REVIEW 4 major objections 5 minor 21 references
Convolution morphisms and Kottwitz conjecture
T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read The paper establishes the Kottwitz conjecture for all inner forms of GL3 with minuscule cocharacters, by comparing étale cohomology of local shtuka spaces.
desk verdict Serious preprint that proves new cases of the Kottwitz conjecture and gives explicit counterexamples; the argument is transparent in structure, rests on heavy Fargues-Scholze machinery, and deserves expert refereeing. 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 central object is the family of moduli spaces $Sht^{\mu_\bullet}_{b,b'}$ of mixed characteristic local shtukas, together with three morphisms between them. The convolution morphism composes successive modifications of $G$-bundles along the legs, and it reduces the cohomology of one shtuka space to a sum, over intermediate $\sigma$-conjugacy classes, of tensor products of cohomologies of simpler shtuka spaces, with coefficients $V^\lambda_{\mu_\bullet}$ supplied by the geometric Satake equivalence. The duality morphism, coming from a standard duality involution relating representations to contragredients, separates the summands corresponding to different $\lambda$; the twist morphism by central cocharacters handles central-character bookkeeping. The decomposition is stated as Proposition 5.1, and the compatibility of the duality involution with the geometric Satake commutativity constraint is Proposition 6.2.
What would settle it
Compute both sides of the isomorphism in Theorem 7.6(4) for $n=3$: take basic $b,b'$ with $\kappa(b)=1$, $\kappa(b')\equiv -1 \pmod 3$, the minuscule cocharacter $\mu=(1,1,0)$, and a discrete L-parameter $\varphi$. If the $R\Hom$ complex is not the predicted $J_b(F)\times W_F$ representation, the theorem is false.
Extended reading notes
Core claim
Conjecture 7.3 of the paper—a derived, representation-valued strengthening of the Kottwitz conjecture for moduli spaces of mixed characteristic local shtukas—is true in the cases covered by Theorem 7.6, and in particular for every inner form of $GL_n$ with $n \le 3$ and minuscule $\mu$. For basic $b,b'$, a discrete local L-parameter $\varphi$ with Langlands correspondents $\pi_b$ and $\pi_{b'}$, the statement is $$R\Hom_{J_{b'}(F)}(R\Gamma_c(Sht^\mu_{b,b'}), \pi_{b'}) \simeq \pi_b \boxtimes (r_\mu \circ \varphi)$$ as representations of $J_b(F) \times W_F$. For $GL_2$, the weaker numerical Conjecture 7.2 holds whenever the L-parameter is cuspidal or the group is not quasi-split, while for non-cuspidal parameters and non-minuscule $\mu$ the unmodified conjecture fails; the failure terms are invisible to traces on regular elliptic elements, so they are compatible with the earlier weak Kottwitz results. The paper also shows that a naive non-minuscule generalization of the Harris–Viehmann conjecture fails in Hodge–Newton irreducible cases.
Load-bearing premise
The load-bearing premise is that the cohomology theory for these moduli spaces has the same base-change, finiteness, and stratification behavior as ordinary étale cohomology; a gap in any of those properties would break the GL3 minuscule result.
Editorial extensions
If this is right
- For every inner form of $GL_3$, every basic $b$, every minuscule $\mu$, and every discrete L-parameter $\varphi$, Conjecture 7.3 holds exactly as stated.
- The proof uses moduli spaces for non-minuscule cocharacters even in minuscule statements, so non-basic intermediate contributions cannot be avoided in these cases.
- For $GL_2$, Conjecture 7.2 holds for all cuspidal L-parameters and all non-quasi-split inner forms, and the inductive formulas make the remaining cases computable.
- The unmodified Kottwitz conjecture fails in general for non-minuscule cocharacters with non-cuspidal L-parameters, and the error terms have zero trace on regular elliptic elements.
- A naive non-minuscule generalization of the Harris–Viehmann conjecture fails in Hodge–Newton irreducible examples.
Reading between the lines
- A natural next step is to test whether the same convolution decomposition yields an inductive proof for higher-rank minuscule cases; because the intermediate strata are automatically non-basic, the non-basic theory built here would be essential.
- The $GL_2$ failure terms are invisible to traces on regular elliptic elements, which suggests that the correct non-minuscule formulation may need a modified $r_\mu$ rather than the usual one.
