{"id":"191379bc-e39f-44de-ab72-7bce195ef28e","arxiv_id":"2507.04838","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A positive characteristic Steenbrink vanishing theorem is proved for rational singularities, giving the vanishing for strongly F-regular threefolds and Q-factorial klt threefolds in large characteristic.","lead":"This paper proves Steenbrink vanishing for rational singularities in positive characteristic, with applications to strongly F-regular threefolds and to Q-factorial klt threefolds in characteristic p>41. The result matters because vanishing theorems for logarithmic differential forms are a standard tool in characteristic zero that had been missing in positive characteristic birational geometry.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Abstract overstates the proven scope: Steenbrink vanishing is proved only for log resolutions whose exceptional divisor supports a pi-ample divisor, while Definition 1.1 requires every log resolution.","rationale":"The reader's formal weakest assumption is the cited [KTT+24] theorem that Q-factorial threefold klt singularities in p>41 are quasi-F-pure. I do not find an internal flaw in Theorem A's proof: Theorem 4.1's use of Serre vanishing under the pi-ampleness hypothesis is standard; Proposition 4.2's long-exact-sequence indexing checks out; F-injectivity implies quasi-F-injectivity through the commutative diagram (2.0.3); and Proposition 4.6's duality argument is intricate but coherent. The external citation is legitimate, so I would not base a conditional verdict on it alone. The load-bearing issue is the mismatch between the abstract and the theorem statements. Definition 1.1 requires every log resolution, but Theorems A and B and the proof strategy only address log resolutions whose reduced exceptional support is pi-ample; [KW24b] gives existence of one such resolution, not a reduction from arbitrary resolutions. Thus the advertised claim for non-Q-factorial strongly F-regular and klt threefolds is unproven in the paper, regardless of the cited theorems. This supports the reader's CONDITIONAL verdict, so I keep it unchanged.","tokens_in":13661,"tokens_out":36930,"duration_ms":410733,"concrete_test":"Take a non-Q-factorial strongly F-regular threefold singularity, for example a 3-dimensional affine toric singularity whose defining cone is non-simplicial and is F-regular in characteristic p, checking F-regularity by the usual combinatorial criterion. Compute, for a log resolution pi:Y->X whose reduced exceptional divisor E does not support a pi-ample divisor, the module R^2 pi_* Omega^2_Y(log E)(-E). If this module is nonzero, the abstract's unrestricted Steenbrink claim fails; if it is always zero, test the alternative possibility that the vanishing is independent of the log resolution, which the paper would then need to state and prove. In the toric setting, the existence of a pi-ample divisor supported on E is a finite linear-programming check in the relative Neron-Severi space.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Definition 1.1 makes Steenbrink vanishing a statement about every log resolution pi:Y->X. Theorem A is proved only under the hypothesis that the reduced exceptional divisor E supports a pi-ample divisor; this hypothesis is used in Theorem 4.1 to choose a pi-ample anti-effective Q-divisor A. Remark 1.2 says the hypothesis is automatic when X is Q-factorial, and otherwise [KW24b, Theorem 1.1] supplies a log resolution with the property, not a proof that every log resolution has it. Consequently Theorem B, as stated, covers only those special log resolutions, and the 'In particular' consequence is restricted to Q-factorial X. Strong F-regularity does not imply Q-factoriality, so the abstract sentence claiming that strongly F-regular threefolds satisfy Steenbrink vanishing is not supported. The same issue arises in Theorem C and the abstract for klt threefolds, where Q-factoriality is assumed. The reliance on [KTT+24, Theorem A] is a normal citation dependency and would not by itself change the verdict; the unresolved gap is the passage from special log resolutions to all log resolutions for non-Q-factorial threefolds.