REVIEW 2 major objections 5 minor 9 references
Computing higher direct images in Macaulay2
T0 review · 2 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read This article presents a computer-algebra package that computes higher direct images of line bundles under surjective toric morphisms, replacing a method that only worked for products of projective spaces.
desk verdict Useful Macaulay2 package that genuinely extends computable higher direct images from products of projective spaces to toric fibrations, but the abstract overstates the input class and leaves the key 'mild assumption' unspecified. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing machinery is the reduced Čech complex of the companion paper [8]—a trimmed Čech complex that computes the same higher direct images without redundant sheaves—together with the character decomposition of $R^i f_* L$ into eigensheaves indexed by characters of the kernel torus $T_K$. For Frobenius pushforwards, the machinery is the line-bundle splitting theorem [9], implemented through the combinatorial description of [1], extended by functoriality and free resolutions so that arbitrary coherent sheaves can be pushed forward.
What would settle it
Take a toric fibration with a hand-computable answer, such as a projective bundle over a complete toric threefold pushed forward with the structure sheaf, and compare the package's $R^i f_* \mathcal{O}_X$ to the classical projective-bundle formula; any mismatch in rank, twist, or filtration would refute the claimed generality. A second test would be to run the method on a map violating the unspecified mild assumption and check whether the output contradicts known local-cohomology vanishings.
Extended reading notes
Core claim
The central claim is that the ToricHigherDirectImages package faithfully implements the companion algorithm [8], which computes $R^i f_* L$ for a line bundle $L$ on a smooth complete toric variety $X$ under a surjective toric morphism $f : X \to Y$ to another smooth complete toric variety. The computation is organized by the kernel torus $T_K = \ker f|_{T_X}$: each $R^i f_* L$ splits into eigensheaves indexed by characters of $T_K$, and the package's computeEigencharacters method returns the divisors on $X$ and $Y$ needed to build those summands. The package also implements a separate Frobenius pushforward routine: for the $p$-th toric Frobenius map, the splitting theorem [9] describes how the pushforward of a line bundle splits into line bundles, and the algorithm from [1] makes this explicit; functoriality lets the user push forward any sheaf by resolving it by line bundles and pushing forward the complex.
Load-bearing premise
The whole package rests on the companion algorithm [8] being valid for every surjective toric morphism between complete toric varieties that satisfies the 'mild assumption'—an assumption this article never actually states.
Editorial extensions
If this is right
- Any complete toric fibration satisfying the mild assumption becomes computationally accessible, not just maps between products of projective spaces.
- The character decomposition output by computeEigencharacters makes the eigensheaf structure of each higher direct image explicit, which is exactly the data needed for further cohomological calculations.
- Frobenius pushforwards of arbitrary coherent sheaves can now be computed, giving a concrete handle on objects of recent interest in birational geometry and derived categories.
- Coherent duality can be checked term by term for line bundles whose higher direct images live in a single degree, as demonstrated on a Fano sixfold fibration over projective four-space.
- The package is distributed with the computer algebra system's version 1.25.05, so the algorithms are available to other users directly.
Reading between the lines
- If the companion algorithm's coverage extends beyond the smooth case, the same character-indexed decomposition should work for singular complete toric varieties; pushing a few known singular examples through the package would be a direct test.
- The Frobenius routine is characteristic-independent by construction, so it could be used to probe phenomena special to positive characteristic, such as p-torsion in pushforwards, without needing a finite-field implementation.
- A full complex-level implementation of shriek functors would turn the term-by-term duality checks into a derived-category tool; the examples here suggest the package already computes all the pieces.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper presents the Macaulay2 package ToricHigherDirectImages, whose central feature is a method for computing (higher) direct images of line bundles under toric morphisms. The package implements algorithms from a companion paper by Roth and Zotine [8] for toric fibrations, and additionally provides a method for computing pushforwards of arbitrary sheaves under toric Frobenius maps, based on a theorem of Achinger [1]. The paper demonstrates the package on several examples, including a projective bundle over P^1, a blowup of P^2, a fibration from a smooth Fano sixfold, a Grothendieck duality verification, and a Frobenius pushforward of the cotangent sheaf on a Hirzebruch surface.
