Professional Documents
Culture Documents
On A Class of Richardson Varieties: Mahir Bilen Can and Reuven Hodges
On A Class of Richardson Varieties: Mahir Bilen Can and Reuven Hodges
1
Tulane University, New Orleans; mahirbilencan@gmail.com
2
Northeastern University; hodges.r@husky.neu.edu
Abstract
We investigate Schubert varieties that are spherical under a Levi subgroup action
in a partial flag variety. We show that if the Schubert variety is nonsingular, then it
is a spherical L-variety, where L is the largest Levi subgroup that acts the Schubert
variety. Moreover, we classify all partial flag varieties that are spherical with respect to
a Levi subgroup action; we show that the closures of L-orbits in a spherical Schubert
L-variety are Richardson varieties.
Keywords: Spherical varieties, Schubert and Richardson varieties, nonsingular vari-
eties.
MSC: 14M15, 14M27
1 Introduction
For simplicity, throughout our paper, we will work with the varieties and algebraic groups
that are defined over the field of complex numbers, C.
Let X be a smooth algebraic variety. If several subvarieties in X intersect properly in a
finite set of points, say S, then the degree of the product of the corresponding cohomology
classes in H ∗ (X, Z) is equal to the number of points, counted with their multiplicities in
S. This fact from algebraic topology is especially useful for studying intersection theory in
homogenous varieties.
Let G be a connected reductive algebraic group, B be a Borel subgroup, P be a parabolic
subgroup such that B ⊆ P ⊆ G. Let W denote the Weyl group of G and let W P denote the
minimal length right coset representatives for the subgroup WP of W that is determined by
P . By definition, a Schubert variety in G/P is the Zariski closure of a B-orbit in G/P for the
left translation action of G × G/P → G/P , (g, hP/P ) 7→ ghP/P . In particular, Schubert
varieties in G/P are indexed by the subset W P . The Poincaré duals of fundamental classes
of Schubert varieties give an additive basis for the integral cohomology ring H ∗ (G/P, Z). If
1
there is no danger for confusion we will refer to G/P , which is a Schubert variety itself, as
a flag variety.
We consider a normal and irreducible complex algebraic variety X together with an
algebraic action
m : G × X → X.
In this case, X is called a spherical G-variety if the restriction m|B×X : B × X → X has
only finitely many orbits. In this case, the action is called a spherical action.
Let L be a Levi subgroup of P . Note that L ∩ B is a Borel subgroup in L. Let XwP
(w ∈ W P ) be a Schubert variety in G/P . Note also that the Schubert varieties are always
normal and irreducible. In this article, we are concerned with the following two questions
although we will be chiefly interested in the first one:
1. What are the conditions on L and w ∈ W P so that the restriction m|L×Xw : L×XwP →
XwP is defined and the Borel subgroup BL of L has finitely many orbits in XwP ?
2. Assuming that XwP is spherical with respect to the restricted action m|L×XwP : L ×
XwP → XwP , what are the intersection multiplicities for the intersections of L-stable
Schubert varieties with XwP ?
In the special case that P is a maximal parabolic subgroup of the general linear group
GLn , the complete answer to the first question is given by the second author and Lakshmibai
in [12]. We will present their answer in the sequel. In the special case that L is a maximal
Levi subgroup of the form L = GLp × GLq in GLn , the complete answer to both of the
questions is given by Wyser in [17]. We will review Wyser’s answer in the sequel also.
Let us mention that L = GLp ×GLq is one of the few examples to a Levi subgroup which
is a “spherical subgroup” in its ambient reductive group; a closed subgroup H in G is called
a spherical subgroup if HB is open and dense in G. The classification of reductive spherical
subgroups in G is due to Brion [2], Krämer [10], and Mikityuk [13]. The classification of
spherical Levi subgroups is due to Brundan [5] and it goes as follows: let L be a Levi subgroup
Q
and let G′ denote the derived subgroup of G. Let ri=1 Gi be the decomposition of G′ as a
commuting product of simple factors and set Li := L ∩ Gi . Then, L is spherical in G if and
only if, for each i, either Gi = Li or (Gi , L′i ) is one of
The reason for the classification symbolism is invoked in the above list is because the property
that a subgroup H in G is spherical is preserved by the isogenies of G, so any representative
of the corresponding root datum in (1.1) gives a pair of reductive group and a spherical Levi
subgroup. In particular, for any pair (G, L), where L is a spherical Levi subgroup in G, any
Schubert variety in G/P (for any parabolic subgroup P ) is a spherical L-variety.
