{"id":"aa9f8699-c2f3-4180-a0bc-f82526b2de25","arxiv_id":"2507.07516","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The number of automorphism orbits of smooth rational curves on an Enriques surface equals a weighted sum of orbit counts of the ADE components of its Nikulin root invariant under the Vinberg group.","lead":"An Enriques surface is a special kind of geometric object whose rational curves can be grouped into symmetry classes. This paper gives a single algebraic formula that counts those classes for every Enriques surface, replacing many case-by-case computations.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 5.3's surjectivity claim and the untabulated order comparison are the load-bearing computational step; a Magma check of the 18 stabilizer orders would settle it.","rationale":"The reader's weakest_assumption identifies exactly the step I would stress: Lemma 5.3 and the accompanying order computations. I agree that this is the most load-bearing, least externally verified point. The rest of the proof is structurally coherent: Theorem 1.2 reduces the orbit count to maximal ADE configurations, Proposition 5.5 gives the required classification, and Section 7 carries the classification through all root-invariant types with a combination of geometric arguments and explicit finite computations. The individual case analyses are long, and some steps are left to the reader (notably Lemma 7.3 and several 'direct computations'), but those are localized and readily checkable, whereas Lemma 5.3 underpins the whole classification. The paper has independent support in the form of the 184 computational examples from [BS22] that motivated the conjecture, but that does not substitute for verifying the finite group orders. I found no internal inconsistency and no sign of data fitting or circular reasoning. The correct response is to keep the reader's CONDITIONAL verdict: the main theorem is plausible and the strategy is sound, but the computational core of Lemma 5.3 should be independently checked before full acceptance. Since the reader already made the verdict conditional on exactly this kind of computational verification, no adjustment to the verdict is needed.","tokens_in":37078,"tokens_out":10877,"duration_ms":126830,"concrete_test":"In Magma, realize E10 with an explicit Gram matrix (e.g. U ⊕ E8), and for each of the 18 irreducible root sublattice types R in Table 1 compute the finite group G_R = image of O(E10)_R in O(E10 ⊗ F2) and H_R = O(E10 ⊗ F2)_{R,{R'}}. Concretely: compute K = R⊥, take generators of the finite group O(K)^♯ extended by the identity on R, reduce each generator modulo 2, and compute H_R directly as the pointwise stabilizer of R and setwise stabilizer of R' in the finite orthogonal group of E10 ⊗ F2. Then check that G_R = H_R for every R except (A9,A9), where [H_R : G_R] = 2. This single computation directly confirms Lemma 5.3 and thereby the O(E10)(2)-classification in Proposition 5.5.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 5.5 is the pivot of the proof: it classifies maximal ADE configurations in E10 up to the 2-congruence subgroup O(E10)(2), and Section 7 uses that classification in every case via Proposition 6.6. Proposition 5.5 depends directly on Lemma 5.3, which asserts that the natural map O(E10)_R → O(E10 ⊗ F2)_{R,{R'}} is surjective for every irreducible negative-definite root sublattice R except (A9,A9), where the image has index 2. The proof of Lemma 5.3 verifies this claim only by comparing orders: the image order is derived from the table of #O(DK(2)) and #O(DK), while the codomain order is asserted to 'agree' by Lemma 5.1. That comparison is not tabulated, and the index-one assertion [O(DK(2)):G2] = 1 is delegated to Miranda-Morrison theory without a detailed computation. A single incorrect row in the table, or a misapplication of Lemma 5.1, would change the O(E10)(2)-orbit classification of maximal configurations and therefore alter the orbit counts in Theorem 1.2. The (A9,A9) exception is explicitly excluded from Proposition 5.5 and handled separately in Proposition 7.10, so the danger is concentrated in the non-exceptional rows: if any of those rows has index not equal to 1, the classification collapses. This is not a claimed error, but it is the least externally supported, most load-bearing computational assertion in the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves Theorem 1.2, a closed formula for the