{"id":"c593b3e6-e441-46b6-bc80-1ba865fb3c74","arxiv_id":"2501.09269","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For each degree, the moduli space of rational curves on a general Artin-Mumford double solid has exactly two (for lines) or four (higher degrees) irreducible components, giving the first multiple-component verification of Geometric Manin's Conjecture.","lead":"This paper enumerates all families of rational curves on Artin-Mumford double solids, finding two families of lines and four families for each higher degree. It provides the first example where Geometric Manin's Conjecture holds with multiple genuinely moving families, reflecting a nontrivial Brauer group.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.9's equality R^{++}=R^{--} rests on an unproved sign-map property: \"one can see\" that the sign map on the 56 lines of S satisfies Lemma 2.12. This is load-bearing, though likely fixable via constancy of algebraic equivalence.","rationale":"The reader's conditional verdict is appropriate. The sign-map hypothesis in Lemma 2.12 is indeed the weakest point in the proof of Theorem 4.9: the assertion 'one can see' is unsupported, and the equality R^{++}=R^{--} depends on it. This is load-bearing because all higher-degree components are obtained by induction from the conic case. However, the property is very plausible and likely follows from the constancy of algebraic equivalence in the family of conics on S, combined with Lemma 3.6 and the fact that there are exactly two algebraic classes in the numerical class 2\\tilde{\\ell}. Thus the gap is a missing justification rather than a demonstrated falsehood. The other concern, reliance on the unpublished preprint [Oka24a] for the classification of breaking morphisms and a-covers, is real but secondary; it affects the proof of Corollary 4.12 and the irreducibility of the higher-degree gluing loci, yet the paper's main theorem would still be credible if those citations are correct. The proposed finite check would settle the sign-map issue directly and is feasible with the combinatorial data already present in the paper. No ad hominem or theatrical language is needed; the verdict CONDITIONAL remains the right one until the missing argument is supplied.","tokens_in":18882,"tokens_out":43127,"duration_ms":382318,"concrete_test":"Verify the sign-parity property by a finite computation: using the classification of the 126 conic bundles on a degree-2 del Pezzo surface in §2.2 and the two algebraic classes of lines from Lemma 3.6, assign to each of the 56 lines its class (a or b), and for each conic bundle check that the parity of the two components in every singular fiber is constant. If the property holds for all 126 cases, Lemma 2.12 applies and R^{++}=R^{--} is justified; if any conic bundle has mixed parity, the base case of Theorem 4.9 is false.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central base case is Theorem 4.9, where the equality R^{++}=R^{--} is obtained by applying Lemma 2.12 to the sign map on the 56 lines of a general hyperplane section S. The paper asserts \"one can see that this map satisfies all the assumptions of Lemma 2.12\" without proof. The first assumption (Geiser involution pairs opposite signs) follows because the Geiser involution on S is ι|_S and ι interchanges M_1^+ and M_1^-; this is standard. The second assumption — that every conic bundle on S has constant sign parity across its six singular fibers — is not automatic and is not demonstrated. If it fails for some conic bundle, Lemma 2.12 need not produce a conic bundle with both (+,+) and (−,−) fibers, and the argument for R^{++}=R^{--} collapses; since the d≥2 description is built by induction on this base case, Theorem 1.1(2) and Corollary 1.3 would lose their foundation. The property is likely true: for a conic bundle on S, the strict transforms of its general fibers on \\tilde{X} form an algebraic equivalence class; by Lemma 3.6 there are exactly two classes in the numerical class 2\\tilde{\\ell}, corresponding to even (same-sign) and odd (opposite-sign) fiber parity. But this argument does not appear in the paper, leaving a genuine gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the moduli spaces of rational curves on Artin-Mumford double solids, which are double covers of P^3 branched along quartic symmetroids. The main theorem (Theorem 1.1) asserts that for a general Artin-Mumford double solid X, the space Mor(P^1,X,1) of H-lines has exactly two irreducible components M_1^+ and M_1^- of expected dimension 5, and for each d >= 2 the space Mor(P^1,X,d) has exactly four irreducible components