{"id":"ba7d95a4-8672-48fb-9697-3858f694fe52","arxiv_id":"2508.18978","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every G-invariant intermediate von Neumann algebra between N and N⊗̄L∞(Poisson boundary) splits as N⊗̄L∞(C) for a (G,μ)-boundary C, when G is a product of two groups and each factor acts ergodically on N.","lead":"This paper proves a noncommutative splitting theorem: for a product of two groups acting on a tracial von Neumann algebra, every intermediate invariant subalgebra between the algebra and its Poisson-boundary extension must be a tensor product. It generalizes a classical measurable rigidity result and settles an open problem about maximal Haagerup subalgebras.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.1 hinges on N^{Gi}=C1: without it the Radon–Nikodym derivatives need not land in L∞(B_i), and the abstract's unqualified claim is not proved.","rationale":"The reader's weakest_assumption correctly identifies the Gi-ergodicity hypothesis as the point where the proof of Theorem 3.1 is most fragile. The containment MRN⊂L∞(B) is the pivot of the argument, and it is obtained by showing h_{g1}^{it}∈(N⊗̄L∞(B))^{G2}=L∞(B1)⊗1. Recomputing the fixed-point algebra gives L∞(B1,N^{G2}), so N^{G2}=C1 is exactly what turns this into L∞(B1). Without it the proof cannot proceed. I also checked the rest of the proof: the construction of the equivariant homomorphism Φ, the use of Lemma 2.3 to identify P with P0, and the final multiplicative-domain argument all appear internally consistent under the stated hypotheses. Section 4 does not actually demonstrate the necessity of the ergodicity assumption, because its non-splitting algebras arise from invariant ideals in a general tensor product, while the Poisson boundary setting of Theorem 3.1 is ergodic and has no such ideals; I therefore do not treat the reader's supporting remark as established, but the core concern stands. Since the theorem as stated in the body includes the ergodicity assumption and the proof supports it, the existing CONDITIONAL verdict remains appropriate, with the caveat that the abstract should be amended to include the hypothesis or the theorem should be extended.","tokens_in":22236,"tokens_out":54271,"duration_ms":503942,"concrete_test":"Work in the simplest case where the ergodicity hypothesis fails: take G2={e} (so B2={pt} and N^{G2}=N), let G1 be a non-amenable group with non-trivial Poisson boundary B1, and let N=L∞(X) be a nontrivial G1-ergodic abelian algebra. Determine whether every G1-invariant intermediate von Neumann algebra M with N⊂M⊂N⊗̄L∞(B1) is of the form N⊗̄L∞(C) for a (G1,μ1)-boundary C. If a non-splitting example exists, it proves the ergodicity assumption is genuinely necessary and the abstract's unqualified claim is false; if all such M split, the proof should be revisable by replacing N with N^{G1}∩N^{G2} in the tensor factorization.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 3.1 has a single load-bearing step: immediately after Lemma 2.4 it needs (N⊗̄L∞(B))^{G2} = L∞(B1)⊗1, obtained via L∞(B1,N)^{G2}=L∞(B1,N^{G2})=L∞(B1,C)=L∞(B1). The first equality is unconditional, but the second uses N^{G2}=C1. If N is only G-ergodic for the full group but not G2-ergodic (say G2 acts trivially on a non-scalar part of N), then h_{g1}^{it} is only forced into L∞(B1,N^{G2}), not L∞(B1), so the G-Radon–Nikodym factor MRN need not be contained in L∞(B) and the whole reconstruction of M as N⊗̄L∞(C) collapses. The abstract advertises the theorem without the G_i-ergodicity hypothesis, so the advertised claim is strictly stronger than what is proved. Section 4 does not settle the necessity of the hypothesis: its non-splitting examples live in M⊗̄N with an invariant ideal, whereas in the Poisson-boundary setting of Theorem 3.1 there are no non-trivial G-invariant ideals (the boundary is ergodic). Thus the assumption is load-bearing for the proof, and whether the theorem can be extended to remove it is open.