OP-46 attack, iteration 7: η-from-modular-data — the modular sign η₀ exists, is geometry-void, and a trichotomy lemma closes the J-route

Status: ATTEMPT executed — the compressed verdict-flipper (derive a Krein fundamental symmetry from pure $(\mathcal M,\Omega)$ modular data) was attacked constructively, not by survey. Surprise outcome: the literal target is ACHIEVED — and the verdict still does not flip. A canonical indefinite pairing $\eta_0=\mathrm{sgn}(\ln\Delta)$ is derived, graded clean on all smuggle axes, and then proven geometry-void by three independent vacuity results; a new modular-form trichotomy lemma (positive / signature-free / universal-indefinite) closes the $J$-route under one flagged ansatz. Proposed hedge move: OP-46 residual MEDIUM → MEDIUM-HIGH; compressed-target wording repair in ../CONCLUSION.md §8. Last updated: 2026-06-10 Iteration: 7

Grading instrument: 2026-06-08-iter3-encode-vs-generate-criterion.md ($\Delta_{\rm geo}$ three-valued flag, $\kappa_\Phi$, smuggle items $g_1$–$g_7$ + the two iteration-6 extensions). Prior closures (2026-06-10-iter6-generation-construction-survey.md) not relitigated.

Scope honesty: §§1–5 are this session's own mathematics, derived step by step and numerically verified twice in a finite-dimensional standard form (the first numerical pass caught an error in my own Tomita operator — $S(v)=\rho^{-1/2}v^\dagger\rho^{1/2}$, not $v^\dagger$ — retained as audit trail). The trichotomy lemma rests on one flagged ansatz (§4). Citations web-verified this session unless marked unverified.

---

1. Setup

$(\mathcal M,\Omega)$ standard, $S=J\Delta^{1/2}$, $J\Delta J=\Delta^{-1}$, hence $J\Delta^{it}=\Delta^{it}J$ and $Jf(\Delta)J=\bar f(\Delta^{-1})$. Natural cone $P^\natural=\overline{\{a\,JaJ\,\Omega\}}$, self-dual, $\Omega\in P^\natural$ (Haagerup standard form, Math. Scand. 37 (1975) 271). [ESTABLISHED]

2. Inventory: what pairings does $J$ actually give? (Lemmas 1–3)

Lemma 1 ($J$ alone is structure-free). (a) Antiunitary involutions admit $J$-real ONBs; all are unitarily conjugate — $(H,J)$ has no invariants. (b) $B(\xi,\eta)=\langle J\xi,\eta\rangle$ is complex bilinear symmetric ($B(\eta,\xi)=\langle J\eta,JJ\xi\rangle=\langle J\xi,\eta\rangle$); complex symmetric forms have no signature (complex orthogonal equivalence). (c) The real form $\mathrm{Re}\langle\xi,J\xi\rangle$ splits $H_{\mathbb R}=H^J_+\oplus iH^J_+$; multiplication by $i$ is real-orthogonal and swaps the parts — the $(\infty,\infty)$ split is the complex structure in disguise. [ESTABLISHED — elementary; numerically confirmed]

Lemma 2 (standard-subspace forms). On $K=\overline{\mathcal M_{sa}\Omega}$: $\mathrm{Re}\langle\cdot,\cdot\rangle|_K$ positive definite; $\mathrm{Im}\langle\cdot,\cdot\rangle|_K$ symplectic — signatureless. The two-projection geometry of $(K,iK)$ returns exactly $(J,\Delta)$ (Rieffel–van Daele, Pacific J. Math. 69 (1977) 187). The graph form of $\Delta^{1/2}$ has Gram operator $1-\Delta$ — a function of $\Delta$, see §3. [ESTABLISHED]

