Download as pdf or txt
Download as pdf or txt
You are on page 1of 48

REGULAR CHARACTERS OF CLASSICAL GROUPS OVER COMPLETE

DISCRETE VALUATION RINGS

SHAI SHECHTER

Abstract. Let o be a complete discrete valuation ring with residue field k and let G be
a classical group, i.e. the group of isometries of a symplectic or orthogonal bilinear form
on a finite-dimensional vector space. For any ` N let G` denote the principal congruence
subgroup of G(o). An irreducible character of the group G(o) is said to be regular if it is
trivial on a subgroup G`+1 for some `, and if its restriction to G` /G`+1 ' Lie(G)(k) consists
of characters of minimal stabilizer dimension. In the present paper we consider the regular
characters of the classical groups over o, and construct and enumerate all regular characters
of G(o). As a result, we compute the regular part of the representation zeta function of
such groups.

Contents
1. Introduction 2
1.1. The basic definitions 2
1.2. Regular Elements and Regular Characters 3
1.3. Regular representation zeta functions 4
1.4. Classification of regular orbits in g(k) 4
1.5. Organization 6
1.6. Acknowledgements 7
2. Notation and basic definitions 7
2.1. The symplectic and orthogonal groups 7
2.2. Artinian local rings 8
2.3. The Greenberg functor 8
2.4. Congruence subgroups and quotients 9
2.5. Properties of the Cayley map 9
2.6. Group characters 10
3. Regular elements and regular characters 10
3.1. Regular elements 10
3.2. Regular characters 17
4. Construction of regular characters 19
4.1. Preparation for the proof of Theorem I 19
4.2. Proof of Theorem I 21
5. The classical groups 24
5.1. Statement of results 25
5.2. Preliminary matters 26
Date: May 30, 2017.
This work is part of the authors Ph.D. requirements under the supervision of Uri Onn at Ben-Gurion
university of the Negev.
1
5.3. Symplectic and odd-orthogonal group 30
5.4. Even orthogonal groups 36
Appendix A. The number of even polynomials of a given degree over Fq 46
References 47

1. Introduction
Let K be a non-archimedean local field, and let o be its valuation ring, with maximal
ideal p and residue field k of odd characteristic. Let q and p denote the cardinality and
characteristic of k, respectively. Fix to be a uniformizer of o. Let G GLN be a
symplectic or an orthogonal group over o, i.e. the group of automorphisms preserving a
fixed non-degenerate symmetric or anti-symmetric o-defined bilinear form. In this essay we
study the set of irreducible regular characters of the group G(o), the definition of which we
now present.

1.1. The basic definitions. Let Irr(G(o)) denote the set of irreducible complex characters
of G(o) which are continuous with respect to the profinite topology. The level of a character
Irr(G(o)) is defined to be the minimal number ` N0 such that the restriction of to
the congruence kernel G`+1 = Ker G(o) G(o/p`+1 ) is trivial. For example, the set of
characters of level 0 is naturally identified with the set of irreducible complex characters of
G(k).

1.1.1. The residual orbit of a character. Let g = Lie(G) glN denote the Lie algebra

scheme of G. The map x 7 1 + ` x is an isomorphism of abelian groups g(k) G` /G`+1 ,
whenever ` 1. Note that the action of G(o) by conjugation on the quotient G` /G`+1
factors through its quotient G(k), and that the isomorphism above is G(k)-equivariant.
Additionally, the semisimplicity of g(k) implies that its underlying additive group can be
naturally identified with its Pontryagin dual in a G(k)-equivariant manner. Consequently,
there exists an isomorphism of G(k)-spaces

Irr G` /G`+1 .

g(k) (1.1)
Let Irr(G) have level ` > 0. We consider the restriction G` of to G` . By Clifford
theory and the definition of level, the restricted character G` is equal to a multiple of the
sum over a single G(k)-orbit of characters of G` /G`+1 . Using (1.1), this orbit corresponds to
a single G(k)-orbit in g(k), which we call the residual orbit of . We denote the residual
orbit of by 1 () Ad (G(k)) \g(k).

1.1.2. Regular characters. Let K = kalg denote the algebraic closure of k. An element of
g(K) is said to be regular if its centralizer in G(K) has minimal dimension among such
centralizers (cf. [21, 3.5]). By extension, an element of g(k) is said to be regular if its
image under the natural inclusion of g(k) into g(K) is regular.
Definition 1.1.1 (Regular Characters). A character Irr(G) of positive level is said to
be regular if its residual orbit 1 () consists of regular elements of g(k).
2
For a general overview of regular elements in reductive algebraic groups over algebraically
closed fields, we refer to [17, Chapter III]. The definition of regular characters goes back to
Shintani [16] and Hill [9]. An overview of the history of regular characters of GLN (o) can be
found in [19]. Also- see [12, 20] and [23] for the analysis of regular characters of isotropic
groups of type An , as well as [15], for a partial treatment of anisotropic groups of type An .

1.2. Regular Elements and Regular Characters. Following the path taken by Hill in
[9], we begin our investigation of regular characters from the study of regular elements in
the finite Lie rings g(om ), where om = o/pm (see Definition 3.1.1).
An important result which enables the construction and enumeration of regular characters
of GLn (o) is [9, Theorem 3.6]. Let us recall the main relevant statements of this theorem.
Given x Mn (o) and m N, let xm Mn (om ) denote the image of x under reduction
modulo pm . The following are equivalent, for x Mn (o).
(1) The element xm Mn (om ) is regular;
(2) The element x1 Mn (k) is regular;
(3) The module onm is cyclic over the ring om [xm ] Mn (om ).
In particular, the above statements imply that the centralizer of a regular element xm Mn (om )
is equal to the polynomial ring om [xm ] and that the reduction map om or (for r m)
induce a surjective map from the centralizer of xm onto the centralizer of xr in Mn (or ).
In Section 3.1 of this paper, we prove variants of this result (Theorem 3.1.2 and Theo-
rem 3.1.3), which imply the following properties of regular elements in classical groups.
(1) An element xm g(o/pm ) is regular if and only if its image x1 g(k) under the
reduction map is regular.
(2) The centralizer CG(o/pm ) (xm ) of a regular element xm g(o/pm ) is an abelian sub-
group of G(o/pm ).
(3) Let 1 r m be integers, and let xm g(o/pm ) and xr g(o/pr ) its image
under the reduction map. The reduction modulo pr map restricts to a surjective map
CG(o/pm ) (xm )  CG(o/pr ) (xr )
As we shall see in Section 3.2.2, Assertion (1) above is pivotal in the classification and
enumeration all characters of certain subquotiens of the group G lying above a given regular
orbit, while Assertions (2) and (3) provide the information needed in order to describe the
inertia subgroup of such a character and enumerate the number of characters of G lying
above the given character.
From this, we deduce the first main statement of this article.
Theorem I. Let o be a local ring of odd residual characteristic, and let G be a classical
group over o of rank n. Let g(k) be an orbit consisting of regular elements and let
` N.
(1) The number of regular characters Irr[`] (G) whose residual orbit is equal to is
|G(k)|
||
q (`1)n .
(2) Any such character has degree || q (`1) , where = dim 2Gn .
Remark. It is worth mentioning that the assertions leading up to the proof of Theorem I
are reliant on specific known properties of the classical groups; namely, the existence of the
Cayley parametrization map, and the structure of the centralizers of regular elements over
3
k. Imposing additional restrictions on the characteristic of o, it is likely to be able to extend
Theorem I to a larger class of groups.
1.3. Regular representation zeta functions. Taking the perspective of representation
growth, one is often interested in understanding the asymptotic behaviour of the sequence
{rm (G)}m=1 , where rm (G) N {0, } denotes the number of irreducible characters of G
of degree m. In the case where the sequence rm (G) is bounded above by a polynomial in m,
the representation zeta function of G is defined to be the Dirichlet generating function

X
G (s) = rm (G)ms , (s C). (1.2)
m=1

Considering a possible strategy for analysing the representation zeta function of G, one
may initially restrict to a description of the regular representation zeta function, i.e. the
Dirichlet function counting only regular characters of G. In this respect, Theorem I actually
shows that the rate of growth of regular characters of G is precisely n , where n and are
as above. Furthermore, we obtain the following corollary.
Corollary 1.3.1. Let X Ad(G(k))\g(k) denote the set of orbits consisting of regular
elements, and let
X |G(k)|
Dg (s) = ||s , (s, C).
X
||
The regular zeta function of G is of the form
Dg (s)
Greg. (s) = (1.3)
1 q ns
where n and are as in Theorem I.
1.4. Classification of regular orbits in g(k). Once Theorem I is proved, the final result
of this manuscript is to compute the regular representation zeta function of the classical
groups over o. In view of Corollary 1.3.1, to do so, one must classify an enumerate the
regular odjoint orbits in g(k), under the action of G(k). This classification is undertaken in
Section 5, and its consequences are summarized in Theorem 5.1.2 and Theorem 5.1.3.
Our results in this section yields a uniform formula for each of the four classical groups,
dependent only on the residual cardinality q of k. Recall that the number of monic irreducible
polynomials ofd degree d over k is given by evaluation at t = q of the function wd (t) =
1 d
P
d m|d m
t , where () is the Mobius function (see, e.g. [6, Ch. 14]. Put
( P
1 d d

d m|d m if d > 1
Pd (t) = (1.4)
q1 if d = 1.
A polynomial f k[t] is said to be even if it satisfies the condition f (t) = f (t). The
number of monic irreducible even polynomials of degree d over k is given by evaluation at
The function Ed (t) is probably well-known. In the lack of a reference, a proof of (1.5) is given in
Appendix A. I wish to thank Jyrki Lahtonen [13] of the Mathematics Stack Exchange network taking part
in computing this formula.
4
t = q of the function
( P
1 m

d r|m , 2- m r
(tr 1) , if d = 2m is even
Ed (t) = r (1.5)
0 , otherwise.

Given n N let Xn denote the set of triplets (r, A, B) Z0 Mn (Z0 ) Mn (Z0 )


satisfying the conditions
Xn
r+ (Ad,e + Bd,e ) = n. (1.6)
d,e=1

For any t Xn we associate the following quantity.


 Pd,e Bd,e Yn  P   P  
1 e Ad,e E2d (q) e Bd,e Pd (q) Ed (q)
Mt = P P .
2 d=1
A d,1 , Ad,2 , . . . , Ad,n e A d,e Bd,1 Bd,2 , . . . , Bd,n e Bd,e
(1.7)
We remark that the value Mt describes the size of a distinguished subset of polynomials over
k, namely, that of polynomials of type t; see Definition 5.1.1.
In the case where G is a symplectic group or an orthogonal group, defined over a free
o-module of odd rank, we obtain the following.
Theorem II. Let o be a complete discrete valuation ring of odd residual characteristic. Let
n N. In the notation of Corollary 1.3.1, the Dirichlet polynomials in the numerator of the
regular zeta functions of Sp2n (o), SO2n+1 (o) are given in equations (1.8) and (1.9).
Let Xn0 denote the set of triplets t = (r, A, B) Xn with r = 0.

Dsp2n (o) =
Qn !s
2i
2n2 i=1 (1 q )
X Y
q n
M(0,A,B) (1 + q d )Ad,e (1 q d )Bd,e q Q d
d,e (1 + q )
A d,e (1 q d )Bd,e
(0,A,B)Xn0 d,e
Qn !s
2i
2n2 i=1 (1 q )
X Y
+4q n M(r,A,B) (1 + q d )Ad,e (1 q d )Bd,e q Q ,
2 d,e (1 + q ) (1 q d )Bd,e
d A d,e
(r,A,B)Xn rXn0 d,e
(1.8)

Dso2n+1 (o) =
Qn !s
2i
2n2 i=1 (1 q )
X Y
qn M(r,A,B) (1 + q d )Ad,e (1 q d )Bd,e q Q d
.
d,e d,e (1 + q )
A d,e (1 q d )Bd,e
(r,A,B)Xn
(1.9)
In the case where G consists of isometries of a symmetric non-degenerate bilinear form over
a free o-module of even rank, our results regarding the regular zeta function are presented in
Theorem III. Recall that a symmetric bilinear form over a finite field of odd characteristic is
determined by the dimension of a maximal isotropic subspace with respect to the form (i.e.
the Witt index of the form). In suitable bases, any non-degenerate symmetric bilinear form
5
on k can be represented by one of the matrices

0 1 0 1
1 0 1 0
.. ..

. .

+

J = 0 1 or J =
0 1 .


1 0


1 0

0 1 1 0
1 0 0
Following standard notation, we write SO+2n for the group of isometries of the bilinear form
represented by J+ , and SO
2n for the group of isometries of the form represented by J .
Theorem III. Let o be a complete discrete valuation ring of odd residual characteristic,
and residual cardinality greater than 3. Let n N. In the notation of Corollary 1.3.1, the

Dirichlet polynomials in the numerator of the regular zeta functions of SO+2n (o), SO2n (o) are
given in equations (1.10) and (1.11) below.
r = 0, and let Xn0,+ denote the
Let Xn0 denote the set of triplets t = (r, A, B) Xn with P
the subset of Xn consisting of element (0, A, B) such that d,e eAd,e is even and Xn0, =
0

Xn0 r Xn0,+ .

Dso+2n (o) =
Qn1 !s
X Y
2n2 n (1 q n ) i=1 (1 q 2i )
qn M(0,A,B) (1 + q d )Ad,e (1 q d )Bd,e q Q d Ad,e (1 q d )Bd,e
d,e (1 + q )
(0,A,B)Xn0,+ d,e
Qn1 !s
X Y
2n2 n (1 q n ) i=1 (1 q 2i )
+2q n
M(r,A,B) (1+q d )Ad,e (1q d )Bd,e q Q d Ad,e (1 q d )Bd,e
,
(r,A,B)Xn rXn0 d,e d,e (1 + q )

(1.10)
and

Dso2n (o) =
Qn1 !s
X Y
2n2 n (1 + q n ) i=1 (1 q 2i )
qn M(0,A,B) (1 + q d )Ad,e (1 q d )Bd,e q Q d Ad,e (1 q d )Bd,e
d,e (1 + q )
(0,A,B)Xn0, d,e
Qn1 !s
X Y
2n2 n (1 + q n ) i=1 (1 q 2i )
+2q n M(r,A,B) (1+q d )Ad,e (1q d )Bd,e q Q d Ad,e (1 q d )Bd,e
.
(r,A,B)Xn rXn0 d,e d,e (1 + q )

(1.11)
1.5. Organization. In Section 2 we present the key players which take part in the analysis.
Section 3 contains basic structural results regarding the regular adjoint orbits of g(o), and
the regular characters of G(o), which are constructed in Section 4. Lastly, in Section 5
we classify the regular adjoint orbits of g(k) and compute the regular representation zeta
function of G.
6
1.6. Acknowledgements. This paper is part of the authors doctoral Thesis. I wish to
thank Uri Onn for guiding and advising this research. I also wish to thank Alexander
Stasinski for carefully reading through a preliminary version of this article and offering some
essential remarks.

2. Notation and basic definitions


2.1. The symplectic and orthogonal groups. The main case under investigation in this
article is that of G a symplectic or orthogonal group defined over o. We fix N N and a
matrix J GLN (o) such that Jt = J with  {1}, and consider the group scheme G
defined by
G(R) = x MN (R) | xt Jx = J and det(x) = 1 ,

(2.1)
where R is an o-algebra and the notation xt stands for the transpose matrix of x. A standard
computation (see, e.g.[25, 12.3]) shows that the Lie-algebra scheme g = Lie(G) is given
by
g(R) = x MN (R) | xt J + Jx = 0 .

(2.2)
Let n denote the absolute rank of G and d the dimension of G. Note that n = b N2 c.

