{"id":"e9423860-b5eb-413a-895d-2d1efa619258","arxiv_id":"2505.16450","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In diagonal Heintze groups with non-scalar A, horospheres split into exactly two isometry and quasi-isometry classes, with the non-Euclidean class having volume growth of order r^k, k=(λ1+...+λd)/λ1.","lead":"A diagonal Heintze group is a curved space built from a diagonal matrix A; this paper shows its horospheres come in exactly two geometric types, one flat and one with volume growth r^k where k is the sum of the eigenvalues divided by the smallest one. The result gives a concrete, computable family where the isometry and quasi-isometry classes of horospheres coincide.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 13's angle estimate (24) is load-bearing and only sketched; an explicit Busemann computation or model-space angle bound should settle whether the controlled-volume transfer is valid.","rationale":"After reviewing the construction, the main components are independently supported: Proposition 3 gives an explicit volume computation for the approximate horosphere, Proposition 5 uses standard projection arguments for the quasi-isometry, and Corollary 4 is a direct consequence of Coulhon-Saloff-Coste. The weakest point is indeed Lemma 13's inequality (24), which is the only substantial geometric estimate asserted rather than derived. I believe the estimate is true: in the constant-curvature model a geodesic at distance at least D from a point subtends visual angle at most 4 arctan(e^{-lambda_1 D}), uniformly in the segment length, and the CAT(-lambda_1^2) inequality transfers the needed distance lower bound from the original triangle to the comparison triangle. But the paper does not show this computation, and the transfer is only implicit. This does not undermine the overall strategy or novelty, and it matches the reader's assessment. The reader's CONDITIONAL verdict is appropriate, so no change is needed.","tokens_in":17970,"tokens_out":41937,"duration_ms":352183,"concrete_test":"Compute the Busemann function for xi- = 0 in G_A: b_0(y,x) = lim_{s -> infinity} dist((y,x),(0,-s)) - s, and set beta^-_p = grad b_0(p). For a concrete diagonal A (e.g., diag(1,2,3)), solve the geodesic from p to 0 numerically or analytically and verify that |<dy, beta^-_p>| >= 1 - C e^{-lambda_1 D} whenever dist(p, alpha(R)) >= D, with C independent of the horizontal coordinate of p. Separately, in H^2_{-lambda_1^2}, prove the visual angle bound 4 arctan(e^{-lambda_1 D}) for a geodesic segment at distance at least D from p, and use the CAT(-lambda_1^2) inequality dist_M(p,s) <= dist_model(pbar,sbar) to transfer the distance lower bound to the comparison triangle. If both hold, Lemma 13 stands; if the numerical constant degrades with horizontal position, the controlled-volume claim is in doubt.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of Theorem 1 depends on transferring the volume growth of the approximate horosphere H to a genuine horosphere H_T via Coulhon-Saloff-Coste (Corollary 4). The transfer needs H_T to have controlled volume, which is exactly Lemma 13. The decisive assertion is inequality (24): for p with dist(p, alpha(R)) >= D, we have |g(dy, beta^-_p(0))| >= 1 - epsilon'(D). The proof sketches a comparison triangle in the model space of curvature -lambda_1^2 and asserts that the angle at p tends to 0 as D -> infinity independently of the segment parameter t, but it does not compute the model-space angle bound nor spell out the CAT(-lambda_1^2) transfer from the original triangle to the comparison triangle. If (24) failed in this uniform form, the bound on |y - y_p| for points in B_{H_T}(p,r) would fail, the differential of Pi_{y_p} would not be nearly isometric on the ball, and H_T would not have controlled volume. In constant curvature the claim is true: a geodesic segment all of whose points are at distance at least D from p subtends angle at most 4 arctan(e^{-lambda_1 D}), independent of segment length. The gap is therefore likely fillable, but it is not filled in the text.