{"id":"52c78375-f3f4-49b7-9923-a6e55b519cad","arxiv_id":"2608.08276","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Very general Horikawa surfaces with p_g ≥ 5 (and first-kind with p_g = 3, 4) contain only finitely many rational or elliptic curves, with explicit counts in most strata.","lead":"This paper proves that very general Horikawa surfaces, a family of complex algebraic surfaces with minimal Chern slope, contain only finitely many rational or elliptic curves. It also gives exact counts of these curves in most moduli strata, a key step toward the Green-Griffiths-Lang hyperbolicity conjecture.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Internal contradiction in moduli input: Theorem 2.5 omits n=4 and declares n=5 second-kind moduli irreducible, while Table 3, Table 5, and §4.7 use two components for both; the 'very general' quantifier in Theorem 1.3 depends on this classification.","rationale":"The central claim is substantial and the proof chain is mostly coherent: log algebraic hyperbolicity bounds for (W,D) are lifted via Lemma 3.8 to double covers, and the explicit curve counts in Tables 4 and 5 are internally consistent, with no fitted parameters or circularity. The reader's weakest assumption—that the Gieseker moduli description in Theorems 2.4–2.5 is correct—is exactly where I find a concrete defect. The internal contradiction for second-kind moduli at n=4 and n=5 is not merely a typo: it concerns which families are irreducible components and therefore which surfaces count as 'very general'. Since Theorem 1.3 asserts a statement for all very general Horikawa surfaces with p_g≥5, and p_g≥5 includes n=4 and n=5 for the second kind, the quantifier 'very general' is not well-supported until the component structure is fixed. A secondary but genuine issue is that several results used in the proof are stated without proof: Proposition 4.10 (F0 case) and Propositions 4.41, 4.44, 4.46 plus Lemmas 4.42–4.43 in Section 4.6. These are load-bearing for the d=0 cases and for the n=4 second-kind component of type (1*); the authors say the proofs are similar to earlier ones, which is plausible but unverified. Because the proof already provides arguments for the disputed n=4 and n=5 components (Corollaries 4.29, 4.31, 4.39, 4.47), I do not see a reason to reject the mathematical claim outright. The verdict should remain conditional: the authors must reconcile the moduli statements with Horikawa/CP25 and supply or reference the omitted proofs. This does not change the reader's CONDITIONAL verdict, hence 'UNCHANGED'.","tokens_in":40977,"tokens_out":17137,"duration_ms":148276,"concrete_test":"Consult Horikawa [Hor76b] and CP25, Theorem 3.5, to list the irreducible components of M^{Hor,2}_4 and M^{Hor,2}_5, with their general types. Determine whether type (1*) is a separate component at n=4 and whether type (2*) is a separate component at n=5 or a lower-dimensional stratum deforming into type (0)/(2). Then reconcile Theorem 2.5 with Table 3, Table 5, and §4.7: if the components match Table 3, amend Theorem 2.5 to state the n=4 and n=5 exceptional cases; if they match Theorem 2.5, correct the tables and verify that every actual component's very general point is still covered by Corollaries 4.29, 4.31, 4.39, and 4.47.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing weakness is the Gieseker moduli description, not any individual inequality. Theorem 2.5(2) says that for the second kind, if 4∤n then M^{Hor,2}_n is irreducible; this applies to n=5. But Table 3, Table 5, and §4.7 treat n=5 as having two components, with the second component having general type (2*), and the proof uses Corollary 4.19 for that component. For n=4, neither Theorem 2.5(2) (since 4|4) nor Theorem 2.5(3) (which requires k≥2) applies, yet §4.7 and Table 3 again assert two components (types (1) and (1*) ← (3')). Thus the paper's foundational statement of which families are components is internally inconsistent and incomplete. Since 'very general Horikawa surface' is defined per component, the central claim (Theorem 1.3) cannot be evaluated for p_g≥5 until this is resolved. If the true moduli has a component whose general branch divisor is not the one analyzed in Section 4, the finiteness proof may miss that component; if Table 3 is correct, Theorem 2.5 must be revised. The proof does cover the alleged n=4 and n=5 components, so the claim is likely salvageable, but the current text asserts contradictory moduli facts.