A cohomological smoothness conjecture for moduli of mixed characteristic local shtukas with one leg

Sean Howe

REVIEW 3 major objections 3 minor 12 references

A cohomological smoothness conjecture for moduli of mixed characteristic local shtukas with one leg

T0 review · 3 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The very special locus in shtuka moduli is exactly the zero Harder–Narasimhan slope locus, and cohomological smoothness holds on its complement in the EL Rapoport–Zink case.

desk verdict Howe proves a genuine extension of Ivanov–Weinstein to all EL Rapoport–Zink spaces, but the proof's last step rests on an identification of inscribed structures that the paper calls 'almost formal' and leaves unproved. read the letter →

arxiv 2508.11595 v1 pith:UKN2VZUF submitted 2025-08-15 math.AG math.NT

Sean Howe This is my paper

classification math.AGmath.NT MSC 11G1814G2214L05

keywords localshtukasRapoport–ZinkspacescohomologicalsmoothnessBanach–ColmezFargues–FontainecurveHarder–Narasimhanslopesinscribedv-sheavesdiamonds

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 isolates a simple numerical condition on the inscribed tangent bundle of moduli of mixed-characteristic local shtukas with one leg and fixed determinant: a geometric point is very special precisely when the pullback of the bundle $E^{\circ}_{\max}$ admits zero as a Harder–Narasimhan slope, equivalently when the corresponding inscribed tangent space is disconnected. The paper conjectures that the structure morphism to $\operatorname{Spd} C_p$ is cohomologically smooth over the open complement of the very special locus, and proves the conjecture in the EL Rapoport–Zink case, generalizing earlier work from the basic case to arbitrary $b$. This matters because cohomological smoothness is the finiteness property needed for $\ell$-adic cohomology on these infinite-level diamonds, so the result turns a structural question about period spaces into a checkable slope computation.

What carries the argument

The central object is the inscribed Banach–Colmez tangent bundle. The paper works with inscribed v-sheaves, a differential enrichment of diamonds in which each space carries a tangent bundle valued in Banach–Colmez spaces. For the local shtuka moduli space, this tangent bundle is the space of global sections of a vector bundle $E^{\circ}_{\max}$ on the relative Fargues–Fontaine curve, formed from the adjoint action of $b$ on the derived Lie algebra. The very-special/non-very-special dichotomy is exactly whether zero appears among the Harder–Narasimhan slopes of $z^*E^{\circ}_{\max}$. In the EL case, comparing this inscribed structure with the Ivanov–Weinstein moduli-of-sections scheme shows

What would settle it

Exhibit a geometric point in a non-basic EL Rapoport–Zink space whose intersection of the two period fibers is non-discrete while $z^*E^{\circ}_{\max}$ has no zero Harder–Narasimhan slope; that mismatch would refute the characterization in Proposition 2.3.5 and the theorem built on it.

Watch

Extended reading notes

Core claim

For a connected reductive group $G$, a conjugacy class $[\mu]$ of cocharacters, and a Kottwitz class $b$, the paper shows that the very special locus of the fixed-determinant infinite-level diamond $M^{\tau}_{b,[\mu]}$ over $\operatorname{Spd} C_p$ is exactly the locus of rank-one geometric points $z$ where the vector bundle $z^*E^{\circ}_{\max}$—built from the adjoint action of $b$ on the derived Lie algebra—admits zero as a Harder–Narasimhan slope (Corollary 2.3.7). Equivalently, those are the points where the intersection of the Hodge and Hodge–Tate period fibers is non-discrete. The paper conjectures that the structure morphism is cohomologically smooth on the open complement of the very

Load-bearing premise

The whole argument hinges on the claim that the differential structure attached to the moduli space by the Ivanov–Weinstein construction is the same as the differential structure computed by the general inscription formalism; if these two enrichments disagreed at even one point, the proof of the theorem would fail.

Editorial extensions

If this is right

Conjecture 1.1.4 now holds for all EL infinite-level Rapoport–Zink spaces, not just the basic case: the non-very-special locus is cohomologically smooth over $\operatorname{Spd} C_p$.
The very special locus has a purely geometric description as the locus where the intersection of the two period fibers is non-discrete, and equivalently where the associated modification of $G$-bundles has extra infinitesimal automorphisms.
In the EL case, very special points are exactly those whose associated $p$-divisible group has extra $B$-linear endomorphisms beyond the center of $B$, matching the earlier definition of the special locus in the basic case.
The Harder–Narasimhan slope condition is semicontinuous, so the very special locus is a closed subdiamond and its complement is an open subdiamond, giving a natural candidate for the cohomologically smooth locus in all cases.