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the intrinsic geometry of horospheres in real diagonal Heintze groups G_A = R ⋉_A R^d, with metric g = dy^2 + ∑_{i=1}^d e^{-2λ_i y} dx_i^2 and 0 < λ_1 ≤ ... ≤ λ_d. Assuming λ_1 < λ_d, the main theorem states that there are exactly two isometry classes of horospheres: the Euclidean ones, with volume growth of order r^d, and the non-Euclidean ones, whose volume growth is of order r^k with k = (λ_1+...+λ_d)/λ_1. The proof constructs an approximate horosphere H as the boundary of a model horoball V ∩ {y ≤ 0}, computes its volume growth in Proposition 3, proves in Proposition 5 that H is quasi-isometric to a genuine non-Euclidean horosphere, and transfers the volume growth by the Coulhon-Saloff-Coste invariance result for controlled-volume spaces. The main technical work is in proving that the approximate and genuine horospheres have controlled volume, respectively in Corollary 3 and Lemma 13.","tokens_in":18246,"tokens_out":29287,"duration_ms":254836,"significance":"If correct, the result is a significant and clean contribution to the little-understood intrinsic geometry of horospheres in homogeneous negatively curved spaces. It gives explicit exact growth exponents for non-symmetric examples, and it shows that the isometry and quasi-isometry classifications of horospheres coincide in this family. The computed exponent matches Pansu's conformal dimension and the L^p-cohomology critical exponent, which is a strong consistency check. The volume computation for the approximate horosphere is explicit and self-contained, and the overall strategy — approximation by a polyhedral hypersurface plus a quasi-isometric transfer — is natural and potentially reusable. The principal weaknesses are local gaps in the transfer argument in Lemma 13; they are likely fillable, but they are load-bearing for the theorem.","major_comments":[{"comment":"Inequality (24) is the decisive estimate: it is what makes the projection Π_{y_p} nearly isometric on B_{H_T}(p,r), and therefore what makes H_T have controlled volume. The proof given in the paragraph after (24) is only a sketch. It asserts, from a comparison triangle in the model space of constant curvature −λ_1^2, that the angle at p 'goes to 0 when D→∞ independently of t', but the model-space angle is not computed and the uniformity in t is not demonstrated. In constant curvature the claim is true and can be bounded by an explicit angle estimate of order e^{−λ_1 D} independent of the length of the segment; the paper should supply such a bound or cite and verify the exact CAT(−λ_1^2) statement that yields it. It should also spell out how [BH99, Proposition 1.7, part 4] transfers the finite comparison-triangle angle to the ideal triangle with vertices p, α(t0) and ξ+. As written, the uniform estimate (24) is asserted rather than proved.","section":"Section 5, Lemma 13, inequality (24)"},{"comment":"After choosing T = D+r and p = (y_p,x_p) ∈ H_T with y_p < y0, the text states: 'Since dist(p,{y≥0}) ≥ |y_p| > D+r, and dist(p,HB_{ξ−}(0)) ≥ T ≥ D+r, we obtain dist(p,α(R)) ≥ D+r.' The two displayed lower bounds are not sufficient by themselves to imply the lower bound on the distance to the geodesic α; one needs a comparison principle such as dist(p,α(R)) ≥ c(b_{ξ+}(p)+b_{ξ−}(p)) for the two Busemann functions, or an explicit coordinate computation. This matters because (25), which is used to apply (24) to every q ∈ B_{H_T}(p,r), is exactly the conclusion of this step. Please provide the missing argument and, if a general lemma is used, state it with a reference.","section":"Section 5, Lemma 13, derivation of (25)"},{"comment":"The proof of Proposition 4 jumps from the pointwise derivative bounds in Lemma 10 to the ball containment (20): B_{H_{ξ+}(y_p)}(Π_{y_p}(p),(1−ε)r) ⊂ Π_{y_p}(B_H(p,r)) ⊂ B_{H_{ξ+}(y_p)}(Π_{y_p}(p),(1+ε)r). The upper inclusion follows from the derivative bound along curves, but the lower inclusion does not follow from a pointwise Jacobian bound alone: one must show that the projection is onto a full Euclidean ball of radius (1−ε)r, for example by a path-lifting or normal-coordinate argument. The same gap is inherited by the final paragraph of Lemma 13, which says that the proof 'finishes by repeating the argument of Proposition 4'. Since (20) is what converts the Jacobian estimate into a volume comparison for the balls, it is load-bearing for controlled volume.","section":"Section 3.4, Proposition 4; Section 5, Lemma 