{"id":"ff3c9359-4bb6-487f-83cb-f75e9da800b1","arxiv_id":"1909.01633","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A transfer trick yields non-vanishing top ℓ2-Betti numbers for Out(Fn), Aut(Fn) and Torelli groups, vanishing results for subgroups of 3-manifold groups, and ergodic dimension d+1 for F2^d × Z.","lead":"Gaboriau and Noûs prove a transfer principle: if a subgroup of a group has a non-zero top-dimensional ℓ2-Betti number on an action, the whole group has one too. This yields new non-vanishing results for Out(Fn) and Aut(Fn), vanishing results for subgroups of 3-manifold groups, and exact ergodic dimensions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The manifold results rest on unpublished [CGMT]; the §7 fallback for Theorem 1.5 says a non-cocompact H3 lattice has geometric dimension ≤ 2, which is impossible (vcd = 3), so Theorem 1.5 is not yet independent of [CGMT].","rationale":"The reader's weakest_assumption correctly identifies the [CGMT] dependence as load-bearing. I agree with that, but I also find that the paper's own escape hatch for Theorem 1.5 is compromised: the §7 alternative asserts without proof that a non-cocompact lattice in Isom(H3) has geometric dimension ≤ 2, which is false as written because such lattices have virtual cohomological dimension 3. Replacing 'geometric' by 'ergodic' would make the claim plausible, but then the alternative is not independent of [CGMT], since the ergodic-dimension bound for these lattice vertex groups is precisely the kind of input [CGMT] was meant to supply. The main transfer theorems and the poly-free computations have independent support in the literature and appear checkable; I do not see a gap in the Out/ Aut/Torelli core. Because the paper itself emphasizes the distinction between ergodic and geometric dimension, this slip is not cosmetic: it marks the exact place where the manifold theorem relies on an external, unpublished bound. Keeping the CONDITIONAL verdict is therefore appropriate; the condition is that [CGMT] be published or that the H3 lattice claim in §7 be repaired and proved.","tokens_in":14771,"tokens_out":18215,"duration_ms":176001,"concrete_test":"Take a non-cocompact lattice Γ'_i ⊂ Isom(H3) and a torsion-free finite-index subgroup Γ_o. Compute cd(Γ_o): since Γ_o acts freely on H3, cd ≤ 3; because the quotient is a finite-volume hyperbolic 3-manifold, cd = 3. Hence vcd(Γ'_i) = 3 and no proper action on a contractible 2-complex exists. If the §7 line means this literally, the alternative proof of Theorem 1.5 fails; if it means ergodic dimension, require an explicit argument or citation proving erg dim ≤ 2 for these vertex groups before accepting the theorem.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central transfer trick (Theorems 1.8 and 5.1) is internally coherent, and the Out(F_n)/Aut(F_n)/Torelli applications follow from the cited poly-free constructions without the disputed input. The load-bearing weakness is the manifold section. Theorem 1.4 is proved by a one-line appeal to the unpublished preprint [CGMT] (ergodic dimension ≤ d−1 for compact aspherical d-manifold groups) followed by Theorem 1.6; as long as [CGMT] is unavailable, that theorem is not verifiable. The last paragraph of §7 attempts an alternative for Theorem 1.5, but it contains a false premise: it states that a non-cocompact lattice Γ'_i in Isom(H3) has geometric dimension ≤ 2. Such a lattice is virtually torsion-free with cohomological dimension 3 (a torsion-free finite-index subgroup is the fundamental group of a finite-volume hyperbolic 3-manifold), so any proper action on a contractible complex has dimension at least 3. Unless 'geometric dimension' is a slip for 'ergodic dimension', the assertion is inconsistent; if it is a slip, then the needed ergodic-dimension bound for the H3 vertex groups is exactly what [CGMT] was supposed to prove, so the alternative still depends on the same unpublished result. Theorem 1.5 therefore remains conditional.