Significance. If the advertised functionality is correct, this package is a useful computational tool that extends the previous state of the art, which only handled maps between products of projective spaces, to arbitrary complete toric varieties satisfying a mild assumption. The examples and included assertions suggest that the implementation works for the exhibited cases, and the package's design around fans and Cox rings is natural for toric geometry. However, the central correctness claim is conditional on an unspecified 'mild assumption' and on the companion paper's algorithm; the paper does not provide an independent proof or a pinned, directly citable version of the code. The Frobenius pushforward feature is also asserted without a detailed correctness argument. Consequently, the significance is real but conditional on the missing hypotheses being stated and the implementation being reproducible.
major comments (2)
- [Abstract and body paragraph beginning 'Our setting is...'] The abstract advertises the central feature as computing higher direct images of line bundles under surjective toric morphisms, but the algorithm described in the body is only justified for toric fibrations, i.e. maps satisfying f_*O_X = O_Y. These two classes are not equivalent: the toric Frobenius map F_p: P^1 -> P^1 is surjective, yet f_*O_{P^1} is a direct sum of line bundles O(-i) for i=0,...,p-1, not O_{P^1}. The paper never states the 'mild assumption' from the abstract, nor does it say whether the package verifies the fibration condition before applying the algorithms from [8]. This is load-bearing because a user cannot know whether the method is applicable to a given ToricMap, and the package's behavior on non-fibrations is unspecified. Please state the precise hypothesis and either restrict the advertised input class to toric fibrations or explain how the package checks the condition and what it does when the condition fails.
- [Toric Frobenius Pushforwards] The claim that the package 'computes the pushforward of any module' under toric Frobenius maps is not supported by the manuscript: no algorithm or correctness statement is given beyond a reference to [1, Theorem 2] and a single example. Since the method is described as 'independent', the paper should at least sketch how the splitting from [9] is combined with a free resolution, and should state any assumptions on the toric variety (e.g., smoothness, completeness). Without this, the advertised functionality of the Frobenius method is not verifiable from the paper.
minor comments (5)
- [Abstract and transcript header] The abstract says the package is included in version 1.25.05 of Macaulay2, but the transcript header reads 'Macaulay2, version 1.24.05'. Please clarify which version was used for the computations and correct the mismatch.
- [Grothendieck duality example] In the derivation of f^!G = K_Z[dim Z], the composition is written 'f ◦ g' where it should be 'g ◦ f' (the structure map Z -> pt is g ∘ f). The intended argument is clear, but the typo should be fixed.
- [Throughout] The paper does not provide a URL or stable repository for the package; adding one would facilitate reproducibility and help readers verify the claimed functionality.
- [Notation in examples] The notation f_*^0, f_*^1, etc. is nonstandard; please note explicitly that these denote the higher direct image functors R^0 f_*, R^1 f_*, and similarly for the package methods.
- [Example divisor inputs] Divisors are given as lists of integers (e.g., D = {0,0,0,0,-2,0,-2}); it would help to state explicitly that these are coefficients on the torus-invariant divisors, since this is not obvious to a reader unfamiliar with the package.
Circularity Check
No significant circularity: the package computations are validated against independent classical results, though the abstract's 'surjective toric morphisms' claim is broader than the body's 'toric fibration' setting.
full rationale
This is a software highlight, not a derivation paper. It contains no fitted parameters, no equation that is defined in terms of its own output, and no known result merely relabeled in new coordinates. The main method implements the algorithms of the companion paper [8] by Roth and the author, but the included examples are checked against independent external mathematics: the projective bundle splitting on F1, the ideal sheaf of a blown-up point, the reproduced example [8, Example 1.8], and the term-by-term Grothendieck-Serre duality identity (1) on a Fano sixfold. The Frobenius pushforward routine is based on Achinger's theorem [1] and Thomsen's splitting theorem [9], both external. Thus the central computations do not reduce by construction to the package's own outputs. The only notable issue is a scoping gap: the abstract advertises computation 'under surjective toric morphisms', while the body defines the working setting as a toric fibration with f_*O_X = O_Y and the promised 'mild assumption' is never stated; toric Frobenius maps are surjective but not fibrations. That is a correctness and documentation risk, not an instance of circularity, because no claim is justified solely by restating its input or by an unverified self-citation chain.