2
Theorem 1. Let P be a standard parabolic subgroup of G determined by the subset I ⊂ ∆.
If L is a standard Levi factor of minimal dimension such that G/P is a spherical L-variety,
then L is of the form L = LJ ′ with J ′ = J − I, where LJ is a spherical Levi subgroup of G.
1. If L is as in Theorem 1, then for any Levi subgroup L̃ such that L ⊂ L̃, the flag variety
G/P is a spherical L̃-variety.
2. Combined with Brundan’s result, the first item and Theorem 1 tell us if a partial flag
variety G/P is spherical with respect to a given Levi subgroup L̃ or not.
3. Theorem 1 enables us to extend Wyser’s results from [17] and [18] to any flag variety
of the form GLn /P, where P is a parabolic subgroup containing Bn . (We postpone
this development to a future publication.)
The second main result of our article shows that nonsingular Schubert varieties are always
spherical (with respect to a Levi subgroup).
• L is maximal in the sense that there is no other Levi subgroup L̃ such that L ( L̃ and
L̃ acts on X,
The appearance of the hypothesis on G2 -factors is because in the proof of our result we
use a result of Carrell from [8], where he characterized the smooth Schubert varieties in G/B
under the same assumptions.
3
Let v and w be two elements from the Weyl group W of G. The Schubert variety op-
posite to XvB is defined as w0 Xw0 vB , where w0 is the maximal element of W with respect
to Bruhat-Chevalley order. The intersection XwB ∩ w0 Xw0 vB , denoted by Xwv , is called a
Richardson variety. The fundamental classes of Richardson varieties determine the multi-
plicative structure of the cohomology ring H ∗ (G/B, Z), see [3].
We go back to our discussion of spherical Schubert varieties. Let L be a Levi subgroup
acting on a Schubert variety X (in some flag variety). Now, let P and Q be two parabolic
subgroups in G and assume that they share a Borel subgroup B, B ⊂ P ∩ Q. Let WQ and
WP denote, respectively, the parabolic subgroups of W corresponding to Q and P . Define
−
XP,Q := the minimal length coset representatives for the double cosets of (WP , WQ ) in W ;
+
XP,Q := the maximal length coset representatives for the double cosets of (WP , WQ ) in W .
We proceed to give a brief overview of our article and explain how we structured our paper.
In Section 2, we set up our notation and present some results that we use in the sequel. In
particular, we review briefly a theorem that is due to the second author and Lakshimibai
on the Levi-spherical Schubert varieties in grassmannians. In the same section we introduce
the lattice of Levi subgroups of G that acts on a given Schubert variety. Section 3 is devoted
to an example of an L-spherical Richardson variety in the grassmannian. In Section 4, we
initiate our study of the intersections of L-spherical Schubert varieties; Proposition 1 and
Theorem 3 are proven. At the beginning of the subsequent (and final) Section 5, we prove
Theorem 1. In the same section, after reviewing some results of Carrell on the smooth
Schubert varieties, we prove Theorem 2.
4
scheme of finite type. Our implicitly assumed symbolism is as follows:
G : a connected reductive group;
T : a maximal torus in G;
X ∗ (T ) : the character group of T ;
B : a Borel subgroup of G s.t. T ⊂ B;
U : the unipotent radical of B (so, B = T ⋉ U);
Φ : the root system determined by (G, T );
∆ : the set of simple roots in Φ relative to B;
W : the Weyl group of the pair (G, T ) (so, W = NG (T )/T );
S : the Coxeter generators of W determined by (Φ, ∆);
R : the set consisting of wsw −1 with s ∈ S, w ∈ W ;
Cw : the B-orbit through wB/B (w ∈ W ) in G/B;
ℓ : the length function defined by w 7→ dim Cw (w ∈ W );
Xw : the Zariski closure of Cw (w ∈ W ) in G/B;
pP,Q : the canonical projection pP,Q : G/P → G/Q if P ⊂ Q ⊂ G;
WQ : the parabolic subgroup of W corresponding to Q, where B ⊂ Q;
WQ : the minimal length coset representatives of WQ in W ;
XwQ : the image of Xw in G/Q under pB,Q .