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a noncommutative analogue of the Bader--Shalom Intermediate Factor Theorem. For G = G1 × G2 with an admissible product measure μ, Poisson boundary B, and a trace-preserving G-von Neumann algebra N that is G_i-ergodic for each i, Theorem 3.1 asserts that every G-invariant intermediate algebra M with N ⊂ M ⊂ N ⊗̄ L∞(B, ν_B) splits as N ⊗̄ L∞(C, ν_C) for some (G, μ)-boundary (C, ν_C). The proof introduces a G-Radon–Nikodym factor M_RN, shows that it lands in the commutative boundary algebra L∞(B), and then constructs a trace-preserving G-equivariant ∗-homomorphism from M onto N ⊗̄ L∞(C) that is the identity on both tensor legs. The paper also establishes auxiliary splitting results (Section 2), shows that invariant ideals in tensor products obstruct splitting (Section 4), and applies a master theorem of Glasner–Weiss to classify intermediate factors for SL2(Z)-actions and to prove a maximal Haagerup subalgebra result (Section 5).","tokens_in":22592,"tokens_out":35671,"duration_ms":363228,"significance":"If the main theorem is correct, it is a substantial contribution: it gives a genuine W∗-version of the Bader–Shalom IFT with a new tool, the equivariant G-Radon–Nikodym factor, and it addresses a natural question about intermediate von Neumann algebras in boundary extensions. The proof is largely self-contained and has a coherent structure: metric ergodicity of the Poisson boundary and the G_i-ergodicity of N are used in a transparent way to force the Radon–Nikodym derivatives into the two boundary legs. The Section 5 application to maximal Haagerup subalgebras is a nice payoff and appears to use the Glasner–Weiss theorem appropriately. The significance is tempered, however, by the fact that the abstract advertises a stronger statement than Theorem 3.1 actually proves, and the missing ergodicity hypothesis is load-bearing for the proof.","major_comments":[{"comment":"The abstract claims the splitting theorem for every trace-preserving G-von Neumann algebra N, but Theorem 3.1 assumes N is G_i-ergodic for i = 1, 2. This hypothesis is used exactly at the line \"Since G2 acts trivially on B1 and ergodically on N, we have L∞(B1,N)^{G2} = L∞(B1,N^{G2}) = L∞(B1,C) = L∞(B1)\". Without N^{G2} = C1, the element h_{g1}^{it} is only shown to lie in L∞(B1,N^{G2}), not in L∞(B1); hence M_RN is not proved to be contained in L∞(B), and the reconstruction of M as N ⊗̄ L∞(C) collapses. Section 4 does not fill this gap: its non-splitting algebras are of the form M + qP inside a tensor product with an invariant ideal, whereas the Poisson-boundary algebra in Theorem 3.1 is ergodic and has no nontrivial invariant ideals. The abstract must be amended to include the G_i-ergodicity hypothesis, or the theorem must be proved without it.","section":"Abstract; Theorem 3.1, proof"},{"comment":"The proof of Proposition 2.8 fixes a countable dense subgroup G0 and asserts M_RN = ⟨h_g^{it}⟩_{g∈G} = ⟨h_{g0}^{it}⟩_{g0∈G0}, citing [30, Proposition 1.14]. This equality requires a continuity or measurability property of the action g ↦ α_g, or of g ↦ h_g. The paper never states the continuity convention for a \"G-von Neumann algebra\". Since Theorem 1.2, and through it the proof of Theorem 3.1, rests on Proposition 2.8, please state the standing assumption on the action and justify the dense-subgroup reduction.","section":"Section 2, Proposition 2.8"},{"comment":"The final identification of the C-leg uses the unstated [37, Lemma 2.2] to pass from Ψ = (τ_N⊗id)∘Φ|_{L∞(C)} = id to Φ(L∞(C)) ⊂ 1⊗L∞(C). This is a load-bearing step, and [37] is an arXiv preprint. Please state the lemma, or give a proof, and verify that its