---
title: Thin Schubert Cells in Finite Chevalley Groups
url: https://www.emergentmind.com/topics/thin-schubert-cells
type: topic
---

# Thin Schubert Cells in Finite Chevalley Groups

Thin Schubert cells arise in the incidence-theoretic study of Schubert cells in flag varieties over finite fields. For a finite Chevalley group $G = G(\mathbb{F}_q)$, a Schubert cell $C_w = BwB/B$ in the full flag variety $G/B$ is an affine space of dimension $\ell(w)$, but the notion of thinness is not attached to $C_w$ merely as a variety. Instead, it is attached to an incidence structure $(X_w)_{ij}$ defined from the image of a single Schubert cell under two projections to maximal partial flag varieties. In that setting, thinness means that every line meets the point set in at most two points, in exact analogy with the first ovoid property $(O1)$ from finite geometry. The central result is a uniform fiber-size formula: if $w = uzv$ is the canonical parabolic factorization, then every line in $(X_w)_{ij}$ contains exactly $q^{\ell(z)}$ points; consequently, thinness occurs only in the cases $z = 1$, or $q = 2$ and $z = s_i$ [1805.03864].

## 1. Algebraic and combinatorial setting

Let $G = G(\mathbb{F}_q)$ be a finite Chevalley group over the finite field $\mathbb{F}_q$. Fix a Borel subgroup $B$, a maximal torus $T \subseteq B$, and the associated Weyl group $W = N/T$, where $N$ is generated by certain elements $n_\alpha$ arising from the Chevalley generators. Write $S = \{s_1, \dots, s_n\}$ for the simple reflections, corresponding to simple roots $\alpha_1, \dots, \alpha_n$ [1805.03864].

The Bruhat order $\leq$ on $W$ is defined by inclusion of inversion sets, and the length function $\ell: W \to \mathbb{N}$ gives the number of positive roots sent to negative ones by an element, equivalently the number of simple reflections in any reduced expression. These two structures organize both the geometry of Schubert cells and the parabolic factorizations that control the incidence structures associated with them.

The relevant root-theoretic notation is as follows. Let $R$ be the root system and $X_a = \{x_a(c)\mid c\in \mathbb{F}_q\}$ the root subgroup for each root $a \in R$. Choose a system of positive roots $R^+$ and simple roots $\alpha_1,\dots,\alpha_n$. Set
$$
U = \langle X_a \mid a \in R^+ \rangle,\qquad
T = \langle h_\lambda(d)\mid \lambda \in \mathfrak{h}^*,\ d\in \mathbb{F}_q^\times \rangle,\qquad
B = UT.
$$
Write $x_i(c) = x_{\alpha_i}(c)$, $n_i = n_{\alpha_i}$, and $s_i = n_iT$. This notation makes it possible to pass directly between the language of buildings and incidence geometry and the language of Chevalley generators and Weyl group combinatorics.

## 2. Schubert cells and explicit parametrization

The Bruhat decomposition is
$$
G = \bigcup_{w\in W} BwB.
$$
In the full flag variety $G/B$, the Schubert cell corresponding to $w \in W$ is
$$
C_w = BwB/B,
$$
and its closure, the Schubert variety, is
$$
X_w = \bigcup_{u\leq w} C_u.
$$
A standard fact is that $C_w$ is isomorphic to affine space of dimension $\ell(w)$:
$$
C_w \cong \mathbb{A}^{\ell(w)}.
$$
Therefore, for the full flag variety,
$$
|C_w(\mathbb{F}_q)| = q^{\ell(w)}.
$$
For partial flag varieties $G/P$, where $P$ is parabolic, the Schubert cells are indexed by minimal coset representatives in $W/W_P$, and again each cell is an affine space with dimension given by the appropriate length; the $\mathbb{F}_q$-point count is $q$ to the power of that dimension [1805.03864].

