REVIEW 2 major objections 4 minor 11 references
Extensions of Abelian Schemes and the Additive Group
T0 review · 2 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash
Pith's one-line read The paper computes the fppf extension sheaves of abelian schemes and the additive group by Gm: generalized Barsotti–Weil, vanishing of Ext^2(A,Gm), and a seminormality criterion for Ext^1(Ga,Gm) over Q.
desk verdict A serious paper with three genuinely new results; the proof of the Ext^2 vanishing has a real but fixable gap at the site comparison in Section 5, so it deserves a serious referee. 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 tool is the Breen–Deligne resolution: a functorial resolution of any abelian group object $G$ by free abelian group objects on powers of $G$, $\cdots \to \bigoplus_j \mathbb{Z}[G^{s_{i,j}}]\to\cdots\to \mathbb{Z}[G^2]\to \mathbb{Z}[G]\to G$, with explicit differentials. Applying $\operatorname{Hom}(-,A)$ to it and taking hypercohomology produces a spectral sequence $E_1^{i,j}=\prod_r H^j(G_T^{s_{i,r}},A_T)\Rightarrow \operatorname{Ext}^{i+j}_T(G,A)$, whose low-degree terms read $0\to H^2_s(G,A)\to \operatorname{Ext}^1(G,A)\to H^1_m(G,A)\to H^3_s(G,A)\to \operatorname{Ext}^2(G,A)$. Here $H^1_m(G,A)$ is the group of primitive, or multiplicative, $A$-torsors over $G$ (classes whose pullback along the group law is the contracted product of the two pullbacks), and $H^2_s,H^3_s$ are symmetric Hochschild-cohomology variants. The hypotheses of Theorems A and C are engineered to make $H^2_s$ and $H^3_s$ vanish, so that $\operatorname{Ext}^1$ is governed by multiplicative torsors; for Theorem C, group-cohomology computations for $\mathbb{G}_a$ over reduced rings or $\mathbb{Q}$-algebras supply this vanishing. For Theorem B the proof instead runs a dévissage from fields to artinian local rings to complete noetherian rings, using torsion of $\operatorname{Ext}^2$ over regular bases, injectivity of restriction to the reduced and completed base (the formal GAGA-type Lemma 4.6), and spreading-out via algebraization.
What would settle it
A direct check: take $R=\mathbb{Q}[t^2,t^3]$, the reduced affine cusp that is not seminormal, and compute $\underline{\operatorname{Ext}}^1(\mathbb{G}_a,\mathbb{G}_m)(\operatorname{Spec}R)$; Theorem 6.1 predicts a nonzero group, with an explicit nonzero class represented by the invertible module $M=(t^2u^2,1+tu)$ over $R[u]$, so a verification that this module is trivial, or that $H^1_m(\mathbb{G}_{a,\operatorname{Spec}R},\mathbb{G}_m)$ vanishes, would falsify the theorem. For the abelian-scheme statement, the surgical test is to find an $I$-adically complete noetherian $R$ and an abelian scheme $A$ for which $\operatorname{Ext}^2_R(A,\mathbb{G}_m)\to\lim_n \operatorname{Ext}^2_{R/I^n}(A,\mathbb{G}_m)$ has a nonzero kernel; Lemma 4.6 asserts no such $R$ and $A$ exist.
Extended reading notes
Core claim
On the paper's own terms, the central discovery is that extension sheaves of the two standard commutative group schemes by $\mathbb{G}_m$ are fully computable and obey clean formulas. Theorem 3.1 says that if $G$ and $A$ are abelian sheaves such that every pointed morphism $G_T^n\to A_T$ is trivial for $n=1,2,3$ after any base change, then $\operatorname{Ext}^1_T(G,A)\to \operatorname{Ext}^1(G,A)(T)$ and $\operatorname{Ext}^1_T(G,A)\to H^1_m(G_T,A_T)$ are isomorphisms; for $A$ an abelian scheme this yields $\operatorname{Ext}^1(A,\mathbb{G}_m)\simeq A^\vee$ and $\operatorname{Ext}^1(A,M)\simeq \mathrm{Lie}(A^\vee)\otimes_{\mathcal{O}_S}M$. Theorem 4.3 identifies $\operatorname{Ext}^2_T(A,\mathbb{G}_m)$ with $H^1(T,A_T^\vee)$ for every $S$-scheme $T$, so that the sheaf $\underline{\operatorname{Ext}}^2(A,\mathbb{G}_m)$ vanishes on the big fppf site, hence on the big étale site; for quasi-coherent $M$ the analogous sheaf $\underline{\operatorname{Ext}}^2(A,M)$ vanishes as well. Theorem 6.1 gives, for quasi-compact quasi-separated $\mathbb{Q}$-schemes $T$, isomorphisms $\underline{\operatorname{Ext}}^1(\mathbb{G}_a,\mathbb{G}_m)(T)\simeq \underline{\operatorname{Ext}}^1(\mathbb{G}_a,\mathbb{G}_m)(T_{\mathrm{red}})\simeq \operatorname{Ext}^1_{T_{\mathrm{red}}}(\mathbb{G}_a,\mathbb{G}_m)\simeq H^1_m(\mathbb{G}_{a,T_{\mathrm{red}}},\mathbb{G}_m)$, with vanishing when $T_{\mathrm{red}}$ is seminormal and nonvanishing when $T_{\mathrm{red}}$ is affine and not seminormal.
