{"id":"a57d7fe1-e29a-465b-ba03-e9f4efd3af99","arxiv_id":"1908.00750","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Finite fiberwise bimeromorphic group actions on holomorphic conic bundles without degree-one divisors are always subgroups of Z/2Z times Z/2Z.","lead":"The paper proves that any finite group of fiberwise bimeromorphic symmetries of a holomorphic conic bundle without a section-like divisor has order at most four, with every non-identity element of order two. This is the complex-analytic analog of a known algebraic theorem and it implies new Jordan-property statements for bimeromorphic transformation groups.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corollary 3.4 is false as stated; it is used in Lemma 3.6 and Corollaries 5.8–5.9, so those proofs are invalid, though Theorem 1.3 itself is unaffected.","rationale":"I read Theorem 1.3 as the central claim. Its proof uses Lemma 4.1, whose fixed-point divisor argument I checked in detail: the eigenvalue splitting is legitimate because the multiplier is a holomorphic function on each irreducible component taking values in the finite set {ζ, ζ^{-1}}, hence locally constant; Lemma 3.2 correctly ensures a faithful holomorphic action on a typical fiber. I found no flaw in Lemma 4.1 or in the deduction of Theorem 1.3. The reader's identified flaw in Corollary 3.4 is real: direct images of line bundles under proper flat maps are not automatically locally free, and the blow-up example over the disk gives a rank jump from 2 to 3. This affects the proof of Lemma 3.6 and hence Corollaries 5.8 and 5.9, and also the use of Corollary 3.4 inside Lemma 3.5, although Lemma 3.5 can likely be repaired via Grauert's theorem because in a P1-bundle the relevant line bundle has constant h^0. Since the main boundedness theorem is independent of the defective statement, the correct disposition remains the reader's CONDITIONAL verdict: accept once the auxiliary corollary is corrected or restricted and the affected applications are re-proved.","tokens_in":8801,"tokens_out":19433,"duration_ms":220210,"concrete_test":"Compute the direct image sheaf φ_*O_X(H−2C2) for X=Bl_p(P^1×Δ) over Δ. If the rank on the punctured disk is 2 and the fiber dimension over 0 is 3, then Corollary 3.4 is false and the proof of Lemma 3.6 as written fails. Then check whether a corrected proof of Lemma 3.6 can be supplied, for example by contracting the component on which D has negative degree; if it cannot, Corollaries 5.8 and 5.9 must be withdrawn.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader's objection is correct and is the main correctness issue. Corollary 3.4 asserts that for a proper holomorphic map with equidimensional fibers (hence flat by Theorem 3.3) and a vector bundle L, the direct image φ_*L is a vector bundle. This fails: flatness of φ does not make φ_*L flat; local freeness of direct images requires constancy of h^0(L|fiber), as in Grauert's theorem. For example, take Y=Δ, X=Bl_p(P^1×Δ) with p on the central fiber, write the central fiber as C1∪C2 with C1 the exceptional curve and C2 the proper transform of the central fiber, and take L=O_X(H−2C2), where H is a horizontal section meeting C2 once and missing C1. On a general fiber L has degree 1 and h^0=2, while on the central fiber the restrictions have degrees −2 on C1 and +3 on C2, giving h^0=3. Thus φ_*L is not locally free. This invalidates the proof of Lemma 3.6 exactly where it applies Corollary 3.4, and the same defective step appears in Lemma 3.5, although Lemma 3.5 can likely be rescued because in a P1-bundle the line bundle restricts to O(1) on every fiber, so h^0=2 is constant. Consequently Corollaries 5.8 and 5.9 are unproven as written. The main theorem, Theorem 1.3, and its supporting Lemma 4.1 do not use Corollary 3.4; the fixed-point divisor argument is independent and appears sound. So the central bound on G stands, but the paper needs repair of the auxiliary statements.