number of Aut(Y)-orbits of smooth rational curves on an Enriques surface Y in terms of the Nikulin root invariant and the Vinberg group GY. The strategy is to identify (-2)-curves with facets of the nef cone, to pass to minimal faces and maximal ADE configurations, to classify these configurations up to the 2-congruence subgroup O(E10)(2), and then to analyze the finitely many root-invariant cases. The formula is sum_{i=1}^9 a_i + sum_{i=1}^8 d_i + 2 d_9 + e_6 + 2 e_7 + 4 e_8 + a'_7 + 2 d'_8, where the a_i, d_i, e_i, a'_7, d'_8 count GY-orbits of connected components of specified types.","tokens_in":37379,"tokens_out":6530,"duration_ms":67325,"significance":"If correct, this is a substantial and satisfying structural result: the orbit count is completely determined by two classical invariants, and the formula is easy to apply in practice. The reduction from Aut(Y)-orbits to GY-orbits on maximal ADE configurations is elegant, and the case analysis gives a conceptual explanation of the exceptional multiplicities 2 and 4. The paper builds on a large body of prior computational work, and the formula is parameter-free and falsifiable against the 184 examples of [BS22] and the 527 elliptic fibrations of [BGA24]. However, several load-bearing computational and case-specific assertions are currently not documented enough for a reader to certify them independently.","major_comments":[{"comment":"The assertion that the image of O(E10)_R in O(E10 ⊗ F2)_{R,{R'}} has index one for all irreducible R except (A9,A9) is the key computational step on which Proposition 5.5 and therefore all of Section 7 depend. The proof computes only the orders #O(DK(2)) and #O(DK) in the table; it does not tabulate #G1, the order of the codomain O(E10 ⊗ F2)_{R,{R'}} from Lemma 5.1, or the index [O(DK(2)) : G2] that is asserted to be 1 by Miranda-Morrison theory. The equality of these orders is stated as 'they agree' without a comparison. Since a single incorrect row would alter the orbit classification and hence Theorem 1.2, please provide the missing comparison, at least as a table of #G1 and #O(E10 ⊗ F2)_{R,{R'}} for each R, or a verifiable computation (e.g. a Magma script) confirming the orders and the Miranda-Morrison index.","section":"Lemma 5.3 (Section 5)"},{"comment":"The statement of Lemma 6.7 says 'at most 5 (resp. 6)' for types (A7,Z/2Z) and (E7,0), while its proof immediately says 'There are 6 (resp. 5) such vectors'. The two numbers are swapped. Moreover Proposition 7.11 invokes the lemma to obtain at most 6 orbits for (A7,Z/2Z) and finds 6 vectors, and Proposition 7.16 invokes it to obtain 5 configurations for (E7,0). As written, the lemma contradicts both its proof and its later uses. This is load-bearing for the coefficients a'_7 and e_7 in Theorem 1.2, and the statement must be corrected.","section":"Lemma 6.7 (Section 6)"},{"comment":"The D9 case is treated in a very compressed way: the proof says 'Following [BS22, (7.4)], we compute' the nef chamber and then 'by looking at the various subgraphs of type eA8 and using Lemma 7.5(3-4) we find enough automorphisms to get at most 2 orbits'. It also says 'If it is generic, then it is a (D9,D9)-generic Enriques surface', but the connection between a component of type (D9,0) and this genericity notion is not established. Since the factor 2d_9 in Theorem 1.2 rests entirely on this paragraph, the case needs a complete argument or a precise citation to a computation whose input and output are described in enough detail to be checked.","section":"Proposition 7.14 (Section 7.2)"},{"comment":"The E8 case, which contributes 4e_8 to Theorem 1.2, relies on several unchecked assertions: 'We check that each orbit meets a configuration B0 in the diagram', 'on may check that each of them occurs in the diagram', and a determinant computation [SY : L] = 4 whose verification is omitted. The lower bound of four orbits also needs a complete argument: the use of the unique elliptic fibration and the fixed curves of the numerically trivial automorphism is only sketched. Please provide the explicit verifications for the five maximal types listed in Proposition 6.3(8), or a reproducible computational transcript for this case.","section":"Proposition 7.17 (Section 7.3)"}],"minor_comments":[{"comment":"The lemma states 'The proof is left to the reader'. Since Lemma 7.4 uses these integer solution sets to control the number of elements of ∆(Y)|B⊥, please include a short verification