hypotheses are satisfied by Φ in the present setting.","section":"Section 3, proof of Theorem 3.1"}],"minor_comments":[{"comment":"The text contains the typo \"∗-homorphism\"; it should be \"∗-homomorphism\".","section":"Section 4, proof of Theorem 4.3"},{"comment":"The proof says \"Using Theorem 4.3\", but both legs are abelian in Proposition 4.4; the invoked conclusion is the symmetric case of Proposition 4.2, not Theorem 4.3 itself.","section":"Section 4, Proposition 4.4"},{"comment":"In the sentence \"for some G-factor map G ↷ (X,ν)→(Z,ζ)\", the target of the action should be (X,μ), not (X,ν).","section":"Section 5, Corollary 5.2"},{"comment":"The statement of Proposition 2.13 contains a stray minus in \"E_N^T : N− → T\"; it should read \"E_N^T : N → T\".","section":"Section 2, Proposition 2.13"},{"comment":"The term \"(G,μ)-boundary\" is used in Theorem 3.1 and elsewhere without a definition; please add the definition, presumably as a G-space with a μ-stationary measure.","section":"Sections 2 and 3"}],"recommendation":"major_revision","confidential_remarks":"The main theorem appears to be correct under the stated G_i-ergodicity hypothesis, and the proof strategy is novel and interesting. However, the abstract overstates the result, and the proof relies on an unpublished lemma ([37, Lemma 2.2]) at a decisive point. The Section 4 examples show limitations of splitting in a different context but do not settle whether the G_i-ergodicity assumption in Theorem 3.1 is necessary; the authors should explicitly acknowledge this open point. These issues are fixable within the manuscript's scope, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Put the theorem's actual hypothesis in the abstract and this is a solid piece. The G-Radon-Nikodym factor with its equivariant conditional expectation is genuinely useful, not a repackaging. Theorem 3.1 is a real lift of Bader-Shalom to tracial von Neumann algebras, and the SL2(Z) application answers the Jiang-Skalski question. The central proof is coherent: the Radon-Nikodym derivatives land in the boundary legs, the boundary C is built from M itself, and the final *-homomorphism is forced to be the identity on both legs. There are no free parameters and no invented entities. Section 2's framework is careful and should be reusable.\n\nWhere the paper is softer: the abstract is stronger than the theorem. Theorem 3.1 assumes N is G_i-ergodic for each factor, and that assumption is load-bearing. When it fails, h_{g1}^{it} is only known to land in L^∞(B1, N^{G2}), not L^∞(B1), so M_RN need not sit inside L^∞(B) and the reconstruction collapses. Section 4 shows invariant ideals obstruct splitting in a different ambient algebra, but that does not settle whether the ergodicity assumption can be removed in the Poisson-boundary setting. So the advertised unqualified claim is not proved. This should be fixed: either state the theorem with its hypothesis and make the abstract match, or prove the stronger claim. Minor issue: Proposition 4.4 cites Theorem 4.3 where it should cite Proposition 4.2.\n\nWould I referee it? Yes. The main theorem is new and the method is reusable. The proof is coherent enough that referee time is well spent, and the issues are fixable with honest presentation rather than structural failure.","headline":"Solid NC-IFT under the stated ergodicity assumptions; the abstract just needs to state them.","tokens_in":23113,"tokens_out":1473,"would_cite":true,"duration_ms":15889,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46L10","46L55","37A15","22D25"],"pacs":[],"model":"deepseek-v4-flash","headline":"Product-group actions force intermediate von Neumann algebras to split.","keywords":["Intermediate Factor Theorem","von Neumann algebras","Poisson boundary","product groups","Radon-Nikodym factor","ergodic