{"id":"cb649715-9c25-484d-b203-234a8d9e0b53","arxiv_id":"2501.09266","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every fixed k and every ε>0, the maximum of λ_k - λ_{k-1} over genus-g hyperbolic surfaces with systole at least ε tends to 1/4 as g→∞.","lead":"This paper proves that on any compact chunk of the moduli space of large-genus hyperbolic surfaces, the largest possible gap between two consecutive Laplacian eigenvalues approaches 1/4. It matters because it extends the celebrated 2023 Hide-Magee result from the whole moduli space to thick parts, where surfaces have a lower bound on their shortest closed geodesic.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; Proposition 3.1's random-input assembly is sound, and the uncited spectral fact about the three-punctured sphere is a citation issue, not a correctness risk.","rationale":"The reader correctly identified Proposition 3.1 as the place where the construction's input is generated, and correctly noted that the base spectral fact is asserted without citation. But the probabilistic combination is sound: two a.a.s. events and one positive-probability event have positive intersection probability in the limit, and connectedness is a.a.s. by Dixon's theorem. The base spectral fact, while uncited, is a classical property of the congruence surface Γ(2) and is needed only in the weak form that there is no discrete eigenvalue below 1/4; even if λ1(S0,3)=1/4 exactly, the argument still works because λ̄1(S0,3)=1/4. I therefore do not view this as a load-bearing correctness risk, only a missing citation. I also checked the main flow: Theorem 1.2 converts the spectral gap of the non-compact surface to a Neumann eigenvalue lower bound on the compact core; Theorem 1.3 gives a uniform boundary-perturbation stability result; the endgame glues k pieces and applies the mini-max principle to force λk−1→0 and λk≥1/4−O(δ); Cheng's upper bound supplies the matching limsup. No evident gap or circularity was found. Hence the reader's conditional verdict should stand unchanged, with the base spectral fact supplied as a citation or short proof.","tokens_in":43751,"tokens_out":27637,"duration_ms":292044,"concrete_test":"Verify by literature search or a rigorous spectral computation that the thrice-punctured sphere H/Γ(2) has no square-integrable eigenvalue in the interval (0,1/4), i.e. λ1(S0,3)≥1/4; if confirmed, the uncited assertion in Proposition 3.1 is harmless and the starting-surface input is secured.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I could not find a load-bearing flaw in the central argument. The most delicate input is Proposition 3.1, which assembles finite-area non-compact starting surfaces with large systole, large cusps, and λ̄1 close to 1/4. The proof combines a.a.s. events (connectedness, spectral gap, large cusps) with a positive-probability event (systole). This combination is legitimate: for finitely many events whose probabilities tend to 1 and one event with positive liminf, the intersection has positive liminf, so existence for each large degree follows. The uncited assertion λ1(S0,3)>1/4 is the only soft spot: it is stated without reference in the proof of Proposition 3.1. However, the construction only needs no discrete eigenvalue below 1/4, and for the congruence surface H/Γ(2) this is a classical Selberg-type fact. Thus the assertion is very likely true and, even in the weaker form λ1(S0,3)≥1/4, would suffice. No circularity or internal inconsistency emerged in the later sections; the estimates in Section 8 are lengthy but appear to be a self-contained deformation argument with uniform constants.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that for every fixed k≥1 and every fixed ε>0, the maximum of λ_k−λ_{k−1} over the ε-thick part of the moduli space of closed hyperbolic surfaces of genus g tends to 1/4 as g→∞. The proof has three main ingredients: (i) a comparison theorem (Theorem 1.2) between the first non-zero spectrum of a finite-area non-compact surface and the first Neumann eigenvalue of a compact core obtained by cutting off large cusps; (ii) a uniform stability theorem (Theorem 1.3) for first Neumann eigenvalues under small changes of boundary lengths of a bordered hyperbolic surface; and (iii) an input from random covers (Proposition 3.1) producing finite-area starting surfaces with large spectral gap, large cusps, and long systole. These ingredients are then combined by gluing several copies of a bordered surface with two boundary components to obtain closed surfaces with large k-th spectral gap. The paper also contains several auxiliary results, including a mass-distribution estimate for eigenfunctions on collars and a detailed deformation argument for pairs of pants.","tokens_in":43962,"tokens_out":52402,"duration_ms":547127,"significance":"If the proof is completed as intended, the result is a natural and significant extension of the Hide–Magee theorem and of the authors' earlier work with Zhu to the compact thick part of moduli space. The paper