{"id":"47dd3ca8-bf5e-4038-8d7c-1f8306dcbc0c","arxiv_id":"2504.16676","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The volume of a divisor on the generic fiber of a flat family equals the infimum of its volumes on any dense family of fibers, yielding upper semicontinuity for reduced irreducible fibers.","lead":"This paper proves that in a flat family of projective varieties, the volume of a divisor on the generic fiber equals the infimum of its volumes on any dense set of fibers. It then shows the volume function is upper semicontinuous when the fibers are reduced and irreducible.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.1's big-case proof depends on a quantitative Fujita error estimate and its spreading to an open base; the cited [Laz04, Thm 11.4.21] is not verified to give the stated (A^{d-1}E)^2 bound, so equation (1) is unsupported.","rationale":"The reader's weakest assumption identifies the same two technical pillars as the proof's soft spot: the precise quantitative form of the Fujita approximation and the spreading-out of the model to an open base. I do not find a clear mathematical contradiction or a known counterexample, and the argument is plausible and likely repairable. However, the heavy reliance on an unstated version of [Laz04, Thm 11.4.21] and on unproved flatness/relative ampleness assertions means the submitted proof is not yet fully rigorous. This matches the reader's CONDITIONAL verdict, so I recommend no change. The concern is about correctness risk, not about novelty or agreement with existing consensus.","tokens_in":3668,"tokens_out":36421,"duration_ms":395677,"concrete_test":"Open [Laz04, Theorem 11.4.21] and its proof and verify whether it contains the quoted inequality (A^{d-1}E)^2 ≤ C H^d(vol(D)-vol(A)) with C independent of ε; if it contains only vol(D)-vol(A)<ε, independently re-derive the stronger inequality from the proof or from the same hypotheses. Then test the spreading step by computing A_t^{d-1}E_t for a non-trivial flat family after shrinking T: if the mixed intersection number is not locally constant for the closed divisor E, equation (1) fails and the infimum equality is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1.1 in the big case reduces to the fiberwise inequality (1), which is obtained by spreading a Fujita decomposition g*D ∼_Q A+E to an open neighborhood U of the base. Two steps are load-bearing. First, the paper cites [Laz04, Theorem 11.4.21] for the inequality (A^{d-1}_η·E_η)^2 ≤ C·H^d_η·(vol(D_η)-vol(A_η)) with a constant C independent of ε. The standard statement of Fujita approximation only asserts that for every ε there is a decomposition with vol(D)-vol(A)<ε; it does not by itself give the quadratic control of A^{d-1}E. If the cited theorem does not contain this quantitative estimate, then the bound A^{d-1}E ≤ C'√ε used in (1) has no proof. Second, the descent and spreading-out step asserts that A is relatively ample, E is flat over T, and g_t is birational on an open neighborhood; the flatness of E (not merely of its support) is needed for the local constancy of A_t^{d-1}E_t. The manuscript asserts these conditions without proof. Since Equation (4) and the final infimum equality are exactly the conclusion of this chain, the central claim is conditional on these two verifications.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proves Theorem 1.1: for a projective flat morphism f:X→T of varieties over an algebraically closed field of characteristic zero and a Cartier divisor D on X, the volume of D on the generic fiber equals the infimum of the volumes of D on fibers over any dense subset T'⊂T. Theorem 1.2 then derives upper semicontinuity of the fiberwise volume function under reduced and irreducible fibers. The proof of Theorem 1.1 first reduces to the case where the generic divisor is big, then applies a Fujita approximation to the geometric generic fiber, spreads the approximation out to an open neighborhood of the base, and uses the Hodge index inequality to control fiberwise volumes. The non-big cases are handled by adding an ample divisor and taking limits in the approximation parameter.","tokens_in":3908,"tokens_out":14180,"duration_ms":136810,"significance":"If the proof is made rigorous, the result is a substantial and clean statement: the generic volume is a sharp lower bound for volumes over any dense set of fibers, and it yields upper semicontinuity as a corollary. The manuscript is self-contained modulo standard references and does not rely on circular arguments. Its main strengths are the elegant reduction to the big-divisor case, the explicit