{"id":"9d43776c-c675-42e1-8c02-dd66a6b01e7b","arxiv_id":"2507.07076","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The Cannon-Thurston map is surjective for metric graph bundles with hyperbolic fibers that are either one-ended or of bounded valence with coarsely surjective barycenter maps.","lead":"This paper proves that, in certain metric graph bundles over a ray with hyperbolic fibers and hyperbolic total space, the Cannon-Thurston map from a fiber's boundary to the total space's boundary is surjective. It generalizes a theorem of Bowditch and answers a question of Lazarovich, Margolis, and Mj.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.6(B) Claim 2 invokes Lemma 2.6 without proving that the flowed paths α_{n,i} have endpoints on opposite sides of the point x_i ∈ β_i; as written the doubly exponential length lower bound is unjustified.","rationale":"The reader's verdict is CONDITIONAL and identifies L-coarse surjectivity of barycenter maps as the weakest assumption. I agree that assumption is load-bearing, since it is used essentially in Corollary 4.13, Theorem 3.2, and Proposition 4.21, and the paper itself cites examples where its failure destroys surjectivity. However, my stress-test found a more specific internal gap in the proof of Theorem 1.6(B), which is the one-ended-fiber case. The gap is that Lemma 2.6 is applied to paths α_{n,i} without proving that the point x_i on β_i lies on the geodesic segment between the endpoints of α_{n,i}. The written construction of α_{n,0} does not force a path that straddles the ball around Σ(0), and the transfer step does not record any straddling data. Without this, the lower bound length(α_{n,i}) ≥ b^{a b^i − K − 1} is unjustified, and the contradiction with the upper bound in Note 1 collapses. This is not necessarily fatal: the gap appears repairable by choosing α_{n,0} to join the two outer components of β0 and by adding an induction lemma showing coarse preservation of straddling under the quasiisometries ϕ_i. For that reason I do not recommend changing the reader's CONDITIONAL verdict, but the author should be asked to supply this missing argument before publication. The reader's rationale already mentions 'the implicit use of a separating point in the one-ended path construction', which is closely related to my concern, but the reader does not elevate it to a load-bearing issue, instead selecting the barycenter assumption. Hence my agreement is partial.","tokens_in":43050,"tokens_out":28816,"duration_ms":281877,"concrete_test":"Re-prove Claim 2 in §5.1 with an explicit induction hypothesis: for each i, the two endpoints of α_{n,i} lie on β_i on opposite sides of the point x_i closest to Σ(i), and d_i(x_i, α_{n,i}) ≥ a b^i − K. Verify that Lemma 2.12 transfers this 'straddling' property from fiber i to fiber i+1 with a uniform additive error. If the property holds, Lemma 2.6 applies and the contradiction goes through; if it cannot be established, the lower bound in (∗∗) is unsupported. A concrete check would be to run this induction explicitly on Bowditch's H^2 stack example, computing the flowed endpoints and confirming that they remain on opposite sides of the flowed barycenter point.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In §5.1, Claim 2, after flaring gives d_i(Σ(i), α_{n,i}) ≥ a b^i, the proof chooses x_i ∈ β_i and applies Lemma 2.6 to conclude length(α_{n,i}) ≥ b^{a b^i − K − 1}. Lemma 2.6 only bounds the length of a path that stays outside a ball centered at a point lying on a geodesic joining the path's endpoints. The proof never establishes that x_i lies between the two endpoints of α_{n,i} on β_i. The initial path α_{n,0} is specified merely as 'a path joining two points belonging to β0 and lying outside the n-radius ball centered at Σ(0)'; such a path could be a short geodesic segment lying entirely on one side of Σ(0) — for instance, in H^2, take endpoints on β0 at distances n+1 and n+2 from the center, joined by the geodesic segment, which stays outside the n-ball and has length 1. For such a choice the claimed exponential lower bound is plainly false. One-endedness does guarantee a path connecting the two outer components of β0, but that choice is not made, and the inductive transfer via Lemma 2.12 does not track whether the two endpoints remain on opposite sides of the corresponding point near Σ(i). Thus the doubly exponential growth of length(α_{n,i}), which is the contradiction with Note 1, does not follow from the stated hypotheses. This gap occurs in the essential Claim 2, so the proof of Theorem 1.6(B) as written is incomplete.