- The failure of the Harris–Viehmann generalization in Hodge–Newton irreducible examples suggests that Hodge–Newton reducibility is not merely a convenient hypothesis but may be necessary for support statements about local shtuka cohomology.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper develops convolution, duality, and twist morphisms between moduli spaces of mixed characteristic local shtukas and uses them to relate the étale cohomology of different moduli spaces. The main structural result is a convolution formula (Proposition 5.1 and Corollary 5.2) expressing a sum of cohomology groups over intermediate σ-conjugacy classes in terms of geometric Satake data. The paper then applies this formalism to prove new cases of the Kottwitz conjecture: for all inner forms of GL₃ with minuscule cocharacters and discrete local L-parameters (Theorem 2 / Corollary 7.7), and for GL₂ in several non-minuscule cases, including a counterexample to the non-minuscule generalization when the L-parameter is not cuspidal. A Harris–Viehmann-type statement is also shown to fail outside Hodge–Newton reducibility. The methods rely on the Fargues–Scholze v-stack cohomology package and on earlier work of Dat, Gaisin–Imai, and Scholze–Weinstein, with several new compatibility and fiber computations supplied in the text.
Significance. If the central arguments are correct, the paper establishes the Kottwitz conjecture in genuinely new settings—notably all inner forms of GL₃ for minuscule μ—and, perhaps more strikingly, produces explicit failures of the conjecture in non-minuscule cuspidal-free cases, thereby refining the expectations in Fargues's geometrization program and in Hansen–Kaletha–Weinstein. The paper is also methodical: it introduces useful convolution and duality morphisms for local shtuka spaces that are likely to be of independent value, and it does not assume the Kottwitz conjecture as an input. The derivation of the main formula from geometric Satake, with no fitted parameters, is a substantial strength. However, the proof as written contains a small number of load-bearing steps that are summarized rather than fully derived; these need to be completed before the advertised theorems can be regarded as established.
major comments (4)
- [§7, proof of Theorem 7.6(4)] The decisive step of the proof is the separation of the two summands in the displayed equality after the application of Proposition 5.1 and Lemma 5.3. The text says only: "Using Proposition 6.2, we can separate the above equality to obtain the claim." This is the only point where a derived-category sum is promoted to individual isomorphisms, and it is precisely where a hidden contragredient twist or an extra scalar could change the conclusion. Please provide a complete derivation of the induced involution on the object R Hom_{J_{ι(b)}(F)}(RΓ_c(Sht^{μ1}_{b,1}) ⊗ RΓ_c(Sht^{μ1}_{1,ι(b)}) ⊗^L_{GL_n(F)} Q_ℓ, π_{ι(b)}) after Lemma 5.3, and identify explicitly the +1 eigenspace. Without this computation, Corollary 7.7 is not verifiable.
- [§6, Proposition 6.2] The proof of Proposition 6.2 is a single diagrammatic paragraph asserting that the displayed diagram is compatible with involutions. It does not spell out how the duality involution θ_b acts after applying R Hom_{J_{ι(b)}(F)}(-, π_{ι(b)}), nor does it show that the identification via Lemma 5.3 intertwines θ_b-induced and σ-induced involutions with the factor id_{π_b}. Since Proposition 6.2 is the load-bearing compatibility that makes Theorem 7.6(4) work, the proof should be expanded into a formal argument, or the statement should be restricted to the precise compatibility needed and proved there.
- [§5, Proposition 5.1 and Lemma 2.6] The finiteness of the index set I^{μ•}_{b0,bm} and the stratification of Sht^{μ•}_{b0,bm} are justified by Lemma 2.6, but Lemma 2.6 is stated under the hypothesis that b' is basic. Proposition 5.1 is stated for arbitrary b0 and bm, and the proof applies Lemma 2.6 to intermediate elements b_i that need not be basic. If a non-basic form of Lemma 2.6 is intended, it should be stated and proved; otherwise the statement of Proposition 5.1 should be restricted to the cases where the required basicity is available, and the consequences for Theorem 1 and its later uses should be reassessed.
- [§8, Proposition 8.2 and Proposition 8.4] Several computations in the inductive formulas are asserted without derivation. For example, in the proof of Proposition 8.2 the equality "the fiber of the natural morphism Sht^{(1,0)}_{b1,b} → Sht^{(1,0)}_{T,b1,b} is isomorphic to B^{φ=...}" is used to compute an R Hom, and in Proposition 8.4 the reduction to Corollary 7.7 is stated in a single sentence after a display. These fiber computations are load-bearing for Theorem 3 and for the claimed counterexamples. Please supply the missing computations or give precise references to the explicit Hecke-stack descriptions in [GI16], [Han16], or [FS21] where they are derived.
minor comments (5)
- [Abstract and §2] There are several typographical errors: "shtuka s" in the abstract, "isomorphim" in the paragraph after (2.1), and "ultilt" in §4. These should be corrected.
