Equivariant geometry of cubic threefolds with non-isolated singularities

Ivan Cheltsov; Lisa Marquand; Yuri Tschinkel; Zhijia Zhang

REVIEW 3 major objections 4 minor 19 references

Equivariant geometry of cubic threefolds with non-isolated singularities

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper proves that every finite group action on a cubic threefold with singular locus a line, plane, or chordal quartic is linearizable, while conic-singular cubics mostly admit non-linearizable actions.

desk verdict Main results are plausible and one is new, but the proofs of the chordal cubic and plane-singularity theorems rest on a false generic-freeness assertion and need repair. read the letter →

arxiv 2505.03986 v1 pith:IJ776MI5 submitted 2025-05-06 math.AG

Ivan Cheltsov , Lisa Marquand , Yuri Tschinkel , Zhijia Zhang This is my paper · ORCID

classification math.AG MSC 14E0714J3014J5014L30

keywords cubicthreefoldnon-isolatedsingularitieslinearizableactionsG-unirationalityfinitegroupchordalno-namelemmaequivariantbirationalgeometry

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

The paper studies finite group actions on cubic threefolds whose singular locus has positive dimension but is not an isolated point. Such threefolds come in four families: singular along a line, along a conic, along a plane, or along a rational normal quartic (the chordal cubic). The paper proves that every finite group action is linearizable — equivariantly birational to a linear action on projective three-space — in the line, plane, and chordal cases, and that in the conic case most finite actions are not linearizable, with even dihedral groups providing explicit obstructions. It also proves that every finite action on all four families is $G$-unirational, meaning a linear representation of the group dominates the threefold equivariantly. The contrast with the isolated-singularity story, where rational K-unstable cubics had linearizable actions, shows that non-isolated singularities introduce genuinely new non-linearizable equivariant behavior.

What carries the argument

The case analysis rests on the classification of non-cone cubic threefolds with non-isolated singularities, which reduces to four normal forms according to the singular locus: a line, a conic, a plane, or a rational normal quartic. For each form the paper identifies the invariant subvariety that carries the argument — the singular line, the plane spanned by the conic, the singular plane itself, or the secant-variety structure of the chordal cubic. The workhorse is the no-name lemma, a theorem saying that a generically free linear group action on a vector bundle over a projective space is equivariantly birational to a trivial bundle; this turns projective bundles over linearizable bases into linearizable varieties. A second lemma converts $G$-unirationality into a statement about twists over nonclosed fields, so an invariant unirational subvariety with rational points forces the whole threefold to be $G$-unirational. In the conic case, unprojection — rewriting the cubic by introducing one new coordinate to eliminate an invariant plane — produces intersections of two quadrics in which the non-linearizable dihedral actions can be recognized by the elliptic curve they stabilize.

What would settle it

Exhibit a finite subgroup of the automorphism group of one of the classified conic-singular cubics that fixes a singular point and whose action is not linearizable; this would refute the asserted reduction and force the theorems to be restricted to actions with no fixed singular points.

Watch

Extended reading notes

Core claim

The central claim, stated as Theorems 3.1, 5.1, 6.2, 4.3, and 4.5, is a near-complete dichotomy for finite group actions. If the singular locus is a line, a plane, or the chordal quartic, every finite subgroup of the automorphism group acts linearizably: after an equivariant birational transformation the threefold becomes a projective bundle over a linearizable base, and the no-name lemma trivializes the bundle. If the singular locus is a conic, the story splits: the paper constructs models as intersections of two quadrics and shows that, for even-order dihedral subgroups generated by a specific rotation and swap, the action is not linearizable, because a Klein four subgroup fixes a cubic surface while a residual involution fixes a smooth elliptic curve. At the same time, every finite action on every one of the four families is $G$-unirational, since the singular plane or line supplies a $G$-invariant unirational subvariety whose twisted forms have rational points. The chordal cubic shows a further subtlety: each finite subgroup linearizes, but the full $\mathrm{PGL}_2$ automorphism group does not.

