{"id":"b3305a87-7894-4206-a0fe-9bbfeec80be5","arxiv_id":"2501.12041","paper_version":3,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper establishes a canonical bundle formula for Fano-type threefold fibrations in large characteristic and proves effective birationality for strongly F-regular weak Fano varieties with bounded Gorenstein index.","lead":"A mathematics paper proves new tools for studying Fano varieties in positive characteristic, including a canonical bundle formula for threefold fibrations and a boundedness result for anticanonical maps of F-regular Fano varieties. If correct, it makes progress on the BAB conjecture, a central open problem about when Fano varieties form a bounded family.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.4 is not established as written: its induction rests on the unproved assertion in Remark 4.1 that any ample Q-divisor with vol(D) ≥ (2d)^d yields a bounded covering family of F-pure centers; without this, the effective birationality conclusion has no engine.","rationale":"The central claim Theorem 4.4 is a positive-characteristic analogue of Birkar's effective birationality. For it to hold, one needs a supply of boundaries with controlled F-pure centers through arbitrary general points. Remark 4.1 is the only statement providing this supply, and it is neither proved nor derived from a cited theorem. The text says 'combine the statements above and [HMX14, 7.1] and some tie breaking arguments', but [HMX14, 7.1] is an ACC result for log canonical thresholds in characteristic 0 and does not, by itself, produce bounded covering families of F-pure centers in characteristic p. The distinction drawn in Remark 4.1 between potential birationality and F-potential birationality is real, but it is a caveat, not a proof. The reader's conditional verdict is therefore appropriate; if anything, the unresolved state of Remark 4.1 and the coefficient mismatch in the induction step mean the burden is on the author to supply a complete argument. I recommend keeping the CONDITIONAL verdict rather than accepting the theorem. The proposed test, a direct check on surfaces, would settle whether the covering-family assertion has content or is simply false in a basic case.","tokens_in":22780,"tokens_out":7862,"duration_ms":79456,"concrete_test":"Independently prove or disprove Remark 4.1 in the minimal case d=2: take X a smooth del Pezzo surface and D = −a K_X with a chosen so that vol(D) = a^2(−K_X)^2 ≥ 16. Compute, for general x,y, whether there is Δ ∼_Q D with (X,Δ) sharply F-pure at x, isolated F-pure center {x}, and not strongly F-regular at y; this can be checked by explicit test-ideal computations for the families Δ = general member of |−aK_X| plus a divisor through y. If a counterexample is found, Remark 4.1 fails and Theorem 4.4 is unsupported. Separately, recompute the coefficient identity in the induction step of Theorem 4.4: verify whether (1−δ)Δ_i + cA with c<1 can be Q-linearly equivalent to −4 l n K_{X_i} when Δ_i ∼_Q −(n+1)K_{X_i} and A ∼_Q −l n K_{X_i}; if not, the induction does not terminate even under Remark 4.1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Remark 4.1 is the engine of Theorem 4.4. It asserts, without proof, that for any ample Q-divisor D with vol(D) ≥ (2d)^d on a d-dimensional variety X there is a bounded covering family {G_i} of subvarieties such that for general x,y there exists Δ ∼_Q D with (X,Δ) sharply F-pure at x, having a unique F-pure center G_i containing x, and not strongly F-regular at y. The paragraph preceding it only explains why 'potentially birational' does not automatically become 'F-potentially birational'; it does not prove the covering-family statement. In characteristic 0, the analogous assertion is a major theorem of Birkar and relies on asymptotic saturation and the ACC of log canonical thresholds. In positive characteristic, F-pure centers are less rigid than lc centers, there is no established boundedness theorem for them, and the cited [HMX14, 7.1] is a characteristic-0 ACC result that does not directly apply. Moreover, even granting Remark 4.1, the induction step in the proof of Theorem 4.4 has a numerical problem: from Δ_i ∼_Q −(n+1)K_{X_i} and A ∼_Q H = −l n K_{X_i}, Lemma 4.3 produces (1−δ)Δ_i + cA with c<1, whose class is −((1−δ)(n+1)+c l n)K_{X_i}; equating this to the stated replacement −4 l n K_{X_i} forces c ≈ 4 for l ≥ 2, contradicting c<1. Thus the proof would need additional scaling or a different choice of boundary, and the induction on dimension is not closed as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies birational geometry in positive characteristic. It proves that boundedness is stable under normalizations (Theorem 2.4), establishes a canonical bundle formula for klt threefold Fano fibrations over positive-dimensional base in sufficiently large characteristic (Theorem 3.3), applies this to non-klt center adjunction (Theorems 3.4-3.5) and to preservation of Fano type under contractions (Corollary 3.6), and proves an effective birationality statement for strongly F-regular weak Fano varieties with bounded Gorenstein index (Theorem 4.4), with a corollary for epsilon-strongly F-regular varieties (Corollary 4.5). The main emphasis is Theorem 4.4, which aims at a positive-characteristic analogue of Birkar's effective birationality in the F-regular, bounded-index setting.","tokens_in":23118,"tokens_out":12882,"duration_ms":130671,"significance":"If Theorem 4.4 were established, it would be a significant step toward effective birationality and BAB-type boundedness in positive characteristic. The earlier sections also contain useful tools: Theorem 2.4 is a clean characteristic-free boundedness observation, Theorem 3.3 is a usable canonical bundle formula for threefold Fano fibrations, and Corollaries 1.4-1.5 are likely to be useful for complement constructions. The paper is not circular: it derives results from external theorems such as the three-dimensional MMP, F-adjunction, and boundedness of geometrically integral del Pezzo surfaces. However, the central Theorem 4.4 is currently a conditional statement, because the covering-family engine in Remark 4.1 is asserted rather than proved, and the induction in its proof has a concrete numerical inconsistency.","major_comments":[{"comment":"The assertion that an ample Q-divisor D with vol(D) >= (2d)^d gives a bounded covering family of F-pure centers is the engine of Theorem 4.4, but it is not proved. The preceding paragraph only explains why potential birationality does not automatically become F-potential birationality, and the sentence 'combine the statements above and [HMX14, 7.1] and some tie breaking arguments' is not a proof. In characteristic 0 the analogous statement is a deep theorem of Birkar and depends on ACC for log canonical thresholds; [HMX14] is a characteristic-0 result and does not directly control F-pure centers in positive characteristic. Moreover, the phrase 'bounded covering family' is not defined in the paper, and the uniformity over a sequence of varieties is unclear. Since Theorem 4.4 invokes this statement for a sequence of varieties of arbitrary dimension, the main theorem is currently conditional on an unproved assertion.","section":"Section 4, Remark 4.1"},{"comment":"The induction step contains a numerical inconsistency. With Delta_i ~Q -(n+1)K_{X_i} and A ~Q H = -lnK_{X_i}, Lemma 4.3 produces (1-delta)Delta_i + cA with 0<c<1, whose divisor class is -((1-delta)(n+1)+cln)K_{X_i}. The proof then asserts a replacement Delta'_i ~Q -4lnK_{X_i}. Equating these classes forces c = 4 - (1-delta)(n+1)/(ln), which for l>=1 and large n is at least about 3, contradicting c<1. Thus the stated application of Lemma 4.3 does not produce the claimed boundary, and the dimension-decreasing induction is not closed as written.","section":"Section 4, proof of Theorem 4.4"},{"comment":"Theorem 4.4 is stated for every dimension d, but its proof uses F-potential birationality and Theorem 4.1, whose definition requires X to admit a resolution of singularities. In positive characteristic, resolution of singularities is not known beyond dimension 3. The k=0 branch of the proof therefore does not apply to the stated generality. Either the theorem should be restricted to d<=3, or a resolution-free argument (for example via alterations) must be supplied.","section":"Section 4, Theorem 4.4 and