Additivity of non-acyclicity classes for constructible \'etale sheaves

Jiangnan Xiong

REVIEW 2 major objections 4 minor 17 references

Additivity of non-acyclicity classes for constructible \'etale sheaves

T0 review · 2 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper proves that non-acyclicity classes add over exact triangles of constructible sheaves.

desk verdict Genuinely new additivity theorem for NA classes, but Theorem 1.12 omits the cohomological smoothness hypothesis that the proof actually uses. read the letter →

arxiv 2505.24345 v1 pith:YAFZP3AN submitted 2025-05-30 math.AG math.CTmath.NT

Jiangnan Xiong This is my paper

classification math.AGmath.CTmath.NT MSC 14F2014F0518N70

keywords non-acyclicityclasscategoricaltracecohomologicalcorrespondencesétalesheavesadditivityconstructibleVerdierdualityexactnine-diagrams

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 establishes that the non-acyclicity class, a cohomological invariant measuring how a constructible sheaf fails to be acyclic along a closed subset relative to a fibration, is additive: for a distinguished triangle $F' \to F \to F''$ of sheaves satisfying the relevant local acyclicity conditions, the class of $F$ equals the class of $F'$ plus the class of $F''$. To prove this, the author rewrites the non-acyclicity class as a categorical trace-like composition using a bivariant version of cohomological correspondences, then shows that any such trace-like pairing is additive on exact triangles by passing through exact nine-diagrams. If correct, the result turns the non-acyclicity class into a genuine characteristic-class-valued invariant, parallel to the known additivity of cohomological characteristic classes, and provides a categorical explanation for why the class behaves like a ramification-theoretic conductor. The main theorem is stated under a cohomological smoothness assumption on the middle fibration.

What carries the argument

The load-bearing objects are the bivariant category of cohomological correspondences—roughly, a category whose objects are pairs of a scheme and a sheaf and whose morphisms are correspondences together with a map of sheaves—and the exact nine-diagram lemma. $\mathrm{BivCohCorr}_S$ is a symmetric monoidal fibration whose object $(X/Y;F)$ packages a scheme $X$ over $Y$ over $S$ together with a sheaf $F$; it organizes the categories $\mathrm{CohCorr}_Y$ for all intermediate bases $Y$ into one structure, so that the comparison maps between the dual over $Y$ and the dual over $S$ become natural isomorphisms under Assumption 2.22. The nine-diagram lemma (Proposition 3.6) takes an exact $3\times 3$ diagram whose rows and columns are exact triangles and produces an exact triangle on alternating sums of the entries; applied to the maps between coevaluation and evaluation associated to an endomorphism of an exact triangle, it yields the trace identity $\mathrm{Tr}(g) = \mathrm{Tr}(f) + \mathrm{Tr}(h)$ that is the heart of additivity.

What would settle it

Take $S = \mathrm{Spec}\,k$, $Y$ a smooth curve over $S$, $X$ a relative curve over $Y$, and an exact triangle $F'\to F\to F''$ of locally constant étale sheaves on $X$ that are ULA over both $Y$ and $S$. Compute the three localized classes $C^Z_{X/Y/S}$ by the defining excision formula: the theorem predicts the middle class is the sum of the two outer classes. A single such computation where the equality fails would falsify Theorem 4.4.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 4.4: under Assumption 2.22, for an exact triangle $F' \to F \to F''$ of constructible complexes of finite Tor-amplitude on $X$, if $F', F, F''$ are dualizable (the categorical form of universal local acyclicity) in $\mathrm{CohCorr}_S$ and their restrictions to an open immersion $U \subset X$ are dualizable in $\mathrm{CohCorr}_Y$, then the localized non-acyclicity classes satisfy $C^Z_{X/Y/S}(F) = C^Z_{X/Y/S}(F') + C^Z_{X/Y/S}(F'')$ in $$H^{0}$_Z(X, K_{X/Y/S})$. Equivalently, the non-localized class $C_{X/Y/S}$ is additive on exact triangles of dualizable objects. The route is a categorical trace-like formula: the class is written as the composition $(X;\Lambda) \to \mathrm{Hom}_Y((X;F),(X;F)) \to \Delta_{Y/S}((X;F) \otimes_S D(F/S)) \to D(X/Y/S)$ in the category of cohomological correspondences over $S$, and the localized version is obtained by excision on the closed complement of $U$. Additivity then follows from a general proposition about trace-like pairings built from such compositions.