2.1.1. Adjoint operators. Let RN denote the N th cartesian power of R, identified with the
space MN 1 (R) of column vectors, and define a non-degenerate bilinear form on RN by
BR (u, v) = ut Jv. One defines an R-anti-involution on MN (R) = EndR (RN ) by
A? = J1 At J (A MN (R)), (2.3)
or equivalently, by letting A? be the unique matrix satisfying BR (A? u, v) = BR (u, Av), for
all u, v RN . In this notation, we have that A G(R) if and only if det(A) = 1 and
A? A = 1, and that A g(R) if and only if A? + A = 0.

2.1.2. The Cayley map. Put


D(R) = {x MN (R) | det (x 1) 6= 0} . (2.4)
The Cayley map cay is obtained from the rational map defined on D(R) by
cayR (x) = (1 + x)(1 x)1 , (2.5)
for any o-algebra R.
It is routine to verify that, in the case R = k, the restriction of the Cayley map to
D(k) g(k) is a k-regular map onto a Zarisky open subset of G(k). Additionally, being a
rational map, the map cayk is equivariant with respect to the conjugation action of GLN (k),
and hence also G(k)-equivariant.
Similarly, for m 1, the Cayley map induces a birational equivalence g(om ) 99K G(om )
which is Ad (G(om ))-equivariant. Upon application of the Greenberg functor (on which we
expand below), this map induces a k-regular birational equivalence between the the smooth
algebraic schemes obtained from g(om ) and G(om ), which is equivariant with respect to the
associated group (see [18, 3] for further information).
7
2.2. Artinian local rings. Let O denote the maximal unramified extension of o, and let
K be the residue field of O and P = O its maximal ideal. We write K alg and K unr for the
algebraic and unramified closures of K, respectively. Given m N we put om := o/pm and
Om := O/Pm and write m : O Om and m,r : Om Or for the reduction maps, for any
1 r m; see Figure 1.

K alg

K unr
r
O Or K

K
r
o or k

Figure 1.

The ring Om (resp. om ) is isomorphic to either K[t]/(tm ) (resp. k[t]/(tm )) in the case
where K and k are of equal characteristic, or the ring of m-dimensional Witt vectors over
K (resp. k), otherwise. In both cases, the canonical splitting map s : K O is a K-rational
map, satisfying that s |K is a homomorphic embedding into O , s(0) = 0 and such that
1 s is the identity map on K (See [14]).
Let : K unr. K unr. be the local Frobenius map whose fixed field is K. Then restricts
to a ring automorphism of O, with fixed subring O = o, and induces a map Om Om
for any m 1 whose fixed subring is om . In particular, in the special case m = 1, the map
: K K is given by the q-power map x 7 xq .
2.3. The Greenberg functor. A large part of our analysis depends on an understanding
of the geometry of the groups and Lie-algebras appearing in Section 2.4. To this end, we use
the Greenberg functor, which allows us to identify reductive group and Lie-algebra schemes
over the quotient rings Or with group and Lie-algebra schemes over the quotient ring K. The
main references we use are Greenbergs original papers [7, 8] and [18].
In its most general form, the Greenberg functor allows us to realize any artinian local
ring R as an algebraic ring over its residue field, and thereby associate to any R-scheme Y
a scheme FR (Y ) over the residue field of R. The association Y 7 FR (Y ) is functorial and
preserves many important properties of Y .
Our application of the Greenberg transform is focused on the artinian rings Om , described
in the previous subsection, and is quite rudimentary. The main properties which we require
are the following. For m N fixed, we have
(1) The rings Om are the K-points of an m-dimensional algebraic ring scheme over K.
(2) The Greenberg functor maps the groups G(Om ) to a dm-dimensional linear algebraic
group over K.
(3) The Greenberg functor maps the Lie-algebras g(Om ) to a d m-dimensional affine
space over K, which is naturally endowed with the structure of a Lie-ring.
8
(4) The natural maps G(Om ) G(Or ) (for r m) are transported via the Greenberg
functor to K-regular group homomorphisms. Similarly for g(Om ) g(Or ).
Additionally, the groups G(Om ) act on g(Or ) for all r m, and the action is commutes
with the natural bijection of g(Or ) and its image under the Greenberg functor [18]. Addi-
tionally, by equivariance of the Cayley map, we have explicit birational maps between the
centralizer of an element x g(Or ) within g(Or ), and the centralizer of x g(Or ) under
the conjugation action of G(Or ).
Given m N we denote the functor FOm by Fm ..
2.4. Congruence subgroups and quotients. We fix the notation G = G(o), g = g(o),
= G(O) and = g(o). Subscript notation is used to denote congruence quotients-
Gr = G(om ), gm = g(om ), m = G(Om ) and m = g(Or ).
The maps r : O Or (resp. m,r : Om Or ) are extended coordinate-wise to MN (O)
(resp. MN (Om )). For {G, } or {g, }, we write r = Ker (r : r ) for the
congruence subgroups and rm = Ker (m,r : m r ) for the congruence subquotient.
The local Frobenius automorphism : O O, is extended coordinate-wise to Mn (O)
and the group G is the group of fixed points of . Similarly, for all congruence subgroups,
quotients, subquotients and Lie-algebras.
2.5. Properties of the Cayley map. Let us recapitulate the main properties of the Cayley
map which will be required in this paper.
Lemma 2.5.1. Let m N, with r m. Let cayOm : m 99K m be the Cayley map, as
defined over Om in Section 2.1.2, and let cayc m denote its image under Fm . The following
hold.
(Cay1) The map cay c m is a birational equivalence Fm (m ) 99K Fm (m ). Furthermore, its
restriction to Fm (rm ) is a homeomorphism, whenever r 1, and is an isomorphism
of abelian groups if 2r m.
(Cay2) The map cay c m is Fm (m )-equivariant with respect to the adjoint action on Fm (m )
and with respect to group conjugation on Fm (m ).
(Cay3) The diagram in Figure 2 commutes.

cay
cm
Fm (m ) Fm (m )

m,r m,r
cay
cr
Fr (r ) Fr (r ).

Figure 2.
Proof. For any r 1, the group rm is contained in the nilradical of m , and in particular the
Cayley map is defined on rm and induces a rational map into the unipotent radical G1m Gm
(see [18, Proposition 4.3]). Furthermore, the Cayley map serves as its own inverse on the
r
image cay
[ m (m ). This implies the second assertion of (Cay1). The first assertion follows by
restricting exact sequence
m,1
0 1m m 1 0
9
to the domain D(Om ) m of cay[ m , and showing that it is a dense open subset of m .
Similarly, one proves the analogous statement for the image of cay
[ m to obtain birational-
ity. Properties (Cay2) and (Cay3) hold by preservation of fibre products by the Greenberg
transform [7, 4, Theorem]. 
As should become apparent in the sequel, the properties described in Lemma 2.5.1 con-
stitute a substantive portion of the required properties in order to complete the analysis of
Section 3.
2.6. Group characters. In general, given finite groups and characters Irr()
and Irr(), we denote by the restriction of to , and by the character in-
1
P
duced from in . We also write hf, gi = || f ()g() for the standard -invariant
inner-product on C[]. Group commutators are denoted by (x, y) = xyx1 y 1 . Lie-algebra
commutators are denoted by [x, y] = xy yx.
The Pontryagin dual of an abelian group is denoted by b = Homcont. (, C ). If is
given with additional structure (e.g. a ring or a Lie-algebra), then b refers to the Pontryagin
dual of the abelian group underlying .

3. Regular elements and regular characters


3.1. Regular elements. The starting point of our analysis of regular characters of is
to develop an understanding of orbits of regular elements in m . The methods which we
apply, influenced by [9], are based on an algebro-geometric analysis of such orbits. Recall
that an element of a reductive algebraic group over an algebraically closed field is said to be
regular if its centralizer is an algebraic group of minimal dimension among such centralizers
[21, 3.5]. Following [9], this definition is extended to elements of m with respect to the
conjugation action of m .
Definition 3.1.1. Let m 1. An element x m is said to be regular if the group
Fm (Cm (x)) is of minimal dimension over K among such centralizers.
Applying an argument similar to [9, Proposition 3.3] it can be shown that the minimal
dimension of such a centralizer is equal to the dimension of a Cartan subgroup of m .
Invoking Theorem 4.5 of [18], this dimension can be shown in this case to be n m, where
n is the absolute rank of the group G (see Corollary 3.1.6 below).
The following theorem lists the main properties of regular elements of m , which will be
proved here.
Theorem 3.1.2. Let G be a classical algebraic group scheme with Lie-algebra g = Lie(G)
and let = G(O) and = g(O). Let x . Given m N write xm = m (x) m .
(1) If xm is a regular element of m , then dimK Fm (Cm (xm )) = m n.
(2) The element xm is regular if and only if x1 is regular in 1 .
(3) Suppose xm m is regular. The restriction of the reduction map m,1 to Cm (xm )
is onto C1 (x1 ).
We note that the proof of Assertion (1) of Theorem 3.1.2 does not rely on any specific
properties of classical groups, and that all key statements appear in [18, Section 4]. Assertion
(2) is the main point in which the assumption of G being a classical group is invoked, whereby
the dimension of the variety 1 (Cm (xm )) is shown to be bounded from below using the
10
properties of the Cayley map. The analogous statement regarding Lie-algebra stabilizers,
however, is quite general and is proved here without any assumption on the group G (see
Lemma 3.1.7 below). For the third assertion of the theorem, we require an explicit description
of the centralizers of regular element in classical groups over K (Proposition 3.1.11), which
we complete in Section 5.2.1.
Once the proof of Theorem 3.1.2 is complete, we return to analyse the case of regular
elements of gm = m . Recall that an element x gm is said to be regular, if it is regular as
an element of m .
Theorem 3.1.3. Let G be a classical group as above and let x g = g(o). Assume
xm = m (x) is regular for some (for all) m N. Then x is a regular element of g and
CG (x) = lim CGm (xm ).

We then deduce the following.
Corollary 3.1.4. Let G be as in Theorem 3.1.3, and x g such that xm = m (x) is a
regular element of gm . Then CGm (xm ) is abelian.
3.1.1. General properties of the groups m . We begin our way towards the proof of Theo-
rem 3.1.2 by examining some basic properties of the group m (m N) and of centralizers
of elements of m , when considered as algebraic varieties over K.
Lemma 3.1.5. (1) The variety Fm (m ) is a connected algebraic group over K of dimen-
sion m d, where d = dim G;
(2) The unipotent radical of Fm (m ) is Fm (1m );
(3) Let T be a K-split maximal torus of G, and let T1 = T(K) 1 . The splitting map
s : 1 m embeds T1 as a maximal torus.
(4) The centralizer of s(T1 ) in m is equal to Tm = T(Om ) and is a Cartan subgroup
and has dimension n m over K, i.e. dimK Fm (Tm ) = n m
Proof. The assertion regarding the dimension of m follows from [7, 1, Proposition 4]. All
other assertions appear, in greater generality, in sections 3 and 4 of [18]. 
Lemma 3.1.5 implies the first assertion of Theorem 3.1.2.
Corollary 3.1.6. Let x m be regular. Then dimK Cm (x) = n m.
Proof. The alternative proof of [17, 3.5, Proposition 1] shows that the minimal centralizer
dimension of such a centralizer is equal to that of a Cartan subgroup of m , given that the
Cartan subgroups of m are abelian and that their union forms a dense subset of m . The
former of these conditions holds by [18, Theorem 4.5], and the latter by [3, IV 12.1]. 

3.1.2. Regularity and the reduction maps. We now turn to prove the second assertion of
Theorem 3.1.2. As mentioned above, the first step towards the proof is an analogous result to
[9, Lemma 3.5] in the Lie-algebra setting. Following this, we use the properties of the Cayley
map in order to transfer the result to the group setting and consequently the equivalence of
regularity of an element of m and of its image in 1 .
Lemma 3.1.7. Let x and put, for any m N, xm = m (x) m . For any m N, the
Lie algebra m,1 (Cm (xm )) is of dimension greater or equal to n, as a vector space over K.
11
Proof. Assume towards a contradiction that the statement of the lemma is false, and let m
be minimal such that dimK 1 (Cm (xm )) < n. Note that, since m,1 r,m = r,1 for all r > m
we also have that dimK r,1 (Cr (xr )) < n for all r m.
Fix r m, and consider the filtration
Cr (xr ) C1r (xr ) . . . Cr1
r
(xr ) 0. (3.1)
We have that
r1
X 
dimK Fr (Cr (xr )) = dimK Cir (xr )/Ci+1
r
(xr ) (3.2)
i=0

From the definition of the maps r0 ,1 (r0 N) and the isomorphisms y 7 i y : Cri (xri )

Cir (xr ), we have that

Cir (xr )/Ci+1
r
(xr ) ' Cri (xri )/C1ri (xri ) ' ri,1 Cri (xri ) ,
for all 0 i r 1. In particular, in (3.2), we get that
r1
X 
dimK Fr (Cr (xr )) = dimK ri,1 Cri (xri )
i=0
m1
X r
X
= dimK i,1 (Ci (xi )) + dimK i,1 (Ci (xi ))
i=1 i=m
d (m 1) + (n ) (r m),
for some integer 1. Note that, for any r N, Fr (Cr (xr )) is the Lie-algebra of the
group Fr (Cr (xr )). In particular, by Corollary 3.1.6, we have that dimK Fr (Cr (xr )) =
dimK Fr (Cr (xr )) r n. Thus, the above equation implies that
n r d (m 1) + (n ) (r m). (3.3)
which means that r d (m 1) m (n ) for all r > m. A contradiction, since r
can be chosen arbitrarily large and the right-hand side of (3.3) is constant. 
Using Lemma 2.5.1, we now pass to the group setting.
Proposition 3.1.8. Let x and xm = m,1 (x) for all m N. The set 1 (Cm (xm )) is a
K-algebraic variety of dimension greater or equal to n.
Proof. Properties (Cay2) and (Cay3) of the Cayley map imply the commutativity of the
following square

cay
cm
Fm (Cm (xm )) Fm (Cm (xm ))

m,1 m,1
cayK
m,1 (Cm (xm )) m,1 (Cm (xm )) .

By (Cay1), and the properties of the fibre product, it follows that the two terms of the
bottom row are of the same dimension as varieties over K. 
12
From this we are now ready to deduce the second assertion of Theorem 3.1.2.
Proposition 3.1.9. Let x and let xm = m (x) for all m N. Then xm is regular if and
only if x1 is regular.
Proof. The proposition is proved by induction on m, the case m = 1 being trivially true.
Consider the following exact sequence
1m
1 / C1m (xm ) / Cm (xm ) / C1 (x1 ). (3.4)
By (Cay1) and (Cay2), we have that C1m (xm ) ' Cm1 (xm1 ) which has the same K-
dimension as Cm1 (xm1 ). 
If x1 is regular then by induction we have that dimK Fm1 Cm1 (xm1 ) = n(m 1) and
hence

dimK Fm (Cm (xm )) dimK Fm1 Cm1 (xm1 ) + dimK C1 (x1 ) = m n.
Conversely, if x1 is not regular,
 then by induction xm1 is not regular, and the dimension
over K of Fm1 Cm1 (xm1 ) is strictly greater than n(m 1). By Proposition 3.1.8, we
get that

dimK Fm (Cm (xm )) = dimK Fm1 Cm1 (xm1 ) +dim 1 (Cm (xm )) > n(m1)+n = nm,
and xm is not regular. 
Before tuning to discuss the final assertion of Theorem 3.1.2, let us observe a simple
corollary of Lemma 3.1.7, which is the Lie-algebra version of the assertion.
Corollary 3.1.10. In the notation of Proposition 3.1.9, assume xm is regular for some
m N. The restriction of m,1 to Cm (xm ) is onto C1 (x1 ).
Proof. Proposition 3.1.9 implies that x1 is regular and hence dimK C1 (x1 ) = n. By Lemma 3.1.7,
the image of Cm (xm ) under m,1 is a subspace of C1 (x1 ) of the same dimension, whence
the assertion. 
3.1.3. The image of m,1 on Cm (xm ). In this section we wish to complete the proof of the
third assertion of Theorem 3.1.2. To do so, we require the following proposition, which
relies on the properties of classical groups. The proof of Proposition 3.1.11 requires some
additional notation, which will be introduced in Section 5, and is therefore postponed until
Section 5.2.1.
Proposition 3.1.11. Let L be either K or K alg (or any other algebraically closed field which
is also an o-algebra), and let H = G(L) and h = g(L) its Lie-algebra. Assume x h is
regular. Then
CH (x) = CH (x) Z(H).
In particular, |CH (x) : CH (x) | 2 and CH (x) is abelian.
Proof of Theorem 3.1.2.(3). Note that, by Proposition 3.1.8 and Chevalleys Theorem, the
image of Cm (xm ) under m,1 contains the connected component C1 (x1 ) of the identity
in C1 (x). Additionally, the center Z(m ) of m is clearly contained in Cm (xm ) and is
mapped under m into Z(1 ). This implies the inclusion
C1 (x1 ) m,1 (Cm (xm )) (C1 (x1 )) Z(1 ).
13
Invoking Proposition 3.1.11, and recalling that 1 = G(K), the inclusion above is actually
an equality. 

