{"id":"9eceb271-2762-424a-9c27-ce6925242ffe","arxiv_id":"1909.01909","paper_version":5,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"K3 surfaces with finite automorphism group and no elliptic fibrations are classified geometrically, with five of eight moduli spaces shown unirational.","lead":"This paper shows that K3 surfaces with finite automorphism group and no elliptic fibrations are exactly special quartics or double covers of the plane branched over sextics with prescribed tangencies. The result makes an abstract lattice classification visible and proves that five of the eight associated moduli spaces are unirational.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unirationality of S1 and S5 moduli rests on an unverified companion irreducibility theorem; the main geometric classification is not affected.","rationale":"The reader's weakest assumption identifies the same point I would flag: the S1 and S5 unirationality proofs rely on the author's separate irreducibility theorem in [26], which is not proved or reproduced in the manuscript. The strongest claim of the paper, the geometric description in Theorem 1, is well supported by lattice-theoretic derivations and explicit examples, so this concern should not overturn the classification. It does, however, make the unirationality part of the paper conditional on an external result that is easy to state but not independently verified here. Since the reader's verdict is already CONDITIONAL and the main classification appears sound, I do not recommend changing the verdict.","tokens_in":27185,"tokens_out":39633,"duration_ms":447790,"concrete_test":"Verify the irreducibility premise directly for the S1 and S5 period domains: for each lattice, write the moduli as an arithmetic quotient of a connected type-IV domain D(N^⊥); then check that the families q1q2q3−f3^2 (for S1) and l1l2q4−f3^2 (for S5) dominate the corresponding quotient by computing the Kodaira–Spencer map at the explicit Examples 22 and 35 and showing its image has dimension 17 and is not contained in a proper Noether–Lefschetz locus. If both checks pass, Corollaries 21 and 34 follow without invoking [26].","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 5's unirationality claims for S1 and S5 are not self-contained. Corollary 21 and Theorem 33 invoke the author's separate result [26] that the moduli spaces of K3 surfaces with finite automorphism group and Picard number at least 3 are irreducible. Without that premise, the dimension computations in Theorem 20 (respectively Theorem 33) only exhibit 17-dimensional subfamilies of the moduli of sextics with the required tangencies: the image could fill one component while another component of the sextic locus, or a component where the double cover has higher Picard number, is missed. This is a genuine load-bearing assumption for the unirationality statement, and it is imported from a self-cited companion paper rather than proved here. The central classification in Theorems 1 and 4 does not depend on this step; it is supported by direct lattice computations, Saint-Donat's theorem, discriminant-group arguments, and the explicit van Luijk examples. The concern is therefore addressable and does not indicate an error in the main geometric description.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the eight Néron–Severi lattices of K3 surfaces with finite automorphism group and no elliptic fibrations (ranks 3 and 4, types S1–S6, L(24), L(27)). For each type the author computes a distinguished big and nef divisor and uses Saint–Donat's theorems to produce either a quartic model in P3 or a double cover of P2 branched over a sextic with prescribed tangencies; Theorem 1 states this geometric description and the resulting automorphism group. Section 4 carries out the same analysis for the two rank-four lattices. Section 5 studies the converse: whether the geometric description forces the Néron–Severi lattice. Proposition 3 and Theorem 4 identify 'twins' for S3, S4, L(24), L(27) and characterize the other four types. The last section also provides explicit polynomial examples whose Picard numbers are checked by reduction modulo primes, and Theorem 5 asserts unirationality of the moduli spaces of S1, S2, S3, S5, S6.","tokens_in":27365,"tokens_out":18748,"duration_ms":177362,"significance":"The lattice computations are explicit and verifiable, and the geometric conclusions, when the classification results are accepted, give a concrete description of