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves an analog for compact complex manifolds of a theorem of Bandman–Zarhin on finite groups acting fiberwise on conic bundles without sections. The main result, Theorem 1.3, states that if φ:X→Y is a proper surjective holomorphic map with typical fiber P1 and there is no divisor on X meeting a typical fiber in one point, then any finite group acting by fiberwise bimeromorphic maps has all non-trivial elements of order 2 and is isomorphic to a subgroup of (Z/2Z)^2. The proof uses a fixed-point analysis of a finite-order fiberwise map on a typical fiber, which produces two divisors of degree one on the fiber, contradicting the no-divisor assumption. The paper also derives applications to strong Jordan property for bimeromorphic automorphism groups of certain conic bundles, especially over complex tori.","tokens_in":9125,"tokens_out":3045,"duration_ms":29530,"significance":"If the main theorem and its proofs are correct, the paper gives a natural complex-analytic analogue of a known algebraic result and establishes boundedness of finite fiberwise bimeromorphic group actions on such conic bundles, with a bound of order 4. The proof of Theorem 1.3 is self-contained, uses a standard fixed-point argument, and does not fit parameters or assume the conclusion; the argument appears sound. I also credit the author for explicitly identifying the key geometric mechanism (the two fixed points with reciprocal tangent eigenvalues) that converts forbidden symmetry into forbidden divisors. However, a significant flaw affects several auxiliary statements used in Section 5, so the paper cannot be accepted in its present form.","major_comments":[{"comment":"Corollary 3.4 is false as stated. The proof asserts that because φ is flat and L is flat, the direct image φ_*L is flat, citing [GPR94, Proposition 2.2.6(2)]. This implication is not valid for direct images: flatness of φ does not make φ_*L flat, and local freeness of φ_*L fails precisely when h^0(L|_fiber) jumps. For example, take Y=Δ, X=Bl_p(P1×Δ) with p on the central fiber, write the central fiber as C1∪C2 with C1 the exceptional curve and C2 the proper transform, and take L=O_X(H−2C2), where H is a horizontal section meeting C2 once and missing C1. On a general fiber L has degree 1 and h^0=2, while on the central fiber the degrees are −2 on C1 and +3 on C2, giving h^0=3, so φ_*L is not locally free. This correction is load-bearing because the corollary is cited in Lemma 3.5 and Lemma 3.6.","section":"§3, Corollary 3.4"},{"comment":"The proof of Lemma 3.5 relies directly on Corollary 3.4 to conclude that E=φ_*L is a vector bundle. Since Corollary 3.4 is false, the proof as written is invalid. In this lemma, however, the conclusion may be salvageable: because φ is a P1-bundle and L restricts to O(1) on every fiber, h^0(L|_F)=2 is constant, so Grauert's theorem would give local freeness. The author should replace the invocation of Corollary 3.4 with Grauert's theorem or another correct argument.","section":"§3, Lemma 3.5"},{"comment":"Lemma 3.6 is unproven as stated because it applies Corollary 3.4 in a genuinely non-constant situation: the fibers of φ are one-dimensional and only the typical fiber is P1, so h^0(L|_F) can jump on special fibers. The direct image φ_*L need not be a vector bundle; it may have skyscraper sheaves. The proof cannot be repaired merely by replacing one citation, because the claimed isomorphism of X with a projectivization up to modification needs a locally free direct image. Consequently the lemma as stated is not established.","section":"§3, Lemma 3.6"},{"comment":"Corollaries 5.8 and 5.9 are deduced from Lemma 3.6, which is not proved. These corollaries are therefore currently unsupported. The main theorem, Theorem 1.3, does not use Corollary 3.4; its proof via Lemma 4.1 is independent and appears sound. The status of the §5 applications, however, needs to be explicitly revisited after the issue with Lemma 3.6 is resolved.","section":"§5, Corollaries 5.8 and 5.9"}],"minor_comments":[{"comment":"The abstract contains a typographical artifact, 'ord ers', where a space splits the word 'orders'.","section":"Abstract"},{"comment":"The notation for the closed analytic set denoted 'Fix(g)' is inconsistent: in the statement an overline appears to have been lost, and the proof refers to Fix(g) as containing Fix(g) as a dense open subset. Please clarify the notation, e.g. use \\(\\overline{\\mathrm{Fix}(g)}\\).","section":"§3, Lemma 3.1"},{"comment":"The statement of Lemma 3.6 says 'the intersection number of D with a fiber of φ equals 1,' but the proof works with typical fibers; since a non-typical fiber may be singular or reducible, the intended hypothesis should be stated in terms of a typical