Load-bearing premise
For the vanishing of $\underline{\operatorname{Ext}}^2(A,\mathbb{G}_m)$ over general bases, the load-bearing premise is Lemma 4.6: if $R$ is $I$-adically complete noetherian, the natural map $\operatorname{Ext}^2_R(A,\mathbb{G}_m)\to\lim_n \operatorname{Ext}^2_{R/I^n}(A,\mathbb{G}_m)$ is injective, and the proof of that lemma identifies the relevant $H^1$ and $H^2$ comparison maps with the paper's formal GAGA results; if that identification gives way, the sheaf-vanishing conclusion does not follow.
Editorial extensions
If this is right
- For an abelian scheme $A$ over any base, $\underline{\operatorname{Ext}}^1(A,\mathbb{G}_m)$ is represented by the dual abelian scheme $A^\vee$, and $\underline{\operatorname{Ext}}^1(A,M)\simeq \mathrm{Lie}(A^\vee)\otimes_{\mathcal{O}_S}M$ for quasi-coherent $M$.
- For every $S$-scheme $T$, $\operatorname{Ext}^2_T(A,\mathbb{G}_m)\simeq H^1(T,A_T^\vee)$; individual extension groups can be nonzero even on affine $T$, although the extension sheaf itself vanishes.
- The sheaf $\underline{\operatorname{Ext}}^2(A,\mathbb{G}_m)$ vanishes on the big fppf site, hence on the big étale site, and $\underline{\operatorname{Ext}}^2(A,M)=0$ for every quasi-coherent $M$; this replaces an earlier argument that had only established torsion over regular bases.
- For quasi-compact quasi-separated $\mathbb{Q}$-schemes $T$, the sheaf $\underline{\operatorname{Ext}}^1(\mathbb{G}_a,\mathbb{G}_m)(T)$ is isomorphic to $H^1_m(\mathbb{G}_{a,T_{\mathrm{red}}},\mathbb{G}_m)$, so it vanishes for seminormal $T_{\mathrm{red}}$ and is nonzero for affine non-seminormal $T_{\mathrm{red}}$; the classical vanishing statement for this sheaf is therefore false in general.
Reading between the lines
- Beyond the paper: the same criterion of trivial pointed maps $G_T^n\to A_T$ suggests a uniform recipe for computing $\operatorname{Ext}^1(G,A)$ for other group schemes $G$ with rigid finite powers, and one could test whether $\operatorname{Ext}^1(\mathbb{G}_a^N,\mathbb{G}_m)$ over $\mathbb{Q}$ is again governed by seminormality of the base.
- Beyond the paper: the formula $\operatorname{Ext}^2_T(A,\mathbb{G}_m)\simeq H^1(T,A_T^\vee)$ invites the analogue $\operatorname{Ext}^k_T(A,\mathbb{G}_m)\simeq H^{k-1}(T,A_T^\vee)$ for $k>2$ whenever the higher extension sheaves are torsion, a torsion statement the paper itself expects to hold in a range depending on residual characteristics.
