Representations of locally compact groups

Fall 2013

Fiona Murnaghan
 Contents

Chapter 1. Topological groups 5

1. Definitions and basic properties 5
2. Locally compact, totally disconnected groups 10
3. Matrix groups 14
4. Haar measure on locally compact Hausdorff groups 16
5. Coset spaces and quasi-invariant measures 21
Chapter 2. Continuous representations of locally compact groups 29
1. Definition of continuous representation 29
2. Representations of compact groups are unitary 32
3. Irreducibility, subrepresentations, etc. 32
4. The regular representation on L2 (G/H) 34
5. Intertwining operators and Schurs Lemma 35
Chapter 3. Representations of compact groups: the general theory 39
1. The Peter-Weyl Theorem 39
2. Faithful representations 43
3. Finite-dimensional representations of locally profinite groups 43
4. Finite-dimensional subrepresentations of the regular representation 44
5. Orthogonality relations for matrix coefficients 47
6. Decomposition of the regular representation 48
7. Decomposition of unitary representations of compact groups 49
8. Characters of finite-dimensional representations 50
9. Convolution, characters and orthogonal projections 53
10. Induced representations of compact groups 56

Chapter 4. Representations of compact Lie groups 59

1. Maximal tori and Weyls Integral Formula 60
2. Weights of finite-dimensional representations 62
3. Relations between representations of compact Lie groups and complex Lie
algebras 63
4. Highest weight theorem 65
5. The Weyl character formula 68
Chapter 5. Representations of reductive groups 69
3
4 CONTENTS

1. The Iwasawa decomposition 69

2. Principal series representations of SL2 (F ) 71
3. Parabolic subgroups of Sp4 (F ) 73
4. Parabolic subgroups and induction 77
5. Discrete series representationsgeneral properties 78
6. Maximal Tori 80
7. Discrete series representations of reductive Lie groups 82
8. Discrete series representations of reductive p-adic groups 85
9. Maximal tori and representations of finite groups of Lie type 88
10. Harish-Chandras Philosophy of Cusp Forms 89
Bibliography 93
 CHAPTER 1

Topological groups

1. Definitions and basic properties

Definition 1.1. A topological group G is a group which is a topological space and has
the property that the map (x, y) 7 xy 1 from G G to G is continuous. (Equivalently, the
maps (x, y) 7 xy and x 7 x1 are continuous.) Here, G G is given the product topology.

Lemma 1.2. Let G be a topological group. Then

(1) The map g 7 g 1 is a homeomorphism of G onto itself.
(2) Fix g0 G. The maps g 7 g0 g, g 7 gg0 , and g 7 g0 gg01 are homeomorphisms of
G onto itself.

Example 1.3. Let F be a field that is also a topological group under addition and
the nonzero elements F in F is a toplogical group under multiplication. (Some standard
examples are R, C and finite extensions of the field Qp of p-adic numbers). The set Mn (F )
2
of n n matrices with entries in F is homoemorphic to F n . Let GLn (F ) (also denoted
by GL(n, F )) be the group of invertible matrices in Mn (F ). Since GL(n, F ) is a subset of
Mn (F ), we can make GLn (F ) into a topological space using the the subspace topology. The
determinant map det : Mn (F ) F is continuous (its a polynomial in the matrix entries). It
is easy to see that the multiplication map (A, B) 7 AB from Mn (F ) Mn (F ) Mn (F ) is
continuous, so the restriction to GLn (F )GLn (F ) is also continuous. Given A GLn (F ), let
adj A be the adjoint of A. Using the fact that A1 = adj A/ det A, we can see that inversion
is continuous. Hence GLn (F ) is a topological group under matrix multiplication. If F is
Hausdorff and locally compact, then GLn (F ) is Hausdorff and locally compact. (Recall that
a topological space X is locally compact if given a point x X there exists a compact subset
U of X such that U contains an open neighbourhood of x.)
More generally, if R is a commutative ring with identity, R is a topological space, and
multiplication and addition are continuous maps from R R to R, let GLn (R) = { A
5
6 1. TOPOLOGICAL GROUPS

Mn (R) | det A R }. Here, R is the group of units in R. Then GLn (R) is a topological
group under matrix multiplication.

Example 1.4. Let B(H) be the set of bounded linear operators on a Hilbert space H.
(Recall that a linear operator T on H is bounded if there exists a positive constant C such
that kT (v)k Ckvk for all v H.) If T B(H) is bijective, then T 1 B(H). The strong
operator topology on B(H) is the weakest locally convex topology such that the evaluation
map sending T to kT (v)k is continuous for each vector v H. If H is infinite-dimensional, the
group of bijective operators in B(H) is not a topological group with respect to the subspace
topology induced by the strong operator topology on B(H). Neither the multiplication map
nor the inversion map is continuous.
If we take the subgroup U (H) of unitary operators on H, this is a topological group with
respect to the subspace topology. To see this, fix T1 , T2 U (H). Fix  > 0 and v H.
Then the set
V := { S U (H) | k(S T1 T2 (v)k <  }
is an open neighbourhood of T1 T2 in the strong operator topology. Let

V 0 = { (S1 , S2 ) U (H) | k(S1 T1 )(T2 (v))k < /2, and k(S2 T2 )(v)k < /2 }.

Then V 0 is an open neighbourhood of (T1 , T2 ) in U (H) U (H). Let (S1 , S2 ) V 0 . Then

k(S1 S2 T1 T2 )(v)k = kS1 (S2 T2 )(v) + (S1 T1 )(T2 (v))k
kS1 (S2 T2 )(v)k + k(S1 T1 )(T2 (v))k
= k(S2 T2 )(v)k + k(S1 T1 )(T2 (v))k < /2 + /2
To see that inversion is continuous, fix T U (H). Then, given v H and  > 0, let
V = { S U (H) | k(S T 1 )(v)k <  }. Set w = T 1 (v). Show that the image of the open
neighbourhood { S 0 U (H) | k(S 0 T )(w)k <  } of T under the inversion map is a subset
of V.

A subgroup H of a topological group G is a topological group in the subspace topology.

Example 1.5. Let (I, ) be a directed partially ordered set. Suppose that {Gi }iI is
a family of finite groups. Assume that whenever i j I, there exists a homomorphism
fij : Gj Gi such that
(1) fii is the identity map.
(2) fik = fij fjk for i j k I.
 1. DEFINITIONS AND BASIC PROPERTIES 7

Then ((Gi )iI , (fij )ijI ) is called an inverse (or projective) system of finite groups and ho-
Q
momorphisms. Let lim Gi be the subset of the direct product iI Gi consisting of elements

(gi )iI such that gi = fij (gj ) for all i j I. It is easy to see that lim Gi is a subgroup of
Q
G i . We give each G i the discrete topology and then we take the product topology on
QiI Q
iI Gi . Then lim
i
G is closed in iI Gi . Furthermore, lim Gi is compact and totally dis-

connected. The details are left as an exercise. Here is a simple concrete example: fix a prime
p, take I to be the natural numbers, Gi = Z/pi Z, and, for i j, set fij (n + pj Z) = n + pi Z,
n Z. Then Zp := lim Z/pi Z is known as the group of p-adic integers. (In fact, Zp is a

ring.)

Lemma 1.6. Any open subgroup of a topological group is closed.

Proof. Let H be an open subgroup of a topological subgroup G. If g G, then gH is

S
open. Hence S := gG\H gH is open. Since H = G \ S, it follows that H is closed. 

Lemma 1.7. If H is a subgroup of a topological group G, the closure H is a subgroup of

G. If H is normal in G, then H is normal in G.

Proof. Let x, y H. Let U be an open neighbourhood of xy. Let SU = { (u, v)

GG | uv U }. Then SU is open and contains (x, y). Let U1 and U2 be open neighbourhoods
of x and y, respectively, such that U1 U2 SU . Because x, y H, Uj H is nonempty,
j = 1, 2. Fix u U1 H and v U2 H. Note that uv U H. We have shown that
U H is nonempty for any open neighbourhood U of xy. Hence xy H.
Let V be an open neighbourhood of y 1 . Let V 0 = { v 1 | v V }. Since V 0 is open
and contains y and y H, the intersection V 0 H is nonempty. The inverse of any point in
V 0 H belongs to V H. Hence V H is nonempty. As before, this implies that y 1 H.
The assertion about normality is left as an exercise. 

Let H be a subgroup of a topological group G, and let q : G G/H be the canonical

mapping of G onto G/H. We define a topology UG/H on G/H, called the quotient topology,
by UG/H = { q(U ) | U UG }. (Here, UG is the topology on G). The canonical map q is
open (by definition) and continuous.

Proposition 1.8. Let H be a subgroup of a topological group G.

(1) If H is compact, then q is a closed map.
(2) G/H is a Hausdorff space if and only if H is closed.
8 1. TOPOLOGICAL GROUPS

(3) If G is locally compact, then G/H is locally compact. If, in addition, H is closed,
then H is locally compact.
(4) If G is Hausdorff and H is a locally compact subgroup of G, then H is closed.
(5) If H is normal in G, then G/H is a topological group.
(6) H is open in G if and only if G/H is a discrete space. If G is compact, then H is
open in G if and only if G/H is a finite discrete space.

Proof. (1): It suffices to show that if S is a closed subset in G, then the set SH is
closed in G. In fact, if S G is closed and T G is compact, then ST is closed. The proof
is left as an exercise.
(2): Suppose that G/H is Hausdorff. Then single points are closed in G/H. In particular
{H} is closed in G/H. By definition of the quotient topology, q 1 ({H}) is closed in G. Next,
suppose that H is closed in G. If X is a topological space, then X is Hausdorff if and only
if X := { (x, x) | x X } is closed in X X. Define f : G/H G/H (G G)/(H H)
by f (g1 H, g2 H) = (g1 , g2 )(H H). Then f is a homeomorphism. It suffices to show that
f (G/H ) is closed. Equivalently, is suffices to show that the set { (g1 , g2 ) G G | g1 g21
H } is closed. This set is the inverse image of H under the continuous map : G G G
defined by (g1 , g2 ) = g1 g21 . Since H is closed, this set is also closed.
(3): Since q is continuous, the image under q of a compact subset of G in G/H is compact.
Since q is open, this implies that if G is locally compact, then G/H is also locally compact.
For the second statement, use the fact that a closed subset of a locally compact topological
space is locally compact (in the subspace topology).
(4): Suppose that H is a locally compact subgroup of G. Let U H be a compact
set such that U contains an open neighbourhood of 1 in H. Because compact subsets of
Hausdorff spaces are closed, U is closed. Fix a closed set V in G such that U = V H and V
contains an open neighbourhood of 1. Since U is compact in H, U is compact in G. Because
G is Hausdorff, this implies that U is closed in G. Fix an open neighbourhood W of 1 in G
such that W W := { ww0 | w, w0 W } is a subset of V .
Let x H. We know that W x1 H is nonempty. Let y W x1 H. Let W 0 be an
open neighbourhood of yx in G. Because y 1 W 0 and xW are both open neighbourhoods of
x in G and x H, the intersection y 1 W 0 xW H is nonempty. Let z y 1 W 0 xW H.
Note that yz W W V , yz H and yz W 0 . Hence yz W 0 (V H) = W 0 U .
 1. DEFINITIONS AND BASIC PROPERTIES 9

This shows that W 0 U is nonempty for every open neighbourhood W 0 of yx in G. Thus

yx belongs to the closure U . As observed above, U is closed. Hence yx U H. Now we
have that x = y 1 yx H.
(5): If g, x G, define Tg (x) = gx. Similarly, we define Tq(g) : G/H G/H, g G.
Observe that q Tg = Tq(g) p and q(x)1 = q(x)1 . Using the fact that q is an open map
and Tg and the map x 7 x1 are continuous maps, we can show that Tq(g) and the map
gH 7 g 1 H are continuous.
(6): Exercise. 

Remark 1.9. In cases where certain G-invariant measures (known as Haar measures)
exist, integration plays a key role in various aspects of representation theory. In order for
such measures to exist, G must be locally compact. This is discussed in Section 4.

Exercise 1.10. Let G be a Hausdorff topological group.

a) Show that the centralizer of a nonempty subset of G is closed.
b) Show that the normalizer of a subgroup of G is closed.

Definition 1.11. The identity component G of G is defined to be the connected com-

ponent of G containing the identity element.

Lemma 1.12. Let G be a topological group.

(1) G is a closed normal subgroup of G.
(2) The quotient group G/G0 is discrete if and only if G is open in G.

Exercise 1.13. Let G be a connected topological group. Show that a proper subgroup
of G cannot contain an open neighbourhood of the identity.

Definition 1.14. If G and G0 are topological groups a continuous homomorphism f :

G G0 is called a homomorphism of topological groups. If f is also a homeomorphism, then
f is an isomorphism of topological groups.

Lemma 1.15. Let f : G G0 be a homomorphism of topological groups. Let H be the

kernel of f .
(1) H is a closed normal subgroup of G.
(2) If G is locally compact, then H and G/H are locally compact.
(3) G/H and f (G) are isomorphic (as topological groups) if and only if f is open when
viewed as map from G to f (G).
10 1. TOPOLOGICAL GROUPS

2. Locally compact, totally disconnected groups

Recall that a topological space X is totally disconnected if given any two distinct elements
x and y in X, there exist open sets U and V with x U , y V , U V = , and X = U V .
Equivalently, there is no connected subset of X with more than one element.

Lemma 2.1. A totally disconnected group is Hausdorff.

The proof is left as an exercise. In fact, if a topological group is T1 (that is, points are
closed sets), then the group is Hausdorff.

Definition 2.2. A topological group G is profinite if G is compact and totally discon-

nected.

Lemma 2.3. If G is profinite, the quotient maps combine into a topological isomorphism
between G and the projective(aka inverse) limit lim G/N , where N runs over the set of open

normal subgroups of G.

Lemma 2.4. . (Chapter I, 1 of [S]) Let G be a topological group. The following are
equivalent:
(1) G is a projective limit of finite groups.
(2) G is a compact, Hausdorff group in which the family of open normal subgroups forms
a fundamental system of open neighbourhoods of the identity.
(3) G is a compact, totally disconnected group.

Example 2.5. One example that has already been mentioned is the set Zp of p-adic
integers (where p is a fixed prime). Another example: Suppose that L is a Galois extension
of a field K of infinite degree. Let F range over all extension fields of K such that F/K is a
finite Galois extension. We can define a partial order on the set of such F using inclusion.
If F2 F1 , we take the homomorphism Gal(F1 /K) Gal(F2 /K) given by restriction to
F2 . Then Gal(L/K) ' lim Gal(F/K). The topology on Gal(L/K) is often referred to as the

Krull topology.

Definition 2.6. If G is a locally compact, totally disconnected group, we say that G is

locally profinite (alternatively G is a group of t.d. type, or a t.d. group).

If G is a locally profinite group, then every open neighbourhood of the identity in G

contains a compact open subgroup of G. A compact open subgroup of a locally compact,
totally disconnected, group is a profinite group.
 2. LOCALLY COMPACT, TOTALLY DISCONNECTED GROUPS 11

Lemma 2.7. A Hausdorff topological group is locally profinite if G has a countable neigh-
bourhood basis at the identity consisting of compact open subgroups, and G/K is a countable
set for every open subgroup K of G.

Some locally profinite groups occur as matrix groups over p-adic fields. Let p be a prime.
Let x Q . Then there exist unique integers m, n and r such that m and n are nonzero and
relatively prime, p does not divide m or n, and x = pr m/n. Set |x|p = pr . This defines a
function on Q , which we extend to a function from Q to the set of nonnegative real numbers
by setting |0|p = 0.

Definition 2.8. The function | |p is called the p-adic absolute value on Q.

The p-adic absolute value is a valuation on Q - that is, it has the properties
(1) |x|p = 0 if and only if x = 0
(2) |xy|p = |x|p |y|p
(3) |x + y|p |x|p + |y|p .
The usual absolute value on the real numbers is another example of a valuation on Q. The
p-adic abolute value satisfies the ultrametric inequality, that is, |x + y|p max{|x|p , |y|p }.
Note that the ultrametric inequality implies property (iii) above. A valuation that satisfies
the ultrametric inequality is called a nonarchimedean valuation.
Note that the set { |x|p | x Q } is a discrete subgroup of R . Hence we say that
| |p is a discrete valuation. The usual absolute value on Q is an example of an archimedean
valuation. Clearly it is not a discrete valuation. Two valuations on a field F are said to be
equivalent is one is a positive power of the other.

Theorem 2.9. (Ostrowski; see Theorem 2.1 of [Cas]) A nontrivial valuation on Q is

equivalent to the usual absolute value or to | |p for some prime p.