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies very general Horikawa surfaces, i.e. minimal general-type surfaces on the Noether boundary, and proves that they are pseudo-Lang algebraically hyperbolic: they contain only finitely many rational or elliptic curves. The main result, Theorem 1.3, covers very general Horikawa surfaces with p_g >= 5 and, in addition, very general Horikawa surfaces of the first kind with p_g in {3,4}. The proof combines Horikawa's classification of these surfaces as double covers of P^2, Hirzebruch surfaces, or their blowups, with log algebraic hyperbolicity inequalities for the corresponding branch pairs (W,D). The paper develops refined inequalities and equality analyses (Propositions 4.3, 4.7, 4.10-4.12, 4.16, 4.18, 4.24, 4.25, 4.36, 4.38, 4.44) and a lifting lemma (Lemma 3.8) to transfer information from curves on W to curves on the double cover. It also gives explicit counts and classifications of the rational and elliptic curves in each moduli stratum, summarized in Tables 4 and 5.","tokens_in":41250,"tokens_out":15390,"duration_ms":146141,"significance":"If the stated result is correct, it is a significant contribution: it establishes the Green-Griffiths-Lang-type finiteness statement for surfaces at the Noether line, where the classical inequalities c_1^2 > c_2 and c_1^2 > (3/5)c_2 are unavailable. The authors go beyond mere finiteness by explicitly characterizing and counting the exceptional rational and elliptic curves in most strata, which sharpens earlier work of Roulleau-Rousseau and Liu. The paper also contains potentially reusable techniques: refined log algebraic hyperbolicity inequalities with equality analysis, and a section-dominating collection argument adapted to the relevant blowups. The main caveat is that the moduli input, taken from the preprint [CP25], is internally inconsistent as presented, and several load-bearing propositions are stated without proof. These issues affect the evaluation of the central claim but appear fixable within the scope of the manuscript.","major_comments":[{"comment":"Theorem 2.5(2) states that M_n^{Hor,2} is irreducible whenever 4 does not divide n. Since 4 does not divide 5, this theorem applies to n=5 and asserts irreducibility. This is directly contradicted by Table 3, Table 5, and §4.7, which describe the second-kind moduli space for n=5 as having two components, with the second component having general type (2*). Moreover, Theorem 2.5 gives no statement for n=4: part (2) fails because 4 divides 4, and part (3) requires k>=2. Nonetheless, §4.7 and Table 3 again use two components for n=4, of types (1) and (1*)<-(3'). A similar omission occurs in Theorem 2.4 for the first kind with n=5, since n-1=4=4*1 falls outside both cases (2) and (3), while §4.3 and Table 4 assert two components. Because 'very general' in Theorem 1.3 is defined component-wise, the central statement cannot be evaluated until this classification is reconciled. The proof itself appears to cover the alleged n=5 and n=4 components, so the issue is likely fixable, but the current text asserts mutually incompatible moduli facts.","section":"§2.2, Theorem 2.5(2), Table 3, §4.7"},{"comment":"In the equality analysis of Proposition 4.7, the text states: 'Since f_b(C_b) is not contained in Delta_0, we must have alpha>=0 and beta>=ad.' This is not correct: an integral curve of class alpha Delta_0 + beta Gamma not equal to Delta_0 satisfies beta >= alpha d, not beta >= ad. For example, a fiber of class Gamma has alpha=0 and beta=1, and there is no reason for beta to be at least ad when a,d>1. The subsequent equality classification uses exactly the classes Gamma and Delta_0+Gamma, which generally violate the displayed condition beta>=ad. This appears to be a typo for beta>=alpha d, but because the equality analysis feeds directly into the explicit curve counts in Corollaries 4.13 and 4.15, the proof must be corrected at this point.","section":"§4.2, proof of Proposition 4.7"},{"comment":"Several propositions that are load-bearing for the main theorem are stated without proof. Proposition 4.10, the analogue of Proposition 4.7 for F_0, is dismissed with 'We omit the proof... since it is similar to and simpler than Proposition 4.7.' It is used in Proposition 4.12 and Corollary 4.15, which cover the F_0 cases contributing to p_g=3,4 and to n=5 first-kind surfaces. In §4.6, Propositions 4.41-4.44 are all introduced with 'We omit the proofs of the following results, as they are similar to those in §§4.4-4.5.' These propositions underlie Corollary 4.47, which handles the n=4 second-kind type (1*) component. These are not peripheral remarks: they are part of the chain of inequalities that establishes Theorem 1.3. The authors should include complete proofs or provide precise references to published results with matching hypotheses and conclusions.","section":"§4.2 and §4.6, Propositions 4.10, 