3.1.4. Returning to the -fixed setting. Our aim in this subsection is to complete the proof
of Theorem 3.1.3. An initial step towards the proof is to show that the third assertion of
Theorem 3.1.2 remains true when replacing the groups m and Lie-algebras m with their
-fixed counterparts, i.e. Gm and gm . Recall that : O O was defined in Section 2.2 to
be the local Frobenius automorphism of O over o, given on its quotient K by () = |k| and
extended A to MN (O). Also, recall that by definition, an element x gm is regular if and
only if it is a -fixed element of m .
The first Proposition we require in order to prove Theorem 3.1.3 is the following variant
of Langs Theorem.
Lemma 3.1.12. Let m N and let xm gm be a regular element, with x1 = m,1 (xm ). For
any g C1 (x1 ) let Fg = {g 0 Cm (xm ) | m,1 (g 0 ) = g} denote the fibre of the restriction of
m,1 to Cm (xm ) lying above g. Given g CG1 (x1 ), define a map L : Fg F1 by
h 7 h (h)1 .
Then L is a surjective map onto F1 .
Proof. Note that the sets Fg0 (g 0 C1 (x1 )) are simply the cosets of the subgroup F1 =
C1m (xm ). In particular, by (Cay1) and (Cay2), the varieties Ag0 are isomorphic as algebraic
varieties to C1m (xm ) ' Cm1 (xm1 ) and hence are affine (m 1)n-dimensional spaces over
K.
Since the maps and m,1 commute and g is assumed fixed by , we have that L is
well-defined. The surjectivity of L now follows as in the proof of the classical Lang Theorem
(see, e.g. [17, I, 2.2]). 
From this we deduce
Corollary 3.1.13. Let xm gm be regular and x1 = m,1 (xm ). The restriction of m,1 to
CGm (xm ) is onto CG1 (x1 ).
Proof. Lemma 3.1.12 implies that g CG1 (x1 ), there exists an element hCm (xm ) such that
m,1 (h) = g and such that L(h) = h(h)1 = 1. In particular, h is fixed under and hence
h CGm (xm ). 
Remark. Note that the analogous statement for the Lie algebras Cgm (xm ) and Cg1 (x1 ) does
not require the machinery of Lemma 3.1.12, and can be deduced from the surjectivity of the
map m,1 : Cm (xm ) C1 (x1 ) and the triviality of the first Galois cohomology group of a
vector space.
Another necessary ingredient in the proof of Theorem 3.1.3 is the connection between the
groups CGm (xm ) and CGr (xr ), where r m and x g is such that xm is regular.
Lemma 3.1.14. Let x g and m N be such that xm = m (x) is regular in gm . Let
1 r m. Then
(1) The map m,r : Cgm (xm ) Cgr (xr ) is surjective.
(2) The map m,r : CGm (xm ) CGr (xr ) is surjective.
14
Proof. (1) We prove the assertion by induction on r, the case r = 1 being given by
Corollary 3.1.10. Consider the following commutative diagram (Figure 3), in which
both rows are exact by induction hypothesis.
m,r1
Cgr1
m
(xm ) Cgm (xm ) Cgr1 (xr1 ) 0
m,r

Cgr1
r
(xr ) Cgr (xr ) r,r1
Cgr1 (xr ) 0

Figure 3.

By the Four Lemma (on epimorphisms), in order to prove the surjectivity of the
map m,r : Cgm (xm ) Cgr (xr ), it would be sufficient to show that the restricted map
m,r : Cgr1
m
(xm ) Cgr1
r
(xr ) is surjective. To show this, consider the commutative
map in Figure 4, in which the maps on the top and bottom rows are given the o-
module isomorphism y 7 r1 y, and the map on the left column is surjective by
Corollary 3.1.10.


Cgmr+1 (xmr+1 ) Cgr1
m
(xm )
1

Cg1 (x1 ) Cgr1
r
(xr )

Figure 4.

(2) In the group setting, one applies a similar argument to the previous case, considering
the commutative diagram in Figure 5, and proving the surjectivity of the left most
arrow by applealing to the surjectivity of the map Cgmr+1 (xmr+1 ) Cg1 and the
bijections y 7 cay( r1 y) from Cgmr+1 (xmr+1 ) to CGr1
m
(xm ) and from Cg1 (x1 ) to
CGr1
r
(x r ).

m,r1
CGr1
m
(xm ) CGm (xm ) CGr1 (xr1 ) 0
m,r

CGr1
r
(xr ) CGr (xr ) r,r1
CGr1 (xr ) 0

Figure 5.


The second assertion of Theorem 3.1.3 now follows.
15
Proposition 3.1.15. Let x g and xm = m (x) for any m N. Assume xm is regular for
some (for all) m. Then
CG (x) = lim CGm (xm ) and Cg (x) = lim Cgm (xm ).

Proof. Given gm CGm (xm ) one inductively invokesLemma 3.1.14 to construct a converging
sequence (gr )rm such that gr CGr (r (x)) and such that r,r0 (gr ) = gr0 for all r r0 m.
The limit g = limr gr is easily verified to be an element of CG (x), which is mapped by m to
gm . The surjectivity of m : Cg (x) Cgm (xm ) follows likewise. 
To finish the proof of Theorem 3.1.3, we now prove that the lift x g of a regular element
xm gm is a regular element of g, i.e. that its centralizer in G(K alg ) has minimal dimension.
Clearly, it suffices to prove this claim for the case m = 1.
To prove the claim, we fix x1 g1 and let x g be such that 1 (x) = x1 . Let Cox be the o-
defined group subscheme of G, defined by the condition of commuting with x, and let Ckx1 be
he k-defined subscheme of G Spec o Spec k, defined by the condition of commuting with x1 .
From the definition of a fibre product, it holds that Ckx1 = Cox Spec o Spec k. Additionally,
it is immediate to verify that the group of o points of Cox (o) is CG (x) and that the group of
k points of Ckx1 is CG1 (x1 ).
The argument we invoke here in order to prove the regularity of x is due to A. Stasinski
Lemma 3.1.16. In the notation above, the dimension of the group scheme Cox is lesser or
equal to that of Ckx1 .
Proof. The argument is based on [5, IV, (13.1.6)]. Let : Cox Spec o be the structure
morphism. By Chevalleys upper semicontinuity theorem [5, IV, Theorem 13.1.3], for any
e N, the set of y Cox such that dim 1 ((y)) e is closed.
Let s be the closed point and t the generic point of Spec o (i.e. t = {0} and s = p),
and let e = dim Cox = dim 1 (t). Chevalleys theorem implies that the set of y Cox
such that dim 1 ((y)) < e is open and does not contain t, hence empty. It follows that
dim 1 ((y)) e for all y Cox and in particular for y in the fibre of over s we have
dim Ckx1 = dim 1 (s) e = dim Cox .

In particular, Lemma Lemma 3.1.16, we have that
dimK alg CG(K alg ) (x) = dim Cox dim Ckx1 = dimK C1 (x1 ) = n,
whenever x1 is regular. The regularity of x in g follows since the minimum value of centralizer
dimension of an element of g is n = rk(G) by [21, 3.5, Proposition 1].
Remark. Note that the converse of Lemma 3.1.16 is in general false. That is to say, if x g is
regular, it is rarely the case that the reduction of x to any of the congruence quotients gm is
regular. As concrete example, consider the case x = ( 1+ 0
0 1 ) gl2 (o) and its coordinatewise
reduction modulo p.
Proof of Theorem 3.1.3. The fact that CG (x) = lim CGm (xm ) was shown in Proposition 3.1.15.

Additionally, given xm gm let x g be a lift of xm and x1 = 1 (x). By Lemma 3.1.7, the el-
ement x is regular in g and hence CG(K alg ) (x) is abelian, by Proposition 3.1.11. Consequently
CG (x) is abelian as well and hence so is its quotient CGm (xm ). 
personal communication
16
3.2. Regular characters. At this point, our description of the regular elements of the Lie-
algebras gm is sufficient in order to initiate our description of regular characters of G, and
to establish some of the properties which we shall require in the proof of Theorem I. Let us
recall some of the basic definitions we require from here on.
Let Irr(G) denote the set of continuous complex-valued irreducible characters of G, and
recall that the level of a character Irr(G) is define to be the minimal ` N{0} for which
the character is trivial on G`+1 . Recall that for any m N, the map x 7 1 + m x induces
an isomorphism between the abelian group underlying g1 and Gm m+1 . Being a semisimple Lie-
algebra over k, g1 is endowed with a non-degenerate G1 -invariant bilinear pairing (x, y).
It follows that the map y 7 y , where y (x) = ((x, y)) (and : k C is a fixed non-
trivial character) is an isomorphism of g1 with its Pontryagin dual gb1 . Given y g1 , define
y Irr(Gm m
m+1 ) by y (1 + x) = y (x). The following holds.

Lemma 3.2.1. Let m N. The map y 7 y above is a G1 -equivariant bijection of g1 onto


Irr(Gm
m+1 ).

Given a character Irr(G) of level ` 1, we consider the restriction of to the group


G``+1' g1 . By Clifford theory and Lemma 3.2.1, there existsPa unique G1 -orbit 1 () g1
such that this restriction decomposes as sum of the form e y1 () `y (e 1).
Definition 3.2.2. A character Irr(G) as above is called regular if the associated
G1 -orbit 1 () g1 consists of regular elements of g1 . In this situation, we say that the
character lies above the orbit 1 (). The orbit 1 () is called the residual orbit of .
Our description of regular characters of G follows the path of [11, 7] and [12], and relies
on Clifford theory in order to describe the irreducible regular characters of G in terms of
the constituents of their restriction to principal congruence subgroups of G. The first step
we take towards this analysis is to consider the irreducible characters of abelien congruence
subquotients of G.
3.2.1. Abelian subquotients of G and their irreducible characters. In this subsection we de-
scribe the set of characters of the group Grm , where m2 r m. Note that, as (Gr , Gr ) Gm ,
the group Grm is abelian. Similarly to the case of g1 , the Lie-algebra g = g(o) can be endowed
with a G-invariant non-degenerate symmetric bilinear form, which we take to be given by
o (x, y) = Tr(x y), (x, y g glN (o)). (3.5)
Let : K C be a fixed character which satisfies { K | |o = {1}} = o; see [2, 5.3]

for the explicit construction of such a character. Given y gm , define a map my : gm C
by
m
m

y (x) = o (
x, y) , (3.6)
where x, y g are lifts of x, y, respectively. We have the following.
Proposition 3.2.3. Let m N.
(1) For any y gm the map my of (3.6) is well-defined.
(2) The correspondence y 7 m
y is a Gm -equivariant bijection of gm onto its Pontryagin
dual gc
m.
A form on g is said to be non-degenerate if (o, o) is not included in p. Equivalently, is non-degenerate
if its representing matrix with respect to some o-basis of g is invertible over o.
17
The proof of Proposition 3.2.3 is omitted. In the succeeding section, the maps m y defined
above are used use in order to describe the set of characters of the group Grm , whenever
m
2
r m, via the isomorphism grm Grm which exists in this case. This alone is enough
in order to allow us to construct all regular character of odd level of G. In order to approach
the general case, we will need to slightly extend the definition to the case where m3 r m.
To this end, we use a truncated form of the exponential map, the definition of which we now
recall.
m
Lemma 3.2.4. Let r, m N with 3
r m. The truncated exponential map, defined by
1
exp(x) = 1 + x + x2 (x grm ), (3.7)
2
is a well-defined bijection of grm onto the group Grm , and is equivariant with respect to the
action of Gm . The inverse of exp is given by
1
log(1 + x) = x x2 (1 + x Grm ). (3.8)
2
In the case of interest for this section, where m2 r, the exponential map is simply given

by exp(x) = 1 + x and defines an isomorphism of abelian groups grm Grm . The conjunction
of Proposition 3.2.3 and Lemma 3.2.4 implies the following.
Corollary 3.2.5. Let m
2
r m. For any y gmr , let y : Grm C be defined by
y (x) = mr r log(x)

y (3.9)
where mr
y g[mr is as in (3.6). The map y 7 y is a well-defined Gm -equivariant bijection
of gmr onto Irr(Grm ).
Before concluding this subsection, let us recall some basic formulas regarding the truncated
exponential and logarithm map. The proofs of these formulas are standard, and can be
deduced in the current case by direct computation.
m
Lemma 3.2.6. Let r, m N be such that 3
r m. For any x, y grm ,
log ((exp(x), exp(y))) = [x, y] , (3.10)
where (exp(x), exp(y)) denotes the group commutator of exp(x) and exp(y) in Grm . Further-
more, the following truncated version of the Baker-Campbell-Hausdorff formula holds
1
log (exp(x) exp(y)) = x + y + [x, y] . (3.11)
2
3.2.2. The inertia subgroups arising from a regular character. Throughout this section we
fix regular character Irr(G) of level ` 1. For notational convenience, we put m = ` 1,
and identify with an irreducible character of Gm . Let r N be such that m2 r m and
let Irr(Grm ) be a constituent of the restriction of to the subgroup Grm . Recall that the
inertia subgroup of in Gm is defined by
IGm () = g Gm | (g 1 xg) = (x) for all x Grm .

(3.12)
By Corollary 3.2.5, there exists a unique y gmr such that = y , in the notation (3.9).
In the case where is regular, we show the following.
18
Proposition 3.2.7. In the setting described above, the inertia subgroup of is of the form
IGm () = Gmr
m CGm (
y ), (3.13)
where y is an arbitrary lift of y to gm .
Note that the inclusion in (3.13) is true regardless of whether is regular. Indeed, if
g Grm is of the form g = h z with h Gmr m , z CGm (y ), then zyz 1 = y, and by
r
definition of y = we have that for any x Gm
  (i)  (ii)
y (g 1 xg) = y z 1 h1 xh z = y z 1 xz = z1 yz (x) = y (x). (3.14)
Here the equality (i) is justified since the commutator subgroup (Gmr r
m , Gm ) is trivial, and
the equality (ii) follows from the Gm -equivariance in Corollary 3.2.5.
To prove the converse inclusion in (3.13) we require a simple lemma.
Lemma 3.2.8. Let y gmr , and let y Irr(Grm ), as in (3.9). Then
1

IGm (y) = m,mr CGmr ( y) .
Proof. Let g Gm be arbitrary, and write g = mr (g). Then
y g 1 (1 + r x)g = y(1 + r x),

g IGm (y) x gmr ,
mr
y (g 1 x
g ) = mr
y (x), x gmr ,
mr mr
gyg1 (x) = y (x), x gmr ,
g = mr (g) CGmr (
y ),
where the second equivalence follows from the definition of y, the third equivalence from
the G-invariance of the form (, ), and the final equivalence from the injectivity of the map
y 7 mr
y . 