Definition 2.10. If F is a field and | | is a valuation on F , the topology on F induced

by | | has as a basis the sets of the form U (x, ) = { y F | |x y| <  }, as x varies over F ,
and  varies over all positive real numbers. A field F 0 with valuation | |0 is a completion of
the field F with valuation | | if F F 0 , |x|0 = |x| for all x F , F 0 is complete with respect
to | |0 (every Cauchy sequence with respect to | |0 has a limit in F 0 ) and F 0 is the closure
of F with respect to | |.

The completion of F is the smallest field containing F such that F 0 is complete with
respect to | |0 .
12 1. TOPOLOGICAL GROUPS

The real numbers is the completion of Q with respect to the usual absolute value on Q.

Definition 2.11. The p-adic numbers Qp is the completion of Q with respect to | |p .

(We denote the extension of | |p to Qp by | |p also).

In Example 1.5, we defined the p-adic integers Zp as lim Z/pi Z. Strictly speaking, before

making the next definition, we should verify that lim Z/pi Z is isomorphic to the group

R = { x Qp | |x|p 1 }. This can be done as follows. It is possible to show that a
proper open subgroup of R has the form pi R for some positive integer i and that R/pi R
is isomorphic to Z/pi Z. Then, since R is profinite (its compact and totally disconnected),
R ' lim R/pi R ' lim Z/pi Z.

Definition 2.12. The p-adic integers Zp is the set { x Qp | |x|p 1 }.

Note that Zp is a subring of Qp (this follows from the ultrametric inequality and the
mulitiplicative property of | |p ), and Zp contains Z. The set pZp (the ideal of Qp generated
by the element p) is a maximal ideal of Zp and Zp /pZp is therefore a field.
Let x Q . Write x = pr m/n with r Z and m and n nonzero integers such that m and
n are relatively prime and not divisible by p. Because m and n are relatively prime and not
divisible by p, the equation nX m(mod p) has a unique solution ar { 1, . . . , p 1 }. That
is, there is a unique integer ar {1, . . . , p 1 } such that p divides m nar . Since |n|p = 1,
p divides m nar is equivalent to |(m/n) ar |p < 1, and also to |x ar pr |p < |x|p = pr .
Expressing x ar pr in the form ps m0 /n0 with s > r and m0 and n0 relatively prime integers,
we repeat the above argument to produce an integer as {1, . . . , p 1} such that

|x ar pr as ps |p < |x ar pr |p = ps .

If s > r + 1, set ar+1 = ar+2 = = as1 = 0, to get

s
X
|x an pn |p < ps .
n=r

Continuing in this manner, we see that there exists a sequence { an | n r } such that
an { 0, 1, . . . , p 1 } and, given any integer M r,
M
X
|x an pn |p < pM .
n=r
P
It follows that n=r an pn converges in the p-adic topology to the rational number x.
 2. LOCALLY COMPACT, TOTALLY DISCONNECTED GROUPS 13

On the other hand, it is quite easy to show that if an { 0, 1, . . . , p 1} and r is an

integer, then n
P
n=r an p converges to an element of Qp (though not necessarily to a rational
number).

Lemma 2.13. A nonzero element x of Qp is uniquely of the form n

P
n=r an p , with an
{ 0, 1, 2, . . . , p 1}, for some integer r with ar 6= 0. Furthermore, |x|p = pr . (Hence x Zp
if and only if r 0).

Lemma 2.14. Zp /pZp ' Z/pZ.

Proof. Let a Zp be nonzero. According to Lemma 2.13, a = n

P
n=r an p for some
sequence { an | n r }, where |a|p = pr 1 implies that r 0. If r > 0, then a pZp .
For convenience, set a0 = 0 when |a|p < 1. If r = 0, then a0 { 1, . . . , p 1 }. If a = 0, set
a0 = 0. Define a map from Zp to Z/pZ by a 7 a0 . This is a surjective ring homomorphism
whose kernel is equal to pZp . 

Definition 2.15. A local field F is a (nondiscrete) field F which is locally compact and
complete with respect to a nontrivial valuation.

The fields R, C, and Qp , p prime, are local fields. If | | is a nontrivial nonarchimedean

valuation on a field F , then pF := { x F | |x| < 1} is a maximal ideal in the ring
oF := { x F | |x| 1 }, so the quotient oF /pF is a field, called the residue class field of F .
The following lemma can be used to check that Qp is a local field.

Lemma 2.16. ([Cas], Corollary on p. 46) Let | | be a nonarchimedean valuation on a

field F . Then F is locally compact with respect to | | if and only if
(1) F is complete (with respect to | |)
(2) | | is discrete
(3) The residue class field of F is finite.

For every integer N , pN Zp is a compact open (and closed) subgroup of Qp . It is not hard
to see that { pN Zp | N 0 } forms a countable neighbourhood basis at the identity element
0. The group Z p of units in the ring Zp is equal to { a Qp | |a|p = 1 }. It follows from
Lemma 2.13 that Q
p ' hpi Zp . Hence Qp /Zp is discrete. It can be shown that any open
subgroup of Qp is of the form pN Zp for some integer N . Thus Qp /K is discrete for every
open subgroup K of Qp . Therefore Qp is a locally profinite group. For more information on
valuations, the p-adic numbers, and p-adic fields, see [Cas].
14 1. TOPOLOGICAL GROUPS

As mentioned in Example 1.3, because Qp is locally compact, the topological group

GLn (Qp ) is also locally compact. In fact GLn (Qp ) is a locally profinite group. If j is a
positive integer, let Kj be the set of g GLn (Qp ) such that every entry of g 1 belongs
to pj Zp . Then Kj is a compact open subgroup, and { Kj | j 1 } forms a countable
neighbourhood basis at the identity element 1.
Let K be an open subgroup of GLn (Qp ). Then Kj K for some j 1. Hence to
prove that G/K is countable, it suffices to prove that G/Kj is countable for every j. For
a discussion of the proof that G/Kj is countable, see [M]. At this point, we can apply
Lemma 2.7 to conclude that GLn (Qp ) is a locally profinite group.
Closed subgroups of locally profinite groups are locally profinite. Hence any closed sub-
group of GLn (Qp ) is locally profinite. Such groups are often called p-adic matrix groups.
This gives a way to generate many examples of locally profinite groups.

3. Matrix groups

3.1. Lie groups. A real Lie group is a topological group G that is a finite-dimensional
real smooth manifold with a group structure in which the multiplication and inversion maps
from G G to G and from G to G are smooth maps. Without referring to the differentiable
manifolds, we may define a matrix Lie group, or a closed Lie subgroup of GLn (R) or GLn (C)
to be a closed subgroup of the topological group GLn (R) or GLn (C), respectively. (This
latter definition is reasonable because GLn (R) and GLn (C) are real Lie groups, and it can
be shown that a closed subgroup of a (real) Lie group is also a (real) Lie group). A complex
Lie group is a complex-analytic manifold G that is also a group in which the multiplication
and inversion maps are holomorphic functions.
A connected matrix Lie group is reductive if it is stable under conjugate transpose, and
semisimple if it is reductive and has finite centre. The book of Hall [Ha] gives an introduction
to matrix Lie groups, their structure, and their finite-dimensional representations. Some
other introductory references on Lie groups and their representations are [BD], [FH], [K1]
and [K2].

Example 3.1. Consider the hermitian form z1 z1 + z2 z2 + + zn zn on Cn . The n n

complex unitary group Un (C) is the group of complex n n matrices that preserve this
hermitian form. It is a closed subgroup of GLn (C), so it is a locally compact group. A
matrix g Un (C) satisfies t gg = In (the columns of the matrix form an orthonormal basis of
Cn with respect to the standard inner form). The group Un (C) (sometimes written Un ) is a
 3. MATRIX GROUPS 15

real Lie group. The group SUn (C) := Un (C) SLn (C) is the n n complex special unitary
group.

Exercise 3.2. Verify that Un (C) is connected and compact.

Example 3.3. Consider the quadratic form x21 + x22 + x2n on Rn . The n n real
orthogonal group On (R) is the group of matrices in Mn (R) that preserve the form. It is a
compact, disconnected Lie group. The real special orthogonal group SOn (R) = SLn (R)
On (R) is the connected component of the identity in On (R). The n n complex orthogonal
group On (C) is the matrices in Mn (C) that preserve the quadratic form z12 + z22 + + zn2
on Cn .

Example 3.4. If F is a field, the 2n 2n symplectic group over F is defined by:

!
0 In
Sp2n (F ) = {g GL2n (F ) | t gJg = J }, where J = .
In 0
If F = R or C, then Sp2n (F ) is a noncompact connected Lie group. There is another
real Lie group, the so-called compact symplectic group Sp(n), which may be defined as
U2n (C) Sp2n (C).

3.2. Linear algebraic groups. A linear algebraic group is an affine algebraic variety
that is also a group, where the multiplication and inversion operations are given by regular
functions on the variety. A linear algebraic group may be realized as a subgroup of some
general linear group.
Many Lie groups are algebraic groups over the real or complex numbers. For example, a
compact Lie group can be regarded as the group of points of a real linear algebraic group.
Other examples of linear algebraic groups include various p-adic groups, that is, groups
occuring as closed subgroups of GLn (F ), where F is a p-adic field (a finite extension of Qp ,
p prime). If F is a finite field, the groups GLn (F ), SLn (F ), Spn (F ) are linear algebraic
groups.
Orthogonal and unitary groups are the groups preserving nondegenerate quadratic and
hermitian forms, respectively, on finite-dimensional vector spaces. If the vector space is
a real vector space, there are several equivalence classes of quadratic forms, so there are
various corresponding orthogonal groups. When the form is positive-definite, the associated
orthogonal group is On (R) (mentioned above). Over finite fields and p-adic fields, there can
be more than one equivalence class of quadratic forms. Similary, different hermitian forms
16 1. TOPOLOGICAL GROUPS

can give rise to nonisomorphic unitary groups. For example, when F = Qp , if n is even,
there are two isomorphism classes of unitary groups and if n is odd, there is one such class.
Almost all of these groups are noncompact.
If we have a topology on F , then GLn (F ), SLn (F ), Spn (F ), as well as many other matrix
groups, are topological groups. When F is a p-adic field, these groups are locally profinite
(with respect to the topology coming from F ). Such groups have another topology, the
Zariski topology (coming from the variety that is the algebraic group). In the representation
theory, the topology arising from the topology on F plays an important role. Results in the
structure theory of the groups depend on the underlying algebraic group structure, hence
are related to the Zariski topology.
The ad`eles AQ = R 0 Qp of Q is a subring of the direct product R p Qp . It
Q Q

consists of elements (a , a2 , a3 , a5 , ...) such that ap Zp for all but finitely many primes.
Q
Consider the set of open neighbourhoods of zero of the form UR p Up , where UR is an
open neighbourhood of zero in R, Up is an open neighbourhood of zero in Qp , and Up = Zp
for all but finitely many primes p. This set forms a base of neighbourhoods of zero for a
topology on AQ . This topology is locally compact. More generally, if F is a number field
(that is, a finite extension of Q), we can define the ring of ad`eles of F similarly. Since AF
is a locally compact topological ring, GLn (AF ) is a locally compact topological group. This
is also the case if G is a linear algebraic group defined over F : G(AF ) is a locally compact
group. Certain irreducible representations of G(AF ), known as automorphic representations,
are basic objects in the theory of automorphic forms. Such representations can be expressed
as restricted tensor products of infinite-dimensional representations of the locally compact
groups G(Fv ), where Fv runs over the set of completions of F with respect to nontrivial
valuations.

4. Haar measure on locally compact Hausdorff groups

Let G be a locally compact Hausdorff group. Integration on G and on various coset

spaces plays a role in the construction of unitary representations, including discrete series
representations, in the construction of induced representations, in relating representations of
G to representations of the Banach space of integrable functions L1 (G) on G, and in defining
characters of infinite-dimensional representations.

Definition 4.1. If X is a topological space, a -ring in X is a nonempty family of

subsets of X having the property that countable unions of elements in the family belong to
 4. HAAR MEASURE ON LOCALLY COMPACT HAUSDORFF GROUPS 17

the family, and if A and B belong to the family, then so does { x A | x / B }. If X is

a locally compact topological space, the Borel ring in X is the smallest -ring in X that
contains the open sets. The elements of the Borel ring are called Borel sets. A function
f : X R is (Borel) measurable if for every t > 0, the set { x X | |f (x)| < t } is a Borel
set.

Definition 4.2. Let be a positive Borel measure on X.

(1) The measure is called regular if, for any Borel subset S of X,

(S) = inf{ (U ) | U S, U open } and (S) = sup{ (C) | C S, C compact }.

(2) The measure is a called a Radon measure if

(i) (C) < for any compact set C X,
(ii) (S) = inf{ (U ) | U S, U open } for any Borel subset S of X,
(iii) (U ) = sup{ (C) | C U, C compact } for any open subset U of X.

A -finite Radon measure is regular. (Recall that -finite means the space is a countable
union of measurable sets, each having finite measure.)

Definition 4.3. A Radon measure on G is left invariant (respectively, right invariant)

if (gS) = (S) (respectively, (Sg) = (S)) for any g G and Borel subset S of G.

Theorem 4.4. ([Hal], [HR], [L]) There exists a nonzero left-invariant Radon measure
` on G. It satisfies ` (U ) > 0 for any nonempty open subset U of G. If ` is any nonzero
left-invariant Radon measure on G, then there exists c > 0 such that ` = c ` .

The measure ` is called a left Haar measure on G. There is also a right Haar measure
r , unique up to positive constant multiples, on G. Right and left Haar measures do not
usually coincide. See Proposition 4.11(3) for more information.

Exercise 4.5. Let ( ! )

x y
G= | x R , y R .
0 1
Show that |x|2 dx dy is a left Haar measure on G and |x|1 dx dy is a right Haar measure on
G.

If X is a locally compact topological space, let Cc (X) be the space of continuous complex-
valued functions on X with compact support. Let Cc+ (X) be the set of real-valued functions
f Cc (X) such that f (x) 0 for all x G.
18 1. TOPOLOGICAL GROUPS
R
If f Cc (G), then f is integrable with respect to ` and we often write G f (g) d` (g)
R
as G f (g) d` g. We use the same simplification for many kinds of functions on G (for exam-
ple, non-negative measurable, Haar integrable, or vector valued-versions of integrability) for
R
which G f (g) d` g makes sense. The left invariance of ` implies that
Z Z
f (g0 g) d` g = f (g)d` g, g0 G.
G G
R
If f Cc+ (G)and f is not identically zero, then G f (g) d` g > 0.
Existence of Radon measures can be expressed in terms of continuous linear functionals
on the space Cc (G). The following remark is useful when working with Haar measure.

Remark 4.6. The notion of (left) Haar measure is essentially equivalent to the notion
of a linear functional on the space Cc (G) such that
(i) For each compact subset C of G, there exists a constant MC such that for every
f Cc (G) with support inside C, |(f )| MC supgC |f (g)|. (If is the linear
functional corresponding to a particular left Haar measure ` , then MC = ` (C).)
(ii) (Lg f ) = (f ) for all g G and f Cc (G), where (Lg f )(x) = f (g 1 x), g, x G.

Definition 4.7. Let Let L1 (G) be the Banach space obtained as the completion of
R
Cc (G) with respect to the norm kf k1 := G |f (g)| d` g. (As usual, two functions in L1 (G) are
regarded as equivalent if and only if they differ on a set of measure zero.) Let L2 (G) be the
R
1/2
completion of Cc (G) with respect to the norm kf k2 := G |f (g)|2 d` g . Since the L2 norm
R
is associated to the inner product (f1 , f2 ) = G f1 (g)f 2 (g) d` g, L2 (G) is a Hilbert space.

Definition 4.8. A (locally compact Hausdorff) group G is called unimodular if each

left Haar measure is also a right Haar measure.
R R
If G is unimodular, we write G f (g) dg instead of G f (g) d` g. Clearly, G is unimodular
if G is abelian.

Lemma 4.9. Let ` be a left Haar measure on G. If : G G is an automorphism of

G and a homeomorphism, then ` is a left Haar measure on G. (Hence there exists a
positive real number c such that ` = c ` . Note that c is independent of the choice of
` .)

If g0 G, let Int g0 : G G be defined by Int g0 (g) = g0 gg01 , g G. Because Int g0 is

an automorphism of G and a homeomorphism, the measure S 7 ` (Int g0 (S)) = ` (Sg01 )
 4. HAAR MEASURE ON LOCALLY COMPACT HAUSDORFF GROUPS 19

(S measurable) is also a left Haar measure. By uniqueness of left Haar measure, there exists
a constant (g0 ) > 0 such that ` (Sg01 ) = (g0 )` (S). This can also be expressed as
Z Z Z
1
(f Int(g0 ))(g) d` g = f (gg0 ) d` g = (g0 ) f (g) d` g, f integrable
G G G

Definition 4.10. A quasicharacter of G is a continuous homomorphism from G to C .

