Pith. sign in

REVIEW 2 major objections 4 minor 5 references

Splitting of Vector Bundles on Toric Varieties

T0 review · 2 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper establishes that on a smooth projective toric variety, a vector bundle whose cohomology matches an ordered sum of line bundles in every twist must itself be that sum.

desk verdict A genuinely new and plausible splitting criterion for vector bundles on smooth projective toric varieties, but the proof's key technical lemma has a load-bearing identification that is asserted rather than demonstrated, so the paper needs a fix before it is fully convincing. read the letter →

arxiv 2412.19793 v1 pith:ONFXS6QS submitted 2024-12-27 math.AG math.AC

Mahrud Sayrafi This is my paper · ORCID
classification math.AGmath.AC MSC 13D0214F0614F08
keywords vectorbundlestoricvarietiessplittingcriteriacohomologyvanishingresolutionofthediagonallineintegraltransformspectralsequence
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves a splitting criterion for vector bundles on arbitrary smooth projective toric varieties. If a vector bundle $E$ has the same cohomology $H^q(X, E \otimes L)$ as a fixed direct sum of line bundles $E' = \bigoplus_i \mathcal{O}(D_i)^{r_i}$ for every degree $q \geq 0$ and every twist $L \in \operatorname{Pic} X$, and the divisors $D_{i+1} - D_i$ are ample for consecutive summands, then $E$ is isomorphic to $E'$. In other words, the full cohomology table of a bundle, once it matches an increasingly ordered line-bundle sum, completely determines the bundle. This generalizes the classical splitting criterion for projective spaces to all smooth projective toric varieties. The proof builds a spectral sequence from a resolution of the diagonal by line bundles and uses it to peel off one line-bundle summand at a time.

What carries the argument

The load-bearing object is the resolution of the diagonal by line bundles constructed in [HHL24], used as the kernel of an integral transform. Its terms are box products $G \boxtimes L$ of a locally free sheaf with a line bundle from a canonical collection on $X$. The key estimate, Lemma 3.1(III), says that any summand $\mathcal{O}(E') \boxtimes \mathcal{O}(E)$ appearing in the $p$-th term has $p \leq \dim P_{-E}$, where $P_{-E}$ is the polytope of global sections of the line bundle $\mathcal{O}(-E)$. Combined with the standard toric vanishing theorem, this makes the resolution cohomologically supported in the ample cone: for every summand of its $p$-th term, $H^q(X, \mathcal{O}(E-D))=0$ for $q<p$ and every ample divisor $D$. The spectral sequence of the integral transform then has vanishing higher diagonals, so the zeroth term $\mathcal{O}^{r_n}$ is a direct summand of $E$, and induction removes the remaining summands.

What would settle it

Construct the diagonal resolution explicitly for a small smooth projective toric surface, such as $\mathbb{P}^1 \times \mathbb{P}^1$, and search for a summand $\mathcal{O}(E') \boxtimes \mathcal{O}(E)$ in the second or higher term with $\dim P_{-E} < p$; such a summand would violate Lemma 3.1(III). Alternatively, produce a vector bundle $E$ and an ordered line-bundle sum $E'$ with equal cohomology in all twists but $E \not\cong E'$, refuting Theorem 1 directly.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1: on a smooth projective toric variety $X$, if $E$ is a vector bundle and $E'=\bigoplus_{i=1}^n \mathcal{O}(D_i)^{r_i}$ is a direct sum of line bundles such that $D_{i+1}-D_i$ is ample for every $0<i<n$, then equality of all cohomology groups $H^q(X,E\otimes L)=H^q(X,E'\otimes L)$ for $q\ge 0$ and all $L\in\operatorname{Pic} X$ forces $E\cong E'$. The proof first establishes a general recipe: whenever a variety has a locally free resolution of the diagonal that is cohomologically supported in a cone $A$ and the line bundles are ordered by consecutive differences in $A$, cohomology equality implies splitting. For toric varieties the needed support statement is proved for $A=\operatorname{Ample}(X)$: every line-bundle summand in the $p$-th term of the diagonal resolution satisfies $p\le \dim P_{-E}$, the dimension of the section polytope of $-E$, and the toric vanishing theorem then kills all higher spectral-sequence terms. The theorem is obtained by induction, peeling the highest line-bundle summand off $E$ and repeating.