all K3 surfaces with finite automorphism group, no elliptic fibrations, and Picard number greater than 2. The paper honestly separates the four cases where the geometric description has a twin. The explicit examples with Weil-polynomial and discriminant checks are a strength. The unirationality claims are the least self-contained part: for S1 and S5 they import an irreducibility theorem from the author's companion paper [26], and for S6 the proof as written contains an inconsistency (cubic versus quartic) that should be corrected. These issues do not appear to affect the classification in Theorems 1 and 4.","major_comments":[{"comment":"The proof of unirationality of the S1 moduli space depends on the assertion from [26, Section 2] that the moduli spaces of K3 surfaces with finite automorphism group and Picard number at least 3 are irreducible. The dimension count in Theorem 20 shows that the family f6 = q1q2q3 − f3^2 is a 17-dimensional subvariety of the locus of sextics with three 6-tangent conics, but without irreducibility it need not dominate the 17-dimensional moduli of type S1 surfaces; it could fill one component while another component is missed. Since this premise is load-bearing for the S1 part of Theorem 5 and is imported from a self-cited companion paper rather than proved here, the unirationality statement as written is not self-contained.","section":"§5.1, Corollary 21; Theorem 5"},{"comment":"The same issue occurs for type S5. The proof of Theorem 33 uses the sentence 'The moduli space of K3 surfaces of type S5 is irreducible and 17 dimensional', again from [26]. The dimension computation only produces a 17-dimensional family of sextics of the form l1l2q4 − f3^2; without the irreducibility premise one can only conclude that the S5 moduli contains a unirational 17-dimensional subvariety. Please either supply a proof of the irreducibility, cite a published source, or state the weaker conclusion.","section":"§5.5, Theorem 33 and Corollary 34; Theorem 5"},{"comment":"In the proof of Proposition 37, the curve C1 is taken to be a rational normal quartic curve in a hyperplane H ⊂ P4, but Section 3.6 shows that in the model |A1 + A5| the curves A1 and A5 are cubic rational normal curves (degree 3); a degree-4 rational curve would have a different intersection matrix and not match (A1, A3, A5). Please correct 'quartic' to 'cubic' throughout this proof and check the dimension statement that follows. The subsequent sentence 'since that family contains the surfaces X...' should also be made precise: it needs to assert that every S6 surface appears in the family, otherwise dominance of the moduli space is not established. As written, the proof of unirationality of the S6 moduli space rests on an inconsistent choice of degree.","section":"§5.6, Proposition 37"}],"minor_comments":[{"comment":"In the paragraph after Corollary 11, 'L = 5A1 + 3A2 + A1' should be 'L = 5A1 + 3A2 + A3'.","section":"§3.2"},{"comment":"The two divisors displayed as D3 are in fact D3 and D4; rename the second one.","section":"§3.4"},{"comment":"In the generating series for ΘX, the term '3T23' sits between T44 and T50; the exponent is likely a typo, perhaps '3T46'.","section":"§3.1"},{"comment":"In the final paragraph of Section 4.1, 'degree 7 with respect to D' refers to a divisor D that has not been introduced; it should be L.","section":"§4.1"},{"comment":"In the proof of Proposition 38, the phrase 'which forces B1B4 = 0' is inconsistent with the displayed matrix M1, where B1B4 = 1; please correct the sentence.","section":"§5.7"},{"comment":"The ten tuples T1,...,T10 are listed in an order such that the later reference to 'cases T5,...,T10' is confusing; T5 belongs to the previous group, so the intended range is probably T6,...,T10.","section":"§5.8"}],"recommendation":"major_revision","confidential_remarks":"The author self-cites [26] for a load-bearing irreducibility result used in the unirationality claims for S1 and S5. The editors may wish to verify whether [26] is published or under review, and to consider whether Theorem 5 should be made conditional on that result if it is not available. The main geometric classification appears sound and is supported by explicit computations."