{"id":"dd789a13-b56b-4b95-a4ec-37eb746a9e02","arxiv_id":"2507.00771","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves an equivalence between Cartier crystals and V-crystals on dual abelian varieties and derives H^0(X,ω_X)≠0, with S^0(X,ω_X)≠0 in the ordinary case, for normal proper varieties of maximal Albanese dimension.","lead":"This paper presents a simplified and improved framework for positive characteristic generic vanishing, centered on a new equivalence between Cartier crystals and V-crystals on dual abelian varieties. It derives a concrete geometric consequence: normal proper varieties of maximal Albanese dimension always have a nonzero canonical section, with a Frobenius-stable version under ordinarity.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Stein factorization step in Theorem 4.3 uses the inclusion f_*ω_X ⊆ ω_Y, which is not standard and is false for general proper birational maps of normal varieties; the reduction to the finite case is therefore unjustified.","rationale":"The reader's weakest assumption pinpoints exactly the same load-bearing step: the reduction in Theorem 4.3 via Stein factorization. I agree that this is where the argument is least secure. My read strengthens the concern slightly: the asserted inclusion f_*ω_X ⊆ ω_Y is not merely unproved but is false for general proper birational maps between normal varieties, as the trace map goes in the opposite direction and can be surjective with nonzero kernel. The theorem might still be true, and the gap might be fixable by using the trace image and proving the required lifting property or by proving that the relevant Y has rational singularities, but the paper supplies none of this. The rest of the paper is careful and the categorical results in Sections 3.1–3.2 appear coherent; there is no machine-checked formal verification, so the proof relies on traditional arguments. Since the reader already assigned a conditional verdict and this concern supports that verdict without moving it, I recommend no change: the paper should be accepted only after the Stein factorization step is either corrected or properly justified.","tokens_in":26594,"tokens_out":23233,"duration_ms":255209,"concrete_test":"Re-derive the reduction step of Theorem 4.3 without invoking f_*ω_X ⊆ ω_Y: define N := Tr(f_*M) ⊆ ω_Y, where Tr is the trace map of the proper birational morphism f. Prove directly that a nonzero section of N pulls back to a nonzero section of M, or find a counterexample. If the implication fails, the reduction is invalid. Equivalently, test the claimed inclusion on the blowup of the affine cone over an elliptic curve: compute H^0(X,ω_X) and H^0(Y,ω_Y); since H^0(X,ω_X) > H^0(Y,ω_Y), the inclusion f_*ω_X ⊆ ω_Y fails, so the proof must supply an additional property of Y that rules out this behavior (e.g., rational singularities) and prove it.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing gap is in the proof of Theorem 4.3 (Section 4). After Stein factorization X→Y→A of the generically finite morphism a, the paper asserts 'f_* fω_X ⊆ f_*ω_X ⊆ ω_Y' and concludes that the statement for the Cartier submodule M := fω_X follows from the statement for f_*M ⊆ ω_Y. The inclusion f_*ω_X ⊆ ω_Y is not a standard consequence of proper birationality for normal varieties. The canonical morphism supplied by Grothendieck duality is the trace map f_*ω_X → ω_Y, which is surjective, not injective. In general f_*ω_X is larger than ω_Y: for the blowup of the affine cone over an elliptic curve, H^0(X,ω_X) has strictly larger dimension than H^0(Y,ω_Y) because the singularity is non-rational. Thus a Cartier submodule of ω_X need not push forward to a submodule of ω_Y. The reader labeled this a standard unproved fact, but without additional hypotheses (e.g., Y having rational singularities) it is false. In the specific situation, Y is the normalization of a subvariety of an abelian variety, and the paper gives no argument that such Y have the needed property in positive characteristic. Since this step is the bridge from the finite case to the general case, the proof of Theorem 4.3 is incomplete as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a unified treatment of positive characteristic generic vanishing theory. It