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves surjectivity of the Cannon–Thurston map for the inclusion of a fiber into a hyperbolic metric (graph) bundle, in two settings: (A) fibers have uniformly bounded valence and controlled hyperbolic fibers, and (B) fibers are one-ended proper hyperbolic spaces. The main results are Theorem 1.6, Corollary 1.2 (fibers uniformly quasiisometric to a fixed nonelementary hyperbolic group), and Theorem 6.1 for arbitrary hyperbolic base, with applications to metric bundles in the appendix. The paper also proves a standalone growth result (Theorem 3.10): a hyperbolic metric graph with bounded valence and coarsely surjective barycenter map has exponential growth, via a quasiisometric embedding of the trivalent tree (Theorem 3.2).","tokens_in":43331,"tokens_out":13163,"duration_ms":150496,"significance":"If the proofs are completed, the paper would answer Question 3.15 of Lazarovich–Margolis–Mj and generalize Bowditch's theorem for hyperbolic-plane fibers to one-ended hyperbolic fibers and to bounded-valence controlled hyperbolic fibers. The growth theorem for metric graphs is of independent interest and the constants in the arguments are explicit. The paper is not circular: the surjectivity statement is derived from hyperbolicity, barycenter coarse surjectivity, bounded valence, or one-endedness, and the acknowledgements and references are appropriate. However, the proof of the one-ended case, which is one of the two main theorems, contains a gap in the key length-growth contradiction; the claim is likely repairable, but the manuscript as written does not establish it.","major_comments":[{"comment":"The lower bound (∗∗), length(α_{n,i}) ≥ b^{ab^i−K−1}, does not follow from Lemma 2.6 as stated. Lemma 2.6 applies to a path γ joining two points p,q when the chosen point x lies on a geodesic segment [p,q] and γ lies outside the n-radius ball centered at x. In the proof, α_{n,i} is only known to join two points on the geodesic line β_i, and the point x_i ∈ β_i satisfying d_i(x_i, α_{n,i}) ≥ ab^i − K is not shown to lie between the endpoints of α_{n,i} on β_i. If the endpoints lie on the same side of x_i, the geodesic segment between them can lie outside the ball and have length O(1), so no doubly exponential lower bound holds for that path. The one-endedness assumption does provide paths connecting points on opposite sides of the ball around Σ(i), but such a choice is not made for α_{n,0} and the inductive transfer via Lemma 2.12 does not record or propagate an 'opposite sides' condition. Consequently the contradiction with Note 1 is not established, and the proof of Theorem 1.6(B) is incomplete as written; the same gap affects Theorems 6.1(B) and 7.8(B), which rely on Theorem 1.6(B). This is a local but load-bearing gap, and the argument appears repairable by explicitly choosing endpoints on opposite sides and proving this property is preserved by the flow.","section":"5.1, Claim 2 (proof of Theorem 1.6(B))"}],"minor_comments":[{"comment":"In the display after the inequality ab^n ≤ ||A||D^{d+1}, the expression (aD^{-([k^2]+2)}b^{-3k})(b^{1/k})^n does not follow from the preceding line; the exponent calculation yields b^{m/k−3} = b^{-3}(b^{1/k})^m, so the constant b^{-3k} appears to be a typo for b^{-3} (with the small-t case handled separately by b^{-3k}). The lemma is still true with corrected constants, but the displayed arithmetic should be fixed.","section":"3.2, Lemma 3.7 proof"},{"comment":"The quantifiers in Claim 2 are imprecise: 'There is a subsequence {n_k} ⊆ N such that d_l(x_l, Σ(l)) ≤ M for some l ∈ N and x_l ∈ α_{n_k,l} where M ≥ 0' does not specify whether M is chosen first or may depend on the subsequence, and it does not state that l may depend on k. The subsequent application to Lemma 4.22 needs: there exists M ≥ 0 and infinitely many n such that for some l and some x_l ∈ α_{n,l}, d_l(x_l, Σ(l)) ≤ M. The current wording should be corrected.","section":"5.1, Claim 2 statement"},{"comment":"The inductive construction of Σ_1 in Claim 1 conflates Σ and Σ_1 in several places (e.g., 'Suppose Σ(j) ∈ α_{n,j}' should read 'Suppose Σ_1(j) ∈ α_{n,j}'), and the