{"id":"1104af32-7437-46ea-997c-9ba8bff9dea3","arxiv_id":"1908.08087","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Under a Hermitian flatness condition on the direct image of the relative log-canonical bundle, a Kähler fibration with klt log Calabi-Yau fibers is locally trivial; a K3 example shows the relative Ricci-flat metric need not be semipositive.","lead":"This paper proves that a flatly varying family of log Calabi-Yau fibers must be locally trivial, and disproves a folklore conjecture by constructing a one-parameter family of elliptic curves whose relative Ricci-flat metric is not semipositive. The results connect the curvature of relative Kähler-Einstein metrics to the birational geometry of the fibration.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.2 rests on the delta-uniform estimates (1.32)-(1.34) quoted from [Gue20] and [GP16]; if these fail uniformly as delta->0, the limiting horizontal lift need not be holomorphic and the local trivialization collapses.","rationale":"The reader's weakest assumption identifies exactly the un-reproduced uniform estimates (1.32)-(1.34) as the load-bearing premise for Theorem 1.2. My reading confirms this: the proof of local triviality hinges on the delta->0 limit of the horizontal lifts v_delta being a holomorphic vector field, and that limit is obtained solely through those quoted estimates. The paper even says 'we will not reproduce here the arguments' for (1.32)-(1.34), so a referee must verify them in [Gue20] and [GP16]. There is an additional subtlety: the delta-family tau_delta is not the same as the epsilon-family treated in [Gue20], so the claim that the estimates 'go through' requires checking the dependence of constants and the adapted norms; also (1.32) must be read off Supp(B), since near the divisor the conic potential has unbounded higher derivatives. I found no internal logical contradiction or counterexample to the theorem; the counterexample in Theorem D appears sound, and the descending argument in Section 3 is consistent. Therefore the conditional verdict is appropriate: accept only after the cited analytic estimates are confirmed. My concern does not shift the verdict; it sharpens what needs checking.","tokens_in":38741,"tokens_out":17542,"duration_ms":190946,"concrete_test":"Check [Gue20, (3.13), Prop. 4.1&4.2] and [GP16, §5.2] line-by-line and verify that the estimates (1.32)-(1.34) follow for the u_delta defined by (1.22) with constants independent of delta, specifically: (i) the C^k bound holds on compact subsets of X_t \\ Supp(B) with C_k independent of delta and t; (ii) the L^2 bound (1.33) is genuinely with respect to tau_delta^n, not a model conic measure with different weights; (iii) the mass vanishing (1.34) is uniform in t. If any cited statement requires an extra hypothesis, such as absence of basepoints or a lower bound on delta relative to distance to B, then the limiting argument for the holomorphic horizontal lift is incomplete and Theorem 1.2 needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The local triviality proof in Sections 1.4.2-1.4.3 depends on three imported estimates: (1.32) uniform C^k bounds for d_t u_delta off Supp(B), (1.33) a uniform L^2 bound for |v_delta|^2 against tau_delta^n, and (1.34) vanishing of the weighted L^2 mass of v_delta near B as delta->0. The paper states 'we will not reproduce here the arguments for (1.32)-(1.34)' and only sketches (1.33). This is load-bearing: Proposition 1.8 uses (1.32) and (1.33)-(1.34) to extract a holomorphic limit w of v_delta|X_t; Corollary 1.13 then identifies w with the horizontal lift of rho. If any of these estimates is not uniform in delta, the limit could have nonzero dbar singularities or fail to be L^2, so the flow of v would not identify nearby fibers, breaking Theorem 1.2. A subtlety is that [Gue20] proves estimates for a different family (the epsilon-twisted conic KE metrics solving (0.2)), not the delta-regularized volume element tau_delta defined by (1.22); the assertion that those estimates 'go through' for u_delta is a non-obvious analytic transfer. Moreover, (1.32) as printed on a coordinate set meeting Supp(B) cannot hold for all k without weights, since conic potentials have derivatives blowing up near B; the proof only uses it away from B, but this discrepancy must be reconciled with the cited statements. No internal contradiction was found in the rest of the argument; the central claim is plausible conditional on these estimates.