R_d^+, R_d^-, N_d^+, N_d^- of expected dimension 2d+3, where N_d^+ and N_d^- parametrize d-sheeted covers of lines in M_1^+ and M_1^-, and R_d^+ and R_d^- generically parametrize embedded very free curves. The paper also proves a strong movable bend-and-break statement (Theorem 1.2/4.10) and derives Geometric Manin's Conjecture for these varieties (Corollary 1.3). The proof combines the geometry of Reye congruences (Enriques surfaces) to classify lines and conics, a combinatorial lemma on conic bundles on degree-2 del Pezzo surfaces (Lemma 2.12), and results from the author's previous preprint [Oka24a] on Fano threefolds with Gorenstein terminal singularities.","tokens_in":19203,"tokens_out":3882,"duration_ms":38194,"significance":"If the main theorem is correct, it provides the first example of Fano varieties whose moduli spaces of rational curves contain multiple Manin components for every sufficiently positive degree, exactly as predicted by Geometric Manin's Conjecture. This is a meaningful advance for the geometric Manin program, which until now has mostly produced examples with a unique Manin component. The paper also contains a self-contained and well-reasoned analysis of the H-line components via Reye congruences, including the non-normality statement in Remark 4.4 and the irreducibility of incidence loci in Lemma 4.6. The main caveats are that the proof of the d=2 base case relies on an unproved assertion about a sign map on the 56 lines of a hyperplane section, and that several key inputs (classification of a-covers, movable bend-and-break, and the Manin-component count) are quoted from the unpublished preprint [Oka24a]. These issues are local and apparently repairable, but they make the current version conditional.","major_comments":[{"comment":"The equality R^{++}=R^{--} is the base step for the four-component description for all d >= 2, and it is obtained by applying Lemma 2.12 to the sign map on the 56 lines of a general hyperplane section S. The proof states only 'one can see that this map satisfies all the assumptions of Lemma 2.12' and verifies the Geiser-involution pairing. It does not prove the second assumption of Lemma 2.12, namely that for every conic bundle on S, if one singular fiber has components of opposite sign then every singular fiber does. This property is not automatic from the first pairing condition and is load-bearing: without it Lemma 2.12 may fail to produce a conic bundle with both a (+,+) and a (−,−) fiber, and the identification R^{++}=R^{--} collapses. The likely proof via algebraic equivalence of the strict transforms of the general fibers on the blow-up ~X is not present. This gap must be closed before the d=2 base case, and hence Theorem 1.1(2) and Corollary 1.3, can be accepted.","section":"Theorems 4.8, 4.11, 4.12 and Corollary 4.12"},{"comment":"The proof of the main theorem depends critically on the unpublished preprint [Oka24a] for several central inputs: the classification of a-covers (used to prove Lemma 4.7 and to control accumulating components), movable bend-and-break for components of degree at least 3 (Theorem 4.8), and the identification of exactly which components are Manin (Corollary 4.12). The manuscript states these results as lemmas/theorems but does not reproduce their proofs or state precisely which parts of [Oka24a] are used. Since [Oka24a] is the author's own unpublished work, the reader cannot currently verify these load-bearing steps. The authors should either include the necessary statements and proofs, or make the dependency explicit and ensure that [Oka24a] is publicly and verifiably available. This is not a mathematical error in the present paper, but it is a serious conditionality issue.","section":"Theorem 4.9, proof of R_{+−} ≠ R_{++}"}],"minor_comments":[{"comment":"The sentence 'Lemma 3.6 shows that ~C^{++} and ~C^{--} are algebraically equivalent, but ~C^{++} and ~C^{+−} are not' is too terse. Lemma 3.6 concerns the numerical class ~ℓ of lines, not conic classes in 2~ℓ. Please spell out the algebraic-equivalence argument for the strict transforms of general conics, or give a direct reference.","section":"Theorem 1.1 and Theorem 4.11"},{"comment":"The phrase 'parametrized-sheeted covers' should be 'parametrize d-sheeted covers' in Theorem 1.1(2), Theorem 4.11, and the abstract. Also, in the proof of Theorem 4.9 there is a typo 'M 0.0(X, 2)' for 'M_{0,0}(X,2)'.","section":"Lemma 4.6"},{"comment":"In the proof of Lemma 4.6, the notation 'a smooth quadric Q ∈ L_{ij}' is potentially confusing because L_{ij} is a line in Bit(D_W), i.e., a pencil of quadrics. It would be clearer to write 