Proposition 4.11. (1) The function : G R

+ is a quasicharacter.
(2) The restriction of to a compact subgroup of G is trivial.
(3) (g)1 d` g is a right Haar measure on G.

Proof. (1) The fact is a homomorphism, follows from the fact that right multiplica-
tion is an action of G on itself. Suppose that f Cc (G). Since f has compact support, f is
uniformly continuous. Given  > 0, there is an open neighbourhood U of the identity 1 in
G such that for u U and g G, |f (gu) f (g)| < . Without loss of generality, we may
assume that the closure U of U is compact. Let S be the support of f . For each u U , the
1
function g 7 f (gu) f (g) is supported in the compact set C := SU S. By continuity
of integration (see Remark 4.6), there exists a constant MC such that
Z Z
| f (gu) d` g f (g) d` g| < MC , u U.
G G

This can be rewritten as

Z
1
|(u) 1| | f (g) d` g| < MC , u U.
G
R
We may choose f such that G f (g) d` g is nonzero. Hence is continuous at 1. The general
case follows because is a homomorphism.
(2): By continuity of , the image of a compact subgroup of G under is a compact
subgroup of R
+ . The trivial subgroup is the only compact subgroup of R+ .
(3): We know that Z Z
(g0 )1 f (g) d` g = f (gg0 ) d` g
G G
for any Haar integrable function f on G (actually its enough to check things for f Cc (G)).
Replacing f by f 1 and multiplying both sides by (g0 ), we obtain
Z Z
1
f (g)(g) d` g = f (gg0 )(g)1 d` g.
G G
20 1. TOPOLOGICAL GROUPS

This shows that (g)1 d` g is right invariant. The rest of the proof is left as an exercise and
involves checking continuity (see Remark 4.6) or verifying that (g)1 d` g is a regular Borel
measure using the fact that is continuous and positive-valued. 

Definition 4.12. The function is called the modular (quasi)character or modular

function of G.

Clearly, G is unimodular if and only if the modular quasicharacter is trivial. If G is

unimodular, we simply refer to Haar measure on G. Reductive groups over local fields
are unimodular. However, many of their closed subgroups, including the proper parabolic
subgroups (see Exercise 4 below), are not unimodular.

Exercise 4.13. (1) Let dX denote Lebesgue measure on Mn (R). This is a Haar
measure on Mn (R). Show that | det(g)|n dg is both a left and a right Haar measure
on GLn (R). Hence GLn (R) is unimodular.
(2) Fix a prime p. Let dt denote a Haar measure on Qp . For convenience, we normalize
dt so that the measure of the compact subgroup Zp is equal to one.
(i) Show that if n is a nonzero integer, the measure of pn Zp is equal to pn . (This
can be done by induction. As a first step, note that the index of pZp in Zp is
p.)
(ii) Show that if x Q p is nonzero, then (xZp ) is equal to |x|p .
(iii) Show that |t| dt is a Haar measure on Q
1
p . (Hint: If c Qp , then c : x 7 cx
is an automorphism of Qp and a homeomorphism. Hence c is a Haar
measure on Qp . Because (cZp ) = |c|p (Zp ), it follows that c = |c|p .)
(3) Fix a prime p. Let dt denote a Haar measure on Qp . The product measure dg =
n
Q
i,j dgij is a Haar measure on the additive group Mn (Qp ). Show that | det(g)|p dg
is both a left and a right Haar measure on GLn (Qp ).
(4) Let n1 and n2 be positive integers such that n1 +n2 = n. P = P(n1 ,n2 ) be the standard
parabolic subgroup of GLn (F ) (F = R or F = Qp ) corresponding to the partition !
g1 X
(n1 , n2 ). That is, P is the matrices in GLn (F ) of the form g = P,
0 g2
with gj GLnj (F ), j = 1, 2, and X Mn1 n2 (F ). Let dgj be Haar measure
on GLnj (F ), and let dX be Haar measure on Mn1 n2 (F ) ' F n1 n2 . Show that
d` g = | det g2 |n2 dg1 dg2 dX and dr g = | det g2 |n1 dg1 dg2 dX are left and right Haar
measures on P (respectively). Here, | | is the usual absolute value if F = R
 5. COSET SPACES AND QUASI-INVARIANT MEASURES 21

and | | = | |p if F = Qp . Hence the modular quasicharacter of P is equal to

(g) = | det g1 |n1 | det g2 |n2 .
(5) Show that the homeomorphism g 7 g 1 turns ` into a right Haar measure. Con-
clude that if G is unimodular, then G f (g) d` g = G f (g 1 )d` g for all integrable
R R

functions f on G.

Proposition 4.14. If G is compact, then G is unimodular and ` (G) < .

Proof. It follows from Proposition 4.11(2) that G is unimodular. Haar measure on any
locally compact group has the property that any compact subset has finite measure. Hence
` (G) < whenever G is compact. 
If G is compact, normalized Haar meaure on G is the unique Haar measure on G
such that (G) = 1. When working with compact groups, we will usually work relative to
normalized Haar measure.

Exercise 4.15. Prove that if G is such that the measure of G (with respect to left Haar
measure) is finite, then G is compact. (Hint: Assume that G is noncompact. Let C be a
compact set in G containing an open neighbourhood of 1. Let U be an open neighbourhood
of 1 such that g 1 U for all g U and gg 0 C for all g, g 0 U . Show that there exist
elements gj G, j N, such that gi U gj U = whenever i 6= j. Use this to show that the
measure of G is infinite.)

5. Coset spaces and quasi-invariant measures

The most important method of constructing representations of locally compact groups is

the procedure of inducing representations from subgroups. Let H be a closed (not necessarily
normal) subgroup of a Hausdorff locally compact group G. If is a representation of H,
then the induced representation IndG H is a representation of G. When is unitary, the
representation IndGH should also be unitary. In order to achieve this, the measure-theoretic
relations between G, H and G/H must be taken into account. In particular, the notion of
quasi-invariant measure on the set of left cosets G/H (with the quotient topology) is key.
Let G/H be the set of left cosets of H in G, with the quotient topology. Fix a left Haar
R
measure `,H on H. For f Cc (G), Define f H (g) = H f (gh) d`,H (h), for g G. By left-
invariance of the Haar measure on H, f H (gh) = f H (g) for all h H and g G. Therefore,
there is a unique function f ] on G/H such that f ] (gH) = f H (g) for all g G.

Lemma 5.1. If f Cc (G), then f ] Cc (G/H).

22 1. TOPOLOGICAL GROUPS

Proof. Fix f Cc (G) and g0 G. Choose an open neighbourhood U of 1. Since G

is locally compact, we may assume that U is compact. Let  > 0. Because f is uniformly
continuous and G is locally compact, there exists an open neighbourhood U 0 of 1 such that
U 0 U and and |f (ug) f (g)|  for all u U 0 and g G.
1
Let u U 0 . Then f (ug0 h) = 0 and f (g0 h) = 0 whenever h H \ (H g01 U supp f ).
1
Since H g01 U supp f is a compact subset of H, continuity of Haar measure on H implies
that there exists a nonnegative constant c (which we can take to be the measure of the set
with respect to the chosen Haar measure) such that
Z 
] ]
 
|f (ug0 H) f (uH)| =  (f (ug0 h) f (g0 h)) d`,H h  c.
H

It follows that f ] is continuous.

Since supp f ] q(supp f ), the function f ] is compactly supported. 

Lemma 5.2. Let q : G G/H be the canonical map. If C is a compact subset of G/H,
then there exists a compact subset C 0 of G such that q(C 0 ) = C.

Proof. Let C be a compact subset of G/H. Since G is locally compact, there exists an
open neighbourhood U of the identity in G such that U is compact. Since q is open, q(U ) is
open in G/H, and so are the translates g q(U ) for g G. The sets { g q(U ) | g G } form an
open cover of C. By compactness of C, there exists a finite subcover { gi q(U ) | 1 i n }.
We know that q 1 (C) is closed in G and, since G is Hausdorff, g1 U gn U is compact
in G. The set C 0 := q 1 (C) (g1 U gn U ) is compact and satisfies q(C 0 ) = C. 

Proposition 5.3. The mapping f 7 f ] is onto Cc (G/H). An element of Cc+ (G/H) is

of the form f ] for some f Cc+ (G).

Proof. Fix F Cc (G/H). Let C = supp F . Fix a compact subset C 0 of G such that
q(C 0 ) = C. Choose Cc (G) such that (g) = 1 for all g C 0 and 0 (g) 1 for all
g G. (This uses Urysohns Lemma see Theorem 2.1 of [K].) Then ] > 0 on C.
Define a function on G/H by

F (gH)/] (gH) if gH C
(gH) =
0 if gH / C.
 5. COSET SPACES AND QUASI-INVARIANT MEASURES 23

Then Cc (G/H). For g G, let f (g) = (gH)(g). Then f Cc (G) and f 0 if F 0.

If g G, then
Z
]
f (gH) = (ghH)(gh) d`,H h = (gH)] (gH) = F (gH).
G

Given a bounded, continuous, complex-valued function f on G, let kf k = supgG |f (g)|.

(kf k is called the supremum norm of f ). If K is a compact subset of G that contains an
open neighbourhood of 1, let CK (G) = { f Cc (G) | supp f K }. We dont need the next
result at the moment, but it may be used later.

Proposition 5.4. Let K be a compact subset of G that contains an open neighbourhood

of 1. Then there exists a positive constant cK such that kf ] k cK kf k , for all f CK (G).

Proof. Fix f Cc (G) with supp f K. Note that f ] (gH) = 0 unless g KH.
Assume that g K and h H. If gh supp f , then h (K 1 K) H. But (K 1 K) H is
compact, so has finite Haar measure. Let cK be the H-Haar measure of (K 1 K) H. Then,
for g K, Z
|f ] (gH)| |f (gh)| d`,H h cK kf k .
H
It follows that kf ] k cK kf k . 

Define (Lg )(g0 H) = (g 1 g0 H) for Cc (G/H) and g, g0 G. Note that (Lg f )] =

Lg (f ] ) for all f Cc (G) and g G.

Proposition 5.5. Let G and H be the modular functions on G and H, respectively.

If f Cc (G), define
Z
f (g) = G (h)H (h)1 f (gh) d`,H (h), g G.
H

Then f is continuous and f (gh) = H (h)G (h)1 f (g) for g G and h H. If f

Cc+ (G), then f (g) 0 for all g G.

Proof. Continuity of f follows from the fact that f has compact support. The trans-
formation property follows from a change of variables. 

Functions of the form f are used to transfer integration between G and G/H.
24 1. TOPOLOGICAL GROUPS

Proposition 5.6. Let be a locally integrable function on G which satisfies (gh) =

H (h)G (h)1 (g) for g G and h H. If f Cc (G) is such that f ] 0, then
Z
f (g) (g) d` g = 0.
G

Proof. The set C := q(supp f ) is a compact subset of G/H. Choose a function in

Cc (G/H) that takes the value 1 on C. By Proposition 5.3, this function is of the form ]
for some Cc (G). In particular, H (g) = 1 for all g supp f . By assumption f H (g) = 0
R
for all g G. So we have G (g)(g)f H (g) d` g = 0. That is
Z Z
0= (g) (g) f (gh) d`,H h d` g.
G H

Consider the function (g, h) 7 (g)(g)f (gh) on G H. If the value at (g, h) is nonzero,
then g supp and h g 1 supp f . This implies (g, h) lies in the compact set

supp ((supp )1 supp f ) H.

That is, the function is compactly supported. Since the function is integrable on G H and
vanishes outside a set of finite measure, we can use Fubinis Theorem.
Below we apply Fubinis theorem, replace g by gh1 , apply Fubinis theorem again, make
the change of variables h 7 h1 , use properties of , and the fact that H (g) = 1 for all
g supp f :
Z Z Z Z
0= (g)(g)f (gh)d`,H h d` g = (g)(g)f (gh)d` g d`,H h
G H H G
Z Z
= (gh1 )(gh1 )f (g)G (h1 )d` g d`,H h
ZH Z G
= G (h)H (h1 )(gh)(gh)f (g)d`,H h d` g
ZG H Z
]
= (g)f (g) (gH)d` g = (g)f (g) d` g
G G

Given a measurable function on G such that 0 and (gh) = H (h)G (h)1 (g)
for h G and g G, we can define a linear functional on Cc (G/H) by (f ] ) =
R
G
f (g)(g) d` g, f Cc (G/H). Note that we can apply Proposition 5.6 to see that is
well-defined.
 5. COSET SPACES AND QUASI-INVARIANT MEASURES 25

Proposition 5.7. Let be as above. Then there exists a regular Borel measure on
G/H such that Z Z
]
f (
g )d g = f (g)(g) d` g, f Cc (G).
G/H G

Proof. According to Proposition 5.3, given Cc+ (G/H), there exists f Cc+ (G) such
R
that f ] = . Note that () = (f ] ) = G f (g)(g)d` g 0, since f Cc+ (G). That is,
is a positive linear functional. Applying the Riesz Representation Theorem, is given by
a regular Borel measure on G/H. (The Riesz Representation Theorem holds for locally
compact Hausdorff spaces. It says that each positive linear functional is given by integration
against a positive measure.) 

Theorem 5.8. There exists a nonzero positive G-invariant regular Borel measure on
G/H if and only if G (h) = H (h) for all h H. When it exists, such a measure d
g is
unique up to positive scalar multiples and can be normalized so that it satisfies
Z Z Z
f (gh) d`,H h d
g= f (g) d` g, f Cc (G).
G/H H G

Proof. Assume that G | H = H . We may apply Proposition 5.7 with (g) = 1 for
all g G. Let dg = d g.
If x G, then
Z Z Z
] 1
f (x g) d
g= ]
(Lx f ) (
g ) d
g= f (x1 g) d` g
G/H G/H G
Z Z
= f (g) d` g = f ] (
g ) d
g, f Cc (G).
G G/H

This shows that d g is G-invariant.

For the converse, assume that d g is a nonzero positive invariant measure on G/H. Set
R ]
(f ) = G/H f ( g for f Cc (G). Then, if x G, we can show that (Lx f ) = (f )
g ) d
(similar to the above calculation). So is a nonzero positive left-invariant linear functional
on Cc (G). By uniqueness of the left Haar measure on G, there exists a positive c such that
Z Z
]
f (
g ) d
g=c f (g) d` g, f Cc (G).
G/H G
26 1. TOPOLOGICAL GROUPS
R
Choose f Cc (G) such that G f (g) d` g = 1 Fix h H. Set h (g) = H (h)f (gh), g G.
Then ]h = f ] (details omitted). Therefore
Z
H (h)G (h) 1 = (f (g)H (h)G (h1 ) f (g)) d` g
1

ZG
= (f (gh)H (h) f (g)) d` g
ZG Z
= h (g) d` g f (g) d` g
G G
Z
=c 1
(]h (
g ) f ] (
g )) d
g = 0.
G/H

Remark 5.9. Let H be a closed normal subgroup of G. Then H = G | H. Let d g

be a left Haar measure on G/H. Then a left Haar measure on G/H is left G-invariant,
R
so f 7 G/H f ] ( g defines a left Haar measure on G. Fix h0 H. Replace f by the
g ) d
function g 7 f (gh0 ). Examining the value of the integral of this new function using the
above realization of left Haar measure on G, we find that H (h) = G (h).

If is a measure on G/H and x G, define a measure x by x (S) := (x S) for any

measurable subset S of G/H. If is a regular Borel measure on G/H, we say that is
quasi-invariant if is nonzero and x for all x G. Here denotes mutual absolute
continuity of measures (essentially, and x have the same sets of measure zero).

Lemma 5.10. There exists a continuous function on G such that (g) > 0 for all g G
and (gh) = H (h)G (h)1 (g) for h G.

Let be as in Lemma 5.10 and let = . For x G, define x by x(S) = (xS) for any
Borel set S G/H. It is easy to see that Lx satisfies (Lx )(gh) = H (h)G (h)1 (Lx )(g)
for all g G and h H and Lx = ( )x1 . Because the function y 7 (xy)/(y) is right
H-invariant, we can view it as a function on G/H:

(x, q(y)) = (xy)/(y), x, y G

The function (x, y) 7 (x, y) is a continuous, strictly positive function on G G/H. We

can easily see that

(5.1) (x, y)(z, x y) = (zx, y), y G/H, x, z G.