fiber to match the proof and the usage in Corollaries 5.8–5.9.","section":"§3, Lemma 3.6"}],"recommendation":"major_revision","confidential_remarks":"The main theorem and its proof are good, but the false Corollary 3.4 affects several auxiliary statements used in Section 5. The author should repair or replace those arguments before publication; if only the main theorem is of interest, the paper can be significantly shortened. I recommend major revision rather than rejection because the central result appears sound and the issues are local and potentially fixable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: Theorem 1.3 is a genuine extension of Bandman–Zarhin’s boundedness theorem from quasi-projective conic bundles to compact complex manifolds, and the proof of that theorem is in good shape. The trouble is in the auxiliary material. Corollary 3.4 is false as stated, and the statements that rely on it need to be re-proved or restricted.\n\nWhat’s actually new: the paper replaces “no rational section” by “no divisor meeting a typical fiber in degree 1” and uses fixed-point divisors. Lemma 4.1 shows an order n > 2 fiberwise bimeromorphic map would create two such divisors, one for the tangent eigenvalue ζ and one for ζ^{-1}; then Theorem 1.3 follows from the fact that all nontrivial elements have order 2, so the group embeds in (Z/2)^2. The argument is short, self-contained, and persuasive. The Jordan-property corollaries are the natural payoff, and they are new in this analytic setting. I see no circularity: [BZ17] and [SV18] are used as prior context, not as inputs to the proof.\n\nThe soft spot is Corollary 3.4. It says that for a proper flat map with equidimensional fibers, φ_*L is a vector bundle. That is not what flatness buys you. The direct image can fail to be locally free when h^0 jumps on a special fiber; the standard counterexample is a deformation of P^1 with a degree-one line bundle degenerating to a reducible fiber. The stress-test example (blow-up of P^1×Δ at a point on the central fiber, twist by O(H−2C2)) is exactly this phenomenon. So Corollary 3.4 is wrong, and Lemma 3.6 uses it directly; Lemma 3.5 and Corollaries 5.8–5.9 inherit the problem. Lemma 3.5 is probably repairable because in a genuine P1-bundle a fiber-degree-1 line bundle has constant h^0=2 and h^1=0, so Grauert’s theorem gives local freeness; but as written the proof still cites the false statement. Corollaries 5.8 and 5.9 are unproven as written.\n\nThe main theorem does not use Corollary 3.4, and I could not find any other load-bearing flaw. The fixed-point divisor argument in Lemma 4.1 is independent and looks sound. Citation pattern is fine; self-citations are to standard facts, not to results that determine the conclusion. No fitted parameters or invented entities.\n\nBottom line: this paper deserves a serious referee. The right outcome is likely “major revision” or “conditional accept”: fix Corollary 3.4 (or replace it with the correct Grauert-type statement), repair Lemma 3.6 and the corollaries, and the core result stands. I would send it back for that revision.","headline":"Theorem 1.3 is a real extension and its proof is sound, but the paper currently has a false auxiliary lemma (Cor. 3.4) that takes down Lemma 3.6 and Corollaries 5.8–5.9 as written; the main theorem survives.","tokens_in":9682,"tokens_out":5939,"would_cite":true,"duration_ms":57622,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["32M05","14E07"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that if a holomorphic conic bundle admits no divisor meeting a typical fiber in one point, then every finite group acting on it by fiberwise bimeromorphic maps is trivial, cyclic of order 2, or the Klein four-group.","keywords":["conic bundle","bimeromorphic map","fiberwise action","finite group action","Jordan property","complex manifolds","projective line bundle"],"falsifier":"Construct or find a compact complex conic bundle with no divisor meeting a typical fiber once that admits a fiberwise bimeromorphic automorphism of order $3$; the theorem predicts the fixed locus of that map must split into two degree-one divisors, so inspecting the fixed locus of such a candidate decides the claim. A natural place to look is a projectivized indecomposable rank-$2$ vector bundle over a complex torus: finding a fiberwise order-$3$ automorphism