Proof of Proposition 3.2.7. It remains to show that IGm () Gmrm CGm (y ) for any lift
y gm of y.
Let g IGm (). By Lemma 3.2.8 we have that m,mr (g) CGmr (y). By Lemma 3.1.14,
and the regularity of y gm , there exists h CGm (
y ) such that m,mr (h) = m,mr (g).
1 mr
Then gh Gm and
g = gh1 h Gmr
m CGm (y ).


4. Construction of regular characters


In this section we will complete the proof of Theorem I and Corollary 1.3.1. To begin
with, we require some preliminaries.

4.1. Preparation for the proof of Theorem I. We begin with an enumerative lemma.
Lemma 4.1.1. Let g1 be a regular G1 -orbit, and m N. Let gm be the preimage

of under the reduction map m,1 . The set is a disjoint union of exactly q n(m1) distinct
Gm -orbits.
19
Proof. Let y be arbitrary. By Proposition 3.1.9, y is regular, and by Proposition 3.1.8
the sequence
1 CG1m (y) CGm (y) CG1 (1 (y)) 1
is exact. As the Cayley map induces a bijection of Cg1m (y) onto CG1m (y) (cf. Lemma 2.5.1)
we deduce that
|G1 | n(m1)
|CGm (y)| = |CG1 (1 (y))| Cg1m (y) = q .
||

1
The claim follows, as = m,1 () = || q d(m1) . 
bm/2c
4.1.1. Characters of Gm . In the previous section, our analysis supplied us with the means
to enumerate and compute the inertia subgroup of the irreducible constituents of the restric-
tion of an irreducible regular character of level ` 1 to the subgroups Gr , where r `1 2
.
In the case where ` is and odd number, this analysis is sufficient in order to complete the
description of such regular characters.
For the case where ` is even, however, we require some additional machinery; namely, the
method of Heisenberg lifts, on which we now expand (see [4, 8]).
Fix m N and odd number and put r = m1 2
. Note that the group Grm is two-step
nilpotent, and that the group Gr+1 r r+1
m is a central subgroup of Gm . Let Irr(Gm ) be fixed,
and let y gmr1 be such that = y , as in Corollary 3.2.5. Define an alternating form
on the group Grr+1 by
r1
B (g1 Gr+1 , g2 Gr+1 ) = y ((g1 , g2 )) = mr1 (g1 , g2 Gr ).

y log ((g 1 , g2 ))
As the exponential map defines an isomorphism of grr+1 ' g(k) onto Grr+1 , by (3.10) and
the definition above, we have that
B (exp(x1 ), exp(x2 )) = mr1 r1 [x1 , x2 ] = 1 y (1 , 2 ) , x1 , x2 grr+1 ,
 
y

where i g1 are such that xi = r i , and y (, ) is the alternating bilinear form defined on
g1 by y (1 , 2 ) = Tr (1 (y) [1 , 2 ]) .
Let ry g1 denote the radical of the form y . Note that by definition
ry Tr (1 (y) [, 0 ]) = Tr ([1 (y), ] 0 ) = 0, 0 g1
Cg1 (1 (y)) (4.1)
by non-degeneracy of the Killing form.
Let Ry grm and Ry Grm be the preimages of ry under the respective reduction maps.
As Ry is abelian, the character exp (y ) = y exp gd r+1 extends to R in |r |-many non-
m y y
equivalent ways. Note that, by Proposition 3.2.3, any such extension is the restriction of a
character x 7 mr
y0 ( r x) of grm to Ry , where y 0 gmr is a lift of y gmr1 . Additionally,
by (3.14), given such an extension 0 R cy , the map 0 = log (0 ) defines a character of the
group Ry .
Any such extension 0 determines a unique character of Grm as follows. The bilinear form
y endows the quotient space g1 /ry with the structure of a symplectic space. Let j g1
be the preimage of some maximal isotropic subspace of g1 /ry . Let J grm and J Grm be
the preimages of j, under the respective reduction maps. As in the case of Ry , the map 0
extends to J and any such extension 00 defines a linear character 00 = exp (00 ) of J, which
extends and lies above the fixed extension 0 Irr(Ry ) (see Figure 6).
20
Grm grm g1
log

J J j
log

Ry Ry ry
log


Gr+1
m gr+1
m 0
log

Figure 6.

The following proposition summarizes the properties of the construction described above
r
Proposition 4.1.2. Let = (00 )Gm be the induced character. Then
(1) The character is irreducible.
(2) The character is independent of the choice of isotropic subspace J and of the choice
of extension of 0 to J.
(3) The character is the unique character of Grm lying above 0 , and any character of
Grm can be obtained in this manner.
4.2. Proof of Theorem I. Fix ` N and m = ` 1. Our first step towards the proof of
Theorem I is ??, which enumerates of all regular characters of the group Grm , where r = b m2 c,
and keeps track of their dimensions.
Proposition 4.2.1. Let g1 be a regular orbit and let m N and r = b m2 c.
(1) The set Irr(Grm | ) of characters of Grm which lie above the regular orbit consists
of exactly q n(mr1) orbits for the action of Gm .
(2) Any character Irr(Grm | ) satisfies
(
1 if m = 2r,
(1) =
(4.2)
q if m = 2r + 1,

where = dn 2
, as in Theorem I.
|G1 | n(r1)
(3) Each character Irr(Grm | ) admits exactly ||
q non-conjugate extensions
to its inertia subgroup IGm ().
Proof. As in [9, 11, 12], we separate the proof of ?? into two cases according to the parity
of m.
Assume first that m is even, so that r = m2 and Grm is abelian and isomorphic to grm .
By Lemma 4.1.1, any element of gives rise to q n(mr1) non-conjugate lifts yi gmr ,
(i = 1, . . . , q n(mr1) ), which determine q n(mr1) distinct orbits of regular characters yi
Irr(Grm ) (as in Corollary 3.2.5). Assertions (1) and (2) in the case of m even follow.
To prove (3), pick a representative z = yi in one of these orbits. By Proposition 3.2.7 and
the fact that m r = r, we have that IGm (z ) = Grm CGm ( z ) where z gm is some lift of
z gr .
21
Let Irr (CGm (z )) be some extension of the restriction of z to CGm ( z ) Grm . The
extension exists by virtue of the commutativity of CGm (
z ) (Theorem 3.1.3), and the number
of such extensions is

|CGm ( z ) Grm | = |CGr (z)| = |CG1 (1 (y))| q n(r1) .


z ) : CGm (

Gluing the characters and z , we define a map g z (gh) = (g)z (h), for g CGm (
z)
and h Grm . Then g z is well-defined and extends z to IGm (z ). Any character extension
of z to IG (z ) is obtained in this manner (see Figure 7, for a schematic description of the
groups discussed so far).

Gm

IGm (z )

CGm (
z) Grm

z ) Grm
CGm (

{1}
Figure 7. Even Case, m = 2r

Assume now that m = 2r + 1 is odd, in which case Grm is a two-step nilpotent group, and
will be handled using the method described in Section 4.1.1.
We start by considering Irr(Gr+1 m ). As in the case where m is even, we have that the
number of characters of Gr+1 m lying above the orbit is q n(mr2) (by Lemma 4.1.1). Any such
character y Irr(Gr+1 r
m ) defines a symplectic form y on the space g1 ' gr+1 , and admits
r
|ry | many non-equivalent extensions to Ry Gm . By (4.1), and the fact that 1 (y) g1
is regular, the number of extensions to Ry is |CG1 (1 (y))| = q n , and each such extension
corresponds to a unique character of Grm which lies above y (see Proposition 4.1.2). As
these are all characters of Grm which lie over the orbit , we deduce that the number of
characters of Grm lying above is precisely q n(mr2)+n = q n(mr1) , which proves (1).
Additionally, the degree of such character equals the index of a maximal isotropic subspace
1
of g1 /ry , which is q 2 (dn) = q . Hence (2).
Aiming towards the proof of (3), we now show that all such characters of Grm extend to
IGm (). To show this, by [10, Corollary 11.31], it suffices to show that extends to all
subgroups IGm () such that /Grm is p0 -Sylow subgroup of IGm ()/Grm , with p0 prime.
In the case where p0 6= p such an extension exists, as |/Grm | is a p0 -group, and hence the
group H 2 (/Gm , C ) is trivial; see [10, Theorem 11.15] and [1, Lemma 4.2].
For the case p0 = p, we Let be a p-Sylow subgroup of IGm (). Note that, as seen in
Section 4.1.1, the extension 0 Irr(Ry ) arises by picking a lift z gmr of y and considering
the the restriction to Ry = log(Ry ) of the character mr z of gmr , defined as in (3.6). Let
22
Gm

IGm ()

CGm (
z) P Grm

PJ Grm

P J

CGrm (
z) Ry

Gr+1
m

CGr+1
m
(
z)

{1}
Figure 8. Odd Case, m = 2r + 1.

z gm be such that mr (
z ) = z. We obtain
IGm () = IGm (0 ) IGm () (by Proposition 4.1.2)
z ) Grm
= CGm ( (by Proposition 3.2.7)
0
IGm ( ) IGm () = IGm (), (4.3)

and in particular IGm () = IGm ().


Note that, as Grm is a p-group, the p-Sylow group is of the form P Grm for a Sylow-
p subgroup P CGm ( z ). By [12, Lemma 3.4], the space gm m+1 /ry admits a P -invariant
maximal isotropic subspace j with respect to y . Let J Grm be the preimage of j and
let 00 Irr(J) be an extension of 0 , as in Section 4.1.1 (see Figure 8, for a schematic
description of the groups discussed in this case).
Pick a character Irr(P ) which lies above the restriction of 0 to P J = CGrm (z) and
g00 be the glued character, defined as in the previous case. Let
let g00 )P Grm be the
= (
induced character. We claim that is the desired extension of .
To show this, first note that the degree of
is
(1) = |P Grm : P J| = |Grm : J| = (1),

where the second equality holds since P J Grm = J.
Additionally, from Frobenius reciprocity, we have
r
h g00 P Gm )J , 00 iJ ,
Grm , iGrm = h(
which is non-zero by Mackeys decomposition. In particular,
Grm = and
is irreducible.
23
Lastly, by [10, Corollary 6.17], the number of extensions of to IGm () is equal to the
number of characters of the group
IGm ()/Grm ' CGm ( z ) Grm ) ' CGr (r (
z )/ (CGm ( z )),
which is |CG1 (1 (y))| q n(r1) . The proof of ?? is now complete. 
Proof of Theorem I. Let m = ` 1 be fixed and r = b m2 c. Let Irr(Grm | ) be a regular
character lying above . By ??, the character extends to its inertia subgroup, and hence
induces irreducibly from IGm () to Gm , by [10, Theorem 6.11]. In particular, the degree of
this extension is
|Gr |
|Gm : IGm ()| (1) = (1) = || q (`1) ,
|CGr (r (
z ))|
by ??.(2). Here z is as in the proof of the proposition.
Additionally, since any character of Gm lying above can be constructed in this manner,
by considering a constituent of its restriction to Grm , we obtain from ??.(3) that

[`]
|G1 | n(r1) |G1 | n(`1)
Irr (G | ) = |Irr(Grm | )| q = q ,

|| ||
where the last equality follows from the first assertion of ??. 

5. The classical groups


Our final goal for this article is to compute the regular representation zeta function of the
classical groups. Following Corollary 1.3.1, to do so, we classify the regular orbits in the
space of orbits Ad(G1 )\g1 and compute their cardinalities, in order to obtain a formula for
the Dirichlet polynomial
X |G1 |
Dg (s) = ||s , (5.1)
X
||
defined in Corollary 1.3.1.
As it turns out, the cases where G is a classical group of type Bn or Cn (i.e. G = Sp2n
or G = SO2n+1 ) can be handled simultaneously and will be analysed in Section 5.3. The
case of the groups of the form Dn , i.e. even-dimensional orthogonal groups, is slightly more
elaborate. The analysis for this case is carried out in Section 5.4. The main difference
between the two cases lies in the fact that regularity of elements of sp2n (k) and so2n+1 (k) is
equivalent to their being a regular matrix (cf. [22, 5]). This equivalence fails to hold for
even-orthogonal groups; see Lemma 5.4.1 below. In both cases, we obtain a classification of
the regular orbits in the Lie-algebra g1 in terms of the minimal polynomial of the elements
within the orbit.
Recall that two matrices x, y MN (k) are said to be similar if there exists a matrix
g GLN (k) such that y = gxg 1 . Our description of regular orbits of g1 follows the
following steps.
(1) Classification of all similarity classes in glN (k) which intersect the set of regular
elements in g1 non-trivially;
(2) Description of the intersection of such a similarity class as a union of Ad(G1 )-orbits;
(3) Computation of centralizer of an element of a regular orbit within G1 .
24
5.1. Statement of results. Let Xn denote the set of triplets t = (r, A, B) Z0
Mn (Z0 ) Mn (Z0 ) which satisfy
n
X
r+ de (Ad,e + Bd,e ) = n. (5.2)
d,e=1

Definition 5.1.1. Given a polynomial f (t) k[t] of degree N satisfying f (t) = (1)N f (t),
and 1 d, e n, let Ad,e (f ) denote the number of distinct monic irreducible even polyno-
mials (t) = (t) of degree 2d which occur in f with multiplicity e, and let Bd,e (f ) denote
the number of pairs { (t), (t)}, with (t) irreducible and coprime to (t), such that is
of degree d and occurs in f with multiplicity e. Let r(f ) be the maximal integer such that
t2r(f ) divides f . Put t(f ) = (r(f ), A(f ), B(f )) Xn , where A(f ) and B(f ) are the matrices
(Ad,e (f ))d,e and (Bd,e (f ))d,e respectively. In this setting we say that f is of type t(f ).
The results of the classical groups of type Bn and Cn are summarized in Theorem 5.1.2
below. The proofs of Assertions (1), (2) and (3) of the theorem re carried out in sections
5.3.1, 5.3.2 and 5.3.3 respectively.
Theorem 5.1.2. Assume char(k) 6= 2. Let V = kN and let B be a non-degenerate bilinear
form which is alternating if N is even, and symmetric otherwise. Let G1 = G(k) be the
group of isometries of B and g1 = g(k) the Lie-algebra of anti-hermitian operators.
Let x MN (k) have minimal polynomial mx k[t].
(1) The element x is similar to a regular element of g1 if and only if mx has degree N
and satisfies mx (t) = (1)N mx (t).
Furthermore, assuming x g1 is a regular element, let = Ad(G1 )x denote its orbit under
G1 .
(2) If N is even and mx (0) = 0, then the intersection of the similarity class of x with
g1 is the disjoint union of two orbits of G1 . Otherwise, any element y g1 which is
similar to x is also Ad(G1 )-conjugate to it.
(3) Let t = t(mx ) = (r(mx ), A(mx ), B(mx )) as in Definition 5.1.1. Then
  Qn 2i
2n2 1 i=1 (1 q )
|| = q Q d d,e x (1 q d )Bd,e (mx )
A (m )
,
2 1d,en (1 + q )

where = 1 if N = 2n is even and mx (0) = 0, and = 0 otherwise.


In the case of of groups of type Dn , the results obtained are summarized in Theorem 5.1.3.
The proof of Assertions (1),(2) and (3) of the theorem appear in sections 5.4.1, 5.4.2 and
5.4.3.
As already mentioned, the analysis undertaken here is valid only for the case where the
residue field of o is of odd characteristic. For technical reasons, we also need to omit the
specific case of k = F3 from the analysis (see Lemma 5.4.9 below). Additionally, the case
n = 1 is omitted as well, as in this case so2 (o) is an abelian group.