reproves Hacon–Patakfalvi's theorem that Cartier modules are GV-sheaves up to nilpotence, proves an equivalence between Cartier crystals on an abelian variety and V-crystals on its dual (Theorem 3.2.1), improves the support-locus approximation theorem of Hacon–Patakfalvi, and derives an application: a normal proper variety of maximal Albanese dimension has nonzero H^0 of its canonical sheaf, with a stronger statement S^0(X,\\omega_X)\\neq 0 when there is a generically finite morphism to an ordinary abelian variety. The paper is partly expository but contains new statements, worked examples, and a discussion of pathologies in the positive-characteristic theory.","tokens_in":26919,"tokens_out":30162,"duration_ms":341170,"significance":"If the proofs can be completed, the categorical equivalence between Cartier crystals and V-crystals is a strong structural result, and the geometric application is a natural positive-characteristic analogue of a characteristic-zero theorem that is likely to be useful for further birational work. The paper is clearly written and gives helpful examples and explicit statements of pathologies. It relies on standard prior theorems rather than circular reasoning. The main concerns are a false or unjustified birational inclusion in the final application and a too-terse spectral sequence argument in the proof of the new equivalence.","major_comments":[{"comment":"The final reduction step uses the inclusion f_*\\omega_X \\subseteq \\omega_Y after Stein factorization f:X\\to Y. For a proper birational morphism f between normal varieties with f_*O_X=O_Y, the natural inclusion goes in the other direction, \\omega_Y \\subseteq f_*\\omega_X, and can be strict: for a resolution of a non-rational surface singularity, f_*\\omega_X is strictly larger than \\omega_Y. The trace morphism supplied by duality is f_*\\omega_X \\to \\omega_Y, not an inclusion. Since Y is only known to be finite over an abelian variety, no argument is given that Y has rational singularities, which would force equality. Therefore the reduction of the generically finite case to the finite case is not justified, and the proof of Theorem 4.3 is incomplete as written.","section":"Section 4, proof of Theorem 4.3"},{"comment":"The proof uses a “hypercohomology spectral sequence of V-modules” with E_2-term H^a FMA(H^b(\\tau_{\\leq -1}M^\\bullet)) supposedly converging to H^{a-b} FMA(\\tau_{\\leq -1}M^\\bullet). The indexing is nonstandard (the usual target is H^{a+b}), and the assertion that this spectral sequence “degenerates at the level of V-crystals” by Theorem 3.1.4 is not a formal consequence of that theorem, which concerns cohomology sheaves of a single Fourier–Mukai transform rather than differentials in a spectral sequence. This step is used to conclude that every H^b(\\tau_{\\leq -1}M^\\bullet) is nilpotent. The argument needs to be rewritten with a precise filtration and a verifiable degeneration statement.","section":"Section 3.2, Proposition 3.2.5"},{"comment":"The proof concludes by “Fujita vanishing” after reducing to H^i(A, \\varphi_L^*M \\otimes L^{p^{es}} \\otimes \\alpha)=0 for all i>0 and all \\alpha\\in Pic^0(A). The standard Fujita theorem gives, for a fixed coherent sheaf and a fixed ample line bundle, a bound for H^i(F\\otimes M^n), but here the ample line bundle varies continuously with \\alpha, namely M=L^{p^{es}}\\otimes\\alpha. A simultaneous bound over the whole family needs an additional uniform-vanishing statement, which is neither proved nor cited. Please state the exact uniform Fujita theorem used, or prove the uniformity.","section":"Section 3.1, Lemma 3.1.11"},{"comment":"The proof uses the isomorphism H^d(X,\\omega_X)\\cong H^0(X,O_X)^\\vee, citing [PZ21, Proposition 2.4]. For a normal but not necessarily Cohen–Macaulay proper variety X, the reflexive hull \\omega_X does not automatically give this Serre duality statement by the standard form of Serre duality. Please state the precise hypotheses of [PZ21, Proposition 2.4] and check that they apply to the varieties considered here; if the proposition is valid for all normal proper varieties, a one-line explanation of why would remove the concern.","section":"Section 4, finite case in Theorem 4.3"}],"minor_comments":[{"comment":"The