Pith. sign in

REVIEW 3 major objections 4 minor 23 references

An explicit decomposition of higher Deligne-Lsuztig representations

T0 review · 3 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 Under $p\ne 2$ and $q>3$, the geometric module $\kappa_\Lambda$ equals, up to a sign, the algebraic type twisted by the quadratic character $\epsilon_\Lambda$, making elliptic higher Deligne-Lusztig decompositions explicit.

desk verdict A strong, useful note that makes Nie's higher Deligne-Lusztig decomposition fully explicit, modulo two genuine proof gaps that are likely fixable. read the letter →

arxiv 2506.12918 v1 pith:N4EP2PGA submitted 2025-06-15 math.RT

Ben Liu , Sian Nie This is my paper · ORCID
classification math.RT MSC 22E5020C33
keywords higherDeligne-LusztigrepresentationsparahoricinductionYutypesWeil-Heisenbergrepresentationquadraticcharactersupercuspidalelliptictorip-adicgroups
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

Higher Deligne-Lusztig representations are cohomological objects attached to parahoric subgroups of $p$-adic groups, and the paper aims to identify the geometric pieces inside them with the algebraic pieces used to build supercuspidal representations. The main theorem states that the geometric module $\kappa_\Lambda$ obtained from a higher Deligne-Lusztig variety differs from the algebraic Weil-Heisenberg module $\kappa(\Lambda)$ by only a character; once the residue field is large enough ($q>3$ always suffices), that character is trivial after multiplication by the quadratic character $\epsilon_\Lambda$. The result turns a structural decomposition from previous work into an explicit formula whose summands are compact inductions of $\kappa(\Lambda)\otimes\epsilon_\Lambda\otimes\rho$, each irreducible and pairwise non-isomorphic. The same equality is then used to show that unramified Yu types occur in higher Deligne-Lusztig cohomology and that unramified regular supercuspidal representations are compact inductions of explicit higher Deligne-Lusztig representations up to a sign.

What carries the argument

The comparison runs on the Heisenberg representation. Both $(-1)^{d_\Lambda}\kappa_\Lambda$ and $\kappa(\Lambda)\otimes\epsilon_\Lambda$ restrict to the same irreducible Heisenberg module $\omega_\Lambda$ on $H_\Lambda^F$, with the same central character $\chi_\Lambda$; a general lemma then forces the two modules to differ by a one-dimensional character $\psi$ of the finite reductive quotient $K_\Lambda^F/H_\Lambda^F\cong (L_\Lambda)^F_0$. The remaining burden is to show $\psi$ is trivial, which is done by checking that $(L_\Lambda)^F_0$ is generated by its commutator subgroup and a maximally split torus $S_0^F$, and then comparing the trace formulas of Propositions 3.2 and 4.3 on $S_0^F$. The threshold $c_\Lambda\in\{2,3,4\}$ comes from a Dynkin-diagram case check that ensures such a torus generates up to commutators.

What would settle it

Take a residue field with $q=2$ or $q=3$, choose a torus element $\gamma$ that is not generic, and compute the two traces $\operatorname{tr}(\gamma;(-1)^{d_\Lambda}\kappa_\Lambda)$ and $\operatorname{tr}(\gamma;\kappa(\Lambda)\otimes\epsilon_\Lambda)$ from the displayed formulas; if they differ while the generic traces still agree, then $\psi$ is nontrivial and the conclusion $\psi=1$, together with Corollary 1.3, fails.

Watch

Extended reading notes

Core claim

The central claim is that the geometric representative $\kappa_\Lambda$ — defined as the alternating cohomology of the higher Deligne-Lusztig variety $Y_\Lambda$ cut by a character — is governed by the same Heisenberg representation that underlies the algebraic construction. Theorem 1.1 asserts that under $p\ne 2$ there is a character $\psi$ of $K_\Lambda^F$ such that $(-1)^{d_\Lambda}\kappa_\Lambda \cong \kappa(\Lambda)\otimes\epsilon_\Lambda\otimes\psi$, and that $\psi=1$ whenever $q\ge c_\Lambda$, with $2\le c_\Lambda\le4$. Since earlier work had already reduced an elliptic higher Deligne-Lusztig representation to an induction of $\kappa_\Lambda$, the case $\psi=1$ produces the explicit decomposition $R^G_{T,r}(\phi)=(-1)^{d_\Lambda}\sum_\rho m_\rho \operatorname{ind}^{G(O_k)}_{K(O_k)} \kappa(\Lambda)\otimes\epsilon_\Lambda\otimes\rho$, with every displayed summand irreducible and pairwise non-isomorphic. The paper also derives that each unramified Yu type appears in the cohomology of a higher Deligne-Lusztig variety, and that each unramified regular supercuspidal representation is the compact induction of a specified higher Deligne-Lusztig representation up to a sign.

