Track G3 (iteration 16): Marrakchi's bicentralizer-flow theorem — proof map, transfer audit, AHHM re-check, technique harvest

Status: Watch-target execution (sharpened target from watch-sweep #1, 2026-07-02-watch-sweep-01.md §2). Outcome: EXECUTED — RESOLVED-NEGATIVE on transfer; watch target RE-SHARPENED; AHHM re-check run and closed (no interaction, and stronger: the structural output is pre-satisfied/vacuous for physics nets). Verdict-neutral. Last updated: 2026-07-03 Iteration: 16

Binding inheritance. Iter-12 corrections respected in full (2026-06-19-iter12-physics-net-bicentralizer-lever.md): the bicentralizer route is NON-load-bearing for the carrier no-go; trivial (relative) bicentralizer is the carrier-FRIENDLY condition; the physics local inclusions M(O) ⊂ M(W) are taken per the binding record as irreducible, infinite-index, non-discrete, with NO normal conditional expectation. Default posture: DOWNGRADE. Every abstract phrase relied on is quoted verbatim; all preprint claims are tagged [ESTABLISHED as preprint claim], never as refereed results.

Source-integrity note (process record). The full TeX source of arXiv:2606.23636 was downloaded from arxiv.org/e-print this session (gzip tarball, 15,197 bytes; single file Ergodicity_of_the_relative_bicentralizer_flow.tex, 774 lines) and every quotation below is taken from it directly. An intermediate PDF-summarization step in the pipeline produced fabricated content (nonexistent "Cartan subalgebra" theorems, a nonexistent "Theorem 2.1 spectral gap" statement) which was caught against the source and discarded. This repeats the watch-sweep #1 referee-integrity pattern and is logged deliberately. [ESTABLISHED — process record]

---

§1. Proof map — where exactly "with expectation" enters

The paper. A. Marrakchi, Ergodicity of the bicentralizer flow and Kadison's problem, arXiv:2606.23636 (v1 2026-06-22, math.OA, 12 pp., un-refereed; MSC 46L10, 46L36, 46L37, 46M07). Abstract verbatim (re-verified live this session, matches the sweep record): "We show that the relative bicentralizer flow of a type III₁ irreducible subfactor with expectation is always ergodic. As a consequence, every irreducible subfactor with expectation in a factor with separable predual contains a maximal abelian subalgebra. This completes the solution to Kadison's problem on maximal abelian subalgebras from 1967." [ESTABLISHED as preprint claim]

The author himself flags the hypothesis as essential. Introduction, verbatim: "It is also essential to assume that M has separable predual and that the inclusion N ⊂ M is with expectation, i.e. that there exists a faithful normal conditional expectaton [sic] from M onto N (see the discussion after [GP98, Corollary 5.1] and [Po21, Problem 7.6])." So the with-expectation restriction is not an artifact of this proof — the author regards it as necessary for the statement, with pointers to Ge–Popa (Duke Math. J. 94 (1998)) and Popa (Commun. Math. Phys. 384 (2021), Problem 7.6). [ESTABLISHED as preprint claim]

Statement structure (letter-theorem numbering from the source):

• Theorem A (MASA): "Let N ⊂ M be an inclusion of von Neumann algebras with expectation and with separable preduals. There is a maximal abelian subalgebra A ⊂ N such that A′ ∩ M = A ∨ (N′ ∩ M). In particular, if N′ ∩ M ⊂ N, then A is maximal abelian in M."
• Theorem B (main; ergodicity + eigenvectors): "Let N ⊂ M be an inclusion of von Neumann algebras with expectation, where N is of type III₁. Let φ be a faithful normal state on N. Then the fixed point algebra of the bicentralizer flow β^φ : ℝ₊ ↷ B(N ⊂ M, φ) is equal to N′ ∩ M."* Plus: x is a β-eigenvector at frequency t ⟺ "xy = σ_t^φ(y)x for all y ∈ N."
• Corollary C: for N type III₁ with expectation in M, "there exists an amenable subfactor with expectation P ⊂ N such that P′ ∩ M = N′ ∩ M" (P can be taken type III₁; type III_λ availability governed by the relative T-invariant).
• Corollary D: same with "amenable subalgebra" for arbitrary N.
• Theorem E: for irreducible subfactors with expectation, existence of a faithful state with ergodic modular flow leaving N globally invariant ⟺ irreducibility of the core inclusion c(N) ⊂ c(M).

