The Carrier Problem, Posed — An Open-Problem Dossier for External Solvers

What this is. The single open analytical gate of the geometry-from-algebra program, distilled from fifteen adversarially-refereed iterations into its sharpest, most-attackable form — a self-contained statement an operator-algebraist, AQFT theorist, or noncommutative-geometer can pick up cold. At terminal analytical saturation, this is what reasoning can still usefully do: pose the problem so the external mathematics that could close it has the best possible target. Every reduction below is web-verified and epistemically tagged; the binding corrections from iterations 9–14 are folded in.

The Net-Naturality Theorem (The Carrier Problem): A Self-Contained Open-Problem Statement for External Solvers

Distilled from 15 adversarially-refereed iterations. This is the single most attackable form of the one open gate in the "geometry-from-algebra" program. An operator-algebraist / AQFT-theorist / noncommutative-geometer can pick this up cold.

0. One-sentence statement

On a Haag-dual, split, Bisognano–Wichmann local net of von Neumann algebras with vacuum, is an Aut(N,Ω)-natural, modular-covariant, localized, indefinite (n,n) (Krein) Hermitian form forced to be signature-inherited from a pre-installed Euler element / causal index-set order (⇒ carrier no-go is unconditional; "Program A encodes but does not generate geometry" is a theorem), or can such a form exist that is not so forced (⇒ a carrier; geometry is generated; the verdict flips)?

1. Setup and definitions

A local net N = {M(O)}_{O∈I} indexed by a partially ordered set I of spacetime regions, with a cyclic-separating vacuum Ω, isotony, locality (commuting algebras for causally disjoint regions), Haag duality, the split property, and the Bisognano–Wichmann (BW) property (the modular group Δ_W^{it} of a wedge acts as the boost). Each local algebra is the unique hyperfinite type III₁ factor.
The modular package m(N) = ({Δ_O^{it}}, {J_O}, {Δ_{O'|O}}, {P_{O',O}}): modular flows, conjugations, relative modular operators, the modular-Berry/Connes connection over the net.
Aut(N,Ω) is the group of vacuum-preserving net automorphisms: invertible maps α sending M(O) → M(αO) compatibly with isotony and the index-set order on I, with α(Ω)=Ω. By Tomita–Takesaki + BW it contains the modular/boost flow {Δ_W^{it}} and the geometric symmetries. "Aut(N,Ω)-natural" (used throughout, incl. §0 and the Conjecture) means: in the locally-covariant (Brunetti–Fredenhagen–Verch) idiom the carrier is a natural transformation of the carrier functor Φ: Net → Krein — i.e. for every α ∈ Aut(N,Ω) the form transforms equivariantly, η_{αx} = α_*η_x (equivalently Ad U_α commutes with the family {η_x}). Naturality under the modular subgroup is exactly the [η, Δ_O^{it}]=0 condition below; naturality under the full Aut(N,Ω) is strictly stronger. [INFERENCE — definitional, after iter-9 §1]
A carrier is a field of Hermitian forms {η_x} (equivalently a fundamental symmetry J=J*, J²=1, signature (n,n)) that is: (modular-natural) [η, Δ_O^{it}]=0; (algebra-compatible) Ad J ∈ Aut(M); (localized, "(loc)") affiliated to M(O) ⊊ M(W) with vanishing cross-coupling between causally disjoint regions; (natural) equivariant under Aut(N,Ω); and (generative) carrying NEW geometric content Δ_geo ≠ 0 — i.e. not a function of Δ and not vacuum/symmetry-inherited.

2. The precise conjecture (the net-naturality theorem)

Conjecture (carrier no-go, net form). Every Aut(N,Ω)-natural, modular-covariant carrier on a Haag-dual split BW net is positive-and-localized, signature-free, or indefinite-and-nonlocalized — never both indefinite and localized. Equivalently: any localized indefinite (n,n) form on such a net has its signature inherited from the net's causal index-set order I (the pre-installed Euler element), not generated by (algebra, state) data.

A proof makes "Program A encodes geometry, does not generate it" a theorem and the no-distinguishing-test corollary unconditional. A counterexample (a generative carrier) flips the verdict: geometry would be an output of algebra+state.