Theorem 5.1.3. Assume |k| > 3 and char(k) 6= 2. Let N = 2n be an even number, with
n 2. Let V = kN and let B + and B be non-degenerate forms on V of Witt index n and
n 1, respectively. For  {+, }, let G1 be the group of isometries of B  and let g1 be the
Lie-algebra of anti-hermitian operators.
Let x MN (k) have minimal polynomial polynomial mx (t).
25

(1) The element x is similar to an element of g+ 1 or g1 if and only if mx (t) = (1)
deg mx
mx (t)
and
(a) mx (0) 6= 0 and mx has degree N ; or
(b) mx (0) = 0 and the characteristic polynomial of x is cx (t) = t mx (t).
Furthermore, the following hold.
(a) If mx hasP degree N , mx (t) = mx and mx (0) 6= 0, put t = t(mx ) = (r(mx ), A(mx ), B(mx ))
and = eAd,e (mx ). Then x is similar to an element of g 1 if and only if is
odd. Otherwise x is similar to an element of g+ 1 .
(b) If mx has degree N 1, mx (t) = mx (t) and the characteristic polynomial of
x is cx (t) = t mx (t), then x is similar to an element of g+ 1 , as well as to an

element of g1 .
Assume now that x g1 is a regular element and let = Ad(G1 )x be its orbit under G1 , for
 {+, } fixed.
(2) In the case where mx (0) = 0, the intersection of the similarity class of x with g1 is the
disjoint union of two orbits of G1 . Otherwise, any element y g1 which is similar
to x is also Ad(G1 )-conjugate P to it.
(3) (a) If mx (0) 6= 0, put = d,e e Ad,e (mx ) as above. Then
(1 + (1) q n ) n1 2i
Q
2n2 i=1 (1 q )
|| = q Qn d A (m ) d
i=1 (1 + q )
d,e x (1 q ) (mx )
Bd,e

(b) Otherwise, if mx (0) = 0 put t = t(t mx ).


(i) The orbit of x under G+ 1 is of order

(1 q n ) n1 2i
Q
2n2 i=1 (1 q )
|| = q Qn d d,e x (1 q d )Bd,e (mx )
A (m )
.
i=1 (1 + q )

(ii) The orbit of x under G 1 is of order

(1 q n ) n1 2i
Q
2n2 i=1 (1 q )
|| = q Qn d Ad,e (mx ) (1 q d )Bd,e (mx )
.
i=1 (1 + q )

Once Theorem 5.1.2 and Theorem 5.1.3 are proved, the proof of ?? can be completed, by
direct computation.
5.2. Preliminary matters. To begin with, we gather some preliminary results which will
be required in the proofs of Theorem 5.1.2 and Theorem 5.1.3, as well as the postponed
proof of Proposition 3.1.11.
5.2.1. Centralizers over algebraically closed field. For this section, let L be an algebraically
closed field which is also an o-algebra, e.g. L = K or L = K alg . In this section we investigate
the centralizer of a regular element x h = g(L) and prove Proposition 3.1.11, which states
that the centralizer of x in H = G(L) can be described as CH (x) = CH (x) Z(H). In
particular, we deduce that CH (x) is abelian and connected modulo center.
Let W = LN be the N th cartesian power of L. The space W is endowed with a non-
degenerate bilinear form BL (, ), which is defined by BL (u, v) = ut Jv (see Section 2.1) and
satisfies BL (v, u) = BL (u, v) ( = 1) for all u, v W . The group H consists of isometries
of the form BL and the Lie algebra consists h of anti-hermitian operators, i.e. elements
x End(W ) such that BL (xu, v) + BL (u, xv) = 0 for all u, v W .
26
We begin with two general lemmas regarding the centralizer of a semisimple element of h.
Lemma 5.2.1. Let s h be a semisimple element.
(1) Let L be an eigenvalue of s. Then is an eigenvalue of s as well.
(2) The space W decomposes as a direct orthogonal sum
M
W = W[] , (5.3)
[]L/{1}

where W[0] = Ker(s) and for any 6= 0


W[] = Ker(s 1) Ker(s + 1). (5.4)
Furthermore, and the restriction of BL (, ) to any of the subspaces W[] is non-
degenerate and the spaces Ker(s 1) on the right hand side of (5.4) are totally
isotropic and orthogonal.
Proof. Let E be an L-basis of eigenvectors of s and let u E be an eigenvector with eigenvalue
L. By non-degeneracy, there exists w E such that BL (u, w) 6= 0. Let L be the
eigenvalue of w. Then BL (su, w) + BL (u, sw) = 0. On the other hand
0 = BL (su, w) + BL (u, sw) = BL (u, w) + BL (u, w) = ( + ) BL (u, w) .
Since B(u, w) 6= 0 by assumption, it follows that = , and therefore (1) follows.
The decomposition in (5.3) is simple the decomposition of W into eigenspaces of s. The
non-degeneracy of the spaces W[] and the orthogonality of the sum follow at once from the
previous argument, once we show that the eigenspaces Ker(s 1) and Ker(s 1) are
orthogonal whenever 6= . This holds, as for any u Ker(s 1), v Ker(s 1) we
have
0 = BL (su, v) + BL (u, sv) = ( + )BL (u, v),
which implies BL (u, v) = 0 if 6= . 
Lemma 5.2.2. Let s h be a semisimple element. For any L put W () = Ker(s 1).
Let 1 , . . . , t L be such that {1 , . . . , t } is the set of all non-zero eigenvalues of s,
with i 6= j whenever i 6= j. Then
t
Y
CH (s) ' GL (W (j )) ,
j=1

where is the group of isometries of a non-degenerate bilinear form on W (0) = Ker(s).


Proof. Lemma 5.2.1 implies that any element g CGLN (L) (s) can be written uniquely as
a commuting product g = g1 . . . gt g0 , where for any j = 1, . . . , t the element gj acts
trivially on the subspaces W[] for any 6= j and preserves W[j ] , and g0 acts trivially on
the subspaces W[j ] for j = 1, . . . , t and preserves W[0] .
By inspecting the action of g on elements of W[] ( L) one easily verifies that the
condition g H implies that any element gj in the decomposition must also act as an
isometry of BL . It follows directly from this that g0 is taken from the group of isometries of
the restriction of BL (, ) to W (0) = W[0] . Additionally, since for any j 6= 0, the element gj
must preserves the subspaces W (j ), W (j ) W[j ] , hence is of the form gj = hj (h?j )1
for hj GLN (L) acting trivially on the subspaces W () for any 6= j and such that
hj |W (j ) = gj |W (j ) . 
27
Lemma 5.2.2 implies the following useful corollary.
Corollary 5.2.3. Assume x h is non-singular. The x is regular in h if and only if x is a
regular element of glN (L).
Proof. Let x = s + h h be the Jordan decomposition of x. By assumption, all eigenvalues
of s are non-zero and hence, by Lemma 5.2.1, CGLN (L) (s) is isomorphic to the direct product
Qt
j=1 GL (W (j )) GL (W (j )) where {1 , . . . , t } are all eigenvalues of s. Further-
more, by Lemma 5.2.2, CH (s) ' tj=1 GL (W (j )). A direct computation shows that in this
Q
case
t
Y 
CGLN (L) (x) = CCGLN (L) (s) (h) = CGL((W (j )) (h |W (j ) CGL((W (j )) (h |W (j )
j=1

and that x is regular in glN (L) if and only if h |W () is regular in gl(W ()) for all
{1 , . . . , t }. By the same token, and by Lemma 5.2.2, it holds that
t
Y
CH (x) = CCH (s) (h) = CGL(W (j )) (h |W (j ) ).
j=1

In particular, if x is a regular element of glN (L) if and only if


t
X N
dim CH (x) = dim CGL(W (j )) (h |W (j ) ) = = n,
j=1
2

which occurs if and only if x is regular in h. 


Remark. Note that the assumption that x is non-singular in Corollary 5.2.3 is crucial, as
the proof of the corollary relies heavily on the fact that the centralizer of a non-singular
semisimple element of h is a direct product of groups of the form GL(W (j )). The same
argumentation would not apply in the case where x is singular, and in fact fails in certain
cases; see Lemma 5.4.1 below.
We are now ready to complete the proof of Proposition 3.1.11.
Proof of Proposition 3.1.11. Let x h be a regular element and let x = s + h be its Jordan
decomposition, with s h semisimple, h h nilpotent and [s, h] = 0. As mentioned in
Corollary 5.2.3, since an element of H commutes with x if and only if it commutes with both
s and h, we have that CH (x) = CCH (s) (n). By Lemma 5.2.2, it follows that
t
Y  
CH (x) = CCH (s) (h) = CGL(W (j )) h |W (j ) C h |W (0) . (5.5)
j=1

Additionally, by [21, 3.5, Proposition 5] and a simple computation, it follows that the
restricted operators h |W (j ) and h |W (0) are regular as elements of the Lie-algebras of
GL(W (j )) and respectively.
By [17, III, 3.2.2] it is known that all factors in (5.5), except for C (h |W (0) ), are connected.
Furthermore, by [17, III, 1.14], we have
 
C h |W (0) = C h |W (0) Z(),
28

as char(L) 6= 2 (see [17, I, 4.3]). Taking into account the fact that Z() = 1W (0) , one
easily deduces from this the equality
CH (x) = CH (x) Z(H).
Furthermore, CH (x) is abelian by [17, Corollary 1.4], and |CH (x) : CH (x) | |Z(H)| =
|{1W }| = 2. The proof of Proposition 3.1.11 is now complete. 
5.2.2. From similarity classes to adjoint orbits. In this section we return to the setting of
groups and Lie-algebras over k and develop the tools required in order to analyse the decom-
position of intersection of the similarity class of a regular element of g1 with into Ad(G1 )-
orbits. The results appearing in this section can be also derived from [24, 2.6]. However,
as the case of regular elements allows for a much simpler argument, we present it below for
the sake of completeness.
Let V = kN be an N -dimensional vector space, identified with the space of N -fold column
vectors. Recall the definition of the non-degenerate bilinear form B = Bk , defined in Sec-
tion 2.1.1. Also, recall the involution ? of End(V ), defined by the rule B(x? u, v) = B(u, xv)
for all u, v V and x End(V ); see (2.3).
Let Sym(?; x) be the set of elements Q CGLN (k) (x) such that Q? = Q and define an
equivalence relation on Sym(?; x) by
Q1 Q2 if there exists a CGLN (k) (x) such that Q1 = a? Q2 a. (5.6)
Let x to be the set of equivalence classes of in Sym(?; x). In the case where Sym(?; x)
is abelian (e.g., when x is a regular element of glN (k)), the set x is simply its quotient by
the image of restriction of w 7 w? w to CGLN (k) (x).
Proposition 5.2.4. Let x g1 and let x denote the intersection of the similarity class of
x with g1 . There exists a map : x x such that y1 , y2 are Ad(G1 )-conjugate if
and only if (y1 ) = (y2 ).
Proof. (1) Construction of . Let y x = (Ad (GLN (k)) x g1 ) and let w GLN (k)
be such that y = wxw1 . Put Q = w? w. Note that, as x, y g1 , by applying the
anti-involution ? to the equation y = wxw1 , we deduce that (w? )1 xw? = y as
well and consequently, that Q = w? w commutes with x. Since Q? = Q, we get that
Q Sym(?; x).
Define (y) to be the equivalence class of Q in x . We claim that is well-defined,
i.e. if w0 GLN (k) is another element such that y = w0 xw01 and Q0 = w0? w0 , then
Q0 Q. Indeed, put a = w1 w0 . Then a commutes with x, and
a? Qa = w0? (w? )1 Qw1 w0 = w0? w0 = Q0 .
(2) Proof that y1 , y2 x are Ad(G1 )-conjugate if (y1 ) = (y2 ). Let w1 , w2 GLN (k)
be such that yi = wi xwi1 , and let Qi = wi? wi (i = 1, 2). Then, by assumption,
there exists a CGLN (k) (x) such that Q2 = a? Q1 a. Put z = w1 aw21 . Note that
zy2 z 1 = y1 . We claim that z G1 . This holds since for any u, v V
B(zu, zv) = B(w1 aw21 u, w1 aw21 v) = B(a? (w1? w1 )aw21 u, w21 v)
= B(a? Q1 aw21 u, w21 v) = B(Q2 w21 u, w21 v) (since Q2 = a? Q1 a)
= B(w2? u, w21 v) = B(u, v).
29
(3) Proof that y1 , u2 x are Ad(G1 )-conjugate only if (y1 ) = (y2 ). Assume now that
z G1 is such that y1 = zy2 z 1 , and let w1 , w2 GLN (k) be such that yi = wi xwi1
(i = 1, 2). Then w1 and zw2 both conjugate x to y1 , and hence, by the unambiguity
of the definition of and fact that z G1 , we have that
(y1 ) = [w1? w1 ] = [w2? (z ? z)w2 ] = [w2? w2 ] = (y2 ).

Thus, in order to analyse the adjoint orbits of regular elements of g1 , in Section 5.3
and Section 5.4, we present an explicit description of the sets x and of the image of the
corresponding map , for x g1 regular. A crucial property of the case of regular elements,
which makes the analysis of adjoint orbits feasible in this case, is that the set x can be
realized within the quotient of an etale algebra over k by the image of the algebra under an
involution. The consequence of the proposition above is that the set x decomposes into
|Im| many Ad(G1 )-orbits, a quantity which does not exceed the value 4 in the regular case.
Before concluding this subsection, we state another general lemma, which will be required
in the description of x .
Lemma 5.2.5. Let C MN (k) be the ring of matrices commuting with x, and let N / C be
a nilpotent ideal. The following are equivalent, for Q1 , Q2 Sym(?; x).
(1) There exists a C such that a? Q1 a = Q2 ;
(2) There exists a C such that a? Q1 a Q2 (mod N ).
Proof. The argument of [24, Theorem 2.2.1] applies to the case where N is any nilpotent
ideal, provided that the required trace condition holds. In the present case the condition
holds since char(k) 6= 2. 
5.2.3. Similarity classes via bilinear forms. We recall a basic lemma which would allows us
to determine when an element of glN (k) is similar to an element of g1 .
Lemma 5.2.6. Let C1 , C2 be two non-degenerate bilinear forms on a vector space V = kN ,
and assume there exists g End(V ) and k such that C1 (gu, gv) = C2 (u, v) for all
u, v V . Let x glN (k) be anti-hermitian with respect to C2 . Then gxg 1 is anti-hermitian
with respect to C1 .
The proof of Lemma 5.2.6 is by direct computation, and is omitted. In our application
we will invoke the lemma in order to prove a general matrix x is similar to an element of
g1 by constructing an non-degenerate bilinear form C on V , with respect to which x is anti-
hermitian, and such that the pair B = Bk (the fixed non-degenerate bilinear form defined
in Section 2.1.1) and C satisfy the hypothesis of the lemma.
5.3. Symplectic and odd-orthogonal group. In this section we consider the symplectic
group on V = kN , with N = 2n even, and the orthogonal group on the space V = kN , where
N = 2n + 1 is odd. The reason these two cases can be uniformly treated by the following
well-known fact.
Lemma 5.3.1. Let  {1} and let N = 2n be even if  = 1 and N = 2n + 1 be odd if
 = 1. Let C1 , C2 be two non-degenerate forms on V = kN such that Ci (u, v) = Ci (v, u) for
all u, v V and i = 1, 2. There exists k and g End(V ) such that C1 (gu, gv) = C2 (u, v)
for all u, v V . Additionally, if  = 1 then can be taken to be equal to 1.
30
Proof. See, e.g., [26, Ch. 3]. 

5.3.1. Similarity classes of regular elements. Our first objective is to classify all regular
elements of g1 upto conjugation by an element of GLN (k), i.e. the similarity classes of
regular elements of g1 . As it turns out, this is equivalent to classifying the regular similarity
classes of glN (k) which meet g1 non-trivially. The following lemma gives a criterion for a
regular matrix to be similar to an element of g1 .