- Beyond the paper: because the paper makes the nonvanishing classes on non-seminormal bases explicit through an algorithmically checkable module, Theorem 6.1 could be turned into machine-verifiable certificates for affine non-seminormal $\mathbb{Q}$-algebras.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This paper studies extension sheaves in the fppf topology. Theorem 3.1 gives a generalized Barsotti--Weil isomorphism Ext^1_T(G,A) ≅ H^1_m(G_T,A_T) under a constancy hypothesis, with Corollaries 3.5--3.6 recovering Ext^1(A,M) ≅ Lie(A^∨)⊗M and Ext^1(A,G_m) ≅ A^∨. Theorem 4.3 computes Ext^2_T(A,G_m) ≅ H^1(T,A_T^∨) and establishes the vanishing of the sheaf Ext^2(A,G_m), by a dévissage whose final step (Lemma 4.6) is proved in Section 5 through a lax-colimit topos and a comparison with [KM23]. Theorem 6.1 computes Ext^1(G_a,G_m) over Q-schemes and shows that it is governed by seminormality of T_red. Sections 2 and 7 develop the Breen--Deligne spectral sequence machinery and the necessary Hochschild cohomology computations.
Significance. If the results are correct, this is a substantial contribution: it repairs a frequently invoked but flawed inference from Breen's torsion theorem to the vanishing of Ext^2(A,G_m), gives a clean and more general formulation of the Barsotti--Weil formula, and settles the subtle Ext^1(G_a,G_m) computation with a sharp seminormality criterion. The paper is generally careful: the spectral sequence setups are explicit, the main external dependencies are named, and the inline Macaulay2 verification is reproducible. However, the proof of Lemma 4.6 contains an unproved site-comparison that is load-bearing for Theorem 4.3, so the manuscript is not yet in publishable form.
major comments (2)
- [Section 5, Corollary 5.5 and footnote 7] The identification of the cohomology groups H^i(A^n,G_m) in the lax-colimit topos X with the formal GAGA cohomology groups of [KM23, Props. 3.2 and 5.3] is asserted in a footnote rather than proved. The objects here are cohomology groups in a lax colimit of fppf topoi, whereas [KM23] works on small étale sites of the relevant formal schemes; an interpretation of H^1 and H^2 in terms of torsors and gerbes does not by itself establish site-invariance for these two sites in this setting. This comparison is the only input that yields the injectivity of Ext^2_R(A,G_m)→Ext^2_X(A,G_m), and hence the whole conclusion of Lemma 4.6; without it the vanishing of Ext^2(A,G_m) in Theorem 4.3 is unsupported. Please supply a proof, for example by establishing a Milnor-type exact sequence 0→lim^1 H^{i-1}(A_n,G_m)→H^i(A,G_m)→lim H^i(A_n,G_m)→0 with vanishing lim^1 terms, together with an identification of the limits with the groups appearing in [KM23]. As written, the footnote explicitly concedes the site difference, so this is an admitted gap rather than a disagreement about an external theorem.
- [Section 4, Theorem 4.1] The proof that the middle row of the Breen--Deligne spectral sequence is exact at H^{1,2} is too compressed. The sentence 'By considering all possible inclusions ... we see that the map ... acts as (x,y)↦(x,0,−y)' hides a four-term diagram chase involving the pullbacks pr*_{1,2}, (id×m)*, (m×id)*, pr*_{2,3} and the Künneth identifications. Since Theorem 4.1 underpins Corollary 4.2 and Lemma 4.5, and hence the reduction to R_red in the proof of Theorem 4.3, this is load-bearing. Please write out the maps in terms of the Künneth components and justify the claimed formula explicitly.
minor comments (4)
- [Section 2, Remark 2.15] There is a typo: 'pursuded' should be 'pursued'.
- [Section 5, Lemma 5.1] In the final sentence of the proof, the notation 'P(ker ι_{n−1})' is imprecise: a base term for I_1 and a consistent indexing of the kernels ker ι_i should be spelled out before claiming that this object is isomorphic to (I_n,ι_n).
- [Section 6, Remark 6.6] The Macaulay2 code is a useful reproducibility check, but the text should state more explicitly that the change-of-basis matrix produced by the computation supplies the coefficients exhibiting 1 as a linear combination of the generators; as printed, the code only displays a Groebner basis.
- [Section 4, proof of Theorem 4.1] The statement that checking injectivity after tensoring with F_p for a single prime p suffices is correct for maps between finite free Z-modules, but it deserves a one-line justification because it can puzzle readers who expect a check at every prime.