Load-bearing premise

The proof depends on Assumption 2.22, that the dualizing complex $D(Y/S)$ is tensor-invertible in the category of cohomological correspondences over $Y$ and the bi-evaluation map is an isomorphism, which holds when $g:Y\to S$ is cohomologically smooth; without this, the trace-like formula for the non-acyclicity class and the comparison maps that make it additive are not available.

Editorial extensions

If this is right

The additivity of localized non-acyclicity classes holds for any exact triangle of constructible sheaves satisfying the ULA hypotheses, so the invariant behaves like a cohomological characteristic class in the relative setting.
The non-localized class $C_{X/Y/S}$ is additive as well, giving a relative analogue of the known additivity of categorical traces for cohomological characteristic classes.
Because the proof is categorical, the same exact-nine-diagram mechanism applies to any trace-like natural transformation in a stable symmetric monoidal $\infty$-category, not only to the étale-sheaf construction.
Under Assumption 2.22, which is satisfied when the middle morphism $g:Y\to S$ is cohomologically smooth, the theorem gives a uniform additivity statement that does not require case-by-case geometric arguments.

Reading between the lines

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

A natural next step, not pursued in the paper, is to use the same trace-like formula to prove additivity for the geometric characteristic cycle class, since both classes are expected to coincide under the cycle-class map.
The bivariant category could serve as the natural home for a relative trace formula; the trace-like composition here is already close to a relative trace, so one might deduce product formulas and decomposition statements by the same adjunction arguments.
The proof only needs the two comparison maps (2.17.1) and (2.17.2) to be isomorphisms; testing whether a weaker assumption than cohomological smoothness of $g:Y\to S$ suffices would enlarge the class of fibrations to which the additivity theorem applies.
In arithmetic settings where the non-acyclicity class recovers classical ramification conductors, additivity would give a unified proof of the additivity of conductors under extensions of sheaves.

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.

Desk Editor's Note

read the letter

Short version: the paper is worth taking seriously, but it needs a revision to align the statements with the assumptions. The main result is genuinely new: additivity for Yang–Zhao non-acyclicity classes has not been proved elsewhere, and the bivariant category of cohomological correspondences is a natural tool for the task. The trace-like formula (Proposition 2.28) and the exact nine-diagram proof (Propositions 4.3–4.4) are substantive. The paper also gives a self-contained proof of the nine-diagram lemma it needs, even though the author learned the trick from a note by Yang; that is fine in my book.

The soft spot, which the stress-test note correctly identifies, is that Theorem 1.12 appears to state additivity without any condition on the map g:Y→S. But Section 4 begins by imposing Assumption 2.22, which requires D(Y/S) to be ⊗-invertible and the bi-evaluation map to be an isomorphism—conditions that hold when g is cohomologically smooth. The proof of the categorical description of the non-acyclicity class and the additivity argument both rely on this assumption. So the theorem as stated, and the abstract's unqualified claim, are overbroad. The fix is straightforward: either carry Assumption 2.22 explicitly into the theorem statements, or prove that the NA class and its additivity make sense without it. This matters, because non-smooth finite morphisms are exactly the kind of case the geometric ramification theory wants to handle.

Minor points: the paper leans on known categorical trace additivity (May, Groth–Ponto–Schulman, Ramzi), so the novelty is in the application and formulation, not the underlying mechanism. That is fine for a paper like this. I did not see circularity: the result is genuinely derived from trace additivity, not assumed. No code or external verification, but that is not expected here.

Overall: the paper deserves peer review, not desk rejection, but only after the authors fix the mismatch between the main theorem and the stated assumptions. I would tell the editor to send it out and ask the referee to check that the trace-like formula really recovers the NA class under the stated hypotheses, and whether additivity can be pushed beyond cohomologically smooth g. A serious referee can handle that.