there would refute the automorphism version, while proving none exist would support it.","tokens_in":8532,"feed_emoji":"📐","tokens_out":15327,"duration_ms":138125,"temperature":0.7,"pith_summary":"This paper establishes a boundedness theorem for finite symmetry groups of holomorphic conic bundles. If $\\varphi\\colon X\\to Y$ is a proper surjective holomorphic map whose typical fiber is $\\mathbb{P}^1$, and if no divisor on $X$ meets a typical fiber in exactly one point, then any finite group acting on $X$ by fiberwise bimeromorphic transformations is either trivial, $\\mathbb{Z}/2\\mathbb{Z}$, or $\\mathbb{Z}/2\\mathbb{Z}\\times\\mathbb{Z}/2\\mathbb{Z}$. This is the complex-analytic counterpart of a theorem for quasi-projective conic bundles without rational sections. The result matters because it converts a potentially wild group of bimeromorphic self-maps into one of three bounded possibilities whenever the fibration carries no degree-one divisor, and the fixed-locus argument behind it feeds directly into strong Jordan-property conclusions for automorphism groups of $\\mathbb{P}^1$-bundles over complex tori.","feed_headline":"Fiberwise finite actions on sectionless conic bundles cap at order 4","feed_subtitle":"Only trivial, order-2, and Klein-four groups can act fiberwise on a sectionless holomorphic conic bundle.","key_machinery":"The load-bearing mechanism is the eigenvalue splitting of the fixed locus in Lemma 4.1. For a finite-order fiberwise bimeromorphic map $g$ of order $n>2$, the restriction to a typical fiber $F\\cong\\mathbb{P}^1$ is a finite-order automorphism with exactly two fixed points; Lemma 3.8 gives a primitive $n$-th root of unity $\\zeta$ such that the tangent actions at those two points are multiplication by $\\zeta$ and $\\zeta^{-1}$. Since $n>2$, $\\zeta\\neq\\zeta^{-1}$, and this split makes the closure of the fixed locus break into two distinct divisor components $\\Sigma_\\zeta$ and $\\Sigma_{\\zeta^{-1}}$, each meeting a typical fiber in exactly one point. The existence of even one such component contradicts the no-degree-one-divisor hypothesis; in the automorphism setting, Lemmas 3.5 and 3.6 convert such components into a projectivization of a rank-$2$ vector bundle (decomposable when there are two disjoint sections). Lemma 3.2 is the bridge that lets bimeromorphic maps be treated as holomorphic on typical fibers.","core_discovery":"The central claim is Theorem 1.3: for a proper surjective holomorphic map $\\varphi\\colon X\\to Y$ of irreducible complex manifolds with typical fiber $\\mathbb{P}^1$, if there is no divisor on $X$ whose intersection with a typical fiber equals $1$, then every non-trivial element of a finite fiberwise bimeromorphic group $G$ has order $2$, and $G$ is $\\{1\\}$, $\\mathbb{Z}/2\\mathbb{Z}$, or $\\mathbb{Z}/2\\mathbb{Z}\\times\\mathbb{Z}/2\\mathbb{Z}$. The proof runs through the fixed locus of a single finite-order element $g$. On a typical fiber $F\\cong\\mathbb{P}^1$, the restriction of $g$ has exactly two fixed points, and the tangent action there is multiplication by $\\zeta$ and $\\zeta^{-1}$ for a primitive $n$-th root of unity. When $n>2$ these two eigenvalues are distinct, so the closure of the fixed locus splits into two irreducible codimension-one divisors, each meeting a typical fiber in exactly one point. The hypothesis forbids even one such divisor, hence $n$ cannot exceed $2$; with all elements involutions, a faithful action on a typical $\\mathbb{P}^1$ restricts $G$ to a subgroup of $\\mathbb{Z}/2\\mathbb{Z}\\times\\mathbb{Z}/2\\mathbb{Z}$. The companion Lemma 4.4 gives the same bound for holomorphic fiberwise automorphisms of a $\\mathbb{P}^1$-bundle that is not the projectivization of a decomposable rank-2 vector bundle.","pith_inferences":["The fixed-locus splitting suggests a sharper statement than the one emphasized: on a smooth conic bundle, existence of a finite-order fiberwise bimeromorphic map of order $>2$ is equivalent to existence of two distinct effective degree-one divisors, so the two-divisor version in Remark 4.2 is probably the natural hypothesis.","The same eigenvalue-splitting mechanism could be tried on fibrations whose typical fiber is a rational surface; there the fixed locus of a finite-order map would be a curve rather than two points, and the number of eigenvalues controlling the splitting would