actions","Haagerup property","tensor product splitting"],"falsifier":"A single example satisfying all hypotheses of Theorem 3.1 — $G=G_1\\times G_2$, $\\mu=\\mu_1\\times\\mu_2$, $\\mathcal N$ tracial and $G_i$-ergodic — but with a $G$-invariant intermediate algebra $\\mathcal M$ not of the form $\\mathcal N\\,\\bar{\\otimes}\\,L^\\infty(C)$ would refute the theorem. Concretely, one can test the proof's key containment $\\mathcal M_{\\mathrm{RN}}\\subseteq L^\\infty(B,\\nu_B)$: if for some $\\mathcal M$ the element $h_{g_1}^{it}$ is not in $L^\\infty(B_1,\\nu_{B_1})$, the splitting argument fails, and such an $\\mathcal M$ would be a counterexample.","tokens_in":21959,"feed_emoji":"🧩","tokens_out":12580,"duration_ms":112295,"temperature":0.7,"pith_summary":"The paper proves a noncommutative analogue of the Intermediate Factor Theorem for product groups. If $G=G_1\\times G_2$ acts trace-preservingly on a tracial von Neumann algebra $\\mathcal N$ that is ergodic for each factor separately, and if $(B,\\nu_B)$ is the Poisson boundary (a compact boundary space encoding the long-term behaviour of the random walk driven by $\\mu$), then every $G$-invariant von Neumann algebra $\\mathcal M$ with $\\mathcal N\\subset \\mathcal M\\subset \\mathcal N\\,\\bar{\\otimes}\\, L^\\infty(B,\\nu_B)$ must split as $\\mathcal N\\,\\bar{\\otimes}\\, L^\\infty(C,\\nu_C)$ for some boundary $(C,\\nu_C)$. The upshot is that boundary rigidity transfers from classical measurable dynamics into the noncommutative world: intermediate algebras cannot be exotic. The paper also shows that invariant ideals can obstruct such splitting, and uses the theorem to settle a problem about maximal Haagerup subalgebras (subalgebras with the Haagerup property, a weak form of amenability) for $\\mathrm{SL}_2(\\mathbb Z)$ actions.","feed_headline":"Product-group actions force intermediate von Neumann algebras to split","feed_subtitle":"No exotic intermediate subalgebras: every invariant one is N tensored with a boundary factor.","key_machinery":"The central object is the $G$-Radon-Nikodym factor $\\mathcal M_{\\mathrm{RN}}=\\langle h_g^{it}\\rangle_{g\\in G,t\\in\\mathbb R}$, the von Neumann algebra generated by the imaginary powers of the Radon-Nikodym derivatives $h_g=(D(\\tau\\circ g):D\\tau)$ of the action. This factor plays the role of an invariant that, for the ambient algebra $\\mathcal N\\,\\bar{\\otimes}\\, L^\\infty(B,\\nu_B)$, collapses to $L^\\infty(B,\\nu_B)$, and for $\\mathcal N$ collapses to scalars. The proof combines three tools: metric ergodicity of the Poisson boundary, which forces $G_i$-invariant elements in $\\mathcal N\\,\\bar{\\otimes}\\, L^\\infty(B_i)$ to lie in $L^\\infty(B_i)$; the bijection between $\\mu$-stationary states and equivariant completely positive maps into $L^\\infty(B,\\nu_B)$; and the chain rule for Radon-Nikodym derivatives, which lets the derivatives for product-group elements factor into the two boundary legs.","core_discovery":"The central result, Theorem 3.1, asserts that under $G_i$-ergodicity of $\\mathcal N$, every $G$-invariant intermediate von Neumann algebra $\\mathcal M$ splits as $\\mathcal N\\,\\bar{\\otimes}\\, L^\\infty(C,\\nu_C)$. The proof isolates the $G$-Radon-Nikodym factor $\\mathcal M_{\\mathrm{RN}}$ of $\\mathcal M$, shows it is contained in $L^\\infty(B,\\nu_B)$, identifies it with $L^\\infty(C,\\nu_C)$ for a boundary, and then builds a trace-preserving $G$-equivariant $*$-homomorphism $\\Phi:\\mathcal M\\to \\mathcal N\\,\\bar{\\otimes}\\, L^\\infty(C,\\nu_C)$ that is the identity on both $\\mathcal N$ and $L^\\infty(C)$, forcing $\\mathcal M=\\mathcal N\\,\\bar{\\otimes}\\, L^\\infty(C,\\nu_C)$. Along the way the authors prove that the conditional expectation onto the $G$-Radon-Nikodym factor is always $G$-equivariant, even when the trace is not