Proof strategy (introduction, verbatim): "The starting point of the proof is the weak relative Dixmier property for the inclusion N ⊂ M proved in [Ma19]… Our strategy is to construct a binormal state Φ ∈ B(L²(M)) whose properties encode the weak Dixmier property of the inclusion and then apply the ultrapower implementation of binormal states technique of [Ma25]. But in order to apply this technique, one needs to control the behaviour of the state Φ with respect to the modular operator. The key novelty of this paper is to use the approximate eigenstate technique of [Ma18] and then carefully manipulate the state Φ in order to make it well-behaved with respect to the modular operator, while preserving its binormality."* No spectral-gap, no Popa deformation/rigidity, no model theory: the toolkit is (i) Dixmier-type averaging over almost-φ-central unitaries, (ii) ultrapower implementation of binormal states, (iii) approximate eigenstates of log Δ_φ with UCP spectral-translation maps, (iv) Følner averaging over the flow, (v) crossed-product promotion. [ESTABLISHED as preprint claim — architecture read directly from source]

The expectation-usage census. The hypothesis enters at EIGHT distinct load-bearing points; the finding is pervasive AND definitional, not once-used-and-removable:

Answer to task (1): the hypothesis is used pervasively — and, decisively, before any proof step begins: the relative bicentralizer flow β^φ (the theorem's subject) is implemented through the compatible state extension φ = φ∘E_N (E0 + E4). The definitional references are to AHHM (arXiv:1804.05706) and Ma25 (Invent. Math. 239 (2025) 79–163), both of which are with-expectation frameworks throughout: the AHHM abstract (re-verified live this session) sets its scope as inclusions "where N is a type III₁ subfactor with normal expectation and φ is a faithful state." [ESTABLISHED — read from source + live abstract]

§2. Transfer analysis — M(O) ⊂ M(W)

Physics-net data (binding, per iter-12 record): M(O) ⊂ M(W) irreducible, infinite-index, non-discrete, no normal conditional expectation; extra structure: split chain M(O) ⊂ N_I ⊂ M(Õ) ⊂ M(W) with N_I a type I factor (Doplicher–Longo), BW modular covariance, vacuum Ω cyclic-separating for every algebra in the chain. [ESTABLISHED per iter-12 record]

(a) Is the relative bicentralizer even the right object without an expectation?

Split answer. The set B(N ⊂ M, φ) — elements of M asymptotically commuting with every uniformly bounded φ-asymptotically-central net in N — can be written down without an expectation (φ a state on N; centrality in N_*; commutation tested σ-strongly in M). [INFERENCE — definitional reading] But the theorem is not about the set; it is about the flow, and the flow is not defined without the expectation:

• The implementation (E4) requires partial isometries v ∈ N^ω that are modular eigenvectors for the ambient state: vφ^ω = e^t φ^ω v with φ ∈ M_*. This presupposes a faithful normal state on M whose modular group globally preserves N (so that its restriction is modular for N). By Takesaki's theorem, such a state exists iff a faithful normal conditional expectation exists. For M(O) ⊂ M(W) no expectation exists, hence no state on M(W) has a modular flow globally preserving M(O) — indeed for the natural (vacuum-class) states the modular flow is the boost (Bisognano–Wichmann), and σ^ω_t(M(O)) = M(Λ_t O) ≠ M(O): the flow moves the small algebra off itself. [ESTABLISHED mechanism; the no-expectation input is the iter-12 record]
• Equivalently at the Hilbert-space level: the projection onto the closure of M(O)Ω exists, but the commutation Δ^it e_N = e_N Δ^it (which every step E1–E5 uses) fails — it is the Takesaki condition.

So the first obstruction is definitional, and it is located exactly at the geometry: the reason the flow cannot be defined is that the ambient modular dynamics is geometric and moves O — smuggle-item n₁ in yet another costume. The watch-sweep's sharpened target ("an ergodicity/rigidity theorem for expectation-free irreducible inclusions") is therefore ill-posed as stated: before any such theorem, one needs a new definition of a relative-bicentralizer-type invariant for expectation-free inclusions. [INFERENCE, high]

(b) Can the substitutes repair each proof step?