 5. COSET SPACES AND QUASI-INVARIANT MEASURES 27

If f Cc (G) and x G, then

Z Z Z
]
f (g ) dx (
g) = f (y) (Lx1 )(y) d` y = f (y) (x, q(y)) (y) dy
G/H G G
Z
= f ] (
g ) (x, y) d(
g ).
G/H

Applying Proposition 5.3, we have

Z Z
(
g ) dx (
g) = (
g ) (x, y) d(
g ), Cc (G/H).
G/H G/H

Theorem 5.11. Let be as in Lemma 5.10. Then the measure (see Proposition 5.7)
is quasi-invariant, the Radon-Nikodym derivative (x, y) = [dx /d] (y ) is continuous and
satisfies equation (5.1). Furthermore,
Z Z Z Z
f (g)(g) d` g = f (gh)d`,H h d (gH) = f ] (gH) d (gH), f Cc (G).
G G/H H G/H

Quasi-invariant measures on G/H are not unique up to scalar multiples, but any two of
them are mutually absolutely continuous. For more details, see Section 1.3 of [KT].

Example 5.12. (Exercise 1 (f)) from p.126 of [R]) Let G = GL2 (R) and let H be the
group of upper triangular matrices in G. Because G is unimodular and H is not unimodular,
there is no G-invariant measure ! on G/H. The quotient G/H can be identified with the
a b
projective line. If g = with ad 6= bc, the coset gH is identified with the line
c d
generated by (a, c) in R2 , or x = a/c R {}. Show that Lebesgue measure on R gives a
quasi-invariant measure on G/H.
 CHAPTER 2

Continuous representations of locally compact groups

From now on, G will denote a locally compact Hausdorff group.

1. Definition of continuous representation

For V be a normed complex vector space, let Aut(V ), B(V ) and GL(V ) be the set of
invertible linear operators on V , the set of bounded linear operators on V , and the set of
invertible elements in B(V ) which have bounded inverses, respectively.

Definition 1.1. If V is a vector space, (, V ) is an abstract representation (of G) if

is a homomorphism from G to Aut(V ). If V is finite-dimensional, say n = dim(V ), we say
that is finite-dimensional, or n-dimensional. When V is infinite-dimensional, we say that
is infinite-dimensional.

Definition 1.2. If V is a complex normed vector space, (, V ) is a continuous rep-

resentation (of G) if : G GL(V ) is a homomorphism and the map G V V is
continuous.

Example 1.3. Let be a one-dimensional representation of R. Then (0) = 1, (t1 +

t2 ) = (t1 )+(t2 ) for all t1 and t2 R, and (t) = (t)1 for all t R. Fix t0 R. Choose
a function f on R such that f is infinitely differentiable, f is zero outside a neighbourhood
R
of t0 , and there is a nonzero constant c such that f (t)(t) dt = c 6= 0. (Here, we have
used that (t) 6= 0 for all t.) Observe that
Z Z
(t1 + t2 )f (t2 ) dt2 = (t1 ) (t2 )f (t2 ) dt2 = (t1 ) c,

and Z Z
1 1
(t2 ) = c (t1 + t2 )f (t2 ) dt2 = c (t2 )f (t2 t1 ) dt2 .

The last integral is infinitely differentiable. Combining this with the property (t1 + t2 ) =
(t1 ) + (t2 ), we have that is a solution of the differential equation 0 (t) = z (t) for some
29
30 2. CONTINUOUS REPRESENTATIONS OF LOCALLY COMPACT GROUPS

complex constant z. It follows that (t) = ezt . Note that is unitary if and only if z is
purely imaginary.

The proof of the next proposition uses the fact that the Banach-Steinhaus Theorem holds
for normed barrelled spaces (or, more generally, locally convex barrelled spaces). Given a
normed barrelled space V and a nonempty family F in B(V ), pointwise boundedness of the
family implies that the family is equicontinuous: If for each v V , sup{kT (v)k | T F } <
, then sup{kT kop | T F } < . Since G is locally compact, this can be applied for
F = { (g) | g U }, where U is a compact neighbourhood of the identity.

Proposition 1.4. Let V be a normed vector space that is barrelled. Suppose that is a
homomorphism from G to Aut(V ). Then (, V ) is a continuous representation of G if and
only if
(1) The map g 7 (g)v is continuous for every v V .
(2) The map v 7 (g)v is continuous for every g G.

Remark 1.5. More generally, we can assume that V is a locally convex barrelled topo-
logical vector space. In a locally convex vector space, the topology is determined by a family
of seminorms.

We can define various topologies on B(V ). Let V = B(V, C) be the space of bounded
(that is, continuous) linear functionals on V .
The weak operator topology on B(V ) is the topology induced by the seminorms T 7
|(T (v))|, v V , V . If  > 0, V and v V , the set { T B(V ) | |(T (v))| <  }
is an open neighbourhood of zero. (Of course, if V is a Hilbert space, then V can be
realized as v 7 hv, wi for some fixed w V .)
The strong operator topology on B(V ) is the topology induced by the seminorms T 7
kT (v)k, v V . If  > 0 and v V , then { T B(V ) | kT (v)k <  } is an open neighbourhood
of zero. This topology has the property that each function T 7 T (v) from B(V ) to V is
continuous for all v V .
In the operator norm topology (or uniform topology) on B(V ), for  > 0, the set { T
B(V ) | kT kop <  } is an open neighbourhood of zero.
If V is finite-dimensional, all three of the above topologies coincide. In the infinite-
dimensional setting, the operator norm topology is stronger than the strong operator topol-
ogy, which in turn is stronger than the weak operator topology.
 1. DEFINITION OF CONTINUOUS REPRESENTATION 31

Lemma 1.6. Let (, V ) be a continuous representation of G. Then the function g 7 (g)

from G to GL(V ) is continuous when GL(V ) is given the subspace topology arising from the
strong operator topology on B(V ).

Proof. Fix g0 G. Given  > 0 and v0 V , the set Ug0 ,v0 , := { T GL(V ) | k(T
(g0 ))(v0 )k <  } is an open neighbourhood of (g0 ) in GL(V ). By assumption, the function
g 7 (g)v0 from G to V is continuous. Therefore, there exists an open neighbourhood U of
g0 in G such that if g U , then k(g)v0 (g0 )v0 k < . That is, for g U , (g) belongs to
the set Ug0 ,v0 , . Thus the function g 7 (g) is continuous at g0 . 

Note that when V is infinite-dimensional, GL(V ) is not a topological group with respect
to the strong operator topology. However, when V is finite-dimensional, the continuous
function g 7 (g) is also a continuous homomorphism. function. In the infinite-dimensional
setting, GL(V ) is a topological group in the uniform topology. But very few continuous
representations (, V ) have the property that the function g 7 (g) is continuous when
GL(V ) has the norm topology. In particular, as shown in [Sg], the trivial representation
is the only uniformly continuous unitary representation of a complex semisimple Lie group,
and all topologically irreducible uniformly continuous representations of real Lie groups are
finite-dimensional. See [Ka] for information about characterizations of uniformly continuous
unitary representations of connected locally compact groups.

Exercise 1.7. Let G = R/Z (or, equivalently, G = SO2 (R)).

a) Let V = C(G) be the space of continuous complex-valued functions on G, equipped
with the supremum norm: kf k = supgG |f (g)|. Then V is a Banach space. The
right regular representation of G on V is given by ((t)f )(x) = f (x + t), f V , x,
t G. Show that the homomorphism t 7 (t) from G to GL(V ) is not continuous
when GL(V ) is given the operator norm topology. (Hint: Viewing G as the interval
[0, 1] with its endpoints identified, suppose that 0 < t < 1/2. Take ft V such that
ft is supported in [0, t], 0 ft (x) 1, and ft (t/2) = 1. Note that kft k = 1. Show
that k(t)ft ft k = 1. This implies that k(t) Ikop 1, where I is the identity
operator on V .)
b) Let W = L2 (G). Let be the right regular representation of G on W . Show that
the homomomorphism t 7 (t) from G to the group of unitary operators U (W ) on
W is not continuous when U (W ) is given the operator norm topology.
32 2. CONTINUOUS REPRESENTATIONS OF LOCALLY COMPACT GROUPS

Proposition 1.8. Let V be a Hilbert space and let U (V ) be the unitary subgroup of
GL(V ). If : G U (V ) is an abstract homomorphism and for each v V , the map
g 7 hv, (g)vi is continuous at the identity, then (, V ) is a unitary representation of G.

It follows from the above proposition that if g 7 (g)v is continuous in the weak topology,
then it is continuous in the strong topology. The converse follows from the Cauchy-Schwarz
inequality: Given u, v V ,

|h(g)u, vi hu, vi| k(g)u ukkvk.

2. Representations of compact groups are unitary

Proposition 2.1. Let (, V ) be a continuous representation of a compact group in a

Hilbert space V . Then there exists a G-invariant positive-definite Hermitian form on V and
defines the same topology on V .
R
Proof. (The general idea). Define (v, w) = G h(g)v, (g)wi dg, v, w G. Show that
is G-invariant, positive, hermitian. Show that the norm v 7 (v, v)1/2 is equivalent to
k k are equivalent norms this involves applying the Banach-Steinhaus Theorem to see that
g 7 k(g)kop is bounded above. 

3. Irreducibility, subrepresentations, etc.

If (, V ) is a representation of G and W is a subspace of V , we say that V is G-invariant

if (g)W W for all g G.

Definition 3.1. An abstract representation (, V ) of a group G is algebraically irre-

ducible if there does not exist a nonzero proper G-invariant subspace of V .

Definition 3.2. A continuous representation (, V ) is topologically irreducible if there

does not exist a nonzero proper closed G-invariant subspace of V .

Recall that every subspace of a finite-dimensional normed space is closed. Hence for con-
tinuous finite-dimensional representations, the notions of algebraic irreducibility and topo-
logical irreducibility coincide.
It is easy to show that if V is an normed vector space, A B(V ), and W is a subspace
of V such that A(W ) W , then A(W ) W .

Lemma 3.3. If (, V ) is a continuous representation of G and W is a G-invariant sub-

space of V , then W is a G-invariant subspace.
 3. IRREDUCIBILITY, SUBREPRESENTATIONS, ETC. 33

Definition 3.4. Let (, V ) be a continuous representation of G. If W is a closed subspace

of V , we can define a continuous representation (W , W ) of G, by taking W (g) to be the
restriction of (g) to W , for g G. In this case, (W , W ) is called a subrepresentation of
(, V ). If (W , W ) is a subrepresentation of G, we can define a continuous representation
(V /W , V /W ) of G on the quotient space V /W (which is also a normed space, since W
is closed in V ) by V /W (g)(v + W ) = (g)v + W , g G, v V . The representation
(V /W , V /W ) is called a quotient (representation) of (, V ). A representation of G that
occurs as (W2 /W1 , W2 /W1 ) where W1 W2 are closed G-invariant subspaces of V is called
a subquotient of (, V ).

Definition 3.5. A representation (, V ) is pre-unitary if V is an inner product space

and h(g)v1 , (g)v1 i = hv1 , v2 i for all g G and v1 , v2 V . If V is a Hilbert space and is
continuous and pre-unitary, we say that is unitary.

Definition 3.6. If (i , Vi ), 1 i n are continuous representations of G, we may define

a continuous representation (i , Vi ) where (i )(g)(v1 , . . . vn ) = (1 (g)v1 , . . . , n (g)vn ).
Here, we use the norm
n
!1/2
X
k(v1 , . . . , vn )k = kvi k2i
i=1
on i Vi , where k ki is the norm on Vi , 1 i n. The representation (i , i Vi ) is called
the direct sum of the representations (i , Vi ), 1 i n. We may also define direct sums
of infinite collections of representations. For example, suppose that (i , Vi ), i I (I some
index set) is a collection of continuous unitary representations of G. Let h, ii be the inner
product on Vi . Let
Y X
iI Vi := { (vi )iI Vi | kvi k2i < }.
iI iI
P
Define h(vi ), (wi )i = iI hvi , wi ii . This defines an inner product on V that makes i Vi into
a Hilbert space. Set (i )(g)(vi ) = (i (g)vi ), g G, (vi ) V . Then the representation
(i , Vi ) is unitary. Representations of the form (i , Vi ) are said to be completely
reducible.

If W is a closed G-invariant subspace of a Hilbert space V , then W is a Hilbert space.

Hence a subrepresentation of a unitary representation is a unitary representation.
34 2. CONTINUOUS REPRESENTATIONS OF LOCALLY COMPACT GROUPS

Lemma 3.7. Let (, V ) be a continuous unitary representation of G. A closed subspace

W of V is G-invariant if and only if the orthogonal complement W of W is G-invariant.
In this case (, V ) is the direct sum of the subrepresentations (W , W ) and (W , W ).

Corollary 3.8. A finite-dimensional unitary representation of G is completely re-

ducible.

Recall that we have shown (see Proposition 2.1) that a representation of a compact group
in a Hilbert space is unitary.

Corollary 3.9. A finite-dimensional continuous representation of a compact group is

completely reducible.

Definition 3.10. If (, V ) is a continuous representation of G, let V = B(V, C) be

the space of continuous linear functionals on V . Given V and v V , the continuous
function g 7 ((g)v) is called a matrix coefficient of . If V is a Hilbert space, a matrix
coefficient of has the form g 7 h(g)v, wi, where v and w are fixed vectors in V .

Exercise 3.11. Let (, V ) be a unitary representation of G. Prove that is topologically

irreducible if and only if the the matrix coefficient g 7 h (g)v, w i is not identically zero for
all nonzero vectors v and w in V .

Example 3.12. Consider the two-dimensional representation (, C2 ) of G = R defined

by (x)(z1 , z2 ) = (z1 + xz2 , z2 ), x G, z1 , z2 R. The subspace W := Span{(1, 0)} is
the unique one-dimensional invariant subspace of C2 . In particular, there is no G-invariant
complementary subspace, so is not completely reducible.

4. The regular representation on L2 (G/H)

Let H be a closed subgroup of G. Let be a quasi-invariant measure on G/H. This

measure has the property that
Z Z
g )d(x g) =
f ( (x, g) f ( g , x G, f Cc (G/H),
g ) d
G/H G/H

where is a nonnegative continuous function on G G/H that satisfies

(y, g)(x, y g) = (xy, g), x, y G, g G/H,

and (x, yH) = (xy)/(y), x, y G, where is a strictly positive continuous function as

in Lemma 5.10.
 5. INTERTWINING OPERATORS AND SCHURS LEMMA 35

If G | H = H , we will take to be a G-invariant measure on G/H. (In this case,

(x, g) = 1 for all x G and g G/H.)
For x G, g G/H and f L2 (G/H, ), set
p
g ) = (x1 , g) f (x1 g).
((x)f )(

It is a simple matter to check that k(x)f k = kf k and (xy)f = (x)(y)f for f

L2 (G/H, ) and x, y G. We refer to as the left regular representation of G on L2 (G/H, ).

Proposition 4.1. Let be as above. Then is a continuous unitary representation of

G.

More will be added later. For the moment, we only mention that given  > 0 and
f L2 (G/H, ), there exists Cc (G/H) such that kf k2 < /3. Using the fact that
each (x) is an isometry, we can show that

k(x)f f k2 < k(x) k2 + 2/3, x G.

Consider the map x 7 Lx from G to Cc (G/H) given by (Lx )( g ) = (x1 g), x G,

g G/H. If Cc (G/H) is given the topology associated to the sup norm, this map is
continuous (details to be added). Combining this fact with continuity of , we can show
that there exists an open neighbourhood U of 1 in G such that k(x) k2 < /3 for all
x U.
In the special case where H = {1}, we have the left regular representation of G on
L2 (G), where we use a left Haar measure on G. For x and y in G and f L2 (G), set
((x)f )(y) = G (x)1/2 f (yx). Then is a continuous unitary representation of G, called the
right regular representation of G.

Exercise 4.2. Let X = GL2 (R)/O2 (R), where

O2 (R) = { g GL2 (R) | (gv, gw) = (v, w) v, w R2 },

where (, ) is the standard inner product. Consider the space L (X) of essentially bounded
measurable functions on X, equipped with the essential supremum norm. Set ((x)f )( g) =
f (x1 g), x G, g X, f L (X). Show that is not a continuous representation of
GL2 (R).

5. Intertwining operators and Schurs Lemma

Definition 5.1. Let (1 , V1 ) and (2 , V2 ) be continuous representations of G.

