---
title: Fine Deligne–Lusztig Varieties
url: https://www.emergentmind.com/topics/fine-deligne-lusztig-varieties
type: topic
---

# Fine Deligne–Lusztig Varieties

Fine Deligne–Lusztig varieties are locally closed subvarieties of partial flag varieties attached to a connected reductive group over a finite field, a parabolic type, and a Frobenius-twisted relative position. They generalize the classical Deligne–Lusztig varieties from the full flag variety to partial flag varieties, and in Coxeter-type settings they also appear as the geometric pieces governing stratifications of affine Deligne–Lusztig varieties and basic loci of Shimura-theoretic moduli spaces. Their study lies at the intersection of finite reductive groups, affine flag geometry, Bruhat–Tits theory, and arithmetic intersection theory [1901.02870][1802.02225].

## 1. Definition in partial flag varieties

Let $\mathbb{F}_q$ be a finite field, let $k=\overline{\mathbb{F}_q}$, and let $\sigma$ be the Frobenius automorphism. If $\mathbb{G}$ is a connected reductive group over $\mathbb{F}_q$, with $G=\mathbb{G}_k$, Weyl group $W$, set of simple reflections $\mathbb{S}$, and a $\sigma$-stable Borel $B$ containing a maximal torus $T$, then for $J\subset \mathbb{S}$ and $w\in {}^J W$ the fine Deligne–Lusztig variety is
\[
X_{J,w}
=
\left\{
gP_J\in G/P_J \mid g^{-1}\sigma(g)\in P_J\cdot_\sigma BwB
\right\}.
\]
Here $P_J$ is the parabolic corresponding to $J$, and ${}^J W$ denotes the minimal-length representatives for $W_J\backslash W$ [1901.02870].

These varieties partition the partial flag variety:
\[
G/P_J=\bigsqcup_{w\in {}^J W} X_{J,w}.
\]
When $J=\emptyset$, the construction recovers the classical Deligne–Lusztig variety $X_w$ in the full flag variety. In this sense, fine Deligne–Lusztig varieties generalize the classical, or coarse, theory by partitioning partial flag varieties rather than only full flag varieties.

The closure relation is controlled by a partial order:
\[
\overline{X_{J,w}}=\bigsqcup_{w'\le_{J,\sigma} w} X_{J,w'}.
\]
This gives a precise replacement, in the partial-flag setting, for the Bruhat-type closure calculus familiar from classical Deligne–Lusztig theory. A common simplification is to identify fine Deligne–Lusztig varieties with ordinary Deligne–Lusztig varieties on a partial flag variety; the formula above shows that the fine theory carries a genuinely richer stratification indexed by ${}^J W$ and the order $\le_{J,\sigma}$.

## 2. $\sigma$-unbranched data, closure stratifications, and parabolic induction

A particularly tractable class arises from a $\sigma$-unbranched datum $(J,\mathscr{L})$. The combinatorial condition is that $\mathbb{S}-J$ is contained in a single $\sigma$-orbit, while $\mathscr{L}$ is a connected, simply-laced sub-diagram of the Dynkin diagram with no branches, each node representing a distinct $\sigma$-orbit, and one end-node lying in $\mathbb{S}-J$ [1901.02870].

Under this hypothesis, the closure of a distinguished fine Deligne–Lusztig variety admits a chain stratification:
\[
\overline{X_{J,w_1}}=\bigsqcup_i X_{J,w_i}.
\]
Each stratum $X_{J,w_i}$ is again a fine Deligne–Lusztig variety. This gives a recursive geometry in which the closure is built from finitely many finer pieces of the same type.

The decisive structural statement is that every such $X_{J,w_i}$ is related by parabolic induction to a classical Deligne–Lusztig variety for a Levi subgroup $G_i$:
\[
X_{J,w_i}\cong \mathbb{G}(\mathbb{F}_q)\times^{\mathbb{P}_i(\mathbb{F}_q)} X^{\mathbb{G}_i}_{w_i}.
\]
Here $\mathbb{P}_i$ is an explicit nested parabolic, and $X^{\mathbb{G}_i}_{w_i}$ is a classical Deligne–Lusztig variety for the finite reductive group $\mathbb{G}_i$. The significance of this formula is methodological as much as geometric: it reduces questions about closures, cohomology, and fixed points of fine Deligne–Lusztig varieties to the classical theory after passage to suitable Levi subgroups.

This inductive description also clarifies the adjective “fine.” The partition of $G/P_J$ is not merely coarse orbit combinatorics; it is compatible with a nested system of Levi-theoretic reductions, so the geometry of a partial flag variety is resolved into layers controlled by finite reductive groups of smaller rank.

## 3. Relation with affine Deligne–Lusztig stratifications

Affine Deligne–Lusztig varieties are defined in affine flag varieties or affine Grassmannians and play a central role in the geometry of moduli spaces of $p$-divisible groups and in the reduction theory of Shimura varieties. For an Iwahori subgroup $I$ and $b\in G(F)$, one has
\[
X_w(b)=\{g\in G(F)/I \mid g^{-1}b\sigma(g)\in IwI\},
\]
with analogous definitions at parahoric level. In this setting, the comparison between fine geometric decompositions and group-theoretic stratifications is especially important [1802.02225].