For a fixed reduced expression
$$
w = s_{i_1}\cdots s_{i_\ell},
$$
Steinberg’s parametrization gives an explicit affine parametrization of the Schubert cell:
$$
BwB = \{x_{i_1}(c_1)n_{i_1}\cdots x_{i_\ell}(c_\ell)n_{i_\ell}B \mid c_1,\dots,c_\ell \in \mathbb{F}_q\}.
$$
This realizes $C_w$ with coordinates $(c_1,\dots,c_\ell)\in \mathbb{F}_q^\ell$. The significance of this parametrization is structural rather than merely enumerative: it provides explicit coordinates in which the fibers of the projection maps to partial flag varieties can be analyzed coordinate-by-coordinate. In particular, the later fiber-size formula is proved by tracking which $x_i(c)$-coordinates survive passage to a given parabolic quotient.

## 3. The incidence structure attached to a Schubert cell

Fix two standard maximal parabolic subgroups $P_i$ and $P_j$, corresponding to the omission of the simple root $\alpha_i$ or $\alpha_j$. Given $w \in W$, consider the Schubert cell $X_w = BwB/B$ inside $G/B$ and the natural projections
$$
\pi_i: G/B \to G/P_i,\qquad \pi_j: G/B \to G/P_j.
$$
The paper defines an incidence structure $(X_w)_{ij}$ whose points are the cosets $gP_i$ lying in the image of $\pi_i|_{BwB}$, whose lines are the cosets $hP_j$ lying in the image of $\pi_j|_{BwB}$, and in which a point $gP_i$ is incident with a line $hP_j$ if there exists $kB \in BwB$ with $\pi_i(kB)=gP_i$ and $\pi_j(kB)=hP_j$ [1805.03864].

Equivalently, $gP_i$ and $hP_j$ are incident if $gh^{-1}\in B$, with $g$ and $h$ chosen as canonical representatives from the Steinberg parametrization of $BwB$. This reformulation makes the incidence relation compatible with explicit root-subgroup coordinates.

The relevant parabolic and Weyl-theoretic notation is:
$$
R_i = \{a\in R^+ \mid X_{-a}\subseteq P_i\},\qquad
R_j = \{a\in R^+ \mid X_{-a}\subseteq P_j\},\qquad
R_{i,j}=R_i\cap R_j.
$$
Define
$$
W_i = \langle s_a \mid a\in R_i\rangle,\qquad
W_j = \langle s_a \mid a\in R_j\rangle,\qquad
W_{i,j}=W_i\cap W_j.
$$
For $z\in W$, the inversion set is
$$
R(z):=\{a\in R^+ \mid X_{za}\nsubseteq B\},
$$
and $\ell(z)=|R(z)|$. The sets of minimal coset representatives are
$$
W^j = \{u\in W \mid R(u)\cap R_j=\varnothing\},
$$
and
$$
(W_j)^{i,j} = \{z\in W_j \mid R(z)\subseteq R_i\ \text{and}\ R(z)\cap R_{i,j}=\varnothing\}.
$$
Every $w\in W$ has a unique factorization
$$
w = uzv,
$$
with
$$
u\in W^j,\qquad z\in (W_j)^{i,j},\qquad v\in W_{i,j}.
$$

In building-theoretic language, the pair of types $\{i,j\}$ determines rank-2 residues controlled by $W_{i,j}$; the fiber structure of $\pi_i$ over a fixed $j$-type coset reflects the local incidence geometry with two types. Representation-theoretically, the Chevalley root subgroups $x_\alpha(t)$ give explicit coordinates, and Proposition 4.2 shows how the $x_i(c)$-coordinates in the $z$-block parametrize the points lying on a given line.