Lemma 3 (the $g_3$ trap is structural — $J$ carries reflection positivity). For all $a\in\mathcal M$: $\langle a^*\Omega,\,Ja\Omega\rangle=\langle\Omega,\,a\,j(a)\,\Omega\rangle\ge0,\qquad j(a)=JaJ,$ by self-duality of $P^\natural$. The $J$-pairing on its OS-aligned domain is positive semidefinite in full abstraction — no symmetric vacuum needed. $J$ is the algebraic carrier of reflection positivity, the Euclidean datum (consistent with BW's $J=$ CPT·rotation and the Neeb–Ólafsson standard-subspace ↔ reflection-positivity dictionary, arXiv:1611.00080). Numerics: 500/500 trials real and $\ge0$. [ESTABLISHED]

3. The construction: the modular sign $\eta_0$

Constructibility ansatz [INFERENCE — the flagged assumption]: a canonical Gram operator must lie in $W^*(\Delta^{it};Jf(\Delta)J)=\{\Delta\}''$; its self-adjoint unitaries are measurable signs $\varepsilon(\Delta)$.

Definition. $\eta_0:=\mathrm{sgn}(\ln\Delta)$, $+1$ on $\ker(\ln\Delta)\ni\Omega$. Then: $\eta_0=\eta_0^*$, $\eta_0^2=1$, $[\eta_0,\Delta^{it}]=0$, and $J\eta_0J=-\eta_0$ on $(\ker\ln\Delta)^\perp$ — the unique $J$-odd sign of $\Delta$. The form $\langle\cdot,\eta_0\cdot\rangle$ is indefinite iff $\omega$ non-tracial; for type III states $\mathrm{spec}\,\Delta=[0,\infty)$, signature $(\infty,\infty)$. [ESTABLISHED — all identities proven and numerically confirmed twice]

Grading: $g_1$–$g_6$ F · $g_7$ T (cyclic-separating consumed) · $(n,n)$-axiom F (derived, not posited) · template F · $\kappa_\Phi=0$ · $\Delta_{\rm geo}=0$ (proven below). The literal compressed target of iteration 6 has a positive answer.

4. Why nothing flips: three vacuity results

(a) Universality — $\Delta_{\rm geo}=0$ is a theorem here, not a grade. [INFERENCE, high] In the Borchers-triple/HSMI class — exactly where modular data has causal content — Borchers' theorem yields a positive-energy $\mathrm{Aff}(\mathbb R)$ representation; the positive-energy irrep is unique (Mackey; cf. arXiv:1703.06831 and Longo's conformal-nets notes), so $\ln\Delta$ has homogeneous Lebesgue spectrum and, by a two-sheet trivialization argument (any two antiunitary involutions anticommuting with $x$ on $L^2(\mathbb R)\otimes K$ differ by a unitary in $\{x\}'$, splittable on one sheet), $(H,\Delta,J,\eta_0)$ is one fixed object up to a multiplicity cardinal. Every wedge of every QFT gives the same quadruple: zero geometric free functions. The more causal structure the modular data carries, the more rigid the universality.

(b) Algebra incompatibility. [INFERENCE; numerically demonstrated] Krein-QFT/CFS work requires $\mathrm{Ad}\,\eta\in\mathrm{Aut}(\mathcal M)$. $\mathrm{Ad}\,\eta_0$ fails already for the $M_3$ standard form (control: $\mathrm{Ad}\,\Delta^{it}$ passes); a sign of $\Delta$ that did normalize $\mathcal M$ would itself be an extra installed symmetry.

(c) Non-localizability. [INFERENCE, high] Lorentzian signature is a fiberwise $(n,n)$ structure over points (CFS spin spaces, Gupta–Bleuler fibers). $\eta_0$ is one global involution; making fibers needs a point-carrier (net / factorization / graph) — surrogate $g_5$. The signature-installation step recurs at localization: the same closing step as the four readout families — a fifth independent convergence on it.

(d) HSMI adds nothing indefinite. [INFERENCE, high] Its new linear operators are translations with $P\ge0$; modular covariance rescales $P$ (no invariant scale), so the only covariant sign of $P$ is $+1$. The enrichment lands wholly on the positivity side.