Reading between the lines

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

If Conjecture 1.1.4 holds beyond the EL case, cohomological smoothness would reduce to a slope computation on the Fargues–Fontaine curve, turning a difficult geometric property into linear algebra of isocrystals; explicit Hodge–Newton reducible cases could serve as a test.
Proposition 2.3.5 suggests a possible strengthening: the cohomologically smooth locus might in general coincide with the locus where the inscribed tangent space is connected, not merely contain it, though the paper only establishes this equivalence pointwise and proves smoothness in the EL case.
The comparison of inscribed structures in Proposition 3.2.3 may extend to other minuscule cocharacter data, since the construction of $Z$ via exact sequences is not obviously limited to EL data; this offers a route toward the conjecture for Hodge-type or abelian-type local Shimura varieties.
The endomorphism characterization in Lemma 3.3.1 is computable in examples: for explicit $p$-divisible groups one can check whether the center of $B$ equals the full endomorphism ring, giving a practical test of the conjecture's prediction about the smooth locus.

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

What Howe does here is concrete: Theorem 1.1.5 removes the basic hypothesis from Ivanov–Weinstein's cohomological smoothness result, covering the full EL Rapoport–Zink case. The new criterion—very special locus equals the locus where the inscribed tangent space is disconnected—is genuinely useful, and Conjecture 1.1.4 is a sensible target. The proof structure is coherent: compute the tangent bundle as BC(E_max), use duality and p-adic Lie group arguments to identify very special points with the zero-slope condition, then invoke the Fargues–Scholze Jacobian criterion. Nothing is fitted, and the conjecture is not assumed; the circularity burden is low.

The soft spot is Proposition 3.2.3. The theorem's final step needs the inscribed moduli-of-sections structure on the Ivanov–Weinstein scheme Z to match the inscription formalism from [2] used in the tangent computation. The text calls the comparison 'almost formal' and 'immediate from the construction in [6, §5.6]', but no proof is given. Inscription is extra structure, not something an isomorphism of underlying diamonds carries automatically. If the two structures disagreed, the equality z*E_max = s*T_Z would fail and the Jacobian criterion would not apply on the non-very-special locus. The reader and the stress-test both flag this, and I think the worry lands. That said, the paper is transparent about the omission—it does not pretend to have written the identification out—and the surrounding argument is clean enough that the gap looks fillable rather than fatal.

I also note the heavy reliance on the author's unpublished preprint [2]. The tangent bundle computation is quoted from there. Self-citation is not a flaw here since [2] is independent work, but it does mean the full verification of Theorem 1.1.5 depends on two checks the present paper does not contain: the inscribed-structure identification and the correctness of [2, Cor. 9.2.3]. A careful referee should look at both.

This is for p-adic geometers working on local Shimura varieties and the geometrization program. I would send it to peer review with a referee who can cross-check the inscription theory. The main theorem is important if true, and the paper is honest about the deferred details.

Recommendation: accept a serious referee; the referee should spend time on Proposition 3.2.3.

Referee Report

3 major / 3 minor

Summary. The paper studies infinite-level moduli of mixed-characteristic local shtukas with one leg and fixed determinant, viewed as inscribed v-sheaves in the sense of the author's preprint [2]. It characterizes a closed 'very special' locus in terms of the Harder–Narasimhan slopes of the vector bundle E^○_max obtained from the inscribed tangent bundle: a rank-one point is very special exactly when zero is an HN slope (Prop. 2.3.5, Cor. 2.3.7). It then conjectures that the structure morphism to Spd C_p is cohomologically smooth on the open non-very-special locus (Conj. 1.1.4), and proves this in the EL Rapoport–Zink case (Thm. 1.1.5, Cor. 3.2.4) by identifying the inscribed moduli space with the moduli of sections of the Ivanov–Weinstein scheme Z and applying the Fargues–Scholze Jacobian criterion. The proof also gives an endomorphism characterization of the very special locus in this case (Lemma 3.3.1).