- [§7, proof of Theorem 7.6] The proof refers to "[Dat07, Tho´ er` eme A]" and "[Dat07, Tho´ er` eme 4.1.2]" with inconsistent accents; the reference formatting should be normalized.
- [§6, diagram preceding Proposition 6.2] The cartesian diagrams defining i1, j1, i2, j2 are visually confusing: the placement of the diagonal morphism from Spd Ě to the ambient product is hard to read, and the subscript Δ2 appears to be attached ambiguously. Replacing the diagrams with a precise description of the subdiamonds would improve readability.
- [Notation, §8] The notation R• Hom and R•Γ c is used interchangeably with R Hom and RΓ c without comment; since these are used in graded and derived settings, a sentence fixing the conventions would avoid ambiguity.
- [References] The reference [Han16] is listed as a preprint without an arXiv identifier or year of the current version; if it has appeared or been updated, the citation should be completed.
Circularity Check
No significant circularity: the Kottwitz cases are derived from the convolution/geometric Satake formalism and external prior results, not assumed as inputs.
full rationale
The central formula, Proposition 5.1/Corollary 5.2, is obtained by applying the geometric Satake equivalence and the section 3 convolution morphisms to a stratification of the moduli spaces; neither the statement nor the proof assumes the Kottwitz conjecture (Conjectures 7.2/7.3). Theorem 7.6(4) reduces the claimed R Hom computation to the minuscule case (2), which is taken from Dat's Theoreme A rather than from the conjecture under proof, and then uses the compatibility in Proposition 6.2 to split a derived-category sum. Even if a reader doubts the completeness of that compatibility proof, that is a technical correctness risk, not circularity: the individual isomorphisms being proved are not fed back as inputs. Self-citations to [GI16] and [Ima20] occur, but only for a technical lemma (Lemma 2.7), a definition of local L-parameters (Definition 7.1), and auxiliary GL2 computations in section 8; none of these is a re-statement or assumed version of the Kottwitz conjecture being established. There are no fitted parameters, no prediction that reduces by construction to an input, and no uniqueness or ansatz imported from the authors' own prior work in a way that forces the conclusion.
Assumptions & free parameters
assumptions (6)
- domain assumption The D^■ solid-sheaf cohomology formalism for small v-stacks and the f_♮ adjoints satisfy the projection formula, proper base change, and duality properties stated in FS21.
- domain assumption The geometric Satake equivalence S: Rep_{Q_ℓ}(\hat G ⋊ Q) → D^■(Hck_G, Q_ℓ) is symmetric monoidal and realizes convolution products with the stated commutativity constraint.
- domain assumption The moduli spaces Sht^{µ•}_{b,b'} are diamonds or spatial diamonds with the stated torsor structure, stratifications, and nonemptiness criterion [b] ∈ B(G, µ, [b']).
- domain assumption Lemma 2.7: for f: \tilde J_b^{>0} → *, one has f_♮(Λ) ≃ Λ[-2N_b].
- domain assumption The local Langlands correspondence for GL_n, the local Jacquet-Langlands correspondence, and Dat07's Theorem A are valid.
- domain assumption The relative homology computations in Lemma 1.3 for perfectoid polydiscs give f_♮Λ ≃ Λ(-d)[-2d] with Frobenius scaling by p^d.
Cite this review
Pith. "Pith review of Convolution morphisms and Kottwitz conjecture." pith.science (2026). https://pith.science/paper/QVJTV2X3
@misc{pith2026190902328,
author = {Pith},
title = {Pith review of: Convolution morphisms and Kottwitz conjecture},
year = {2026},
howpublished = {\url{https://pith.science/paper/QVJTV2X3}},
note = {Machine review of arXiv:1909.02328}
}
abstract
We define etale cohomology of the moduli spaces of mixed characteristic local shtukas so that it gives smooth representations including the case where the relevant elements of the Kottwitz set are both non-basic. Then we relate the etale cohomology of different moduli spaces of mixed characteristic local shtukas using convolution morphisms, duality morphisms and twist morphisms. As an application, we show the Kottwitz conjecture in some new cases including the cases for all inner forms of $\mathrm{GL}_3$ and minuscule cocharacters. We study also some non-minuscule cases and show that the Kottwitz conjecture is true for any inner form of $\mathrm{GL}_2$ and any cocharacter if the Langlands parameter is cuspidal. On the other hand, we show that the Kottwitz conjecture does not hold as it is in non-minuscule cases if the Langlands parameter is not cuspidal. Further, we show that a generalization of the Harris--Viehmann conjecture for the moduli spaces of mixed characteristic local shtukas does not hold in Hodge--Newton irreducible cases.