For a basic element $b$, let
\[
J_b(F)=\{g\in G(F)\mid g^{-1}b\sigma(g)=b\}
\]
be the $\sigma$-centralizer. Chen–Viehmann’s $J$-stratification is defined using relative position with respect to elements of $J_b(F)$:
\[
S_{J_b}(w)
:=
\Big\{
x\in G(F)/K \ \Big|\ \forall j\in J_b(F): \operatorname{inv}_K(j,x)=w_j
\Big\}.
\]
The resulting strata are locally closed, and only finitely many of them meet any quasi-compact subset. At the same time, the theory records an important limitation: in general, the $J$-stratification need not be a stratification in the strict sense, since closures of strata are not necessarily unions of strata.

In Coxeter type, however, a stronger picture emerges. Görtz proves that for $(G,\mu,K)$ of Coxeter type and $b$ basic, the $J$-stratification coincides with the Bruhat–Tits stratification of $X(\mu,b)_K$. Each Bruhat–Tits stratum is the image of a classical Deligne–Lusztig variety for the reductive quotient of a parahoric in $J_b$, and the closure of a stratum is a union of strata. In the corresponding Rapoport–Zink setting, the fine Ekedahl–Oort stratification coincides with the Bruhat–Tits stratification. This is one of the principal mechanisms by which fine Deligne–Lusztig geometry enters arithmetic geometry: the local pieces of an affine or moduli-theoretic space are literally classical Deligne–Lusztig varieties indexed by building-theoretic data.

## 4. Finite Coxeter type and hyperspecial affine Grassmannians for $GL_n$

A complementary development appears in the study of affine Deligne–Lusztig varieties for $G=GL_n$ and basic $b$ at hyperspecial level. For a dominant cocharacter $\lambda\in X_*(T)_+$,
\[
X_\lambda(b)=\{gK\in G(L)/K \mid g^{-1}b\sigma(g)\in Kt^\lambda K\}.
\]
The paper of 2022 introduces a new condition on $\lambda$, called being of finite Coxeter type, which is a variant of Coxeter type based on the finite Weyl group $W_0=S_n$ rather than the affine Weyl group [2204.10799].

If ${}^S(\lambda)^\circ_{\mathrm{cox}}$ denotes the subset of ${}^S(\lambda)^\circ$ whose $W_0$-parts are (twisted) Coxeter elements, then $\lambda$ is of finite Coxeter type if, for the unique basic $\sigma$-conjugacy class $[b]$ in $B(G,\lambda)$,
\[
X_\lambda(b)=\bigsqcup_{w\in {}^S(\lambda)^\circ_{\mathrm{cox}}}\pi(X_w(b)).
\]
The main classification theorem states that, up to central translates $m=(m,\ldots,m)$, $\lambda_{\mathrm{ad}}$ is of finite Coxeter type if and only if it has one of the following forms:
\[
\left\{
\begin{aligned}
&\big((n-1)r+\kappa,-r,-r,\ldots,-r\big),\\
&\big(r,r,\ldots,r,-(n-1)r-\kappa\big),\\
&\big((n-1)r+1+\kappa,-r,\ldots,-r,-r-1\big),\\
&\big(r+1,r,\ldots,r,-(n-1)r-1-\kappa\big),
\end{aligned}
\right.
\]
for $0\le \kappa<n$, with $r\ge 1$ if $\kappa=0$ and $r\ge 0$ if $1\le \kappa<n$.

For such $\lambda$, the geometric structure simplifies drastically. If $X_\lambda(b)\neq\emptyset$, then
\[
X_\lambda(b)\cong \bigsqcup_{J/K} Y\times \mathbb{A}^d,
\]
where $J=\{g\in G(L)\mid g^{-1}b\sigma(g)=b\}$, $Y$ is a classical finite Deligne–Lusztig variety associated to a Coxeter element in $W_0$, and $\mathbb{A}^d$ is a finite-dimensional affine space. In the more general family one has
\[
X_\lambda(b)\cong \bigsqcup_{j=1}^{n-1}\bigsqcup_{J/K} Y_j\times \mathbb{A}^{d_j}.
\]
Each irreducible component is locally closed in a Schubert cell and is individually described as $Y\times \mathbb{A}^d$.

Within the affine literature, this is one of the explicit senses in which “fine” Deligne–Lusztig geometry appears: the affine variety decomposes into classical Deligne–Lusztig factors times affine spaces. The proof leverages Iwahori-level descriptions using the work of Chan–Ivanov and then shows that, under the finite Coxeter type condition, the pushforward to the hyperspecial quotient preserves the product decomposition. The dimension is governed by
\[
\dim X_\lambda(b)=\langle \rho,\lambda-\nu_b\rangle-\frac12\mathrm{def}(b),
\]
so the affine-space factor can be read off from the standard group-theoretic invariants after subtracting the dimension of the finite Deligne–Lusztig factor.

