{"id":"411bacab-6c3e-4d9c-b5b2-ac754e3f85c0","arxiv_id":"2508.17592","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every positive element of a von Neumann algebra that is integrable for an operator valued weight can be averaged by conjugating unitaries so that the weak closure of its orbit's convex hull meets the relative commutant, generalizing Marrakchi's theorem.","lead":"This mathematics paper proves a new averaging theorem for inclusions of infinite-dimensional operator algebras, the first version that works when only a generalized weight, rather than a trace or an exact projection, is available. The theorem yields a trace-free reformulation of Popa's intertwining technique, new solidity results for type III factors, and the answer to an open question about intermediate subfactors.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Semifinite step in Theorem 3.3 is unsupported: Proposition 3.11 proves only m_φ⊂D for an invariant weight φ=Tr_A∘E_A, while Theorem 3.3 needs m_EA⊂D; these differ when A is properly infinite (e.g. A=B(H), x=1).","rationale":"The reader identified the type III part of Proposition 3.13 and its dependence on [Ma23, Theorem 5.2] as the weakest assumption. That is a legitimate concern about reliance on deep published machinery, but the published theorem appears to apply exactly as stated for type III A, and the minimal-idempotent extension step can be repaired once the typo (D vs. DSG) is corrected: if \tilde E extends E:A→Z(A) to M and Φ is minimal, then Φ∘\tilde E is minimal and its restriction to A is Φ|_A∘E=E, because elements of DSG(Ad_A) fix Z(A) pointwise. The more serious gap is internal and occurs already in the semifinite case, which Theorem 3.3 needs for every semifinite summand. Proposition 3.11 proves only m_φ⊂D for an invariant weight φ, not m_{E_A}⊂D; for a properly infinite semifinite A and φ=Tr_A∘E_A these domains genuinely differ. The proof of Theorem 3.3 nevertheless invokes Proposition 3.11 to conclude m_{E_Az}⊂D. Since D is only shown to be norm closed, approximation from m_φ does not close the gap. The theorem may well be true in this case—for example, Dixmier's property for A itself is available—but the written proof does not provide the required argument. Because this concern affects the central theorem directly, it reinforces the reader's CONDITIONAL verdict; it does not by itself show the theorem is false, so I would not move the verdict to REJECT. I disagree with the reader's choice of weakest point because the semifinite properly infinite case is an unaddressed step in the proof as written, rather than an external hypothesis that is visibly satisfied.","tokens_in":47493,"tokens_out":35984,"duration_ms":401130,"concrete_test":"Check the implicit assertion m_{E_A}⊂m_{Tr_A∘E_A} used in applying Proposition 3.11. Take A=B(ℓ2) with the usual trace, M=A, E_A=id, and x=1_A. Then E_A(x)=1_A is bounded but Tr_A(E_A(x))=∞, so x∈m_{E_A}∖m_{Tr_A∘E_A}; Lemma 3.10 cannot be applied to this x. To test whether the gap is cosmetic, supply the missing “standard Hilbert space methods” step for a properly infinite semifinite A: for instance, with A=B(H)⊗1⊂B(H⊗K), E_A=(id⊗Tr) restricted to m_{E_A}, and x=p⊗e where p is an infinite projection and e is rank-one, verify directly that the σ-weak closure of K(x,A) meets A'∩M. If the paper contains no such argument, the proof of Theorem 3.3 is incomplete in the properly infinite semifinite case.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Theorem 3.3 (§3.4), after decomposing A into a semifinite summand A_z and a type III summand A_{z^⊥}, the text states: “By Propositions 3.11 and 3.13, we have m_{E_Az}⊂D(Ad_{Az}) and m_{E_Az^⊥}⊂D(Ad_{Az^⊥}).” The type III part is Proposition 3.13, but the semifinite part does not follow from Proposition 3.11 as stated. Proposition 3.11 assumes a faithful normal semifinite weight φ with A⊂M^φ and concludes only m_φ⊂D(Ad_A). In the second bullet, φ is taken to be Tr_A∘E_A, where Tr_A is a faithful normal semifinite trace on A. For x∈m_{E_A}, E_A(x) is a bounded positive element of A, so x∈m_φ would require Tr_A(E_A(x))<∞. This is false in general when A is properly infinite: for A=B(H) with