3. Established reductions and lemmas (the spine)

Lemma N2 (the engine — [INFERENCE, high; numerically demonstrated twice]). A localized indefinite natural form has a natural sign-locus Λ ⊂ I carried by every automorphism, hence by the modular flow. By the Borchers relation the modular flow acts as a free translation p ↦ e^{-2πt} p, which has no fixed locus in the open index set; the only flow-invariant sign-pattern is a global constant (definite). The half-line dilation's fixed points p=0,∞ are the wedge edge — a geometric boundary outside the open I, i.e. smuggle-item n₁. Hence localized + natural + indefinite is impossible at finite locus. The only invariant indefinite object is sgn(modular frequency) — natural but nonlocal (Hilbert tail ~1/sep).

Lemma N1 (constructibility — [downgraded to weakening, NOT a theorem]). Modular-naturality forces η ∈ {Δ_O}' (commutant), not into the abelian W*({Δ_O}); for type III₁ the modular spectrum is degenerate so these differ. The iter-9 referee struck the claimed discharge to a theorem. What survives is the commutant-to-generated gap: η lands in the commutant of the modular flow, and whether it is forced to be a function of Δ (geometry-void, η₀-class) or can be a genuinely new generated object is exactly the open residual.

The exact residual (where the whole problem lives). Two-fold and located: (i) the localization-template hypothesis n₁ — "(loc)" is natural only relative to I's causal order, and that order is the geometry (BGL: positivity ⟺ isotony; Neeb–Ólafsson: the order needs a pre-installed Euler element); removing it removes the meaning of "localized"; (ii) the commutant-to-generated gap — the surviving sliver of algebra-compatible carriers in {Δ}' \ W*(Δ) not provably reducible to a function of Δ. iter-10 confirmed the gap does not close: ergodicity of the BW vacuum empties only the centralizer, not the representation-space multiplicity gap, which retains the indefinite η₀=sgn(ln Δ); the strongest candidate carrier clears the indefinite + algebra-compatible bars by an inner automorphism yet is geometry-void.

The baseline obstruction (iter-7). The canonical Krein fundamental symmetry η₀ = sgn(ln Δ) does exist, smuggle-free, from pure (M,Ω) data, is genuinely indefinite — but is provably geometry-void: unitarily universal across the entire Borchers/HSMI class (Δ_geo = 0, proven), generically algebra-incompatible, and non-localizable. Any carrier must be genuinely not of this class.

4. The FIVE converging routes (count is FIVE, never six)

All five are facets of ONE problem; new framings are re-expressions, not new routes:

stress-tensor-flux / Sorce–CKV readout;
modular-Berry / zero-mode readout;
Lorentzian-NCG / η-foliation-twist readout;
causal-fermion-system operator-spectral dichotomy (the (≤n,≤n) eigenvalue axiom IS the signature carrier);
the modular-sign / net-naturality route (this one), of which the following are equivalent forms (each a re-expression of the SAME route, not a sixth):
- the reverse-Weinberg–Witten theorem (a class-separating observable realizable only without fundamental metric d.o.f.) — its theorem direction is co-extensive with the carrier no-go; a witness is strictly harder than a positive carrier solution. [INFERENCE — the equivalence is this program's reduction; the reverse-WW witness itself is OPEN as of 2026-06, no construction exists];
- the completeness of (net + vacuum) as an invariant = Weiner's algebraic Haag theorem (arXiv:1006.4726, M. Weiner, An algebraic Haag's theorem, Comm. Math. Phys. 2011) [ESTABLISHED]: under split + geometric modular action of the wedge algebras, the vacuum-sector net (double cones based on a fixed spacelike hyperplane) is a complete invariant — no assumption that the connecting unitary intertwines vacua or symmetry representations — but its hypotheses are the geometric inputs (n₁, the collar);
- the indefinite-metric / Krein-modular route (iter-12, CLOSED-NEGATIVE): every Krein modular structure either installs J by hand (runs Krein→modular), is a gauge artifact modded out on passing to observables, or is η₀-equivalent / complex-pseudo-entropy.