Definition 4.1"},{"comment":"The assertion that 'l is bounded' after F-adjunction is too terse. Here l_i is the smallest integer with vol(-lnK_{X_i}|_{G_i}) > k^k, and a uniform bound requires a lower bound on vol(-K_{X_i}|_{G_i}) for all i. The proof mentions that IK_X|_G is Cartier but does not state the volume lower bound that would follow from this. This is probably repairable, but it should be written out explicitly.","section":"Section 4, proof of Theorem 4.4, boundedness of l"}],"minor_comments":[{"comment":"In the proof, the bound p0 >= 2/min(Phi) is used, but if 0 is in Phi the minimum is 0 and p0 is undefined. The statement should either require Phi to be contained in (0,1] or the proof should discard zero coefficients.","section":"Section 3, Theorem 3.3(1)"},{"comment":"The reference to 'Theorem 3.8' is incorrect: the F-adjunction statement used here is Theorem 3.7. Please correct the cross-reference.","section":"Section 4, proof of Theorem 4.4"},{"comment":"The introduction refers to 'Theorem 3.5, 3.6', while the body numbers the relevant statements as Theorem 3.4 and Theorem 3.5. The numbering should be harmonized.","section":"Section 1, Corollary 1.4"},{"comment":"The statement has a typo: 'a family P of projective varieties k' should read 'over k'. The proof of the positive-characteristic part is dense; in particular, the key openness assertion for geometric normality should be stated explicitly with a precise reference.","section":"Section 2, Theorem 2.4"},{"comment":"The application of [BCRG+17, 4.8] gives that [1/epsilon+1]!K_X is Cartier, which may be divisible by p. Since the hypothesis is that the Gorenstein index is not divisible by p, the proof should explicitly pass to the actual index, which divides this factorial, rather than applying Theorem 4.4 with the factorial itself.","section":"Section 4, Corollary 4.5"},{"comment":"There are small textual issues: 'can be defined defined' should be 'can be defined', and the reference [Hot22] spells the author's name 'Hotchster' instead of 'Hochster'.","section":"Section 2, Remark 2.2 and references"},{"comment":"The notation T^0(Z,D_Z) is used without definition. It appears to be the intersection of images of trace maps over finite covers, but it should be defined explicitly before use.","section":"Section 4, Theorem 4.1"}],"recommendation":"major_revision","confidential_remarks":"The paper contains several useful tools, but the headline result is not ready in its current form. The unproved Remark 4.1 is effectively a new theorem about F-pure centers; if the authors cannot prove it, Theorem 4.4 should be restated as conditional. The numerical gap in the induction is concrete and must be repaired. I would not reject the paper outright, because the earlier sections contain valuable results and the overall strategy is plausible, but Section 4 needs substantial reworking."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know about this one. It has two genuinely new pieces: Theorem 2.4 (boundedness is stable under normalization in char p) and Theorem 3.3 (a canonical bundle formula for threefold Fano fibrations in large characteristic). And it has a headline theorem, Theorem 4.4 (effective birationality for strongly F-regular weak Fano varieties), which is not proved as written. I would not send the effective birationality to the bank yet, but the other results deserve a serious look.\n\nTheorem 2.4 is a clean observation: boundedness of a family of projective varieties in char p survives normalization. The Frobenius-base-change argument and the conductor stabilization are credible, and the statement is useful. Theorem 3.3 is the real substance. For a threefold klt Fano-type fibration to a surface or curve, it gives a canonical bundle formula for p above a bound depending on the coefficients or the Gorenstein index. The proof is compressed and leans on Witaszek, Benozzo, and Bernasconi-Martin, but the strategy—control the geometric generic fiber, kill inseparable base change with a large-p bound, then invoke known CBF results—is coherent. I did not find a fatal gap there, though a referee will want the abbreviated steps expanded.