The Weyl group W operates on X ∗ (T ) and the action leaves Φ stable. In particular, Φ
spans a (not necessarily proper - depends on G) euclidean subspace V of X ∗ (T ) ⊗Q Z; ∆
is a basis for V . The elements of S are called simple reflections and the elements of R are
called reflections. The elements of S are in one-to-one correspondence with the elements of
∆, and furthermore, S generates W .
If there is no danger for confusion, for the easing of our notation, we will use the notation
w to denote a coset in NG (T )/T as well as to denote its representative nw in the normalizer
F
subgroup NG (T ) in G. The Bruhat-Chevalley decomposition of G is G = w∈W BwB, and
the Bruhat-Chevalley order on W is defined by
w ≤ v ⇐⇒ BwB ⊆ BvB.
Here, the bar indicates Zariski closure in G. Then (W, ≤) is a graded poset with grading
given by ℓ.
In this case, that is w ∈ W Q , the preimage in G/B of XwQ is equal to Xww0,Q , where
5
and the restriction
pB,Q |Xww0,Q : Xww0,Q → XwQ
is a locally trivial fibration with generic fiber Q/B.
Remark 2. We mention in passing that for w ∈ W Q , the products ww0,Q are the maximal
length coset representatives of WQ in W . Also, if u is any element in the right coset wWQ ,
then XuQ = XwQ .
The standard parabolic subgroups with respect to B are determined by the subsets of S;
let I be a subset in S and define PI by
PI := BWI B. (2.1)
LI := hT, Uα : α ∈ ΦI i, (2.2)
UI := hUα : α ∈ Φ+ − Φ+ I i. (2.3)
Then LI is a reductive group and UI is a unipotent group. The Weyl group of LI is equal
to WI . The relationship between LI , UI , and PI is given by
PI = LI ⋉ UI ;
6
parabolic P − such that L ⊂ P − . Moreover, any two parabolic subgroups opposite to P are
conjugate by a unique element of Ru (P ). See [1, Proposition 14.21]. If P is of the form
P = PI for some I ⊂ ∆, then its opposite, denoted by PI− is the parabolic subgroup
where
+ +
UI− := hUα : α ∈ Φ− − Φ−
I i = hUα : −α ∈ Φ − ΦI i.
Uv− := U ∩ vU − v −1 .
fv : Uv− −→ G/B
u 7−→ uvB/B
We apply the following conventions throughout our paper without mentioning them.
Convention 2.5. Note that in general, for simplicity of notation, given a parabolic subgroup
P with Levi factor L, we shall denote the Weyl group of L by WP . In the same way, the set
of minimal length coset representatives of WP in W will be denoted W P .
Convention 2.6. If the flag variety under consideration, G/Q, is clear from the context,
then we will write Xw in place of XwQ .
7
the finiteness condition can be rephrased in terms of representation theory; X is a spherical
G-variety if and only if its coordinate ring, C[X] is multiplicity-free as a G-module. For
non-affine varieties, there is an instrumental result of Sumihiro [15], which states that if X
is a normal G-variety, then X is covered by open G-stable and quasi-projective subvarieties.
Moreover, any such variety admits a G-equivariant embedding π : X ֒→ P(V ), where V
is a finite dimensional G-module. Let X c denote the affine cone over π(X) in V . If it is
necessary, by passing to a central extension of G, we make the following observation: Since
the canonical projection π ′ : Xc → π(X) is a GIT quotient by a one-dimensional torus C∗ ,
′
π is an open G-equivariant map. In particular, X c is a spherical G-variety if and only if
X is a spherical G-variety. Line bundles on a Schubert variety are the restrictions of line
bundles on G/P . Let λ be a dominant weight for (G, T ) such that StabG (λ) = P . The
weights satisfying the latter condition are called P -regular. In this case the corresponding
line bundle Lλ on G/P is very ample, hence it gives a G-equivariant projective embedding
into P(Vλ ), where Vλ is the irreducible G-module with the highest weight λ. Let L be a Levi
subgroup and let X be a Schubert variety which admits the restriction of the left translation
action l : L × G/P → G/P . In the light of our previous discussion, if the homogenous
L
coordinate ring m≥0 H 0 (X, Lλ |X (m)) of X is a multiplicity-free L-module, then X is a
spherical Schubert L-variety.