Referee Report

2 major / 4 minor

Summary. The paper develops a bivariant ∞-categorical framework of cohomological correspondences to give a trace-like formula for the non-acyclicity classes introduced by Yang and Zhao, and then derives an additivity theorem for these classes under distinguished triangles. Section 2 constructs the bivariant category BivCohCorr_S and defines the non-localized and localized NA classes via categorical traces; Section 3 proves a nine-diagram lemma; Section 4 applies the lemma to prove additivity. The abstract and Theorem 1.12 state the additivity theorem without any hypothesis on the morphism g: Y → S, but the proof in Section 4 is carried out under Assumption 2.22, which is satisfied, for example, when g is cohomologically smooth.

Significance. The categorical trace formula is a genuinely useful reformulation: it packages the NA class as an instance of the same construction that gives relative characteristic classes, and the additivity theorem is the expected 'characteristic class' property. The paper contains a self-contained proof of Proposition 3.6, and the exact nine-diagram technique is a workable substitute for earlier homotopy-theoretic trace-additivity results. If the hypothesis issue is resolved, this will be a solid contribution to geometric ramification theory. The main value is conditional: the current statement overclaims what is proved, because the proof as written requires Assumption 2.22 on g.

major comments (2)

[§4.1, Assumption 2.22; Theorem 1.12; Abstract] The additivity theorem is proved only under Assumption 2.22, which requires D(Y/S) to be ⊗_Y-invertible and the bi-evaluation map (2.19.1) to be an isomorphism, but the abstract and Theorem 1.12 state the result without this hypothesis. Section 4 opens by imposing Assumption 2.22, and Propositions 4.3 and 4.4 depend on (2.17.1) and (2.17.2) being isomorphisms, which are derived from Assumption 2.22 via Proposition 2.21. As written, the central claim is overbroad; the theorem should either carry Assumption 2.22 explicitly as a hypothesis, or the author should prove that the trace formula and the additivity argument survive without it.
[§4.4, diagram (4.4.2)] The equality '∆_{Y/S}(E_F ⊗_S Hom_S(E_F,S)) = i_* i^! ∆_{Y/S}(E_F ⊗_S Hom_S(E_F,S))' is not justified and is generally false, since the object on the left need not be supported on Z. The proof of Corollary 4.4 needs to explain explicitly how the class in H^0_Z is obtained: the vanishing of the U-restriction (Definition 2.30) should force the relevant trace to land in the Z-supported part. As written, this is a gap in the proof of part (2) of Theorem 1.12.

minor comments (4)

[Throughout] The abbreviation 'c.f.' should be 'cf.' throughout, and there is a typo 'equiality' in Proposition 1.17.
[Definition 2.30] The assertion that ∆_{Y/S}(F|_U ⊗ Hom_S(F|_U,S)) = 0 uses the dualizability assumption on U and should be stated explicitly as a consequence of (2.23.1).
[Acknowledgments; Proposition 3.6] The paper relies on an unpublished note for Proposition 3.6; although a proof is included, adding the note's URL to the bibliography rather than only in the acknowledgments would improve traceability.
[§1.20] The proof of Proposition 1.17 is only sketched; since the same nine-diagram technique is used later, a few more details there would help the reader.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: NA-class additivity is derived from categorical trace additivity; the only notable issue is a missing hypothesis in Theorem 1.12, which is a correctness gap, not a circular reduction.