Load-bearing premise

The load-bearing premise is that in the diagonal resolution from [HHL24], every line-bundle summand in the $p$-th term comes from a $p$-dimensional stratum whose label matches the corresponding stratum in the resolution of a point, which yields the bound $p \leq \dim P_{-E}$; if this correspondence fails, the resolution need not be cohomologically supported in the ample cone and the induction cannot start.

Editorial extensions

If this is right

  • A vector bundle on any smooth projective toric variety that has the same cohomology as an ordered line-bundle sum in every twist is actually isomorphic to that sum, so cohomology tables determine such bundles completely.
  • The classical splitting criterion for projective space appears as the special case in which the Picard group is cyclic and every line-bundle sum can be ordered by degree; the theorem recovers it without an inductive restriction argument.
  • The theorem gives a new obstruction to indecomposability: a bundle that matches an ordered line-bundle sum cohomologically must be decomposable.
  • The recipe in Section 2 turns any resolution of the diagonal supported in a cone $A$ into a splitting criterion for line bundles ordered inside $A$, so new resolutions immediately yield new criteria on other varieties.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Implicit in the proof: the ordering hypothesis is used only to locate the top summand inside a cone; if a toric variety admitted a diagonal resolution supported in a larger cone, the same argument would give a splitting criterion with weaker hypotheses.
  • Testable extension: on a smooth projective toric surface of Picard rank two, one could search computationally for a vector bundle whose cohomology matches a non-ordered line-bundle sum; finding one would show the ample-ordering hypothesis is necessary.
  • The dimension bound $p \leq \dim P_{-E}$ can be checked directly for small examples by writing out the diagonal resolution and comparing the stratification labels; any violation would pinpoint where the support argument fails.
  • If the theorem is correct, multigraded cohomology tables on toric varieties are rigid enough to certify splitting, which suggests a practical route to automatic splitting certificates in computer algebra systems.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper proves a Horrocks-type splitting criterion for vector bundles on smooth projective toric varieties: if a vector bundle E has the same cohomology as a sum E' of line bundles whose successive difference divisors are ample, then E is isomorphic to E' (Theorem 1). The proof is organized around a general Fourier-Mukai recipe in Section 2, which reduces the splitting statement to the existence of a diagonal resolution that is `cohomologically supported' in the ample cone. The main technical input, Proposition 3.3, claims that the Hanlon-Hicks-Lazarev resolution of the diagonal on a toric variety has this support property, using the dimension bound in Lemma 3.1(III).

Significance. If the proof is completed, the result would be a substantial generalization of Horrocks' criterion to arbitrary smooth projective toric varieties, under an ordered-ampleness hypothesis analogous to the Eisenbud-Erman-Schreyer criterion for products of projective spaces. The general recipe in Proposition 2.4 is a useful organizational framework, and the intended use of the Hanlon-Hicks-Lazarev diagonal resolution is natural and promising. The paper is concise and clearly written, but the main theorem currently depends on a toric stratification assertion that is not fully established.