Operator-valued weight. Haagerup's existence criterion: an n.s.f. operator-valued weight T : M → N exists iff there exist f.n.s. weights φ on N, ψ on M with σ^ψ|_N = σ^φ — the same Takesaki-type compatibility, at weight level. Whether M(O) ⊂ M(W) admits such a T at all is, to this track's knowledge, [OPEN] (leaning no: it would supply an ambient weight whose modular flow globally preserves M(O), and no such non-geometric invariant weight is known for BW nets; but no citable theorem excludes it — the iter-12/task framing "an operator-valued weight in place of an expectation [Longo; ILP]" is hereby graded as unverified for this specific inclusion and should not be repeated as established). Even granting T, the proof machinery is irreparably state-based: (i) the normalization Φ(e_N) = 1 (E2) has no unit-preserving analogue for an unbounded T; (ii) the almost-central-unitary averaging with ‖uφ − φu‖ ≤ ε (E3) needs a state; (iii) Ma18's approximate-eigenstate decomposition (E5, [Ma18 Thm 3.2]) decomposes states, with unit vectors ξ_i ∈ dom(X); (iv) the accumulation-point/Følner arguments live in state space. The substitution fails at E4 at the latest, and plausibly already at existence. [INFERENCE, high]

Split-property intermediate type I factor. The chain relocates the problem, it does not solve it:

• N_I ⊂ M(W) is with expectation — indeed many: since N_I is a type I factor, M(W) ≅ N_I ⊗̄ (N_I′ ∩ M(W)) and every normal state ψ on the relative commutant gives a slice expectation id ⊗ ψ. None is canonical (state-dependent). [ESTABLISHED — standard tensor splitting]
• M(O) ⊂ N_I is expectation-free for type reasons, irreparably: a normal conditional expectation defined on a type I algebra has type I (atomic) image (Tomiyama), while M(O) is type III₁. [INFERENCE, high — standard structure theory]
• Therefore no composition of expectations along the chain can reach M(O), and the split property cannot restore the hypothesis: it concentrates the entire expectation-free-ness in the leg where it is provably permanent.
• What Marrakchi's theorems DO give on the with-expectation legs is vacuous for our purpose: Theorem B does not even apply to N_I ⊂ M(W) (it needs the small algebra type III₁; N_I is type I, where bicentralizer theory trivializes), and Theorem A applied there outputs a MASA A ⊂ N_I with A′ ∩ M(W) = A ∨ (N_I′ ∩ M(W)) — a restatement of the tensor splitting. [INFERENCE, high]

Vacuum standardness / BW covariance. Joint cyclicity-separation of Ω for the chain gives standardness of every inclusion but buys none of E0–E8 (the machinery needs the state extension property, not a standard vector). BW covariance points the wrong way: it is precisely the mechanism that makes every natural candidate flow geometric, i.e., non-N-preserving. [INFERENCE, high]

(c) Verdict: exact first obstruction, and the strongest transferable residue

Exact first obstruction (one sentence). The relative bicentralizer flow of M(O) ⊂ M(W) is undefined: its construction requires a faithful normal state on M(W) whose modular group globally preserves M(O) (Takesaki-equivalent to the existence of a normal expectation, which the physics inclusion lacks), and for BW nets every natural candidate modular flow is a boost that moves O — the "with expectation" hypothesis is consumed at the level of the object, before the first proof step. [INFERENCE, high; mechanism ESTABLISHED]

The strongest transferable residue is not the flow but the relative T-invariant. Marrakchi's Definition (verbatim): "T(N ⊂ M) := { t ∈ ℝ | there exists a nonzero x ∈ M such that xy = σ_t^φ(y)x for all y ∈ N } where φ is any faithful normal semifinite weight on N" — this needs only a weight on N, no expectation, and is well-defined for M(O) ⊂ M(W). By Theorem B, in the with-expectation world the flow's eigenvectors at frequency t are exactly the T-realizing intertwiners; the expectation-free shadow of the eigenvector theorem is therefore a statement about T. Precise transferable conjecture:

Conjecture (T-triviality of local inclusions). For a Haag-dual, split, BW net and O ⊊ W with M(O)′ ∩ M(W) = ℂ1 (per program record): T(M(O) ⊂ M(W)) = {0}. Equivalently, no nonzero x ∈ M(W) satisfies xy = σ^ω_t(y)x for all y ∈ M(O) and some t ≠ 0. (Mechanism sketch: polar decomposition + irreducibility of both M(O) and its boosted image would upgrade x to a unitary u ∈ M(W) with Ad(u)|{M(O)} = σ^ω_t|{M(O)} — an inner-in-M(W) implementation of a boost displacement of a local algebra, which standard AQFT rigidity should exclude.) [SPECULATIVE — proposed adjacent question; not verdict-bearing either way: a proof yields a rigidity statement about unitary intertwiners, not an indefinite pairing; a counterexample yields a modular-covariant localized intertwiner, still not a carrier (unitary, not a fundamental symmetry).]