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a positive-characteristic Steenbrink-type vanishing theorem. Theorem A states: if X is a normal d-dimensional variety over a perfect field of positive characteristic, d >= 3, with rational singularities (and F-injective when d = 3), then for every log resolution pi: Y -> X whose reduced exceptional divisor E supports a pi-ample divisor one has R^{d-1} pi_* Omega^{d-1}_Y(log E)(-E) = 0. The proof is split into Theorem 4.1, an annihilation result for R^{d-1} pi_* Omega^i_Y(log E)(-E) under the action of iterated inverse Cartier operators, and Propositions 4.2 and 4.6, injectivity results under rational-singularity and quasi-F-injectivity assumptions. Theorem B derives the vanishing for all pairs (i,j) with i+j > 3 for strongly F-regular threefolds, but only for log resolutions whose reduced exceptional divisor supports a pi-ample divisor; a Q-factorial hypothesis then yields the full Definition 1.1 statement. Theorem C gives full Steenbrink vanishing for Q-factorial klt threefolds in characteristic p > 41, relying on [KTT+24, Theorem A].","tokens_in":13898,"tokens_out":16679,"duration_ms":180123,"significance":"The central two-step mechanism is attractive and appears sound: the inverse iterated Cartier operators provide a uniform way to kill the obstruction in top degree, and injectivity under rational singularities is proved by local duality and the quasi-F-injective condition. The result is new and relevant for birational geometry in positive characteristic, and the dependence on [KTT+24, Theorem A] is explicit and is a normal citation dependency. The main defect is that the abstract and parts of the introduction claim more than the theorems prove: full Steenbrink vanishing for all strongly F-regular threefolds is not established, only the special log resolutions with a pi-ample exceptional support. This is a statement-level overreach rather than a flaw in the internal algebra.","major_comments":[{"comment":"The abstract's opening claim that strongly F-regular threefolds and klt threefolds in characteristic p > 41 satisfy Steenbrink vanishing is not supported by the theorems as stated. Definition 1.1 requires the vanishing for every log resolution. Theorem B proves the vanishing only for log resolutions pi: Y -> X whose reduced exceptional divisor E supports a pi-ample divisor, and its final sentence explicitly uses Q-factoriality to conclude the full Definition 1.1 statement; since strong F-regularity does not imply Q-factoriality, the advertised consequence does not follow. Theorem C does prove full Steenbrink vanishing, but only under Q-factoriality, which the abstract's klt sentence omits. Remark 1.2(1) explains that the pi-ample condition is automatic for Q-factorial X and that [KW24b, Theorem 1.1] supplies a log resolution with the property, but it does not show that every log resolution has it. Please either add a reduction argument from arbitrary log resolutions to those with E supporting a pi-ample divisor, or reformulate the claims with the special-resolution hypothesis made explicit.","section":"Abstract; §1.1, Theorem B; §1.2, Theorem C; Remark 1.2"},{"comment":"The gap between special and arbitrary log resolutions is load-bearing. Theorem A and Theorem B are not statements of Definition 1.1 for non-Q-factorial X; the proof of Theorem 4.1 uses the existence of a pi-ample anti-effective Q-divisor A supported on E, so the vanishing is genuinely proved only for that class of resolutions. If the authors believe the full Steenbrink vanishing statement follows for all log resolutions once it is known for one resolution with the pi-ample property, that implication is absent from the paper and should be proved. Otherwise the abstract, the introduction, and the theorem announcements should consistently use a qualified phrase such as 'Steenbrink vanishing for log resolutions whose reduced exceptional divisor supports a pi-ample divisor'.","section":"Theorem A; §1.3; Theorem 4.1"}],"minor_comments":[{"comment":"There is a repeated typo: the superscript 'n-i' in Lemma 3.2(2) should read 'd-i', and in the proof the two occurrences of 'B^{n-1}Omega^{d-1}_Y' should be 'B^{n-1}Omega^{d-i}_Y'.","section":"Lemma 3.2(2) and proof"},{"comment":"The sentence 'the top horizontal arrow is surjective H^d_m(O_X) -> H^d_m(Q_{X,l}) is injective' is garbled; it should say that the horizontal map H^{d-1}_m(Q_{X,l}) -> H^{d-1}_m(B_{l,X}) is surjective and that H^d_m(O_X) -> H^d_m(Q_{X,l}) is injective.","section":"Proposition 4.6 proof"},{"comment":"There are several typographical errors: 'OY -moudle' in §3.0.1, 'deinition' in the proof of Theorem 4.1, and the spacing in 'p >41' in the abstract.