A closed subgroup H in G is called a spherical subgroup if G/H is a spherical variety.
Let X be an arbitrary spherical G-variety, let x be a point from the open orbit and let
H denote the stabilizer subgroup of x in G. Then X is an equivariant embedding of the
spherical homogenous variety G/H. Any spherical variety that is G-equivariantly birational
to G/H is called an embedding of G/H. An embedding is called simple if it has a unique
closed G-orbit. Every spherical variety has an open covering by the simple embeddings and
the simple embeddings are parametrized by “strictly convex colored fans.”
8
form Sd × Sn−d ; in one-line notation the elements of W P have the form w = w1 . . . wn ,
with a unique descent at the dth position. The Bruhat-Chevalley ordering on W P takes a
particularly simple form as well, with v = v1 . . . vn ≤ w = w1 . . . wn if and only if vj ≤ wj for
1 ≤ j ≤ d.
P
Let D = (d1 , . . . , ds ) be a sequence of positive integers such that si=1 di = n. Following
the standard combinatorial terminology, we will call such a sequence an integer composition
Ps−1
of n. The map d1 + · · · + dr 7→ {d1 , d1 + d2 , . . . , i=1 di} is a bijection between the set of all
integer compositions of n and the set of all subsets of the set {1, . . . , n − 1}.
A subset I = {αi1 , . . . , αir } ⊆ ∆ corresponds to a subset (i1 , . . . , ir ) ⊆ {1, . . . , n − 1}.
As noted above, this is turn corresponds to an integer composition D = (d1 , . . . , dr+1 ) with
dj = ij −ij−1 for 1 < j ≤ r and d1 = i1 , dr+1 = n−ir . Then the standard parabolic subgroup
PI consists of matrices of the form
Ad ∗ ... ∗
1
0
Ad2 . . . ∗
.
.. .. .. .. ∈ GLn ,
. . .
0 0 . . . Adr+1
In [12, 9], Hodges and Lakshmibai determine the Schubert varieties Xw in the Grass-
mannian for which the action of the Levi subgroup L is spherical in the case where L is the
maximal Levi which acts stably on Xw . We briefly describe this result. LetL be a Levi sub-
group; as remarked above, the Levi subgroups GLn are indexed by subsets of {1, . . . , n − 1}.
In particular, there is an I = {i1 , . . . , ir } ⊂ {1, . . . , n − 1} with i1 < · · · < ir such that
The Levi L induces a set partition of {1, . . . , n} denoted BlockL,1 , . . . , BlockL,r+1 where
the maximal element of BlockL,k is ik (or n for BlockL,r+1). For example for I = {3, 6}
and L = GL3 × GL3 × GL3 ⊂ GL9 then BlockL,1 = {1, 2, 3}, BlockL,2 = {4, 5, 6}, and
BlockL,3 = {7, 8, 9}. Let bL denote the number of blocks (it is equal to |I| + 1). For
w = w1 . . . wn a Grassmann permutation with a descent at the dth position define the
numbers h1 , . . . , hbL by
9
Note that checking sphericity of Xw in the cases where w1 = 1 or wd 6= n can always be
reduced to checking the sphericity of the Schubert variety Xw̃ in a smaller Grassmannian
˜
˜ , for some Grassmann permutation w̃ = w̃1 . . . w̃ñ with the descent at the d-th position
Gd,ñ
and w̃1 6= 1 and w̃d˜ = ñ(see [9]).
This result motivates our next subsection.
Definition 2.8. Let Y be a Schubert variety in G/Q, where Q is a parabolic subgroup. Let
L be an element from Levi(Y ). We call Y an L-spherical Schubert variety if the action of L
on Y is spherical. We denote the subset of elements L ∈ Levi(Y ) such that Y is L-spherical
by Levis (Y );
Proof. The proof follows from the fact that if L acts spherically on Y , then so does any Levi
subgroup L̃ in Levi(Y ) containing L.