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a transfer principle for top-dimensional ℓ2-Betti numbers: if a subgroup (or sub-equivalence relation) has non-vanishing top-degree ℓ2-Betti number for a proper action (or a discrete complex), then the ambient group (or equivalence relation) has non-vanishing top-degree ℓ2-Betti number as well; equivalently, vanishing in the top degree for the ambient object forces vanishing for all subobjects. This is proved in a topological version (Theorem 1.8), a measured version for equivalence relations (Theorem 5.1), and applied in three directions: non-vanishing for Out(F_n), Aut(F_n), and their Torelli subgroups in their virtual cohomological dimensions; vanishing of β_2 and β_3 for all subgroups of fundamental groups of compact 3-manifolds; and computation of the ergodic dimension of products Λ×B where B is infinite amenable. The paper is concise and mostly clearly written, but the manifold applications rely on the unpublished preprint [CGMT], and the alternative argument offered in §7 contains a false geometric-dimension assertion.","tokens_in":15029,"tokens_out":11122,"duration_ms":107942,"significance":"The transfer principle is a genuinely useful idea: it is simple, self-contained given standard ℓ2-invariant machinery, and it converts subgroup non-vanishing into ambient non-vanishing without coamenability or finite-index assumptions. The applications to Out(F_n), Aut(F_n), and the Torelli subgroups are explicit and do not depend on unpublished work; the poly-free subgroup constructions are concrete and the computations using the cited results of Lück and Sauer–Thom appear correct. If the manifold theorems were fully established, they would also be significant, and the ergodic-dimension corollaries are nice. However, the manuscript currently does not provide a verifiable unconditional proof of Theorems 1.4 and 1.5: Theorem 1.4 is a direct appeal to an unpublished preprint, and the alternative proof in §7 rests on a false geometric-dimension claim. The core transfer results and the automorphism-group applications are sound, but the advertised manifold applications need substantial repair.","major_comments":[{"comment":"Theorem 1.4 is proved by a one-line appeal to the unpublished preprint [CGMT] for the assertion that the fundamental group of a compact connected aspherical d-manifold has ergodic dimension at most d−1. This assertion is load-bearing: without it, Theorem 1.4 does not follow from Theorem 1.6. Since [CGMT] is not available to the reader, Theorem 1.4 is not currently verifiable. The authors should either supply a proof of the needed ergodic-dimension bound, cite a published version, or explicitly state Theorem 1.4 as conditional on [CGMT].","section":"§7, proof of Theorem 1.4"},{"comment":"The alternative argument intended to avoid [CGMT] states that a non-cocompact lattice Γ'_i in Isom(H^3) has geometric dimension at most 2. This is false: Γ'_i is virtually torsion-free, and a torsion-free finite-index subgroup has cohomological dimension 3 because it is the fundamental group of a finite-volume hyperbolic 3-manifold; hence any proper action of Γ'_i on a contractible complex must have dimension at least 3. If the intended word was 'ergodic dimension' rather than 'geometric dimension', then the needed bound is exactly the content of [CGMT], so the alternative still depends on the unpublished result. Thus the proof of Theorem 1.5 as written is incomplete and does not provide an unconditional proof.","section":"§7, final paragraph of the proof of Theorem 1.5"},{"comment":"The main proof of Theorem 1.5 also relies on [CGMT] for the statement that π1(M) has ergodic dimension at most 2 for every compact 3-manifold M. The introduction's statement of the [CGMT] result is formulated only for compact aspherical manifolds, and the extension to all compact 3-manifold groups is not justified in the manuscript. Since [CGMT] is unpublished, this constitutes another load-bearing gap in the proof of Theorem 1.5.","section":"§7, proof of Theorem 1.5"}],"minor_comments":[{"comment":"The notation F_2^{2n-4} is not explicitly defined; from the construction in §3 it is the direct product of 2n−4 copies of the free group F_2, but this should be stated explicitly to avoid confusion with the free group of rank 2n−4.","section":"§1.3 and §3"},{"comment":"The expression 'F_2^d × Z' in the abstract and §1.3 is ambiguous; if the intended group is the free group F_d, the notation should be corrected to F_d × Z.","section":"Abstract and §1.3"},{"comment":"Question 7.1 reads 'of dimension?' and appears to be missing the dimension variable; presumably it should refer to 'of dimension d'.","section":"§7, Question 7.1"},{"comment":"The construction of the S-exhaustion Θ_i inside the R-exhaustion Ω_i by taking intersections is only sketched; a sentence explaining why the intersection remains a good S-exhaustion would improve clarity.","section":"§5, proof of Theorem 5.1"},{"comment":"Several references are marked 'in preparation' or 'preprint', including [CGMT], [AG20], and [Gab20]; the dependence of the main theorems on unpublished references should be clearly flagged in the text.