Lemma 5.3.2. Let x glN (k), with minimal polynomial mx (t) k[t].
(1) If x g1 then mx (t) satisfies mx (t) = (1)deg mx mx (t).
(2) Conversely, if x is a regular element of glN (k) (and hence deg mx = N ) such that
mx (t) = (1)N mx (t), then x is similar to an element of g1 .

Proof. For the first assertion, note that for any r N we have that B(xr u, v) = B(u, (1)r xr v)
for all u, v V . Invoking the non-degeneracy of B, we deduce that (1)deg mx mx (t) is a
monic polynomial of degree deg mx which vanished at x, and hence equal to mx (t).
To prove the second assertion, in view of Lemma 5.2.6 and Lemma 5.3.1, we shall construct
a non-degenerate bilinear form C on V such that C(u, v) = C(v, u), where  = (1)N , and
such that C(xu, v) + C(u, xv) = 0 for all u, v V .
By [21, 3.5, Proposition 2], the assumption that x is a regular matrix is equivalent to
V being a cyclic module over the ring k[x]. In particular, there exists v0 V such that
(v0 , xv0 , . . . , xN 1 v0 ) is a k-basis for V . Let N 1 : V k denote the projection onto basis
element xN 1 v0 . Given u1 , u2 V let p1 , p2 k[t] be polynomials such that ui = pi (x)v0
and define
C(u1 , u2 ) = N 1 (p1 (x)p2 (x)v0 ) . (5.7)
Note that given p01 , p02 k[t], another pair of polynomials such that p0i (x)v0 = ui , it necessarily
holds that p0i pi (mod mx ) for i = 1, 2 and, by the assumption mx (t) = (1)N mx (t), it
follows that p01 (t)p02 (t) p1 (t)p2 (t) (mod mx (t)), whence that C(, ) is well-defined. It
is also obvious from the definition that C(, ) is bilinear, that C(xu, v) + C(u, xv) = 0 for
all u, v V , and that C(u, v) = (1)N 1 C(v, u) = C(v, u) for all u, v V . Let us verify
that C is non-degenerate.
Let u V be non-zero, and let p(t) be such that p(x)v0 = u. By unambiguity of the
definition of C, we may assume that deg p(t) < N . Let v = xN 1deg p v0 V . Then

C(u, v) = N 1 ((1)N 1deg p xN 1deg p p(x)v0 )

is non-zero, since tN 1deg p p(t) is a polynomial of degree N 1. 

Note that Lemma 5.3.2 gives a criterion for a regular element of glN (k) to be similar to
an element of g1 , but a-priori, not necessarily to a regular element of g1 . We will shortly
see that it is indeed the case that the similarity class of such x meets g1 at a regular orbit.
Before proving this, let us consider a pivotal example.

Example 5.3.3 (Regular nilpotent elements). Let x glN (k) be a regular nilpotent element,
i.e. mx (t) = tN . Picking a generator v0 for V over k[x] and putting E = (v0 , xv0 , . . . , xN 1 v0 ),
31
the element x is represented in the basis E by the matrix
0 1

0 1
=
. .
.. ..

. (5.8)
0 1
0
The bilinear form C of Lemma 5.3.2 is represented in this basis by the matrix
1

1
1

c= . (5.9)
.. .

(1)N 1
To show that is similar to a regular element of g1 we now pass to the algebraic closure
of k and compute the dimension of the centralizer of zz 1 in 1 , where z1 z 1 g1 . Note
that the centralizer of in GLN (K) consists of upper triangular Toplitz matrices,

a0 a1 . . . aN 1



... ... ..

. N

CGLN (K) () = | a 0 , . . . , a N 1 K, a0 6
= 0 ' k[t]/(t ) .


a0 a1


a0
Additionally, the map g 7 zgz 1 induces an isomorphism of C1 (zz 1 ) onto the subgroup
of elements y CGLN (K) () which preserve C(, ), i.e. such that y t dy = d. Computing the
dimension of this subgroup (e.g. by passing to its Lie-algebra), one easily verifies that it is
of dimension n over K, and hence is similar to a regular element of g1 .
We now deduce the first assertion of Theorem 5.1.2.
Proposition 5.3.4. Let x g1 . Then x is a regular element of g1 if and only if x is regular
in glN (k).
Proof. By definition of regularity, we need to prove that dimK C1 (x) = n if and only if
dimK CGLN (K) (x) = N . Let x = s + h be the Jordan decomposition of x over K, with s
semisimple, h nilpotent, and [s, h] = 0. Let W = KN be as in Section 5.2.1. By Lemma 5.2.1,
the space W decomposes as an orthogonal direct sum W1 W2 with respect to the bilinear
form BK , where W0 = Ker(s) and s |W1 is non-singular. Let 1 be the subgroup of
elements acting trivially on W0 and preserving W1 , and let be as in Lemma 5.2.2. Then
C1 (x) = C (x) C (x)
and
CGLN (K) (x) = CGL(W1 ){1W } (x) C{1W }GL(W0 ) (x)
0 1

and therefore the proof reduces to the cases where x is non-singular and where x is a nilpotent
element acting on W0 . The first case follows from Corollary 5.2.3, whereas the second
case follows from Example 5.3.3 and from the uniqueness of a regular nilpotent orbit over
algebraically closed fields [21, III, Theorem 1.8] 
32
In particular, Proposition 5.3.4 implies that the map x 7 mx , where mx is the minimal
polynomial of x, induces a bijection between the set of similarity classes which intersect
g1 at regular elements, and the set of polynomials f (t) k[t] such that deg(f ) = N and
f (t) = (1)N f (t).

5.3.2. From similarity classes to adjoint orbits. Our next goal for this section is to analyse
the decomposition of the intersection of such a similarity class with g1 into Ad(G1 )-orbits. To
this end, we will invoke Proposition 5.2.4. Before doing so, let us introduce some notation.
Notation 5.3.5. Given a polynomial f (t) k[t] we write k(f ) for the quotient ring k[t]/(f ).
For example, if f is an irreducible polynomial over k then k(f ) stands for the splitting field
of f . It will also be convenient to write GL1 (k(f )) for the group of units of k(f ).
In the case where f (t) = f (t), we put f to denote the k-involution f : k(f ) k(f ),
induced from t 7 t. In this case we put U1 (k(f )) for the group of elements k(f ) such
that f () = 1.
Proposition 5.3.6. Let x g1 be a regular element. If x is singular and N is even, then
the similarity class of x meets g1 at two distinct Ad(G1 )-orbits. Otherwise, any element of
g1 which is similar to x is also Ad(G1 )-conjugate to it.
Proof. The proof of the proposition follows in two steps (notation of Proposition 5.2.4).
(1) A computation of the cardinality of x , namely- we will show that |x | = 2 if x is
singular and 1 otherwise.
(2) A description of the image of the map in x .
By Lemma 5.3.2, the minimal polynomial mx of x is of degree N and satisfies mx (t) =
(1)deg(mx ) mx (t). Thus, it can be expressed uniquely as the product of pairwise coprime
factors
k2
Y k3
Y
mx (t) = tk1 i (t)mi i (t)ri , (5.10)
i=1 i=1

where the polynomials 1 , . . . , k2 are irreducible, monic and even, and 1 , . . . , k3 are of the
form i (t) = i (t) i (t) with i (t) monic, irreducible and coprime to (t). The centralizer
C = CMN (k) (x) is isomorphic to the ring k(mx ) and the restriction of the involution ? to C is
transferred via this isomorphism to the map mx defined above. By the Chinese remainder
theorem, we get
Yk2 Yk3
k1 m1
C ' k(t ) k(i (t) ) k(i (t)ri ). (5.11)
i=1 i=1

Furthermore, the restriction of the involution above to each of the factors k(f ), for f
rj
tk1 , m

i , j
i
coincides with the respective involution f , induced from t 7 t. A quick
computation shows that the nilpotent radical of C is isomorphic to the direct product of the
nilpotent radicals of all factors on the right hand side of (5.11), and that the quotient C/N
is isomorphic to the etale algebra
k2
Y k3
Y
r
K=k k(i ) k(i ), (5.12)
i=1 i=1
33
where r = 1 if k1 > 0 (i.e. if x is singular) and equals 0 otherwise. Let denote the the
involution induced on the k-algebra K in (5.12) from the restriction of ? to C. From the
observation regarding the action of ? on C above, we deduce the following properties of the
involution on K.
(D1) The involution preserves the factor kr and acts trivially on it.
(D2) The involution preserves the factors k(i ) and coincides with the non-trivial field
involution i .
(D3) The involution preserves the factors k(i ) ' k(i (t)) k(i (t)) and maps a pair
(, ) k(i (t)) k(i (t)) to the pair (1 (), ()), where : k(i (t)) k(i (t)) is
the isomorphism induced from t 7 t.
Let Sym() be subgroup of K of elements fixed by . Note that, as K ' C/N is a
commutative ring, by Lemma 5.2.5, the set x is can be identified quotient of Sym() by the
image of the map z 7 z z : K Sym().
By (D2) and the theory of finite fields, the restriction of the map z 7 z z to the factors
k(i ) coincides with the field norm onto the subfield of element fixed by . Furthermore,
by (D3), it is evident that an element (, ) k(i (t)) k(i (t)) is fixed by if an only
if = (), in which case (, ) = (, 1) (, 1). Lastly, by (D1) it holds that the image of
the restriction of z 7 z z to the multiplicative group of kr is either trivial, if r = 0, or the
group of squares in k , otherwise. It follows from this that the set x is in bijection with the
quotient (k /(k )2 ), and hence of cardinality 2 if x singular, and is a singleton otherwise.
This completes the first step of the proof.
For the second step, in order to describe the image of , we divide the analysis according
to the parity of N .
(1) N even. In this case we claim that is surjective, and hence the proposition. To do
so, let Q Sym(?; x). Note that, by assumption Q? Q and Q GLN (k), the form
(u, v) 7 B(u, Qv) is alternating and non-degenerate. By Lemma 5.3.1, there exists
w GLN (k) such that Q = w? w. In order to show that Q = (wxw1 ) we only
need to verify that y = wxw1 g1 . This holds, as

y ? = (w? )1 x? w? = (w? )1 (QxQ1 )w? = wxw1 ,

since Q is assumed to commute with x.


(2) N odd. Note that in this case, all elements of g1 are non-singular and hence |x | = 2
for all x g1 , and it suffices to prove that the map is not surjective in this case.
Note that by definition of the equivalence class , if Q1 , Q2 Sym(?; x) are such that
Q1 Q2 , then det(Q1 )1 det(Q2 ) is a square in k . This holds since det(a? ) = det(a)
for all a MN (k). By the same token, it holds that the det(w? w) is a square in k
for all w GLN (k).
Therefore, to show that is not surjective, it suffices to show that Sym(?; x)
contains elements whose determinant is not a square in k. One may take, for example,
the element Q = 1N , for k non-square.


Here it is understood that the ring k0 is the trivial algebra.


34
5.3.3. Centralizers of regular elements. The final objective in this section is to compute the
order of the centralizer of a regular element of g1 . The analysis we propose is analogous to
[12, Proposition 4.4].
Lemma 5.3.7. Let x g1 be regular with minimal polynomial
k2
Y k3
Y
mx (t) = t k1
m
i
i
iri ,
i=1 i=1

where the product on the right hand side is as in (??), with i (t) = i (t)i (t). The deter-
minant map induces a short exact sequence
k2 k3
det
Y Y
1 CG1 (x) U1 (k(tk1 )) U1 (k(m
i ))
i
GL1 (k(iri )) Z 1 (5.13)
i=1 i=1

where Z k is a group of order 2 if N is odd and trivial otherwise.
Proof. As mentioned in the proof of Proposition 5.3.6, the centralizer of x in GLN (k) is
isomorphic to group of units of the ring C, i.e. the direct product
k2
Y k3
Y
k1
CGLN (k) (x) ' GL1 (k(t )) GL1 (k(m
i ))
i
GL1 (k(iri )).
i=1 i=1

Furthermore, the involution ? of GLN (k) restricts to an involution of CGLN (k) (x) which is
 k by t 7
transferred via this isomorphism to the involution mx , induced t, and restricts to
the involution f on each of the factors GL1 (k(f )) for f t , i , j .
1

The additional condition z ? z = 1, and the fact that ? preserves all factors in the decom-
position (5.11), imply that the centralizer of x in G1 is embedded in the group
k2
Y k3
Y
k1
U1 (k(t )) U1 (k(i (t)) GL1 (k(iri )).
i=1 i=1

Similarly to Proposition 5.3.6, the map iri acts on GL1 (k(i (t)ri )) ' GL1 (k(i (t)ri ))
GL1 (k(i (t)ri )) as (, ) 7 (1 (), ()), where : k(i (t)ri ) k(i (t)ri ) is the isomor-
phism induced from t 7 t. It follows from this that (, ) U1 (k(iri )) if and only if
() = 1 , and hence that U1 (k(iri )) ' GL1 (k(iri )).
Lastly, since for any w GLN (k) we have that det(w? ) = det(w), it holds that the
condition w? w = 1 implies that det(w) {1}. Thus, to complete the lemma, we need to
show that both determinants are possible in the case of N odd, and that only 1 is possible for
N even. Both statements are well-known. The former can be proved simply by considering
the elements 1 GLN (k), while the latter can be deduced by considering the Pfaffian of
the matrix w? Jw = J. 
Corollary 5.3.8. Let x g1 be a regular element with minimal polynomial mx k[t]. Let
t(mx ) = (r(mx ), A(mx ), B(mx )) Xn be the type of mx (see Definition 5.1.1). Then
Y A (mx ) B (mx )
|CG1 (x)| = 2 q n 1 + q d d,e 1 q d d,e ,
d,e

where = 1 in the case where N is even and r(mx ) > 0, and = 0 otherwise.
35
Proof. Let mx = tk1 ki=1
Q 2 m1 Qk3 ri
i i=1 i be a decomposition of mx as in (??), with i even
and irreducible, and i (t) = i (t)i (t) with i (t), i (t) irreducible and coprime. Note that
by definition of t(mx ) we have that r(mx ) = b k21 c.
In view of Lemma 5.3.7 it would suffice to show the following three assertions.