In general, the converse of Proposition 2 is false. As a simple example, consider the
Schubert variety Y corresponding to the permutation w = 24681357 in the Grassmann
variety of 4-planes in C8 . Note that Y is a singular Schubert variety. It is easy to see that a
maximal Levi subgroup L ⊂ GL8 that acts on Y is conjugate-isomorphic to GL2 × GL2 ×
GL2 × GL2 , and so bL = 4. Hence, by Theorem 4, Y can not be L-spherical.
10
Theorem 4 implies that the converse of Proposition 2 holds for a smooth Schubert variety
in a Grassmannian. In Section 5, we extend this result and show that the converse holds for
a smooth Schubert variety in G/Q, under mild assumptions on G.
3 A motivating example
Before working out our example we relate the sphericity of the opposite Schubert variety
X w to the sphericity of the Schubert variety Xw0 w .
Lemma 3.1. Let P be a standard parabolic subgroup and L a Levi subgroup of G. The
opposite Schubert variety X w in G/P is L-spherical if and only if the Schubert variety Xw0 w
in G/P is w0 Lw0 -spherical.
11
Xw (w ∈ W P ) be a Schubert variety in G/P . For convenience, if there is no danger for
L
confusion, we will denote by C[Xw ] the homogenous coordinate ring m≥0 H 0 (Xw , Lλ |Xw ).
Lemma 4.1. Let L be a Levi subgroup in G and let Xw be an L-spherical Schubert variety
from G/P . If Xv is a Schubert variety that admits the restriction of the left translation
action L × G/P → G/P , then the intersection Xw ∩ Xv is a union of L-spherical Schubert
varieties.
Proof. The idea of the proof is to use an L-equivariant projection from the homogenous
coordinate ring of Xw to the homogenous coordinate ring of Xw ∩ Xv .
One one hand we have the multiplicity-free L-module C[Xw ]. On the other hand, the
map C[Xw ] → C[Xw ∩ Xv ], which is induced by the restriction of the line bundle Lλ , is
surjective. This follows from [4, Corollary 4 (2)]. Since Xw ∩ Xv is stable under L action, the
homogenous coordinate ring of C[Xw ∩Xv ] is a quotient of C[Xw ] as an L-module. Therefore,
the canonical projection
C[Xw ] −→ C[Xw ∩ Xv ]
is L-equivariant. Since the domain, that is C[Xw ] is multiplicity-free, by Schur’s lemma, the
target is a multiplicity-free L-module as well.
It is well known that the intersection of two Schubert varieties is scheme theoretically
reduced. Moreover, C[Xw ∩Xv ] is the coordinate ring of the union of the maximal dimensional
Schubert subvarieties in Xw ∩ Xv . We will show that each of these (maximal dimensional
Schubert) subvarieties is a spherical L-variety. Indeed, Xw ∩ Xv is a union of finitely many
L-orbits (since it is a subset of Xw ). In particular, if O is a maximal dimensional L orbit in
Xw ∩ Xv , then the closure O is an irreducible L stable subvariety of Xw ∩ Xv , hence, it is a
spherical L variety. Clearly, the closure of the dense Borel orbit is equal to O. Conversely, if
O is the closure of a maximal dimensional Borel orbit in Xw ∩ Xv , then it is an irreducible
component of Xw ∩ Xv , therefore, it is equal to one of the L orbit closures. This finishes the
proof of our lemma.
In fact, the proof shows that the intersection of an L-spherical Schubert variety by any
L-stable subvariety in G/P is L-spherical.
Let P and Q be two standard parabolic subgroups in G, with L the Levi factor of P . It
is well known that the P × Q-orbits in G, or equivalently, P -orbits in G/Q (with respect to
left multiplication action) are parametrized by the double cosets WP \W/WQ . See [1, Section
21.16]. It follows that the P -invariant (and hence L-invariant) Schubert varieties in G/Q are
parametrized by the maximal length coset representatives of the double cosets WP \W/WQ .
In light of Lemma 4.1 and this fact, the proof of the following corollary is evident.
12
Next, we prove another corollary of Lemma 4.1.
Corollary 2. Let P (resp. L) be a parabolic (resp. a Levi subgroup) in G. Let v and w be
two elements from W P . If the Schubert variety Xw in G/P is L-spherical and the opposite
Schubert variety X v is L-stable, then the Richardson variety Xwv is L-spherical as well.