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"I think the central transfer principle and the automorphism-group applications are correct and publishable, and the paper could be a nice contribution after revision. The main obstacle is the manifold section: Theorem 1.4 is conditional on an unpublished preprint, and the alternative proof of Theorem 1.5 contains a false geometric-dimension claim that must be corrected. The authors should either give a correct unconditional proof or explicitly mark the manifold theorems as conditional; the current wording overstates what has been established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: the transfer trick is the real deal, and the Out(Fn)/Aut(Fn) results are solid. Second thing: the 3-manifold theorems rest on the unpublished [CGMT], and the paper's attempt to bypass it in §7 contains a false premise. So read the paper for the trick and the free-group applications; treat the manifold section as conditional.\n\nWhat's new: Theorems 1.8 and 5.1 give a short, honest argument: if a subgroup (or subrelation) has nonzero top-dimensional ℓ2-Betti number in a d-dimensional proper action (or contractible R-complex), then the ambient group/relation does too. The proof is a clean dimension-count with von Neumann modules; nothing circular. The applications to Out(Fn), Aut(Fn), and the Torelli groups are genuine: the poly-free subgroups F2 ⋉ F2^(2n−4) are explicit, and the cited Lück / Sauer–Thom computations are standard. Those non-vanishing results are new and likely to be useful.\n\nThe soft spot is exactly where the reader flagged: Theorem 1.4 is one line from [CGMT] plus Theorem 1.6, so without [CGMT] it has no support. The authors know this and offer an alternative for Theorem 1.5 in §7, but that passage says a non-cocompact lattice in Isom(H3) has geometric dimension ≤2. That can't be right: such a lattice is virtually torsion-free with cohomological dimension 3 (a finite-index torsion-free subgroup is the fundamental group of a finite-volume hyperbolic 3-manifold). So the alternative fails; if 'geometric dimension' was a slip for 'ergodic dimension', then the needed bound is exactly what [CGMT] claims, so the dependence remains. This isn't a small typo — it's load-bearing for the claim that Theorem 1.5 is independent of [CGMT].\n\nOther minor wrinkles: a few references are in preparation ([AG20], [BG20], [Gab20]), but the core transfer proof doesn't depend on them. The paper is honest about the [CGMT] dependence, which I appreciate.\n\nWho it's for: anyone working on ℓ2-Betti numbers, measured group theory, or automorphism groups of free groups. The transfer lemma is worth having even if the manifold applications need another source for ergodic dimension.\n\nRecommendation: send it to a serious referee. The right referee can verify the transfer argument quickly and judge whether the free-group applications justify publication on their own. The manifold section should not be accepted in its current form; either wait for [CGMT] to become available or rewrite §7 with a correct argument.","headline":"Genuinely new transfer trick with solid applications to Out(Fn)/Aut(Fn); the 3-manifold theorems are conditional on unpublished [CGMT] and the §7 fallback contains a false geometric-dimension claim.","tokens_in":15578,"tokens_out":3623,"would_cite":true,"duration_ms":32683,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37A20","19K56","20F28","20E15","57Mxx"],"pacs":[],"model":"deepseek-v4-flash","headline":"A subgroup's nonzero top Betti number forces the group's too.","keywords":["ℓ2-Betti numbers","top-dimensional Betti numbers","Aut(Fn)","Out(Fn)","Torelli groups","3-manifold groups","ergodic dimension","measured equivalence relations"],"falsifier":"A single counterexample would settle it: find a countable group $\\Gamma$ acting properly on a $d$-dimensional simplicial complex $L$, and a subgroup $\\Lambda$, with $\\beta_d^{(2)}(\\Lambda\\curvearrowright L)\\neq 0$ while $\\beta_d^{(2)}(\\Gamma\\curvearrowright L)=0$. Theorem 1.8 says this cannot happen. A reader could look for the analogous manifold example: a compact aspherical 4-manifold whose fundamental group has vanishing fourth $\\ell^2$-Betti number but contains a