phrase “Fix L ample on A” appears to be a typo; in light of Definition 3.1.9 it should probably be “on \\hat A”, or the notation should be harmonized with the rest of the section.","section":"Section 3.1, Lemma 3.1.11"},{"comment":"In the displayed statement of the two functors, the second functor sends a V-module N to H^0(FMA(N)); to make sense as a Cartier module on A, this should be H^0(FM_{\\hat A}(N)), and the proof uses FM_{\\hat A}.","section":"Theorem 3.2.1"},{"comment":"The arrows in the displayed diagram should be checked; as printed, the top and bottom rows do not form the commutative diagram of exact triangles described in the text.","section":"Section 3.2, equation (3.2.5.c)"},{"comment":"The description of W^i_F(\\omega_A) could use a sentence explaining how Serre duality and the p-rank r determine the Frobenius action on H^i(A,\\omega_A), since the displayed formula is not immediate.","section":"Example 3.4.1"},{"comment":"There are several typos, e.g., “an other commutative diagram” in the proof of Proposition 3.2.5 and “philoshophy” in Section 1.1; these should be corrected in revision.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper's main new results are plausible and the exposition is strong, but the proof of Theorem 4.3 contains a birational inclusion that is false in general and cannot be repaired by a citation, and the spectral sequence step in the proof of Theorem 3.2.1 is too terse to verify. I would ask for a substantive revision rather than reject, because the finite case and the categorical statements may well be salvageable with a different reduction and a clarified spectral sequence argument."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe paper does two things well. First, it gives a clean, self-contained account of the Hacon–Patakfalvi generic vanishing machinery, simplifying the proofs of [HP16] and [HP22] without changing the main lines. Second, it states and proves a genuinely new categorical equivalence: Theorem 3.2.1, the contravariant equivalence between Cartier crystals on an abelian variety and V-crystals on its dual. That result is formal once you have the nilpotence theorems, but it is a useful and satisfying statement, and it gives the field a single reference for this circle of ideas. The support-locus material in Section 3.3 is also a genuine improvement over [HP22], with cleaner constructions and a decidable example.\n\nThe soft spot is Theorem 4.3, and it is not a minor gap. After the Stein factorization X→Y→A, the proof asserts the inclusion f_*ω_X ⊆ ω_Y. That is backwards. The trace map f_*ω_X → ω_Y is surjective, and for normal varieties with non-rational singularities f_*ω_X is strictly larger than ω_Y, not smaller. For example, the blowup of the affine cone over an elliptic curve has H^0(X,ω_X) strictly larger than H^0(Y,ω_Y). The author gives no argument that Y, the normalization of a subvariety of an abelian variety, has rational singularities in positive characteristic. So the bridge from the finite case to the generically finite case is not justified. The theorem may be true, but the proof as written does not establish it. The H^0 part can likely be recovered from existing results, and the S^0 statement in the ordinary case needs a different argument.\n\nThere are smaller issues: the Fujita vanishing step in Lemma 3.1.11 needs uniformity over Pic^0(A), which is asserted without proof, and the inverse-system notation in that lemma is sloppy. These are fixable. The categorical equivalence stands on its own and is the real content of the paper.\n\nI would send this to a serious referee. It deserves refereeing because the equivalence is important and the exposition is valuable, but the referee should catch the Stein factorization error and require a corrected proof of Theorem 4.3 or an explicit acknowledgment that the reduction is incomplete. I would also bring it to a reading group: the categorical part is worth knowing, and the error is a useful cautionary tale about dualizing sheaves and birational maps in positive characteristic.