Reference graph
Works this paper leans on
-
[1]
Averaging functors in Fargues' program for GL_n
J. Ansch\"utz and A.-C. Le Bras, Averaging functors in Fargues' program for GL_n , 2021, arXiv:2104.04701
work page Pith review arXiv 2021
-
[2]
I. N. Bern s te n and A. V. Zelevinski , Representations of the group GL(n,F), where F is a local non- A rchimedean field, Uspehi Mat. Nauk 31 (1976), no. 3, 5--70
work page 1976
-
[3]
Casselman, A new nonunitarity argument for p -adic representations, J
W. Casselman, A new nonunitarity argument for p -adic representations, J. Fac. Sci. Univ. Tokyo Sect. IA Math. 28 (1982), no. 3, 907--928
work page 1982
-
[4]
A. Caraiani and P. Scholze, On the generic part of the cohomology of compact unitary S himura varieties, Ann. of Math. (2) 186 (2017), no. 3, 649--766
work page 2017
-
[5]
Dat, Th\' e orie de L ubin- T ate non-ab\' e lienne et repr\' e sentations elliptiques, Invent
J.-F. Dat, Th\' e orie de L ubin- T ate non-ab\' e lienne et repr\' e sentations elliptiques, Invent. Math. 169 (2007), no. 1, 75--152
work page 2007
-
[6]
Fargues, Geometrization of the local L anglands correspondence: An overview, 2016, arXiv:1602.00999
L. Fargues, Geometrization of the local L anglands correspondence: An overview, 2016, arXiv:1602.00999
arXiv 2016
-
[7]
L. Fargues and P. Scholze, Geometrization of the local L anglands correspondence, 2021, arXiv:2102.13459
arXiv 2021
-
[8]
I. Gaisin and N. Imai, Non-semi-stable loci in H ecke stacks and F argues' conjecture, 2016, arXiv:1608.07446
arXiv 2016
Show all 21 references
-
[9]
Hansen, Moduli of local shtukas and H arris's conjecture, 2016, preprint
D. Hansen, Moduli of local shtukas and H arris's conjecture, 2016, preprint
2016
-
[10]
Hansen, On the supercuspidal cohomology of basic local Shimura varieties, 2021, preprint
D. Hansen, On the supercuspidal cohomology of basic local Shimura varieties, 2021, preprint
2021
-
[11]
Hansen, T
D. Hansen, T. Kaletha and J. Weinstein, On the Kottwitz conjecture for local shtuka spaces, 2017, arXiv:1709.06651
2017 arXiv
-
[12]
Imai, Local L anglands correspondences in -adic coefficients, 2020, arXiv:2003.14154
N. Imai, Local L anglands correspondences in -adic coefficients, 2020, arXiv:2003.14154
2020 arXiv
-
[13]
R. E. Kottwitz, Isocrystals with additional structure, Compositio Math. 56 (1985), no. 2, 201--220
1985
-
[14]
R. E. Kottwitz, Isocrystals with additional structure. II , Compositio Math. 109 (1997), no. 3, 255--339
1997
-
[15]
Prasad, Generalizing the MVW involution, and the contragredient, Trans
D. Prasad, Generalizing the MVW involution, and the contragredient, Trans. Amer. Math. Soc. 372 (2019), no. 1, 615--633
2019
-
[16]
u rich, 1994), Birkh\
M. Rapoport, Non- A rchimedean period domains, in Proceedings of the I nternational C ongress of M athematicians, V ol. 1, 2 ( Z \" u rich, 1994), Birkh\" a user, Basel, 1995 pp. 423--434
1994
-
[17]
Renard, Repr\'esentations des groupes r\'eductifs p -adiques, vol
D. Renard, Repr\'esentations des groupes r\'eductifs p -adiques, vol. 17 of Cours Sp\'ecialis\'es, Soci\'et\'e Math\'ematique de France, Paris, 2010
2010
-
[18]
Rapoport and E
M. Rapoport and E. Viehmann, Towards a theory of local S himura varieties, M\"unster J. Math. 7 (2014), no. 1, 273--326
2014
-
[19]
Scholze, Etale cohomology of diamonds, 2017, arXiv:1709.07343
P. Scholze, Etale cohomology of diamonds, 2017, arXiv:1709.07343
2017 arXiv
-
[20]
Scholze and J
P. Scholze and J. Weinstein , Berkeley lectures on \(p\)-adic geometry , vol. 207, Princeton, NJ: Princeton University Press, 2020
2020
-
[21]
Zhu, Affine G rassmannians and the geometric S atake in mixed characteristic, Ann
X. Zhu, Affine G rassmannians and the geometric S atake in mixed characteristic, Ann. of Math. (2) 185 (2017), no. 2, 403--492
2017
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.