likely set a higher but still explicit order bound.","For the torus applications, the mechanism predicts that a projectivized indecomposable rank-$2$ vector bundle over a complex torus admits no fiberwise automorphism of order $>2$; this can be checked directly on explicit bundles, and finding a counterexample would pinpoint exactly where the Jordan-property conclusion fails."],"forward_implications":["On a holomorphic conic bundle with no degree-one divisor on typical fibers, finite fiberwise bimeromorphic group actions are bounded in the strongest sense: there are only three possible groups, so the group order is at most $4$.","For automorphisms of a $\\mathbb{P}^1$-bundle, the same order-$2$ dichotomy holds whenever $X$ is not a projectivization of a decomposable rank-$2$ vector bundle; consequently $\\operatorname{Aut}(X;\\varphi)$ is strongly Jordan, meaning every finite subgroup has a normal abelian subgroup of bounded index and bounded generator number, whenever $\\operatorname{Aut}(Y)$ is strongly Jordan.","For $\\mathbb{P}^1$-bundles over complex tori, $\\operatorname{Bim}(X)$ is strongly Jordan whenever $X$ is not a projectivization of a rank-$2$ vector bundle on the torus, because every bimeromorphic self-map of $X$ is automatically fiberwise over the torus.","Contrapositively, any finite-order fiberwise bimeromorphic map of order greater than $2$ produces two distinct degree-one divisors on $X$; checking for such divisors is a direct certificate that no such symmetry exists."],"supporting_citations":[{"why":"It proves the algebraic analog for conic bundles over fields containing all roots of unity, supplying the group dichotomy that the paper extends to compact complex manifolds.","marker":"[BZ17]"},{"why":"It supplies the complex-analytic foundations—flatness criteria, direct images of line bundles, indeterminacy loci, and meromorphic graph facts—used in Lemmas 3.2, 3.5, and 3.6.","marker":"[GPR94]"},{"why":"It gives the fact, used at the end of Lemma 4.1, that a non-trivial holomorphic automorphism of finite order acts non-trivially on the tangent space at a fixed point.","marker":"[Akh95]"},{"why":"It provides an alternative reference for the same tangent-space fact and is used in the Jordan-property applications.","marker":"[PS17]"}],"fun_headline_variants":["Fiberwise groups on sectionless conic bundles: only 1,2,4","Sectionless conic bundles restrict fiberwise symmetries to Klein four","No sections means fiberwise automorphisms are tiny on conic bundles","Conic bundles without sections: fiberwise finite groups capped","Bounded fiberwise transformations on sectionless holomorphic conic bundles"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the two fixed points of a finite-order fiberwise map on a typical $\\mathbb{P}^1$ fiber extend to two codimension-one fixed components in $X$, each meeting a typical fiber once, with the tangent scaling factors $\\zeta$ and $\\zeta^{-1}$ staying distinct along them.","fun_headline_variants_meta":{"raw":{"variants":["Fiberwise groups on sectionless conic bundles: only 1,2,4","Sectionless conic bundles restrict fiberwise symmetries to Klein four","No sections means fiberwise automorphisms are tiny on conic bundles","Conic bundles without sections: fiberwise finite groups capped","Bounded fiberwise transformations on sectionless holomorphic conic bundles"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000264,"raw_usage":{"total_tokens":1582,"prompt_tokens":901,"completion_tokens":681,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":586}},"tokens_in":517,"tokens_out":681,"duration_ms":6209,"temperature":1.0,"reasoning_tokens":586,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:36:00.207544+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct or find a compact complex conic bundle with no divisor meeting a typical fiber once that admits a fiberwise bimeromorphic automorphism of order $3$; the theorem predicts the fixed locus of that map must split into two degree-one divisors, so inspecting the fixed locus of such a candidate decides the claim. A natural place to look is a projectivized indecomposable rank-$2$ vector bundle over a complex torus: finding a fiberwise order-$3$ automorphism there would refute the automorphism version, while proving none exist would support it.","supporting_citations":[],"review_version":1}