{"id":"c8e1711f-86f8-45cf-9283-fe2dbcce940d","arxiv_id":"2501.13234","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Theorems A, B, and C construct new reducibly geometrically finite subgroups of mapping class groups, including right-angled Artin subgroups and free products of reducible subgroups.","lead":"This paper gives conditions under which subgroups of the mapping class group built from powers of mapping classes, or from families of reducible subgroups, are reducibly geometrically finite (RGF), a relative form of geometric finiteness. Its three main theorems supply many new examples of RGF and PGF subgroups, useful for testing the emerging theory of geometric finiteness in mapping class groups.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem A's proof of Claim 7.2 fails when two distinct generators share the same support subsurface, a case the admissibility condition permits.","rationale":"The reader's weakest assumption was the non-nesting condition in Theorem A, correctly identifying Claim 7.2 as load-bearing. My stress-test refines this: the proof's 'equal supports' branch is only valid when equal supports correspond to the same generator. The admissibility condition permits duplicate support surfaces with distinct mapping classes, so Claim 7.2 is false in that case. This is a concrete missing case in the proof of a central theorem, though likely patchable by treating a block of generators on one support as a single loxodromic syllable. The rest of the machinery (Bass–Serre trees, L-displacing families, Theorem B) appears internally consistent, and no independent fatal flaw was found. The verdict should be conditional acceptance pending a fix or a clarification that supports are pairwise distinct.","tokens_in":35822,"tokens_out":55451,"duration_ms":497093,"concrete_test":"Let S be a closed surface, Y a proper subsurface, and let f_1,f_2 be two independent pseudo-Anosov mapping classes fully supported on Y. Let f_3 be fully supported on a subsurface W with d_S(∂Y,∂W)≥3 and with W overlapping Y. Set S_1=S_2=Y and S_3=W; take Γ1={1,2}, Γ2={3}. This list is admissible, Γ=Γ1⊔Γ2, each G_k is reducible, and condition (3) holds. For large p, test whether ⟨f_1^p,f_2^p,f_3^p⟩ ≅ F_2 * Z is RGF relative to ⟨f_1^p,f_2^p⟩ and ⟨f_3^p⟩. In particular, check the lower bound of Lemma 5.1 for the type-1 vertices v(⟨f_1^p,f_2^p⟩) and v((f_1^p f_2^p f_3^p)⟨f_1^p,f_2^p⟩): the tree distance is 6, and the curve-graph distance is d_S(∂Y, f_3^p(∂Y)), which by the BGIT is at least c p - C. If this bound holds for all p≥N, the conclusion is true but the proof of Claims 7.2–7.3 must be revised; if it fails for some p, Theorem A is false.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Claim 7.2, after concluding that S_{ν(j)} and S_{ν(t)} are equal, the proof asserts that g_j and g_t are powers of the same generator f_{ν(j)}=f_{ν(t)}. This is only valid if ν(j)=ν(t). The admissibility condition forbids nesting only when S_i≠S_j, so the list S_1,…,S_n may contain duplicates: two distinct vertices i,t with S_i=S_t=Y. If f_i and f_t are independent pseudo-Anosovs on Y, then g_j and g_t do not commute, and β_t does not fix Y_j or preserve projections to Y_j. Consequently, the adjacent claim that every β_t with ι(j)<t<τ(j) preserves Y_j is false, and the computation d_{Y_j}(Y_{ι(j)},Y_{τ(j)}) = d_{S_{ν(j)}}(S_{ν(ι(j))}, g_j S_{ν(τ(j))}) in Claim 7.3 breaks. The Behrstock hypothesis for Corollary 2.8 is therefore not established in the presence of duplicate supports. Since Theorem A's statement does not rule out S_i=S_j, the proof as written has a genuine missing case.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a theory of reducibly geometrically finite (RGF) subgroups of mapping class groups and contributes several construction theorems. Theorem A asserts that, under an admissibility condition on a family of fully supported mapping classes whose realization graph splits into at least two subgraphs, with each subgraph generating a reducible subgroup and with 3-separation between supports in distinct subgraphs, sufficiently high powers generate a free product of right-angled Artin subgroups and are RGF relative to those factors. Theorem C states that pairwise 5-separated infinite reducible subgroups admit finite-index subgroups with the property that every further infinite subgroup generates an RGF free product. Theorem B gives a combination theorem for D-separated and A-misaligned torsion-free reducible subgroups. The paper also contains examples showing that several hypotheses in the theorems are necessary.","tokens_in":36065,"tokens_out":15905,"duration_ms":166419,"significance":"If the main construction is sound, the paper substantially enlarges the supply of RGF subgroups of mapping class groups and ties together RAAG subgroups, combination theorems, and the newer geometric finiteness notions in a useful way. The proofs are detailed and largely use established machinery as black boxes, including curve-graph