4.41-4.44"}],"minor_comments":[{"comment":"Table 5 contains question marks for the n=4 type (3') stratum and for the strata d=(n+3)/3 in the second kind, and the text says the results do not apply there. This should be stated explicitly in the abstract or introduction, since the abstract's phrase 'explicit characterization and count' could otherwise be read as covering every stratum.","section":"Table 5 and §4.7"},{"comment":"There are several typographical errors that should be corrected: 'isomophism' in §2.2, 'becasue' in §4.2, 'projecitve' in §4.4, and the spelling 'HORIKA W A' and 'WERN YEONG' in the header.","section":"Throughout"},{"comment":"The text says 'the moduli space has two components' for n=5 first kind, but Theorem 2.4 contains no statement for n=5. Please either add the missing case to Theorem 2.4 or explain why the component decomposition is known independently of the theorem as stated.","section":"§4.3, n=5 first kind"},{"comment":"The summary for n=5 second kind says that the first component has 80 or 81 elliptic curves and the second component has none. This depends on the two-component structure that is contradicted by Theorem 2.5(2), so the numerical summary should be reorganized after the moduli issue is resolved.","section":"§4.7, n=5 second kind"}],"recommendation":"major_revision","confidential_remarks":"The internal inconsistency between Theorem 2.5 and Tables 3/5 is the most serious issue; the authors will need to align the manuscript with the precise statement in [CP25] or with Horikawa's original classification. Since Theorems 2.4 and 2.5 are quoted from a preprint, the published version should indicate which formulation is being used. The paper also relies heavily on several same-group preprints ([CRY22], [ATY24], [IMRY25]), and the referee cannot fully verify the adaptation of those techniques from the present text; for a journal of this level, the omitted proofs flagged above should be supplied rather than deferred."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main theorem is genuinely new and the curve counts are a nice bonus. The paper proves that very general Horikawa surfaces with p_g ≥ 5 (and first kind with p_g = 3, 4) are pseudo-Lang algebraically hyperbolic, resolving questions from Roulleau–Rousseau. The techniques—refined Chen inequalities plus section-dominating collections—are extended to blowups of Hirzebruch surfaces, which is a real step beyond what was there. The enumerative tables are consistent with the inequality analysis and look plausible.\n\nThe moduli input is where the trouble is. Theorem 2.5(2) says M^{Hor,2}_n is irreducible when 4 ∤ n, which covers n = 5, and Theorem 2.5 has no case for n = 4 at all. But Table 3, Table 5, and §4.7 treat n = 4 and n = 5 as having two components each. The same gap appears in Theorem 2.4 for the first kind: n = 5 is covered by neither case (2) nor (3). Since a very general Horikawa surface is defined per component, the central claim cannot be evaluated until this is resolved. This is not a minor typo; it is a load-bearing inconsistency in the paper's own foundation.\n\nThere are also omitted proofs: Proposition 4.10 (the F_0 case) and the Section 4.6 results are stated without proof. For a preprint that is tolerable, but they are needed for the summary tables and should be supplied. The paper also leans on the CP25 preprint for the moduli description, which is fine, but the contradiction above is internal to this text.\n\nThat said, the core inequality chain is coherent, the lifting lemma is sound, and the proof does not assume what it wants to prove. The n = 4 and n = 5 components that the tables use are actually analyzed in Section 4, so the theorem is likely salvageable with a corrected moduli statement.\n\nThis paper deserves a serious referee. The referee should ask the authors to fix the moduli inconsistency and supply the missing proofs; after that, the result should be accepted.","headline":"A genuinely new theorem with explicit curve counts, but the moduli input has an internal contradiction that must be fixed before the main claim is fully stated.","tokens_in":701,"tokens_out":752,"would_cite":true,"duration_ms":31453,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14J29","32Q45","14E20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Very general Horikawa surfaces with geometric genus at least 5 contain only finitely many rational or elliptic curves, and the same holds for first-kind Horikawa surfaces with geometric genus 3 or 4.","keywords":["Horikawa surfaces","algebraic hyperbolicity","pseudo-Lang algebraic hyperbolicity","Green–Griffiths–Lang conjecture","surfaces of general type","log pairs","kernel