Significance. If fully substantiated, the result is a meaningful advance: it removes the basic hypothesis from the Ivanov–Weinstein cohomological smoothness theorem and, more importantly, proposes a clean geometric characterization of the smooth locus that is conjecturally valid for all reductive groups. The paper is honest about its architectural dependence on the author's own unpublished inscription theory [2], and the equivalence chain in Prop. 2.3.5 is a useful conceptual contribution in its own right. The main risk is not internal inconsistency but whether the omitted comparison of inscribed structures in Prop. 3.2.3 can be supplied. The paper would be publishable once that comparison and the vector-bundle step in Cor. 3.2.4 are proved in detail.

major comments (3)

[§1.2 and Prop. 3.2.3] The load-bearing step is the asserted identification between (a) the inscribed structure on M^τ_{b,[μ]} coming from [2, §9.4] and (b) the inscribed moduli-of-sections structure attached to the Ivanov–Weinstein scheme Z. The text calls this 'almost formal' and the proof of Prop. 3.2.3 says it is 'immediate from the construction in [6, §5.6]'. This is not a routine consequence of an isomorphism of underlying v-sheaves: inscription is extra structure, and the two objects are presented differently (moduli of exact sequences vs. moduli of sections of a scheme). The equality is exactly what converts the tangent computation TM^τ_{b,[μ]} = BC(E^○_max) from Lemma 2.2.1 into the Jacobian-criterion identity BC(s^*T_Z) = BC(z^*E^○_max) in Cor. 3.2.4. Without a proof, the identification could fail and the theorem would not follow. Please provide a complete argument, not a citation to [6] and [2] alon
[Cor. 3.2.4, final paragraph] The deduction from equality of Banach–Colmez spaces to equality of vector bundles is too compressed. The text says BC(s^*T_Z) = BC(z^*E^○_max), then uses full faithfulness on the non-negative-slope part and a dimension equality to conclude z^*E^○_max = s^*T_Z. But for this to work one must know that s^*T_Z has no negative-slope summands (or otherwise justify that equality of BC spaces upgrades to equality of vector bundles after passing to the non-negative part). If s^*T_Z has a negative-slope summand, BC(s^*T_Z) is typically insensitive to it, and the claimed equality with the non-very-special locus M^sm_Z is not established. This step is necessary for the identification of M^τ,non-vsp with the locus where the Jacobian criterion applies.
[Lemma 2.2.1 and [2]] The paper's central computation of the tangent bundle is inherited from [2, Cor. 9.2.3 and §9.4], and the comparison in Prop. 3.2.3 relies on [2, §4.4]. Since [2] is an unpublished preprint by the same author, the present manuscript is conditional on an unreviewed body of work in a way that is not merely cosmetic. I am not objecting to the use of the author's own prior work, but the referee cannot fully verify the main theorem without access to the details of the inscription formalism. The text should either state these computations explicitly enough to be checked, or clearly frame the theorem as conditional on [2] and provide the cited statements in an appendix.

minor comments (3)

[Lemma 2.2.2, proof] Typo: 'such a morphism does does not exist' should be 'does not exist'.
[Lemma 2.3.3, proof] The displayed commutative diagram contains garbled arrow notation ('/leftr⫯g⊸tl⫯ne'), which should be cleaned up.
[General notation] The symbol M^τ_{b,[μ]} is used both for the inscribed v-sheaf and for its underlying diamond; the difference is essential to the proof (e.g. in §2.1 and Prop. 3.2.3). Please use visibly distinct notation or explicitly say when the distinction is being suppressed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 1.1.5 is not assumed; the proof derives the very-special-locus criterion from prior parameter-free tangent computations and the Fargues-Scholze Jacobian criterion. Proposition 3.2.3 is an unproved compatibility check, but it is a proof gap rather than a circular reduction.

full rationale

The paper's central claim (Theorem 1.1.5 / Corollary 3.2.4) is not assumed anywhere. Conjecture 1.1.4 is stated but explicitly not used in the proof. The very special locus is first defined geometrically (a point is very special if it is not isolated in the intersection of its two period fibers), and Proposition 2.3.5 derives equivalence with the Harder-Narasimhan slope condition from the tangent-bundle computation imported from [2, Cor. 9.2.3] together with standard facts about dual bundles and p-adic Lie groups. This is a derivation, not a renaming or a fit. The main use of the author's own prior work is Lemma 2.2.1 and Lemma 2.3.2, which cite [2] for the computation of the inscribed tangent bundle BC(E^o_max). That citation is load-bearing but independent: [2] is a parameter-free prior theory of inscription and Banach-Colmez tangent bundles whose stated assumptions do not include the present conjecture or cohomological smoothness, so it functions as external support rather than circularity. The delicate step is Proposition 3.2.3, which asserts that the inscribed moduli-of-sections structure attached to Ivanov-Weinstein's scheme Z equals the inscribed structure on M^tau_{b,[mu]}. Section 1.2 calls this 'almost formal', and the proof says 'The identification with M^tau_{b,[mu]} is immediate from the construction in [6, §5.6]'. This is an asserted compatibility check and a genuine proof gap: if the two inscriptions disagreed, Corollary 3.2.4 would not follow. But it is not circular, because M^tau_{b,[mu]} is not defined as the moduli of sections of Z, and Z is not constructed to force the equality; the proposition compares two prior constructions from [6] and [2]. No fitted parameter is relabeled as a prediction, and no uniqueness theorem from the same authors is invoked to forbid alternatives. The theorem is also checked against the external benchmark [6, Theorem 1.0.1] in Remark 1.1.6 and Lemma 3.3.1, where the new geometric very-special condition is proven equivalent to Ivanov-Weinstein's endomorphism definition. Thus the paper's derivation chain is not circular; its principal risk is the unproved identification in Proposition 3.2.3, which is a completeness issue rather than a circularity issue.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