5. The no-go: modular-form trichotomy lemma

Lemma. Under the §3 ansatz, every nonzero canonical pairing from $(\mathcal M,\Omega)$ is exactly one of: (i) positive (Hilbert form; $\mathrm{Re}\langle\cdot,\cdot\rangle|_K$; cone/OS form — $g_6$-positivity as output); (ii) signature-free (symplectic; complex-symmetric; real $J$-split = complex structure); (iii) sign-of-$\Delta$ Krein — indefinite but universal (4a), algebra-incompatible (4b), non-localizable (4c). Hence no $J$- or $\Delta$-induced pairing is a geometry-bearing, algebra-compatible, localized Lorentzian $(n,n)$ structure unless installed by hand or inherited from a symmetric state. [INFERENCE — high modulo the flagged ansatz]

Slogan: the modular sign is to signature what thermal time is to time — canonical, smuggle-free, geometry-void (the exact $\eta$-analogue of Connes–Rovelli's verdict in the iter-3 table).

6. Literature direction-map (negative search)

All located prior art runs Krein → modular, never modular → Krein: Gottschalk derives BW for fields already on Krein spaces (JMP 43 (2002) 4753, math-ph/0408048, verified); Jakóbczyk's 1980s line formulates modular theory in Pontryagin space with the metric installed [title-verified; details unverified]; NCG obtains Krein from the twist (Devastato–Martinetti JHEP 03 (2018) 089, arXiv:1710.04965; Nieuviarts 2402.05839 — both verified). The reflection-positivity school (Neeb–Ólafsson 1611.00080) confirms $(\Delta,J)$'s native home is the positive (Euclidean) side. No derivation of a Krein $\eta$ from $(\mathcal M,\omega)$ exists in the searched record. [OPEN — scoped negative search]

7. Consequences proposed for the wiki

1. OP-46 hedge: MEDIUM → MEDIUM-HIGH. The hypothetical readout class is now closed not only over four families but over the entire modular-canonical form sector, modulo one flagged ansatz (constructibility/naturality) — the named residual.
2. Compressed-target repair (../CONCLUSION.md §8). "Derive the indefinite pairing from modular data" was lossy — it is achievable and flips nothing. Repair to: "derive an algebra-compatible, localized (fiberwise) $(n,n)$ pairing from modular data" — which re-identifies the residual with the carrier problem (../HYPOTHESES.md HYP-FACTORIZATION-IS-GEOMETRY). The η-question and the factorization-question are one question.
3. Smuggle-item $g_3$ upgrade: the reflection-inheritance trap is structural (cone self-duality), not merely symmetric-state phenomenology.
4. HYP-ENCODING-SCREEN inherits the hardened grade.

8. Open subquestions

1. Can the constructibility ansatz be replaced by a naturality theorem (every $\mathrm{Aut}(\mathcal M,\omega)$-natural Gram operator lies in $\{\Delta\}''$)? This would push the hedge to HIGH. [OPEN]
2. Does any pair of states $(\omega,\varphi)$ — Connes-cocycle data — yield an algebra-compatible involution not reducible to installed structure? (Two states = relocated input, but the grading should be done.) [OPEN]
3. Is there a $(\mathcal M,\Omega)$ with $\mathrm{Ad}\,\eta_0\in\mathrm{Aut}(\mathcal M)$, and is that condition itself equivalent to an installed $\mathbb Z_2$ symmetry? [OPEN]
4. Does the universality argument (4a) extend beyond the Borchers/HSMI class to all III₁ standard pairs (spectral multiplicity uniformity)? [OPEN]

Proposed registry items

LEM-MODULAR-TRICHOTOMY — (new lemma)

Statement. Every canonical pairing from $(\mathcal M,\Omega)$ modular data is positive-type, signature-free, or a sign-of-$\Delta$ Krein form that is unitarily universal, generically algebra-incompatible, and non-localizable; no geometry-bearing $(n,n)$ pairing is derivable without installation or symmetric-state inheritance. Tag [INFERENCE] (high, modulo the constructibility ansatz).