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the mathematical core is sound conditional on [KTT+24] and [KW24b], but the abstract should be aligned with the theorem statements. The author's own earlier papers are heavily used, but the dependency is explicit and not circular. I recommend major revision; if the author chooses to narrow the claims to the special-resolution version, the revised version would be acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main result, Theorem A, is a real new vanishing statement: for a d-dimensional normal variety with rational singularities (F-injective if d=3), the top higher direct image of Ω^{d-1}(log E)(-E) vanishes under a natural 'ample exceptional divisor' condition. The proof splits into an annihilation theorem via inverse iterated Cartier operators and an injectivity theorem under rational singularities. That decomposition is clean and the result is new even for smooth X. Example 4.7 shows the range is sharp. This is a solid contribution to positive-characteristic MMP.\n\nWhere it gets shaky is the abstract. Definition 1.1 defines Steenbrink vanishing as the vanishing for every log resolution. Theorem A only proves it for log resolutions whose reduced exceptional divisor E supports a pi-ample divisor. That is automatic when X is Q-factorial, and otherwise Remark 1.2 cites a result giving existence of one such resolution—not a proof that all resolutions have the property. So Theorem B proves the vanishing only for those special resolutions, not that strongly F-regular threefolds satisfy Steenbrink vanishing as defined. The abstract sentence is unsupported. The fix is not cosmetic: either the main theorems should be phrased as a property of a chosen log resolution, or the author needs an argument showing the vanishing is independent of the resolution. I did not see that argument.\n\nThe klt application (Theorem C) is fine, though it depends on the recent quasi-F-purity result [KTT+24], which is cited, not proved. That is a normal dependency.\n\nMinor issues: in the introduction's Theorem E the phrase 'Moreover, X is quasi-F-injective in addition' needs an 'if'; and some notation in Lemma 3.2 is dense but the proofs go through.\n\nBottom line: the paper deserves a serious referee. The main theorem is new, the proof seems sound, and the overclaim is local to the abstract and the wording of Theorem B. I would ask for a revision that either narrows the claims or proves the independence; then recommend acceptance. I would cite the main theorem in my own work on positive-characteristic singularities.","headline":"New positive-characteristic vanishing theorem with a clean proof, but the abstract overstates the scope: full Steenbrink vanishing is only proved for Q-factorial threefolds or special log resolutions.","tokens_in":14400,"tokens_out":3453,"would_cite":true,"duration_ms":36437,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14F10","13A35","14F17","14B05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that rational singularities of dimension at least three over a perfect field of positive characteristic satisfy Steenbrink vanishing in the central coefficient range, with full Steenbrink vanishing following for strongly…","keywords":["Steenbrink vanishing","rational singularities","positive characteristic","differential forms","Cartier operator","F-injective","klt threefolds","vanishing theorems"],"falsifier":"The claim would be settled by finding a normal variety $X$ of dimension $d\\ge 3$ with rational singularities, F-injective when $d=3$, over a perfect field of positive characteristic, with a log resolution $\\pi:Y\\to X$ whose