the usual trace, M=A, E_A=id, and x=1_A, we have 1_A∈m_{E_A}=M but Tr_A(1_A)=∞, so 1_A∉m_{Tr_A}. Thus Proposition 3.11 does not imply m_{E_A}⊂D. Corollary 3.5 only gives norm closedness of D, not σ-weak closedness, so one cannot simply approximate elements of m_{E_A} by elements of m_φ and pass to the limit. The introductory remark that the semifinite case follows by standard Hilbert space methods is not a substitute for the omitted argument: the proof that is actually written leaves the properly infinite semifinite case uncovered. This is an internal gap in the proof of the central Theorem A, independent of the deep external theorem [Ma23, Theorem 5.2] used for the type III part.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proves a localized weak relative Dixmier theorem for inclusions A⊂M admitting a faithful normal semifinite operator valued weight E_A:M→A: every positive x∈M with E_A(x)<∞ can be averaged, via the σ-weak closure of the convex hull of {uxu*:u∈U(A)}, into A'∩M. The proof is organized around the notion of weak Dixmier subspaces, decomposes A into semifinite and type III parts, and applies Marrakchi's compact semigroup machinery. The theorem is then used to derive several applications: a Popa intertwining criterion without tracial assumptions, relative solidity theorems for type III crossed products, and a Galois correspondence for crossed products by totally disconnected groups, resolving a question of Boutonnet and Brothier.","tokens_in":47786,"tokens_out":23594,"duration_ms":250963,"significance":"If fully established, the main theorem is a substantial generalization of Marrakchi's weak relative Dixmier theorem from conditional expectations to operator valued weights, and the applications are broad. The paper is well structured, states precise elementwise conditions (Theorems 3.3 and A), and makes honest and explicit use of published deep results ([Ma19], [Ma23, Theorem 5.2]). The proofs are largely self-contained after those inputs, with no hidden parameter fitting or circularity. However, the central proof has a gap in the semifinite properly infinite case that must be repaired before the applications can be regarded as consequences of the stated argument.","major_comments":[{"comment":"The proof does not cover the properly infinite semifinite summand. It invokes Propositions 3.11 and 3.13 to conclude m_{E_{A_z}}⊂D(Ad_{A_z}) and m_{E_{A_{z⊥}}}⊂D(Ad_{A_{z⊥}}). Proposition 3.11, in the second bullet, is applied with φ=Tr_{A_z}∘E_{A_z} and yields only m_φ⊂D(Ad_{A_z}). For a properly infinite A_z, m_{E_{A_z}} is not contained in m_φ: for A=B(H), M=A, E_A=id, and x=1, one has x∈m_{E_A} but Tr(1)=∞. Corollary 3.5(2) gives only norm closedness of D(α), not σ-weak closedness, so the missing inclusion cannot be obtained by approximation in the manner suggested by the phrase 'standard Hilbert space methods' in the introduction. Since Theorem 3.3 is the engine for Theorem A, Corollary B, and the applications in Sections 4–7, this is a load-bearing gap; a separate argument for the properly infinite semifinite case (or a reduction to the finite case) is required.","section":"§3.4 (Proof of Theorem 3.3)"}],"minor_comments":[{"comment":"The first paragraph uses the notation `\\tilde E`∈D(Ad_A:U(A)↷A) for what should be an element of DSG(Ad_A:U(A)↷M) extending E; D(α) is a subspace of the algebra, not a set of maps. Please correct the notation and explain why replacing Φ by Φ∘\\tilde E yields Φ|_A=E.","section":"§3.3 (Proposition 3.13)"},{"comment":"The displayed equality pK(x,Ad_A)p=K(pxp,Ad_{Ap})=K(pxp,Ad_{Ap}) contains a duplicated right-hand side; one occurrence should refer to the reduced action K(pxp,Ad_{A_p}).","section":"Lemma 3.8(1)"},{"comment":"The line 'As in the proof of [HI15a, Lemma 4.4], we can embed K(d,α)→L^2(M,φ) and the range is closed in the L^2-norm topology' is not evident from the text; since it is used to justify existence of the minimizer, a brief proof or a more precise reference to this closedness statement would improve readability.","section":"§3.2 (Lemma 3.10)"},{"comment":"The proof refers to 'reduced word in (B_0)_{a*}LZ' without defining the notation (B_0)_{a*} and LZ; please define these or add a reference.","section":"§5.3 (Lemma 5.11)"}],"recommendation":"major_revision","confidential_remarks":"This is a serious paper; conditional on the semifinite properly infinite gap being fixed, I would expect it to be accepted. The dependence on [Ma23, Theorem 5.2] is heavy but legitimate; the author should be encouraged to state precisely where that theorem is used and what would be needed to make the paper more self-contained."