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: Roulleau gives explicit geometric models for all eight K3 surfaces with finite automorphism group, Picard rank 3 or 4, and no elliptic fibrations: smooth quartics with special hyperplane sections (S2, S3) or double covers of the plane branched over sextics with prescribed tangencies (S1, S4, S5, S6, L(24), L(27)). That is genuinely new relative to the Nikulin–Vinberg lattice classification, and the converse/twin analysis is real: four types are characterized by their geometric description, four have a unique twin. The paper also proves unirationality for five moduli spaces. This is a useful paper and deserves a serious referee.\n\nWhat it does well: the core classification is backed by explicit lattice computations—intersection matrices, rank and discriminant arguments, Saint-Donat, Torelli—and those parts are checkable. The examples use the van Luijk–Elsenhans–Jahnel reduction method, with Weil polynomials given and Picard numbers concluded from discriminant comparisons. The generating series of big nef classes are a nice byproduct, and the paper is honest about which claims come from the companion paper [26].\n\nThe main soft spot is exactly the one flagged: unirationality of the S1 and S5 moduli spaces is not self-contained. In Theorem 20/Corollary 21 and Theorem 33, the dimension count shows the constructed family is 17-dimensional; to conclude it fills the whole moduli space you need the irreducibility result imported from the author's separate preprint [26]. That premise is load-bearing for Theorem 5, though not for Theorems 1 and 4. It is addressable—either by making [26] available or by proving irreducibility directly—so it is a gap in presentation, not a defect in the main classification. The stress-test note lands on this and does not overstate it.\n\nMinor issues: the Magma computations and Edgar Costa's zeta-function computations are not shipped, so the numerical examples are not independently reproducible from the paper alone. There are also a few small typos and notational slips, but nothing that blocks reading. The self-citation to [26] is not a problem per se; the problem is that a key premise lives there.\n\nWho is this for? Anyone working on K3 surfaces, automorphism groups, or Mori dream spaces. The geometric descriptions will be cited. Send it to a competent referee; the central claims are sound and the load-bearing dependency can be fixed in revision.","headline":"Solid geometric classification of the eight finite-automorphism K3 types without elliptic fibrations; the unirationality claims for S1 and S5 lean on a companion irreducibility theorem that is not proved here.","tokens_in":27873,"tokens_out":2817,"would_cite":true,"duration_ms":28446,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14J28"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that every projective K3 surface with finite automorphism group and no elliptic fibrations is either a smooth quartic in P3 or a double cover of the plane branched over a smooth sextic with prescribed tangencies.","keywords":["K3 surfaces","finite automorphism groups","Néron–Severi lattices","elliptic fibrations","double covers of the plane","sextic curves","unirational moduli spaces","(-2)-curves"],"falsifier":"Compute the dimension of the space of smooth plane sextics with three 6-tangent conics; the paper claims it is exactly 17, so a larger dimension would show the constructed families do not cover the full space. Alternatively, exhibit a smooth sextic with three 6-tangent conics whose double cover has Picard number 3 but whose pull-back curves have intersection numbers different from the unique rank-3 case in Proposition 19, which would contradict the claimed classification.","tokens_in":26947,"feed_emoji":"","tokens_out":15099,"duration_ms":119116,"temperature":0.7,"pith_summary":"Roulleau sets out to turn the lattice-theoretic classification of K3 surfaces with finite automorphism group and no elliptic fibrations into concrete projective geometry. He shows that every such surface, one of eight Néron–Severi types, appears either as a smooth quartic surface in $\\mathbb{P}^3$ with one of two special hyperplane-section configurations, or as a double cover of the plane branched over a smooth sextic curve that is tangent in prescribed ways to lines, conics, or cuspidal cubics. The tangency description is shown to characterize the surface exactly for four of the eight types, while the other four admit a unique 'twin' with the same equations but a different Néron–Severi lattice. As a by-product, five of the eight moduli spaces are shown to be unirational by explicitly constructing the corresponding families of sextics and quartics.","feed_headline":"K3s