indices in the backward/forward extension step, involving ϕ_j(α_{n,j−1}) versus ϕ_{j+1}(α_{n,j+1}), are written in a confusing way. The intended construction is clear but should be rewritten carefully.","section":"5.1, Claim 1 proof"},{"comment":"The constant K is used both for the good-section constant from Corollary 4.13 and for the qi-section constant K' in the proof of Theorem 1.6(B); this overloads notation and makes the line 'K'' = max{K', K}' hard to parse. Renaming one of the constants would improve readability.","section":"5.1, text after Claim 1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is careful and the main results are plausible, but the proof of Theorem 1.6(B) has a genuine gap in Claim 2: the application of the exponential divergence lemma is not justified without an 'opposite sides' condition that is neither chosen nor propagated. This gap is local and likely fixable, so I recommend major revision rather than rejection. Please ask the author to repair this point, to clarify the quantifiers in Claim 2, and to correct the small arithmetic typo in Lemma 3.7. I do not see grounds for concern about circularity or credit."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Main point: this paper gives a real positive answer to LMM's Question 1.5, and the standalone tools are worth having. The embedding of the trivalent tree under coarse barycenter surjectivity (Theorem 3.2) and the exponential growth criterion (Theorem 3.10) are new and clean. The proof of Theorem 1.6(A) and Corollary 1.2 is solid: reduce to uniform exponential growth of the fibers, then use the flow argument. The boundary description via good qi-sections (Proposition 4.21) is carefully laid out, and I see no circularity or fitted parameters. The soft spot is in the proof of Theorem 1.6(B), specifically Claim 2 in Section 5.1. The contradiction aims for doubly exponential growth of the flowed paths alpha_{n,i}. It applies Lemma 2.6 to the point x_i in beta_i, using d_i(x_i, alpha_{n,i}) >= a b^i - K. But Lemma 2.6 only gives a length lower bound for paths that stay outside a ball centered at a point lying on a geodesic joining the path's two endpoints. The proof never establishes that x_i lies between the endpoints of alpha_{n,i} on beta_i. The initial path alpha_{n,0} is chosen as any path outside the n-ball around Sigma(0); such a path can have both endpoints on the same side of Sigma(0). In H^2, take endpoints at distances n+1 and n+2 from the center on the same ray, joined by the geodesic segment; that path stays outside the n-ball and has length 1. One-endedness does allow choosing the initial endpoints on opposite sides, but the paper doesn't make that choice, and the inductive transfer via Lemma 2.12 doesn't track the side information. So as written, the doubly exponential lower bound does not follow. This is an essential gap: without it, Claim 2 fails and the contradiction with Note 1 collapses. Minor issues worth mentioning: Lemma 3.7 has a typo in the exponent (b^{-3} should replace b^{-3k} in the definition of a_{3.7}); the quantifiers in Claim 2 are loose; and the one-ended path construction would be easier to follow if the projection of Sigma(i) to beta_i were introduced explicitly. Overall, the main theorems are likely true, and the new tools justify a serious look. I would send this to a careful referee. The author should be asked to fix the gap in Claim 2, choosing alpha_{n,0} with endpoints on opposite sides of the projected center and tracking the sides through the induction. Once that is done, the paper deserves publication. This is for people working on Cannon-Thurston maps and metric bundles; Theorems 3.2 and 3.10 are worth citing independently.","headline":"Genuinely useful results and a mostly clean proof of Theorem 1.6(A), but Theorem 1.6(B) has a real gap in Claim 2 that needs fixing before publication.","tokens_in":789,"tokens_out":958,"would_cite":true,"duration_ms":84889,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20F65","20F67"],"pacs":[],"model":"deepseek-v4-flash","headline":"For hyperbolic metric bundles, fiber inclusions extend continuously to boundary maps, and this paper proves those boundary maps are surjective whenever the fibers are one-ended or have uniformly bounded valence.","keywords":["Cannon–Thurston map","metric graph bundles","metric bundles","Gromov boundary","barycenter