Load-bearing premise

The paper assumes without proof that any finite group action fixing a singular point of the cubic is linearizable; the reduction to the fixed-point-free case for all four singularity types depends on this claim.

Editorial extensions

If this is right

Every finite group action on a cubic threefold singular along a line is linearizable (Theorem 3.1).
Every finite group action on a cubic threefold singular along a plane is linearizable (Theorem 6.2).
Every finite subgroup action on the chordal cubic is linearizable, although the full automorphism group action is not (Theorem 5.1 and Remark 5.2).
For conic-singular cubics, the even dihedral subgroups generated by $a=\zeta_n$ and the swap are not linearizable, so the full automorphism group action is not linearizable (Theorem 4.3).
Every finite subgroup action on any non-cone cubic threefold with non-isolated singularities is $G$-unirational (Theorem 4.5).

Reading between the lines

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

The paper establishes a sharp separation between $G$-unirationality, which holds for every finite action, and linearizability, which fails for conic dihedral actions; a testable expectation is that non-linearizable finite actions on rationally connected threefolds often come from subgroups with an elliptic curve in their fixed locus.
The remaining open linearizability questions in the conic family are concentrated in finitely many parameter values; computing whether the finite groups not covered by Theorem 4.3 admit equivariant birational maps to $\mathbb{P}^3$ would complete the classification.
If the asserted reduction that actions fixing a singular point are linearizable ever fails, the theorems would still apply to the fixed-point-free actions explicitly analyzed, but the statements quantifying over all finite subgroups would need to be narrowed.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Here's the short version: the conic case (Theorem 4.3) is a genuine new result, and the unirationality theorem (4.5) goes through cleanly. But the proofs of Theorems 5.1 and 6.2 contain a false claim about generically free actions, so those two theorems are unsupported as written. The normal-form and automorphism computations are extensive and look right, but they are largely asserted via 'direct computation'—a minor issue by comparison.

The stress-test note is correct. For any involution in PGL2, the induced action on the Veronese surface S ≅ P2 has eigenvalues 1, -1, 1 on Sym^2, so it fixes a line pointwise. Hence the assertion in Section 5 that 'for all such G, the induced action on S is generically free' is false for every finite subgroup containing an involution, which includes all the non-cyclic groups relevant there. Proposition 2.1's linearizability claim explicitly requires generic freeness on the base, so the no-name lemma argument does not apply. The same problem occurs in Section 6: the action on the P2 factor fixes a line for any element with eigenvalue -1, so it is not generically free for, say, C2. This is not a gap in one edge case; it breaks the stated proofs of two main theorems.

The theorems may still be true. For the chordal cubic, the structure of the Schwarzenberger bundle or other methods might give linearizability, but the paper would need to supply a correct argument. The automorphism-group classifications (Propositions 3.2, 3.3, 4.1, 4.2, 6.1) are plausible and probably checkable, but the 'direct computation' leaves a lot out. Minor note: the introduction's reduction to actions fixing no singular points is asserted without proof; for the line and conic cases it excludes some components, and a reader shouldn't have to take it on faith.

Bottom line: I'd send this to referees, but with a clear warning that the generic-freeness claim must be fixed before acceptance. The conic result alone is worth a serious look. I'd wait for a corrected version before citing the chordal- or plane-singularity linearizability results.

Referee Report

3 major / 4 minor

Summary. The paper studies finite group actions on cubic threefolds in P^4 whose singular locus has dimension at least one. It classifies the possible automorphism groups for the four singularity types (line, conic, plane, rational normal quartic) and proves G-linearizability in the line, plane, and chordal cases, non-linearizability for most conic cases, and G-unirationality for all cases. The methods combine normal forms, blowups, projections, the no-name lemma, and twists over nonclosed fields.