Note Marrakchi's countability proof for T (E7) uses the expectation — so even the T-invariant's basic properties need new proofs in the physics setting. This is a well-posed, expectation-free, genuinely new adjacent question — the only object in the paper that transfers.

§3. Load-bearing re-check — the AHHM hyperfinite-subfactor output (the never-run check)

The AHHM equivalence (abstract re-verified live this session): "For general irreducible inclusions N ⊂ M, we relate the ergodicity of the flow β^φ to the existence of irreducible hyperfinite subfactors in M that sit with normal expectation in N." [Ando–Haagerup–Houdayer–Marrakchi, Math. Ann. 376 (2020) 1145–1194, arXiv:1804.05706] In Marrakchi's phrasing (source, intro): "[AHHM18, Theorem C] says that when N is of type III₁, there exists an amenable subfactor with expectation P ⊂ N such that P is irreducible in M if and only if the so-called relative bicentralizer flow β^φ : ℝ₊ ↷ B(N ⊂ M, φ) is ergodic."* [ESTABLISHED — published, Math. Ann.]

The check the prior iterations never ran: if Marrakchi's ergodicity (or a hypothetical transfer) forces such subfactors inside the local inclusion, does that structural output interact with the carrier problem's localization bar? Answer: no interaction — and stronger than expected: the output is VACUOUS for physics nets. Three blocks, in order of decisiveness:

1. Blocked at level 0 (category). The AHHM equivalence itself lives entirely in the with-expectation category — its scope line (verified live) already assumes "N … type III₁ subfactor with normal expectation" — and its left-hand side (flow ergodicity) is undefined for M(O) ⊂ M(W) by §2(a). Nothing can "force" anything: the implication cannot fire on the physics inclusion. [ESTABLISHED]

2. Pre-satisfied even counterfactually (the new, decisive observation). Suppose a full expectation-free transfer existed — a new flow-like invariant, proved ergodic, with an AHHM-type equivalence attached. Its structural output for N = M(O), M = M(W) would be: an amenable (hyperfinite) subfactor P ⊂ M(O), with normal expectation into it, irreducible in M(W) (P′ ∩ M(W) = ℂ). But this is already trivially true with P = M(O) itself: M(O) is hyperfinite/amenable [ESTABLISHED — split ⟹ hyperfinite, iter-12 §2], the identity expectation is normal, and M(O)′ ∩ M(W) = ℂ per the binding irreducibility record. (If one weakens the record to the P′ ∩ M = N′ ∩ M form of Corollary C, P = N pre-satisfies again.) The hyperfinite-subfactor existence output carries zero new structural information about the physics inclusion — it is not merely non-load-bearing, it is empty. [INFERENCE, high]

3. No interaction in either direction even where non-vacuous. (i) Carrier-killing direction: the no-go rests on the iter-7/9/10 geometry-void, naturality (Lemma N2 boost-no-fixed-locus), and commutant-gap bars, none of which reference subfactor structure; adding hyperfinite irreducible subfactors adds no bar. (ii) Carrier-friendly direction: the objects the Marrakchi/AHHM machinery produces — amenable subfactors P, MASAs A — are constructed by ultrapower and averaging arguments: state-dependent and non-canonical, hence not Aut(N,Ω)-natural (they fail the naturality leg of the carrier definition before any signature question arises), and the MASA output is abelian = definite, structurally the opposite of an indefinite (n,n) pairing (watch-sweep referee point, re-confirmed). Any fundamental symmetry affiliated to such a P ⊂ M(O) is a fortiori affiliated to M(O), so Lemma N2 applies verbatim — no better-localized home for a carrier is created. [INFERENCE, high]

On the record: the AHHM hyperfinite-subfactor channel is closed. Expected answer confirmed ("no interaction / still non-load-bearing"), with the sharpened reason: for local inclusions the equivalence's right-hand side is pre-satisfied by P = M(O), so even a full expectation-free transfer of Marrakchi's theorem would force nothing.