Load-bearing premise

The proof that the character $\psi$ is trivial assumes that a trace identity established only for generic torus elements extends to every torus element, yet no lemma or cited result in the paper states that extension.

Editorial extensions

If this is right

  • Corollary 1.3 gives a fully explicit irreducible decomposition of every elliptic higher Deligne-Lusztig representation once $q\ge c_\Lambda$, with summands indexed by the irreducible constituents of a classical Deligne-Lusztig representation.
  • Theorem 1.5 shows that every unramified Yu-type supercuspidal representation appears in the cohomology of a higher Deligne-Lusztig variety, and that suitable elliptic tori and depth-zero characters always exist.
  • Corollary 1.7 realizes each unramified regular supercuspidal representation as the compact induction of a specified higher Deligne-Lusztig representation, up to an explicit sign and the character twist $\epsilon[\phi]$.
  • For toral characters the threshold is $c_\Lambda=2$, so the explicit decomposition holds with no restriction on $q$ beyond $p\ne2$ and the standing hypotheses on $p$.
  • The identification $\kappa_\Lambda\cong\kappa(\Lambda)\otimes\epsilon_\Lambda$ lets one transfer computations between the cohomology of higher Deligne-Lusztig varieties and the explicit algebraic formulas of the type construction.

Reading between the lines

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

  • The uniform bound $c_\Lambda\le4$ implies the main equality is unconditional for every residue field of size at least $5$; only $q=2,3$ can require case-by-case checks, so the theorem's true domain may be wider than the stated thresholds.
  • The unproved extension of Proposition 3.2 off the generic torus elements is the one place a counterexample could hide; checking the trace identity on a single non-generic element for $q=2$ or $q=3$ would test whether the full decomposition survives there.
  • Carrying the same character comparison to modular coefficients, as the paper suggests in Remark 1.2, would turn the explicit decomposition into a modular decomposition, with the sign $d_\Lambda$ and the quadratic twist $\epsilon_\Lambda$ controlling what happens under reduction modulo $\ell$.
  • The sign and quadratic twist likely record the difference between geometric compact induction and the packet normalization of the algebraic construction, so the equality may also serve as a bridge for endoscopic character identities.
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.

Referee Report

3 major / 4 minor

Summary. The paper compares two objects attached to an unramified generic datum Λ: the geometrically defined higher Deligne-Lusztig representation κ_Λ (cohomology of the variety Y_Λ) and the algebraic Weil-Heisenberg representation κ(Λ) twisted by the Fintzen-Kaletha-Spice quadratic character ϵ_Λ. The main result (Theorem 1.1) asserts that these differ by a character ψ of Z_G(k)K(O_k), and that under a mild condition q ≥ c_Λ with 2 ≤ c_Λ ≤ 4 this character is trivial. Combining with the second author's prior decomposition theorem [22], the paper derives an explicit irreducible decomposition of elliptic higher Deligne-Lusztig representations (Corollary 1.3) and applications to unramified Yu types and regular supercuspidal representations (Theorem 1.5, Corollary 1.7). The proof strategy is to show that both representations extend the same Heisenberg module ω_Λ, so Proposition 5.2 yields a character ψ; the remaining task is to prove ψ is trivial by comparing trace formulas on a maximally split torus.

Significance. If the comparison is correct, the paper gives a fully explicit algebraic characterization of elliptic higher Deligne-Lusztig representations, identifying the geometric κ_Λ with κ(Λ)⊗ϵ_Λ up to a character that vanishes under a mild hypothesis on q. The approach is conceptually clean and the trace formulas are explicit and nonzero. The paper also gives concrete applications relating unramified Yu types and Kaletha's regular supercuspidal representations to higher Deligne-Lusztig cohomology. The reliance on prior work [22] and the relation to concurrent work [8] are acknowledged honestly. The main unresolved issue is the extension of the trace identity from the very regular locus to the full torus, which is load-bearing for the triviality of ψ and hence for the explicit decomposition.