hyperbolicity, the Bounded Geodesic Image Theorem, the Behrstock inequality, the distance formula, and Bass-Serre trees; the paper does not circularly rely on its own conclusions. The examples in Section 9 that probe necessity of the hypotheses are a valuable feature. The main caveat is a missing case in the proof of Theorem A, discussed below, which requires either a strengthened hypothesis or an additional argument.","major_comments":[{"comment":"The proof of Theorem A does not cover the case of duplicate support surfaces, a case that the admissibility hypothesis explicitly permits. The admissibility condition only restricts nesting when S_i ≠ S_j, so two distinct indices i and t may have S_i = S_t = Y. If f_i and f_t are independent pseudo-Anosovs on the same Y, the inference in Claim 7.2 that equality of S_{\\nu(j)} and S_{\\nu(t)} forces g_j and g_t to be powers of the same generator f_{\\nu(j)} = f_{\\nu(t)} is false unless \\nu(j) = \\nu(t). In the duplicate case, g_j and g_t need not commute, \\beta_t need not preserve Y_j, and the equality d_{Y_j}(Y_{\\iota(j)},Y_{\\tau(j)}) = d_{S_{\\nu(j)}}(S_{\\nu(\\iota(j))}, g_j S_{\\nu(\\tau(j))}) in Claim 7.3 is not justified. Since Theorem A's statement does not rule out S_i = S_j for distinct indices, the proof as written has a genuine missing case. Please either add the hypothesis that the supports S_i are pairwise distinct, or supply an argument that handles equal supports, for instance by treating \\Psi(A(\\Gamma_k)) as a reducible peripheral subgroup rather than spelling its internal RAAG structure in the curve-graph estimate.","section":"Section 7, Claim 7.2 and Claim 7.3"}],"minor_comments":[{"comment":"The sentence 'In answering Proposition 1.3 we will formulate our conditions...' should refer to Question 1.3, not Proposition 1.3.","section":"Section 1, after Question 1.3"},{"comment":"The text says 'In Proposition 2.2 it is not hard to see...' but the relevant item is Definition 2.2, the Bowditch definition of relative hyperbolicity.","section":"Remark 2.3"},{"comment":"The assertion that G satisfies the bounded coset penetration property is stated without proof or reference; since this example is used to justify the torsion-free assumption in Theorem B, please provide an argument or a precise citation.","section":"Example 9.4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is well within the scope of the journal and the overall project is valuable. The gap in Theorem A is localized and appears repairable, either by restricting the statement or by a more careful treatment of equal supports, so I would not recommend rejection. The positive results of Theorems B and C appear to be unaffected by this issue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this paper is worth a serious referee. It delivers new RGF examples via three combination theorems that genuinely extend the literature. Theorem B generalizes Loa's free-product result from two multitwist groups to arbitrary torsion-free reducible subgroups and to any finite number of factors; Theorem C shows that 5-separation is enough after passing to finite-index subgroups; Theorem A gives the first RGF analysis of the Koberda/CLM RAAG construction. The canonical reducing system for subgroups (Section 3) is a useful tool in its own right.\n\nThe proofs are careful and use established machinery (curve graph hyperbolicity, Behrstock inequality, BGIT, distance formula) as black boxes. The examples in Section 9 are not filler: they show that the separating, misaligned, and torsion-free hypotheses in Theorem B are necessary. The paper is also honest about its limitations: boundary surfaces are excluded because of the center, and the constants D and A in Theorem B are ineffective.\n\nI checked the stress-test worry about Claim 7.2 — distinct generators sharing the same support subsurface. It does not hold up. The theorem's map Ψ sends each vertex x_i of Γ to f_i^{p_i}; if two different indices had the same support, they would be the same vertex, and the map would be ill-defined unless the f_i were equal. So the intended assumption is that the S_i are pairwise distinct, and under that reading the proof's inference 'if S_{ν(j)} = S_{ν(t)} then they are the same generator' is valid. The admissibility wording could be clearer, but the gap is not real.\n\nMinor quibbles: the paper uses 'irredundant' without defining it (relying on Koberda), and the proof of Lemma 3.6 could be more explicit, but these are not problems.