gives a genuinely new compactification procedure that avoids the long thin collars produced by earlier Buser–Burger–Dodziuk-type methods. Theorem 1.2 and Theorem 1.3 are of independent interest: the latter provides a uniform spectral stability statement with explicit constants that is not available elsewhere. The use of random covers is appropriate, and the proof is largely self-contained modulo the cited probabilistic theorems. The manuscript is carefully written, with detailed computations in Section 8.","major_comments":[{"comment":"The systole bound for the glued surfaces is not justified for arbitrary ε>0. In Step 4, a new simple closed geodesic δ in Y_{g,2} that intersects a glued boundary component γ'_j of length ε is ruled out by 'by (9) we find that it always has length ≥ε.' However, equation (9) gives only ℓ(δ) ≥ 2 arcsinh(1/sinh(ε/2)), which for ε>2 arcsinh 1 is strictly smaller than ε. The presence of an embedded half-collar of width ε/2 does not force a geodesic crossing the seam to reach the outer boundary of that half-collar; a geodesic can cross the seam while staying arbitrarily close to it, so the half-collar alone does not supply the claimed lower bound. The same gap appears in the final gluing of k copies of Y_{g,2} in the proof of Theorem 1.1, where Z_g is asserted to lie in M^{≥ε}_g. As written, the argument establishes the claimed systole control only for ε ≤ 2 arcsinh 1. A twist-selection argument, or an additional estimate controlling the shortest geodesic crossing the identified boundary components, is needed for the theorem as stated for all fixed ε>0.","section":"Proposition 1.4, Step 4, and proof of Theorem 1.1"}],"minor_comments":[{"comment":"The proof states without citation that λ1(X)>1/4 for X=S0,3, the three-punctured sphere. This is a classical Selberg-type fact and is very likely true, but a reference should be provided. Moreover, the argument only needs that S0,3 has no eigenvalues in [0,1/4−δ], so the statement could be relaxed to 'λ1(S0,3) ≥ 1/4' or 'no discrete eigenvalue in [0,1/4−δ]'.","section":"Proposition 3.1, proof"},{"comment":"Lemma 8.28 states that inequality (92) holds with δ0 = 54e^{2\\bar h}δd, while Lemma 8.17 uses δ0 = 54e^{2\\bar h}√δd. The former is stronger and the combination is valid, but the notation should be reconciled to avoid apparent inconsistency.","section":"Lemma 8.28 and Lemma 8.17"},{"comment":"The strict inequality ds²_X < ds²_{X^fu} on the stated subsurface is asserted in the statement of Theorem 4.2, but the proof only cites Lemma 4.5 together with Lemma 4.4(3), which give (1−δ)ds²_{X^δ} ≤ ds²_{X^fu} and ds²_{X^δ} ≥ ds²_X. The strict inequality follows from the domain monotonicity of Poincaré metrics (Ahlfors–Schwarz), and this should be stated explicitly.","section":"Theorem 4.2, Part (3)"},{"comment":"In the estimate of the second term in (67), the paper uses dvol_X ≤ dvol_Y on the collars C(√w). This is true because X^fu ⊂ X and the Poincaré metric of a subdomain dominates the ambient one, but it is not explicitly justified at that point.","section":"Lemma 6.4"}],"recommendation":"major_revision","confidential_remarks":"The paper is strong and the main strategy is convincing, but the large-ε systole control after gluing is a genuine gap in a load-bearing part of the argument. The authors should be asked to either prove a twist-selection lemma ensuring that gluing along boundary components of length ε does not create geodesics shorter than ε, or explain a different argument for the case ε>2 arcsinh 1. The remaining issues are minor and local."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper proves the thick-part analogue of Buser's conjecture: for any fixed k and any fixed epsilon, the maximum of the k-th spectral gap (lambda_k - lambda_{k-1}) over the epsilon-thick part of moduli space tends to 1/4 as genus goes to infinity. The k=1 case on the thick part was already known via the Hide-Magee appendix plus Cheng's upper bound, and the paper says so honestly. The new content is the higher-gap statement for k at least 2, which is not in the cited literature. That alone would make the paper worth reading, but the method is also its own contribution: a new cusp-cutting compactification that keeps systole bounded below, a comparison theorem between the original and truncated metrics, and a uniform Neumann eigenvalue stability theorem with explicit constants. The proof is detailed and appears structurally sound.\n\nThe soft spots are minor. The proof of Proposition 3.1 states without citation that the three-punctured sphere has lambda_1 > 1/4. This is a classical fact (or follows from Selberg-type reasoning for the congruence surface H/Gamma(2)), and the construction only needs no discrete eigenvalue below 1/4, so even the weaker statement would suffice. It is a citation issue, not a correctness risk. The heavy computations in Section 8, comparing hyperbolic pairs of pants, are not machine-checked but are standard hyperbolic trigonometry worked out with explicit bounds; I did not find a hidden error. The combination of random-cover results in Proposition 3.1 is legitimate: several asymptotically almost sure events and one positive-probability event intersect with positive liminf, giving existence for each large degree. No circularity or fitted constants enter the argument.