use of Fujita approximation, and the transparent derivation of the fiberwise volume inequality. The claimed theorems are plausible and would be useful in birational geometry. However, several load-bearing technical steps are only sketched, so the current version is a solid proof outline rather than a complete proof.","major_comments":[{"comment":"The proof relies on the estimate (A^{d-1}_η·E_η)^2 ≤ C·(H^d_η)·(vol(D_η)-vol(A_η)), attributed to [Laz04, Theorem 11.4.21]. The standard Fujita approximation theorem as usually stated only asserts the existence of a decomposition with vol(D)-vol(A)<ε for every ε>0; it does not directly give the quadratic control of A^{d-1}E in terms of vol(D)-vol(A). Since this estimate is used to obtain the crucial bound A^{d-1}_η·E_η ≤ C'√ε and then equation (1), the author must either quote the exact statement of Theorem 11.4.21, prove the needed estimate from the construction of the Fujita approximation, or give a precise alternative reference. Without this verification, the central inequality in the big-divisor case is unsupported.","section":"Proof of Theorem 1.1, big-divisor case, after Eq. (1)"},{"comment":"The transitions γ→0+ in Case 1 and γ→ρ+ in Case 2 pass limits inside the equality vol(D_η+γL_η) = inf_{t∈T'} vol(D_t+γL_t). This interchange is not justified. The infimum of a family of continuous functions need not be continuous in general, so the author should supply an argument: for instance, using continuity of volume along the ray γ↦D_η+γL_η, a corresponding continuity or semicontinuity statement for the fiberwise volumes, and the fact that vol(D_η+ρL_η)=0 at the pseudo-effective threshold. As written, the proof establishes Theorem 1.1 only under the additional hypothesis that D_η is big.","section":"Proof of Theorem 1.1, reductions for non-big D_η"},{"comment":"The passage from the geometric generic model g_η:Y_η→X_η to a global model over an open neighborhood of T asserts, after shrinking T, the existence of a compactification Y→T with Y flat and reduced/irreducible fibers, a relatively ample divisor A, a flat divisor E (or at least flat support), and a fiberwise birational morphism g:Y→X over T. These assertions are not immediate: they require generic flatness, spreading-out of morphisms and divisors, and control of the ample property of A over an open base. The flatness of E is particularly important because it is used to conclude that the intersection numbers A_t^{d-1}·(A_t+E_t) are locally constant in t, which is needed for equation (1). The author should provide a detailed proof or precise references (e.g., EGA IV) for each of these spreading-out claims.","section":"Proof of Theorem 1.1, spreading-out paragraph"}],"minor_comments":[{"comment":"The phrase \"algebraically closed number fields\" is imprecise, since number fields are not algebraically closed; presumably \"algebraically closed fields of characteristic zero\" is intended.","section":"Abstract and first line of Section 1"},{"comment":"The inference from vol(D_t+ρL_t)<vol(ρL_t) to \"D_t is not pseudo-effective\" is stated without justification. A short argument that adding an ample divisor to a pseudo-effective divisor cannot decrease the volume would make the step clear.","section":"Proof of Theorem 1.1, Case 2"},{"comment":"The sentence \"by choosing ε and ε′ small enough in Equation (4) and shrinking T accordingly, there always exists t∈T′...\" is terse. It would help to state explicitly that for each ε'' one chooses ε, ε′ with the bound below vol(D_η)+ε'' and that the resulting t may depend on them, which is sufficient to conclude the infimum inequality.","section":"Proof of Theorem 1.1, final paragraph"},{"comment":"The title contains a typo: \"F amilies\" should be \"Families\".","section":"Title"}],"recommendation":"major_revision","confidential_remarks":"The main risk is the attribution to [Laz04, Theorem 11.4.21] of a quantitative Fujita error estimate. I could not verify from the manuscript's citation alone that the stated quadratic bound is a theorem in Lazarsfeld's book; if it is not, the big-divisor case breaks. The spreading-out and limit-interchange steps are standard in spirit but need to be written out. If the author supplies the missing statements and references, the paper could be a nice contribution; the current version is too sketchy for acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The theorem is the kind of clean statement people will want: for a flat family, the volume on the generic fiber is the infimum of volumes over any dense set of fibers, and as a consequence volume is upper semicontinuous when fibers are reduced and irreducible. That is a real step beyond the known jumping examples, and it deserves attention.