§4. Technique harvest — mappings onto the commutant-to-generated gap ({Δ}′ vs W*(Δ))

The gap (dossier §3, "the exact residual"): carriers are forced into {Δ_O}′ but not provably into W*(Δ_O); the surviving sliver {Δ}′ \ W*(Δ) is where any counterexample must live. The paper's techniques, mapped, with one-line feasibility grades:

Absent from the paper (contrary to plausible guesses): no spectral-gap arguments, no Popa deformation/rigidity, no intertwining-by-bimodules, no model theory. The toolkit is averaging + ultrapowers + modular spectral analysis. [ESTABLISHED — read from source]

§5. Cluster sweep (window 2026-06-22 → 2026-07-03) + Houdayer–Marrakchi selflessness

Citations of arXiv:2606.23636 in-window: NONE found. Semantic Scholar (queried live 2026-07-03): citationCount: 0, citations: []. Web searches for citing/extending math.OA postings in the window return nothing. Eleven days post-v1; the null is expected and dated here as the baseline for the next sweep. [ESTABLISHED as of 2026-07-03]

Adjacent cluster items surfaced by the sweep (all PRE-window; background only, not findings): Ando–Goldbring, Model theory and Connes' bicentralizer problem, arXiv:2605.12776 (May 2026 — already in the dossier §5); Weak relative Dixmier property and Popa's intertwining technique for type III subfactors, arXiv:2508.17592 (2025); Cocycle perturbations and ergodicity for actions on type III factors, arXiv:2512.12931 (Dec 2025); Masuda, arXiv:1812.02877 (PRIMS 56 (2020) — relative bicentralizer flow vs relative flow of weights, finite-index/discrete cases).

Houdayer–Marrakchi, arXiv:2511.11409 — Selfless W*-probability spaces and Connes' bicentralizer problem (v1 2025-11-14, v2 2026-05-01; 5 pp.; "To appear in J. Math. Soc. Japan"). Abstract verbatim (fetched live): "We introduce the notion of selfless W*-probability space and study its connection with Connes' bicentralizer problem. In particular, we show that if M is a separable type III₁ factor with trivial bicentralizer, then (M, φ) is selfless for every faithful normal state φ ∈ M_*." [ESTABLISHED as preprint claim; acceptance noted]

Relevance grade to expectation-free inclusions (one paragraph): LOW / not relevant. It is a single-factor statement, not an inclusion statement — no relative bicentralizer, no expectation-free content, no localization structure; wrong category for the watch target. Its only physics-net consequence is automatic: since the local algebras are injective III₁ with trivial bicentralizer (Haagerup, unconditional — iter-12 §2), every (M(O), φ) is selfless for every faithful normal state [INFERENCE, high — direct application] — a new structural property of single local algebras with no indefinite/signature or carrier content in sight. One precision repair for the record: watch-sweep #1 called this the "Houdayer–Marrakchi selflessness equivalence"; the abstract as fetched states one implication ("we show that if … then …"). Whether the converse is in the body is unverified; downgrade the sweep's phrasing to "implication" until checked. [ESTABLISHED — verbatim comparison]

§6. Outcome and consequences proposed

Outcome label: WATCH-TARGET EXECUTED — TRANSFER RESOLVED-NEGATIVE; TARGET RE-SHARPENED. Verdict-neutral (no verdict-relevant claims; nothing moves the headline, the hedge, or the FIVE count).

1. Watch-target resolution. The sweep-#1 sharpened target ("an ergodicity/rigidity theorem for the relative bicentralizer flow of expectation-free irreducible inclusions") is retired as ill-posed: the flow does not exist without the expectation (§2(a) — the obstruction is definitional and sits at Takesaki compatibility = the geometric flow moving O = n₁). Replace it with the re-sharpened three-prong target: (α) any new definition of a relative-bicentralizer-type invariant for expectation-free inclusions (e.g., extension-independent set-level B, or via Longo's canonical endomorphism) together with any theorem about it; (β) movement on bicentralizer triviality (unchanged from sweep #1); (γ) any computation of the expectation-free-definable relative T-invariant for QFT-type inclusions.
2. AHHM channel closed on the record (§3). Even a full expectation-free transfer would force only a hyperfinite irreducible subfactor with expectation inside M(O) — pre-satisfied by P = M(O) itself. Vacuous output; no interaction with the localization bar in either direction. The iter-12 "non-load-bearing" grade is thereby strengthened, not merely confirmed.
3. New adjacent open question proposed (integrator's discretion, OPEN_PROBLEMS candidate, explicitly non-load-bearing): the T-triviality conjecture T(M(O) ⊂ M(W)) = {0} (§2(c)) — the unique object in the paper that transfers expectation-free.
4. Candidate future lever logged (non-urgent): the T3 single-algebra Θ_t-averaging attack on the commutant-to-generated gap over M(W) (§4) — expectation-free, bounded scope, expected outcome no-go-supporting (η₀-classification reconfirmation).
5. Citation-precision repairs for the wiki ledgers: (i) the preprint's own bibliography misprints Haagerup's bicentralizer paper as "Acta Math. 69 (1986)" — the correct datum (already in the iter-12 ledger) is Acta Math. 158 (1987) 95–148; do not import the preprint's version. (ii) Marrakchi's Inventiones-track Kadison paper (arXiv:2308.15163) is now cited by its author as published: Invent. Math. 239 (2025) 79–163 — update the sweep's "Inventiones-track" phrasing. (iii) Downgrade "selflessness equivalence" → "selflessness implication" (§5). (iv) Process record: one pipeline PDF-summarization fabricated theorem statements for 2606.23636; caught against the TeX source (header note).
6. Confirmation counter: no bump proposed from this track (watch-target execution; verdict untouched). HYP-CKV-VACUITY stays -R6 in grade and condition; carrier-convergence count stays FIVE.