or at least the factorization of the relevant quadratic forms.","section":"Lemma 7.3 (Section 7.1)"},{"comment":"The proof begins 'As in Lemma 7.13', which is a self-reference; it should presumably be 'As in Lemma 7.12' or refer to the coordinate setup of an earlier lemma.","section":"Lemma 7.13 (Section 7.2)"},{"comment":"In the proof of item (10), the sentence 'By Proposition 6.3, K := R' ∩ R'⊥ ⊆ σ⊥' should refer to Lemma 6.2(3), not to the proposition being proved.","section":"Proposition 6.3 (Section 6)"},{"comment":"The phrase 'Since 2 ∤ det E8' is confusing: the relevant statement is that R′ is a regular subspace of E10 ⊗ F2 because E8 has odd determinant; please rephrase for clarity.","section":"Lemma 5.4 (Section 5)"}],"recommendation":"major_revision","confidential_remarks":"I am sympathetic to the paper and believe the main theorem is likely correct; the obstacles are verification and presentation. The swapped bounds in Lemma 6.7 and the undocumented computational assertions in Lemma 5.3, Proposition 7.14, and Proposition 7.17 are the main issues. If the authors supply the missing computations and fix the internal inconsistencies, I would be glad to support acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper delivers a genuine theorem: a closed formula for the number of Aut(Y)-orbits of (-2)-curves on any Enriques surface in characteristic not 2, in terms of the Nikulin root invariant and the Vinberg group. This was previously known only for special cases (1-nodal by Cossec-Dolgachev) or by computer enumeration of 184 examples. The formula is new and the proof strategy is sound: reduce orbit counting on curves to orbit counting on maximal ADE configurations, classify those configurations up to the level-2 congruence subgroup O(E10)(2), then handle the geometric distinctions (weights 2 and 4 for D9/E8 etc.) case by case. The conceptual reduction in Sections 4 and 6 is the real substance; the case analysis in Section 7 is long but mostly concrete. I believe the theorem.\n\nWhere are the soft spots? The load-bearing step is Lemma 5.3: surjectivity of O(E10)_R → O(E10⊗F2)_{R,{R'}} for all irreducible R except (A9,A9). The proof compares orders but never tabulates the comparison; it asserts the codomain order 'agrees' with the computed image order. That's a checkable computation—the table of #O(DK(2)) is there, and Lemma 5.1 gives the formula—so this is a gap in exposition, not an evident error. A referee should ask for the comparison table, or a Magma verification of the 18 stabilizer orders. Also 'direct computation' appears in a few places (Lemma 7.3 leaves its proof to the reader, and Proposition 7.10, Lemma 7.13 have coordinate checks); none of these look wrong, but they are exactly where a subtle error would hide. The reliance on [BGA24] for the semi-symplectic assumption and some fibration facts is fine—that's a preprint, but it's the authors' own prior work and the facts are stated explicitly.\n\nI don't share the deeper worry about circularity: the formula is derived from the root invariant and Vinberg group, not fitted to the orbit counts. The only place where the result is used to conjecture the formula is in the introduction, not the proof.\n\nWho is this for? Anyone working on Enriques surfaces, K3 automorphisms, or cone conjectures. It deserves a serious referee, and with minor revisions—mainly supplying the order-comparison data and a few of the 'one checks'—it will be a clean paper.","headline":"A new and likely correct closed formula for orbit counts of (-2)-curves on Enriques surfaces, with a proof whose main debt is a few checkable but untabulated order computations.","tokens_in":37942,"tokens_out":1993,"would_cite":true,"duration_ms":21860,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14J28","14J50"],"pacs":[],"model":"deepseek-v4-flash","headline":"For every Enriques surface, the number of automorphism-orbits of smooth rational curves is computed by a closed formula from the Nikulin root invariant and the Vinberg group.","keywords":["Enriques surfaces","smooth rational curves","automorphism orbits","Nikulin root invariant","Vinberg group","ADE root systems","nef cone","(-2)-curves"],"falsifier":"A concrete check is to take one explicit Enriques surface in each exceptional class, for example a $(D_9,0)$ surface with its 12 rational curves drawn, an $(E_8,0)$ surface with the 10-vertex dual graph displayed in Section 7.3, and an $(A_7,\\mathbb{Z}/2\\mathbb{Z})$ or $(D_8,\\mathbb{Z}/2\\mathbb{Z})$ surface, then compute the actual $\\mathrm{Aut}(Y)$-orbits of the visible $(-2)$-curves directly and compare the orbit count with the formula; a mismatch of the 2- or 4-fold multiplicity in any one of these graphs would refute the theorem.","tokens_in":36866,"feed_emoji":"🧮","tokens_out":8997,"duration_ms":100733,"temperature":0.7,"pith_summary":"The paper establishes a closed formula for the number of orbits of smooth rational curves on an Enriques surface under its automorphism group. It shows that this number is completely determined by two algebraic ingredients: the Nikulin root invariant, which records the connected components of the mod-2 reduction of all $-2$-curves together with the kernel of the reduction, and the Vinberg group, which encodes the surface symmetries modulo 2. The formula is a weighted sum of counts of these components by ADE type, with weights $2$ and $4$ appearing only for the exceptional types $(D_9,0)$, $(E_7,0)$, $(E_8,0)$, $(A_7,\\mathbb{Z}/2\\mathbb{Z})$, and $(D_8,\\mathbb{Z}/2\\mathbb{Z})$. Since orbits of rational curves coincide with orbits of facets of the nef cone, the result gives a complete, computable answer to the finiteness question for these orbits predicted by the cone conjecture for Enriques surfaces.","feed_headline":"A 9-term formula counts rational-curve orbits","feed_subtitle":"On Enriques surfaces, the count depends only on the Nikulin root invariant and the Vinberg group.","key_machinery":"The central working objects are the Nikulin root invariant and the Vinberg group $G_Y$, the image in $O(S_Y \\otimes \\mathbb{F}_2)$ of the modular stabilizer generated by the Weyl group and $\\mathrm{Aut}^*(Y)$. The set $\\Delta(Y)$ of mod-2 reductions of splitting $(-2)$-roots carries a graph whose connected components are ADE root systems; the Nikulin root invariant is the direct sum of the corresponding root lattices together with the kernel of the reduction map. The proof's load-bearing part is the classification of root sublattices of the $E_{10}$ lattice up to the action of the level-2 congruence subgroup $O(E_{10})(2)$: a key lemma shows that, except for the family $(A_9,A_9)$, the stabilizer of a root sublattice acts surjectively on its mod-2 stabilizer, so the classification of $G_Y$-orbits of maximal ADE configurations in $S_Y$ follows. Lemmas 4.5 and 4.6 connect these configurations to the facets of the nef cone, completing the bridge from the algebraic invariants to the geometric orbit count.","core_discovery":"The central claim is Theorem 1.2: for an Enriques surface $Y$, the number of $\\mathrm{Aut}(Y)$-orbits of $(-2)$-curves equals $$\\sum_{i=1}^{9} a_i + \\sum_{i=1}^{8} d_i + 2d_9 + e_6 + 2e_7 + 4e_8 + a'_7 + 2d'_8,$$ where $a_n$, $d_n$, $e_n$, $a'_7$, and $d'_8$ are the numbers of $G_Y$-orbits on connected components of $\\Delta(Y)$ of types $(A_n,0)$, $(D_n,0)$, $(E_n,0)$, $(A_7,\\mathbb{Z}/2\\mathbb{Z})$, and $(D_8,\\mathbb{Z}/2\\mathbb{Z})$. The proof reduces $\\mathrm{Aut}(Y)$-orbits of $(-2)$-curves to orbits of maximal ADE configurations of such curves, classifies these configurations up to the level-2 congruence subgroup of the $E_{10}$ lattice, and then analyzes the eight possible root-invariant types case by case. The exceptional weights in the formula are explained by concrete geometric phenomena: numerically trivial automorphisms, special elliptic fibrations, and finite automorphism groups.","pith_inferences":["The same reduction, from facets of a hyperbolic nef cone modulo automorphisms to a level-2 congruence classification of configurations, should yield analogous orbit-count formulas for other hyperbolic root lattices of rank 10 whenever the corresponding stabilizer-surjectivity lemmas hold.","The formula implies a rigidity statement: the orbit count is constant on the locus of Enriques surfaces with a fixed Nikulin root invariant and Vinberg group; this could be tested by deforming surfaces and checking that the count does not change.","A concrete next step would be to evaluate the right-hand side for all 184 root-invariant types and compare with the existing computer-aided tables, turning the theorem into a complete orbit-count table by root invariant alone.","The exceptional weights appear tied to fixed loci of numerically trivial automorphisms, so one might look for analogous weighted counts on other quotients of K3 surfaces or on Enriques surfaces