\n\nBest,","headline":"A useful consolidation with one genuinely new categorical equivalence, and a broken reduction in the effectivity theorem.","tokens_in":27454,"tokens_out":4595,"would_cite":true,"duration_ms":50045,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14K05","14G17","14F17"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper establishes an equivalence between Cartier crystals and V-crystals on dual abelian varieties in positive characteristic, and uses it to prove that normal varieties of maximal Albanese dimension always carry nonzero canonical…","keywords":["generic vanishing","Cartier crystals","V-crystals","positive characteristic","abelian varieties","Albanese dimension","Fourier-Mukai transform","cohomological support loci"],"falsifier":"Exhibit a non-nilpotent Cartier module $M$ on an abelian variety whose degree-zero Fourier-Mukai transform $H^0 FM_A(M)$ is nilpotent as a V-module; by Theorem 3.2.1 this cannot happen, so such an example would refute the equivalence. Concretely, this can be tested by computing the Tor groups defining $W^i_V$ for the injective hull of a candidate module such as the one appearing in Example 3.4.2.","tokens_in":26395,"feed_emoji":"📐","tokens_out":7939,"duration_ms":79211,"temperature":0.7,"pith_summary":"The paper develops a unified treatment of generic vanishing in characteristic p, where the usual vanishing theorems fail. Its central new result is an equivalence of categories between Cartier crystals on an abelian variety and V-crystals on its dual, induced by taking degree-zero cohomology of the Fourier-Mukai transform. From this it derives that a normal proper variety of maximal Albanese dimension satisfies $H^0(X, \\omega_X) \\neq 0$, and that if it admits a generically finite morphism to an ordinary abelian variety then the Cartier-stable subspace $S^0(X, \\omega_X)$ is nonzero. These results give a positive-characteristic replacement for characteristic-zero generic vanishing and new effectivity statements for canonical forms.","feed_headline":"Cartier crystals are dual to V-crystals on abelian varieties","feed_subtitle":"In positive characteristic, this equivalence forces nonzero canonical forms on maximal-Albanese varieties.","key_machinery":"The central object is the symmetric Fourier-Mukai transform $FM_A: D^b_{\\mathrm{coh}}(A)^{\\mathrm{op}} \\to D^b_{\\mathrm{coh}}(\\hat{A})$, together with the truncation $H^0$. The paper shows that Cartier modules, coherent sheaves with an $F_*$-linear structural morphism, are sent by $H^0 FM_A$ to V-modules, coherent sheaves with a Verschiebung-linear morphism, and that nilpotence of Cartier modules corresponds to nilpotence of V-modules. The key technical step is the theorem that for any Cartier module $M$, the higher cohomology sheaves $H^i FM_A(M)$ are nilpotent V-modules for $i\\neq 0$; this is what allows the categories to be taken modulo nilpotence and yields the equivalence.","core_discovery":"The central claim is that the symmetric Fourier-Mukai transform, taken at the level of $H^0$, induces an anti-equivalence between Cartier crystals on an abelian variety $A$ and V-crystals on its dual $\\hat{A}$. Cartier crystals are coherent sheaves with a Frobenius-linear structural map, considered up to nilpotence; V-crystals are the dual notion, with a Verschiebung-linear map. The theorem says that working up to nilpotence is exactly the right amount of coarsening for the transform to become an equivalence. As a geometric consequence, the paper proves that any normal proper variety $X$ with a generically finite map to an abelian variety has $H^0(X, \\omega_X) \\neq 0$, and when the abelian variety is ordinary the same holds for the subspace $S^0(X, \\omega_X)$ of forms fixed by the Cartier operator.","pith_inferences":["The equivalence suggests that a full derived anti-equivalence may hold between the corresponding derived categories of crystals, with $H^0$ appearing as a truncation; the paper does not pursue that extension.","If the same $H^0$ Fourier-Mukai construction works relatively for families of abelian varieties, it could yield relative generic vanishing statements for fibrations in positive characteristic, beyond the absolute setting treated here.","The $S^0(X, \\omega_X) \\neq 0$ statement for ordinary Albanese targets gives a concrete numerical test for how closely 'ordinary' approximates the characteristic-zero behavior of maximal Albanese