invariant. The paper further establishes that invariant ideals in tensor products of an abelian and a non-commutative factor generate intermediate algebras that do not split, and classifies all intermediate factors for the $\\mathrm{SL}_2(\\mathbb Z)$ Bernoulli shift times torus action.","pith_inferences":["A natural extension is to classify intermediate algebras for trace-scaling actions, where the Radon-Nikodym factor may carry modular data rather than only boundary coordinates.","The obstruction theorem suggests a recipe for building non-splitting intermediate algebras from any $G$-invariant central projection in a tensor product, possibly yielding a full classification in terms of such projections.","A testable strengthening would replace $G_i$-ergodicity by relative ergodicity over a smaller subalgebra; if splitting still holds with a relative boundary, the method would extend to non-ergodic coefficients.","The $\\mathrm{SL}_2(\\mathbb Z)$ application exploits an entropy gap between the Bernoulli leg and the distal torus leg, so the same template may prove maximal Haagerup rigidity for other group actions with a similar gap."],"forward_implications":["Every $G$-invariant intermediate algebra between $\\mathcal N$ and $\\mathcal N\\,\\bar{\\otimes}\\,L^\\infty(B)$ is a tensor product $\\mathcal N\\,\\bar{\\otimes}\\,L^\\infty(C)$ with $C$ a $(G,\\mu)$-boundary; there are no exotic intermediate von Neumann algebras under the hypotheses.","The $G$-Radon-Nikodym factor admits a $G$-equivariant trace-preserving conditional expectation even when the trace is not $G$-invariant, so the same splitting method is available beyond measure-preserving actions.","Invariant ideals form a genuine obstruction: if one factor is abelian and the other non-commutative, the intermediate algebra generated by an invariant ideal fails to split, showing the ergodicity hypotheses in the main theorem are not superfluous.","For $G=\\mathrm{SL}_2(\\mathbb Z)$, every intermediate factor of the Bernoulli shift times the standard torus action splits as a product, which makes $L^\\infty(Y)\\rtimes G$ a maximal Haagerup subalgebra of the full crossed product.","The affirmative resolution of the second part of a previously posed problem on maximal Haagerup subalgebras follows as a corollary."],"supporting_citations":[{"why":"the measurable Intermediate Factor Theorem for product groups, whose proof strategy is adapted to von Neumann algebras.","marker":"[8]"},{"why":"introduces the Radon-Nikodym factor for nonsingular group actions, the classical template for the G-Radon-Nikodym factor.","marker":"[30]"},{"why":"defines the Radon-Nikodym factor for W*-inclusions, used to justify the noncommutative version and its crossed-product form.","marker":"[36]"},{"why":"proves metric ergodicity of the Poisson boundary, the property that forces invariant elements into the base algebra.","marker":"[7]"},{"why":"provides the chain rule for Radon-Nikodym derivatives and uniqueness of trace-preserving conditional expectations, used throughout the proof of Theorem 3.1.","marker":"[35]"},{"why":"contains the argument used as Lemma 2.4 that equivariant maps from a metrically ergodic space into a tracial algebra are essentially constant.","marker":"[6]"},{"why":"the Master theorem classifying intermediate factors of products of disjoint systems, applied in the SL2(Z) section.","marker":"[20]"},{"why":"poses the problem on maximal Haagerup subalgebras that the corollary answers.","marker":"[25]"},{"why":"describes intermediate operator algebras in crossed products by free actions, reducing the maximal-Haagerup claim to the factor classification.","marker":"[33]"}],"fun_headline_variants":["Every invariant intermediate algebra splits for product-group actions","Splitting theorem: product-group actions force intermediate