reduced exceptional divisor $E$ supports a $\\pi$-ample divisor, and such that $R^{d-1}\\pi_*\\Omega^{d-1}_Y(\\log E)(-E)\\neq 0$. For Theorem C it would suffice to exhibit a Q-factorial klt threefold over a perfect field of characteristic $p>41$ with $R^2\\pi_*\\Omega^2_Y(\\log E)(-E)$ nonzero.","tokens_in":13466,"feed_emoji":"🧮","tokens_out":16723,"duration_ms":154315,"temperature":0.7,"pith_summary":"Steenbrink vanishing is a local vanishing statement for the higher direct images $R^j\\pi_*\\Omega^i_Y(\\log E)(-E)$ on a log resolution $\\pi:Y\\to X$, meaning a resolution whose reduced exceptional divisor has simple normal crossings; it is the logarithmic analogue of Akizuki–Nakano vanishing and is known in characteristic zero but can fail in positive characteristic. The paper establishes the first positive-characteristic instance in the hardest range, $R^{d-1}\\pi_*\\Omega^{d-1}_Y(\\log E)(-E)=0$, for rational singularities of dimension $d\\ge 3$, assuming F-injectivity (injective Frobenius action on local cohomology) when $d=3$ and assuming the reduced exceptional divisor supports a $\\pi$-ample divisor. The proof routes the higher direct image through the inverse iterated Cartier operator and shows the map is simultaneously zero and injective. As consequences, strongly F-regular threefolds and Q-factorial klt (Kawamata log terminal) threefolds in characteristic $p>41$ satisfy Steenbrink vanishing, a statement the paper notes is new even when $X$ is smooth.","feed_headline":"Steenbrink vanishing holds for rational singularities in char p>0","feed_subtitle":"A positive-characteristic proof now covers strongly F-regular and klt threefolds.","key_machinery":"The load-bearing object is the inverse iterated Cartier operator $C_n^{-1}$, built from the logarithmic Cartier isomorphism on the smooth log pair $(Y,E)$. It maps $\\Omega^i_Y(\\log E)(-E)$ into the quotient $G_n\\Omega^i_Y(\\log E)(-E)=F^n_*\\Omega^i_Y(\\log E)(-E)/B_n\\Omega^i_Y(\\log E)(-E)$, and Theorem D shows that for large $n$ the induced map on $R^{d-1}\\pi_*$ is zero whenever $E$ supports a $\\pi$-ample divisor. Propositions 4.2 and 4.6 then show that under rational singularities, with quasi-F-injectivity (a Witt-vector generalization of F-injectivity) added for the $R^2$ case, the same induced map is injective for the $i=d-1$ sheaves. A module lying in the kernel of a map that is also injective must be zero, so the two statements lock together to force the vanishing in Theorem A.","core_discovery":"The paper's central claim is Theorem A: over a perfect field of characteristic $p>0$, if $X$ is a normal variety with rational singularities, $\\dim X\\ge 3$, and $X$ is F-injective when $\\dim X=3$, then for every log resolution $\\pi:Y\\to X$ whose reduced exceptional divisor $E$ supports a $\\pi$-ample divisor, $R^{d-1}\\pi_*\\Omega^{d-1}_Y(\\log E)(-E)=0$. This is exactly the $(i,j)=(d-1,d-1)$ case, the one not covered by Grauert–Riemenschneider-type arguments. With known inputs, Theorem A yields Theorem B, that every strongly F-regular threefold — a positive-characteristic singularity class analogous to klt — over a perfect field satisfies Steenbrink vanishing, and Theorem C, that every Q-factorial klt threefold in characteristic $p>41$ satisfies Steenbrink vanishing. The author's stated view is that Theorem A is new even when $X$ is smooth.","pith_inferences":["If the paper's zero-plus-injective dichotomy for the inverse iterated Cartier operator is robust, the same two ingredients could be used to attack Steenbrink-type vanishing for other values of $i$ in any dimension, provided duality computations like Lemma 3.2 can be pushed through.","The prime bound $p>41$ in Theorem C is inherited entirely from the cited quasi-F-purity theorem; if that result's bound improves, Theorem C should improve with it without changing the present argument.","A natural sharpness test is to construct rational singularities that fail to be quasi-F-injective, to see whether the F-injectivity assumption in dimension three is truly needed for Theorem A."],"forward_implications":["Every Q-factorial strongly F-regular threefold in positive characteristic satisfies full