major comments (2)
  1. [§3, Lemma 3.1(III)] The proof of the bound p ≤ dim P_{-E} is not complete. The paragraph in Lemma 3.1(III) asserts that the Bondal stratification on the kernel L_R agrees with the Bondal stratification on M_{X,R}/M_X used for the resolution of a point, and that a stratum labeled O(E')⊠O(E) in the diagonal resolution receives the label O(E) in the point resolution; this label projection is the step that makes [FH22, Lem. 5.6] applicable. The cited [HHL24, Exa. 3.13] only establishes that the kernel inherits a stratification, not the equality of labels, and Remark 3.2 explicitly notes a subtle difference between the HHL and FH22 stratifications. Since Proposition 3.3 needs p ≤ dim P_{-E} to convert Batyrev-Borisov vanishing into H^q(O(E-D))=0 for q<p, this is load-bearing for Theorem 1. Please supply a detailed derivation or a precise reference for the label correspondence, including the claim in Remark 3.2 that the union of HHL strata with a given label is the unique FH22 stratum.
  2. [§2, Proposition 2.4] In the proof of Proposition 2.4, the sentence 'E0,0_1(E) = E0,0_1(E') = O^{r_n}_X' is asserted without justification. The term E0,0_1(E') is a sum over the K^0 summands G⊠O(E) of G ⊗ H^0(X, E'⊗O(E)); it is not automatic that only the O_X⊠O_X summand contributes. This equality is used to extract the first summand in the induction, so the proof needs an explicit argument. For instance, the author should justify that, after twisting, all other possible contributions to H^0 vanish, or add a hypothesis on K^0 (such as K^0 having O_X⊠O_X as its only contributing summand) to Proposition 2.4.
minor comments (4)
  1. [§3, Proposition 3.3] The divisor D+(1-ǫ)~E-E is a Q-divisor; the appeal to Batyrev-Borisov vanishing should state the version for round-ups of Q-divisors, since [CLS11, Thm. 9.3.5(b)] is usually stated for Cartier divisors.
  2. [§2, diagram (2.1)] The diagram in (2.1) refers to 'dotted diagonals with k > 0' without defining k; please define k (for example, k = p - q or the total degree) so that Definition 2.1 is unambiguous.
  3. [References] The reference [Tho00] appears in the bibliography but is not cited in the body of the text.
  4. [Title/Abstract] There are minor typographical issues in the rendered text, such as 'V arieties' in the title and 'vari eties' in the abstract.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the theorem's conclusion is not an input, the recipe and resolutions come from independent sources, and the proof's load-bearing Lemma 3.1(III) is a proof gap rather than a circular reduction.

full rationale

The derivation is self-contained against external benchmarks rather than circular. The theorem's hypothesis is cohomology equality between E and a line-bundle sum E'; the conclusion E is isomorphic to E' is not used in the proof, so no self-definitional step appears. Proposition 2.4 is proved in the paper, with citations to [BS24] and [EES15] only for analogous arguments; those citations are not the load-bearing mechanism because the spectral-sequence argument is written out. The toric input comes from the independent HHL24 resolution and FH22 stratification lemmas, not from a self-citation chain. There is no fitted parameter renamed as a prediction and no known result presented under a new name. The only significant weakness is Lemma 3.1(III), whose proof asserts that the label of a p-dimensional stratum in the diagonal resolution is the same as the label in the resolution of a point on X; this identification is stated without a full derivation and is load-bearing for Proposition 3.3 and hence Theorem 1. But that is an omitted proof or potential correctness gap, not circularity: the asserted correspondence is not equivalent to the theorem's conclusion, and no equation of the paper reduces the target to an input. For that reason the circularity score is 0.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new entities or free parameters. It imports the HHL resolution of the diagonal and the Thomsen collection, then applies a general Fourier-Mukai recipe. The central new content is the proof that this resolution is cohomologically supported in the ample cone, which relies on the listed background results.