13"}],"minor_comments":[{"comment":"In the formula for ρ, the coordinate y′ is written as ŷ; the notation should be made consistent.","section":"Section 3.1, definition of ρ"},{"comment":"The geodesic α is introduced as α(t) = (0,t), which is not a geodesic for the metric (2) and contradicts the subsequent formula β^+_p(t) = (x_p,y_p+t); it should be the vertical geodesic α(t) = (t,0), or an explicitly stated equivalent.","section":"Section 5, Lemma 13"},{"comment":"The statement 'for all tangent vectors v ∈ T_pH_i^±' should quantify v ∈ T_qH (or T_qH_i^±) for q ∈ B_H(p,r); as written the quantification is inconsistent with the proof.","section":"Section 3.4, Lemma 10"},{"comment":"In the comparison-triangle paragraph, the parameter t_0 is used without definition; it should be defined by α(t_0) being the closest point of α(R) to p.","section":"Section 5, Lemma 13, comparison triangle"},{"comment":"The curvature bounds are stated as −λ_d^2 ≤ sec ≤ −λ_1^2 in Corollary 2, but the text preceding (4) writes 'between −λ_d and −λ_1'; the squares should be used consistently.","section":"Throughout, curvature bounds"},{"comment":"The references [Pan89a] and [Pan89b] are dated 1889; the correct year is 1989.","section":"References"},{"comment":"The observation that \\hat H and H differ on a compact set is used without proof; a short justification using Lemma 4 and the explicit definitions would help the reader.","section":"Section 4, Proposition 5"}],"recommendation":"major_revision","confidential_remarks":"The main result is likely correct, and the gaps are local and standard to fill; I would not reject. The paper would be strengthened by making Lemma 13's angle estimate an explicit lemma with a complete proof, and by adding the missing projection/covering argument in Proposition 4. No concerns about attribution or scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper delivers what it claims: a direct computation showing that non-Euclidean horospheres in a diagonal Heintze group have volume growth r^k with k = (λ1+...+λd)/λ1, and that isometry and quasi-isometry classes of horospheres coincide. The exponent was already known as Pansu's conformal dimension and as the L^p-cohomology critical exponent, but nobody had computed the horosphere growth itself, and the two-class statement is new. The authors cite Pansu as confirmation, not as input; there is no circularity.\n\nWhat is good: the volume growth of the approximate horosphere is computed explicitly, with clean estimates on each face and a tight ball-vs-level-set comparison. The controlled volume argument for the approximate horosphere is also solid. The quasi-isometry between the approximate and real horospheres is handled by projection arguments that are mostly standard and carefully checked.\n\nThe main soft spot is Lemma 13. Inequality (24) is load-bearing: it controls the differential of the projection onto the horizontal horosphere, and without it the transfer of volume growth via Coulhon–Saloff–Coste collapses. The proof sketches a comparison-triangle argument in constant curvature −λ1^2 and asserts that the relevant angle tends to zero uniformly, but it does not actually compute the bound. The stress-test note is right that in constant curvature the claim is true — a geodesic segment staying at distance ≥ D from p subtends angle at most about 4 arctan(e^{−λ1 D}) — so the gap is local and likely repairable. But as written it is a genuine omission, and the referee should ask for the details. Lemma 9 has a similar flavor but is less risky; its estimate is explicitly computed.\n\nThe citation pattern looks honest. The self-citations are about horospherical rigidity and flat holonomies, used as background, not as evidence for the main theorem. No invented entities, no free parameters.