Significance. If the results are correct, they provide a fairly complete equivariant birational classification for this class of rational threefolds, extending the authors' earlier work on isolated singularities. The negative result for conic singularities (Theorem 4.3) is particularly interesting as a counterpoint to the isolated-singularity case. The paper also gives explicit normal forms and automorphism group lists that could be useful for moduli and arithmetic questions. However, the reliance on asserted automorphism group classifications and on a reduction whose proof is omitted means the significance is conditional on filling those gaps.

major comments (3)

[Section 1, second paragraph] The claim that an action fixing a singular point of X is linearizable is used to reduce all four singularity types to the fixed-point-free case, but no proof or reference is provided. Since Theorems 3.1, 4.5, 5.1, and 6.2 assert statements for arbitrary finite subgroups, the authors need to justify this reduction. The natural argument via projection from the fixed point should be written out (or a reference given), because without it the proofs do not cover groups that fix a singular point.
[Propositions 3.2, 3.3, 4.1, 4.2, 6.1] The automorphism group classifications are asserted after 'direct computation' with the computations largely omitted. These classifications are load-bearing: they are used to reduce the linearizability arguments to the listed groups. The authors should include the computations, at least in an appendix or supplementary material, or provide a clear algorithmic description that allows verification. As written, a reader cannot check that the lists are complete or that the stated generators are correct.
[Section 5, proof of Theorem 5.1] The proof invokes an undefined 'Condition (A)' and asserts that the induced action on S is generically free and that the P1-bundle is G-linearized. Generic freeness is indeed true (the induced PGL2 action on the Veronese surface is faithful), but the lifting of the action to the vector bundle (the content of Condition (A)) is not established for all finite subgroups G, and the connection to Proposition 2.1 is too terse. Please expand this proof. The stress-test concern that involutions violate generic freeness is a misunderstanding: a fixed line in S does not prevent generic freeness, since the generic point of S is not on that line; the same clarification applies to the proof of Theorem 6.2.

minor comments (4)

[General] There are several typos, including 'EQUIV ARIANT' in the header and 'group s' in the introduction.
[Equation (6.2)] The matrix display has entries like 'b 2' and 'a 2' where superscripts appear to be missing; these should be b^2, a^2, c^2, and d^2.
[Section 5] The term 'Condition (A)' is used without a definition in this paper; it should be either defined or an explicit reference to a previous paper should be given.
[Theorem 4.3] The proof is extremely brief and relies on [6, Proposition 2.6] without explaining how the hypotheses of that proposition are satisfied; a few more sentences would help the reader.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the derivation chain uses external classifications and standard birational tools; self-citations are imported theorems with independent proofs.

full rationale

The paper's arguments are not circular. The classification of cubic threefolds with non-isolated singularities is imported from Yokoyama [19, Proposition 4.2], an external source. The main reduction tools, Proposition 2.1 (the no-name lemma, cited to Hajja--Kang [13]) and Proposition 2.3 (proved via Duncan--Reichstein [12] and Kollár [15]), are stated with proofs and do not presuppose the linearizability conclusions they are used to derive. The same-author citations, such as [6, Proposition 2.6] in Theorem 4.3, are used as black-box theorems with their own independent proofs, not as restatements of the present claims, so they do not make the argument circular. There is no fitting of parameters later re-labeled as predictions, no definitional equivalence between inputs and outputs, and no uniqueness theorem imported from the authors to force a choice. The unproved introductory assertion that actions fixing a singular point are linearizable is a gap/risk, and the generic freeness claim in Section 5 is false for involutions, but these are correctness concerns rather than circularity. Accordingly, the appropriate circularity score is 0.