major comments (3)
  1. [§5.2, Theorem 5.4] The proof applies Proposition 3.2 to an arbitrary element γ∈S^F, although Proposition 3.2 is stated only for unramified very regular γ. No argument is given that the identity tr(γ;κ(Λ)⊗ϵ_Λ)=(-1)^{r(S,Λ)}tr(γ;κ_Λ) extends from the very regular locus to all of S^F, for instance by showing both sides are polynomial class functions in the sense of DeBacker-Spice. This extension is the only mechanism in the proof that forces ψ(γ)=1 for every γ, so the conclusion ψ=1, and with it Corollary 1.3 and Corollary 1.7, does not follow from the text as written. The first half of Theorem 1.1, which relies only on Proposition 5.2, is not affected.
  2. [§4.2, Proposition 4.3] The proof asserts that \bar{H}^{γ_s}_Λ = \bar{S}^{0+} for the semisimple part γ_s of an arbitrary γ∈S^F. For a non-very-regular γ_s, the centralizer of γ_s in H_Λ/K^+_Λ can be strictly larger than \bar{S}^{0+}, which would change both the dimension factor |((H_Λ/K^+_Λ)^γ)^F|^{1/2} and the parity r(S,Λ,γ). The proposition therefore appears to require a very-regular hypothesis that is not stated; either justify the equality for all γ_s or restrict Proposition 4.3 and adjust the proof of Theorem 5.4 accordingly.
  3. [§5.2, Proposition 5.3] The proof asserts, via a 'direct case-by-case analysis on the irreducible Dynkin diagrams,' that the condition α(z)≠1 for all roots α is satisfied for q≥4, and hence c_Λ≤4. This case analysis is not displayed, and the second part of Theorem 1.1 as well as Corollaries 1.3 and 1.7 depend on the resulting threshold q≥c_Λ. Please include the case analysis, or replace it with a uniform verifiable argument; even the uniform bound q≥4 would suffice if proved.
minor comments (4)
  1. [Abstract and title] There are several typos: 'Deligne-Lsuztig' should be 'Deligne-Lusztig'; 'differs' should be 'differ'; 'paly' should be 'play'; 'sytandard' should be 'standard'; 'cupidal' should be 'cuspidal'; 'shcemes' should be 'schemes'.
  2. [§1.1] The phrase 'attached to attached to elliptic tori' is duplicated, and the sentence beginning 'ϕ:T(k)→ Q_ℓ^× of depth ⩽ r ∈ Z_≥0' is grammatically incomplete.
  3. [§3.3, Proposition 3.2] The term 'unramified very regular' is not defined in this paper; please add a definition or a precise reference to [1] or [12] so that the scope of the trace formula is unambiguous.
  4. [§5.2, Proposition 5.3] The notation for the opposite maximal unipotent subgroups V_0 and \bar{V}_0 is introduced without explaining how they are chosen to be F-stable and normalized by S_0; a brief justification or citation would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the geometric κ_Λ and algebraic κ(Λ)⊗ϵ_Λ are defined independently, and the comparison is a theorem relying on cited prior work rather than on its own conclusion.

full rationale

The paper's central claim compares a geometrically defined virtual module κ_Λ, obtained from the cohomology of the higher Deligne-Lusztig variety Y_Λ, with an algebraically defined module κ(Λ)⊗ϵ_Λ, built from Yu's Weil-Heisenberg representation and the Fintzen-Kaletha-Spice quadratic character. Neither object is fitted or defined in terms of the other, so the identification is not self-definitional, and no data are fitted and then renamed as a prediction. The main dependence is on the second author's prior paper [22] for the decomposition R^G_T,r(ϕ) = ind κ_Λ⊗R^{G0}_{T,0}(ϕ_-1) and for the structural properties of κ_Λ used in Theorem 4.1; these are cited theorems with stated hypotheses and proofs, not unverified assertions smuggled in by citation, so they are legitimate mathematical dependencies. The proof of Theorem 5.4 does contain a non-circular gap: Proposition 3.2 is stated only for unramified very regular elements γ∈S^F, yet it is applied to every γ∈S^F to conclude ψ(γ)=1, and Proposition 5.3's Dynkin-diagram case check is not displayed. These are correctness or completeness concerns rather than circularity. Overall, the derivation chain has independent content and the central claim does not reduce to its own inputs, so the circularity score is 0.