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the variation of relative Ricci-flat Kähler metrics for families of log Calabi-Yau pairs. In Section 1, assuming c1(K_{X_y}+B_y)=0 and Hermitian flatness of the Narasimhan-Simha metric on p_*(m(K_{X/Y}+B)), Theorem 1.2 proves local triviality of the fibration over the regular locus, while Theorem 1.1 establishes positivity and canonical extension of the relative Ricci-flat current. Corollary 1.3 (a Kähler version of an Ambro-type statement) and Corollary B (bigness of the direct image under injective Kodaira-Spencer maps) are derived. In Section 2, for fibers with Kodaira dimension zero, Theorem C gives transverse regularity and Lipschitz variation of the relative Kähler-Einstein potentials away from Supp(B+E), based on new weak Sobolev and Poincaré inequalities. Section 3 and the appendix by Tosatti disprove a folklore conjecture by constructing an elliptic fibration of a K3 surface whose relative Ricci-flat metric is not semipositive (Theorem D). The counterexample part is self-contained and uses an explicit K3 surface with two elliptic fibrations.","tokens_in":39116,"tokens_out":7639,"duration_ms":75986,"significance":"If the proofs are completed, Theorem 1.2 is a substantial local triviality criterion for families of log Calabi-Yau pairs, giving a differential-geometric route to results in the direction of Viehweg's C_{n,m} conjecture. Theorem D settles a folklore conjecture negatively with an explicit, verifiable example, and the appendix by Tosatti is a clean and independent construction. The paper is commendably explicit about its limitations: it acknowledges the weak form of the Sobolev and Poincaré inequalities in Section 2 and it openly states that the crucial δ-uniform estimates (1.32)-(1.34) are quoted from [GP16] and [Gue20] rather than proved. The overall logical structure is clear and no circularity is evident. However, the central theorem currently rests on an unverified analytic transfer, so the significance is conditional on closing that gap.","major_comments":[{"comment":"The proof of Theorem 1.2 depends in an essential way on the three δ-uniform estimates (1.32)-(1.34), which are quoted from [Gue20] and [GP16] with the explicit statement 'we will not reproduce here the arguments for (1.32)–(1.34)'. As the stress-test note correctly observes, the transfer is not automatic: [Gue20] proves estimates for the twisted conic Kähler-Einstein metrics ρ_ε satisfying (0.2), whereas τ_δ is defined by the regularized Monge-Ampère equation (1.22). These are different equations, with different normalizations, and the sentence 'The estimates [Gue20, (3.13), Prop. 4.1&4.2] go through for u_δ' needs justification. Without uniformity in δ, the limit w extracted in Proposition 1.8 could have non-zero ∂̄ w, and the flow argument identifying nearby fibers in Corollary 1.13 and the end of Theorem 1.2 would collapse. Please either provide a proof of (1.32)-(1.34) for u_δ, or give a precise statement in [Gue20] or [GP16] that literally applies to this family, and explain how each estimate is used in the limiting argument.","section":"Section 1.4.2, equations (1.32)-(1.34)"},{"comment":"As printed, (1.32) asserts sup_{t∈Δ} ||∂_t u_δ||_{C^k(Ω∩X_t)} ≤ C_k for any coordinate set Ω, with no restriction on Ω. If Ω meets Supp(B), this cannot hold uniformly in δ for general k: conic potentials have derivatives that blow up near B as δ→0, unless the norm is measured with weights or the set Ω is taken away from Supp(B). The later application in the proof of Proposition 1.8 only needs C^k bounds on compact subsets of X_t \\ Supp(B), so the intended statement is recoverable, but the displayed inequality as written is misleading and should be reformulated with the correct quantification over Ω (or with a conic norm) and reconciled with the cited estimates from [Gue20].","section":"Section 1.4.2, display (1.32)"},{"comment":"The uniform bound |∫_{X_t} c(τ_δ) τ_δ^n| ≤ C is stated as 'a by-product of the considerations in the article [Gue20, (5.3) & Prop. 5.4]' and only sketched via equation (1.49), where the identification of V_δ(V_δ(u_δ)) with c(τ_δ) is described as 'the same' up to controlled terms. This bound is load-bearing: Corollary 1.13 uses it, together with Proposition 1.11, to obtain local uniform boundedness of c(τ_δ), then C^k bounds in the fiber directions, and finally C^{1,α} convergence of u_δ. Since Proposition 1.12 is not proved in the text and the cited statement in [Gue20] concerns a different family, the proof of Corollary 1.13 currently has a gap. Please supply