bundles","rational and elliptic curves"],"falsifier":"Exhibit a very general Horikawa surface with $p_g\\ge5$ whose canonical model contains infinitely many rational or elliptic curves, or contains a rational or elliptic curve not lying in the paper's explicit list of preimages of bitangent lines, tangent fibers, zero-section components, and exceptional divisors. A concrete numerical check: for $n=2$, the paper's count of 1320 elliptic curves on the degree-8 plane double cover is exactly the number of bitangent lines to a smooth plane octic, so a very general octic with a different bitangent count would falsify that corollary.","tokens_in":40716,"feed_emoji":"📐","tokens_out":8046,"duration_ms":71518,"temperature":0.7,"pith_summary":"This paper proves that very general Horikawa surfaces—the minimal surfaces of general type sitting on the boundary of the Noether inequality, with the smallest possible Chern slope—are pseudo-Lang algebraically hyperbolic: they contain only finitely many rational or elliptic curves. The main theorem covers every very general Horikawa surface with geometric genus $p_g\\ge5$, and also those of the first kind with $p_g\\in\\{3,4\\}$, which includes double covers of $\\mathbb{P}^2$ branched over a degree 8 curve whose elliptic curves had been found but not shown to be finite. The proof identifies each such surface, up to birational equivalence, as a double cover of a rational surface branched over a very general divisor, and then shows that any rational or elliptic curve in the cover must lie over one of a short list of base curves. In nearly every stratum of the moduli space the paper goes further and counts these curves exactly, for example 1320 elliptic curves and no rational curves for the degree-8 plane double cover with $p_g=3$.","feed_headline":"Most Horikawa surfaces have finitely many rational or elliptic curves","feed_subtitle":"A new proof pins down every such curve explicitly, giving exact counts in almost every stratum.","key_machinery":"The mechanism is log algebraic hyperbolicity of the pair $(W,D)$, where $W$ is the rational surface that the Horikawa surface double-covers and $D$ is its branch divisor. For each family the divisor decomposes as $D=D'+F$ with $D'$ basepoint-free and $F$ fixed; the paper constructs a section-dominating collection of line bundles for $D'$ and proves that the log tangent bundle $T_W(-\\log F)$ is pseudonef outside a small exceptional set. Proposition 3.15 then bounds the degree of the log normal sheaf of a curve by the degree of a restricted kernel bundle, and Proposition 3.17 identifies the equality cases with sections that force the curve to lie in a few explicit classes: fiber classes, zero-section components, exceptional divisors, or bitangent lines. These sharp inequalities convert a finiteness statement into the explicit curve counts recorded in Tables 4 and 5.","core_discovery":"The central discovery is that the Green–Griffiths–Lang prediction can be verified at the extreme boundary of the geography of surfaces of general type: even at minimal Chern slope, a very general Horikawa surface does not admit infinitely many low-genus curves. Theorem 1.3 states that a very general Horikawa surface with $p_g\\ge5$ is pseudo-Lang algebraically hyperbolic, and so is a very general Horikawa surface of the first kind with $p_g\\in\\{3,4\\}$. The proof establishes refined log algebraic hyperbolicity inequalities for the base pairs $(\\mathbb{P}^2,D)$, $(F_d,D)$, and blowups of $F_d$ at one or two points on a fiber; equality in these inequalities forces the curve into an explicit finite list of classes, and Lemma 3.8 transfers the finiteness to the double cover. The result is therefore not just a finiteness theorem but a complete geometric description of the rational and elliptic curves on very general Horikawa surfaces in almost every stratum.","pith_inferences":["The equality-case analysis through kernel bundles is likely to transfer to other families of cyclic covers of rational surfaces on the Noether boundary where the classical Chen bounds are not sharp.","The explicit counts suggest an enumerative interpretation: the elliptic curves are governed by bitangent and tangency formulas on the branch divisors, so the numbers in Tables 4 and 5 could be checked independently by intersection theory.","The theorem is conditional on the moduli description taken from the cited preprint; if a component of the Gieseker moduli space turned out to have a different general type or a special branch locus, the finiteness conclusion for that component would need to be re-examined.","A natural next step, which the paper leaves open, is to decide whether the finitely