Proof. The opposite Schubert variety X v is equal to w0 Xw0 vw0,P . Therefore, the proof follows
from Lemma 4.1.
We will further discuss spherical L-spherical Richardson varieties, so we recall an obser-
vation of Wyser on the Levi subgroup that acts on a Richardson variety [17, Proposition
2]:
Proposition 3 (Wyser). Let P and P − be two parabolic subgroups opposite to each other
and let L denote the standard Levi factor P ∩P − . If u is a maximal length coset representative
of WP \W and v is a minimal length coset representative of WP \W , and u ≥ v, then the
Richardson variety Xvu is invariant under L.
The inclusion poset of P -orbit closures in G/Q is given by the Bruhat-Chevalley order
on WP \W/WQ . This order is well-known (see [14, Proposition 1.8]). Let
π : W → WP \W/WQ
denote the canonical projection onto double cosets. It turns out that the preimage in W of
every double coset in WP \W/WQ is an interval with respect to Bruhat order, hence it has
a unique maximal and a unique minimal element. Moreover, if [w1 ], [w2 ] ∈ WP \W/WQ are
two double cosets represented by the maximal elements w1 , w2 of the corresponding intervals,
then [w1 ] ≤ [w2 ] in WP \W/WQ if and only if w1 ≤ w2 in W .
The set of distinguished (WP , WQ )-double coset representatives is defined as follows. Let
w be a representative of a double coset in WP \W/WQ such that ℓ(w) is as small as possible.
It turns out that such w are precisely the elements of w ∈ P W ∩ W Q , where P W is the set of
−
minimal length coset representatives for WP \W . We denote P W ∩ W Q by XP,Q . In the same
way, let w be a representative of a double coset in WP \W/WQ such that ℓ(w) is as large as
+
possible. We denote the set containing such w by XP,Q . This set can be characterized as
the intersection of the set of maximal length coset representatives for WP \W and the set of
maximal length coset representatives for W/WQ .
− +
Remark 4. The Bruhat orders on XP,Q and XP,Q are isomorphic. Indeed, the spherical
products P \G × G/Q and P \G × G/Q are isomorphic, where P − stands for the parabolic
− −
13
Proof. In the light of the two paragraphs before the theorem, the first part of the proof is
vQ
essentially a consequence of Corollary 2 and Proposition3. For the second part, since XuQ is
vQ
L-spherical, there is an open and dense BL -orbit, denoted by X0 , in XuQ . Let x0 be a point
vQ
from X0 . Then L · x0 = BL · x0 = XuQ .
Remark 5. A Richardson variety in XuQ need not be an L-orbit closure. Indeed, the
GLp × GLq -orbits in GLp+q /Bp+q are parametrized by combinatorial objects called “(p, q)-
clans.” (See the related development in [6], also.) It is noted in [17, Remark 3.9] that
the Richardson varieties that are L-orbit closures in GLp+q /Bp+q correspond to (1, 2, 1, 2)-
avoiding (p, q)-clans.
5 Smoothness
In the previous sections we fixed a Levi subgroup L and we investigated the Schubert
varieties which are L-spherical. Our goal in this section is to analyze the complementary
situation: we fix a Schubert variety Y in a flag variety and determine Levis (Y ). We first
handle the easiest case: Y = G/P , where P is a parabolic subgroup.
Proof. Let L be a Levi subgroup of the form LJ ′ , where J ′ = J − I and LJ is a spherical Levi
subgroup in G. Let BL denote the Borel subgroup L ∩ B and let BL− denote the opposite
Borel subgroup L ∩ B − in L. To show that G/P is an L-spherical flag variety it suffices to
show that the image of the map
BL− × P −→ G (5.1)
(g, h) 7−→ gh (5.2)
is open and dense. On one hand, it is well known that U − × U → G given by (u, v) 7→ uv has
a dense and open image (the big-cell). On the other hand, the sets of root subgroups that
are contained in BL− and P are given by {Uα : −α ∈ J − I} and {Uα : −α ∈ I or α ∈ ∆},
respectively. Therefore the set of root subgroups that are contained in the image is {Uα :
−α ∈ J or α ∈ ∆}, which is precisely the set of root subgroups that are contained in BL−J ·B,
which is open and dense in G. This argument also shows that for any Levi subgroup LK
with K properly contained in the set J − I the dimension of BLK · P is strictly smaller than
dim G, hence the image of the action of LK is not dense in G/P . This finishes the proof.