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis paper does two real things. It states a natural generalization of Marrakchi's weak relative Dixmier property: for an inclusion A ⊂ M with a faithful normal semifinite operator valued weight E_A, every positive x with E_A(x) < ∞ can be averaged by unitaries of A into A' ∩ M. And it shows this local averaging statement can drive Popa intertwining, relative solidity, and a Galois correspondence for totally disconnected group actions, including type III_0 cases. If Theorem A stands, the applications are genuinely new and the paper is important.\n\nWhat is genuinely new: the operator valued weight version is not in Marrakchi's paper, and the author correctly identifies that the averaging map need no longer be faithful. Proposition 3.13, where he studies the ucp map on m_{E_A}, is a real technical contribution. I checked the main inequalities there; they are coherent. The applications are not corollaries of earlier work, since previous statements assumed conditional expectations and usually finite targets.\n\nThe soft spot is in the proof of Theorem 3.3, as written. Proposition 3.11 only gives m_φ ⊂ D(Ad_A) for φ = Tr_A ∘ E_A, but Theorem 3.3 needs m_{E_A} ⊂ D. These sets differ when A is properly infinite: take A = M = B(H), E_A = id, x = 1. Then x ∈ m_{E_A} but Tr_A(1) = ∞. Corollary 3.5 gives norm closedness of D, not σ-weak closedness, so the trace-class approximation does not close. This is a load-bearing gap, since Theorem A is the engine for the rest of the paper. My guess is the statement is true and the gap is fixable with a separate Hilbert-space argument for properly infinite semifinite A, but a referee needs to see that argument before the applications can be trusted.\n\nMinor things: the first line of Proposition 3.13 has a typo (the extension should be in DSG, not D). Lemmas 5.3 and 5.4 are delegated to [Ma16]/[Ma23] as minor modifications but are load-bearing for Theorem D; they should be checked. The type III half of Theorem A inherits the deep [Ma23, Theorem 5.2], which is a dependency to be aware of rather than a flaw.\n\nThis is a paper for people working in type III factors, Popa intertwining, and rigidity. It deserves a serious referee. I would not desk reject it; I would send it out and ask for a fix to the semifinite case.","headline":"Significant and likely-true generalization of the weak relative Dixmier property to operator valued weights, but the written proof of Theorem A has a genuine gap in the semifinite case that needs a separate argument.","tokens_in":48486,"tokens_out":5312,"would_cite":false,"duration_ms":56550,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46L10","46L36","46L40"],"pacs":[],"model":"deepseek-v4-flash","headline":"Averaging unitary orbits hits the relative commutant, no trace needed","keywords":["weak relative Dixmier property","operator valued weights","type III factors","intertwining-by-bimodules","relative solidity","Galois correspondence","weak Dixmier semigroup","crossed products"],"falsifier":"Take any genuinely type III subalgebra $A$ inside a larger von Neumann algebra $M$ with a faithful normal semifinite operator valued weight that is not a conditional expectation, choose a positive $x$ with $E_A(x)<\\infty$, and compute whether the $\\sigma$-weak closure of $\\mathrm{conv}\\{uxu^*\\mid u\\in\\mathcal U(A)\\}$ intersects $A'\\cap M$; if it misses $A'\\cap M$, Theorem A is false, and the paper predicts this cannot happen.","tokens_in":47108,"feed_emoji":"🔁","tokens_out":8389,"duration_ms":86505,"temperature":0.7,"pith_summary":"This paper proves a localized averaging theorem for inclusions of von Neumann algebras: whenever a faithful normal semifinite operator