\n\nWho this is for: geometric group theorists working on mapping class groups, relative hyperbolicity, and RAAG subgroups. It would be a good reading group paper. I'd cite it in my own work. Send it to a serious referee.","headline":"Solid, useful paper with real generalizations; the flagged duplicate-support gap in Theorem A dissolves when you read the statement as requiring distinct supports.","tokens_in":36598,"tokens_out":8637,"would_cite":true,"duration_ms":78792,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20F65","57K20","20F67"],"pacs":[],"model":"deepseek-v4-flash","headline":"High powers of mapping classes generate reducibly geometrically finite subgroups when their supports are separated in the curve graph.","keywords":["mapping class group","reducibly geometrically finite","right-angled Artin subgroup","curve graph","free product","relative hyperbolicity","canonical reducing system","subsurface projection"],"falsifier":"To test the sharpness of Theorem A, build an admissible family with two supports lying in different subgraphs but with curve-graph distance exactly 2, and choose fully supported partial pseudo-Anosovs f_1, f_2. If for all large p, q the product f_1^p f_2^q is reducible, with bounded orbit in the curve graph, then the conclusion fails, since RGF requires every element outside the peripheral factors to be pseudo-Anosov. The paper's Section 9 examples exhibit exactly such behavior when separation drops to 2.","tokens_in":35654,"feed_emoji":"🧩","tokens_out":5560,"duration_ms":53716,"temperature":0.7,"pith_summary":"The paper establishes combination theorems that produce reducibly geometrically finite (RGF) subgroups of the mapping class group: subgroups that are hyperbolic relative to a family of reducible subgroups while their coned-off Cayley graph quasi-isometrically embeds into the curve graph. In one theorem, mapping classes fully supported on an admissible family of subsurfaces, with supports split into separated subgraphs of the realization graph, generate, after passing to sufficiently high powers, a subgroup isomorphic to a free product of right-angled Artin subgroups that is RGF relative to those factors. In another, any sufficiently separated and misaligned family of torsion-free reducible subgroups generates their free product and is RGF with no powering. A third theorem shows that merely 5-separated reducible subgroups can be made RGF by passing to finite-index subgroups. Together these supply many new examples and clarify exactly which separation hypotheses are needed.","feed_headline":"High powers of mapping classes make RGF subgroups","feed_subtitle":"New combination theorems: separation in the curve graph decides when reducible subgroups generate reducibly geometrically finite free…","key_machinery":"The carrying object is the Bass–Serre tree T of the abstract free product H_1 ∗ ... ∗ H_m, together with the equivariant coarse map ϕ: T → C(S) that sends a coset vertex gH_i to the translated canonical reducing system g·∂H_i. The map is a quasi-isometric embedding exactly when the factors are sufficiently separated in the curve graph; the proofs verify this by applying the Behrstock inequality and the bounded geodesic image theorem to sequences of translated support subsurfaces, and use the local-to-global principle for hyperbolic spaces to convert pointwise Gromov-product bounds into linear lower bounds on distance. For Theorem A, the normal form of a word in the right-angled Artin group is matched with a sequence of translated supports whose consecutive distances are at least 3, forcing the total distance in the curve graph to grow linearly.","core_discovery":"The paper's central discovery is that relative geometric finiteness in the mapping class group Mod(S) is governed by how far apart supports, or canonical reducing systems, are in the curve graph. Theorem A states that if mapping classes f_i are fully supported on an admissible family of subsurfaces whose realization graph Γ splits as a disjoint union Γ_1 ⊔ ... ⊔ Γ_m with m ≥ 2, each cluster generating a reducible subgroup, and supports in distinct subgraphs are at least 3-separated in the curve graph, then for all sufficiently large exponents p_i the subgroup generated by $f_i^{{p_i}}$ is isomorphic to the free product Ψ(A(Γ_1)) ∗ ... ∗ Ψ(A(Γ_m)) and is RGF relative to these factors. Theorem B gives constants D, A such that any D-separated and A-misaligned family of torsion-free reducible subgroups generates their free product and is RGF relative to the family. Theorem C shows that any 5-separated family of infinite reducible subgroups admits finite-index subgroups whose further infinite subgroups generate an RGF free product. The examples in the final section confirm that the separation, misalignment, and torsion-free hypotheses are necessary.","pith_inferences":["Theorem A's combination of RGF, freeness, and undistortion suggests these right-angled Artin subgroups could serve as building blocks for further combination theorems, for instance in constructing relatively hyperbolic surface group extensions of Mod(S).","Theorem C indicates that the only real obstruction to relative hyperbolicity in a finite family of reducible subgroups is local proximity of their reducing systems; once separated by distance 5, finite-index passage removes all other obstructions, so a similar phenomenon may hold for other hierarchically hyperbolic groups.","A testable extension is to relax the pairwise non-nested admissibility condition