with no elliptic fibrations are quartics or sextic double covers","feed_subtitle":"All eight possible Néron–Severi lattices get explicit models as quartics or branched double covers.","key_machinery":"The load-bearing mechanism is the bounded fundamental domain $F_X$ of the Weyl group acting on the hyperbolic space of the Néron–Severi lattice. When $F_X$ is compact, the paper's algorithm enumerates all big nef classes of bounded square by projecting onto the negative-definite orthogonal complement of an ample class; the unique class $D$ with $D^2=2$ or $D^2=4$ then produces, through a classical theorem on linear series of K3 surfaces cited as [28], the double cover or the quartic embedding. The finitely many $(-2)$-curves are recovered by a standard lattice sieve, and their intersection pairing forces the tangency conditions on the branch sextic or on the special hyperplane sections. In the converse direction, the same lattice data show that a sextic with the required tangencies yields pull-back curves whose intersection matrix determines the Néron–Severi lattice uniquely, with a single exceptional matrix in four of the eight cases.","core_discovery":"The central theorem states that a K3 surface with Picard number greater than 2, finite automorphism group, and no elliptic fibrations has trivial automorphism group exactly when it is a smooth quartic in $\\mathbb{P}^3$ with either three quadric sections splitting into six rational quartic curves (type S2) or two hyperplane sections splitting into four conics (type S3). It has automorphism group $\\mathbb{Z}/2\\mathbb{Z}$ exactly when it is the double cover of $\\mathbb{P}^2$ branched over a smooth sextic $C_6$ with one of six tangency configurations: three 6-tangent conics (S1), a tritangent line plus a 6-tangent conic (S4), two tritangent lines (S5), a 6-tangent conic plus two cuspidal cubics (S6), three tritangent lines (L(24)), or one tritangent line plus three 6-tangent conics (L(27)). The converse study, stated as Theorem 4, shows that this geometric description is sufficient to determine the Néron–Severi lattice for types S1, S2, S5, S6, while types S3, S4, L(24), L(27) each admit exactly one twin surface with the same projective description but a different lattice. Theorem 5 records that the moduli spaces of types S1, S2, S3, S5 and S6 are unirational.","pith_inferences":["Because the twin surfaces share the tangency description but have different Néron–Severi lattices, the paper's discriminant computations suggest a practical test: reduce a candidate double cover modulo two primes, compare the square classes of the resulting discriminants, and the type (or its twin) will be determined.","The equation forms $f_6=q_1q_2q_3-f_3^2$ for type S1 and $f_6=l_1l_2q_4-f_3^2$ for type S5 indicate that the corresponding moduli spaces of sextics are rational images of spaces of quadrics, cubics, and quartics, which may lead to birational parametrizations of those moduli spaces rather than merely unirational ones.","The same bounded-fundamental-domain enumeration could be applied to the remaining K3 surfaces with finite automorphism group that do admit elliptic fibrations, producing explicit projective models for the lattices S112, S111, S113, S114 and L(25) as well."],"forward_implications":["Every K3 surface in the eight classified types admits an explicit projective model: a quartic in $\\mathbb{P}^3$ or a double cover of the plane branched over a sextic with prescribed tangencies.","For types S1, S2, S5 and S6, the tangency description alone fixes the Néron–Severi lattice, so a generic sextic or quartic with those properties yields exactly the intended surface.","The moduli spaces of types S1, S2, S3, S5 and S6 are unirational, since their points are built from parameter families of plane sextics or quartics.","The twin surfaces of types S3, S4, L(24) and L(27) all carry elliptic fibrations, so the absence of elliptic fibrations is not visible from the tangency configuration alone but is encoded in the richer Néron–Severi lattice.","The computed generating series $\\Theta_X$ of primitive big nef classes begins with the listed coefficients for each type, giving an enumerative fingerprint of the surface."],"supporting_citations":[{"why":"Supplies the six rank-3 Néron–Severi lattices S1–S6 with compact fundamental domains.","marker":"[21]"},{"why":"Supplies the two rank-4 lattices L(24) and L(27) with compact fundamental domains.","marker":"[30]"},{"why":"Provides the theorem