subgroup with non-vanishing fourth $\\ell^2$-Betti number.","tokens_in":14559,"feed_emoji":"📈","tokens_out":14384,"duration_ms":126431,"temperature":0.7,"pith_summary":"This note introduces a transfer trick for top-dimensional $\\ell^2$-Betti numbers: if a subgroup of a countable group has non-zero $\\ell^2$-Betti number in the top dimension $d$ of a proper action, then the ambient group has a non-zero $d$-th $\\ell^2$-Betti number for that same action. The contrapositive says that vanishing for the group forces vanishing for every subgroup in that degree. The same principle is proved for probability-measure-preserving equivalence relations, with geometric dimension playing the role of complex dimension. The paper applies it to show non-vanishing of the $\\ell^2$-Betti numbers of $\\mathrm{Out}(F_n)$, $\\mathrm{Aut}(F_n)$, and their Torelli subgroups in their virtual cohomological degrees, and to prove that every subgroup of a compact 3-manifold group has vanishing second and third $\\ell^2$-Betti numbers. It also determines the ergodic dimension of products such as $F_2^d\\times\\mathbb{Z}$ and $\\mathrm{Out}(F_n)\\times\\mathbb{Z}^k$.","feed_headline":"A subgroup's nonzero top Betti number forces the group's too","feed_subtitle":"The transfer trick settles nonvanishing for Aut(Fn), Out(Fn), and vanishing for 3-manifold subgroups.","key_machinery":"The load-bearing object is the top-dimensional reduced $\\ell^2$-homology. In a $d$-dimensional complex the $(d+1)$-chains vanish, so $\\bar H_d^{(2)}$ is exactly the kernel of the boundary map $\\partial_d$, not a quotient. Inclusions of subcomplexes inject these kernels, so if a $\\Lambda$-cocompact exhaustion of $L$ has a non-zero kernel in degree $d$, its $\\Gamma$-saturation has a non-zero kernel too; the faithfulness of the dimension function converts that into non-vanishing of the $\\ell^2$-Betti number. In the measured setting the same inclusion argument runs on direct integrals of $\\ell^2$-chain complexes over the probability space, with the dimension function attached to the ambient operator algebra of the equivalence relation. The paper also uses explicit poly-free subgroups—$F_2\\ltimes F_2^{2n-4}$ inside $\\mathrm{Out}(F_n)$ and its pullback inside $\\mathrm{Aut}(F_n)$—whose $\\ell^2$-Betti numbers were already computable, as the nonzero subgroups that trigger the transfer.","core_discovery":"The central claim is Theorem 1.8: let $\\Gamma$ act properly by simplicial automorphisms on a $d$-dimensional complex $L$, and let $\\Lambda\\leq \\Gamma$ be a subgroup. If $\\beta_d^{(2)}(\\Lambda\\curvearrowright L)\\neq 0$, then $\\beta_d^{(2)}(\\Gamma\\curvearrowright L)\\neq 0$. The measured analogue, Theorem 5.1, says that whenever a probability-measure-preserving equivalence relation $\\mathcal{R}$ has geometric dimension $\\leq d$ and $\\beta_d^{(2)}(\\mathcal{R},\\mu)=0$, every sub-equivalence relation $\\mathcal{S}\\leq\\mathcal{R}$ also has $\\beta_d^{(2)}(\\mathcal{S},\\mu)=0$. These statements are proved by looking at the kernel of the boundary map in top degree, which is the whole reduced $\\ell^2$-homology in that degree. The applications are: $\\beta_{2n-3}^{(2)}(\\mathrm{Out}(F_n))>0$, $\\beta_{2n-2}^{(2)}(\\mathrm{Aut}(F_n))>0$, analogous non-vanishing for Torelli groups, vanishing $\\beta_2^{(2)}=\\beta_3^{(2)}=0$ for all subgroups of compact 3-manifold groups, and ergodic-dimension computations including $F_2^d\\times\\mathbb{Z}$ having ergodic dimension $d+1$.","pith_inferences":["The transfer principle is not tied to free groups or 3-manifolds: it applies to any group with a top-dimensional classifying complex and any subgroup with a computable non-vanishing top $\\ell^2$-Betti number, so other families of automorphism groups are natural testing grounds.","Because the proof identifies the non-vanishing classes as images of classes from explicit poly-free subgroups, it suggests a route to construct explicit top-dimensional cycles in the spine of outer space, rather than only existence statements.","If the cited ergodic-dimension bound for aspherical manifold groups were sharpened toward a middle-dimensional bound, the same theorem would force additional vanishing for all subgroups of such groups; the paper leaves that sharpening as an open question."],"forward_implications":["The groups $\\mathrm{Out}(F_n)$ and $\\mathrm{Aut}(F_n)$ have non-vanishing $\\ell^2$-Betti numbers in degrees $2n-3$ and $2n-2$ respectively, at their virtual cohomological