a complete proof or a precise citation that applies directly to τ_δ.","section":"Proposition 1.12 (page 16)"}],"minor_comments":[{"comment":"The word 'folkore' appears twice in the abstract and introduction; it should be 'folklore'.","section":"Abstract and Introduction"},{"comment":"Equation numbering in Section 1.4.2 is confusingly duplicated: (1.31)-(1.34) are first used for the quasi-isometry estimate (a) and the [Gue20] estimates (b), and then reused inside the proof of Proposition 1.8 for the geodesic curvature equation, its curvature term, the integrated inequality, and the limit (1.34). This makes cross-referencing the argument unnecessarily hard; please renumber or use distinct labels.","section":"Section 1.4"},{"comment":"The notation 'X_y := p^{-1}(X_y)' should presumably read 'X_y := p^{-1}(y)'.","section":"Page 5, line after (1.1)"},{"comment":"The phrase 'extends canonically' is explained only after the statement of Theorem 1.1 as local potentials locally bounded above on X \\ X^\\circ. It would help to include this clarification directly in Theorem 1.1 or in the definition of ρ, since the introduction's version 'extends canonically to a closed positive current' is stronger-sounding than what is proved.","section":"Theorem 1.1 and Theorem A"}],"recommendation":"major_revision","confidential_remarks":"The main concern is the unproved transfer of the δ-uniform estimates (1.32)-(1.34) from [Gue20] and [GP16] to the family τ_δ defined by (1.22). This is not a matter of taste or presentation: without those estimates the proof of Proposition 1.8 and hence Theorem 1.2 does not close. The counterexample part, including the appendix, is convincing and self-contained. I recommend major revision rather than rejection because the central claim is plausible and the gap is local and potentially fixable; however, the authors should be pressed to either prove the estimates or identify a precise published statement that covers τ_δ. The duplicated equation numbers and the minor typographical issues are easy to fix during revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this is a serious research preprint with three genuine results: a local triviality theorem under Hermitian flatness of the direct image (Theorem A/1.2), a Lipschitz regularity theorem for families with Kodaira dimension zero (Theorem C), and a counterexample to the folklore semipositivity conjecture (Theorem D). The counterexample is the cleanest part: a non-isotrivial elliptic K3 with a second elliptic fibration, so the relative Ricci-flat metric cannot be semipositive. The appendix by Tosatti gives the required K3 and the argument is self-contained.\n\nWhat is actually new: the flatness-to-triviality theorem goes beyond earlier algebraic results, and the counterexample refutes a conjecture that was floating around since Schumacher. The paper is honest about its limitations, including weak Sobolev/Poincaré inequalities and the reliance on prior estimates.\n\nThe soft spot is in the proof of Theorem 1.2. The crucial limit of horizontal lifts depends on estimates (1.32)-(1.34), quoted from [GP16] and [Gue20] and not reproduced. The paper says this explicitly, which is fine, but the transfer is not a formality: [Gue20] treats ε-twisted conic Kähler-Einstein metrics, while the δ-regularized volume element in (1.22) is a different approximation. A referee needs to verify that the quoted bounds are uniform enough to force the limiting vector field to be holomorphic. If they are not, the flow argument collapses. Also, (1.32) as printed cannot literally hold for all k on sets meeting Supp(B) because conic potentials have singular derivatives; the proof uses it only away from B, so it is probably a harmless wording issue, but it should be reconciled.\n\nI did not find circularity or fitted parameters. The counterexample is independent of the main theorems. The analytical parts are long and I could not machine-check them, but the structure is clear.