\n\nWho is this for: geometric group theorists and people working on negative curvature and horospheres. The main theorem is interesting at a subfield level. It deserves a serious referee — an editor should send it out, not desk reject it. The missing angle estimate should be fixed, but the architecture of the proof is sound.","headline":"A genuine first computation of horosphere volume growth in diagonal Heintze groups, with a clean two-class result; the one sketched angle estimate in Lemma 13 is a real gap but almost certainly fillable.","tokens_in":18771,"tokens_out":1484,"would_cite":true,"duration_ms":13705,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["53C21","53C24","53C30"],"pacs":[],"model":"deepseek-v4-flash","headline":"A diagonal Heintze group's horospheres form exactly two growth classes.","keywords":["Heintze groups","horospheres","volume growth","negative curvature","quasi-isometry","Busemann function","controlled volume","conformal dimension"],"falsifier":"Take $A=\\operatorname{diag}(1,2)$ in $G_A=\\mathbb{R}\\ltimes_A\\mathbb{R}^2$ and compute, numerically or analytically, the volume of intrinsic balls on the horosphere centered at $\\xi_-=(0,-\\infty)$. The theorem predicts exponent $k=(1+2)/1=3$; any other polynomial exponent, or a direct violation of inequality (24) for points with large $|x|$, would refute the claim.","tokens_in":17784,"feed_emoji":"📐","tokens_out":8608,"duration_ms":71149,"temperature":0.7,"pith_summary":"Working in a real diagonal Heintze group $G_A = \\mathbb{R}\\ltimes_A \\mathbb{R}^d$, with $A=\\operatorname{diag}(\\lambda_1,\\ldots,\\lambda_d)$ and $\\lambda_1<\\lambda_d$, the paper proves that its horospheres fall into exactly two isometry classes. The class centered at the point at infinity is isometric to Euclidean $\\mathbb{R}^d$; every other horosphere has volume growth of order $r^k$ with $k=(\\lambda_1+\\cdots+\\lambda_d)/\\lambda_1$. Because volume growth is invariant under quasi-isometries between spaces with controlled ball volumes, the same two classes are also the quasi-isometry classes. The insight is that horosphere geometry in a non-symmetric negatively curved homogeneous space can be computed explicitly, and the growth exponent matches the conformal dimension of the boundary.","feed_headline":"Horospheres in Heintze groups split into two growth classes","feed_subtitle":"Non-flat horospheres grow with exponent (λ1+···+λd)/λ1, and isometry classes match quasi-isometry classes.","key_machinery":"The load-bearing object is an approximate horosphere $\\mathcal{H}=\\partial(V\\cap\\{y\\leq 0\\})$, where $V=\\{(y,x):\\max_i e^{-\\lambda_i y/2}|x_i|\\leq 1\\}$. It is a union of $2^d+1$ smooth faces, it lies between two genuine horospheres, and its volume can be computed in coordinates: the total mass of balls of radius $r$ is comparable to $r^k$ with $k=(\\lambda_1+\\cdots+\\lambda_d)/\\lambda_1$. Convexity of $V$ supplies 1-Lipschitz orthogonal projections onto convex sets, giving a quasi-isometry between the approximate and genuine horospheres, while a controlled-volume condition (uniform upper and lower bounds on the volume of every radius-$r$ ball) makes both spaces locally doubling. A transfer principle of Coulhon and Saloff-Coste then turns the quasi-isometry plus controlled volume into comparability of ball volumes at every scale, so the explicit exponent of the approximation becomes the exponent of the actual horosphere.","core_discovery":"On the paper's own terms, the central discovery is Theorem 1: in a real diagonal Heintze group satisfying $\\lambda_1<\\lambda_d$, there are precisely two isometry classes of horospheres—the horizontal ones $H_{\\xi_+}(t)=\\{t\\}\\times\\mathbb{R}^d$, which are Euclidean, and the ones centered at any boundary point $x\\in\\mathbb{R}^d$, which are all isometric to one another and have volume growth of order $r^k$ where $k=(\\lambda_1+\\cdots+\\lambda_d)/\\lambda_1$. A direct corollary is that the quasi-isometry classification coincides with the isometry classification. The proof is constructive: an explicit piecewise-smooth approximate horosphere is shown to be quasi-isometric to the genuine horosphere and to have controlled volume, so the growth exponent transfers from the approximation to the true horosphere.","pith_inferences":["The same approximation-by-piecewise-smooth-horoball scheme could be run for non-diagonal or non-symmetric solvable extensions, where the exponentials mix coordinates; a natural guess is that the growth exponent is again trace$(A)/\\lambda_{\\min}$ whenever a single fastest-contracting direction dominates.","Because the exponent equals the conformal dimension of the boundary, one could test in larger classes of homogeneous negatively curved spaces whether