## 4. Fiber size and the characterization of thinness

The fundamental structural result is the theorem of Bamberg–Ram–Xu: let $w\in W$ and write
$$
w = uzv
$$
with
$$
u\in W^j,\qquad z\in (W_j)^{i,j},\qquad v\in W_{i,j}.
$$
Then, for any line $gP_j$ in $(X_w)_{ij}$, the number of points incident to $gP_j$ is
$$
q^{\ell(z)}.
$$
In particular, the number of points incident to any line depends only on the middle factor $z$ in the parabolic factorization of $w$ [1805.03864].

The proof idea proceeds through Steinberg parametrization. One writes elements in $BwB$ as products of $x_i(c)n_i$ in a fixed reduced word for $w$. Passing to $G/P_j$ via $\pi_j$ kills coordinates associated to simple reflections in $W_j$. The fiber $\pi_i(\pi_j^{-1}(gP_j))$ over any line decomposes as a disjoint union of affine pieces indexed by $y\in W_j$, and, more precisely, by the middle factor $z$ in $w=uzv$. Proposition 4.2 shows that the coordinates associated to the simple reflections in $z$ give a bijection
$$
\pi_i(\pi_j^{-1}(gP_j)) \cong \mathbb{F}_q^{\ell(z)}.
$$
Thus $q^{\ell(z)}$ is the uniform line size in $(X_w)_{ij}$.

Thinness is defined by abstraction from ovoid theory. In finite geometry, following Tits, an ovoid $O$ in a projective space or related ambient geometry is a set of points satisfying $(O1)$ thinness, namely that every line of the ambient geometry contains $0$, $1$, or $2$ points of $O$, and $(O2)$ maximality. In the present setting, only the first property is retained: thinness means that, in the incidence structure under consideration, every line meets the point set in at most two points. Applied to $(X_w)_{ij}$, this means that every line $hP_j$ has at most two incident points $gP_i$.

Since each line has exactly $q^{\ell(z)}$ points, thinness requires
$$
q^{\ell(z)}\leq 2.
$$
Hence either $\ell(z)=0$, that is $z=1$, for any $q$, or $\ell(z)=1$ and $q=2$. The paper observes that the only element of $(W_j)^{i,j}$ of length $1$ is $s_i$. Therefore the Schubert incidence structures $(X_w)_{ij}$ such that every line meets the point set in at most two points are exactly those with
- $w\in W^j\cdot W_{i,j}$ if $q>2$, equivalently $z=1$;
- $w\in (W^j\cdot W_{i,j})\cup (W^j\cdot s_i\cdot W_{i,j})$ if $q=2$, equivalently $z\in \{1,s_i\}$.

Equivalently,
$$
(X_w)_{ij}\ \text{is thin}\ \Longleftrightarrow\ q^{\ell(z)}\leq 2\ \Longleftrightarrow\
\begin{cases}
z=1 & \text{for any } q,\\
z=s_i & \text{and } q=2.
\end{cases}
$$

## 5. Examples and explicit computations

In type $A_n$, one has $G = SL_{n+1}(\mathbb{F}_q)$ and $W \cong S_{n+1}$. The simple reflections are adjacent transpositions, and $P_i$, $P_j$ correspond to stabilizers of $i$- and $j$-dimensional subspaces. For $w\in W$, with parabolic factorization $w = uzv$ where $u\in W^j$, $z\in (W_j)^{i,j}$, and $v\in W_{i,j}$, every line in $(X_w)_{ij}$ has size $q^{\ell(z)}$ [1805.03864].

A small-rank example is $A_2$, where $G = SL_3(\mathbb{F}_q)$ and $W\cong S_3$ with generators $s_1,s_2$. Take $i=1$ and $j=2$. Then
$$
(W_2)^{1,2}=\{1,s_1\}.
$$
If $w = uzv$ with $z=1$, every line has exactly one point. If $q=2$ and $z=s_1$, every line has exactly two points. For $q>2$ and $z=s_1$, lines have $q>2$ points, so the incidence structure is not thin.