Assumptions & free parameters
assumptions (4)
- standard math Toric varieties and morphisms are encoded by fans and linear maps, and coherent sheaves by graded modules over the Cox ring.
- domain assumption The algorithm of [8] correctly computes higher direct images for surjective toric morphisms between complete toric varieties satisfying a mild assumption.
- domain assumption Achinger's Theorem 2 [1] provides the splitting of Frobenius pushforwards, and the paper's adapted algorithm is correct.
- standard math For proper morphisms, Grothendieck duality as stated in [6] holds and is applied term-by-term in the example.
Cite this review
Pith. "Pith review of Computing higher direct images in Macaulay2." pith.science (2026). https://pith.science/paper/AD3E6PKV
@misc{pith2026250522835,
author = {Pith},
title = {Pith review of: Computing higher direct images in Macaulay2},
year = {2026},
howpublished = {\url{https://pith.science/paper/AD3E6PKV}},
note = {Machine review of arXiv:2505.22835}
}
read the original abstract
This article highlights the ToricHigherDirectImages package in Macaulay2. The central feature is a method for computing (higher) direct images of line bundles under surjective toric morphisms.
Reference graph
Works this paper leans on
-
[8]
Reduced \v{C}ech complexes and computing higher direct images under toric maps
Mike Roth and Sasha Zotine. Reduced ˇCech complexes and computing higher direct images under toric maps. arXiv, 2025. Available at arXiv:2504.12903
work page Pith review arXiv 2025
-
[1]
A characterization of toric varieties in characteristic p
Piotr Achinger. A characterization of toric varieties in characteristic p. Int. Math. Res. Not. IMRN , 16:6879–6892, 2015
work page 2015
-
[9]
Frobenius direct images of line bundles on toric varieties
Jesper Funch Thomsen. Frobenius direct images of line bundles on toric varieties. J. Algebra, 226(2):865– 874, 2000. Sasha Zotine, Department of Mathematics & Statistics, McMaster University, Hamilton, On- tario, Canada, L8S 4K1 E-mail: zotinea@mcmaster.ca 5
work page 2000
-
[2]
Ballard, Christine Berkesch, Michael K
Matthew R. Ballard, Christine Berkesch, Michael K. Brown, Lauren Cranton Heller, Daniel Erman, David Favero, Sheel Ganatra, Andrew Hanlon, and Jesse Huang. King’s conjecture and birational geometry. arXiv, 2024. Available at arXiv:2501.00130
-
[3]
Derived categories of toric varieties
Alexey Bondal. Derived categories of toric varieties. Oberwolfach Report 5/2006, No. 1, 284-286, 2006
work page 2006
-
[4]
David A. Cox, John B. Little, and Henry K. Schenck. Toric Varieties. Graduate studies in mathematics. American Mathematical Society, 2011
work page 2011
-
[5]
Relative Beilinson monad and direct image for families of coherent sheaves
David Eisenbud and Frank-Olaf Schreyer. Relative Beilinson monad and direct image for families of coherent sheaves. Trans. Amer. Math. Soc. , 360(10):5367–5396, 2008
work page 2008
-
[6]
Residues and duality, volume No
Robin Hartshorne. Residues and duality, volume No. 20 ofLecture Notes in Mathematics. Springer-Verlag, Berlin-New York, 1966. Lecture notes of a seminar on the work of A. Grothendieck, given at Harvard 1963/64, With an appendix by P. Deligne
work page 1966
Show all 9 references
-
[7]
Algebraic Geometry
Robin Hartshorne. Algebraic Geometry. Graduate Texts in Mathematics. Springer, 1977
1977
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.