valued weight exists from a larger algebra to a subalgebra, every positive element of finite weight can be moved by averages of unitary conjugations from the subalgebra into the relative commutant. The result extends the weak relative Dixmier property from the conditional-expectation setting to the more flexible operator valued weight setting, where no expectation need exist. The paper then uses this elementwise criterion to reformulate the intertwining-by-bimodules technique without any trace or finite subalgebra assumption, to obtain relative solidity for crossed products by biexact groups, and to prove a Galois correspondence for crossed products by totally disconnected groups, resolving a question about intermediate subfactors.","feed_headline":"Averaging unitary orbits hits the relative commutant, no trace needed","feed_subtitle":"Extends the weak Dixmier property to operator valued weights, powering new type III rigidity and Galois theorems.","key_machinery":"The engine is the weak Dixmier semigroup $\\mathrm{DSG}(\\mathrm{Ad}_A)$, the compact convex semigroup of completely positive maps obtained by closing convex combinations of unitary conjugations $u\\cdot u^*$ with $u\\in\\mathcal U(A)$ in the point-$\\sigma$-weak topology. Inside it, the proof isolates a minimal idempotent $\\Phi$ whose restriction to $A$ is the expectation onto the center, then shows $\\Phi$ acts like the identity on its own image inside the finite-weight domain $m_{E_A}$; a parallelogram-law argument forces that image to consist of elements commuting with $A$. The operator valued weight $E_A$ plays the role of keeping the finite-weight domain invariant under averaging, substituting for the trace or conditional expectation that earlier arguments required.","core_discovery":"The central claim is Theorem A: for an inclusion $A\\subset M$ with a faithful normal semifinite operator valued weight $E_A:M\\to A$, every positive $x\\in M$ with $E_A(x)<\\infty$ satisfies the weak relative Dixmier property in the strong sense that the $\\sigma$-weak closure of the convex hull $K(x,A)=\\mathrm{conv}^{\\mathrm{weak}}\\{uxu^*\\mid u\\in \\mathcal U(A)\\}$ intersects $A'\\cap M$. In its precise form, the paper shows $m_{E_A}\\subset D(A\\subset M)$: every element on which the weight is finite lies in the weak Dixmier subspace, so for every averaging map one can find nonempty fixed points of the convex hull of its unitary orbit. In particular, there is a ucp map $\\Psi$, obtained as a point-$\\sigma$-weak limit of convex combinations of conjugations by unitaries of $A$, with $\\Psi(m_{E_A})\\subset A'\\cap M$. This elementwise, weight-finite version is the engine that makes the later applications possible.","pith_inferences":["Beyond the paper, the elementwise formulation suggests a quantitative version: the speed at which convex combinations of conjugations approach the relative commutant may be controlled by the value of $E_A(x)$, which could give uniform approximation bounds for whole classes of elements.","Since the intertwining reformulation needs no finiteness of $A$ or $B$, it should extend rigidity arguments to free Araki-Woods factors and other non-tracial constructions where earlier criteria stalled.","The Galois correspondence for totally disconnected groups may hold under weaker proper-outerness hypotheses than the one stated; the exact boundary of the condition is a testable extension.","The same averaging machinery may yield a relative bicentralizer computation for inclusions with operator valued weights, paralleling the conditional-expectation bicentralizer results.","A concrete search for counterexamples could target the basic construction of a genuinely type III subfactor: if any positive element of finite weight has a convex hull of unitarily conjugated elements missing the relative commutant, Theorem A would be false."],"forward_implications":["Popa's intertwining criterion $A\\preceq_M B$ is characterized, for type III algebras or a properly infinite $B$, by the absence of a ucp map on the basic construction that annihilates the Jones projection, with no tracial assumption needed.","Bernoulli crossed products by actions with finite