The paper’s $A_3$ computation is more explicit. Take $G = GL_4(\mathbb{F}_q)$ with $i=1$ and $j=2$. Then
$$
W \cong S_4,\qquad
W_1 = S_1\times S_3,\qquad
W_2 = S_2\times S_2,\qquad
W_{1,2}=S_1\times S_1\times S_2,
$$
and
$$
(W_2)^{1,2}=\{1,s_1\}.
$$
Choose $w = uzv$ with
$$
u = s_1s_3s_2,\qquad z=s_1,\qquad v=s_3.
$$
In one-line notation,
$$
w=(s_1s_3s_2)(s_1)(s_3).
$$
For
$$
g = x_1(c_1)n_{11}\,x_3(c_2)n_{31}\,x_2(c_3)n_{21},
$$
the fiber computation shows
$$
\pi_i(\pi_j^{-1}(gP_2))
=
\{x_1(c_1)n_{11}\,x_3(c_2)n_{31}\,x_2(c_3)n_{21}\,x_1(d_1)n_{11}P_i \mid d_1\in \mathbb{F}_q\},
$$
confirming that each line has $q$ points, since here $\ell(z)=1$. Thus thinness holds only when $q=2$; for $q>2$, $(X_w)_{12}$ is not thin.

The non-thin cases follow immediately from the theorem. Whenever $\ell(z)\geq 2$, lines in $(X_w)_{ij}$ have at least $q^2$ points, so thinness fails for all $q\geq 2$. When $\ell(z)=1$ and $q>2$, lines have $q>2$ points, so thinness also fails. No exceptional small-$q$ phenomena beyond $q=2$, $\ell(z)=1$ occur in this framework.

## 6. Relation to ovoids, closure phenomena, and broader significance

The ovoid motivation is precise. In finite geometry, the thinness condition $(O1)$ for ovoids says that each line meets the point set in at most two points. For the Schubert-cell incidence structure $(X_w)_{ij}$, this criterion is exactly mirrored by
$$
q^{\ell(z)}\leq 2.
$$
The result therefore shows that Schubert cells provide, in this incidence-theoretic construction, only trivial thin examples, where every line meets the point set in a single point and $z=1$, or the degenerate binary case, where every line meets the point set in two points and $q=2$, $z=s_i$ [1805.03864].

This has a clear negative consequence for the original motivating question: Schubert cells do not yield new rich families of ovoids via this thinness property alone. At the same time, the work is structurally significant because it bridges finite geometry’s lattice/incidence language with representation theory’s Chevalley/Bruhat toolkit. A plausible implication is that the main contribution is methodological as much as classificatory: the paper imports explicit root-subgroup coordinates and parabolic factorization into a finite-geometric setting in which local line sizes can be computed uniformly.

The incidence structure $(X_w)_{ij}$ is defined using the Schubert cell $BwB/B$, not its closure. Passing to the Schubert variety $X_w = \bigcup_{u\leq w} C_u$ generally increases the images of $\pi_i$ and $\pi_j$ and mixes $z$-factors from multiple $u$, making the uniform line-size formula $q^{\ell(z)}$ no longer applicable to the closure. Thus, thinness is a property of the cell-based incidence structure as defined, and does not automatically persist under passage to closures.

Several extensions are identified as worth exploring: incidence structures arising from non-maximal parabolics or more than two types, and whether analogous uniform fiber-size formulas persist; twisted or non-split groups, where the combinatorics of $W$ and parabolics differ; geometric conditions beyond thinness, such as maximality $(O2)$, that might better align Schubert geometry with classical ovoid constructions in polar spaces or generalized quadrangles; and whether other representation-theoretic subvarieties, including Richardson varieties or opposite (double) Schubert cells, yield more interesting finite incidence geometries. The primary outcome remains the structural one: thinness occurs in Schubert cell incidence structures only in trivial or binary cases, and it is governed completely by the parabolic factor $z$ in the factorization $w = uzv$.

Source: https://www.emergentmind.com/topics/thin-schubert-cells