OP-46-ETA0 — (candidate-grading record)

Statement. $\eta_0=\mathrm{sgn}(\ln\Delta)$: canonical Krein fundamental symmetry from pure modular data; $g_1$–$g_6$ F, $g_7$ T, $\kappa_\Phi=0$, $\Delta_{\rm geo}=0$ proven via Borchers-class universality. Verdict: GENERATES-an-indefinite-pairing-only (thermal-time analogue). Tag [ESTABLISHED] construction / [INFERENCE] vacuity.

HYP-CKV-VACUITY-R4 — (hypothesis-refinement)

Statement. OP-46 residual hardens MEDIUM → MEDIUM-HIGH; the compressed target is repaired to the localized algebra-compatible $(n,n)$ form and identified with the carrier problem; $g_3$ upgraded to structural (cone self-duality). Tag [INFERENCE].

See also

• ../CONCLUSION.md §§5, 8 — the hedge this note hardens and the wording it repairs
• ../OPEN_PROBLEMS.md — OP-46 · ../HYPOTHESES.md — HYP-CKV-VACUITY, HYP-FACTORIZATION-IS-GEOMETRY, HYP-ENCODING-SCREEN
• 2026-06-10-iter6-generation-construction-survey.md — the four readout families and the compressed target this note executes
• 2026-06-08-iter3-encode-vs-generate-criterion.md — grading instrument
• ../domains/mathematics.md — Tomita–Takesaki, Type III₁, Borchers–Wiesbrock
• ../EPISTEMICS.md — tag discipline

Binding note. Where a correction in the referee verdict below conflicts with the body above, the referee correction governs; the body is the pre-referee submission, retained for the audit trail per house convention.

Referee verdict

Adversarial referee, track A1-indefinite-pairing-from-modular-data, iteration 7 (2026-06-10). Default stance REFUTE; stance not sustained — the submission survives with corrections. All mathematics re-derived step by step; all numerics independently replicated in fresh code with a fresh RNG (S aΩ = aΩ to 3·10⁻¹⁵; S = JΔ^{1/2}; Lemma 1b bilinearity/symmetry; Lemma 3 cone positivity, 500 trials, max |Im| ≈ 9·10⁻¹⁶, min Re ≈ 1.34; η₀² = 1, self-adjointness, J-oddness off-kernel, indefiniteness; Ad η₀ breaks right-multiplication commutation by O(10) while the Ad Δ^{it} control passes at 10⁻¹⁴). Citations audited live: Rieffel–van Daele PJM 69 (1977) 187–221 ✓; Haagerup Math. Scand. 37 (1975) 271–283 ✓; Morinelli arXiv:1703.06831 ✓; Neeb–Ólafsson arXiv:1611.00080 ✓; Gottschalk JMP 43 (2002) 4753, arXiv:math-ph/0408048 ✓; arXiv:1710.04965 = JHEP 03 (2018) 089 ✓ but four authors (Devastato–Farnsworth–Lizzi–Martinetti — fix the label); Nieuviarts 2402.05839 v6 ✓; Jakóbczyk upgraded to record-verified (OSTI 5135323, 1985). No fabricated source. No standard no-go bears adversely; the Reeh–Schlieder/cyclic-separating consumption is correctly flagged as g₇.*

A1-F1 (η₀ = sgn(ln Δ) answers the literal compressed target): KEEP — severity MINOR — two binding corrections. The construction, its smuggle-free grading (g₁–g₆ F per the iter-3 checklist definitions, checked item-by-item; C2 equivariance holds), and indefinite-iff-non-tracial are all verified. Correction 1 (uniqueness is false): η₀ is not the unique J-odd self-adjoint unitary in {Δ}″ — any measurable sign function odd in ln λ qualifies; the referee's counterexample ε(Δ) = sgn(sin(ln Δ)) was numerically confirmed (J-odd, ε² = 1, ε ≠ η₀). Uniqueness requires an additional modular-rescaling-invariance axiom ε(λˢ) = ε(λ); as submitted, "canonical" means "a distinguished member of an infinite natural family." This strengthens the geometry-void verdict (every member is equally canonical and equally void). Correction 2: "spec Δ = [0,∞) for any type III state" holds only for III₁ (Connes S(M) = [0,∞) characterizes III₁); for III_λ the spectrum may be {0} ∪ λ^ℤ. The (∞,∞) Krein signature survives for all type III.