## 5. Cohomology, character formulas, and arithmetic applications

The paper “Fine Deligne-Lusztig varieties and Arithmetic Fundamental Lemmas” proves a character formula for some closed fine Deligne–Lusztig varieties and applies it to fixed-point computations arising from basic loci of Shimura varieties of Coxeter type [1901.02870].

For a regular element $g\in \mathbb{G}(\mathbb{F}_q)$, the trace on the cohomology of the closure $\overline{X_{J,w_1}}$ is expressed as
\[
\operatorname{tr}(g, J, \mathscr{L})
=
\sum_i \sum_{\gamma\in \Gamma_i}
\#\mathcal{M}_i^{g,\gamma}\cdot
\frac{|\mathbb G_{i,\gamma}(\mathbb{F}_q)|}{|\mathbb G_{i,\gamma}^0(\mathbb{F}_q)|}
\cdot
\big|T_i\cap ({}^{\mathbb G_i(\mathbb{F}_q)}\gamma)\big|.
\]
The summation runs over the finitely many strata in the closure and over representatives in suitable tori of Levi subgroups. This formula reduces the trace computation for a closed fine Deligne–Lusztig variety to explicit group-theoretic data attached to Levi pieces, thereby formalizing the inductive mechanism suggested by the parabolic-induction description.

In the minuscule cases relevant to arithmetic applications, the core geometric move is the reduction of an intersection problem on a Rapoport–Zink space to a fixed-point problem on a finite variety $\mathcal V$. By Lefschetz,
\[
(g)=\operatorname{tr}(\bar g\mid \mathbf H^*(\mathcal V)).
\]
The variety $\mathcal V$ is stratified by fine Deligne–Lusztig pieces, and the character formula evaluates the trace by passing through those strata.

This leads to arithmetic intersection formulas for diagonal cycles on unitary and GSpin Rapoport–Zink spaces arising from the arithmetic Gan–Gross–Prasad conjectures. In particular, the arithmetic fundamental lemma in the minuscule case is proved without assumptions on the residual characteristic. The explicit formula quoted in the summary is
\[
(g)=\frac{m_{Q_0}+1}{2}\cdot \deg Q_0\cdot \mathsf{NSR}_2(f),
\]
where $Q_0$ is the unique monic irreducible self-reciprocal polynomial in the characteristic polynomial $f$ of $\bar g$ with odd multiplicity. The arithmetic significance is direct: intersection lengths are computed by cohomological traces on fine Deligne–Lusztig varieties or their closures.

## 6. Combinatorial and topological perspectives

Recent work places the affine side of the subject into an explicit combinatorial framework via the double Bruhat graph. For a finite Weyl group $W$, this graph has vertices $W$ and edges
\[
u\xrightarrow{\alpha} us_\alpha
\]
for $u\in W$ and $\alpha\in \Phi^+$, together with increasing paths determined by a reflection order and weights in the coroot lattice. Schremmer uses this language to describe structure constants of the Iwahori–Hecke algebra and, as an application, nonemptiness and dimension of affine Deligne–Lusztig varieties for most elements of the affine Weyl group and arbitrary $\sigma$-conjugacy classes [2306.06873].

The geometric applications include the criterion that $X_x(b)$ is nonempty if and only if there exists an increasing path in the double Bruhat graph with the required weight, and, when nonempty,
\[
\dim X_x(b)
=
\frac12\left(\ell(x)+\max(E)-\langle \nu(b),2\rho\rangle-\mathrm{defect}(b)\right),
\]
where $E$ is the multiset of lengths of the relevant paths. The summary explicitly notes that by focusing on the “fine” data of the double Bruhat graph—weights, paths, and recursive relations—the framework enables answers to fine questions about geometry such as explicit enumeration of components or the description of generic versus non-generic Newton strata.

A complementary topological perspective comes from the description of connected components of minuscule affine Deligne–Lusztig varieties. For hyperspecial level and Hodge–Newton indecomposable $(b,\mu)$, Chen, Kisin, and Viehmann prove that
\[
\pi_0(X_\mu(b))\cong C_{b,\mu}\subset \pi_1(G)_\Gamma,
\]
unless $[b]=[\mu]$ is central, in which case $X_\mu(b)=G(F)/G(\mathcal O_F)$ is discrete. They also show that $J_b(F)$ acts transitively on $\pi_0(X_\mu(b))$ [1307.3845]. The same summary makes the analogy with finite Deligne–Lusztig varieties explicit: in the finite case, connected components correspond to orbits under the abelianized fundamental group or component group, whereas in the affine setting the Kottwitz homomorphism plays the role of the component map and $\pi_1(G)_\Gamma$ replaces the finite component group.

Taken together, these developments show that fine Deligne–Lusztig varieties occupy two closely related roles. In the finite setting they are the intrinsic strata of partial flag varieties, with closures and cohomology controlled by Levi induction. In the affine setting they are the finite-type pieces through which more complicated loop-group varieties become accessible, whether through Bruhat–Tits and $J$-stratifications, finite Coxeter type decompositions for $GL_n$, or path combinatorics in the double Bruhat graph.

Source: https://www.emergentmind.com/topics/fine-deligne-lusztig-varieties