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

The paper introduces no free parameters or invented entities. The central claim rests on a cited classification of singular loci, a stated but unproved reduction about fixed singular points, and standard theorems from birational geometry. The omitted automorphism computations are derivation gaps rather than additional axioms.

assumptions (6)

standard math k is an algebraically closed field of characteristic zero
Throughout the paper; the no-name lemma, versality criterion, and classification results require this.
domain assumption Classification of cubic threefolds with non-isolated singularities from Yokoyama [19, Proposition 4.2]: the singular locus is a plane, a line possibly with other components, a conic, or a rational normal quartic (the chordal cubic)
The normal forms in Sections 3, 4, 5, and 6 derive from this classification; it is cited, not re-proved.
domain assumption If a finite group G fixes a singular point of X, then the G-action is linearizable
Stated in the introduction, second paragraph, without proof; used to reduce to the case where no singular point is fixed.
standard math No-name lemma and projective no-name lemma (Hajja-Kang [13])
Used in Proposition 2.1 to establish unirationality and linearizability of P-bundles over generically free bases.
standard math Duncan-Reichstein versality theorem [12, Theorem 1.1]
Used in Proposition 2.3 and Theorem 4.5 to connect G-unirationality to K-unirationality of twisted forms.
standard math Kollar's theorem: cubic hypersurfaces with a rational point are unirational [15]
Used in the proof of Proposition 2.3.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Equivariant geometry of cubic threefolds with non-isolated singularities." pith.science (2026). https://pith.science/paper/IJ776MI5

@misc{pith2026250503986,
  author       = {Pith},
  title        = {Pith review of: Equivariant geometry of cubic threefolds with non-isolated singularities},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IJ776MI5}},
  note         = {Machine review of arXiv:2505.03986}
}

read the original abstract

We study linearizability of actions of finite groups on cubic threefolds with non-isolated singularities.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 16 canonical work pages

[1]

D. Allcock. The moduli space of cubic threefolds. J. Algebr. Geom. , 12(2):201– 223, 2003

work page 2003
[2]

Blanc, A

J. Blanc, A. Fanelli, and R. Terpereau. Automorphisms of P1-bundles over rational surfaces. ´Epijournal G´ eom. Alg´ ebrique, 6:Art. 23, 47, 2022

work page 2022
[3]

Blanc, A

J. Blanc, A. Fanelli, and R. Terpereau. Connected algebraic grou ps acting on three-dimensional Mori ﬁbrations. Int. Math. Res. Not. IMRN , (2):1572–1689, 2023

work page 2023
[4]

Casalaina-Martin, S

S. Casalaina-Martin, S. Grushevsky, K. Hulek, and R. Laza. Com plete moduli of cubic threefolds and their intermediate Jacobians. Proc. Lond. Math. Soc. (3), 122(2):259–316, 2021. 18 I. CHELTSOV, L. MARQUAND, Y. TSCHINKEL, AND ZH. ZHANG

work page 2021
[5]

Cheltsov, L

I. Cheltsov, L. Marquand, Yu. Tschinkel, and Zh. Zhang. Equiva riant geometry of singular cubic threefolds II, 2024. arXiv:2405.02744

arXiv 2024
[6]

Cheltsov, Yu

I. Cheltsov, Yu. Tschinkel, and Zh. Zhang. Equivariant geometr y of singular cubic threefolds. Forum Math. Sigma , 13:Paper No. e9, 2025

work page 2025
[7]

Cheltsov, Yu

I. Cheltsov, Yu. Tschinkel, and Zh. Zhang. Equivariant geometr y of the Segre cubic and the Burkhardt quartic. Selecta Math. (N.S.) , 31(1):Paper No. 7, 36, 2025

work page 2025
[8]

Cheltsov, Yu

I. Cheltsov, Yu. Tschinkel, and Zh. Zhang. Equivariant uniration ality of Fano threefolds, 2025. arXiv:2502.19598

arXiv 2025

Show all 19 references

[9]

Colliot-Th´ el` ene and A

J.-L. Colliot-Th´ el` ene and A. Pirutka. Certaines ﬁbrations en s urfaces quadriques r´ eelles, 2024.arXiv:2406.00463

2024
[10]

Colliot-Th´ el` ene and P

J.-L. Colliot-Th´ el` ene and P. Salberger. Arithmetic on some sin gular cubic hy- persurfaces. Proc. London Math. Soc. (3) , 58(3):519–549, 1989

1989
[11]