(1) U1 (k(tk1 )) = 2q r(mx ) ;
1 1
(2) |U1 (k(m i ))| = q
i m deg i
2 i (1 + q 2 deg i );
(3) |GL1 (k(iri ))| = q ri deg i (1 q deg i )ri .
The last of these assertions is clear, as GL1 (k(i )ri ) is isomorphic to the group of units of the
ring of truncated polynomials of degree at most ri 1 with residue field k(i ), of order q deg i .
Additionally, since the groups GL1 (k(m k1
i )) and GL1 (k(t )) are abelian, the corresponding
i

mi k1
groups U1 (k(i )) and U1 (k(t )) coincide with the kernels of the homomorphisms f 7
m 1 (f ) f and f 7 tk1 (f ) f . Therefore, to deduce the first and second assertions it would
i
be sufficient to compute the image size of these maps. To this end, we note the following.
Claim. Let f k[t] be an irreducible polynomial with f (t) = f (t), and let s N. Let
y GL1 (k(f s )). There exists z GL1 (k(f s )) such that y = f s (z) z if and only if
(1) f s (y) = y; and
(2) there exists z GL1 (k(f s )) such that y f s (z)z (mod f ).
The claim is trivial for s = 1, and follows as in Lemma 5.2.5 for s > 1. In the case of
f = i and s = mi , the claim implies that the image of z 7 m i
i (z)z is simply the group

of invertible truncated polynomials of degree < mi over the subfield of k(i ) fixed by .
Assertion (2) follows easily. For the first assertion, in the case where f (t) = t and s = ki , the
claim implies that the image of z 7 tk (z)z consists of truncated even polynomials of degree
< ki over k, whose constant term is a square in k , and hence of order q1 2
q k1 r(mx )1 .
The claim now follows by multiplying the orders of all factors in the centralizer and invoking
Lemma 5.3.7. 
5.4. Even orthogonal groups. Let N = 2n be even and let B(, ) be a non-degenerate
symmetric bilinear form on V .
As it were, up until now it has been the case that regularity of an element of the classical
Lie algebras discussed so far is equivalent to its being a regular element of the ambient matrix
algebra. The following lemma shows that this is not a general phenomenon.
Lemma 5.4.1. Let N = 2n be even and let x glN (k) be a regular nilpotent element.
Then x is not anti-hermitian with respect to any non-degenerate symmetric bilinear form on
V = kN .
Proof. Note that, as x is conjugate to an N N nilpotent Jordan block, the kernel of x is
one dimensional. Assume towards a contradiction that C(, ) is a symmetric non-degenerate
bilinear form on V such that C(xu, v) + C(u, xv) = 0 for all u, v V . Consider the form
F (u, v) = C(u, xv). By assumption, F is anti-symmetric, and the radical of F coincides with
the kernel of x, by non-degeneracy of C. By properties of antisymmetric forms, it follows
that the kernel of x is even-dimensional. A contradiction. 
Nonetheless, regular nilpotent elements in the case of even orthogonal groups are well
known to exist. In Lemma 5.4.2 below we shall construct such an element and compute its
centralizer.
36
Recall that non-degenerate symmetric bilinear forms on V = kN are classified by the
dimension of a maximal totally isotropic subspace of V with respect to the form (i.e. the
Witt index), and that over a finite field of characteristic not 2 there are exactly two such
forms, upto isometry. It is convenient to modify our notation for this section. We fix B +
and B to be bilinear forms on V of Witt index n and n 1, respectively. In suitable bases,
the forms B + and B are represented by the matrices

0 1 0 1
1 0 1 0
... ...



+

J =
0 1
and J =
0 1 ,
(5.14)

1 0


1 0

0 1 1 0
1 0 0

respectively, where k is a fixed non-square.


Given  {+, }, let G1 = SON (k) and g1 = soN (k) be the group of isometries of
determinant 1 and anti-hermitian operators of the form B  . We will also occasionally use
the colloquial notation G + +
1 = G1 G1 and g1 = g1 g1 . For example, the phrase x is a
regular element g +
1 indicates that x is either a regular element of g1 or of g1 .

5.4.1. Similarity classes of regular elements. Our first objective in this section is to prove
the first assertion of Theorem 5.1.3, i.e. to classify the similarity classes of glN (o) which
intersect g
1 at regular elements. The second goal to be addressed in this section is to give

a criterion for when such a similarity class intersects either g+
1 or g1 .
Note that if x glN (k) is a non-singular element whose minimal polynomial mx is even
and has degree N then, by applying the argument of Lemma 5.3.2.(2) verbatim, we have
that x is anti-hermitian with respect to a non-degenerate symmetric bilinear form and hence
similar to an element of g 1 . By Corollary 5.2.3, all non-singular regular elements of g1 are
obtained in this manner. Furthermore, by inspecting the decomposition of x into primary
rational canonical forms, as the kernel of x g 1 is even-dimensional (by Lemma 5.2.1), to

classify the similarity classes intersecting g1 at singular elements, it is sufficient to consider
the case where x is a nilpotent matrix.

Lemma 5.4.2. Let x glN (k) have minimal polynomial mx (t) = tN 1 . Then x is similar

to a regular nilpotent element of g+
1 , as well as to a regular nilpotent element of g1 .

Proof. By considering the Jordan normal form of such an element x, there exist elements
v0 , u0 V with u0 Ker(x) and such that E = v0 , xv0 , . . . , xN 2 v0 , u0 is a k-basis for V .
Let E 0 = v0 , . . . xN 2 v0 and V 0 = Spank E 0 . By Proposition 5.3.4, the element x |V 0


gl(V 0 ) has minimal polynomial tN 1 and hence is regular. Furthermore, by the proof of
Lemma 5.3.2, there exists a non-degenerate symmetric bilinear form C 0 on V 0 , with respect
to which x |V 0 is anti-hermitian. We wish to extend C 0 to a non-degenerate symmetric
bilinear form on V , with respect to which x is anti-hermitian. This is equivalent to finding
an invertible matrix d MN (k), whose leading N 1 N 1 submatrix coincides with the
37
matrix c of Example 5.3.3, and such that
0 1

.. ..

. .

t
d + d = 0 where = [x]E =
0 1 . (5.15)
0 0
0
A quick computation shows that d can be taken to be the matrix
1

1

. .

d = d =
.
, (5.16)
1

1

with k . Furthermore, by applying a signed permutation basis E, one may verify easily
that d is congruent to the matrix J+ of (5.14) if is a square, and to J otherwise. Thus,

x is similar in this case to both and element of g+ 1 and of g1 .
Lastly, need to verify that x is similar to a regular element of g
1 . To do so, we pass to the
algebraic closure K of k and compute the centralizer in 1 of an element zxz 1 g1 which
is similar to it. Working in the basis E, by direct computation, onesees that the centralizer
of x in MN (K) can be identified with the set of matrices y = uAt vr , where
(1) A MN 1 (K) and commutes with the restriction of to SpanK E 0 ,
(2) u, v KN 1 are elements of the kernel of and t , respectively, and hence of the
t t
form v = v1 0 . . . 0 and u = 0 . . . 0 uN 1 ,
(3) r K is arbitrary.
As in Example 5.3.3, the centralizer of zxz 1 g1 is isomorphic to the group
y CGLN (K) () | yt dy = d .


Computing its Lie-algebra, which consists of matrices y MN (K)() satisfying yd d+dy = 0,


we get the additional three conditions
(1) At c + cA = 0, where c is as in Example 5.3.3,
(2) u + cv = 0, i.e. v1 = uN 1 , and
(3) 2r = 0, and hence r = 0.
It easily follows that C1 (zxz 1 ) is n-dimensional, and hence x is regular. 

In order to streamline the analysis of the nilpotent regular orbits in the sequel, let us fix
some notation which will come in handy shortly.
Notation 5.4.3. Given a matrix A MN 1 (k), column vectors v, u kN 1 and r k, let
(A, v, u, r) denote the N N matrix
 
A v
(A, v, u, r) = .
ut r
38
We also write A for the matrix cAt c, where c is as in Example 5.3.3. Note that, in the
case where d = d is the representing matrix for the symmetric bilinear form given on V ,
we have that
A
 
cu
?
(A, v, u, r) = = (A , cu, 1 cv, r).
1 vt c r

The next step of the computation is to classify when an element x glN (k) which is
similar to a regular element of g +
1 is similar to either g1 or g1 . We first consider two specific
cases, depending on the minimal polynomial of x.
Lemma 5.4.4. Let x glN (k) have minimal polynomial mx . Assume x is similar to a
regular element of g
1.
(1) If mx (t) = f (t)f (t) for some polynomial f k[t] with f (0) 6= 0. Then x is similar

to an element of g+ 1 , and not to an element of g1 .
(2) If mx = r for k[t] an even irreducible polynomial and r N odd, then x is
similar to a regular element of g +
1 and not to an element of g1 .

Proof. Let C be a non-degenerate symmetric bilinear form on V such that C(xu, v) +


C(u, xv) = 0 for all u, v V . We will show that C necessarily has Witt index n in the
first case and n 1 in the second case.
(1) Note that by the assumption mx (0) 6= 0 and Corollary 5.2.3, it follows that x is also
a regular element of glN (k), and hence the space V is cyclic as a k[x] module. Put
W = f (x)V . Then W is isomorphic, as a k[x]-module to V /f (x)V , and hence is of
dimension n = N2 over k. Additionally, for any u, v V we have C(f (x)u, f (x)v) =
C(f (x)f (x)u, v) = 0, and hence W is totally isotropic.
(2) Let us first consider the case where r = 1 and hence V is isomorphic to the field
extension k() of k. Furthermore, the map : V V , induced from t 7 t
is a field involution of V over k, with fixed field W . Invoking the separability of
the extension V /k and the symmetry of C one finds an element c W such that
C(u, v) = C( (u)v, 1) = TrV /k ( (u)v) = 0 for all u, v. By the theory of finite fields,
there exists an element d W such that c = (d)d. It follows that multiplication
by d is an isometry of C with the trace pairing (u, v) 7 TrV /k ( (u)v), and we may
assume without loss of generality that C is the trace pairing.
Note that an element u V is isotropic if and only if (u)u is a traceless element
of W . If follows from this that the number of non-zero isotropic element of V is
(q n + 1)(q n1 1). The fact that C is of Witt index n 1 now follows as in [26,
3.7.2].
0
For the case r > 1, put r0 = b 2r c and U = (x)r +1 V . Then, similarly to (1), U is
0
an isotropic subspace of V , with perpendicular space U = (x)r V . Moreover, the
form C reduces to a non-degenerate symmetric bilinear form on the quotient space
U /U , on which x acts as an anti-hermitian operator with minimal polynomial .
By the case r = 1, we find a two-dimensional anisotropic subspace L U /U , whose

pull-back to U contains a two-dimensional anisotropic subspace of V . Therefore,
the Witt index of C is n 1.

39
Having Lemma 5.4.4 at hand, we need one more basic tool in order to complete the
classification of similarity classes containing regular elements of g
1 . Let us introduce some
notation which streamlines the statement of the following lemma.
Notation 5.4.5. Given a finite, even-dimensional space U over k with a non-degenerate
symmetric bilinear form C, put U = 1 if U is of Witt index 21 dimk U and U = 1 otherwise.
Lemma 5.4.6. Let U, W be finite, even dimensional vector-spaces over k with non-degenerate
symmetric bilinear forms CU and CV respectively. Let U W be endowed with the form
CU W (u + w, u0 + w0 ) = CU (u, u0 ) + CW (w, w0 ) where u, u0 U and w, w0 W . Then
U W = U W .
Proof. Since the direct sum of two isotropic subspaces is again isotropic, the only non-trivial
case to be checked is when U = W = 1. Without loss of generality, we may assume here
that dim U = dim W = 2 and the the forms CU and CW are anisotropic with orthogonal
bases (u1 , u2 ) and (w1 , w2 ) of U and W respectively.
Let f : U W k be the quadratic form associated to CU W , i.e. f (v) = CU W (v, v)
for all v U W . Note that the set {f (u1 ), f (u2 ), f (w1 ), f (w2 )} must contain two elements
of the same coset of k /(k )2 . Without loss of generality we may assume f (u1 ) f (w1 )
(mod (k )2 ). By general properties of finite fields, there exist 1 , 2 , 1 , 2 k such that
12 f (u1 ) + 12 f (w1 ) = f (u2 ) and 22 f (u1 ) + 22 f (w1 ) = f (w2 ).
It follows easily that the set {1 u1 + 1 w1 + u2 , 2 u1 + 2 w1 + w2 } is linearly independent
and consist of isotropic vectors. 
We are now ready to complete the proof of the first assertion of Theorem 5.1.3.
Proposition 5.4.7. Let x glN have minimal polynomial mx . Assume mx (t) = (1)deg mx mx (t)
and let
k2
Y k3
Y
mx (t) = tk1 m
i
i
iri
i=1 i=1
a decomposition as in (??), with i (t) even and irreducible, and i (t) = i (t)i (t) with i
irreducible and coprime to i (t).
(1) If k1 > 0 and deg mx = N 1 then x is similar to a regular element of g 1 . Moreover,
in this case x is similar to an element of g1 as well as to an element of g
+
1.
(2) Otherwise, if k1 = 0 then x is similar to a regular element of g 1 if and only if
deg mx P= N . In this case, we have the following.
(a) If ki=12
mi is even, then x is similar to an element of g+
1 and not to an element

of g1 .
(b) Otherwise, if ki=1 mi is odd, then x is similar to an element of g
P2
1 and not to
an element of g 1 .
Proof. Considering the primary canonical form of x, the space V decomposes as a k[x]-
invariant direct sum V = Wtk1 ki=1
L2 Lk3
Wm i
i=1 Wi i , where the restriction of x to the
r
i
spaces Wf has minimal polynomial f (t), with f (t) = t , mk1
i
i
or iri .
For any f (t) 6= tk1 , the restriction of x to Wf is a regular element of gl(Wf ). By Corol-
lary 5.2.3, the space Wf is endowed with a non-degenerate symmetric bilinear form on which
40
x |Wf acts as an anti-hermitian operator. Furthermore, by Lemma 5.4.4, in the case where
f = iri for i = 1, . . . , k3 or f = m
i
i
with mi even, then Wf = +1. Otherwise, if f mi
i
with
mi odd, Wf = 1. Assertion (2), where k1 = 0, now from Lemma 5.4.6.
In the case where k1 > 0, by the assumption deg mx = N 1 and by Lemma 5.4.2, the

restriction of x to Wtk1 is similar to an element of g+ 1 as well as to an element of g1 . Thus
Wtk1 can be taken to be +1 or 1, and hence, by Lemma 5.4.6, x is similar to an element

of g+
1 as well as to an element of g1 . 
5.4.2. From Similarity classes to adjoint orbits. Our next goal, once the similarity classes
containing regular elements of g 1 have been classified, is to estimate the decomposition of
the intersection of the similarity class of an element of g1 with g1 into orbits under Ad(G1 ),
for  {+, } fixed.
Proposition 5.4.8. Assume |k| > 3. Fix  {+, } and let x g1 . If x is singular, then
the similarity class of x meets g at two distinct Ad(G1 )-orbits. Otherwise, any element of
g1 which is similar to x is also Ad(G1 )-conjugate to it.
Proof. In the notation of Proposition
 5.2.4, let x = (Ad(GL N (k))x g1 ) and x the set of
?
equivalence classes in Sym(?; x) = Q CGLN (k) (x) | Q = Q under . Let : x x
be the map wxw1 7 [w? w] x , for y = wxw1 x .
In the case where x is non-singular, by applying the argument of Proposition 5.3.6 for
non-singular elements verbatim, we have that x consists of a single element and therefore
that x = Ad(G1 )x.
Additionally, by considering the decomposition of x into primary rational canonical forms
and that restriction to the maximal subspace on which the minimal polynomial is of the
form tk1 (which is even dimensional and endowed with a non-degenerate symmetric bilinear
form; see Lemma 5.2.1), we may assume that x is nilpotent.
In this case, by the uniqueness of a regular element in 1 , we may invoke Lemma 5.4.2
and fix a basis E, with respect to which x is represented by the matrix
0 1
.. ..
= . . ,
0 1
0 0
0
and that the non-degenerate symmetric bilinear form is represented in E by the matrix
1

1

. .

d = d = .

1

1

where k is a square if  = + and non-square otherwise.
The centralizer C of in MN (k) is isomorphic to the ring of k[x]-endomorphisms of
k[x] k, and can be realized as the set of matrices (A, v, u, r) (see Notation 5.4.3) with v
and u elements of the kernel of and t respectively, and A MN 1 (k) an upper triangular
Toplitz matrix. Note that the ideal generated by elements of the form (0N 1 , v, u, 0) C
is nilpotent and in particular is included in the nilpotent radical N of C. It follows that the
41
quotient ring C/N is isomorphic to the etale algebra kk. Additionally, by Lemma 5.2.5, we
have that (A, v, u, r) (A0 , v0 , u0 , r0 ) if and only if there exists a block matrix (q, 0, 0, s)
such that ?    0
A v0
  
q A v q
= .
s ut r s u0t r0
Applying a similar argument as in the nilpotent case of Proposition 5.3.6, we have that
the involution ? restricts to the identity map on C/N and hence that the quotient x of
Sym(?; x) by the relation , defined in Section 5.2.2, is isomorphic to the group quotient
k /(k )2 k /(k )2 and is of order 4.
The final step of the proof is to compute the image of the map . Recall that maps
an element wxw1 x = Ad(GLN (k))x g1 to the equivalence class of w? w in x . As
in the odd orthogonal case, two elements which are equivalent with respect to must have
determinant in the same coset of k /(k )2 . In particular, as w? w has square determinant,
the image of in x is included in the subset of equivalence classes in x , containing block
matrices (A, 0, 0, r) with det A r (mod (k )2 ).
To complete the proof that |Im()| = 2 it suffices to find an element w GLN (k)
such that wxw1 g1 and such that w? w is a block matrix of the form (A, 0, 0, r) with
det A, r / (k )2 .
Let k be as above put = (1)(N 2)/2 . Let (k )2 and k r (k )2 be
such that = ; see Lemma 5.4.9 below. Let 0 k be such that 02 = , and put
z = 01 . Let w GLN (k) be represented in E by the matrix


..
.