5. Adjacent named open problems and the exact geometric input each framework smuggles

Connes' bicentralizer problem (type III₁ factors; OPEN in general as of 2026, Ando–Goldbring arXiv:2605.12776). The located commutant-gap residual's vocabulary is RELATED TO (not logically reducible to — the iter-11 equivalence claim was major-corrected) centralizer/bicentralizer structure theory. A trivial (relative) bicentralizer is the carrier-friendly condition, so resolving it either way cannot by itself close the no-go (iter-12, non-load-bearing).
Factorization-algebra / functorial-QFT (iter-15, E1, REPRODUCES-SCREEN). The Benini–Carmona–Grant-Stuart–Schenkel categorical equivalence AQFT ≅ time-orderable prefactorization algebra on globally hyperbolic Lorentzian manifolds (arXiv:2412.07318, Lett. Math. Phys. 2025; building on arXiv:1903.03396; the Lorentzian pAQFT factorization-algebra result is Gwilliam–Rejzner arXiv:2212.08175) is a NEW precise object but not a handle: both its sides are functors over the SAME fixed orthogonality / causal-disjointness relation ⊥ (= smuggle-item n₁), plus the Lorentzian time-orientation, and on the functorial side the O(d)-structured bordism category carries the signature (cobordism hypothesis installs geometric structure as a source parameter). [Published as Lett. Math. Phys. 116, 13 (2026), DOI 10.1007/s11005-025-02035-7; the "2025" above is the acceptance-year DOI stamp, not the publication volume.] It transports the carrier question verbatim across the equivalence — a fourth independent witness to the same obstruction.
Connes–Atiyah–Singer index / K-theory / coarse geometry (iter-15, E2, REPRODUCES-SCREEN). Every index pairing factors through the Dirac operator D (the metric datum; in spectral triples D is the Connes distance) or its bounded phase F=sign(D). In the Lorentzian/indefinite case D is Krein-self-adjoint on a pre-installed Krein space whose indefinite inner product IS the (n,n) carrier η — installed, not generated (arXiv:1611.07062, "indefinite spectral triples over Krein spaces"). On a single local hyperfinite III₁ factor K-theory is structureless (no finite projections, no trace) — but this is only the single-algebra shadow; all carrier content survives in the relations between algebras + the vacuum. The coarse/Roe route smuggles the metric coarse structure and discards fine-scale localization. The spectral localizer (Loring–Schulz-Baldes / Kaad arXiv:2508.08668) still requires a Dirac/first-order operator before yielding a numerical index.
Higher-categorical / homotopical / non-unitary tensor-categorical AQFT (iter-15, E3, REPRODUCES-SCREEN). The orthogonality relation ⊥ = causal disjointness is the base of operadic AQFT (Benini–Schenkel–Woike arXiv:1709.08657); the Lorentzian bordism category is the site (Bunk–MacManus–Schenkel arXiv:2308.01026); the free symmetric monoidal bordism source of the cobordism hypothesis is geometry-in-the-site. An indefinite (n,n) signature can exist categorically only OUTSIDE the positivity-axiomatic unitary world, and there only as installed reflection/dagger data (Müller–Stehouwer arXiv:2301.06664) — the iter-12 Krein problem in categorical vocabulary. No natural functor out of the modular data supplies a localized indefinite carrier; the reformulation does not close the commutant-to-generated gap.

6. The exact common smuggle-item, stated once

Across Tomita–Takesaki, Krein/indefinite-metric, factorization-algebra, index/K-theory, and higher-categorical mathematics, the geometric input enters as one of two interchangeable objects: the causal-disjointness / orthogonality / index-set order n₁ (the localization template), OR a metric/Dirac/Krein datum η (the signature carrier installed on a pre-given space). The open problem is precisely: exhibit, or rule out, a structure in which n₁ (the causal order) and the signature η are OUTPUTS of the algebra+state rather than indexing data of the source category / installed data on the carrier space.

7. What a proof either way would imply for a universal theory of physics

No-go proven (the honest expected outcome, now robust across all modern frameworks): "causal/algebraic/entanglement structure precedes metric geometry" is, for our Λ>0 universe, an encoding strategy, not a generative mechanism — a definitive theorem. By the HYP-ENCODING-SCREEN corollary it follows unconditionally that the central wager has no in-principle distinguishing experimental test (an encode-only program's observational content equals that of the EFT it encodes; the reverse-WW theorem holds). The program is permanently "not yet physics" absent a different idea; a unified GR+QFT framework remains POSSIBLE (AdS/CFT existence proof) but geometry-from-algebra is not the road to deriving it.
Carrier exhibited (a generative counterexample): geometry — causal order and Lorentzian signature — is generated from (algebra, state) data. This would be the first construction outputting a Lorentzian causal/metric structure installing no fundamental symmetry, η-form, foliation, twist, (n,n)-flag axiom, or signature-valued template by hand. It would make the reverse-Weinberg–Witten witness possible in principle (a necessary, not sufficient, prerequisite for a distinguishing test), reopen the entire verdict, and convert the central wager from a research strategy into a candidate result — the first step toward a derived (not merely consistent) universal theory.