36 2. CONTINUOUS REPRESENTATIONS OF LOCALLY COMPACT GROUPS

(1) An intertwining operator from 1 to 2 is a linear transformation A B(V1 , V2 )

such that A 1 (g) = 2 (g) A for all g G. The notation HomG (1 , 2 ) (or
HomG (V1 , V2 )) will be used for the set of intertwining operators from 1 to 2 .
(2) We say that 1 and 2 are equivalent if HomG (1 , 2 ) contains a homoemorphism.
In this case, we write ' 2 .
(3) If 1 and 2 are unitary representations, we say that 1 and 2 are unitarily equiv-
alent if HomG (1 , 2 ) contains a homeomorphism that is an isometry.
It is easy to see that equivalence of representations is an equivalence relation. For the
rest of this section, assume that (1 , V1 ) and (2 , V2 ) are continuous representations of G.
Proposition 5.2. Suppose that A HomG (1 , 2 ). Let W1 be the kernel of A and let
W2 be the closure of the range of A. Then
(1) W1 and W2 are G-invariant.
(2) If 1 and 2 are finite-dimensional and irreducible and A is nonzero, then A is an
isomorphism.
(3) If 1 and 2 are unitary, let (1 , W1 ) be the restriction of 1 to W1 and let (20 , W2 )
be the restriction of 2 to W2 . Then 1 and 20 are unitarily equivalent.
More will be added later. For the third part, we use the polar decomposition of A: A = A1 A2 ,
where A1 is a partial isometry of V1 onto W2 with kernel W1 and A2 = (A A)1/2 . We can
show that A1 HomG (1 , 20 ). Since the restriction of A1 to W1 is an isometry of W1 onto
W2 , 1 and 20 are unitarily equivalent.

Corollary 5.3. Assume that 1 and 2 are unitary and there exists A HomG (1 , 2 )
such that A is one-to-one and the range of A is dense in V2 , then 1 and 2 are unitarily
equivalent.
Corollary 5.4. Assume that 1 and 2 are unitary and topologically irreducible. If
HomG (1 , 2 ) is nonzero, then 1 and 2 are unitarily equivalent.
Corollary 5.5. Assume that 1 and 2 are unitary. Then HomG (1 , 2 ) is nonzero if
and only if some subrepresentation of 1 is (unitarily) equivalent to a subrepresentation of
2 .
Example 5.6. Let G = C n C and V = C2 (written ! as column vectors). Let (, V )
t x
be the representation of G such that (t, x) = , t C , x C. The subspace
0 1
 5. INTERTWINING OPERATORS AND SCHURS LEMMA 37

W = Span(e1 ) is a unique one-dimensional G-invariant

! suspace of V . Let ( 0 , V ) be the
t 0
representation of G such that 0 (t, x) = , t C , x C. Then HomG (, 0 )
0 1
and HomG ( 0 , ) are both one-dimensional. A nonzero element of HomG ( 0 , ) intertwines a
subrepresentation of 0 and the unique subrepresentation of . On the other hand, a nonzero
element of HomG (, 0 ) does not interwine a subrepresentation of and a subrepresentation
of 0 . Instead, it interwines the unique nonzero quotient of and a subrepresentation of 0 .

Exercise 5.7. Let (, V ) be a unitary representation of G. Suppose that { Wi | i I }

is a family of closed G-invariant subspaces of V such that
(1) Wi Wj if i 6= j
(2) iI Wi is total in V (that is, the set of finite linear combinations of vectors from
iI Wi is dense in V )
For g G, let i (g) be the restriction of (g) to Wi . Prove that (, V ) is equivalent to
(i , Wi ).

Proposition 5.8. (Schurs Lemma) Let V be a Hilbert space and let F be a collection
of operators in B(V ) such that T F if and only if T F. Assume that F is topologically
irreducible. (That is, {0} and V are the only closed subspaces of V which are invariant
under all of the operators in F.) Then { S B(V ) | ST = T S T F. } consists of scalar
multiples of the identity operator.

Further comments will be added here. It is easy to see that if S B(V ) commutes with all
operators in F, then S + S and i(S S ) (which are both self-adjoint) also commute with
all operators in F. The spectral theorem for self-adjoint operators can be applied to show
that S + S and i(S S ) are both scalar multiples of the identity operator.

Corollary 5.9. (Schurs Lemma for unitary representations) Let (, V ) be a unitary

representation of G. Then is topologically irreducible if and only if HomG (, ) consists of
scalar multiples of the identity.

Lemma 5.10. If (, V ) is a (nonunitary) finite-dimensional irreducible representation of

G, then HomG (, ) consists of scalar multiples of the identity operator.

Proof. Let A be a nonzero element of HomG (, ). Because V is finite-dimensional, A

has an eigenvalue, say . Note that A I HomG (, ). Since the kernel of A I is
nonzero and G-invariant and is irreducible, the kernel must equal V . That is, A = I. 
38 2. CONTINUOUS REPRESENTATIONS OF LOCALLY COMPACT GROUPS

The next example shows that the converse of the lemma does not necessaily hold in the
nonunitary setting.

Example 5.11. Consider the representation (, V ) from Example 5.6. We can see that,
although is not irreducible, HomG (, ) consists of scalar multiples of the identity operator
on V .

Corollary 5.12. Assume that 1 and 2 are unitary and topologically irreducible (or
finite-dimensional, not necessarily unitary, and irreducible). Then

1 if '
1 2
dim HomG (1 , 2 ) =
0 otherwise.

Proof. Suppose that HomG (1 , 2 ) 6= 0. Let A HomG (1 , 2 ) be nonzero. Because

A is nonzero, the kernel of A is a closed G-invariant subspace of V1 , and 1 is topologically
irreducible, we have that A is one-to-one. In the unitary case, note that A HomG (2 , 1 ).
It follows that AA HomG (2 , 2 ). Since 2 is topologically irreducible, AA must be a
scalar multiple of the identity operator on V2 . This can be used to see that A is onto. In the
finite-dimensional case, the range of A is a nonzero closed G-invariant subspace of V2 , so, by
irreducibility of 2 , A is onto. We have shown that nonzero elements of HomG (1 , 2 ) are
invertible. It follows that 1 and 2 are equivalent if and only if HomG (1 , 2 ) is nonzero.
It remains to show that the dimension of HomG (1 , 2 ) is one whenever 1 ' 2 . Let A1 ,
A2 HomG (1 , 2 ) be such that A2 is nonzero. Then we know that A2 is invertible. Since
A1 1
2 A1 HomG (1 , 1 ), by Schurs Lemma, A2 A1 is a scalar multiple of the identity. That
is, A1 = c A2 for some complex number c. Hence dim HomG (1 , 2 ) = 1. 

Example 5.13. The left regular representation of G on L2 (G) is not topologically irre-
ducible. When G is compact, it is easy to see that the trivial representation of G occurs
as a subrepresentation. However, if G is noncompact, it can be shown that the left regular
representation has no finite-dimensional subrepresentation. Instead, show that the opera-
tors of the right regular representation intertwine the left regular representation. Apart from
the identity element, they are not scalar multiples of the identity operator. Apply Schurs
Lemma.
 CHAPTER 3

Representations of compact groups: the general theory

Throughout this chapter, unless we explicitly say otherwise, G is assumed to be a compact

group. Recall (Proposition 2.1) that a continuous representation of G in a Hilbert space is
unitary (with respect to some inner product on the space that defines a norm which is
equivalent to the Hilbert space norm).
The Peter-Weyl Theorem is a collection of results about continuous representations
of G, which may be viewed as analogues of results from the representation theory of finite
groups.

1. The Peter-Weyl Theorem

Recall that G is unimodular and has finite measure with respect to any Haar measure
on G. We fix a Haar measure on G such that the measure of G is equal to one. Let
C(G) be the space of continuous complex-valued functions on G. With the supremum norm
kf k = supgG |f (g)|, C(G) is a Banach space.
Using continuity of integration and the fact that the measure of G is equal to one, we
can see that the L2 norm of a continuous function is bounded above by the supremum norm
the function:

Lemma 1.1. If f C(G), then kf k2 kf k .

According to Proposition 4.1, the left regular representation of G on L2 (G) is continuous.

Unitarity follows from G-invariance of Haar measure. These properties will be used below.

Definition 1.2. If f1 , f2 C(G), we define the convolution f1 f2 of f1 with f2 by:

(f1 f2 )(x) = G f1 (y)f2 (y 1 x) dy, x G.
R

We can define convolution more generally, for f1 and f2 L2 (G). Let denote the left
regular representation of G on L2 (G). Let f2 be the function defined by f2 (y) = f2 (y 1 ),
y G. Note that (f1 f2 )(x) = G f1 (y)((x)f2 )(y) dy = h(x)f2 , f1 i. Therefore f1 f2 is the
R

matrix coefficient of attached to the vectors f2 and f1 . By continuity of , f1 f2 C(G).

39
40 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

The next lemma follows from the Cauchy-Schwartz inequality and the fact that is
unitary.

Lemma 1.3. If f1 , f2 L2 (G), then k(f1 f2 )k kf1 k2 kf2 k2 .

Lemma 1.4. If f2 L2 (G), define a linear transformation Kf2 : L2 (G) C(G) by

Kf2 (f1 ) = f1 f2 . This is a compact transformation.

Proof. A compact transformation is one that maps bounded sets to relatively compact
sets (sets having compact closure). Let F = {Kf2 (f1 ) | f1 L2 (G), kf1 k2 1 }. To see that
Kf2 is compact, it suffices to prove that F is a relatively compact subset of C(G).
According to Ascolis Theorem (see p.7 of [R] for the general statement), F is relatively
compact if and only if F is equicontinuous and each of the sets F(x) = { Kf2 (f1 )(x) | f1
L2 (G), kf1 k2 1 } is relatively compact in C.
To see that F(x) is relatively compact, apply Lemma 1.3 to conclude that kKf2 (f1 )(x)|
kf2 k2 whenever kf1 k2 1. This shows that F(x) is bounded. A bounded subset of C is
relatively compact.
To verify that F is equicontinuous, note that if x1 , x2 G and kf1 k 1, then
|Kf2 (f1 )(x1 ) Kf2 (f1 )(x2 )| = |h(x1 )f2 h(x1 )f2 , f1 i|
k(x1 )f2 (x2 )f2 k2 kf1 k2 k(x1 )f2 (x2 )f2 k2 .
Let  > 0. By continuity of , there exists an open neighbourhood U of x1 such that whenever
x2 U , we have k(x1 )f2 (x2 )f2 k2 < . That is, |Kf2 (f1 )(x1 ) Kf2 (f1 )(x2 )| < for all
f1 such that kf1 k 1. This says that F is equicontinuous. Now we may apply Ascolis
Theorem to conclude that Kf2 is compact. 

Corollary 1.5. Kf2 is a compact operator on L2 (G).

Proof. A subset S of a complete normed space X is relatively compact if and only if S

is totally bounded. By definition, S is totally bounded if for each  > 0, there exist finitely
many s1 , . . . , sn S such that

S ni=1 { x X | kx si k <  }.

Let F be as in Lemma 1.4. By Lemma 1.4, F is a totally bounded subset of C(G). Using
Lemma 1.1, we can show that it follows that F is a totally bounded subset of L2 (G). Hence
Kf2 (f1 ) is a compact operator on L2 (G). 
 1. THE PETER-WEYL THEOREM 41

Theorem 1.6. (Peter-Weyl Theorem 1) If g0 G and g0 6= 1, then there exists an

irreducible finite-dimensional representation of G such that (g0 ) is not equal to the identity
operator.

Proof. Let V be an open neighbourhood of the identity in G such that g0 / V 2 and

V = V 1 . Fix a function C(G) such that (1) > 0, (g 1 ) = (g) 0 for all g G,
and the support of is contained in V . Let K (f ) = f , f L2 (G). We can check that
Z

K (f )(g) = f (x)(g 1 x) dx,
G

which, by properties of , is equal to K (f )(g). So K is a compact self-adjoint operator on

L2 (G). Applying the Hilbert-Schmidt Spectral Theorem, we have that L2 (G) is the Hilbert
space direct sum M
L2 (G) = Vi ,
(K )

where (K ) is the spectrum of K , and for i > 0, Vi = Ker(K i I) is finite-dimensional.

P
Applying K to , we have K () = i (K ),i >0 i i , where i C(G) Vi . Using
properties of , we can verify that K ()(1) is nonzero and K ()(g0 ) = 0. Hence there
exists i such that (g0 )i 6= i .
Next, observe that (K f )(g) = h(g), fi, f L2 (G). (Here, is the left regular
representation of G.) This can be used to see that K HomG (, ). This implies K
i I HomG (, ). We conclude that Vi is G-invariant. The restriction i of to the
finite-dimensional G-invariant subspace Vi satisfies i (g0 ) 6= IVi . We know that (i , Vi ) is
completely reducible. Hence exists at least one irreducible subrepresentation of i under
which g0 does not act as the identity on the space of the subrepresentation. 

Corollary 1.7. A compact group G is abelian if and only if all of its irreducible finite-
dimensional continuous representations are one-dimensional.

Proof. Let G0 be the derived group of G. (That is, G0 is the smallest closed subgroup
of G containing all commutators elements of the form xyx1 y 1 , x, y G.) Note that any
continuous one-dimensional representation of G is trivial on G0 . Therefore, if all irreducible
finite-dimensional representations of G are one-dimensional, we can apply Theorem 1.6 to
conclude that G0 = {1}, that is, G is abelian.
The converse follows from Schurs Lemma. 
42 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

Before stating the next theorem, we need a few basic facts about matrix coefficients of
finite-dimensional continuous representations of G.
The following definition is valid for noncompact groups and the representations dont
have to be continuous.
Definition 1.8. Let (1 , V1 ) and (2 , V2 ) be finite-dimensional representations of groups
G1 and G2 . Let V = V1 V2 . We may define a representation 1 2 of G1 G2 by setting

(1 (g1 ) 2 (g2 ))(v1 v2 ) = 1 (g1 )v1 2 (g2 )v2 , vi Vi , gi Vi ,

and then extending by linearity. This is called the external tensor product of 1 and 2 .
When G = G1 = G2 , we can restrict the external direct product representation to the
subgroup { (g, g) | g G } of G G. This representation of G is also written as 1 2 :
(1 2 )(g)(v1 v2 ) = 1 (g)v1 2 (g)v2 . It is called the internal tensor product of 1 and
2 .
Lemma 1.9. The product of a matrix coefficient of 1 and a matrix coefficient of 2 is a
matrix coefficient of 1 2 .
Definition 1.10. Let (, V ) be a representation of G and let V be the space of contin-
uous linear functionals on V . The contragredient of , is the representation of G in V
defined by ( (g))(v) = ((g 1 v)), g G, v V .
Lemma 1.11. Let (, V ) be a finite-dimensional continuous representation of G. Then
(1) is irreducible if and only if is irreducible.
(2) A matrix coefficient of is the complex conjugate of a matrix coefficient of .
Note that the last part of the next lemma is an immediate consequence of Theorem 1.6.
Lemma 1.12. Let A(G) be the span of the set of matrix coefficients of irreducible repre-
sentations of G. Then
(1) The characteristic function of G belongs to A(G).
(2) The product of two functions in A(G) belongs to A(G).
(3) The complex conjugate of a function in A(G) belongs to A(G).
(4) If g1 6= g2 G, there exists f A(G) such that f (g1 ) 6= f (g2 ).
Theorem 1.13. (Peter-Weyl Theorem 2) A(G) is dense in C(G).
Proof. Apply Lemma 1.12 and the (complex version of the) Stone-Weierstrass Theorem.

 3. FINITE-DIMENSIONAL REPRESENTATIONS OF LOCALLY PROFINITE GROUPS 43

Corollary 1.14. If U is a neighbourhood of 1 in G, there exists a finite-dimensional

representation of G whose kernel is contained in U .

2. Faithful representations

Definition 2.1. A representation (, V ) of a group G is faithful if (g) 6= I for all g G

such that g 6= 1.

Proposition 2.2. Let G be a compact group. The following are equivalent

(1) There exists a neighbourhood U of 1 in G that does not contain a nontrivial subgroup
of G.
(2) There exists a faithful finite-dimensional representation of G.
(3) G is a real Lie group.

Proof. To be added... 
Comments about faithful and almost-faithful representations of Lie groups... to be
added...

3. Finite-dimensional representations of locally profinite groups

Locally profinite groups (for example SL2 (Qp )) are not necessarily compact. However,
due to the fact that they have so many compact open subgroups, the representation theory
of locally profinite groups is heavily influenced by properties of representations of compact
groups.

Proposition 3.1. Let (, V ) be a continuous finite-dimensional representation of a lo-

cally profinite group. G. Then
(1) The kernel of is an open (normal) subgroup of G.
(2) For v V , let StabG (v) = { g G | (g)v = v }. Then StabG (v) is an open
subgroup of G.
(3) If G is profinite, then G/Ker is finite. (Hence (G) is finite.)