full rationale

The paper does not assume the additivity it proves. The non-acyclicity class is imported from Yang--Zhao as an external object, and the paper gives a new categorical description in Definition 2.28 and 2.30, explicitly noting that unwinding the definitions recovers the earlier constructions. The additivity result in Section 4 is then obtained by applying the independently proved exact-nine-diagram trace additivity (Proposition 3.6 and 4.3) to this description, not by quoting the conclusion or by fitting parameters. No fitted data are called predictions, no uniqueness theorem is imported from the authors' prior work, and no ansatz is smuggled in via citation: Assumption 2.22 is an explicit hypothesis, not a hidden consequence of the conclusion. The paper's self-citations amount to relying on Yang--Zhao's definition of the NA class and on standard categorical-trace references, which provide independent content. The most serious issue is that Theorem 1.12 states additivity without carrying Assumption 2.22, whereas Section 4.1 explicitly assumes g satisfies that assumption and the proofs of Propositions 4.3 and 4.4 depend on it. This is a missing-hypothesis / overstatement problem, not a circularity: the body does not derive the unqualified theorem from itself. Accordingly, the circularity score is 0.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The central claim rests on heavy background from [9] and [17], and on Assumption 2.22 restricting to cohomologically smooth morphisms. There are no fitted numerical parameters. The bivariant category is a formal definition rather than an invented empirical entity.

assumptions (4)

standard math The six-functor formalism for etale cohomology on S with Lambda coefficients is a symmetric monoidal infinity-categorical formalism satisfying the expected adjunctions, base change, and projection formulas.
Invoked throughout Section 2 and in the notation and conventions; not proved in the paper.
domain assumption Constructible complexes of finite Tor-amplitude on X are dualizable in CohCorr_S if and only if they are h-ULA, as established by Lu and Zheng.
Used in 2.16 and in the hypotheses of the additivity theorems; cited from [9].
domain assumption D(Y/S) is a tensor-invertible object in CohCorr_Y and the bi-evaluation map (2.19.1) is an isomorphism, which holds when g: Y to S is cohomologically smooth.
Assumption 2.22 is load-bearing for the trace-like formula and for making (2.17.1) and (2.17.2) isomorphisms.
domain assumption The non-acyclicity class C_Z_{X/Y/S}(F) is the class defined by Yang and Zhao in [17], and the paper's categorical formula recovers [17, (4.12)] and [17, (4.14)].
The central object is imported from [17]; the paper reviews but does not independently re-establish the original definition.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Additivity of non-acyclicity classes for constructible \'etale sheaves." pith.science (2026). https://pith.science/paper/YAFZP3AN

@misc{pith2026250524345,
  author       = {Pith},
  title        = {Pith review of: Additivity of non-acyclicity classes for constructible \'etale sheaves},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YAFZP3AN}},
  note         = {Machine review of arXiv:2505.24345}
}

read the original abstract

Using a bivariant version of cohomological correspondences, we establish a categorical trace-like formula for the non-acyclicity classes introduced by Yang and Zhao (arXiv:2209.11086). As an application, we prove the additivity for the non-acyclicity classes.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

17 extracted references · 16 canonical work pages

[1]

Abbes, T

A. Abbes, T. Saito,The characteristic class and ramification of an ℓ-adic ´ etale sheaf, Invent. Math., 168: 567-612 (2007). ↑1

work page 2007
[2]

Beilinson,Constructible sheaves are holonomic , Sel

A. Beilinson,Constructible sheaves are holonomic , Sel. Math. New Ser. 22: 1797-1819 (2016).↑1

work page 2016
[3]

A. Dold, D. Puppe, Duality, trace and transfer , Proceedings of the International Conference on Geometric Topology (Warsaw, 1978), PWN, Warsaw: 81-102 (1980)

work page 1980
[4]

Ferrand,On the non additivity of the trace in derived categories , preprint available at arXiv: 0506589.↑3

D. Ferrand,On the non additivity of the trace in derived categories , preprint available at arXiv: 0506589.↑3

work page
[5]

Groth, K

M. Groth, K. Ponto, M. Schulman,The additivity of traces in monoidal derivators , Journal of K-theory, 14(3): 422-494 (2014). ↑3

work page 2014
[6]

Heyer, L

C. Heyer, L. Mann,6-functor formalisms and smooth representations , preprint available at arXiv:2410.13038

arXiv
[7]

Illusie,Around the Thom-Sebastiani theorem, with an appendix by Weizhe Zheng , Manuscr

L. Illusie,Around the Thom-Sebastiani theorem, with an appendix by Weizhe Zheng , Manuscr. Math., 152: 61-125 (2017)

work page 2017
[8]