14
1. The stabilizer in G of the point wQ/Q is the parabolic subgroup
NG (wQ/Q) = wQw −1 .
α ∈ ∆ ∩ w(ΦQ ).
α ∈ ∆ ∩ w(ΦQ ∪ Φ− ).
15
Define
Lemma 5.4. [7, Proposition 1] Every T -invariant curve in G/B is of the form Cα,x for
suitable α and x. Moreover, Cα,x ⊆ Xw if and only if x, rα x ≤ w. Consequently, if x ≤ w,
then Φ(x, w) ⊆ Ω(x, w).
From now on, for simplicity, if x = id, then we will write Cα instead of Cα,id . The above
lemma can be restricted to the Schubert varieties. Let Xw be a Schubert variety in G/B.
Put
E(Xw ) = the set of T -stable curves that are containing eB/B,
and set M
T E(Xw ) := Te (Cα ).
Cα ∈E(Xw )
Thus, T E(Xw ) = Θ(w) if and only if Φ(w) = H(w) ∩ Φ− . Moreover, if T E(Xw ) is B-stable,
then T E(Xw ) = Θ(w).
16
Another useful fact is the following.
Lemma 5.8. [8, Theorem 6] Let us assume that Xw is a rationally smooth variety in G/B
and G does not contain any G2 -factors. Then Xw is smooth if and only if Φ(w) = H(w)∩Φ− .
By Lemmas 5.8 and 5.7, we know that if Xw is smooth, then T E(Xw ) = Θ(w) = Te (Xw )
is a B-stable submodule of Te (G/B) with weights Φ(w) = H(w) ∩ Φ− . Now let us slightly
modify the notation and write
if v ≤ w. Thus, Θ(w) = Θ(e, w). At the same time, by smoothness, for any v ≤ w we have
that
Θ(v, w) ∼
= Θ(w).
Also as a consequence of smoothness we see that the set of roots that occur as a weight of
the T -module TvB/B (Xw ) satisfy
In fact, by [7, Proposition 6], we know that for any reflection r ∈ W such that rw < w,
Ω(rv, w) = rΩ(v, w). Therefore, the reduced tangent cone at the identity Θ(e, w) is B-
equivariantly isomorphic to Θ(v, w) for any v ≤ w under the assumption that Xw is smooth.
In summary we have the following fact:
Lemma 5.10. Suppose that G does not have any G2 -factors and let Xw be a smooth
Schubert variety. Then
T E(Xw , v) = Θ(v, w) = Tv (Xw ),
where T E(Xw , v) is the linear span of tangent spaces at vB/B of the T -stable curves that
are containing vB/B. Moreover, Φ(v, w) = H(v, w) ∩ Φ− , where H(v, w) is the convex hull
of Φ(v, w) in the dual space of Lie(T ) (over R).
The final preparatory fact that we need from Carrell’s work is the following lemma. For
r ∈ R let Sr denote the group generated by the root subgroups Uαr and U−αr . Note that
any conjugate of r in W is a reflection as well.
Lemma 5.11. [7, Proposition 4] Let x ∈ W, r ∈ R. If t = x−1 rx, then the curve Cαt ,x =
Sr xB/B. Moreover, Sr xB ⊆ Xw if and only if x, rx ≤ w.
Although the above results of Carrell are for the Schubert varieties in G/B, similar
statements hold true modulo Q. Indeed, if w ∈ W Q , then the Schubert variety XwQ in G/Q
is isomorphic to XwB in G/B. We will use this fact in the proof of our next result without
further mentioning it.
Theorem 7. Let G be a semisimple linear algebraic group with no G2 -factor. Let X be
a smooth Schubert variety in G/Q. If L is a maximal element from Levi(X), then X is a
maximally L-spherical Schubert variety.