Remark 3.2. A representation (, V ) of a locally profinite group G having the property

that StabG (v) is open for all vectors v V is called a smooth representation.

The next lemma is false for noncompact locally profinite groups. Later in the course, we
will see that the trivial representation of SL2 (Qp ) is the only irreducible finite-dimensional
continuous (complex) representation of SL2 (Qp ).
44 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

Corollary 3.3. If G is a profinite group and K is an open compact normal subgroup of

G, there exists a finite-dimensional continuous representation of G such that Ker = K.

4. Finite-dimensional subrepresentations of the regular representation

Several of the results here are valid for finite-dimensional representations of abstract
groups.

Lemma 4.1. Let CG be the space of complex-valued functions on G. Let be the right
regular representation of G on CG . Suppose that V is a finite-dimensional G-invariant
subspace of CG . If f V , there exists A End(V ) such that f (g) = tr(A(g)), g G.

Definition 4.2. If (, V ) is a finite-dimensional representation of G, define a represen-

tation (` , End(V )) by ` (g)(A) = (g) A, A End(V ). Define T : End(V ) CG by
T (A)(g) = tr(` (g)(A)) = tr((g) A).

Lemma 4.3. T HomG (` , ).

If V and v V , then we define Av End(V ) by Av (w) = (w)v, w V . The

map v Av extends to a linear isomorphism between V V and End(V ).
The next lemma shows that the image of T consists of linear combinations of matrix
coefficients of .

Lemma 4.4. If V and v V , then T (Av )(g) = ((g)v), g G.

Lemma 4.5. Every finite-dimensional subrepresentation of (, CG ) that is equivalent to

(, V ) occurs as a subrepresentation of (, T (End(V )).

Proof. Let (, W ) be a finite-dimensional subrepresentation of (, CG ) that is equivalent

to (, V ). Then there exists a linear isomorphism C : V W such that C(g)C 1 =
(g) | W for all g G. Let f W . By Lemma 4.1, there exists A End(V ) such that
f (g) = tr(A(g)) for all g G. To complete the proof, verify that T (CAC 1 )(g) = f (g) for
g G. It follows that W T (End(V )). 

Theorem 4.6. (Burnsides Theorem) If (, V ) is an irreducible finite-dimensional rep-

resentation of a group G, then End(V ) = span{ (g) | g G }.

As the next example shows, it is easy to see that when is not irreducible, span{ (g) | g
G } is a proper subspace of End(V ).
 4. FINITE-DIMENSIONAL SUBREPRESENTATIONS OF THE REGULAR REPRESENTATION 45
!
t x
Example 4.7. Let G = R n R. Define a representation of G by (t, x) = ,
0 1
t R , x R. Clearly, span{ (t, x) | t R , x R } consists of the 2 2 matrices whose
(2, 1) entry is equal to zero.

Proof. More details to be added at a later time.

Step 1 Given v V , define v : End(V ) V by v (A) = A(v). Show that v
HomG (` , ).
Step 2 Show that if (` , W0 ) is a (nonzero) irreducible subrepresentations of (` , End(V )),
then for each nonzero A W0 , tr((g)A) 6= 0 for some g G.
Step 3 Let W 0 = { A End(V ) | tr(AB) = 0 B span{ (g) | g G } }. Prove that W 0
is a G-invariant subspace of End(V ).
Let W 0 be as above. Suppose that W 0 is nonzero. Since W 0 is G-invariant, there exists
a (nonzero) irreducible subrepresentation (` , W0 ) of (` , W 0 ). Let A W0 be nonzero.
According to the result of step 2, the function g 7 tr((g)A) is nonzero. This contradicts
the fact that A belongs to W 0 . Hence W 0 = {0}. To complete the proof, use nondegeneracy
of the bilinear form (A, B) 7 tr(AB). 

Lemma 4.8. The external tensor product is equivalent to the representation

(, End(V )) of G G defined by (g1 , g2 )(A) = (g1 )A(g2 )1 , g1 , g2 G, A End(V ).

Proof. The isomorphism between V V and End(V ) defined above belongs to HomG (
, ). 

The next two results are corollaries of Burnsides Theorem.

Corollary 4.9. If (, V ) is a finite-dimensional irreducible unitary representation of

G, then the representation of G G defined in Lemma 4.8 is irreducible.

Proof. Let T HomGG (, ). Then, for g G and A End(V ),

T ((g)A) = T ((g, 1)(A)) = (g, 1)(T (A)) = (g)T (A)

Taking A = I, we have T ((g)) = (g)T (I) for g G.

Let A End(V ). By Burnsides Theorem, since is irreducible, there exist finitely
P
many elements xi G and complex numbers ci such that A = i ci (xi ). Since T ((xi )) =
(xi )T (I) for each i, we have T (A) = A T (I).
46 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

Similarly, T ((g)) = T (I)(g) for all g G. It follows that (g)T (I) = (g)T (I) for all
g G. That is, T (I) HomG (, ). By irreducibility of , there exists C such that
T (I) = I. Therefore T (A) = A T (I) = A I = A, for A End(V ).
This shows that HomGG (, ) consists of scalar multiples of the identity operator on
End(V ). Since is unitary and irreducible, we may apply Schurs Lemma to conclude that
is irreducible. 

Remark 4.10. The corollary still holds when is nonunitary, but the proof has to be
adjusted, as the last part of the argument cannot be applied in the nonunitary setting.

Corollary 4.11. Let be an irreducible finite-dimensional representation of G. Then

T (see Definition 4.2) is one-to-one.

Proof. Let A End(V ) be such that T (A) = 0. By definition of T , tr((g)A) = 0 for

every g G. According to Burnsides Theorem, End(V ) = span{ (x) | x G }. It follows
that tr(BA) = 0 for all B End(V ). Nondegeneracy of trace forces A = 0. 

Definition 4.12. The biregular representation of GG on CG is defined by ((g1 , g2 )f )(x) =

f (g11 xg2 ). (Of course, when G is a compact group, we may restrict the biregular represen-
tation to a continuous representation of G on C(G) or on L2 (G).)

Lemma 4.13. Let be as in Lemma 4.8 and T be as in Definition 4.2. Then T

HomGG (, ). The composition of the isomorphism between V V and End(V ) with T
belongs to HomGG ( , ).

Corollary 4.14. Let be a finite-dimensional irreducible unitary representation of G.

Then the restriction of the biregular representation of G G to the subspace of CG spanned
by the matrix coefficients of is equivalent to .

Corollary 4.15. Let be a finite-dimensional irreducible unitary representation of G.

Then (, T (End(V )) is equivalent to the direct sum of dim V copies of (, V ).

Proof. Restrict the biregular representation to the subgroup {(1, g) | g G } to obtain

the right regular representation . If we examine the intertwining isomorphism between
and (, T (End(V )) as it restricts to this subgroup of G G, we find that it becomes
an equivalence between the representation of G on V V defined by (g)(v) = (g)v,
V , v V . It is easy to see that is isomorphic to the direct sum of dim V copies of
(, V ). 
 5. ORTHOGONALITY RELATIONS FOR MATRIX COEFFICIENTS 47

Remark 4.16. Once we have established the orthogonality relations for matrix coeffi-
cients of representations of compact groups, we will see that when G is compact, the direct
sum in the above corollary is an orthogonal direct sum.

Now we assume that G is compact and (, V ) is an irreducible finite-dimensional unitary

representation of G. Then T (End(V )) C(G) L2 (G).

Definition 4.17. Let L2 (G, ) be the algebraic direct sum of the irreducible subrepre-
sentations of (, L2 (G)) which are equivalent to .

Lemma 4.18. L2 (G, ) = T (End(V )) and dim L2 (G, ) = (dim V )2 and L2 (G, ) is the
subspace of C(G) spanned by the matrix coefficients of .

Proof. According to Lemma 4.5, L2 (G, ) T (End(V )). By Corollary 4.15, (, T (End(V ))
is equivalent to the direct sum of dim V copies of . In particular, each irreducible subrep-
resentation of (, T (End(V )) is equivalent to . The final statement of the lemma follows
from Lemma 4.4 and continuity of . 

Definition 4.19. The unitary dual of G is the set G

b of equivalence classes of irreducible
unitary finite-dimensional representations of G.

Remark 4.20. Later we will show that all irreducible unitary representations of G are
finite-dimensional.

Our goal is to prove that L2 (G) is the Hilbert space direct sum of the subrepresenta-
tions L2 (G, ) as ranges over G.b One part of this involves showing that if and are
inequivalent finite-dimensional irreducible representations of G, then L2 (G, ) and L2 (G, )
are orthogonal in L2 (G).

5. Orthogonality relations for matrix coefficients

In this section, G is a compact group. More will be added to this section soon.

Lemma 5.1. Let (, V ) and (, V ) be continuous representations of G. Let A B(V , V ).

Define A\ = G (g)A(g)1 dg. Then A\ HomG (, ).
R

Proposition 5.2. Let (, V ) be a finite-dimensional irreducible unitary representation

of G. Then A\ = (tr A/ dim V ) I.

Proof. By Lemma 5.1, A\ HomG (, ). By Schurs Lemma, A\ is a scalar multiple

of I. To complete the proof, evaluate the trace of A\ . 
48 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

Remark 5.3. Let (, V ) and (, V ) be equivalent finite-dimensional irreducible repre-

sentations of G. It is easy to show that the set of matrix coefficients of coincides with the
set of matrix coefficients of .

Theorem 5.4. (Schur orthogonality relations) Let (, V ) and (, V ) be finite-dimensional

irreducible unitary representations of G. Let h, i h, i be G-invariant inner products on
V and V , respectively. For v, v 0 V and u, u0 V ,

0 if 6'
Z
h(g)v, v 0 i h(g)u, u0 i dg =
G hv, ui hv 0 , u0 i if = .

Proof. Define A B(V , V ) by A(w) = hw, ui v. Show that

Z
hA (u ), v i = h(g)v, v 0 i h(g)u, u0 i dg.
\ 0 0
G

According to Proposition 5.1, A HomG (, ). When 6' , we know that A\ = 0.

\

Suppose that = . Show that tr A = hv, ui . Then apply Proposition 5.2. 

6. Decomposition of the regular representation

b let L2 (G, ) be as in Definition 4.17.
If G,

Theorem 6.1. (Peter-Weyl Theorem 3) Let G be a compact group. Then L2 (G) is the
Hilbert space direct sum of the finite-dimensional subrepresentations L2 (G, ) as ranges
over G.
b

Proof. Let and be inequivalent finite-dimensional irreducible unitary representa-

tions of G. We know (see Lemma 4.18) that L2 (G, ) (resp. L2 (G, )) is the space spanned
by the matrix coefficients of (resp. ). By the Schur orthogonality relations, we have that
L2 (G, ) and L2 (G, ) are orthogonal in L2 (G).
Take the algebraic direct sum Gb L2 (G, ). In view of the orthogonality of the spaces
L2 (G, ), it suffices to show that Gb L2 (G, ) is dense in L2 (G) (see Exercise 5.7), Chap-
ter 2). Since L2 (G, ) is the span of the matrix coefficients of , we have that Gb L2 (G, )
is the space A(G) consisting of the span of all matrix coefficients of finite-dimensional ir-
reducible representations of G. By the Peter-Weyl Theorem 2, A(G) is dense in C(G). It
follows that it is dense in L2 (G). 

Exercise 6.2. Let G be a compact Lie group. Let be a faithful finite-dimensional

representation of G. Consider the set of irreducible subrepresentations of representations of
 7. DECOMPOSITION OF UNITARY REPRESENTATIONS OF COMPACT GROUPS 49

the form 1 n , where i is equivalent to an irreducible subrepresentation of or

to the contragedient of the subrepresentation. Prove that this set contains all irreducible
finite-dimensional representations of G (up to equivalence).

Exercise 6.3. Suppose that H is a closed subgroup of a compact abelian group G.

Let : H C be a character of H (that is, a unitary one-dimensional continuous
representation). Show that extends to a character of G. (Hint: Consider the space
V = { f L2 (G) | f (hg) = (h)f (g) }. To show that V is nonzero, note that if C(G),
then f (g) = G (hg)(h1 ) dh defines an element of V . Show that there exists such that
R

f 6= 0.)

7. Decomposition of unitary representations of compact groups

Proposition 7.1. Let (, V ) be a unitary representation of a compact group G. Then

has a finite-dimensional irreducible subrepresentation.

Proof. Let h , i be a G-invariant inner product on V . Let w V . Define Tw : V

C(G) by Tw (v) = h (g)v, w i. Show that

kTw (v)k = sup |Tw (v)(g)| kvkkwk, v V.

gG

It follows that Tw B(V, C(G)) B(V, L2 (G)). In particular, Tw : V L2 (G) is a

continuous linear transformation. Note that

Tw ((x)v)(g) = h (g)(x)v, w i = Tw (v)(gx) = ((x)Tw (v))(g), x, g G, v V.

Hence Tw HomG (, ). If w 6= 0, then Tw (w)(1) = kwk2 6= 0. Hence Tw (w) is nonzero.

Since Tw (w) C(G) L2 (G), there exists an irreducible finite-dimensional representation
of G such that Tw (w) L2 (G, ). Let P be the orthogonal projection of V onto L2 (G, ).
Then P HomG ((, L2 (G)), (, L2 (G, )).
The composition P Tw belongs to HomG (, (, L2 (G, )). By an earlier result concerning
interwining of unitary representations, there exists a subrepresentation of that is equivalent
to a subrepresentation of (, L2 (G, )). Since L2 (G, ) is finite-dimensional, this shows that
(, V ) has a finite-dimensional subrepresentation. 

Corollary 7.2. A topologically irreducible unitary representation of a compact group

is finite-dimensional.
50 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

Theorem 7.3. (Peter-Weyl Theorem 3) Let (, V ) be a unitary representation of a com-

pact group G. Then V is a Hilbert space direct sum of finite-dimensional irreducible repre-
sentations of G.

Proof. Without loss of generality V is nonzero. By Proposition 7.1, (, V ) has at

least one irreducible (finite-dimensional) subrepresentation. Let be the collection of sets
of pairwise orthogonal finite-dimensional irreducible subrepresentations of V . (That is, if
S and (1 , V1 ), (2 , V2 ) belong to S, then 1 and 2 are irreducible subrepresentations
of and V1 V2 .) By Zorns Lemma, has a maximal element, say S.
To prove the theorem, it suffices to show that finite linear combinations of vectors from
subspaces in the set S are dense in V (see Exercise 5.7, Chapter 2). Let V 0 be the orthogonal
complement of the span of vectors from subspaces in S. Then V 0 is G-invariant and closed.
Assume that V 0 is nonzero. By Proposition 7.1, V 0 contains a finite-dimensional irreducible
subrepresentation, say (, W ). We can append W to S, contradicting maximality of S. It
follows that V 0 = { 0 }. 

Definition 7.4. Let (, V ) be a unitary representation of G. We say that a vector

v V is G-finite if v belongs to a finite-dimensional subrepresentation of (, V ). (The set
of G-finite vectors in V is a subspace of V .)

Corollary 7.5. Let (, V ) be a unitary representation of G. The subspace of G-finite

vectors in V is dense in V .

In a later section, we will see how to use characters and integration to construct or-
thogonal projections onto finite-dimensional subrepresentations of unitary representations of
compact groups.

8. Characters of finite-dimensional representations

Definition 8.1. Let (, V ) be a finite-dimensional representation of a group G. The

character of is the function : G C defined by (g) = tr (g), g G.

Definition 8.2. A function f : G C is a class function if f (gxg 1 ) = f (x) for all g

and x G.

Clearly the character of a finite-dimensional representation is a class function.

Proposition 8.3. Let (, V ) and (, W ) be finite-dimensional representations of a group

G. Then
 8. CHARACTERS OF FINITE-DIMENSIONAL REPRESENTATIONS 51

(1) If ' , then = .

(2) = + .
(3) = .
(4) (g) = (g 1 ), g G.
(5) If is unitary, then (g 1 ) = (g), g G.
(6) (1) = dim V .
(7) is a sum of matrix coefficients of .
(8) If is continuous, then is a continuous class function.
! !
t x t 0
Example 8.4. Let G = R nR. Define (t, x) = and (t, x) = , t R ,
0 1 0 1

x R. Then (t, x) = (t, x) = t + 1, t R , x R. The characters of and coincide,
but and are not equivalent.

Theorem 8.5. Let (, V ) and (, V ) be continuous finite-dimensional unitary represe-

nations of a compact group G. Then
R
(1) G (g) dg = dim{ v V | (g)v = v g G }.
(2) h , iL2 (G) = dim HomG (, ).
(3) If and are irreducible, then

1, if V ' V ,

h , iL2 (G) =
0, otherwise.