No free parameters: the paper is pure mathematics with no fitted constants. The axioms are the heavy external machinery: [2]'s inscription theory (author's own, 2025 preprint), [6]'s smooth scheme, [1]'s Jacobian criterion, [7]'s full faithfulness, and the Scholze-Weinstein classification. The new work is the equivalence chain of Proposition 2.3.5 and the comparison in Section 3. No new entities (particles, forces, conserved quantities) are introduced; the very special locus is a defined subdiamond, not a postulated object requiring independent evidence.

assumptions (7)

domain assumption Inscribed tangent bundle computation of [2, Cor 9.2.3]: T M_{b,[mu]} = BC(E_max), and the derivatives of the period maps identify the tangent space of the fiber intersection with BC(E_min).
Quoted in Lemmas 2.2.1 and 2.3.2 from the author's own 2025 preprint [2]; not re-derived in this paper.
domain assumption Fargues-Scholze Jacobian criterion and definition of the smooth locus M^sm_Z ([1, Def IV.4.1, Thm IV.4.2]).
Used in Corollary 3.2.4 to convert the positive-slope tangent condition into cohomological smoothness.
domain assumption Ivanov-Weinstein construction of a smooth quasi-projective scheme Z over X^alg_{C_p^flat} with moduli of sections isomorphic to M^tau_{b,[mu]} ([6, Sec 5.6]).
Proposition 3.2.3 says the identification is 'immediate' from this construction; the paper assumes [6]'s scheme exists on all EL Newton strata, not only the basic case.
domain assumption Full faithfulness of the functor from vector bundles of non-negative slope to Banach-Colmez spaces ([7, Thm 1.2]).
Used in Corollary 3.2.4 to pass from equality of Banach-Colmez spaces to equality of vector bundles.
standard math Killing form provides an isomorphism g^o =~ (g^o)^* for the semisimple Lie algebra g^o (Lemma 2.2.3).
Standard Lie theory; Remark 2.2.4 notes this is why G is required to be reductive.
standard math Semi-continuity of the Harder-Narasimhan polygon ([1, Thm I.3.4]).
Used in Corollary 2.3.7 to conclude the very special locus is closed.
domain assumption Scholze-Weinstein classification of p-divisible groups ([10]).
Used in Lemma 3.3.1 to identify Lie H_z with extra B-linear endomorphisms A_z/F.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A cohomological smoothness conjecture for moduli of mixed characteristic local shtukas with one leg." pith.science (2026). https://pith.science/paper/UKN2VZUF

@misc{pith2026250811595,
  author       = {Pith},
  title        = {Pith review of: A cohomological smoothness conjecture for moduli of mixed characteristic local shtukas with one leg},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UKN2VZUF}},
  note         = {Machine review of arXiv:2508.11595}
}

read the original abstract

We give a simple geometric characterization of the locus where the inscribed Banach--Colmez Tangent Spaces of moduli of mixed characteristic local shtukas with one leg and fixed determinant are connected. We conjecture that the structure morphism for the underlying diamond is cohomologically smooth over this locus and, applying the Fargues--Scholze Jacobian criterion, we prove this conjecture in the case of EL infinite level Rapoport--Zink spaces, generalizing a result of Ivanov--Weinstein in the basic case.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

12 extracted references · 11 canonical work pages

[2]

Inscription, twistors, and p-adic periods

Sean Howe. Inscription, twistors, and p-adic periods. arXiv, 2025

work page 2025
[6]

Ivanov and Jared Weinstein

Alexander B. Ivanov and Jared Weinstein. The smooth locus in infinite-level Rapoport-Zink spaces. Compos. Math., 156(9):1846–1872, 2020. 10 SEAN HOWE

work page 2020
[1]