'a smooth quadric Q in the pencil parametrized by L_{ij}'.","section":"Remark 4.4"},{"comment":"Remark 4.4 asserts that M_1^+ and M_1^- are non-normal, which is an interesting observation. It would help the reader if the proof were expanded by one or two sentences, since the non-normality is used implicitly later when discussing smooth points of the incidence loci.","section":"References"},{"comment":"The paper cites [Oka24a] as an arXiv preprint. If the present paper is to be published before [Oka24a] appears in a refereed venue, the dependency should be flagged in the introduction or in a footnote, and the relevant results should be stated precisely.","section":"Section 2.2"}],"recommendation":"major_revision","confidential_remarks":"The main concern is the unproved sign-map property in Theorem 4.9; this is load-bearing for the entire d >= 2 statement. The reliance on [Oka24a] is also substantial, and since it is the author's own unpublished preprint, the editor should verify whether it is already under review or publicly available in a stable form. The paper is otherwise well structured and the geometric arguments around Reye congruences are convincing. I recommend major revision rather than rejection, because the gap appears fixable within the manuscript's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper delivers what it promises: a complete description of the irreducible components of the moduli spaces of rational curves on general Artin-Mumford double solids, and the first example of Fano varieties where Geometric Manin's Conjecture is verified with two Manin components. The result is genuinely new and the main architecture is sound. The base case for lines via Reye congruences (Theorem 4.3) is clean and well exposed; that part is a real contribution in itself.\n\nThe soft spots are real but not fatal. The biggest is in Theorem 4.9. The equality R^{++}=R^{--} rests on the assertion that the sign map on the 56 lines of a general hyperplane section satisfies all assumptions of Lemma 2.12. The paper says 'one can see' this without proof. As the stress-test note points out, this is load-bearing: if the constant-parity condition fails for some conic bundle, the induction for all d≥2 loses its base. The property is likely true via algebraic equivalence—the two classes in the numerical class 2ℓ~ should force constant sign parity—but it is not demonstrated in the text. A referee should ask the author to prove it. This is a gap in exposition, not a reason to distrust the result.\n\nThe second issue is the heavy reliance on the author's unpublished preprint [Oka24a] for a-covers, movable bend-and-break, and even part of the final corollary. Corollary 4.12 says 'the argument in [Oka24a, Theorem 1.4] is valid' without giving details. That is a bit thin, but it is a common practice when building on one's own prior work; the referee will need to check [Oka24a] carefully.\n\nI agree with the reader's assessment: conditional acceptance is the right call. The paper deserves a serious referee, and the main result is significant enough that the refereeing process should fix the gap rather than reject the work. I would bring it to a reading group and cite it if I worked in this area.","headline":"First genuine multiple-Manin-component example for Geometric Manin's Conjecture, with a mostly solid proof, but one load-bearing 'one can see' in Theorem 4.9 needs to be written out.","tokens_in":19713,"tokens_out":1450,"would_cite":true,"duration_ms":15512,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14J30","14J28","14J45","14C25"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a general Artin-Mumford double solid, the moduli space of lines has exactly two components, and for every degree $d\\ge2$ the moduli space of rational curves has exactly four components, giving the first Fano case of Geometric Manin's…","keywords":["Artin-Mumford double solids","moduli of rational curves","Geometric Manin's Conjecture","Fano threefolds","Reye congruences","Enriques surfaces","del Pezzo threefolds","Brauer group"],"falsifier":"Take a general Artin-Mumford double solid, choose a general hyperplane section $S$, and determine which of the two components $M^+_1$, $M^-_1$ contains each of the 56 lines on $S$. Run through the 126 conic bundles on $S$ classified in Section 2.2: if any one has a singular fiber whose two components have opposite signs and another singular fiber whose two components have the same sign, the hypothesis of Lemma 2.12 fails, the equality $R^{++}=R^{--}$ has no basis, and the four-component description for all degrees