on linear series used to turn the low-square nef class into a double cover or an embedding.","marker":"[28]"},{"why":"Gives the result that the automorphism group of a general low-rank K3 surface is trivial or generated by a non-symplectic involution.","marker":"[13]"},{"why":"Provides the two-prime reduction method for proving the Picard number of the constructed examples.","marker":"[15]"},{"why":"Refines the two-prime reduction method and supplies the discriminant-comparison technique used in the examples.","marker":"[8]"},{"why":"Proves irreducibility of the moduli spaces of K3 surfaces with finite automorphism group and Picard number at least 3, on which the unirationality arguments for S1 and S5 depend.","marker":"[26]"},{"why":"Contributes the finiteness and structure results for automorphism groups of hyperbolic lattices that underpin the classification.","marker":"[20]"}],"fun_headline_variants":["No elliptic fibrations: K3s are quartics or sextic double covers","Quartics or sextic double covers: K3s without elliptic fibrations","Finite automorphism K3s without elliptic fibrations: quartic or sextic cover","K3s with finite automorphism & no elliptic fibrations: quartic or sextic double cover"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The unirationality proofs for types S1 and S5 rely on the author's earlier result that the moduli spaces of K3 surfaces with finite automorphism group and at least three independent divisor classes are irreducible; without that irreducibility, the explicitly constructed 17-parameter families would cover only part of the space, and the unirationality statement would not follow for those two types.","fun_headline_variants_meta":{"raw":{"variants":["No elliptic fibrations: K3s are quartics or sextic double covers","Quartics or sextic double covers: K3s without elliptic fibrations","Finite automorphism K3s without elliptic fibrations: quartic or sextic cover","K3s with finite automorphism & no elliptic fibrations: quartic or sextic double cover"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002926,"raw_usage":{"total_tokens":11159,"prompt_tokens":1043,"completion_tokens":10116,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":659,"completion_tokens_details":{"reasoning_tokens":10023}},"tokens_in":659,"tokens_out":10116,"duration_ms":66376,"temperature":1.0,"reasoning_tokens":10023,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:05:18.531625+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the dimension of the space of smooth plane sextics with three 6-tangent conics; the paper claims it is exactly 17, so a larger dimension would show the constructed families do not cover the full space. Alternatively, exhibit a smooth sextic with three 6-tangent conics whose double cover has Picard number 3 but whose pull-back curves have intersection numbers different from the unique rank-3 case in Proposition 19, which would contradict the claimed classification.","supporting_citations":[{"cited_title":"Steklov Institute of Math","cited_arxiv_id":null,"evidence_quote":"Supplies the six rank-3 Néron–Severi lattices S1–S6 with compact fundamental domains."},{"cited_title":"Moscow Math","cited_arxiv_id":null,"evidence_quote":"Supplies the two rank-4 lattices L(24) and L(27) with compact fundamental domains."},{"cited_title":"Saint-Donat, Projective models of K3 surfaces, Amer","cited_arxiv_id":null,"evidence_quote":"Provides the theorem on linear series used to turn the low-square nef class into a double cover or an embedding."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the result that the automorphism group of a general low-rank K3 surface is trivial or generated by a non-symplectic involution."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the two-prime reduction method for proving the Picard number of the constructed examples."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Refines the two-prime reduction method and supplies the discriminant-comparison technique used in the examples."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves irreducibility of the moduli spaces of K3 surfaces with finite automorphism group and Picard number at least 3, on which the unirationality arguments for S1 and S5 depend."},{"cited_title":"Soviet Math., 22 (1983), 1401–1476","cited_arxiv_id":null,"evidence_quote":"Contributes the finiteness and structure results for automorphism groups of hyperbolic lattices that underpin the classification."}],"review_version":1}