\n\nWho should read it: anyone working on variation of Kähler-Einstein metrics, direct images, or hyperbolicity of bases. It deserves a serious referee, with instructions to focus on the transfer of the quoted estimates. I would not desk reject.","headline":"Solid, original paper; Theorem 1.2's local triviality is conditional on quoted estimates that a referee should verify, but the K3 counterexample is convincing.","tokens_in":39641,"tokens_out":2357,"would_cite":true,"duration_ms":24212,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14J10","14J32","32Q20"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that flatness of the Narasimhan-Simha metric on the direct image of the relative log pluricanonical bundle forces a Kähler fibration of log Calabi-Yau manifolds to be locally trivial, while also giving a regularity…","keywords":["Kähler fiber space","log Calabi-Yau manifolds","conic Kähler metrics","relative Ricci-flat metrics","Narasimhan-Simha metric","local triviality","Kodaira dimension zero","singular Monge-Ampère equations"],"falsifier":"Exhibit a proper Kähler fibration of log Calabi-Yau fibers satisfying the flatness hypothesis on $p_*(m(K_{X/Y}+B))$ whose fibers are not locally isomorphic; Theorem 1.2 predicts such an example does not exist, so even one would refute the main theorem.","tokens_in":38561,"feed_emoji":"📐","tokens_out":11751,"duration_ms":106929,"temperature":0.7,"pith_summary":"This paper studies how the metric geometry of the fibers of a holomorphic family of log Calabi-Yau manifolds controls the global structure of the family. Its main theorem says that if the direct image $p_*(m(K_{X/Y}+B))$ is Hermitian flat with respect to the Narasimhan-Simha metric on the regular part of the base, then the fibration is locally trivial: every point of the smooth locus has a neighborhood over which the family is a product of a fiber with the base. The proof shows that the relative Ricci-flat conic Kähler current is semipositive and extends as a positive current, and that it admits holomorphic horizontal vector fields whose flow identifies nearby fibers while preserving the boundary divisor. A further theorem proves Lipschitz regularity of relative Kähler-Einstein potentials when the fibers have Kodaira dimension zero, and a final example shows that the relative Ricci-flat metric need not be semipositive on a Calabi-Yau family.","feed_headline":"Flat direct-image metric forces Calabi-Yau fibrations locally trivial","feed_subtitle":"For Calabi-Yau families, a flat natural metric on the direct image forces nearby fibers to be isomorphic.","key_machinery":"The main objects are the Narasimhan-Simha metric on $F_m=p_*(m(K_{X/Y}+B))$, defined by fiber integrals $\\|\\sigma\\|^2=V^{m-1}\\int_{X_y}|\\sigma|^2|\\Omega|^{-2(m-1)/m}e^{-\\varphi_B}$, and the horizontal lift $v_\\rho$ of a base vector field with respect to the relative Kähler metric. The proof approximates the singular conic Ricci-flat metric by smooth metrics $\\tau_\\delta$, whose geodesic curvature $c(\\tau_\\delta)$ satisfies $-\\Delta_{\\tau_\\delta}c(\\tau_\\delta)=|\\bar\\partial v_\\delta|^2-\\Theta(K_{X/\\Delta})(v_\\delta,\\bar v_\\delta)$. Uniform estimates imported from earlier work control $v_\\delta$ and its $\\bar\\partial$, so as $\\delta\\to 0$ the limiting vector field $v_\\rho$ is holomorphic; the identity $L_{v_\\rho}\\rho=0$ then shows its flow preserves $\\rho$ and $B$, giving the local product decomposition. For the Kodaira dimension zero case, the key technical device is a weak Sobolev and Poincaré inequality adapted to volume forms with zeros, which controls base-direction derivatives of the relative potentials.","core_discovery":"The central discovery is rigidity from flatness. For a proper holomorphic map $p:(X,B)\\to Y$ between Kähler manifolds whose fibers $(X_y,B_y)$ are klt pairs (mildly singular log pairs) with $c_1(K_{X_y}+B_y)=0$, the vanishing of the curvature of $F_m=p_*(m(K_{X/Y}+B))$ with respect to the Narasimhan-Simha metric forces $p$ to be locally trivial over the regular locus, and under mild extra assumptions over all of $Y$. The mechanism is that the relative Ricci-flat conic Kähler metric admits horizontal lifts of base vector fields that are holomorphic and preserve the boundary, so their flows give explicit isomorphisms between nearby fibers. In the Kodaira dimension zero case, where basepoints may appear, the paper constructs a relative Kähler-Einstein current whose fiberwise Ricci curvature is prescribed by the divisors $-[E_y]+[B_y]$, with Lipschitz potentials away from the support. The paper also constructs a one-dimensional family of elliptic curves whose relative Ricci-flat metric is not semipositive, disproving a folklore conjecture.","pith_inferences":["Editorial extension: the same horizontal-lift mechanism suggests that the kernel of the Narasimhan-Simha curvature foliates the base by locally trivial subfamilies; a curvature decay estimate could replace exact flatness and give a graded rigidity statement.","Editorial extension: Theorem C's Lipschitz regularity is likely a step toward a