§7. Verified-citation ledger (live this session, 2026-07-03)

1. arXiv:2606.23636 (Marrakchi, Ergodicity of the bicentralizer flow and Kadison's problem, v1 2026-06-22, math.OA, 12 pp.) — abstract page fetched live (verbatim match with sweep record) and full TeX source downloaded from arxiv.org/e-print (gzip, 15,197 B; Ergodicity_of_the_relative_bicentralizer_flow.tex, 774 lines). All §1 quotations verbatim from source. [ESTABLISHED as preprint claim — un-refereed]
2. arXiv:1804.05706 (Ando–Haagerup–Houdayer–Marrakchi, Math. Ann. 376 (2020) 1145–1194) — abstract re-fetched live; flow-ergodicity ⟺ hyperfinite-subfactor equivalence and the with-expectation scope line quoted verbatim. [ESTABLISHED — published]
3. arXiv:2511.11409 (Houdayer–Marrakchi, Selfless W*-probability spaces and Connes' bicentralizer problem, v2 2026-05-01, 5 pp., to appear J. Math. Soc. Japan) — abstract fetched live, quoted verbatim. [ESTABLISHED as preprint claim; acceptance per comments field]
4. Semantic Scholar API, paper f3bdf0dd9113bd1b3734ca723216acb1e3c023e2 = arXiv:2606.23636 — citationCount 0, citations [] (queried live 2026-07-03).
5. arXiv:2308.15163 = Invent. Math. 239 (2025) 79–163 (Marrakchi, Kadison's problem for type III subfactors and the bicentralizer conjecture) — per the 2606.23636 bibliography ([Ma25]) + Springer DOI s00222-024-01299-5 surfaced live in search. [ESTABLISHED — published]
6. Inherited from the iter-12 verified ledger (not re-fetched): Marrakchi–Vaes arXiv:2305.14217 (Crelle 809 (2024) 247); Haagerup, Acta Math. 158 (1987) 95–148; Doplicher–Longo, Invent. Math. 75 (1984) 493–536 [volume repaired per R1-C7/R2: the iter-12 ledger's "73" was a typo]; Buchholz–D'Antoni–Fredenhagen, CMP 111 (1987) 123–135.
7. Background items surfaced but NOT relied on for any claim: arXiv:2605.12776, arXiv:2508.17592, arXiv:2512.12931, arXiv:1812.02877 (all pre-window; §5).

Un-verified items flagged: the existence of an n.s.f. operator-valued weight M(W) → M(O) is [OPEN] (§2(b)) — the program should stop asserting it as available structure until sourced; Tomiyama's atomic-image fact for normal expectations on type I algebras is used at [INFERENCE, high] without a page-level citation.

See also

• 2026-07-02-watch-sweep-01.md §2 — the three-referee adjudication and the (now-retired-and-replaced) sharpened watch target
• 2026-06-19-iter12-physics-net-bicentralizer-lever.md — the binding non-load-bearing record + watch-mode addendum this track discharges
• 2026-06-19-iter15-carrier-problem-dossier.md §3 — the exact residual the §4 harvest maps onto

---

Referee verdict — R3 (binding)