collapses.","tokens_in":18678,"feed_emoji":"📐","tokens_out":15387,"duration_ms":137469,"temperature":0.7,"pith_summary":"Artin-Mumford double solids are threefolds realized as double covers of $\\mathbb{P}^3$ branched along a quartic symmetroid, the determinant surface of a symmetric matrix of linear forms; they are Fano threefolds with ten nodes. This paper proves that on a general such threefold $X$, the moduli space of lines has exactly two irreducible components, and for every degree $d \\ge 2$ the moduli space of rational curves of degree $d$ has exactly four irreducible components: two families of $d$-sheeted covers of lines and two families of embedded, very free curves. A corollary is that Geometric Manin's Conjecture holds for $X$, with exactly two Manin components in every sufficiently positive degree, matching the order-two unramified Brauer group. This is the first Fano example for which the conjecture holds with multiple Manin components rather than a unique one. The proof proceeds through the Reye congruence, an Enriques surface of bitangent lines to the branch quartic, and through conic bundles on degree-2 del Pezzo surfaces.","feed_headline":"Two line families spawn four curve families on Artin-Mumford solids","feed_subtitle":"First Fano example where Geometric Manin's Conjecture holds with multiple Manin components per degree.","key_machinery":"The load-bearing object is the pair of line components $M^+_1, M^-_1$, produced by the double-cover involution and controlled by the Reye congruence, an Enriques surface giving a birational normalization of the space of bitangent lines to the branch quartic. On a general hyperplane section $S$, the $56$ lines carry a sign according to which component contains them. Lemma 2.12 says that for any conic bundle on a degree-2 del Pezzo surface, if one singular fiber has components of opposite signs then every singular fiber does. Applied to $S$, this forces the loci $R^{++}$ and $R^{--}$ of unions of two same-sign lines to coincide as components, giving the component $R^+_2$; its conjugate $R^-_2$ comes from mixed-sign unions. The two-component distinction is carried by algebraic equivalence classes of $1$-cycles on the blow-up of $X$ at its ten nodes, where curves can be numerically equivalent but not algebraically equivalent; this separates $R^+_d$ from $R^-_d$. For $d \\ge 3$, the description is propagated by movable bend-and-break and the previously established framework for Fano threefolds with Gorenstein terminal singularities.","core_discovery":"The central discovery is a complete description of the irreducible components of $\\mathrm{Mor}(\\mathbb{P}^1,X,d)$ and $\\overline{M}_{0,0}(X,d)$ for a general Artin-Mumford double solid $X$. The space of $H$-lines consists of two components $M^+_1$ and $M^-_1$, each of expected dimension $5$, swapped by the covering involution. For each $d \\ge 2$, the space $\\mathrm{Mor}(\\mathbb{P}^1,X,d)$ has exactly four irreducible components $R^+_d$, $R^-_d$, $N^+_d$, $N^-_d$ of expected dimension $2d+3$: the $N^\\pm_d$ parametrize $d$-sheeted covers of the lines in $M^\\pm_1$, and the $R^\\pm_d$ generically parametrize embedded, very free curves. The Kontsevich spaces $\\overline{M}_{0,0}(X,d)$ have the same component structure with dimension lowered by $3$. The paper also proves that every component generically parametrizing birational stable maps contains unions of free curves using one line from each of the two line components. From this, Geometric Manin's Conjecture follows: for $d \\ge 2$ the Manin components are exactly $R^+_d$ and $R^-_d$, the line components and cover components being accumulating, and their number equals $|\\mathrm{Br}_{\\mathrm{nr}}(k(X)/k)| = 2$.","pith_inferences":["The unproved sign-map assertion could be settled by a finite check: since the 126 conic bundle types on a degree-2 del Pezzo surface are classified, one can directly test the sign property on the 56 lines of a general hyperplane section.","If the component structure deforms with the web $W$, the four-component description should hold for every excellent web and the two Manin components should form an irreducible family; the paper proves the general case, and extending to boundary webs is an inference.","The pairing of Manin components suggests that any rationally connected threefold with unramified Brauer group of order 2 will show paired components in each sufficiently positive curve class, one from each Brauer class; higher-dimensional Fano varieties with 2-torsion may exhibit the same doubling.","The geometric doubling in this paper may have an arithmetic counterpart: arithmetic Manin's conjecture would