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

All heavy machinery is imported from prior published or preprint work: Nie's decomposition and concentration theorems [22], Yu's type construction [23], Kaletha's Howe factorizations [19], the Fintzen-Kaletha-Spice quadratic character [15], and the Adler-Spice/DeBacker-Spice character formulas [1], [12]. Each is a parameter-free derivation with stated hypotheses, so this is dependence, not circularity. The one genuinely ad hoc premise is the unstated extension of the very-regular trace formula to all of S^F. There are no data-fitted free parameters and no invented entities: κ_Λ, κ(Λ), ϵ_Λ, and ψ are all canonically constructed objects within existing formalism.

assumptions (6)
  • domain assumption Condition (*): p ≠ 2 is not a bad prime for G and p does not divide |π_1(G_der)|·|π_1(Ĝ_der)|.
    Invoked throughout §1.1 and §6 to guarantee the decomposition theorem of [22], existence of Howe factorizations [19], and Yu's construction [23] apply. It is a standard tame hypothesis, not proved in this note.
  • domain assumption Every depth character admits a Howe factorization (Λ, ϕ_{-1}) in the sense of [19, §4].
    Used in §6.2 to write ϕ = ∏_{i=-1}^{d} ϕ_i|_{T^F} with a generic datum Λ; the entire decomposition framework depends on this existence result.
  • domain assumption The decomposition theorem of [22, Theorem 1.6]: R^G_{T,r}(ϕ) = ind^{G^F}_{K^F_Λ} κ_Λ ⊗ R^{L_Λ}_{T,0}(ϕ_{-1}).
    Prior theorem of the second author, restated as Theorem 6.1; it is the geometric engine of the paper. It is a parameter-free derivation with stated hypotheses, so reliance on it is dependence rather than circularity, but it is not reproved here.
  • domain assumption Concentration and parity statements of Theorem 4.1: H^i_c(Y_Λ^γ, Q_ℓ)[χ_Λ] is nonzero in exactly one degree N, with N ≡ r(S,Λ,γ) mod 2 and dimension |((H_Λ/K^+_Λ)^γ)^F|^{1/2}.
    Proved in §4.1 only in sketch: the parity and dimension formulas defer to 'arguments in [22, §6.3]'. Proposition 4.3 and hence the main comparison rest on it.
  • standard math The trace formula for κ(Λ) quoted as Proposition 3.2, from [1, Proposition 3.8] and [12, Proposition 4.3.8].
    Published character formulas for supercuspidal-type representations; this is the algebraic side of the comparison. The proposition is stated for unramified very regular γ, which matters for how it is used in Theorem 5.4.
  • ad hoc to paper Unstated extension: a trace identity proved on the very regular locus of S^F determines the character ψ on all of S^F ⊂ (L_Λ)^F_0.
    The proof of Theorem 5.4 needs ψ = 1 on the full finite torus but cites only the very-regular formula; the extension is neither proved nor cited in the text. This is the most fragile implicit premise in the note.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An explicit decomposition of higher Deligne-Lsuztig representations." pith.science (2026). https://pith.science/paper/N4EP2PGA

@misc{pith2026250612918,
  author       = {Pith},
  title        = {Pith review of: An explicit decomposition of higher Deligne-Lsuztig representations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/N4EP2PGA}},
  note         = {Machine review of arXiv:2506.12918}
}
abstract

In a previous paper, the second named author obtains a decomposition of an elliptic higher Deligne-Lusztig representation into irreducible summands, which are built in the same way as Yu types using a geometric analog $\kappa'$ of the Weil-Heisenberg representation $\kappa$. In this note, we show that $\kappa'$ and $\kappa$ differs by a character $\chi$. Moreover, under a mild condition on the cardinality $q$ of the residue field (for instance $q > 3$), we show that $\chi$ equals the quadratic character constructed by Fintzen-Kaletha-Spice, which gives an explicit irreducible decomposition result on elliptic higher Deligne-Lusztig representations. As an application, we deduce (under the mild condition on $q$) that each unramified Yu type appears in the cohomology of higher Deligne-Lusztig varieties, and each unramified Kaletha's regular supercuspidal representation is the compact induction of a specified higher Deligne-Lusztig representation up to a sign.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 22 canonical work pages

  1. [22]

    Decomposition of higher Deligne-Lusztig representations

    S. Nie,Decomposition of higher Deligne-Lusztig representations, arXiv:2406.06430

    work page Pith review arXiv

  2. [8]

    ,Green functions for positive-depth Deligne-Lusztig induction, arXiv:2506.04449

    arXiv

  3. [1]