D. F. Coray and M. A. Tsfasman. Arithmetic on singular Del Pezz o surfaces. Proc. London Math. Soc. (3) , 57(1):25–87, 1988

1988
[12]

Duncan and Z

A. Duncan and Z. Reichstein. Versality of algebraic group action s and rational points on twisted varieties. J. Algebr. Geom. , 24(3):499–530, 2015

2015
[13]

Hajja and M

M. Hajja and M. Ch. Kang. Some actions of symmetric groups. J. Algebra , 177(2):511–535, 1995

1995
[14]

I.-K. Kim, J. Park, and J. Won. K-polystability of the First Secan t Varieties of Rational Normal Curves. Int. Math. Res. Not. IMRN , (7):rnaf088, 2025

2025
[15]

Koll´ ar

J. Koll´ ar. Unirationality of cubic hypersurfaces. J. Inst. Math. Jussieu , 1(3):467–476, 2002

2002
[16]

Liu and C

Y. Liu and C. Xu. K-stability of cubic threefolds. Duke Math. J. , 168(11):2029– 2073, 2019

2019
[17]

Pinardin, A

A. Pinardin, A. Sarikyan, and E. Yasinsky. Linearization problem for ﬁnite subgroups of the plane Cremona group, 2024. arXiv:2412.12022

2024 arXiv
[18]

Viktorova

S. Viktorova. On the classiﬁcation of singular cubic threefolds, 2025. To appear: Trans. Am. Math. Soc, arXiv:2304.10452

2025 arXiv
[19]

Yokoyama

M. Yokoyama. Stability of cubic 3-folds. Tokyo J. Math. , 25(1):85–105, 2002. Department of Mathematics, University of Edinburgh, UK Email address : I.Cheltsov@ed.ac.uk Courant Institute, 251 Mercer Street, New York, NY 10012, US A Email address : lisa.marquand@nyu.edu Courant...

2002

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

D. Allcock. The moduli space of cubic threefolds. J. Algebr. Geom. , 12(2):201– 223, 2003

work page 2003

[2] [2]

Blanc, A

J. Blanc, A. Fanelli, and R. Terpereau. Automorphisms of P1-bundles over rational surfaces. ´Epijournal G´ eom. Alg´ ebrique, 6:Art. 23, 47, 2022

work page 2022

[3] [3]

Blanc, A

J. Blanc, A. Fanelli, and R. Terpereau. Connected algebraic grou ps acting on three-dimensional Mori ﬁbrations. Int. Math. Res. Not. IMRN , (2):1572–1689, 2023

work page 2023

[4] [4]

Casalaina-Martin, S

S. Casalaina-Martin, S. Grushevsky, K. Hulek, and R. Laza. Com plete moduli of cubic threefolds and their intermediate Jacobians. Proc. Lond. Math. Soc. (3), 122(2):259–316, 2021. 18 I. CHELTSOV, L. MARQUAND, Y. TSCHINKEL, AND ZH. ZHANG

work page 2021

[5] [5]

Cheltsov, L

I. Cheltsov, L. Marquand, Yu. Tschinkel, and Zh. Zhang. Equiva riant geometry of singular cubic threefolds II, 2024. arXiv:2405.02744

arXiv 2024

[6] [6]

Cheltsov, Yu

I. Cheltsov, Yu. Tschinkel, and Zh. Zhang. Equivariant geometr y of singular cubic threefolds. Forum Math. Sigma , 13:Paper No. e9, 2025

work page 2025

[7] [7]

Cheltsov, Yu

I. Cheltsov, Yu. Tschinkel, and Zh. Zhang. Equivariant geometr y of the Segre cubic and the Burkhardt quartic. Selecta Math. (N.S.) , 31(1):Paper No. 7, 36, 2025

work page 2025

[8] [8]

Cheltsov, Yu

I. Cheltsov, Yu. Tschinkel, and Zh. Zhang. Equivariant uniration ality of Fano threefolds, 2025. arXiv:2502.19598

arXiv 2025