\n\nNow the soft spot. Theorem 4.4 depends on Remark 4.1, an unproved assertion that any ample divisor with volume at least (2d)^d yields a bounded covering family of F-pure centers with the separation property. In characteristic 0 this is a deep theorem; in char p, no such boundedness result for F-pure centers is established, and the cited [HMX14, 7.1] is a characteristic-0 ACC statement. So the engine is missing. On a second pass I also see a numerical problem in the induction: Lemma 4.3 gives a new boundary (1−δ)Δ_i + cA with c<1 and A∼_Q −l n K_{X_i}, whose class is roughly −(n+1+c l n)K_{X_i}; equating that to the stated replacement −4 l n K_{X_i} forces c≈4 for l≥2, contradicting c<1. The induction step is not closed. And the theorem is stated for every dimension, while Theorem 4.1 assumes resolution of singularities, which is only known up to dimension 3 in char p. A missing reference: the proof cites Theorem 3.8 for F-adjunction but the paper only has Theorem 3.7; minor.\n\nNet: the author is honest, knows the literature, and the paper is not circular. The canonical bundle formula part is valuable. But Theorem 4.4, as written, is unsupported. If I were the editor, I would send it to a serious referee—there is enough real content to justify the time—and expect a major revision that either proves the covering-family assertion or restricts the theorem to dimensions where resolution is known and the numeric issues can be repaired. I would not cite the effective birationality in its current form.","headline":"New results on normalization and the threefold canonical bundle formula are worth taking seriously, but the effective birationality theorem is not proved as written: it rests on an unproved bounded-covering statement and the induction step has a numerical gap.","tokens_in":23696,"tokens_out":4677,"would_cite":false,"duration_ms":42388,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14E30","14J45","14G17","13A35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper establishes effective birationality for strongly F-regular weak Fano varieties in characteristic p>0: from dimension and Gorenstein index alone, fixed n,m give vol(-nK_X)>(2d)^d and a birational |-mK_X|.","keywords":["Fano varieties","positive characteristic","effective birationality","F-singularities","strong F-regular","canonical bundle formula","boundedness","BAB conjecture"],"falsifier":"To settle the central claim, one could look for a strongly F-regular weak Fano variety $X$ of dimension $d$ with $IK_X$ Cartier and $p\\nmid I$ such that the minimal $m$ with $|-mK_X|$ birational is not bounded in terms of $d$ and $I$, or such that $\\operatorname{vol}(-nK_X)\\le(2d)^d$ for all $n$ in the range allowed by the proof. More directly, one can test the unproved assertion of Remark 4.1: exhibit an ample $\\mathbb{Q}$-divisor $D$ on some $X$ with $\\operatorname{vol}(D)\\ge(2d)^d$ and two general points $x,y$ for which every $\\Delta\\sim_{\\mathbb{Q}}D$ either fails to be F-pure at $x$ with a unique F-pure center in the claimed finite family, or is F-regular at $y$.","tokens_in":22507,"feed_emoji":"📐","tokens_out":11101,"duration_ms":102938,"temperature":0.7,"pith_summary":"The paper works in characteristic $p>0$, where the standard boundedness tools from characteristic $0$ fail because vanishing theorems break down. Its central result is an effective birationality statement for weak Fano varieties with good F-singularities: if $X$ is strongly F-regular, of fixed dimension $d$, and has Gorenstein index $I$ not divisible by $p$, then there are integers $n$ and $m$, depending only on $d$ and $I$, such that $\\operatorname{vol}(-nK_X)>(2d)^d$ and the linear system $|-mK_X|$ defines a birational map. This gives a positive-characteristic analogue of the effective birationality predicted by the BAB conjecture, restricted to the F-regular setting. Along the way the paper proves that boundedness is stable under normalization, and it establishes a usable canonical bundle formula for threefold Fano-type fibrations over bases of positive dimension in large characteristic, together with adjunction formulas on non-klt centers.","feed_headline":"Uniform birational maps exist for F-regular Fano varieties in char