map","exponential growth","one-ended hyperbolic spaces","controlled hyperbolic fibers"],"falsifier":"A concrete falsifier would be an f-metric graph bundle π:X→[0,∞) satisfying all hypotheses of Theorem 1.6(A) or (B) whose good qi section endpoints are not all limits of F_0-sequences: by Lemma 2.15 this is exactly a point of ∂good X outside Λ_X(F_0), so the Cannon–Thurston map misses it and surjectivity fails. The flow estimates in Theorem 5.1 make this checkable numerically: in a counterexample, the number of vertices in Fl_1(A)∩F_i would have to violate the upper bound ||A||c^i while the flaring lower bound grows doubly exponentially.","tokens_in":42790,"feed_emoji":"🌐","tokens_out":7074,"duration_ms":73729,"temperature":0.7,"pith_summary":"The paper proves that, in a metric graph bundle with hyperbolic fibers over [0,∞), when the total space is hyperbolic, the Cannon–Thurston map from a fiber's Gromov boundary to the total space's boundary hits every boundary point, provided either the fibers have uniformly bounded valence or the fibers are one-ended proper metric spaces. This answers a question posed in [LMM24] and generalizes the fiber-hyperbolic-plane result of [Bow13]. The result matters because surjectivity of the Cannon–Thurston map says the whole boundary of the bundle is visible from a single fiber, a strong constraint on the geometry of hyperbolic group extensions and their boundaries.","feed_headline":"Cannon–Thurston maps go onto the whole boundary in two new settings","feed_subtitle":"Hyperbolic fibers that are one-ended or of bounded valence make every boundary point of the total space a limit point of the fiber.","key_machinery":"The central object is a metric (graph) bundle with controlled hyperbolic fibers: each fiber is δ-hyperbolic and its barycenter map ∂³F_i→F_i, sending an ideal triangle in the boundary to a point near all three sides, is L-coarsely surjective. Through every point of the bundle there is a good qi section, obtained by fixing an ideal triangle in one fiber and taking barycenters of its images in every other fiber; the boundary of the total space splits as the fiber limit set plus the endpoints of such good sections. In case (A), a key theorem (Theorem 3.10) shows that a hyperbolic metric graph with bounded valence and coarsely surjective barycenter map has exponential growth, and the resulting uniform exponential growth is combined with a flow-counting estimate to force every good section endpoint to be a limit of the initial fiber. In case (B), one-endedness gives paths in the initial fiber that stay outside arbitrarily large balls, and flowing these paths while controlling their length against exponential divergence yields the same conclusion.","core_discovery":"For an f-metric graph bundle π:X→[0,∞) with δ-hyperbolic fibers, L-coarsely surjective barycenter maps ∂³F_i→F_i, and δ-hperbolic total space, the Cannon–Thurston map ∂$π^{{-1}}$(0)→∂X is surjective under either (A) uniformly bounded vertex valence in every fiber or (B) one-ended proper fibers. In particular, if all fibers are uniformly quasiisometric to a fixed nonelementary hyperbolic group and have bounded valence, the Cannon–Thurston map is surjective (Corollary 1.2). The proof describes ∂X as Λ_X(F_0) ∪ ∂good X and then shows each endpoint of a good qi section over [0,∞) is a limit point of F_0—in case (B) by flowing curves that avoid large balls in one-ended fibers, in case (A) by using uniform exponential growth of fibers and a flow-counting estimate. This answers Question 1.5 of [LMM24].","pith_inferences":["Beyond the paper, the flow-counting mechanism suggests that the bounded-valence hypothesis in case (A) could be weakened to uniform exponential growth of the fibers; Theorem 5.1 already isolates that condition, and the paper leaves open whether it is also necessary.","Beyond the paper, the boundary description ∂X = Λ_X(F_0) ∪ ∂good X may give a direct way to construct Cannon–Thurston laminations from pairs of good sections, so the non-injectivity statements in Section 6.3 could be proved without first passing through fiber boundary pairs.","Beyond the paper, the same coarse-geometric strategy may transfer to other families of hyperbolic spaces with uniformly coarsely surjective barycenter maps, such as strongly proper actions, yielding surjectivity of Cannon–Thurston maps for commensurated subgroups beyond the examples