Adversarial referee R3, iteration 16. Default stance REFUTE. Independent re-verification performed this session (2026-07-03): the arXiv:2606.23636 e-print was re-downloaded by the referee directly from arxiv.org/e-print (gzip tarball, 15,197 bytes; single file Ergodicity_of_the_relative_bicentralizer_flow.tex, 774 lines — byte-level match with the submission's process record). Every §1 quotation (E0–E8), the abstract, Theorems A/B/E, Corollaries C/D, the proof-strategy paragraph, the T-invariant definition, and both bibliography items at issue were checked verbatim against the source by the referee, not taken from the submission. The abstract page, AHHM (arXiv:1804.05706), Houdayer–Marrakchi (arXiv:2511.11409), and the Semantic Scholar record (citationCount 0, reproduced live) were fetched independently. The submission's fabrication-catch is corroborated: the source contains NO Cartan-subalgebra theorem and NO spectral-gap argument (the only regex hit for either term is a commented-out bibitem, line 704).

Overall stance: SUSTAINED (verdict-neutral, as submitted). Two claims carry binding corrections/caveats. Headline, hedge HYP-CKV-VACUITY (-R6), and carrier-convergence count FIVE untouched — confirmed correct.

Per-claim adjudication

(1) "With expectation" is pervasive AND definitional — SUSTAINED. All eight census points verified verbatim in the source: E0 (line 298: "We choose a state φ ∈ M_* such that φ = φ∘E_N"), E1 (line 301), E2 (line 318, the (1 ⊗ e_N)^ω localization), E3 (line 365, the almost-central averaging with ‖uφ − φu‖ ≤ ε_i), E4 (line 375, the eigen-isometry implementation of β^φ — verbatim as quoted), E6 (line 482, "Observe that N ⊂ M̃ is with expectation"), E7 (lines 507–510, including the "condition expectation" typo), E8 (line 597). The author's own essentiality flag is verbatim at line 236, including the "expectaton" [sic] typo the submission faithfully reproduced. The flow's implementation (E4) does presuppose the Takesaki-compatible ambient state; "consumed before any proof step" is the correct reading. The AHHM scope line was independently re-fetched and matches.

(2) Transfer IMPOSSIBLE AS POSED; sweep-#1 target ILL-POSED AS STATED — SUSTAINED-WITH-CORRECTION. The Takesaki contrapositive is airtight conditional on the standing record: for a faithful normal STATE φ on M(W), global σ^φ-invariance of M(O) is equivalent (Takesaki 1972; semifiniteness of φ|_N automatic for states) to the existence of a φ-preserving normal conditional expectation — excluded by the binding record. Hence β^φ is undefined for M(O) ⊂ M(W), and a "theorem about the flow of expectation-free inclusions" is ill-posed as literally worded in watch-sweep-01 §2. The correction to the standing record is ALLOWED; the replacement three-prong target (α/β/γ) is well-posed (α asks for a definition + theorem, a legitimate watch shape; γ is expectation-free-definable, verified under claim (4)). Binding caveat (provenance): the load-bearing record datum "NO normal conditional expectation M(W) → M(O)" is asserted at watch-addendum/sweep level; the iter-12 note's body (§3) argues only infinite-index/non-discreteness. The referee did not find a page-level published source for the all-states exclusion in the standing ledgers. The Takesaki logic is valid GIVEN the record; the integrator must (i) keep the tag at [ESTABLISHED per record], never [ESTABLISHED — published], and (ii) open a small sourcing chore: attach a page-level citation or a written-out derivation (e.g., uniqueness of the expectation for irreducible inclusions + BW geometric flow) to the "no normal expectation" record line. This does not weaken the present resolution — the record is binding — but the chain of custody must be visible.

(3) AHHM structural output VACUOUS for physics nets — SUSTAINED. Quantifiers checked against the independently fetched AHHM abstract: the relation is to "the existence of irreducible hyperfinite subfactors in M that sit with normal expectation in N" — expectation into P from the SMALL algebra, exactly as the submission requires; Marrakchi's Theorem C phrasing (source line 250: "amenable subfactor with expectation P ⊂ N such that P is irreducible in M") matches. Pre-satisfaction check: P = M(O) is hyperfinite/amenable (standing record, split ⟹ hyperfinite), sits in itself with the (trivially normal) identity expectation, and is irreducible in M(W) per the standing iter-12 §3 record (M(O)′ ∩ M(W) = ℂ1 — confirmed to be the record's claim). Neither the AHHM abstract nor Marrakchi's statement imposes properness on P. The Corollary-C form (P′ ∩ M = N′ ∩ M, source line 269) is likewise pre-satisfied by P = N. Block 1 (category-level block: the equivalence's own scope is with-expectation) independently verified. The vacuity finding stands and is the sharpest formulation on record; the counterfactual framing of block 2 is clearly labeled and legitimate.