many exceptional curves can be removed so that the surfaces become pseudo-Demailly algebraically hyperbolic rather than merely pseudo-Lang."],"forward_implications":["Every very general Horikawa surface with $p_g\\ge5$ satisfies the Green–Griffiths–Lang prediction in its pseudo-Lang form, since it contains only finitely many rational or elliptic curves.","The previously open cases with $p_g=3,4$ of the first kind are settled: for example, the degree-8 plane double cover has no rational curves and exactly 1320 elliptic curves, all preimages of bitangent lines.","In almost every stratum of the moduli space, the rational and elliptic curves are explicitly characterized as preimages of bitangent lines, tangent fibers, zero-section components, or exceptional divisors, with exact counts given in Tables 4 and 5.","The finiteness claim holds for very general members of every component with $p_g\\ge5$, even though the explicit count is left open for a few lower-dimensional strata such as type $d=(n+3)/3$ of the second kind."],"supporting_citations":[{"why":"Classifies Horikawa surfaces of the first kind as double covers of $\\mathbb{P}^2$ or Hirzebruch surfaces, fixing the linear systems of the branch divisors.","marker":"[Hor76a]"},{"why":"Classifies Horikawa surfaces of the second kind as double covers of Hirzebruch surfaces and their blowups, fixing the branch linear systems used in Section 4.","marker":"[Hor76b]"},{"why":"Supplies the Gieseker moduli description (Theorems 2.4 and 2.5) that identifies the irreducible components and the very general branch locus for each geometric genus.","marker":"[CP25]"},{"why":"Provides the orbifold-curve framework that turns curves on the double cover into orbifold curves on $(W,\\frac12 D)$, and gives the earlier hyperbolicity results that this paper sharpens.","marker":"[RR13]"},{"why":"Establishes the base log algebraic hyperbolicity inequalities for $(\\mathbb{P}^2,D)$ and $(F_d,D)$ that the paper refines with equality cases.","marker":"[Che01]"},{"why":"Supplies the variational argument and the kernel-bundle mechanism for detecting equality in the log algebraic hyperbolicity bounds.","marker":"[CRY22]"},{"why":"Introduces section-dominating collections of line bundles and the surjection from kernel bundles used throughout the proof.","marker":"[CR23]"},{"why":"Provides the positivity criterion for log tangent bundles outside exceptional sets that underpins Proposition 3.15 for the blowup cases.","marker":"[IMRY25]"}],"fun_headline_variants":["Almost all Horikawa surfaces have finitely many low-genus curves","Very general Horikawa surfaces are algebraically hyperbolic","Horikawa surfaces: finite low-genus curves, explicitly listed","Most Horikawa surfaces have only finitely many rational/elliptic curves","Horikawa surfaces: explicit finite list for low-genus curves"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The moduli-space description used as input—that each irreducible component for a fixed geometric genus is one of the families in Table 3 and that a very general point has a branch divisor very general in its linear system—must be correct, because every hyperbolicity bound in Section 4 is proved for those families and that generality.","fun_headline_variants_meta":{"raw":{"variants":["Almost all Horikawa surfaces have finitely many low-genus curves","Very general Horikawa surfaces are algebraically hyperbolic","Horikawa surfaces: finite low-genus curves, explicitly listed","Most Horikawa surfaces have only finitely many rational/elliptic curves","Horikawa surfaces: explicit finite list for low-genus curves"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002331,"raw_usage":{"total_tokens":8939,"prompt_tokens":851,"completion_tokens":8088,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":467,"completion_tokens_details":{"reasoning_tokens":8002}},"tokens_in":467,"tokens_out":8088,"duration_ms":52378,"temperature":1.0,"reasoning_tokens":8002,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:13:59.376890+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a very general Horikawa surface with $p_g\\ge5$ whose canonical model contains infinitely many rational or elliptic curves, or contains a rational or elliptic curve not lying in the paper's explicit list of preimages of bitangent lines, tangent fibers, zero-section components, and exceptional divisors. A concrete numerical check: for $n=2$, the paper's count of 1320 elliptic curves on the degree-8 plane double cover is exactly the number of bitangent lines to a smooth plane octic, so a very general octic with a different bitangent count would falsify that corollary.","supporting_citations":[],"review_version":1}