stabilizers are solid relative to the base inclusion, giving a spectral gap rigidity theorem for type III factors.","Crossed products by biexact group actions on amenable von Neumann algebras are solid relative to the base algebra, even when the base is type III.","For properly outer actions of totally disconnected groups, every intermediate subfactor between $B$ and $B\\rtimes G$ is of the form $B\\rtimes H$ for a closed subgroup $H\\le G$, settling a previously open question.","Corollary B gives a dichotomy: if every nonzero positive element of $A'\\cap M$ has infinite $E_A$-value, then every positive element of finite weight can be averaged to $0$ inside the convex hull of its unitary orbit."],"supporting_citations":[{"why":"Supplies the weak Dixmier semigroup and minimal-idempotent method, and proves the weak relative Dixmier property for conditional expectations, which this paper extends to operator valued weights.","marker":"[Ma19]"},{"why":"Provides the faithful normal expectation of a type III algebra onto its center lying in the averaging semigroup, a step imported and extended in Proposition 3.13.","marker":"[Ma23]"},{"why":"Introduces the intertwining-by-bimodules criterion whose type III reformulation is Theorem C.","marker":"[Po01]"},{"why":"Basis for the spectral gap rigidity and malleable deformation arguments generalized in Theorem D.","marker":"[Po03]"},{"why":"Defines solidity and proves the original solid von Neumann algebra theorem that Theorem E extends to type III targets.","marker":"[Oz03]"},{"why":"Supplies relative solidity for crossed products by actions on abelian algebras, the setting generalized to arbitrary amenable base algebras.","marker":"[Oz04]"},{"why":"Poses the Galois-correspondence question for totally disconnected groups and provides the method and partial results that Theorem F completes.","marker":"[BB17]"},{"why":"Supplies the biexact von Neumann algebra framework and boundary-piece construction used in the proof of Theorem E.","marker":"[DP23]"},{"why":"Provides type III intertwining formulations and basic construction notation on which Theorem 4.2 builds.","marker":"[Is19]"}],"fun_headline_variants":["Weak Dixmier property now holds for operator-valued weights","Operator-valued weights extend weak Dixmier and type III tools","Popa intertwining without traces via weak Dixmier averaging","Type III relative solidity and Galois via weight Dixmier","Weak Dixmier property under weights yields type III results"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof's load-bearing premise is that a type III algebra admits a faithful normal expectation onto its center that can be obtained as a limit of unitary conjugations; if that deep fact fails for the subalgebra in question, or the extension of such an expectation to the whole algebra cannot be made minimal, the main theorem is not established.","fun_headline_variants_meta":{"raw":{"variants":["Weak Dixmier property now holds for operator-valued weights","Operator-valued weights extend weak Dixmier and type III tools","Popa intertwining without traces via weak Dixmier averaging","Type III relative solidity and Galois via weight Dixmier","Weak Dixmier property under weights yields type III results"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001014,"raw_usage":{"total_tokens":4283,"prompt_tokens":948,"completion_tokens":3335,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":564,"completion_tokens_details":{"reasoning_tokens":3255}},"tokens_in":564,"tokens_out":3335,"duration_ms":24560,"temperature":1.0,"reasoning_tokens":3255,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T17:06:18.010048+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take any genuinely type III subalgebra $A$ inside a larger von Neumann algebra $M$ with a faithful normal semifinite operator valued weight that is not a conditional expectation, choose a positive $x$ with $E_A(x)<\\infty$, and compute whether the $\\sigma$-weak closure of $\\mathrm{conv}\\{uxu^*\\mid u\\in\\mathcal U(A)\\}$ intersects $A'\\cap M$; if it misses $A'\\cap M$, Theorem A is false, and the paper predicts this cannot happen.","supporting_citations":[],"review_version":2}