17
Proof. Let B denote the Borel subgroup in Q. Without loss of generality we assume that L
is the standard Levi factor of the stabilizer subgroup P := NG (Xw ) of Xw . Note that P is a
standard parabolic subgroup with respect to B.
Let w ∈ W Q be such that X = XwQ . Recall that the dense B-orbit is given by the image
of fw : Uw− → BwQ/Q, fw (u) = uwQ/Q. The unipotent group Uw− is generated by the root
subgroups Uα , such that
α ∈ Φ+ and w −1(α) ∈ Φ− − Φ− Q.
Sr wQ/Q ⊆ Xw ⇐⇒ w, rw ≤ w.
Since Cαt ,w is T -stable curve in Xw containing the T -fixed point wQ/Q, Lemma 5.4 implies
that tw ≤ w, and thus rw ≤ w. Hence Sr wQ/Q ⊆ Xw .
But Sr contains both of the unipotent subgroups Uα and U−α . This means that U−α
stabilizes BwQ/Q, hence it stabilizes the Schubert variety Xw . In particular, U−α ⊂ P .
Since the root subgroups that corresponding to negative roots lie in the Levi subgroup L of
P , we see that U−α ⊂ L, hence that Uα ⊂ L. This finishes the proof.
References
[1] Armand Borel. Linear algebraic groups, volume 126 of Graduate Texts in Mathematics.
Springer-Verlag, New York, second edition, 1991.
[2] Michel Brion. Classification des espaces homogènes sphériques. Compositio Math.,
63(2):189–208, 1987.
[3] Michel Brion. Lectures on the geometry of flag varieties. In Topics in cohomological
studies of algebraic varieties, Trends Math., pages 33–85. Birkhäuser, Basel, 2005.
[4] Michel Brion and Venkatramani Lakshmibai. A geometric approach to standard mono-
mial theory. Represent. Theory, 7:651–680, 2003.
18
[5] Jonathan Brundan. Dense orbits and double cosets. In Algebraic groups and their
representations (Cambridge, 1997), volume 517 of NATO Adv. Sci. Inst. Ser. C Math.
Phys. Sci., pages 259–274. Kluwer Acad. Publ., Dordrecht, 1998.
[6] Mahir Bilen Can and Özlem Uğurlu. The genesis of involutions (polarizations and lattice
paths). https://arxiv.org/abs/1703.09881, 2016.
[7] James B. Carrell. The span of the tangent cone of a Schubert variety. In Algebraic groups
and Lie groups, volume 9 of Austral. Math. Soc. Lect. Ser., pages 51–59. Cambridge
Univ. Press, Cambridge, 1997.
[8] James B. Carrell. Smooth Schubert varieties in G/B and B-submodules of g/b. Trans-
form. Groups, 16(3):673–680, 2011.
[9] Reuven Hodges. Schubert Singularities and Levi Subgroup Actions on Schubert Varieties.
ProQuest LLC, Ann Arbor, MI, 2017. Thesis (Ph.D.)–Northeastern University.
[11] Venkatramani Lakshmibai and Justin Brown. Flag varieties, volume 53 of Texts and
Readings in Mathematics. Hindustan Book Agency, New Delhi, 2009. An interplay of
geometry, combinatorics, and representation theory.
[12] Venkatramani Lakshmibai and Reuven Hodges. Levi subgroup actions on schubert
varieties, induced decompositions of their coordinate rings, and sphericity consequences.
2018. https://doi.org/10.1007/s10468-017-9744-6.
[14] John R. Stembridge. Tight quotients and double quotients in the Bruhat order. Electron.
J. Combin., 11(2):Research Paper 14, 41, 2004/06.
[15] Hideyasu Sumihiro. Equivariant completion. J. Math. Kyoto Univ., 14:1–28, 1974.
[16] Dmitry A. Timashev. Homogeneous spaces and equivariant embeddings, volume 138 of
Encyclopaedia of Mathematical Sciences. Springer, Heidelberg, 2011. Invariant Theory
and Algebraic Transformation Groups, 8.
[17] Benjamin J. Wyser. Schubert calculus of Richardson varieties stable under spherical
Levi subgroups. J. Algebraic Combin., 38(4):829–850, 2013.
[18] Benjamin J. Wyser. The Bruhat order on clans. J. Algebraic Combin., 44(3):495–517,
2016.
19