\n\nWho gets value from this paper: spectral geometers, people working on moduli spaces, and anyone following the recent random-cover constructions of high-eigenvalue surfaces. The paper is technically demanding but self-contained in the sense that the new analytic steps are carried out rather than outsourced. It deserves a serious referee, and with the small citation fixed it would be a solid publication. I would recommend engaging with it.","headline":"A serious and mostly convincing proof that thick parts of moduli space contain surfaces with spectral gaps approaching 1/4, genuinely new for higher eigenvalues and built on a careful new compactification.","tokens_in":690,"tokens_out":850,"would_cite":true,"duration_ms":23826,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["58J50","30F60","32G15"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that in every thick part of moduli space, the maximum $k$-th spectral gap of closed hyperbolic surfaces tends to $\\tfrac14$ as the genus tends to infinity.","keywords":["spectral gaps","hyperbolic surfaces","moduli space","thick part","Laplacian eigenvalues","random covers","Neumann eigenvalues","three-punctured sphere"],"falsifier":"Compute the discrete spectrum of the three-punctured sphere: if it has any non-zero eigenvalue below $\\tfrac14$, then Proposition 3.1 cannot hold for small $\\delta$, since random covers inherit the base spectrum below $\\tfrac14-\\delta$; a direct numerical check of this base spectral fact would settle the input to the construction.","tokens_in":43531,"feed_emoji":"📐","tokens_out":12189,"duration_ms":110428,"temperature":0.7,"pith_summary":"This paper proves that on every thick part of the moduli space of closed hyperbolic surfaces—the compact region where the shortest closed geodesic has length at least a fixed $\\varepsilon>0$—the largest possible $k$-th spectral gap tends to $\\tfrac14$ as the genus $g$ goes to infinity, for every fixed $k\\geq1$. Previously, the value $\\tfrac14$ was known as the asymptotic supremum over the whole moduli space, but the surfaces realizing it degenerated toward the boundary. Here the gap $\\lambda_k-\\lambda_{k-1}$ is shown to approach $\\tfrac14$ for surfaces whose systole is bounded below, so the extremal surfaces stay in the interior. The proof constructs such surfaces by cutting off large cusps from random covers of the three-punctured sphere, preserving the spectral bottom in the first Neumann eigenvalue, then gluing $k$ copies together.","feed_headline":"Thick-part surfaces reach the universal spectral gap 1/4","feed_subtitle":"For every fixed k, the maximum k-th Laplacian gap over surfaces with systole ≥ ε tends to 1/4 as genus grows.","key_machinery":"The argument runs on a cusp-compactification comparison. Starting from a finite-area non-compact hyperbolic surface $X$ with large cusps of length $l$, one removes the cusp ends below a fixed horocycle length $\\epsilon_0$ and takes the compact convex core $Y$ of the remaining infinite-area surface; Theorem 1.2 shows that when $l$ is large, the first Neumann eigenvalue $\\sigma_1(Y)$ is at least $(1-4\\delta)\\bar\\lambda_1(X)-\\delta$, so the spectral bottom of the non-compact surface is almost preserved. Theorem 1.3 then perturbs the boundary lengths of $Y$ by factors $1+\\delta_i$ through bi-Lipschitz maps whose deviation and Neumann-eigenvalue change are $O(\\sqrt{\\delta})$, independent of the surface, so the boundaries can be made equal and glued.","core_discovery":"The central discovery is that the universal bound $\\tfrac14$ for spectral gaps is attained inside every thick part: for any fixed $\\varepsilon>0$ and fixed $k$, one has $\\lim_{g\\to\\infty}\\max_{X_g\\in M_g^{\\geq\\varepsilon}}(\\lambda_k(X_g)-\\lambda_{k-1}(X_g))=\\tfrac14$. For $k=1$ this says the maximum of $\\lambda_1$ over the $\\varepsilon$-thick part tends to $\\tfrac14$; for general $k$, the constructed surfaces have $\\lambda_k\\to\\tfrac14$ while the first $k-1$ eigenvalues are pushed down to zero, so the $k$-th gap alone carries the whole spectral radius. The matching upper bound comes from the standard eigenvalue comparison estimate, so the limit is exact.","pith_inferences":["The constant $\\tfrac14$ is the bottom of the continuous spectrum of non-compact finite-area hyperbolic surfaces; this suggests that the extremal spectral behaviour of large-genus thick-part surfaces is inherited from an idealized cusp-like spectrum even though the final closed surfaces have no cusps.","Theorem 1.3 is a genus-independent stability statement for Neumann eigenvalues under boundary-length changes, so it may be useful for other problems where bordered hyperbolic surfaces with slightly different boundary geometry must be