listed in Section 6.2."],"forward_implications":["For any metric graph bundle over a hyperbolic base with controlled hyperbolic fibers satisfying either of the two assumptions, the Cannon–Thurston map from a fiber into any qi-embedded sub-bundle is surjective (Theorem 6.1).","When the fibers are uniformly quasiisometric to a fixed nonelementary hyperbolic group and have bounded valence, the surjective Cannon–Thurston map is not injective, so the Cannon–Thurston lamination is nonempty (Theorem 6.11).","The same surjectivity holds for metric bundles, not just metric graph bundles, when the fibers are uniformly strongly proper or one-ended proper (Theorem 7.8).","The combinatorial horoball example shows that surjectivity can hold even when the fibers are not uniformly quasiisometric to a fixed hyperbolic space and the barycenter maps are not uniformly coarsely surjective, so the paper's hypotheses are sufficient but not necessary."],"supporting_citations":[{"why":"Introduces metric (graph) bundles, proves the existence of the Cannon–Thurston map for fiber inclusions, and supplies the flaring conditions and qi-section barycenter machinery the paper builds on.","marker":"[MS12]"},{"why":"Proves surjectivity of the Cannon–Thurston map when the fibers are hyperbolic planes and supplies the one-endedness strategy and boundary description adapted here.","marker":"[Bow13]"},{"why":"Defines controlled hyperbolic fibers, poses the question answered by Theorem 1.6, and gives an example where the Cannon–Thurston map is not surjective without coarse surjectivity of the barycenter maps.","marker":"[LMM24]"},{"why":"Provides the detailed boundary description and the reduction of an arbitrary hyperbolic base to a geodesic ray, along with the Cannon–Thurston lamination criterion used in Section 6.","marker":"[KS20]"},{"why":"Supplies the result that nonelementary hyperbolic groups have exponential growth, which drives the growth argument in Corollary 1.2.","marker":"[Kou98]"},{"why":"Establishes the existence of Cannon–Thurston maps for trees of hyperbolic metric spaces and for group extensions, giving the general existence result the paper relies on throughout.","marker":"[Mit98b]"}],"fun_headline_variants":["Cannon–Thurston map surjective for one-ended or bounded-valence fibers","Cannon–Thurston surjectivity in two new settings for hyperbolic bundles","Surjective Cannon–Thurston maps when fibers are one-ended or have bounded valence","Generalizing Bowditch: Cannon–Thurston surjectivity for two fiber classes","New result: Cannon–Thurston maps onto boundary in two settings"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the fibers' barycenter maps are uniformly coarsely surjective—every point of every fiber lies within a fixed distance of the barycenter of some ideal triangle—because without it the Cannon–Thurston map can fail to be surjective, as an example cited in the paper shows.","fun_headline_variants_meta":{"raw":{"variants":["Cannon–Thurston map surjective for one-ended or bounded-valence fibers","Cannon–Thurston surjectivity in two new settings for hyperbolic bundles","Surjective Cannon–Thurston maps when fibers are one-ended or have bounded valence","Generalizing Bowditch: Cannon–Thurston surjectivity for two fiber classes","New result: Cannon–Thurston maps onto boundary in two settings"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000742,"raw_usage":{"total_tokens":3304,"prompt_tokens":934,"completion_tokens":2370,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":550,"completion_tokens_details":{"reasoning_tokens":2262}},"tokens_in":550,"tokens_out":2370,"duration_ms":67394,"temperature":1.0,"reasoning_tokens":2262,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:50:00.318116+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete falsifier would be an f-metric graph bundle π:X→[0,∞) satisfying all hypotheses of Theorem 1.6(A) or (B) whose good qi section endpoints are not all limits of F_0-sequences: by Lemma 2.15 this is exactly a point of ∂good X outside Λ_X(F_0), so the Cannon–Thurston map misses it and surjectivity fails. The flow estimates in Theorem 5.1 make this checkable numerically: in a counterexample, the number of vertices in Fl_1(A)∩F_i would have to violate the upper bound ||A||c^i while the flaring lower bound grows doubly exponentially.","supporting_citations":[],"review_version":1}