Proof. Let W = { v V | (g)v = v g G }. Clearly, W is a subrepresentation of

R
. Let P (v) = G (g) v dg, v V . By invariance of Haar measure on G, pi(g)P (v) = P (v)
for all v V . Hence P (V ) W . It follows from the definition of P that P (v) = v for all
v W . Hence W = P (V ). Moreover, P 2 (v) = P (P (v)) = P (v), since P (v) W . So P is a
projection of V onto W . (Indeed, we can see that P is the orthogonal projection of V onto
W .) Hence the trace of P is equal to dim W . Observe that the trace of P is also equal to
R
(g) dg.
G
Define a representation of G on HomC (V , V ) by (g)A = (g)A(g 1 ). Note that

HomG (, ) = { A HomC (V , V ) | (g)A = A }

Given V and v V , define Av HomC (V , V ) by Av (w) = (w)v, w V . Note

that
A (g)(g)v = (g)Av (g 1 ), g G.
52 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

Extending this to an equivalence between and , we have

= = = .

Applying part (1) to the representation (, HomC (V , V )), we see that (2) holds.
Part (3) is a consequence of (2). It can also be obtained as a consequence of the Schur
orthogonality relations for matrix coefficients(Theorem 5.4), since each character is a sum of
matrix coefficients. 

Proposition 8.6. Let (, V ) be a finite-dimensional unitary representation of a compact

group G. Then is determined up to equivalence by .

Proof. We know that is completely reducible. Let { j } be the set of irreducible

P
constituents of . Then there exist positive integers nj such that = j nj j . By
Theorem 8.5, we have nj = hj , iL2 (G) . Consequently, if (, W ) is a finite-dimensional
unitary representation of G such that = , we can see that, up to equivalence, the
representations j are the irreducible constituents of , and the multiplicity of j in is
equal to hj , iL2 (G) = nj . Hence ' . 

Corollary 8.7. Let (, V ) be a finite-dimensional unitary representation of a compact

group G. Then h , iL2 (G) = 1 if and only if is irreducible.

Theorem 8.8. The span of the characters of the irreducible unitary representations of a
compact group G is dense in the space of continuous class functions on G.

Proof. Let C(G) be a class function. Fix  > 0. By the Peter-Weyl Theorem 2,
there exists f A(G) such that k f k < . For x G, let (x) = G f (gxg 1 ) dg. Then
R
Z
k k = sup |(g) (g)| sup |(xgx1 ) f (xgx1 )| dg k f k < .
gG gG

To complete the proof, it suffices to show that is a linear combination of characters

of irreducible representations of G. Since f A(G), there exist finitely many irreducible
unitary representations (j , Vj ) of G, together with vj Vj , j Vj , such that f (g) =
P
j j (j (g)vj ), g G. Observe that, thinking of j (x) as an element of End(Vj ), we
have j (x)\ = G j (gxg 1 ) dg HomG (j , j ) (see Lemma 5.1). Since j is irreducible, by
R

Proposition 5.2, j (x)\ = (tr pij (x)/ dim Vj )IVj . Therefore

XZ X
(x) = j (j (gxg 1 )vj ) dg = j (j (x)\ vj ) = (j (x) / dim Vj )j (vj ).
j G j
 9. CONVOLUTION, CHARACTERS AND ORTHOGONAL PROJECTIONS 53

That is, is a linear combination of the characters j . 

9. Convolution, characters and orthogonal projections

Let f C(G) and let (, V ) be a unitary representation of G, with h , i a G-invariant

R
inner product on V . Define (f ) : V V by (f )v = G f (g)(g)v dg, v V . Recall
that if f1 , f2 G, the convolution f1 f2 belongs to C(G) and is defined by (f1 f2 )(x) =
f (g)f2 (g 1 x) dg, x G. For f C(G), define fe C(G) by fe(g) = f (g 1 ), g G.
R
G 1

Lemma 9.1. Let (, V ) be a unitary representation of G.

(1) If f C(G), then (f ) B(V ).
(2) If f1 , f2 C(G), then (f1 f2 ) = (f1 )(f2 ).
(3) If v, w V and f C(G), then h (f )v, w i = h v, (fe)w i.
(4) If f C(G) and g G, then (g)(f ) = (Lg f ). (Here, (Lg f )(x) = f (g 1 x),
x G.)

Exercise 9.2. Prove Lemma 9.1.

Exercise 9.3. If x G and f C(G), define (x)f by ( (x)f )(g) = f (x1 gx), g G.
Show that (x)(f ) = ( (x)f )(x).

Lemma 9.4. Let (, V ) be a unitary representation of G. Let f C(G).

(1) If f is a class function, then (f ) HomG (, ).
(2) If f = fe and f f = f , then (f ) is an orthogonal projection.

Proof. The first statement follows from Exercise 9.3 and Lemma 9.1(1).
For the second statement, f f = f implies (f ) is a projection. Then, from fe = f
and Lemma 9.1(3), we have that (f ) is self-adjoint. Self-adjoint projections are orthogonal
projections. 

Corollary 9.5. If f C(G) is a class function such that f f = f and fe = f , then

(f ) is an orthogonal projection of V onto a subrepresentation of (, V ). Furthermore, if f1
and f2 are two such functions and f1 f2 = 0, then (f1 )V and (f2 )V are orthogonal.

Proposition 9.6. Let (, V ) and (, W ) be irreducible unitary representations of G.

Then
(dim V )1 , if '

=
0, otherwise.
54 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

Proof. Let h , i be a G-invariant inner product on W . Let { v1 , . . . , vn } be an orthonor-

mal basis of W . For g and x G,
n
X n
X
1 1 1
(g x) = tr((g )(x)) = hvj , (g )vi ih(x)vi , vj i = h(g)vi , vj ih(x)vi , vj i
i,j=1 i,j=1

Here we have used the fact that the matrix of (g) with respect to the orthonormal basis is
a unitary matrix.
Z m Z
X 
1
( )(x) = (g) tr((g )(x)) dg = (g)h(g)vi , vj i dg h(x)vi , vj i.
G i,j=1 G

Because is a sum of matrix coefficients of , if is not equivalent to , by Theorem 5.4,

the L2 (G) inner product of and the matrix coefficient g 7 byh(g)vi , vj i is equal to zero.
Suppose that ' . Then = , so we may assume that = and replace
(g) and (x) by (g) and (x) above. We have (g) = n`=1 h(g)v` , v` i. By the Schur
P

Orthogonality Relations (Theorem 5.4),

Z n Z
X n
X
(g)h(g)vi , vj i dg = h(g)v` , v` ih(g)vi , vj i = hv` , vi ihv` , vj i/m,
G `=1 G `=1

which equals 1 if i = j and 0 if i 6= j. It follows that

n
X
( )(x) = 1/n(h(x)vj , vj i) = (dim V )1 (x).
j=1

Exercise 9.7. Let (, V ) and (, W ) be irreducible unitary representations of G.

(1) Suppose that f and f are matrix coefficients of and , respectively and 6' .
Prove that f f = 0.
(2) Let h , i be a G-invariant inner product on the space of . For vectors u, v V ,
set fu,v (g) = h(g)u, vi, g G. Let u, v, u0 and v 0 be vectors in V . Prove that
fu,v fu0 ,v0 = (hu, v 0 i/ dim V )fu0 ,v .

Definition 9.8. Let (, V ) and (, W ) be unitary representations of G. Assume that

is irreducible. (Then is finite-dimensional.) Let V be the largest subrepresentation of
(, V ) having the property that every irreducible subrepresentation of (, V ) is equivalent
to . The space V is called the -isotypic (or -isotypical ) subspace of .
 9. CONVOLUTION, CHARACTERS AND ORTHOGONAL PROJECTIONS 55

According to the Peter-Weyl Theorem 3 (Theorem 7.3), (, V ) is equivalent to the Hilbert

space direct sum ( | V , V ) (described in Exercise 5.7 of Chapter 2), where ranges
over the irreducible unitary representations of G such that V is nonzero.

Proposition 9.9. Let (, V ) be a unitary representation of G. Let (, W ) be an irre-

ducible unitary representation of G such that V is nonzero. Let e = dim(W ) . Then
R
(e ) is the orthogonal projection of V onto V . That is, v 7 (dim W ) G (g)(g)v dg is
the orthogonal projection of V onto V .

Proof. By Proposition 8.3, e is a class function and ee = e . According to Proposi-

tion 9.6, we have e e = e . Hence (see Corollary 9.5), (e ) HomG (, ) and is an
orthogonal projection.
We need to verify that (e )V = V . Let ( 0 , V 0 ) be an irreducible subrepresentation of
(, V ) that is equivalent to . Let { v1 , . . . , vn } be an orthonormal basis for V 0 . Then
Z Xn Z 
0 0
(e )vi = (dim W ) (g) (g)vi dg = (dim W ) (g)h (g)vi , vj i dg vj
G j=1 G
n Z
X 
= (dim W ) 0 (g)h 0 (g)vi , vj i dg vj
j=1 G
n Z
X 
0
= (dim W ) h 0 (g)v` , v` ih (g)vi , vj i dg vj
j,`=1 G

n 
X 
= (dim W ) hvi , v` ihvj , v` i/ dim W vj = vi .
j,`=1

(Above, we have used Theorem 5.4). Because (e ) is an orthogonal projection, (e )vi = vi

is equivalent to vi (e )V . Hence ( 0 , V 0 ) (e )V . That is, V (e )V .
To complete the proof, it suffices to show that any irreducible subrepresentation of
(, (e )V ) is equivalent to (, W ). Equivalently, its enough to show that an irreducible
subrepresentation ( 0 , V 0 ) of (, V ) that is not equivalent to belongs to the kernel of (e ).
As above, let { v1 , . . . , vn } be an orthonormal basis for V 0 . Then
Z Xn Z 
0 0
(e )vi = (dim W ) (g) (g)vi dg = (dim W ) (g)h (g)vi , vj i dg vj .
G j=1 G

Since is a sum of matrix coefficients of and 6' 0 , it follows from Theorem 5.4 that
(g)h 0 (g)vi , vj i = 0 for all i and j. That is, (e )vi = 0 for all i.
R
G

56 3. REPRESENTATIONS OF COMPACT GROUPS: THE GENERAL THEORY

Remark 9.10. Let f be a matrix coefficient of . We can show that f e = e f = f

(see Exercise 9.7). It follows that (f ) = (e f ) = (e )(f ). So (f )V V

Example 9.11. Consider the right regular representation (, L2 (G)). Let (, W ) be an

irreducible unitary representation of G. For f L2 (G) and x G,
Z Z
1 1 1
(f e )(x) = (dim W ) f (g) (g x) dg = (dim W ) f (xg) ] (g 1 ) dg
ZG Z G

= (dim W )1 (g) ((g)f )(x) dg = e (g)((g)f )(x) dg = ((e )f )(x)

G G

That is, f e = (e )f . So in this example, the orthogonal projection onto the -isotypic
subspace is given by convolution with (dim W )1 .

10. Induced representations of compact groups

Definition 10.1. Let W be a topological vector space and let C(G, W ) be the space of
continuous functions from G to W . If K is a compact subset of G and U is an open subset of
W , define V(K, U ) = { f C(G, W ) |f (K) U }. The compact-open topology on C(G, W )
is the topology for which the sets V(K, U ) form a subbasis.

Lemma 10.2. C(G, W ) is a topological vector space and the map G C(G, W )
C(G, W ) given by (g, f )
7 Lg f is continuous.

Definition 10.3. Let (, W ) be a continuous representation of a closed subgroup H of

G. Let iGH (W ) = { f C(G, W ) | f (gh) = (h)1 f (g) g G, h H }. For f iGH (W )
1
and g G, set (iG H (g)f )(x) =
G G
f (g x). The representation (iH , iH (W )) is called the
representation of G induced from (, W ).

Lemma 10.4. When iG G G

H (W ) is given the compact-open topology, (iH , iH (W )) is a con-
tinuous representation of G.

Theorem 10.5. (Frobenius Reciprocity) Let H be a closed subgroup of G. Let (, V )

and (, W ) be continous representations of G and H, respectively. Then HomG (, iG H ) '
G G
HomH (resH , ). (Here, resH is the representation of H obtained by restriction of to H.)

Proof. (Sketch) Let A HomG (, iG G

H ). Define : iH (W ) W by (f ) = f (1). Check
that A 7 A is a map from HomG (, iG G
H ) to HomH (resH , ).
 10. INDUCED REPRESENTATIONS OF COMPACT GROUPS 57

0 0
Next, suppose that B HomH (resG
H , ). Define B : V C(G, W ) by B (v)(g) =
B((g)1 v), v V , g G. Then

B 0 (v)(gh) = B((gh)1 v) = B((h)1 (g)1 v) = (h)1 B 0 (v)(g), g G, h H.

Hence B 0 (v) iG
H (W ). Next, note that

B 0 ((g)v)(x) = B((x1 g)v) = B 0 (v)(g 1 x) = ((iG 0

H )(g)B (v))(x), g, x G, v V.

Thus the map B 7 B 0 goes from HomH (resG G

H , ) to HomG (, iH ).
To complete the proof, check that the maps A 7 A and B 7 B 0 are inverses of each
other. 
For another approach to induction of representations, assume that (, W ) is a unitary
representation of a closed subgroup H of G. Let L2 (G, W ) be the completion of the space
R
1/2
C(G, W ) with respect to the norm kf k = G kf (g)k2W dg . Then, for the induced rep-
resentation, we take the left regular representation on the space { f L2 (G, W ) | f (gh) =
(h)1 f (g) g G, h H }. This is the approach taken in [R].
 CHAPTER 4

Representations of compact Lie groups

Throughout this chapter, G is a compact connected Lie group. The representations

considered are finite-dimensional and continuous. Although outlines of a few proofs are
included, most of the results are stated without proof. We focus on the parametrization
of the irreducible representations in terms of dominant weights and the characters of the
irreducible representations.
The Borel-Weil Theorem (and its generalization the Borel-Weil-Bott Theorem), which
provides concrete models for the irreducible representations of compact Lie groups, is not
discussed. The text [Se] includes a section on the Borel-Weil Theorem.
There are many references on representations of compact Lie groups, for example, [BD],
[Ha], [Se] and [Si].
The first section contains a summary of results about maximal tori in compact Lie groups
and the statement of Weyls Integral Formula.
In Section 2, we define weights of finite-dimensional representations and list a few of their
properties.
At the beginning of Section 3, we recall some definitions related to Lie algebras and
describe how to attach a (real) representation of the Lie algebra g of G to a finite-dimensional
(continuous) representation of G. Then we complexify to produce a complex representation
of the complexification gC of g.
A choice of maximal torus T in G determines a root space decomposition of the complex
Lie algebra gC , which is described in Section 4. This decomposition has the form gC =
L
tC g , where tC is the complexification of the Lie algebra of T , is a finite collection
of linear functionals, called roots, and each g is a subspace of gC determined by the root
. Certain properties of the representation of gC determined by a given finite-dimensional
representation of G can be used to obtain information about the weights of the representation
of G. In particular, each irreducible finite-dimensional representation has a uniquely defined
highest weight.

59
60 4. REPRESENTATIONS OF COMPACT LIE GROUPS

In Section 5, we comment on how Weyls integral formula, together with information

about weights of representations, is applied to obtain a formula for the character of an
irreducible finite-dimensional representation of G. In addition, we discuss how to use the
Peter-Weyl Theorem to see that every dominant weight occurs as the highest weight of
some irreducible finite-dimensional representation.

1. Maximal tori and Weyls Integral Formula

Proofs of results stated in this section can be found in Chapter IV of [BD] (and various
other references).

Definition 1.1. A Lie group is a compact torus if it is isomorphic to Rn /Zn for some
integer n > 0. (Here, n is the dimension of the torus.) A subgroup T of a Lie group is
a maximal torus in a compact Lie group G if T is a compact torus and there is no other
compact torus T 0 with T ( T 0 G.

Since compact tori are connected, if T and T 0 are compact tori in G with T ( T 0 , then
the dimension of T is strictly less than the dimension of T 0 . Hence maximal tori exist. A
maximal torus in G is a maximal connected abelian subgroup of G.

Definition 1.2. Let T be a maximal torus in G. Let NG (T ) be the normalizer of T in

G. That is NG (T ) = { g G | gtg 1 T t T }. The group W = NG (T )/T is the Weyl
group of T .

Theorem 1.3. (1) Any two maximal tori in G are conjugate.

(2) Every element of G is contained in a maximal torus.
(3) The Weyl group of a maximal torus is finite.