\n\nThe big-divisor case, which is the heart of the paper, follows a sensible route: Fujita-approximate the generic divisor, spread the approximation to the family, then use Hodge index to get a fiberwise volume bound. The finite-base-change reduction is legitimate given standard invariance of volume under base extension. The overall plan is coherent and plausibly correct.\n\nThe main problem sits in equation (1), where the paper needs a quantitative estimate of the form (A^{d-1}E)^2 ≤ C·H^d·(vol(D)-vol(A)). This is attributed to [Laz04, Theorem 11.4.21]. The standard statement of that theorem in Lazarsfeld's book gives, for each epsilon, a decomposition f*D = A+E with A ample, E effective, and vol(D)-vol(A) < epsilon. It does not, as far as I know, give control on A^{d-1}E. If the book contains such a quantitative version, the author should point to the exact statement; otherwise the inequality needs a proof. This is load-bearing: it is exactly what produces the uniform sqrt(epsilon) error in inequality (4). Without it, the fiberwise bound does not follow.\n\nThere are two smaller gaps. First, in Cases 1 and 2 of the non-big situation, the proof passes gamma -> 0+ or gamma -> rho+ inside an infimum without justification; the equality between the limit of infima and the infimum of limits needs a short argument. Second, the spreading-out step asserts that after shrinking T the divisors are flat, but only says \"Supp(E) flat\"; for the constancy of intersection numbers you need flatness of the divisor itself, not just its support.\n\nThese are identifiable, probably fixable holes rather than signs that the theorem is wrong. The citation pattern is thin but not self-serving. The paper is honest about relying on standard tools.\n\nMy recommendation: send it to peer review. A good referee can ask the author to clarify the quantitative Fujita inequality and tighten the limit interchanges. If the quantitative estimate turns out not to be available from the cited source, the author may still rescue the proof with a direct argument, but that is a necessary step before the result can be considered fully proved.","headline":"A plausible and genuinely new theorem about volumes in families, but the proof leans on a quantitative Fujita-type inequality that the cited reference does not obviously supply.","tokens_in":4464,"tokens_out":7456,"would_cite":false,"duration_ms":78950,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14C20","14D06","14E30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that, in a flat projective family of varieties, the volume of a divisor on the generic fiber equals the infimum of its volumes on fibers over any Zariski dense subset of the base, and derives upper semicontinuity of the…","keywords":["volume of divisors","flat families","upper semicontinuity","generic fiber","Fujita approximation","pseudo-effective divisors","birational geometry","Cartier divisors"],"falsifier":"In the example cited in the introduction of flat families whose divisor volumes jump over a Zariski dense set, compute the actual infimum of the fiber volumes over that dense set and compare it with the generic volume: if the infimum is strictly larger than the generic volume, Theorem 1.1 fails.","tokens_in":3415,"feed_emoji":"📐","tokens_out":5519,"duration_ms":53285,"temperature":0.7,"pith_summary":"The paper establishes that the volume of a divisor on the generic fiber of a flat projective family is exactly the infimum of the volumes on fibers over any dense set of base points. This means the generic volume is a sharp lower bound that cannot be improved by restricting to a dense family of special fibers. The proof spreads a Fujita approximation of the generic divisor over the whole family and controls the error uniformly. As a consequence, for families with reduced and irreducible fibers, the volume function is upper semicontinuous: the set of fibers with volume below any threshold is open.","feed_headline":"Generic volume equals infimum over dense subsets of fibers","feed_subtitle":"In flat families, volumes over dense parameter sets cannot dip below the generic value, and upper semicontinuity follows.","key_machinery":"The load-bearing object is the spread-out Fujita approximation. One approximates the big generic divisor Dη on the geometric generic fiber by g_η^*D_η∼Q A_η+E_η with A_η ample and E_η effective, then spreads this model to a fiberwise birational morphism g:Y→X over T, obtaining a relatively ample divisor A, an effective flat divisor