0
z

w= ,

1

. ..



1
1 0
z 1
where the upper-left scalar block is N22 N22 .
 

Recalling that w? is represented by the matrix d1 wt d, one verifies by direct computa-


tion that w? w is given by the diagonal matrix (1N 1 , 0, 0, 1 ), and consequently, that
wxw1 g1 , that w? w Sym(?; x) and that w? w is not equivalent to 1N under the relation
. 
Lemma 5.4.9. Assume |k| > 3 and char(k) 6= 2. For any element k there exist
, k such that (k )2 , k r (k )2 and such that = .
Proof. Let k be a non-square, and let K = k(t2 ) be the splitting field of t2 , with
0 K a square root of . The norm map NrK|k : K k is surjective and has fibres of
order q + 1. In particular, there exist 0 , 0 k such that
NrK|k ( 0 + 0 0 ) = 02 02 = .
We claim that 0 and 0 can be taken to be both non-zero.
42
(1) Case 1, k r (k )2 . Note that in this case we must have that 0 6= 0, as otherwise
= 02 (k )2 . Furthermore, if 0 = 0 for any pair ( 0 , 0 ) such that 02 02 =
then Nr1 0
K|k () k , and in particular has order smaller than q. A contradiction.
(2) Case 2, (k )2 . Consider the set Nr1
K|k () r k . Note that, as the cardinality
of Nr1
K|k () k is 2 (namely, it consists of the two roots of in k), the order of
Nr1
K|k () r k

is exactly q 1. Assume towards a contradiction that there is no
solution ( , ) k k for the equation
0 0

02 02 = NrK|k ( 0 0 0 ) = .
This implies that any solution not in k {0} is an element of {0} k , or in other
words, that Nr1 0
K|k () r k k . By considering the cardinality of the two sets, we
deduce that this inclusion is in fact an equality. In particular, this implies that for
any 0 k ,
NrK|k ( 0 0 ) = 02 =
and in particular, that the set of squares in k equals the singleton set { 1 }.
This contradicts the assumption |k| > 3.
The lemma follows by taking = 02 and = 02 .


5.4.3. Centralizers of regular elements.


Lemma 5.4.10. Let  {+, }. Let x g1 be regular, with minimal polynomial
k2
Y k3
Y
mx (t) = tk1
m
i
i
iri ,
i=1 i=1

a decomposition as in (??), with i = i (t)i (t) and i (t) irreducible and coprime to i (t).
(1) If k1 > 0, then the determinant map induces a short exact sequence
k2 k3
det
Y Y

1 CG1 (x) A U1 (k(m
i ))
i
GL1 (k(iri )) {1} 1. (5.17)
i=1 i=1

where n o
 t
A = w CGLk1 +1 (k) (x) | w d w = d ,
with d Mk1 +1 (k) is as in Lemma 5.4.2.
(2) Otherwise, the group CG1 (x) is isomorphic to ki=1
Qk3
U1 (k(m ri
Q2
i )) i=1 GL1 (k(i )).
i

Proof. Similarly to Lemma 5.3.7, in order to prove the lemma, it is sufficient to compute
the possible determinants in the middle term of (5.17). For the first assertion it is sufficient
to verify that both +1 and 1 are obtained as determinant  of elements A , for which it is
enough to consider block diagonal matrices of the form 1k01 1 0
A .
For the second assertion, we need to verify that any element w CGLN (k) (x) such that
?
w w = 1 has determinant 1. Since any element of CGLN (k) (x) preserves the invariant factors
of the decomposition of V as a k[x]-module, it is sufficient to consider the following cases of
the minimal polynomial of x.
43
(1) Case 1. Assume mx (t) = (t)m , with k[t] irreducible and even and m N. Let
x = s + h be the additive Jordan decomposition of x, with s, h g1 , s semisimple,
h nilpotent and [s, h] = 0. As mx (0) 6= 0, the space V is cyclic as a k[x] module
and hence CMN (k) (x) ' k[x] = k[s][h] and can be identified with the space of upper
triangular m m Toplitz matrices over the field extension k(i ) which is generated
by s over k. Furthermore, the determinant map of is translated via this identification
to the composition of the determinant onto k(i ) and the field norm Nrk(i )|k , and
the involution ? induces a k-automorphism i of k(i ) with fix field L k(i ). The
condition z ? z, for z CGLN (k) (x), implies that the determinant of z over i is lies
in the group of elements of field norm 1 in the extension k(i ) | L, and therefore is
mapped to 1 in k.
(2) Case 2. Assume mx (t) = (i (t) i (t))r , for i (t) irreducible and coprime to (t).
In this case, by the cyclicity of the k[x] module V , we have that CGLN (k) (x) '
GL1 (k( (t)r )) GL1 (k( (t)r )). Moreover, the map ? restricts to the map (, ) 7
(1 (), ()), where : k( (t)r ) k( (t)r ) is the isomorphism induced from t 7
t. Furthermore, since is a ring-isomorphism which preserves k, we have that
det(()) = det() for all k( (t)r ). In particular, if (, )? = (, ) then = ()1
and hence, det((, )) = det() det()1 = 1.

Corollary 5.4.11. Let x g 1 be regular with minimal polynomial mx (t). Let cx denote the
characteristic polynomial of x, i.e. cx = mx if x is non-singular, and cx = tmx otherwise.
Let t(mx ) = (r(mx ), A(mx ), B(mx )) Xn be the type of mx (see Definition 5.1.1). Then
Y A (mx ) B (mx )
|CG1 (x)| = q n 1 + q d d,e 1 q d d,e .
d,e

Proof. In the case where x is non-singular the assertion follows verbatim as in Corollary 5.3.8.
Otherwise, if x is singular, by decomposing x into its primary rational canonical forms, it is
sufficient to compute the order of centralizer of a nilpotent regular element x.
Without losing any generality, we fix the basis E of Lemma 5.4.2, with respect to which
the ambient symmetric form B  ( {+, }) is represented by the matrix d = d , for some
k , and x is represented by the matrix . Let A = z CGLN (k) () | z t dz = d , as

in Lemma 5.4.10. Let N A be the subgroup consisting of elements of the form
1 2x2 2x

1
X(x) =

. ..

(x k).

1
2x 1
Note that N acts as a one-parameter subgroup of A, and has order |k| = q. Addition-
ally, A is the map under the Cayley map of the ideal generated by elements of the form
(0N 1 , u, v, 0) g1 , and hence is normal.
Let H A be the subgroup of block diagonal matrices (A, 0, 0, r). Note that, in the
notation of Notation 5.4.3, A A = 1N 1 and r = det(A)1 , which implies that A commutes
with the restriction of to the subspace spanned by the first N 1 elements of E and hence,
Proposition 5.3.6 and Corollary 5.3.8 that |H| = U1 (k(tk1 1 )) = 2q r(cx )1 .
44
Given an arbitrary element (A, v, u, r) A , it holds that A must be invertible, and
that v = du for some k. In particular, v = 0 iff u = 0. It follows from this, and by
direct computation, that   
v1 A v
X H.
a1,1 ut r
In particular, we have that A = H N and hence, since H N = {1}, that
|A/N | = |H/H N | = |H| = 2q r(cx )1 .
k1 1
To conclude, we have that |A| = 2q r(cx ) = 2q 2 , and the result follows from Lemma 5.4.10.


45
Appendix A. The number of even polynomials of a given degree over Fq
Assume p 6= 2 is prime and q = p , for N. As above, we put k = Fq . We wish to
enumerate the number of even irreducible polynomials over k of a given degree 2m.
We call an element x K even over k if its minimal polynomial over k is even. The set
of even irreducible polynomials of degree 2m is naturally in bijection with the set of Galois
orbits of non-zero even elements x K such that k(x)/k is an extension of degree 2m, and
any such orbit ha cardinality 2m. In view of this, in the sequel we will enumerate the number
of such elements x K.
We begin with a criterion for an element of x to be even.
Lemma A.0.1. Let 0 6= x K have minimal polynomial f (t) over k. Then f is even if and
m
only if there exists m N such that f (t) divides tq + t.
Proof. Assume f is even. Then f (x) = f (x) = 0 and hence x and x are Galois
conjugate over k. In particular, by the theory of finite fields, this implies that x =
m m
xq for some m N. Thus x is a root of tq + t, and hence, since f is its minimal
m
polynomial, f (t) | tq + t.
m m
Let Gal(K | k) be the map (y) = y q . Then, by the assumption f (t) | tq + t,
we have that
m
(x) = xq = x
Define a polynomial g(t) = 21 (f (t) f (t)) . Then g(t) is a monic odd polynomial
(i.e. g(t) = g(t)) of degree smaller or equal to deg(f ). Additionally
1 1
g(x) = (f (x) f (x)) = (f (x) (f (x))) = 0,
2 2
as the coefficients of f are fixed under . This implies that either g = 0 or g = f . But
g = f is impossible, since the condition g(t) = g(t) implies that g(0) = f (0) = 0,
and in particular f is not irreducible. Thus g = 0 and f (t) = f (t).

m
Thus, Lemma A.0.1 asserts that a non-zero element x K is even if and only if xq +x = 0
for some m N.
We now wish to classify those even elements x K which generate a degree 2d extension
of k. We first note the following.
m
Lemma A.0.2. Let m N, and let 0 6= x K be a root of tq + t. Then |k(x) : k| = 2d for
some d N such that d | m and md is an odd integer. Furthermore, in this case we have that
d
xq + x = 0.
m
Proof. By Lemma A.0.1, the assumption that xq + x = 0 implies that x is even and hence
has an even minimal polynomial, say of degree 2d. Hence |k(x) : k| = 2d. Also, note that
2m m q m
xq = xq =x
2m
and so k(x) is fixed under the map y 7 y q , whence a subfield of Fq2m . This gives us that
d | m. Additionally, since f (x) = f (x) = 0, there exists an element Gal(k(x) | k)
such that (x) = x. In particular, 2 (x) = x and hence is an involution of k(x). As
46
the Galois group Gal(k(x) | k) is cyclic of order 2d and generated by the Frobenius map
d
(y) = y q (y k(x)), it follows that (y) = d (y) = y q for all y k(x) and hence
d
xq = (x) = x.
m
Lastly, we show that r = d
is odd. This follows since
m
x = xq = m (x) = (d )r (x) = (1)r x,
and hence r is odd. 
Note that the converse of Lemma A.0.2 is true as well. Namely, if d | m and r = md is an
d
odd integer then any non-zero x K which satisfies xq + x = d (x) + x = 0 also satisfies
m
xq + x = (d )r (x) + x = (1)r x + x = 0.
Thus, we obtain the following.
Corollary A.0.3. For any m N let Sm denote the set of non-zero roots of the polynomial
m
tq + t. Then
(1) Sd Sm if and only if d | m and md is an odd integer.
(2) The set non-zero even elements x K which generate an extension of k of degree 2m
is [
Sm r Sd .
m
d|m1, d
is odd
m
As the set Sm has cardinality q m 1 (since the field k is perfect and the roots of tq + t
are all simple), by exclusion-inclusion we deduce that the non-zero even elements of K which
generate a degree 2m extension of k is
X m
qd 1 ,


m
d
d|m , 2- d

where is the mobius function.

References
[1] N. Avni, B. Klopsch, U. Onn, and C. Voll, Arithmetic groups, base change and representation
growth, Preprint at ArXiv: 1110.6092, (2014).
[2] , Similarity classes of integral p-adic matrices and representation zeta functions of groups of type
A2 , Preprint at ArXiv: 1410.4533, (2014).
[3] A. Borel, Linear Algebraic Groups, Graduate Texts in Mathematics, Springer New York, 1991.
[4] C. Bushnell and A. Fro hlich, Gauss sums and p-adic division algebras, Lecture notes in mathe-
matics, Springer, 1983.
[5] J. Dieudonne and A. Grothendieck, El ements de geometrie algebrique, Inst. Hautes Etudes
Sci.
Publ. Math., 4, 8, 11, 17, 20, 24, 28, 32 (19611967).
[6] D. S. Dummit and R. M. Foote, Abstract Algebra - 3rd Edition, John Wiley and Sons, Inc., 3 ed.,
2004.
[7] M. J. Greenberg, Schemata over local rings, Annals of Mathematics, 73 (1961), pp. 624648.
[8] , Schemata over local rings: II, Annals of Mathematics, 78 (1963), pp. 256266.
[9] G. Hill, Regular elements and regular characters of GLn (O), Journal of Algebra, 174 (1995), pp. 610
635.
[10] I. M. Isaacs, Charachter theory of Finite Groups, Academic Press Inc., 1976.
47
[11] A. Jaikin-Zapirain, Zeta function of representations of compact p-adic analytic groups, J. Amer.
Math. Soc., 19 (2006), pp. 91118.
[12] R. Krakovski, U. Onn, and P. Singla, On regular representations of GLn (o) and GUn (o), (2016).
[13] J. Lahtonen, Number of even irreducible monic polynomials of a given degree over a finite field.
Mathematics Stack Exchange http://math.stackexchange.com/q/1833813.
[14] J. Serre, Local Fields, Graduate Texts in Mathematics, Springer New York, 1995.
[15] S. Shechter, Characters of the norm-one units of local division algebras of prime degree, Journal of
Algebra, (2016), pp. .
[16] T. Shintani, On certain square integrable irreducible unitary representations of some -adic linear
groups, Proceedings of the Japan Academy, Series A, Mathematical Sciences, 44 (1968), pp. 13.
[17] T. Springer and R. Steinberg, Conjugacy classes, in Seminar on Algebraic Groups and Related
Finite Groups, vol. 131 of Lecture Notes in Mathematics, Springer Berlin Heidelberg, 1970, pp. 167266.
[18] A. Stasinski, Reductive group schemes, the greenberg functor, and associated algebraic groups, Journal
of Pure and Applied Algebra, 216 (2012), pp. 1092 1101.
[19] , Representations of gl_n over finite local principal ideal rings - an overview., in Around Langlands
Correspondences : 17-20 June 2015, Paris, France ; proceedings., Contemporary mathematics, American
Mathematical Society, January 2017.
[20] A. Stasinski and S. Stevens, The regular representations of GLN over finite local principal ideal
rings, 2016.
[21] R. Steinberg, Conjugacy classes in algebraic groups, Lecture notes in mathematics, Springer, 1974.
[22] K. Takase, Regular irreducible characters of a hyperspecial compact group and weil representations
over finite fields, 2015.
[23] K. Takase, Regular characters of GLn (O) and Weil representations over finite fields, J. Algebra, 449
(2016), pp. 184213.
[24] G. E. Wall, On the conjugacy classes in the unitary, symplectic and orthogonal groups, Journal of the
Australian Mathematical Society, 3 (1963), pp. 162.
[25] W. C. Waterhouse, Introduction to Affine Group Schemes, Graduate Texts in Mathematic, Springer,
New York, Heidelberg, Berlin, 1979.
[26] R. Wilson, The Finite Simple Groups, Graduate Texts in Mathematics, Springer London, 2009.

Department of Mathematics, Ben Gurion University of the Negev, Beer-Sheva 84105,


Israel
E-mail address: shais@post.bgu.ac.il

48

You might also like