degeneration theory for relative Kähler-Einstein metrics with basepoints, and could be used to study the behavior of the metric at the base locus $E$.","Editorial extension: the K3 counterexample indicates semipositivity of relative Ricci-flat metrics fails precisely when the fibration has variation; testing other non-isotrivial elliptic K3s with two transverse fibrations would show whether this is a general phenomenon."],"forward_implications":["If the Narasimhan-Simha curvature vanishes, the family $(X,B)$ is locally trivial over the regular locus $Y^\\circ$; in particular all nearby fibers are isomorphic as log pairs.","Generic injectivity of the logarithmic Kodaira-Spencer map forces $F_m=p_*(m(K_{X/Y}+B))^{**}$ to be big, giving a logarithmic analogue of Viehweg's $Q_{n,m}$ conjecture.","For compact Kähler $X$, if $-(K_X+B)$ is nef then $-K_Y$ is pseudo-effective; if additionally $c_1(K_X+B)=0$ and $c_1(Y)=0$, then $p$ is locally trivial, including the Albanese map.","When $\\kappa(K_{X_y}+B_y)=0$, the relative Kähler-Einstein current exists with $\\operatorname{Ric}\\theta_y=-[E_y]+[B_y]$ and its potentials are Lipschitz away from $\\operatorname{Supp}(B+E)$.","The relative Ricci-flat metric on a Calabi-Yau fibration need not be semipositive: a non-isotrivial elliptic K3 fibration provides an explicit counterexample."],"supporting_citations":[{"why":"Supplies the psh variation and uniform derivative estimates for families of conic Kähler-Einstein metrics used to prove semipositivity of the relative current and to control horizontal lifts.","marker":"[Gue20]"},{"why":"Provides the uniform regularized conic estimates and C2 a priori bounds for singular Monge-Ampère equations that underpin the approximation argument with $\\tau_\\delta$.","marker":"[GP16]"},{"why":"Contributes the geodesic curvature identity and the semipositivity technique for canonically polarized families that the proof adapts; it is also the source of the folklore conjecture disproved in the paper.","marker":"[Sch12]"},{"why":"Supplies the family version of Kołodziej's L∞ bounds used for the potentials $u_\\delta$ and $\\phi$.","marker":"[DDG+14]"},{"why":"Kołodziej's stability theorem gives the convergence $u_\\delta\\to\\phi$ used to identify the limiting current.","marker":"[Ko05]"},{"why":"Provides the argument that a vector field preserving the metric in the appropriate sense is holomorphic, used to prove $v_\\rho$ is holomorphic.","marker":"[Ber09]"},{"why":"Sets the framework of fiberwise singular Kähler-Einstein metrics that this paper continues.","marker":"[CGP17]"}],"fun_headline_variants":["Flat direct-image metric forces Calabi-Yau fibration to be locally trivial","Counterexample: relative Ricci-flat metric on elliptic curves not semipositive","Rigidity from flatness: Calabi-Yau families with flat direct image are trivial","Semipositivity conjecture for relative Ricci-flat metrics disproved","Flat Narasimhan-Simha metric implies isomorphic fibers in Calabi-Yau families"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of local triviality rests on uniform estimates, quoted from two earlier papers, for the approximated horizontal vector fields as the smoothing parameter tends to zero; if those bounds were not uniform, the limiting vector field could fail to be holomorphic and the flow trivialisation would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Flat direct-image metric forces Calabi-Yau fibration to be locally trivial","Counterexample: relative Ricci-flat metric on elliptic curves not semipositive","Rigidity from flatness: Calabi-Yau families with flat direct image are trivial","Semipositivity conjecture for relative Ricci-flat metrics disproved","Flat Narasimhan-Simha metric implies isomorphic fibers in Calabi-Yau families"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000225,"raw_usage":{"total_tokens":1478,"prompt_tokens":972,"completion_tokens":506,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":588,"completion_tokens_details":{"reasoning_tokens":401}},"tokens_in":588,"tokens_out":506,"duration_ms":5404,"temperature":1.0,"reasoning_tokens":401,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:28:15.295970+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a proper Kähler fibration of log Calabi-Yau fibers satisfying the flatness hypothesis on $p_*(m(K_{X/Y}+B))$ whose fibers are not locally isomorphic; Theorem 1.2 predicts such an example does not exist, so even one would refute the main theorem.","supporting_citations":[],"review_version":1}