in characteristic 2 if the cone conjecture holds there."],"forward_implications":["The number of $\\mathrm{Aut}(Y)$-orbits of smooth rational curves can be computed from the K3 cover's anti-invariant lattice and the Vinberg group, without enumerating curves by hand.","For any Enriques surface whose Nikulin root invariant and Vinberg group are known, the formula gives the exact orbit count; in particular the count is finite, recovering the finiteness that the cone conjecture predicts for these surfaces.","Corollary 1.4 bounds the orbit count by $\\rho-10$, where $\\rho$ is the Picard number of the K3 cover, so surfaces with small Picard number have few orbit classes.","The exceptional coefficients $2$ and $4$ have uniform geometric explanations coming from numerically trivial automorphisms, elliptic fibrations with specific fibers, and finite automorphism groups, so the exceptions are not arbitrary.","The formula unifies earlier partial results: generic nodal Enriques surfaces have a single orbit, and the computer-aided orbit counts for many explicit surfaces all satisfy the same weighted formula."],"supporting_citations":[{"why":"Supplies the classification of the 184 O(E10)-orbits of root sublattices by type, used to realize each connected component of Δ(Y) inside the numerical lattice.","marker":"[Shi21]"},{"why":"Establishes that the modular stabilizer is W(Y)⋊Aut*(Y), contains the level-2 congruence subgroup, and provides elliptic-fibration facts used in Section 7.","marker":"[BGA24]"},{"why":"Provides the computer-aided orbit counts on 184 Enriques surfaces that motivated and test the formula.","marker":"[BS22]"},{"why":"Underlies the discriminant-form and primitive-extension computations in Lemma 5.3 and the order computations.","marker":"[Nik79]"},{"why":"Basis for the surjectivity and index statement in Lemma 5.3 via embeddings of integral quadratic forms.","marker":"[MM09]"},{"why":"Describes numerically trivial automorphisms and their fixed loci, explaining the weights 2 and 4 for (E8,0), (E7,0), and (D8,Z/2Z).","marker":"[DK25]"},{"why":"Classifies zero-entropy Enriques surfaces and their unique positive-rank elliptic fibration, used in the E8 case.","marker":"[MMV24]"},{"why":"Gives the structure of Enriques surfaces with finite automorphism group of type II, used for the (D9,0) case.","marker":"[Kon86]"}],"fun_headline_variants":["9-term formula counts rational-curve orbits on Enriques","Enriques curve orbits counted by 9-term formula","Closed formula for Enriques rational-curve orbit count","9-term count: Enriques surface rational-curve orbits","Rational-curve orbits on Enriques now a 9-term count"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is a group-theoretic computation: a certain group of symmetries of the ten-dimensional lattice that fix a root pattern must surject onto the corresponding group fixing the pattern's binary reduction, except for one pattern where it is exactly half the group; if this computation is wrong, the list of possible configurations is incomplete and the final count changes.","fun_headline_variants_meta":{"raw":{"variants":["9-term formula counts rational-curve orbits on Enriques","Enriques curve orbits counted by 9-term formula","Closed formula for Enriques rational-curve orbit count","9-term count: Enriques surface rational-curve orbits","Rational-curve orbits on Enriques now a 9-term count"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00057,"raw_usage":{"total_tokens":2637,"prompt_tokens":826,"completion_tokens":1811,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":442,"completion_tokens_details":{"reasoning_tokens":1730}},"tokens_in":442,"tokens_out":1811,"duration_ms":15445,"temperature":1.0,"reasoning_tokens":1730,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T18:39:09.387460+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete check is to take one explicit Enriques surface in each exceptional class, for example a $(D_9,0)$ surface with its 12 rational curves drawn, an $(E_8,0)$ surface with the 10-vertex dual graph displayed in Section 7.3, and an $(A_7,\\mathbb{Z}/2\\mathbb{Z})$ or $(D_8,\\mathbb{Z}/2\\mathbb{Z})$ surface, then compute the actual $\\mathrm{Aut}(Y)$-orbits of the visible $(-2)$-curves directly and compare the orbit count with the formula; a mismatch of the 2- or 4-fold multiplicity in any one of these graphs would refute the theorem.","supporting_citations":[],"review_version":1}