dimension in birational geometry.","One could probe sharpness of the approximation theorem by checking on the Example 3.4.2 configuration whether the $W^i$ loci are indeed as fine as the Tor computations predict; a mismatch there would reveal a gap in the crystal-level claims."],"forward_implications":["If the equivalence holds, generic vanishing for Cartier modules is governed entirely by the V-crystal $H^0 FM_A(M)$, so cohomological support loci can be computed as Tor groups of an associated injective V-module.","Theorem 4.3 gives a new effectivity statement: every normal proper variety of maximal Albanese dimension in positive characteristic has a nonzero global canonical form.","When the variety admits a generically finite morphism to an ordinary abelian variety, the Cartier operator has a nonzero fixed global form, so $S^0(X, \\omega_X) \\neq 0$.","The refined support loci $W^i$ and $Z^i$ of Theorem 3.3.5 satisfy the expected codimension bounds and are stable under the Frobenius pullback map $p^s$, giving a tighter approximation of the nonclosed loci $W^i_F(M)$.","The equivalence in Theorem 3.2.1 formally contains the statements of Hacon-Patakfalvi's Theorem 5.2 and Corollary 5.3, so those earlier results follow as corollaries."],"supporting_citations":[{"why":"Supplies the core vanishing statement (Theorem 3.1.4) that higher Fourier-Mukai cohomology of Cartier modules is nilpotent, the foundation of the crystal equivalence.","marker":"[HP16]"},{"why":"Defines V-modules and contains the approximation theorem for support loci that the paper improves, along with the statement of Theorem 5.2 that Theorem 3.2.1 implicitly contains.","marker":"[HP22]"},{"why":"Introduces the Fourier-Mukai transform and its basic duality properties, including the computation for $\\omega_A$ used in the equivalence.","marker":"[Muk81]"},{"why":"Supplies the symmetric Fourier-Mukai transform formalism and the compatibilities (inverse property, translation, pushforward) used throughout.","marker":"[Sch22]"},{"why":"Establishes the characterization of GV-sheaves via concentration of the Fourier-Mukai transform, the conceptual backdrop for the nilpotence analogue.","marker":"[PP11a]"},{"why":"Provides the Serre duality statement $H^d(X,\\omega_X)\\cong H^0(X,\\mathcal{O}_X)^\\vee$ used in the finite case of Theorem 4.3.","marker":"[PZ21]"},{"why":"Defines the Cartier operator that gives $\\omega_X$ its Cartier module structure, the central object whose effectivity is the geometric conclusion.","marker":"[Car57]"}],"fun_headline_variants":["Cartier-V duality forces nonzero canonical forms","Positive char: Cartier crystals dual to V-crystals","Generic vanishing simplified via Cartier-V duality","Nonvanishing canonical forms from Cartier-V duality"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that the finite-map case implies the general case of Theorem 4.3 assumes that pushing the dualizing sheaf forward along the birational part of the Stein factorization lands inside the target's dualizing sheaf, an inclusion that is standard but neither proved nor cited.","fun_headline_variants_meta":{"raw":{"variants":["Cartier-V duality forces nonzero canonical forms","Positive char: Cartier crystals dual to V-crystals","Generic vanishing simplified via Cartier-V duality","Nonvanishing canonical forms from Cartier-V duality"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001178,"raw_usage":{"total_tokens":4784,"prompt_tokens":776,"completion_tokens":4008,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":392,"completion_tokens_details":{"reasoning_tokens":3948}},"tokens_in":392,"tokens_out":4008,"duration_ms":31105,"temperature":1.0,"reasoning_tokens":3948,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T21:13:00.571377+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a non-nilpotent Cartier module $M$ on an abelian variety whose degree-zero Fourier-Mukai transform $H^0 FM_A(M)$ is nilpotent as a V-module; by Theorem 3.2.1 this cannot happen, so such an example would refute the equivalence. Concretely, this can be tested by computing the Tor groups defining $W^i_V$ for the injective hull of a candidate module such as the one appearing in Example 3.4.2.","supporting_citations":[],"review_version":1}