E, and g^*D∼Q A+E. The key estimate is ($A^{{d−1}}$·E)^2 ≤ C·H^d·(vol(D_η)−vol(A_η)) ≤ Cε·H^d, which gives a uniform bound $A_t^{{d−1}}$·(A_t+E_t)≤A_t^d+C′√ε on every fiber. Combining this with the generalized Hodge inequality forces the fiberwise volume inequality from which the infimum equality follows by letting approximation errors tend to zero.","core_discovery":"Theorem 1.1: for a projective flat morphism f:X→T of varieties and a Cartier divisor D on X, writing Dη for the restriction to the generic fiber and Dt for the restriction to a closed fiber, the equality vol(Dη)=inf_{t∈T′}vol(Dt) holds for every dense subset T′ of T. The proof first handles the big case by spreading a Fujita approximation of the geometric generic divisor to a fiberwise birational model over an open neighborhood of the base, then uses Hodge-type inequalities to bound fiberwise volumes; the non-big cases are reduced to the big case by adding ample divisors. Theorem 1.2 follows: when all fibers are reduced and irreducible, the set {t∈T | vol(Dt)<a} is open.","pith_inferences":["Over a one-dimensional base, every infinite subset is dense, so the theorem implies that all but finitely many fibers have volume at least the generic volume; only upward jumps, at finitely many points, can occur.","The uniform spreading technique used here may apply to other asymptotic invariants of divisors, such as restricted volumes or numerical dimensions, in flat families.","The theorem suggests that volume, despite not being constructible, has a semicontinuity structure controlled entirely by the generic fiber, which could simplify computations in moduli problems."],"forward_implications":["For flat families with reduced and irreducible fibers, the volume function t↦vol(D_t) is upper semicontinuous: the sublevel set {t | vol(D_t)<a} is open for every real a.","The generic volume is a sharp lower bound over every dense parameter set: for any ε>0 and any dense T′, there is a fiber t∈T′ with vol(D_t)≤vol(D_η)+ε.","If the generic divisor is not pseudo-effective, then every dense subset of the base contains a fiber whose divisor is not pseudo-effective, since the infimum would otherwise be finite or positive.","Upward jumps in the volume can occur only along non-dense loci; over any dense set, the generic value is the infimum."],"supporting_citations":[{"why":"Supplies the uniform error estimate on the spread Fujita approximation, bounding (A^{d−1}·E)^2 by a multiple of H^d times the volume difference.","marker":"[Laz04, Theorem 11.4.21]"},{"why":"Provides the generalized Hodge inequality used to relate intersection numbers on the fiberwise model and force the fiberwise volume inequality.","marker":"[Laz04, Theorem 1.6.1]"},{"why":"Gives the known example of volumes jumping on infinitely many fibers, which motivates why the generic-fiber infimum statement is nontrivial.","marker":"[PS16]"}],"fun_headline_variants":["Divisor volumes drop only at closed sets in flat families","Generic divisor volume is infimum over dense parameter sets","Upper semicontinuity of divisor volumes in flat families","Infimum over dense fibers equals generic volume for divisors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on the assumption that a Fujita approximation of the geometric generic divisor can be spread out to a whole open neighborhood of the base, with the error bound ($A^{{d−1}}$·E)^2 ≤ Cε remaining uniform on all fibers; if that spreading-out fails on every dense open set, the infimum equality is not established.","fun_headline_variants_meta":{"raw":{"variants":["Divisor volumes drop only at closed sets in flat families","Generic divisor volume is infimum over dense parameter sets","Upper semicontinuity of divisor volumes in flat families","Infimum over dense fibers equals generic volume for divisors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000641,"raw_usage":{"total_tokens":2849,"prompt_tokens":745,"completion_tokens":2104,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":361,"completion_tokens_details":{"reasoning_tokens":2039}},"tokens_in":361,"tokens_out":2104,"duration_ms":13926,"temperature":1.0,"reasoning_tokens":2039,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:04:02.910467+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"In the example cited in the introduction of flat families whose divisor volumes jump over a Zariski dense set, compute the actual infimum of the fiber volumes over that dense set and compare it with the generic volume: if the infimum is strictly larger than the generic volume, Theorem 1.1 fails.","supporting_citations":[],"review_version":1}