A1-F2 (universality / Δ_geo = 0 proven): KEEP — severity MINOR. The Borchers→Mackey→homogeneous-Lebesgue chain and the two-sheet trivialization were re-derived in full and are valid (W = JJ₀ ∈ {x}′ with W(x)·conj(W(−x)) = 1; free choice on one sheet; {x = 0} null). Scope correction: the result substantially coincides with the known uniqueness of non-degenerate standard pairs (every standard pair is a multiple of the unique irreducible one — Longo–Witten line), extended by the trivial remark that any fixed ε(Δ) rides along. File as a repackaging with a new-to-the-wiki corollary, which makes it more reliable.

A1-F3 (trichotomy lemma): KEEP — severity MINOR. Under the flagged ansatz the sesquilinear leg is complete (self-adjoint unitary Grams in {Δ}″ = sign functions of Δ). Two scope notes: (a) the signature-free prong is an enumeration of known non-sesquilinear form types, not a consequence of the ansatz — the residual is correctly flagged but should name this; (b) prong (iii) does not need J-oddness — J-even sign functions are also fundamental symmetries and are equally universal by the same trivialization, so the trichotomy strengthens once stated for all ε. Algebra-incompatibility replicated numerically; "generically" correctly tagged INFERENCE (finite-dim demonstration only).

A1-F4 (J carries reflection positivity — the g₃ trap is structural): KEEP — severity NONE. Exact: aj(a)Ω ∈ P^♮, self-duality, done. The strongest single finding; the upgrade of the g₃ trap from symmetric-state inheritance to abstract cone positivity is correct and well-cited.

A1-F5 (direction-map: prior art runs Krein → modular only): KEEP — severity MINOR. All four cited lines verified live and run in the claimed direction; the referee's own search for any published η-from-modular-data derivation also came up empty. Fix the author truncation (Devastato–Farnsworth–Lizzi–Martinetti); upgrade Jakóbczyk to record-verified.

A1-F6 (hedge hardening + target repair): KEEP — severity MINOR. The repair of OP-46/CONCLUSION §8 is mandatory, not optional: HYP-CKV-VACUITY-R3's sentence "no such derivation exists or is sketched" is now literally false, and the faithful residual is the algebra-compatible, localized (fiberwise) (n,n) pairing — identified via prong (5c) with the carrier problem (HYP-FACTORIZATION-IS-GEOMETRY), the fifth independent route converging on the carrier/localization step. The MEDIUM → MEDIUM-HIGH hardening is accepted conditional on the modular-constructibility ansatz at both ends; HYP-ENCODING-SCREEN inherits with the same condition attached. Consistency with the house closures checked: no conflict with dS-cardinality (η₀ fixes no η = 1/4G), OP-48c (the algebra-incompatibility prong echoes, and does not contradict, the split/collar logic), OP-49, or the four-readout-family structure of HYP-CKV-VACUITY-R3 (the new trichotomy classifies pairings, the old one classifies Lorentzian-datum entry points — distinct, compatible instruments).

Net. The extraordinary-claim test was applied: the claimed positive answer to the literal compressed target is genuine (η₀ exists, is constructible from (M, Ω) alone, and passes every smuggle item including the two iteration-6 additions — the (n,n) axiom and the signature template are derived, not posited), and precisely because it is genuine and provably void, the track outcome RESOLVED-NEGATIVE is upheld: the verdict-flipper was executed, produced its object, and the object carries zero geometric information. Errors found are minor and corrective, and the principal one (non-uniqueness of η₀) points further in the negative direction. Headline (PARTIAL / encodes-not-generates / not-yet-physics) unchanged.