized” and “Argued, not formalized” are not formalized; the eigenvalue range in $\gamma$ is a numerical scan.