Circularity Check
No significant circularity: the central theorems are derived from external inputs and prior work that do not assume the target results.
full rationale
The paper's main results — Theorem 3.1 (generalized Barsotti–Weil), Theorem 4.3 (vanishing of Ext^2(A,G_m)), and Theorem 6.1 (computation of Ext^1(G_a,G_m) over Q) — are each proven from stated external inputs rather than from the conclusions being derived. Theorem 3.1 is proved directly via the Breen–Deligne resolution and a constancy argument for maps G^n -> A; the identification with the dual abelian scheme in Corollary 3.6 uses standard Picard-scheme facts from [GW23]. Theorem 4.3 is a dévissage: Lemma 4.4 cites [SGA 7I] for vanishing of Ext^1(A[n], G_m); Lemma 4.5 reduces square-zero thickenings to the quasi-coherent vanishing Theorem 4.1, which is proved independently by explicit spectral-sequence row computations; Lemma 4.6 is proved in Section 5 using the Breen–Deligne spectral sequence and [KM23, Props. 3.2 and 5.3] as external formal GAGA inputs. None of these inputs is the target vanishing of Ext^2(A,G_m). The footnote on page 28 admitting a site difference with [KM23] flags a possible technical gap in identifying H^i(bA^n,"G_m) with the cohomology groups of [KM23]; that is a correctness concern, not circularity, because the cited result is independent and does not assume the paper's conclusions. Theorem 6.1 uses [Ros23, Lem. 2.2.5 and 2.2.9] and [Bha22] for computations of formal sections and Picard groups; those are prior published results whose stated hypotheses do not include the Ext^1(G_a,G_m) vanishing or nonvanishing being proved. The self-citations to [Ros23] and [Rib24] supply background and independent lemmas, not the load-bearing conclusion. No parameter is fitted to a data subset, no target statement is renamed as an input, and no uniqueness theorem from the authors' own prior work is invoked to force a choice. The derivation chain is therefore not circular.
Assumptions & free parameters
assumptions (9)
- standard math Breen-Deligne resolution exists functorially for abelian groups in a topos (Prop 2.4).
- standard math For an abelian scheme, R^i p_* O_A is finite locally free and compatible with base change ([GW23, Thm. 27.203]).
- standard math Ext^2_S(A,Gm) is torsion when S is regular (Breen's theorem, [Bre69a, §7]).
- standard math Formal GAGA for gerbes: comparison maps of [KM23, Props. 3.2, 5.3] apply to the big fppf site via the lisse-étale site.
- standard math H^n(R, \hat{G}_a)=0 for n=1,2 and Γ(U,\hat{G}_a)=0 for reduced U (de Jong, via [Bha22, Rem. 2.2.18] and [Ros23, Lem. 2.2.5]).
- standard math If R is not seminormal, there exists s in R^sn \ R with s^2,s^3 in R ([Swa80, Lem. 2.6]).
- standard math For seminormal T, pullback Pic(T)→Pic(Ga,T) is an isomorphism ([Sad21, Lem. 4.3]).
- standard math H^3_0(Ga,Q,Gm,Q) vanishes ([HW58, Thm. 1]).
- standard math Ext^1(A[n],Gm)=0 ([SGA 7I, Exp. VIII, Prop. 3.3.1]).
Cite this review
Pith. "Pith review of Extensions of Abelian Schemes and the Additive Group." pith.science (2026). https://pith.science/paper/6I6P5OQA
@misc{pith2026250617393,
author = {Pith},
title = {Pith review of: Extensions of Abelian Schemes and the Additive Group},
year = {2026},
howpublished = {\url{https://pith.science/paper/6I6P5OQA}},
note = {Machine review of arXiv:2506.17393}
}
abstract
We compute extension sheaves of abelian schemes and of the additive group by the multiplicative group in the fppf topology. Our main results include a generalized and streamlined proof of the Barsotti--Weil formula, the vanishing of $\underline{\operatorname{Ext}}^2(A,\mathbb{G}_m)$ for an abelian scheme $A$ over a general base, and a description of $\underline{\operatorname{Ext}}^1(\mathbb{G}_a,\mathbb{G}_m)$ in characteristic zero.
Reference graph
Works this paper leans on
-
[1]
Torsion free and projective modules
[Bas62] Hyman Bass. “Torsion free and projective modules”. In:Trans. Amer. Math. Soc. 102.2 (1962), pp. 319–327 (cit. on p. 30). 37 [BB09] Luca Barbieri-Viale and Alessandra Bertapelle. “Sharp de Rham realization”. In:Adv. Math.222.4 (2009), pp. 1308–1338 (cit. on p. 4). [Ber14] Alessandra Bertapelle. “Generalized 1-motivic sheaves”. In:J. Algebra420 (201...