then predict two leading contributions to the height zeta function, a possibility the paper does not address."],"forward_implications":["Geometric Manin's Conjecture holds for general Artin-Mumford double solids: for each $d \\ge 2$ there are exactly two Manin components, and the count matches the order $2$ of the unramified Brauer group.","The moduli space of rational curves of every degree is fully known: the only components are the two cover families and the two embedded, very free families.","Every component of $\\overline{M}_{0,0}(X,d)$ that generically parametrizes birational maps contains reducible curves built from lines of both line components, so movable bend-and-break continues to hold inside every component even though the line space has two pieces.","This is the first Fano case with multiple Manin components per degree, confirming the conjecture's prediction that the number of Manin components equals $|\\mathrm{Br}_{\\mathrm{nr}}|$ in a non-unique setting."],"supporting_citations":[{"why":"Defines the Artin-Mumford double solids, proves they are unirational but not stably rational, and supplies the nontrivial Brauer group whose order is the predicted number of Manin components.","marker":"[AM72]"},{"why":"Gives the construction of the double cover branched along a symmetric determinantal quartic, which is the starting point of the paper.","marker":"[Bea16]"},{"why":"Introduces Reye congruences as Enriques surfaces, the geometry used to describe the two line components.","marker":"[Cos83]"},{"why":"Provides the modern treatment of Reye congruences and bitangent lines used in Theorem 4.2 and Lemma 4.5.","marker":"[DK24]"},{"why":"Supplies the framework for rational curves on Gorenstein terminal Fano threefolds: classification of a-covers, movable bend-and-break for higher degrees, and identification of non-Manin components.","marker":"[Oka24a]"},{"why":"Defines Manin components and Geometric Manin's Conjecture, and contributes deformation lemmas used in the induction step.","marker":"[LT19]"},{"why":"Supplies foundational dimension estimates, freeness criteria, and the smoothing results used to show general members are embedded and very free.","marker":"[Kol96]"},{"why":"Proves algebraic and homological equivalence of 1-cycles coincide on rationally connected threefolds, a step in identifying two algebraic classes per numerical class.","marker":"[BS83]"},{"why":"Proves the integral Hodge conjecture for 1-cycles on rationally connected threefolds, completing the identification of the Brauer group with two algebraic classes that distinguishes $R^+$ from $R^-$.","marker":"[Voi06]"}],"fun_headline_variants":["Four curve families per degree on Artin-Mumford solids","First Fano proof of Geometric Manin with multiple components","Artin-Mumford double solids: rational curve components found","Two line components lead to four curve components per degree"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on an unproved assertion: on a general hyperplane section, the assignment of each of the 56 lines to one of the two line components has the property that within any conic bundle, either every singular fiber has components of opposite types or none do; if that property fails, the proof that the two same-type conic components coincide has no basis.","fun_headline_variants_meta":{"raw":{"variants":["Four curve families per degree on Artin-Mumford solids","First Fano proof of Geometric Manin with multiple components","Artin-Mumford double solids: rational curve components found","Two line components lead to four curve components per degree"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000754,"raw_usage":{"total_tokens":3329,"prompt_tokens":899,"completion_tokens":2430,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":515,"completion_tokens_details":{"reasoning_tokens":2362}},"tokens_in":515,"tokens_out":2430,"duration_ms":17632,"temperature":1.0,"reasoning_tokens":2362,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:07:22.092760+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a general Artin-Mumford double solid, choose a general hyperplane section $S$, and determine which of the two components $M^+_1$, $M^-_1$ contains each of the 56 lines on $S$. Run through the 126 conic bundles on $S$ classified in Section 2.2: if any one has a singular fiber whose two components have opposite signs and another singular fiber whose two components have the same sign, the hypothesis of Lemma 2.12 fails, the equality $R^{++}=R^{--}$ has no basis, and the four-component description for all degrees collapses.","supporting_citations":[],"review_version":1}