dimensions.","The Torelli subgroups $T_n$ and $K_n$ have non-vanishing $\\ell^2$-Betti numbers in degrees $2n-4$ and $2n-3$ respectively.","Every subgroup of the fundamental group of a compact 3-manifold has vanishing second and third $\\ell^2$-Betti numbers; if the subgroup is infinite, its $\\ell^2$-Euler characteristic lies in $[-\\infty,0]$.","A group with ergodic dimension at most $d$ and vanishing $\\beta_d^{(2)}$ cannot contain a subgroup with non-vanishing $\\beta_d^{(2)}$; consequently $F_2^d\\times\\mathbb{Z}$ has ergodic dimension $d+1$, and multiplying by an infinite amenable group raises the ergodic dimension by one whenever the relevant top Betti number is non-zero.","By the approximation theorem for $\\ell^2$-invariants, the normalized rational homology of $\\mathrm{Out}(F_n)$ and $\\mathrm{Aut}(F_n)$ in their top degrees has positive limit along residual chains of finite-index subgroups."],"supporting_citations":[{"why":"Sets up L2-Betti numbers for pmp equivalence relations and the ergodic dimension, and supplies the framework in which the measured transfer theorem is proved.","marker":"[Gab02]"},{"why":"Provides the definition of $\\ell^2$-Betti numbers for arbitrary groups, the top-dimension kernel identification, and the gluing formulas used in the proofs.","marker":"[CG86]"},{"why":"Supplies the spine of outer space, a contractible $2n-3$-dimensional complex with a proper $\\mathrm{Out}(F_n)$-action, the ambient complex to which the trick is applied.","marker":"[CV86]"},{"why":"Gives the virtual cohomological dimensions of the Torelli groups, fixing the degrees in Theorem 1.3.","marker":"[BBM07]"},{"why":"The unpublished cited result that gives the ergodic-dimension bound $d-1$ for aspherical $d$-manifold fundamental groups; Theorem 1.4 depends on it.","marker":"[CGMT]"},{"why":"Provides the spectral-sequence mechanism used to compute the $\\ell^2$-Betti numbers of the poly-free subgroups.","marker":"[ST10]"},{"why":"Supplies the vanishing of the second $\\ell^2$-Betti number for aspherical 3-manifolds, used in the proof of Theorem 1.5.","marker":"[LL95]"},{"why":"Its geometrisation argument rules out exceptional pieces in the 3-manifold decomposition, part of the support for Theorem 1.5.","marker":"[Per02]"},{"why":"Establishes that infinite amenable groups have ergodic dimension 1, used to compute the ergodic dimension of products with amenable factors.","marker":"[OW80]"}],"fun_headline_variants":["Subgroup nonvanishing forces group nonvanishing in top Betti","Transfer trick settles top ℓ²-Betti for Aut(Fn) and Out(Fn)","Vanishing ℓ²-Betti numbers pass to subgroups in top degree","3-manifold subgroups have zero ℓ²-Betti in degrees 2 and 3","Top-degree ℓ²-Betti: subgroup implies group, three applications"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's manifold applications hinge on an unpublished cited result asserting that a compact aspherical $d$-manifold group has ergodic dimension at most $d-1$; if that result is unavailable, Theorem 1.4 has no proof except in dimension 3.","fun_headline_variants_meta":{"raw":{"variants":["Subgroup nonvanishing forces group nonvanishing in top Betti","Transfer trick settles top ℓ²-Betti for Aut(Fn) and Out(Fn)","Vanishing ℓ²-Betti numbers pass to subgroups in top degree","3-manifold subgroups have zero ℓ²-Betti in degrees 2 and 3","Top-degree ℓ²-Betti: subgroup implies group, three applications"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001174,"raw_usage":{"total_tokens":4866,"prompt_tokens":970,"completion_tokens":3896,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":586,"completion_tokens_details":{"reasoning_tokens":3791}},"tokens_in":586,"tokens_out":3896,"duration_ms":27323,"temperature":1.0,"reasoning_tokens":3791,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:12:53.543053+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A single counterexample would settle it: find a countable group $\\Gamma$ acting properly on a $d$-dimensional simplicial complex $L$, and a subgroup $\\Lambda$, with $\\beta_d^{(2)}(\\Lambda\\curvearrowright L)\\neq 0$ while $\\beta_d^{(2)}(\\Gamma\\curvearrowright L)=0$. Theorem 1.8 says this cannot happen. A reader could look for the analogous manifold example: a compact aspherical 4-manifold whose fundamental group has vanishing fourth $\\ell^2$-Betti number but contains a subgroup with non-vanishing fourth $\\ell^2$-Betti number.","supporting_citations":[],"review_version":1}