splits","Product-group actions: invariant intermediate algebras always split","Splitting phenomenon for all invariant intermediate algebras","All invariant intermediate algebras split under product groups"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that $\\mathcal N$ is ergodic for each factor $G_i$ separately; without it the Radon-Nikodym derivatives of the action can live inside $\\mathcal N$ instead of the Poisson boundary, and the paper's Section 4 shows invariant ideals then produce intermediate algebras that do not split.","fun_headline_variants_meta":{"raw":{"variants":["Every invariant intermediate algebra splits for product-group actions","Splitting theorem: product-group actions force intermediate splits","Product-group actions: invariant intermediate algebras always split","Splitting phenomenon for all invariant intermediate algebras","All invariant intermediate algebras split under product groups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001225,"raw_usage":{"total_tokens":5123,"prompt_tokens":1118,"completion_tokens":4005,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":734,"completion_tokens_details":{"reasoning_tokens":3933}},"tokens_in":734,"tokens_out":4005,"duration_ms":27017,"temperature":1.0,"reasoning_tokens":3933,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:55:11.779894+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A single example satisfying all hypotheses of Theorem 3.1 — $G=G_1\\times G_2$, $\\mu=\\mu_1\\times\\mu_2$, $\\mathcal N$ tracial and $G_i$-ergodic — but with a $G$-invariant intermediate algebra $\\mathcal M$ not of the form $\\mathcal N\\,\\bar{\\otimes}\\,L^\\infty(C)$ would refute the theorem. Concretely, one can test the proof's key containment $\\mathcal M_{\\mathrm{RN}}\\subseteq L^\\infty(B,\\nu_B)$: if for some $\\mathcal M$ the element $h_{g_1}^{it}$ is not in $L^\\infty(B_1,\\nu_{B_1})$, the splitting argument fails, and such an $\\mathcal M$ would be a counterexample.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"the measurable Intermediate Factor Theorem for product groups, whose proof strategy is adapted to von Neumann algebras."},{"cited_title":"Zimmer, Rigidity of Furstenberg entropy for semisimple Lie group actions, Ann","cited_arxiv_id":null,"evidence_quote":"introduces the Radon-Nikodym factor for nonsingular group actions, the classical template for the G-Radon-Nikodym factor."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"defines the Radon-Nikodym factor for W*-inclusions, used to justify the noncommutative version and its crossed-product form."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"proves metric ergodicity of the Poisson boundary, the property that forces invariant elements into the base algebra."},{"cited_title":"II , Encyclopaedia of Mathematical Sciences, vol","cited_arxiv_id":null,"evidence_quote":"provides the chain rule for Radon-Nikodym derivatives and uniqueness of trace-preserving conditional expectations, used throughout the proof of Theorem 3.1."},{"cited_title":"3, 929–985","cited_arxiv_id":null,"evidence_quote":"contains the argument used as Lemma 2.4 that equivariant maps from a metrically ergodic space into a tracial algebra are essentially constant."},{"cited_title":"On intermediate factors of a product of disjoint systems","cited_arxiv_id":"2312.03329","evidence_quote":"the Master theorem classifying intermediate factors of products of disjoint systems, applied in the SL2(Z) section."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"poses the problem on maximal Haagerup subalgebras that the corollary answers."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"describes intermediate operator algebras in crossed products by free actions, reducing the maximal-Haagerup claim to the factor classification."}],"review_version":2}