to allow nesting with a uniform depth bound; the normal-form and separation argument might survive if nested supports are controlled by an explicit constant.","The 3-separation threshold in Theorem A is exactly what forces products of generators from different clusters to be pseudo-Anosov, so analogous thresholds are likely to appear in other relative hyperbolicity results for mapping class groups."],"forward_implications":["The right-angled Artin subgroups constructed by Koberda and by Clay–Leininger–Mangahas are RGF whenever the defining graph splits into at least two components whose supports are 3-separated, and the embedding can be made injective and undistorted (Corollary 1.1).","Free products of arbitrary torsion-free reducible subgroups are RGF once the family is sufficiently far apart and misaligned, generalizing Loa's theorem for multitwist groups to families of any size.","Even without large separation, 5-separated reducible subgroups can be made RGF after passing to finite-index subgroups, so raising elements to high powers is closely analogous to passing to finite-index subgroups.","Every element of the constructed subgroup that is not conjugate into a peripheral factor is pseudo-Anosov.","The separation constants are sharp: the distance-3 condition in Theorem A cannot be weakened to distance 2, and the misalignment and torsion-free assumptions in Theorem B are necessary."],"supporting_citations":[{"why":"Supplies the isomorphism between the right-angled Artin group A(Γ) and the subgroup generated by high powers f_i^r for irredundant families.","marker":"[Kob12]"},{"why":"Provides an independent proof that the RAAG subgroup is isomorphic to A(Γ) and equivariantly quasi-isometrically embeds into Mod(S).","marker":"[CLM12]"},{"why":"Gives effective exponent bounds and extends undistortion to allow Dehn twist powers, confirming the setup used in Corollary 1.1.","marker":"[Run21]"},{"why":"Proves free products of two multitwist subgroups are PGF when their reducing systems are far apart; Theorem B generalizes this to arbitrary torsion-free reducible subgroups and larger families.","marker":"[Loa21]"},{"why":"Establishes hyperbolicity of the curve graph, the ambient space into which the coned-off Cayley graph must quasi-isometrically embed.","marker":"[MM99]"},{"why":"Provides the distance formula and the bounded geodesic image theorem, both used to promote subsurface projection bounds to curve-graph distance bounds.","marker":"[MM00]"},{"why":"Gives the Behrstock inequality that controls how projections to overlapping subsurfaces can be simultaneously large, used in Corollary 2.7 and the proof of Theorem A.","marker":"[Beh06]"},{"why":"Supplies the canonical reducing system construction that the paper extends from elements to reducible subgroups.","marker":"[HT85]"},{"why":"Provides the Nielsen–Thurston classification for subgroups and the uniform power bound that puts elements into normal form.","marker":"[Iva92]"},{"why":"Gives residual finiteness of Mod(S), which is used in Lemma 6.5 to pass to finite-index subgroups with large displacement.","marker":"[Gro75]"}],"fun_headline_variants":["Curve graph separation builds RGF subgroups","High powers plus separation yield RGF free products","When supports separate, mapping classes make RGF","RGF subgroups from separated reducible families","Finiteness in mapping class groups via curve graph"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the admissibility condition that distinct supporting subsurfaces are pairwise non-nested; without it, the proof that non-overlapping supports are either disjoint or equal (Claim 7.2) fails, and with it the normal-form spelling and the quasi-isometric embedding bound collapse.","fun_headline_variants_meta":{"raw":{"variants":["Curve graph separation builds RGF subgroups","High powers plus separation yield RGF free products","When supports separate, mapping classes make RGF","RGF subgroups from separated reducible families","Finiteness in mapping class groups via curve graph"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0002,"raw_usage":{"total_tokens":1371,"prompt_tokens":936,"completion_tokens":435,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":378}},"tokens_in":552,"tokens_out":435,"duration_ms":5087,"temperature":1.0,"reasoning_tokens":378,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T16:20:05.443787+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"To test the sharpness of Theorem A, build an admissible family with two supports lying in different subgraphs but with curve-graph distance exactly 2, and choose fully supported partial pseudo-Anosovs f_1, f_2. If for all large p, q the product f_1^p f_2^q is reducible, with bounded orbit in the curve graph, then the conclusion fails, since RGF requires every element outside the peripheral factors to be pseudo-Anosov. The paper's Section 9 examples exhibit exactly such behavior when separation drops to 2.","supporting_citations":[],"review_version":1}