compared.","The random-cover input suggests a concrete numerical probe: sample random covers of the three-punctured sphere, cut off long cusps, adjust boundary lengths, and check empirically whether $\\lambda_k-\\lambda_{k-1}$ approaches $\\tfrac14$; such simulations could indicate how large the genus must be before the asymptotic is visible.","The paper hints at a possible extension to arithmetic hyperbolic surfaces; a testable question is whether explicit arithmetic sequences with uniformly bounded systole can realize the $\\tfrac14$ gap."],"forward_implications":["For every $\\varepsilon>0$ and every fixed $k$, there are closed hyperbolic surfaces of arbitrarily large genus, with systole at least $\\varepsilon$, whose $k$-th Laplacian gap is within any prescribed tolerance of $\\tfrac14$; the first $k-1$ gaps are simultaneously pushed down to zero.","The maximum of $\\lambda_1$ over the $\\varepsilon$-thick part converges to $\\tfrac14$, so the known optimal first-eigenvalue bound holds uniformly on compact subsets of moduli space rather than only near the boundary.","The matching upper bound shows the value $\\tfrac14$ is sharp: for large genus no surface in the thick part can have a $k$-th gap larger than $\\tfrac14$.","The compactification used here avoids the older procedure that forces closed geodesics to become very short, so the resulting surfaces remain in the thick part by construction.","For the first two distinct eigenvalues, the gap between the second and first distinct eigenvalue also tends to $\\tfrac14$ on the thick part."],"supporting_citations":[{"why":"Proves random covers of the three-punctured sphere inherit the base spectrum below $\\tfrac14-\\delta$, providing the large spectral bottom required in Proposition 3.1.","marker":"[HM23]"},{"why":"Shows that random covers have large cusps of any fixed length asymptotically almost surely, a hypothesis needed for the cusp-compactification comparison.","marker":"[KM24]"},{"why":"Independently proves the same large-cusp statement for random covers and is cited as an alternative basis for Theorem 3.3.","marker":"[Mag24]"},{"why":"Gives the limiting probability that a power of a random permutation has no fixed points, the core input for the systole lower bound.","marker":"[Nic94]"},{"why":"Establishes asymptotic independence of fixed-point counts for several words, allowing the systole bound to apply to all short geodesics at once.","marker":"[PZ24]"},{"why":"Supplies the degenerating mini-max principle and the gluing argument that this paper adapts to turn bordered surfaces with large $\\sigma_1$ into closed surfaces with a large $k$-th gap.","marker":"[WZZ24]"},{"why":"Provides the method of comparing Poincaré metrics and spectra after cutting off large cusps, used in Theorem 4.2 and Theorem 1.2.","marker":"[Bro99]"},{"why":"Contributes the handle lemma and compactification technique that underlies the comparison between non-compact spectra and spectra of compact bordered surfaces.","marker":"[BM01]"},{"why":"Gives the eigenvalue upper bound $\\limsup_g \\sup_{M_g}\\lambda_k \\leq \\tfrac14$ used to match the lower bound and force the limit.","marker":"[Che75]"}],"fun_headline_variants":["Thick-part surfaces' spectral gaps converge to 1/4","Spectral gap limit 1/4 on thick moduli spaces","Universal 1/4 gap emerges on thick parts in genus limit","Maximum spectral gap on thick parts tends to 1/4"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof needs, for every $\\varepsilon$, $l$, $\\delta$ and all sufficiently large $i$, a finite-area non-compact hyperbolic surface of Euler characteristic $-i$ with systole at least $\\varepsilon$, cusps of length at least $l$, and no non-zero Laplacian spectrum below $\\tfrac14-\\delta$; this is assembled from four separate probabilistic theorems about random covers of the three-punctured sphere, and the proof also uses without citation that this base surface has no small eigenvalue.","fun_headline_variants_meta":{"raw":{"variants":["Thick-part surfaces' spectral gaps converge to 1/4","Spectral gap limit 1/4 on thick moduli spaces","Universal 1/4 gap emerges on thick parts in genus limit","Maximum spectral gap on thick parts tends to 1/4"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000278,"raw_usage":{"total_tokens":1560,"prompt_tokens":757,"completion_tokens":803,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":373,"completion_tokens_details":{"reasoning_tokens":729}},"tokens_in":373,"tokens_out":803,"duration_ms":8154,"temperature":1.0,"reasoning_tokens":729,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:07:54.374421+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the discrete spectrum of the three-punctured sphere: if it has any non-zero eigenvalue below $\\tfrac14$, then Proposition 3.1 cannot hold for small $\\delta$, since random covers inherit the base spectrum below $\\tfrac14-\\delta$; a direct numerical check of this base spectral fact would settle the input to the construction.","supporting_citations":[],"review_version":1}