assumptions (6)
  • domain assumption Hanlon-Hicks-Lazarev resolution of the diagonal for a smooth projective toric variety X exists, with terms in the Thomsen collection and satisfying Lemma 3.1(I)-(III).
    Used throughout Section 3. Imported from [HHL24]; the proof of Lemma 3.1 only cites properties of this construction, notably around eq. (18) in [HHL24].
  • standard math Batyrev-Borisov vanishing: for a nef Cartier divisor D on a complete toric variety, H^q(X,O(D))=0 for q < dim P_D.
    Invoked in Proposition 3.3 via [CLS11, Thm. 9.3.5(b)] to obtain vanishing in the cohomological support proof.
  • standard math The Fourier-Mukai transform with kernel a locally free resolution of the diagonal is isomorphic to the identity functor on the derived category.
    Section 2, the starting point for the spectral sequence argument. Standard derived category fact.
  • standard math For a line bundle O(E) in the Thomsen collection, the union of strata labeled E in Bondal's stratification is (P_{-E} minus the union of boundary polytopes P_{-E-D_rho}) modulo the lattice, and has dimension dim P_{-E}.
    Used in Lemma 3.1(III) via [FH22, Lem. 5.6] to relate the dimension of a stratum to the dimension of the section polytope.
  • standard math Summands of Kp in the HHL construction correspond to labels on p-dimensional strata of Bondal's stratification.
    Used in Lemma 3.1(III) from [HHL24, eq. (18)] and its surrounding construction.
  • standard math Thomsen collection line bundles satisfy: -E is an effective Cartier divisor and O(E) is a summand of a high toric Frobenius pushforward of O_X.
    Used to establish Lemma 3.1(I) and (II), from [HHL24, §5].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Splitting of Vector Bundles on Toric Varieties." pith.science (2026). https://pith.science/paper/ONFXS6QS

@misc{pith2026241219793,
  author       = {Pith},
  title        = {Pith review of: Splitting of Vector Bundles on Toric Varieties},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ONFXS6QS}},
  note         = {Machine review of arXiv:2412.19793}
}
read the original abstract

We prove a Horrocks-type splitting criterion for arbitrary smooth projective toric varieties under an additional hypothesis similar to the case of products of projective spaces by Eisenbud--Erman--Schreyer.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

5 extracted references · 5 canonical work pages

  1. [1]

    Acad´ emie des Sciences

    [AM11] Marian Aprodu and Marius Marchitan, A note on vector bundles on Hirzebruch surfaces , Comptes Rendus Math´ ematique. Acad´ emie des Sciences. Paris349 (2011), no. 11-12, 687–690. MR2817392 ↑1 [And23] Reginald Anderson, A Resolution of the Diagonal for Smooth Toric Varieties , arXiv,

    work page 2011

  2. [6]

    ↑1, 4 5 [Bei78] A. A. Beilinson, Coherent sheaves on Pn and problems of linear algebra , Rossiiskaya Akademiya Nauk. Funktsionalnyi Analiz i ego Prilozheniya 12 (July 1978), no. 3, 214–216. ↑3 [BES20] Christine Berkesch, Daniel Erman, and Gregory G. Smith, Virtual resolutions for a product of projective spaces, Algebraic Geometry 7 (2020), no. 4, 460–481....

    work page 2020

  3. [13]

    4, 313–323

    ↑1 [FM11] Mihai Fulger and Marius Marchitan, Some splitting criteria on Hirzebruch surfaces , Bulletin math´ ematique de la Soci´ et´ e des Sciences Math´ ematiques de Roumanie 54 (102) (2011), no. 4, 313–323. ↑1 [Har74] Robin Hartshorne, Varieties of small codimension in projective space , Bulletin of the American Mathematical Society 80 (1974), 1017–103...

    work page 2011

  4. [2011]

    Homotopy Path Algebras

    ↑5 [CM05] L. Costa and R. M. Mir´ o-Roig,Cohomological characterization of vector bundles on multi projective spaces, Journal of Algebra 294 (2005), no. 1, 73–96. MR2179715 ↑1 [EES15] David Eisenbud, Daniel Erman, and Frank-Olaf Schreyer, Tate resolutions for products of projec- tive spaces, Acta Mathematica Vietnamica 40 (2015), no. 1, 5–36. ↑1, 3 [EFS03...

    work page Pith review arXiv 2005

  5. [2023]

    ↑2 [BE24] Michael K

    ↑1 [BCHS21] Juliette Bruce, Lauren Cranton Heller, and Mahrud Sayra fi, Characterizing Multigraded Regularity and Virtual Resolutions on Products of Projective Spaces (2021). ↑2 [BE24] Michael K. Brown and Daniel Erman, A short proof of the Hanlon-Hicks-Lazarev theorem , Forum of Mathematics. Sigma 12 (2024), Paper No. e56,

    work page 2021

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.