Steenbrink vanishing.","Every Q-factorial klt threefold over a perfect field of characteristic $p>41$ satisfies Steenbrink vanishing.","For isolated singularities satisfying the hypotheses, the logarithmic extension property for one-forms follows, because the obstruction $H^1_E(Y,\\Omega^i_Y(\\log E))$ vanishes by duality as noted in Remark 4.8.","Theorem A supplies a Steenbrink-type vanishing statement that is new even when $X$ is smooth, since smooth varieties have rational singularities in all characteristics.","In the threefold theorems the essential case $(i,j)=(2,2)$ is now handled, while the other ranges with $i+j>3$ were already known for klt threefolds when $p>5$."],"supporting_citations":[{"why":"Guarantees, when log resolutions exist, a resolution whose reduced exceptional divisor supports a pi-ample divisor, the standing hypothesis of Theorem A.","marker":"[KW24b, Theorem 1.1]"},{"why":"Defines quasi-F-injectivity and supplies the Witt-vector module constructions and the criterion that quasi-F-purity implies quasi-F-injectivity.","marker":"[KTT+22]"},{"why":"Provides the commutative diagram comparing Frobenius on Witt vectors with the Cartier operator, used to identify the higher direct image of B_n Omega^1_Y with B_{n,X}.","marker":"[KTT+24, Lemma 6.7]"},{"why":"Shows Q-factorial three-dimensional klt singularities over a perfect field of characteristic p>41 are quasi-F-pure, the bridge from Theorem A to Theorem C.","marker":"[KTT+24, Theorem A]"},{"why":"Shows three-dimensional klt singularities are rational in characteristic p>5, an input for Theorem C.","marker":"[ABL22, Corollary 1.3]"},{"why":"Supplies the companion vanishing and rationality results for klt threefolds in p>5, covering the non-essential (i,j) cases in Theorem C.","marker":"[BK23]"},{"why":"Shows strongly F-regular threefolds have rational singularities, converting Theorem A into Theorem B.","marker":"[BKR25, Theorem 3.5]"},{"why":"Gives the duality isomorphism for logarithmic forms that turns local cohomology statements into statements about higher direct images.","marker":"[KW24a, Lemma 2.8]"},{"why":"Supplies the logarithmic Cartier isomorphism on which the inverse iterated Cartier operator is built.","marker":"[Har98, Lemma 3.3]"}],"fun_headline_variants":["Steenbrink vanishing for rational singularities: char p proof","Char p Steenbrink vanishing: the missing case for rational singularities","Strongly F-regular threefolds satisfy Steenbrink vanishing","Klt threefolds in char p>41 satisfy Steenbrink vanishing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument relies on there being a log resolution whose reduced exceptional divisor supports a $\\pi$-ample divisor, and for the klt threefold theorem it also relies on the cited result that Q-factorial klt threefolds in characteristic $p>41$ are quasi-F-pure.","fun_headline_variants_meta":{"raw":{"variants":["Steenbrink vanishing for rational singularities: char p proof","Char p Steenbrink vanishing: the missing case for rational singularities","Strongly F-regular threefolds satisfy Steenbrink vanishing","Klt threefolds in char p>41 satisfy Steenbrink vanishing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001168,"raw_usage":{"total_tokens":4756,"prompt_tokens":794,"completion_tokens":3962,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":410,"completion_tokens_details":{"reasoning_tokens":3881}},"tokens_in":410,"tokens_out":3962,"duration_ms":30245,"temperature":1.0,"reasoning_tokens":3881,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T19:39:12.545293+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The claim would be settled by finding a normal variety $X$ of dimension $d\\ge 3$ with rational singularities, F-injective when $d=3$, over a perfect field of positive characteristic, with a log resolution $\\pi:Y\\to X$ whose reduced exceptional divisor $E$ supports a $\\pi$-ample divisor, and such that $R^{d-1}\\pi_*\\Omega^{d-1}_Y(\\log E)(-E)\\neq 0$. For Theorem C it would suffice to exhibit a Q-factorial klt threefold over a perfect field of characteristic $p>41$ with $R^2\\pi_*\\Omega^2_Y(\\log E)(-E)$ nonzero.","supporting_citations":[],"review_version":1}