    Adler, L

    J. Adler, L. Spice,Supercuspidal characters of reductivep-adic groups, Amer. J. Math. 131 (2009), 1137–1210

    work page 2009

  4. [2]

    Boyarchenko and J

    M. Boyarchenko and J. Weinstein,Maximal varieties and the local Langlands correspondence forGL(n), J. Amer. Math. Soc. 29 (2016), 177–236

    work page 2016

  5. [3]

    Chan,The cohomology of semi-infinite Deligne-Lusztig varieties, J

    C. Chan,The cohomology of semi-infinite Deligne-Lusztig varieties, J. Reine Angew. Math. 768 (2020), 93–147

    work page 2020

  6. [4]

    ,The scalar product formula for parahoric Deligne-Lusztig inductions, arXiv:2405.00671

    work page arXiv

  7. [5]

    Chan and A

    C. Chan and A. Ivanov,Cohomological representations of parahoric subgroups, Represent. Theory 25 (2021), 1–26

    work page 2021

  8. [6]

    Chan and A

    C. Chan and A. Ivanov,On loop Deligne-Lusztig varieties of Coxeter-type for inner forms ofGL n, Camb. J. Math. 11 (2023), 441–505

    work page 2023

Show all 23 references
  1. [7]

    Chan and M

    C. Chan and M. Oi,Geometric L-packets of Howe-unramified toral supercuspidal representations, J. Eur. Math. Soc. (2023), 1–62

    2023

  2. [9]

    Chen and A

    Z. Chen and A. Stasinski,The algebraisation of higher Deligne-Lusztig repre- sentations, Selecta Math. (N.S.) 23 (2017), 2907–2926

    2017

  3. [10]

    ,The algebraisation of higher level Deligne-Lusztig representations II: odd levels, arXiv: 2311.05354

    arXiv

  4. [11]

    Deligne and G

    P. Deligne and G. Lusztig,Representations of reductive groups over finite fields, Ann. Math. 103 (1976), 103–161

    1976

  5. [12]

    DeBacker, L

    S. DeBacker, L. Spice,Stability of character sums for positive-depth supercuspi- dal representations, J. Reine Angew. Math. 742 (2018), 47–78

    2018

  6. [13]

    Fintzen,On the construction of tame supercuspidal representations, Compos

    J. Fintzen,On the construction of tame supercuspidal representations, Compos. Math. 157 (2021), 2733–2746

    2021

  7. [14]

    ,Types for tamep-adic groups, Ann. Math. (2) 193 (2021), 303–346

    2021

  8. [15]

    Fintzen, T

    J. Fintzen, T. Kaletha and L. Spice,A twisted Yu construction, Harish-Chandra characters, and endoscopy, Duke Math. J. 172 (12), 2241–2301

  9. [16]

    G´ erardin,Construction de s´ eries discr` etesp-adiques, Lecture Notes in Math- ematics, Vol

    P. G´ erardin,Construction de s´ eries discr` etesp-adiques, Lecture Notes in Math- ematics, Vol. 462. Springer-Verlag, Berlin-New York, 1975

    1975

  10. [17]

    Ivanov, S

    A. Ivanov, S. Nie,The cohomology ofp-adic Deligne-Lusztig shcemes of Coxeter type, arXiv:2402.09017. 15

    arXiv

  11. [18]

    ,Convex elements and deep level Deligne-Lusztig varieties, arXiv:2503.13412

    arXiv

  12. [19]

    Kaletha,Regular supercuspidal representations, J

    T. Kaletha,Regular supercuspidal representations, J. Amer. Math. Soc. 32 (2019), 1071–1170

    2019

  13. [20]

    Kim,Supercuspidal representations: an exhaustion theorem, J

    J.-L. Kim,Supercuspidal representations: an exhaustion theorem, J. Amer. Math. Soc. 20 (2007), 273–320

    2007

  14. [21]

    Lusztig,Some remarks on the supercuspidal representations ofp-adic semisimple groups, Proc

    G. Lusztig,Some remarks on the supercuspidal representations ofp-adic semisimple groups, Proc. Sympos. Pure Math. 1979, 171–175

    1979

  15. [23]

    Yu,Construction of tame supercuspidal representations, J

    J.-K. Yu,Construction of tame supercuspidal representations, J. Amer. Math. Soc. 14 (2001), 579–622. Academy of Mathematics and Systems Science, Chinese Academy of Sci- ences, Beijing 100190, China Email address:liubenmath@gmail.com Academy of Mathematics and Systems Science, ...

    2001

Pith tools

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