8. Status

Standing verdict unchanged 14 consecutive iterations (terminal analytical saturation, grade-derivative zero since iter-9, condition-derivative zero since iter-10). Every verdict-bearing residual is now decidable only by external input: a resolution of the net-naturality theorem (equivalently the carrier no-go / commutant-gap closure / a reverse-WW witness), an external resolution of Connes' bicentralizer problem, or a genuinely new mathematical readout not yet tried. The problem is posed for external resolution.

Status append (2026-07-03, iteration 16 — the residual restructured; the posed problem above stands verbatim)

The §3 "exact residual" is superseded in its second component by the iteration-16 multi-wedge closure (LEM-NET-NATURALITY-G, referee-verified; 2026-07-03-iter16-G1-multiwedge-naturality.md): (G-loc) every localized natural carrier is ±1 (proof-grade); (G-glob) every fully-natural non-localized algebra-compatible carrier is a gauge-symmetry implementer with Poincaré-invariant eigenprojections, Δ_geo = 0 (proof-grade) — the commutant-to-generated gap is non-empty ((−1)^N, η₀) and classified, NOT closed. The residual is now: (i) the localization-template hypothesis n₁ (unchanged, irreducible); (ii) hypothesis (E_O) — vacuum ergodicity on massive double-cone algebras, $M(O)_\omega = \mathbb{C}1$ , [OPEN]. Direction: (E_O) ⟹ the conjecture in §2 is a theorem on the stated hypotheses (+ weak additivity); ¬(E_O) is necessary but NOT sufficient for a counterexample fiber. §6's two-item interchangeability now has a published theorem-level statement, BW-conditionally: Morinelli–Neeb, Adv. Math. 458 (2024) 109960 (arXiv:2312.12182). For external solvers the sharpest open questions are now (E_O) and the DCNG definability lever — see 2026-07-03-iter16-synthesis.md §6. [INFERENCE, high — referee-verified package]

Status append (2026-07-07, iteration 17 — (E_O) attacked directly; the posed problem stands verbatim)

The §3 residual (E_O) received a dedicated five-lens assault (three binding referees; 2026-07-07-iter17-EO-assault-synthesis.md). It stands [OPEN] — neither proved nor disproved; HYP-CKV-VACUITY stays R7. For external solvers the sharpest decidable sub-target is now concrete: for the free massive scalar, (E_O) reduces to purely continuous spectrum (no eigenvectors) of the double-cone one-particle modular Hamiltonian $\ln\delta_O$ — via second quantization, continuous spectrum ⟹ weak mixing ⟹ $M(O)_\omega=\mathbb C1$ . Correction folded in: the naive "no eigenvalue 0 ⟹ ergodic" reading is INSUFFICIENT (bosonic $d\Gamma(A)$ are modular-fixed regardless of point spectrum); the correct sufficient condition is continuous spectrum. The spectral fact is numerical-only (Bostelmann–Cadamuro–Minz arXiv:2209.04681; Cadamuro arXiv:2312.08525; no closed form — Longo–Morsella arXiv:2012.00565, massive-ball result "contained a gap, and have been removed"), leaning-true, unproven. For general class (H), all three structure-theory routes (weak-mixing-from-clustering; wedge→double-cone HSMI; Marrakchi–Vaes genericity) provably fail, and the adversarial disproof fails on eight candidate mechanisms — so the posed problem stands verbatim, sharpened not solved. [INFERENCE, high — referee-verified package]

Status append (2026-07-11, iterations 18–20 — for external solvers: the sharpest form yet)

The free-field (E_O) case is now maximally compressed. Proven: the E=0 half (no eigenvalue 1 of δ_O — Figliolini–Guido 1989 + an independent factoriality re-proof, LEM-K0); every putative eigenfunction lies in H¹₀(I) and vanishes on no open subinterval (iteration-20 rigidity theorems); embedded eigenvalues occur in the ambient operator family (BKT), so only geometry-specific analysis can decide. Numerics (three independent signatures): ladder density of states ε_k → (2k+1)π²/lnL; no pinned mode; edge-divergent eigenfunctions — all pointing to purely continuous spectrum. For a fractional-UCP specialist, the entire remaining question is LEM-A1′: for every c > 1, if ωF = 0 on I, ω(χ_E F + cχ_I F) = 0 on E, and χ_I F ∈ H¹₀(I) (ω = (m²−Δ)^{1/2}), then F ≡ 0 — a coupled Zaremba corner-indicial problem via Caffarelli–Silvestre extension (Fall–Felli/Rüland toolset). Its proof closes free-field (E_O) and makes the net-naturality no-go a theorem on the free-massive-scalar subclass. [OPEN — the posed problem's sharpest externally-attackable form]