work page 1962
-
[119]
Generic vanishing for holonomicD-modules: a study via Cartier duality
Lecture Notes in Mathematics. Springer Berlin Heidelberg, 1970 (cit. on p. 21). [Rib24] Gabriel Ribeiro. “Generic vanishing for holonomicD-modules: a study via Cartier duality”. PhD thesis. Institut polytechnique de Paris, 2024 (cit. on p. 4). [Ros23] Zev Rosengarten. “Tate duality in positive dimension over function fields”. In:Mem. Amer. Math. Soc.290.1...
work page 2023
-
[145]
American Mathematical Society, 2013 (cit. on p. 32). [Wei91] CharlesA.Weibel.“Picisacontractedfunctor”.In: Invent.Math. 103.1(1991), pp. 351–377 (cit. on p. 32). Gabriel Ribeiro, Department of Mathematics, ETH Zurich, 8092 Zurich, Switzerland Email address:gabriel.ribeiro@math.ethz.ch Zev Rosengarten, Einstein Institute of Mathematics, The Hebrew Universi...
work page 1991
-
[170]
The classifying topos of a continuous groupoid. I
Cambridge studies in advanced mathematics. Cambridge University Press, 2017 (cit. on p. 8). [Moe88] Ieke Moerdijk. “The classifying topos of a continuous groupoid. I”. In:Trans. Amer. Math. Soc.310.2 (1988), pp. 629–668 (cit. on p. 23). [Ols07] Martin Olsson. “Sheaves on Artin stacks”. In:J. Reine Angew. Math.2007 (2007), pp. 55–112 (cit. on pp. 11, 12). [...
work page 1988
-
[179]
Grundlehren der mathe- matischen Wissenschaften. Springer Berlin Heidelberg, 1971 (cit. on pp. 8, 9). [GW23] Ulrich Görtz and Torsten Wedhorn.Algebraic Geometry II: Cohomology of Schemes. Springer Studium Mathematik - Master. Springer Spektrum Wies- baden, 2023 (cit. on pp. 14, 17, 18, 20, 21, 23). [HW58] Robert Heaton and George Whaples. “Polynomial cocy...
work page 1958
-
[269]
Springer Berlin Heidelberg, 1972 (cit
Lecture Notes in Mathematics. Springer Berlin Heidelberg, 1972 (cit. on pp. 6, 7). [SGA 4II] Michael Artin, Alexander Grothendieck, and Jean-Louis Verdier.Théorie des Topos et Cohomologie Étale des Schémas : Séminaire de Géométrie Algébrique du Bois-Marie 1963–1964. Tome
work page 1972
-
[270]
Springer Berlin Heidelberg, 1972 (cit
Lecture Notes in Mathematics. Springer Berlin Heidelberg, 1972 (cit. on p. 13). 39 [SGA 4III] Michael Artin and Pierre Deligne.Théorie des Topos et Cohomologie Étale des Schémas : Séminaire de Géométrie Algébrique du Bois-Marie 1963–1964. Tome
work page 1972
-
[288]
LectureNotesinMathematics.SpringerBerlinHeidelberg,1972(cit.onpp.2, 7, 17, 22). [Stacks] The Stacks Project Authors.Stacks Project. Available athttps://stacks. math.columbia.edu. 2018 (cit. on pp. 6, 7, 9, 11, 12, 14, 17, 22–25, 29, 33). [Swa80] Richard Swan. “On seminormality”. In:J. Algebra67.1 (1980), pp. 210–229 (cit. on pp. 29, 30). [Wei13] Charles A...
work page 1980
Show all 11 references
-
[305]
Springer Berlin Heidelberg, 1973 (cit
Lecture Notes in Mathematics. Springer Berlin Heidelberg, 1973 (cit. on p. 1). [SGA 7I] Alexander Grothendieck et al.Groupes de Monodromie en Géométrie Algébrique : Séminaire de Géométrie Algébrique du Bois-Marie 1967–1969. Tome
1973
-
[1996]
Sur les groupes de Lie formels à un paramètre
arXiv:alg- geom/9603004 [alg-geom] (cit. on p. 1). [Laz55] Michel Lazard. “Sur les groupes de Lie formels à un paramètre”. In:Bull. Soc. Math. Fr.83 (1955), pp. 251–274 (cit. on pp. 34, 35). [Mil17] James S. Milne.Algebraic groups: the theory of group schemes of finite type ov...
1955
-
[2023]
Transformation de Fourier généralisée
arXiv: 2305.19114 [math.AG] (cit. on pp. 5, 28). [Lau96] Gérard Laumon. “Transformation de Fourier généralisée”
Reviewed August 15, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.