(4) Relative T-invariant expectation-free; conjecture well-posed, non-load-bearing — SUSTAINED. Definition verified verbatim (source: "where φ is any faithful normal semifinite weight on N", with Connes-cocycle independence at line 499) — only a weight on N is needed; no expectation. The conjecture T(M(O) ⊂ M(W)) = {0} is well-normalized (0 ∈ T always, via x = 1) and well-posed. The countability proposition's expectation-dependence (E7) verified — the submission's warning that even T's basic properties need new proofs in the physics setting is correct. Non-load-bearing grading (intertwiners, not indefinite pairings; neither resolution moves the verdict) is sound; the mechanism sketch is properly tagged [SPECULATIVE].

(5) Precision repairs — SUSTAINED (all three; one addition). (i) Houdayer–Marrakchi arXiv:2511.11409 abstract independently fetched: "we show that if M is a separable type III₁ factor with trivial bicentralizer, then (M, φ) is selfless…" — an implication. Watch-sweep-01's "selflessness equivalence" is WRONG; the submission's correction to "implication (converse unverified)" is adjudicated CORRECT and binding. (ii) The "Acta Math. 69 (1986)" misprint is real (source line 673 — note the year is also wrong: 1986 vs the correct 1987); correct datum Acta Math. 158 (1987) 95–148 per the iter-12 ledger — do not import the preprint's version. (iii) Ma25 = Invent. Math. 239 (2025) 79–163 confirmed at source line 721; "Inventiones-track" phrasing update approved. Referee addition: the source's [Ma18] bibitem (line 716) gives "Invent. Math. 222(1) (2018), 375–398" — the year appears to be a second internal misprint (that volume is 2020); nothing in this submission relies on it, but flag it so nothing is imported from that bibitem either without independent verification.

Binding corrections and integration instructions

1. Watch-sweep-01 §2 / ROADMAP clock (1) — replace the sharpened watch target with exactly this wording: "Watch target (clock 1, re-sharpened iter-16 after full-source audit of arXiv:2606.23636): the relative bicentralizer FLOW is undefined for expectation-free inclusions (Takesaki compatibility fails for M(O) ⊂ M(W)); the prior target ('ergodicity/rigidity for the flow of expectation-free irreducible inclusions') is retired as ill-posed. Watch instead for: (α) any new definition of a relative-bicentralizer-type invariant applicable to expectation-free inclusions, together with any theorem about it; (β) any movement on bicentralizer TRIVIALITY itself (unchanged); (γ) any computation of the expectation-free relative T-invariant T(N ⊂ M) for QFT-type inclusions — including the T-triviality conjecture T(M(O) ⊂ M(W)) = {0} (iter-16 G3, non-load-bearing)."
2. Do not relitigate the iter-12 record; but open the provenance chore described under claim (2): page-level source or written derivation for "no normal conditional expectation M(W) → M(O)". Until closed, that datum's tag ceiling is [ESTABLISHED per record].
3. Operator-valued weight: binding. The program stops asserting an n.s.f. operator-valued weight M(W) → M(O) as available structure; grade [OPEN] (submission §2(b)/§7 upheld). Any prior "Longo/ILP operator-valued weight in place of an expectation" phrasing in standing files is to be annotated as unverified-for-this-inclusion.
4. T-triviality conjecture: admit to OPEN_PROBLEMS as an adjacent, explicitly non-load-bearing item, wording per §2(c), tags as given (SPECULATIVE mechanism, OPEN question). No hedge, no counter movement.
5. AHHM channel: record as CLOSED-VACUOUS per §3 (strengthening, not merely confirming, the iter-12 non-load-bearing grade). The iter-12 note may take a one-line addendum pointing here.
6. Ledger repairs (i)–(iii) + the referee's [Ma18]-year addition: apply to BIBLIOGRAPHY/notes ledgers. Retain the §6.5(iv) process record (pipeline PDF-summarization fabrication, caught against source — now twice-corroborated: the referee independently confirmed the fabricated content is absent from the source).
7. T3 lever (§4): log as candidate future lever, cross-linked to iter-16 G4's DCNG watch-item — two DISTINCT attacks on the same dossier-§3 residual (ii); do not merge, do not double-count as separate residuals.
8. Counters: no confirmation bump (watch-target execution, not an analytical iteration) — submission's own accounting confirmed correct.