p","feed_subtitle":"For fixed dimension and Gorenstein index, one anti-canonical system |-mK_X| separates points on all such varieties.","key_machinery":"The load-bearing mechanism is the interplay between F-pure centers and F-adjunction. An F-pure center is a closed subvariety where a pair $(X,\\Delta)$ is sharply F-pure but not strongly F-regular; it is, roughly, the minimal locus where the Frobenius splitting degenerates. For such a center $W$, F-adjunction gives a canonical divisor relation $(K_W+\\Delta_W)=(K_X+\\Delta)|_W$, with the Cartier index of $K_W+\\Delta_W$ controlled by that of $K_X+\\Delta$. The paper's engine is Remark 4.1, which asserts that an ample $\\mathbb{Q}$-divisor $D$ with $\\operatorname{vol}(D)\\ge(2d)^d$ yields, for any two general points $x,y$, a boundary $\\Delta\\sim_{\\mathbb{Q}} D$ that is F-pure at $x$ with a unique F-pure center from a fixed finite family covering $X$ and not F-regular at $y$; this lets the argument induct on the center's dimension via Lemma 4.3, which cuts the center down by adding an ample divisor whose restriction to the center has degree $>k^k$.","core_discovery":"The central discovery is Theorem 4.4: for fixed natural numbers $d$ and $I$ with $p\\nmid I$, there exist integers $n$ and $m$ depending only on $d$ and $I$ such that every strongly F-regular weak Fano variety $X$ of dimension $d$ with $IK_X$ Cartier satisfies $\\operatorname{vol}(-nK_X)>(2d)^d$ and $|-mK_X|$ gives a birational map. The proof produces, for each such $X$, a bounded covering family of F-pure centers and then runs an induction on the dimension of these centers: an ample-divisor cutting lemma shrinks the F-pure center while preserving sharp F-purity at one general point and non-F-regularity at a second, until the center is a point, at which point an F-potentially birational divisor theorem forces the relevant anti-canonical linear system to separate points. A corollary replaces the Gorenstein-index hypothesis by a lower bound on the F-signature, giving $m$ depending only on the dimension and the F-signature bound.","pith_inferences":["Because the proof only needs sharp F-purity and F-adjunction, a natural extension would be to replace strong F-regularity by klt or lc with a uniform Cartier index in large characteristic; if Remark 4.1 can be proved in that setting, effective birationality would follow for $\\epsilon$-lc weak Fano varieties, closely matching the BAB prediction.","The dependence of the canonical bundle formula on a characteristic threshold $p_0=p_0(\\Phi)$ suggests that any counterexample to effective birationality in char $p$ would have to live in small characteristic relative to the boundary coefficients or the Gorenstein index; a systematic search over small $p$ for Fano threefolds with unbounded $m$ would test this.","The normalization-stability theorem removes a known obstruction to reducing boundedness questions to normal varieties in char $p$, so one could combine it with the canonical bundle formula to attempt a full BAB-type boundedness statement for $\\epsilon$-lc threefolds by first normalizing and then applying the induction on F-pure centers; the author does not carry out this step.","The volume threshold $(2d)^d$ and the cutting condition $H^k\\cdot Z>k^k$ appear to be the exact analogues of the characteristic-0 thresholds, suggesting the same effective-birationality constants could hold uniformly in all characteristics once the missing assertion is supplied."],"forward_implications":["For fixed $d$ and $I$ with $p\\nmid I$, all strongly F-regular weak Fano $d$-folds with $IK_X$ Cartier share a single anti-canonical linear system $|-mK_X|$ that is birational, so their images in projective space have bounded degree.","Corollary 4.5: if instead of fixing the Gorenstein index one assumes $X$ is $\\epsilon$-strongly F-regular, the same birationality holds with $m$ depending only on $d$ and $\\epsilon$, because a uniform multiple of $K_X$ is Cartier.","Since volumes and Gorenstein indices are bounded for bounded families, the theorem gives a new boundedness statement: the set of strongly F-regular weak Fano varieties with fixed dimension and Gorenstein index not divisible by $p$ is