a single horosphere's volume growth always computes the conformal dimension, giving a geometric probe of that quasi-isometry invariant.","A direct test with $A=\\operatorname{diag}(1,2)$ (predicted exponent $3$) would either confirm or refute the transfer step in isolation, before relying on the full theorem.","If the same controlled-volume transfer were available for horospheres covering closed negatively curved manifolds, it would likely resolve Question 1 in the affirmative, since recurrence of the strong-stable foliation would then force a single growth order; this is speculative and not proved here."],"forward_implications":["Every non-Euclidean horosphere in $G_A$ has ball volumes comparable to $r^k$, with the same exponent $k=(\\lambda_1+\\cdots+\\lambda_d)/\\lambda_1$.","There is no intermediate growth class: a horosphere in these groups is either flat or has this single non-Euclidean growth order.","Isometric and quasi-isometric classifications of horospheres agree inside a single Heintze group, closing the gap between coarse and exact geometry for this family.","The exponent coincides with the conformal dimension of the boundary at infinity, so horosphere volume growth gives a direct, computable way to see that invariant.","When $\\lambda_1<\\lambda_d$, the existence of a flat horosphere does not force constant negative curvature, even in the homogeneous setting."],"supporting_citations":[{"why":"supplies the transfer principle (Proposition 2.2) that converts quasi-isometry plus controlled volume into comparability of all ball volumes, carrying the growth exponent from the approximation to the true horosphere.","marker":"[CSC95]"},{"why":"gives the horosphere distance comparison and Jacobi-field comparison estimates used in the projection and angle arguments (Lemmas 3, 12, 13).","marker":"[HIH77]"},{"why":"provides existence and 1-Lipschitzness of orthogonal projections onto convex sets and the comparison-triangle estimates behind the quasi-isometry and angle bounds.","marker":"[BH99]"},{"why":"establishes the solvable-Lie-group model and curvature criterion that define the Heintze groups studied here.","marker":"[Hei74]"},{"why":"identifies the computed exponent as the conformal dimension of the boundary at infinity, linking the volume-growth value to a known quasi-isometry invariant.","marker":"[Pan89b]"}],"fun_headline_variants":["Two horosphere classes in Heintze groups","Heintze horospheres: exactly two isometry types","Volume growth splits Heintze horospheres into two types","Explicit growth rates for Heintze horosphere classes","Two isometry classes match quasi-isometry for Heintze horospheres"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on Lemma 13's angle estimate (inequality (24)): for points far from the vertical axis, the geodesic joining $p$ to the past endpoint is nearly horizontal, so projection onto the horizontal horosphere is almost an isometry; if that comparison-triangle estimate were wrong, the controlled-volume comparison between approximate and genuine horospheres would break.","fun_headline_variants_meta":{"raw":{"variants":["Two horosphere classes in Heintze groups","Heintze horospheres: exactly two isometry types","Volume growth splits Heintze horospheres into two types","Explicit growth rates for Heintze horosphere classes","Two isometry classes match quasi-isometry for Heintze horospheres"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000814,"raw_usage":{"total_tokens":3492,"prompt_tokens":791,"completion_tokens":2701,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":407,"completion_tokens_details":{"reasoning_tokens":2616}},"tokens_in":407,"tokens_out":2701,"duration_ms":18714,"temperature":1.0,"reasoning_tokens":2616,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T15:02:36.964465+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take $A=\\operatorname{diag}(1,2)$ in $G_A=\\mathbb{R}\\ltimes_A\\mathbb{R}^2$ and compute, numerically or analytically, the volume of intrinsic balls on the horosphere centered at $\\xi_-=(0,-\\infty)$. The theorem predicts exponent $k=(1+2)/1=3$; any other polynomial exponent, or a direct violation of inequality (24) for points with large $|x|$, would refute the claim.","supporting_citations":[],"review_version":1}