Geometrization of the local Langlands correspondence

Laurent Fargues and Peter Scholze. Geometrization of the local Langlands correspondence. To appear in Asterisque, arXiv:2102.13459

arXiv
[3]

Admissible pairs and p-adic Hodge structures III: Variation and unlikely intersection

Sean Howe and Christian Klevdal. Admissible pairs and p-adic Hodge structures III: Variation and unlikely intersection. In preparation

work page
[4]

Admissible pairs and $p$-adic Hodge structures I: Transcendence of the de Rham lattice

Sean Howe and Christian Klevdal. Admissible pairs and p-adic Hodge structures I: Transcendence of the de Rham lattice. To appear in Algebra and Number Theory, arXiv:2308.11065, 2023

work page Pith review arXiv 2023
[5]

Admissible pairs and $p$-adic Hodge structures II: The bi-analytic Ax-Lindemann theorem

Sean Howe and Christian Klevdal. Admissible pairs and p-adic Hodge structures II: The bi-analytic Ax-Lindemann theorem. arXiv:2308.11064, 2023

work page Pith review arXiv 2023
[7]

Espaces de Banach-Colmez et faisceaux coh´ erents sur la courbe de Fargues- Fontaine

Arthur-C´ esar Le Bras. Espaces de Banach-Colmez et faisceaux coh´ erents sur la courbe de Fargues- Fontaine. Duke Math. J., 167(18):3455–3532, 2018

work page 2018
[8]

Rapoport and Th

M. Rapoport and Th. Zink. Period spaces forp-divisible groups, volume 141 of Annals of Mathematics Studies. Princeton University Press, Princeton, NJ, 1996

work page 1996

Show all 12 references

[9]

Towards a theory of local Shimura varieties

Michael Rapoport and Eva Viehmann. Towards a theory of local Shimura varieties. M¨ unster J. Math., 7(1):273–326, 2014

2014
[10]

Moduli of p-divisible groups

Peter Scholze and Jared Weinstein. Moduli of p-divisible groups. Camb. J. Math., 1(2):145–237, 2013

2013
[11]

Berkeley Lectures on p-adic Geometry: (AMS-207)

Peter Scholze and Jared Weinstein. Berkeley Lectures on p-adic Geometry: (AMS-207). Princeton University Press, 2020

2020
[12]

Semistable models for modular curves of arbitrary level

Jared Weinstein. Semistable models for modular curves of arbitrary level. Invent. Math., 205(2):459– 526, 2016

2016

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [2]

Inscription, twistors, and p-adic periods

Sean Howe. Inscription, twistors, and p-adic periods. arXiv, 2025

work page 2025

[2] [6]

Ivanov and Jared Weinstein

Alexander B. Ivanov and Jared Weinstein. The smooth locus in infinite-level Rapoport-Zink spaces. Compos. Math., 156(9):1846–1872, 2020. 10 SEAN HOWE

work page 2020

[3] [1]

Geometrization of the local Langlands correspondence

Laurent Fargues and Peter Scholze. Geometrization of the local Langlands correspondence. To appear in Asterisque, arXiv:2102.13459

arXiv

[4] [3]

Admissible pairs and p-adic Hodge structures III: Variation and unlikely intersection

Sean Howe and Christian Klevdal. Admissible pairs and p-adic Hodge structures III: Variation and unlikely intersection. In preparation

work page

[5] [4]

Admissible pairs and $p$-adic Hodge structures I: Transcendence of the de Rham lattice

Sean Howe and Christian Klevdal. Admissible pairs and p-adic Hodge structures I: Transcendence of the de Rham lattice. To appear in Algebra and Number Theory, arXiv:2308.11065, 2023

work page Pith review arXiv 2023

[6] [5]

Admissible pairs and $p$-adic Hodge structures II: The bi-analytic Ax-Lindemann theorem

Sean Howe and Christian Klevdal. Admissible pairs and p-adic Hodge structures II: The bi-analytic Ax-Lindemann theorem. arXiv:2308.11064, 2023

work page Pith review arXiv 2023

[7] [7]

Espaces de Banach-Colmez et faisceaux coh´ erents sur la courbe de Fargues- Fontaine

Arthur-C´ esar Le Bras. Espaces de Banach-Colmez et faisceaux coh´ erents sur la courbe de Fargues- Fontaine. Duke Math. J., 167(18):3455–3532, 2018

work page 2018

[8] [8]

Rapoport and Th

M. Rapoport and Th. Zink. Period spaces forp-divisible groups, volume 141 of Annals of Mathematics Studies. Princeton University Press, Princeton, NJ, 1996

work page 1996