birationally bounded.","The canonical bundle formula of Theorem 3.3 implies that a contraction of a threefold of Fano type is again of Fano type in large characteristic when the general fibers are normal and a uniform multiple of the relative canonical divisor is Cartier.","Theorem 2.4 shows boundedness is stable under normalization in arbitrary characteristic, so future boundedness arguments may assume varieties are normal without loss."],"supporting_citations":[{"why":"Supplies the characteristic-0 effective birationality framework and boundedness criterion (Lemma 2.3) that the positive-characteristic argument adapts.","marker":"[Bir19]"},{"why":"Provides the F-adjunction formula used to transfer F-purity and Cartier-index control from X to its F-pure centers.","marker":"[Sch09]"},{"why":"Gives the F-version of Nadel vanishing and test-ideal lifting of sections used to show the linear system separates points.","marker":"[BST11]"},{"why":"Provides openness and generic freeness of F-singularities in families, used in the cutting lemma to choose divisors avoiding F-pure centers.","marker":"[PSZ13]"},{"why":"Bertini-type theorems for F-singularities, used to preserve sharp F-purity when adding divisors in Lemmas 4.2 and 4.3.","marker":"[SZ13]"},{"why":"Its ACC result for log canonical thresholds is invoked in Remark 4.1 to produce the bounded covering family of F-pure centers.","marker":"[HMX14]"},{"why":"Shows that a uniform lower bound on F-signature makes a uniform multiple of K_X Cartier, which converts Corollary 4.5 into Theorem 4.4.","marker":"[BCRG+17]"},{"why":"Supplies the tie-breaking strategy used in Theorem 4.1 to construct boundaries with controlled F-pure centers.","marker":"[Wan22]"}],"fun_headline_variants":["Uniform anti-canonical birational maps for F-regular weak Fanos","One m works for all: effective birationality for F-regular Fanos in char p","Bounded Fanos, uniform birational anti-canonical systems in char p","F-signature bound gives birational |-mK| for weak Fano varieties","Effective birationality: F-regular weak Fanos have uniform separating maps"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on an unproved assertion, Remark 4.1, that from any ample $\\mathbb{Q}$-divisor $D$ with $\\operatorname{vol}(D)\\ge(2d)^d$ one can always find, for any two general points $x$ and $y$, a boundary $\\Delta\\sim_{\\mathbb{Q}} D$ whose only F-pure center containing $x$ is a subvariety from a fixed finite family covering $X$, while the pair is not F-regular at $y$; if that assertion fails, the effective birationality conclusion is unsupported.","fun_headline_variants_meta":{"raw":{"variants":["Uniform anti-canonical birational maps for F-regular weak Fanos","One m works for all: effective birationality for F-regular Fanos in char p","Bounded Fanos, uniform birational anti-canonical systems in char p","F-signature bound gives birational |-mK| for weak Fano varieties","Effective birationality: F-regular weak Fanos have uniform separating maps"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00108,"raw_usage":{"total_tokens":4467,"prompt_tokens":844,"completion_tokens":3623,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":460,"completion_tokens_details":{"reasoning_tokens":3516}},"tokens_in":460,"tokens_out":3623,"duration_ms":25969,"temperature":1.0,"reasoning_tokens":3516,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:35:09.721577+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"To settle the central claim, one could look for a strongly F-regular weak Fano variety $X$ of dimension $d$ with $IK_X$ Cartier and $p\\nmid I$ such that the minimal $m$ with $|-mK_X|$ birational is not bounded in terms of $d$ and $I$, or such that $\\operatorname{vol}(-nK_X)\\le(2d)^d$ for all $n$ in the range allowed by the proof. More directly, one can test the unproved assertion of Remark 4.1: exhibit an ample $\\mathbb{Q}$-divisor $D$ on some $X$ with $\\operatorname{vol}(D)\\ge(2d)^d$ and two general points $x,y$ for which every $\\Delta\\sim_{\\mathbb{Q}}D$ either fails to be F-pure at $x$ with a unique F-pure center in the claimed finite family, or is F-regular at $y$.","supporting_citations":[],"review_version":1}