F. Jin, E. Yang,K¨ unneth formulas for motives and additivity of traces, Advances in Mathematics, 376: 107446 (2021). ↑3 20 JIANGNAN XIONG

work page 2021

Show all 17 references

[9]

Q. Lu, W. Zheng,Categorical traces and a relative Lefschetz-Verdier formula , Forum of Mathematics, Sigma, Vol. 10: 1-24 (2022).↑2, ↑5, ↑9, ↑18

2022
[10]

Lurie,Kerodon, available at kerodon.net.↑5

J. Lurie,Kerodon, available at kerodon.net.↑5
[11]

Lurie,Higher topos theory , Annals of Mathematics Studies 170, Princeton (2009).↑5, ↑14, ↑15, ↑17

J. Lurie,Higher topos theory , Annals of Mathematics Studies 170, Princeton (2009).↑5, ↑14, ↑15, ↑17

2009
[12]

Lurie,Higher algebra, 2017, preprint available at Lurie’s homepage.↑5, ↑14, ↑15

J. Lurie,Higher algebra, 2017, preprint available at Lurie’s homepage.↑5, ↑14, ↑15

2017
[13]

J. P. May,The additivity of traces in triangulated categories , Advances in Mathematics, Volume 163, Issue 1: 34-73 (2001). ↑3

2001
[14]

Ramzi,The additivity of traces in stable ∞-categories, preprint available at arXiv: 2109.01512.↑4

M. Ramzi,The additivity of traces in stable ∞-categories, preprint available at arXiv: 2109.01512.↑4

arXiv
[15]

Saito, The characteristic cycle and the singular support of a constructible sheaf , Invent

T. Saito, The characteristic cycle and the singular support of a constructible sheaf , Invent. Math., 207: 597-695 (2017). ↑1

2017
[16]

E. Yang, Y. Zhao,On the relative twist formula of ℓ-adic sheaves, Acta Math. Sin. Engl. Ser. 37: 73-94 (2021). ↑2

2021
[17]

E. Yang, Y. Zhao,Cohomological Milnor formula and Saito’s conjecture on characteristic classes , Invent. Math., Volume 240: 123-191 (2025).↑1, ↑2, ↑5, ↑7, ↑8, ↑9, ↑12, ↑13, ↑14

2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Abbes, T

A. Abbes, T. Saito,The characteristic class and ramification of an ℓ-adic ´ etale sheaf, Invent. Math., 168: 567-612 (2007). ↑1

work page 2007

[2] [2]

Beilinson,Constructible sheaves are holonomic , Sel

A. Beilinson,Constructible sheaves are holonomic , Sel. Math. New Ser. 22: 1797-1819 (2016).↑1

work page 2016

[3] [3]

A. Dold, D. Puppe, Duality, trace and transfer , Proceedings of the International Conference on Geometric Topology (Warsaw, 1978), PWN, Warsaw: 81-102 (1980)

work page 1980

[4] [4]

Ferrand,On the non additivity of the trace in derived categories , preprint available at arXiv: 0506589.↑3

D. Ferrand,On the non additivity of the trace in derived categories , preprint available at arXiv: 0506589.↑3

work page

[5] [5]

Groth, K

M. Groth, K. Ponto, M. Schulman,The additivity of traces in monoidal derivators , Journal of K-theory, 14(3): 422-494 (2014). ↑3

work page 2014

[6] [6]

Heyer, L

C. Heyer, L. Mann,6-functor formalisms and smooth representations , preprint available at arXiv:2410.13038

arXiv

[7] [7]

Illusie,Around the Thom-Sebastiani theorem, with an appendix by Weizhe Zheng , Manuscr

L. Illusie,Around the Thom-Sebastiani theorem, with an appendix by Weizhe Zheng , Manuscr. Math., 152: 61-125 (2017)

work page 2017

[8] [8]

F. Jin, E. Yang,K¨ unneth formulas for motives and additivity of traces, Advances in Mathematics, 376: 107446 (2021). ↑3 20 JIANGNAN XIONG

work page 2021