Example 1.4. For the group G = U (n) = { g GLn (C) | t gg = In } (resp. G0 =

SU (n) = U (n) SLn (C)), the group of diagonal matrices in G (resp. G0 ) is a maximal torus.
In both cases, W is isomorphic to the symmetric group on n letters.

Because any two maximal tori in G are conjugate, we can see that if T1 and T2 are
maximal tori in G, then the Weyl groups of T1 and T2 are isomorphic. For this reason, we
refer to the Weyl group of a maximal torus in G as the Weyl group of G.
For the rest of the section, we fix a maximal torus T in G.

Definition 1.5. let g and t be the Lie algebras of g and t, respectively. For t T
and X g, define Ad(t)(X) = tXt1 . Note that Ad(t)(X) = X for all X t and t T .
 1. MAXIMAL TORI AND WEYLS INTEGRAL FORMULA 61

Hence we may view Ad(t) (and hence also Ad(t1 ) I) as an element of EndR (g/t). Let
D(t) = | det(Ad(t1 ) I)g/t |, t T . (In fact, the absolute value is not needed here, as the
endomorphism Ad(t1 ) I of g/t has no nonzero real eigenvalues.)

If z1 , . . . zn C, let d(z1 , . . . , zn ) be the diagonal matrix whose (j, j) entry is equal to zj ,

1 j n.

Example 1.6. For G = SU (3) = { g GL3 (C) | t gg = I3 }, the set T of diagonal

matrices in G, that is, the matrices of the form d(e2it1 , e2it2 , e2i(t1 +t2 ) ), where t1 , t2 R,
is a maximal torus in G. The Lie algebra of G is g = { X M3 (C) | t X = X, tr(X) = 0 }
and t is the diagonal matrices in M3 (C) with purely imaginary entries and trace zero. So
g/t can be identified with the (real) subspace of g consisting of those matrices of the form

0 z1 z2
z1 0 z3 , where z1 , z2 , z3 C.

z2 z3 0

The following integral formula is an important tool in the proof of Weyls character
formula.

Theorem 1.7. (Weyls Integral Formula) Let f C(G). Then

Z Z  Z 
1
|W | f (g) dg = D(t) f (gtg ) dg dt.
G T G

If f is a class function, then

Z Z
|W | f (g) dg = D(t) f (t) dt.
G T

Lemma 1.8. Let t1 , t2 T . Then t1 and t2 are conjugate in G if and only if t1 and t2
are conjugate in NG (T ).

Definition 1.9. The group NG (T ) acts on the space C(T ) (continuous complex-valued
functions on T ) as follows: Given g NG (T ) and f C(T ), set g f (t) = f (g 1 tg), t T .
When g T , g f = f , so the action of NG (T ) factors to an action of W on C(T ). Let C(T )W
be the W -invariant functions in C(T ).

Definition 1.10. The representation ring R(G) of G be the set of class functions of the
P
form i ni i , where ni Z and i is the character of a finite-dimensional representation of
G. Let R(T )W be the W -invariant functions in R(T ).
62 4. REPRESENTATIONS OF COMPACT LIE GROUPS

Clearly, the restriction of a continuous class function on G to T belongs to C(T )W .

Lemma 1.11. (1) The restriction map f 7 f | T from C(G) to C(T )W is onto.
(2) For f R(G), the restriction f | T belongs to R(T )W .
(3) The map f 7 f | T from R(G) to R(T )W is one-to-one.

Remark 1.12. The map from R(G) to R(T )W is actually an isomorphism.

2. Weights of finite-dimensional representations

Let (, V ) be a finite-dimensional representation of a compact Lie group G. Let T be

a maximal torus in G. The restriction resG T of to T is a direct sum of one-dimensional
representations (these are also called characters) of T .
Let g and t be the Lie algebras of G and T , respectively. Let gC = g ig and tC = t it
be the complexifications of g and t, respectively. Set tR = it.
The map X 7 e2iX from tR to T is a surjective homomorphism from Rd onto T , where
d is the dimension of T . (Here, e denotes the usual matrix exponential map.) Let K be the
kernel of this homomorphism. Then K ' Zd and T ' Rd /K.

Example 2.1. Continuing with Example 1.6, recall that t = { d(it1 , it2 , i(t1 +t2 )) | t1 , t2
R. So we have tR = { d(t1 , t2 , t1 t2 )) | t1 , t2 R. For convenience, we set Xt1 ,t2 =
d(t1 , t2 , t1 t2 ), t1 , t2 R. Here, K = { Xt1 ,t2 | t1 , t2 Z }. The complexfication
gC is equal to the Lie algebra sl3 (C) of 3 3 matrices whose trace is zero. Note that
sl3 (C) = sl3 (R) isl3 (R), so sl3 (C) is also the complexification of the Lie algebra sl3 (R).
That is, g = su(3) and sl3 (R) are real forms of sl3 (C).

Definition 2.2. The weight lattice Y is defined to be { tR | (X) Z X K }.

Given Y, define e (X) = e2i(X) , X tR .

For Y, since e (X) = 1 for all X K, we may view e as a one-dimensional

representation of the abelian group tR /K. Since T ' tR /K, we also view e as a one-
dimensional representation of T .

Definition 2.3. If (, V ) is a finite-dimensional unitary representation of G and Y,

set V = { v V |(e2iX ) = e (X)v | X tR }.

We have V = Y V . Since V is finite-dimensional, V is nonzero for finitely many .

Definition 2.4. The set Y () = { Y | V 6= {0} } is called the set of weights of .

If Y (), let m = dim(V ). We refer to m as the multiplicity of in .
3. RELATIONS BETWEEN REPRESENTATIONS OF COMPACT LIE GROUPS AND COMPLEX LIE ALGEBRAS
63

If g G, then conjugation by g preserves g and gC . If g NG (T ), then conjugation by

g preserves t, tC and tR . If g belongs to T , the restriction of conjugation by g to tC is the
identity map that is g commutes with every matrix in tC . In Definition 1.9, we defined
actions of NG (T ) and W = NG (T )/T on C(T ).
If Y, and g NG (T ), then, viewing e as an element of C(T ), we have already
defined g e (see Definition 1.9). It follows from the definitions that g e (X) = e (g 1 Xg),
X tR .

Definition 2.5. If Y and w = gT W , we set (w)(X) = (g 1 Xg), X tR .

1
Because g e2iX g 1 = e2igXg for X g and g G, it follows that if g NG (T ), then
gKg 1 = K.

Lemma 2.6. Let w W and Y, then w Y and w e = ew for w W .

Lemma 2.7. If w W and Y (), then m = mw .

Proof. Choose g G such that w = gT . For t = e2iX T (X tR ) and v V ,

(t)(g)v = (g)((g 1 tg)v = (g)e (g 1 Xg)v = ew (X)(g)v.

Hence (g)V Vw . Similarly, (g 1 )Vw V . Thus m = mw . 

Example 2.8. In Example 1.6, SU (3) is realized as a closed subgroup of GL3 (C). We
can define a 3-dimensional faithful representation of SU (3) by (g) = g, g SU (3). We
call this the self-representation. Define 1 (Xt1 ,t2 ) = t1 and 2 (Xt1 ,t2 ) = t2 , for t1 , t2 R.
Since
(e2iXt1 ,t2 ) = e2iXt1 ,t2 = d(e2it1 , e2it2 , e2i(t1 t2 ) ),
we see that Y () = { 1 , 2 , 1 2 }.

3. Relations between representations of compact Lie groups and complex Lie

algebras

If G is a matrix Lie group occurring as a closed subgroup of GLn (C), then the Lie algebra
g of G is given by g = { X Mn (C) | etX G t R }. For X, Y Mn (C), the Lie algebra
bracket [X, Y ] is defined by [X, Y ] = XY Y X. If X,Y g, then [X, Y ] g.

Definition 3.1. Let g be a Lie algebra.

(1) g is abelian if [X, Y ] = 0 for all X and Y g.
64 4. REPRESENTATIONS OF COMPACT LIE GROUPS

(2) A subspace h of g is a Lie subalgebra of g if [X, Y ] h for all X and Y h.

(3) A subspace I of g is an ideal if [X, Y ] I whenever X g and Y I.
(4) The Lie algebra g is simple if g is nonabelian and {0} and g are the only ideals in
g.
(5) The Lie algebra g is semisimple if g does not contain any nonzero abelian ideals.
(6) The Lie algebra g is reductive if g is the direct sum of a semisimple subalgebra and
an abelian subalgebra. (Equivalently, g = [g, g] z, where z is the centre of g.)
(7) If X g, adX : g g is defined by adX(Y ) = [X, Y ], Y g.
(8) The Killing form is defined by K(X, Y ) = tr(adXadY ), X, Y g. It is a symmetric
bilinear form on g.

Lemma 3.2. (1) K([X, Y ], Z) = K(X, [Y, Z]) for all X, Y , Z g.

(2) K(gXg , gY g 1 ) = K(X, Y ) for all X, Y g and g G.
1

Lemma 3.3. (Cartans criterion) A Lie algebra g is semisimple if and only if the Killing
form is nondegenerate.

Lemma 3.4. If G is a compact Lie group, then

(1) The Killing form is negative semidefinite.
(2) The Lie algebra g of G is reductive.

Proposition 3.5. Let G and H be matrix Lie groups, with Lie algebras g and h, re-
spectively. Let : G H be a homomorphism of Lie groups. Then there exists a unique
R-linear map d : g h such that (eX ) = ed(X) for all X g. The map d has the
following additional properties:
(1) d(gXg 1 ) = (g)d(X)(g)1 , g G, X g.
(2) d([X, Y ]) = [d(X), d(Y )], X, Y g.
(3) d(X) = d/dt(etX ) |t=0 , X g.

Let G be a compact Lie group with Lie algebra g. Let (, V ) be a finite-dimensional

unitary representation of G. Then g 7 (g) is a continuous homomorphism from G to
U (n).

Proposition 3.6. If G and H are Lie groups and : G H is a continuous homo-

morphism, then is a homomorphism of Lie groups (that is, is smooth).

Applying the above propositions to the homomorphism g 7 (g), we have an R-linear

map d : g u(n) such that (eX ) = ed(X) for all X g. Furthermore, d([X, Y ]) =
 4. HIGHEST WEIGHT THEOREM 65

[d(X), d(Y )] for X, Y g. That is, d is a real representation of the Lie algebra g. It is
conventional to use the notation , rather than d, for the representation of g determined by
the group representation . Since gC = g ig, we can extend to a complex representation
of the complex semisimple Lie algebra g. The notation will be used for this extension.
As discussed in the next section, properties of the representation of gC can be used to
analyze the weights Y ().

Remark 3.7. If Y() and v V , then (X)v = (X)v, X tR .

Example 3.8. Let be the self-representation of SU (3). Then the representation of

sl3 (C) determined by satisfies (X) = X, X sl3 (C).

4. Highest weight theorem

For simplicity, we assume that G is a semisimple compact Lie group. That is, we assume
that the centre of G is finite. In this case, the Killing form (see Definition 3.1 and Lemma 3.3)
is nondegenerate.
Let , T , Y, K, etc, be as in previous sections.
We define an inner product h , i on gC by h X, Y i = K(X, Y ), X, Y gC . If X tR ,
adX is self-adjoint with respect to this inner product. So { adX | X tR } is a space of
commuting self-adjoint operators on gC . Therefore the space gC has a decomposition into
simultaneous eigenspaces.

Definition 4.1. A root is a nonzero tR such that there exists a nonzero Y gC

with adX(Y ) = (X)Y for all X tR . Let be the set of roots. If , the root space
corresponding is

g := { Y gC | adX(Y ) = (X)Y X tR }.

There exist a unique H tR , called a root vector for , such that (X) = K(H , X) for all
X tR .

Proposition 4.2. (1) If , then dim g = 1.

(2) If , and + , then [g , g ] is a nonzero subset of g+ .
(3) If + and + / , then [g , g ] = {0}.
(4) If , then .
(5) SpanR () = tR
(6) SpanR { H | } = tR .
66 4. REPRESENTATIONS OF COMPACT LIE GROUPS

(7) There exist elements X g , , such that [X , X ] = 2H /K(H , H ).

L
(8) gC = tC g .
P
(9) If X, Y tR , then K(X, Y ) = (X)(Y ).

Example 4.3. Returning to our basic example G = SU (3), gC = sl3 (C), and tR =
{ Xt1 ,t2 | t1 , t2 R }, we can easily describe the roots, root vectors and root spaces. We can
check that K(Xt1 ,t2 , Xt1 ,t2 ) is 6 times the sum of the squares of the diagonal entries of Xt1 ,t2 .
This can be rewritten as K(Xt1 ,t2 , Xt1 ,t2 ) = 12(t21 + t1 t2 + t22 ). Let 1 (Xt1 ,t2 ) = t1 t2 and
2 (Xt1 ,t2 ) = t1 + 2t2 . Then

= { 1 , 2 , (1 + 2 ) }.

The root vectors for 1 and 2 are as follows: H1 = d(1, 1, 0)/6 and H2 = d(0, 1, 1)/6.
The matrices
0 1 0 0 0 0 0 0 1
0 0 0 , 0 0 1 , 0 0 0 ,

0 0 0 0 0 0 0 0 0
belong to the root spaces g1 , g2 and g3 , respectively. Their transposes belong to g1 ,
g2 and g3 , respectively.

Let X0 tR be such that (X0 ) 6= 0 for all . Let + = { | (X0 ) > 0 } and
= { | (X0 ) < 0 }.

Definition 4.4. A root is said to be simple if + and 6= + for roots ,

+ . Let be the set of simple roots.

Example 4.5. For G = SU (3), let X0 = d(1, 0, 1). Then + = { 1 , 2 , 1 + 2 } and

= { 1 , 2 }.

Proposition 4.6. (1) If , are distinct then K(H , H ) < 0.

(2) SpanR () = tR .
(3) is linearly independent.
(4) If + , then = i ni i , where each ni is a nonnegative integer.
P

Lemma 4.7. Y.

Definition 4.8. A weight Y is dominant if (Hi ) 0 i . Let Y d be the set

of dominant weights in Y.
 4. HIGHEST WEIGHT THEOREM 67

Example 4.9. For G = SU (3), we have tR = SpanR { 1 , 2 }, Y = { k1 + `2 | k, `

Z }, and Y d = { k1 + `2 | k/2 ` 2k, k, ` Z }. Note that 1 + 2 is the only dominant
root. Among the weights 1 , 2 and 1 2 of the self-representation of SU (3), only 1 is
dominant.

Definition 4.10. Let 1 , 2 Y. We say that 1 is higher than 2 (or 2 is lower than
P
1 ) if 1 2 = i ci i , where ci 0 for all i. The notation 1 < 2 or 2 4 1 is used.
This defines a partial ordering on Y.

Example 4.11. The weights of the self-representation of SU (3) satisfy 1 2 = 1 and

2 (1 2 ) = 1 + 2 2 = 2 , so 1 < 2 < 1 2 .

If , let Y g . For example, we could take Y = X , where X is as in

Proposition 4.2. The next result describes the images of the weight spaces V , Y (),
under the operators (Y ). In particular, if V is not a subspace of the kernel of (Y ),
then + Y (). This is used to obtain relations between various weights in Y ().

Proposition 4.12. If , let Y g and Y (), then (Y )V = {0} if

+/ Y () and (Y )V V+ if + Y ().

Proof. Let Y (), v V , X tR , and Y g . Then t := e2iX belongs to

T and

(t)(Y )v = (t)(Y )(t)1 (t)v = (t)(Y )(t)1 e (X)v = e (X)(tY t1 )v.

Here, we have used the relation from Proposition 3.5(1). Next, since t = e2iX , we have

= ead 2iX (Y ) =
X
1
tY t (ad 2iX)n (Y )/n!,
n=0

which, since ad X(Y ) = (X)Y , can be rewritten as tY t1 = e2i(X) Y = e (X)Y . Hence

(t)(Y )v = e (X)e (X)(Y )v = e+ (X)(Y )v.

That is, the vector (Y )v belongs to V+ . 

Theorem 4.13. Let (, V ) be an irreducible unitary representation of G. Then there

exists a unique dominant weight Y () that is maximal among all weights in Y ().
Moreover,
(1) If i , then + i
/ Y ().
68 4. REPRESENTATIONS OF COMPACT LIE GROUPS
P
(2) Every weight in Y () has the form i ni i , with ni 0.
(3) Every weight in Y () belongs to the convex hull of the images of under W .
(4) m = 1.

The maximal weight is called the highest weight of .

5. The Weyl character formula

 CHAPTER 5

Representations of reductive groups

Let G be a connected reductive linear algebraic group defined over a field F and let
G = G(F ) be the F -rational points of G. We will consider three cases:

F = R: In this case, G is a also a Lie group.

F is a nonarchimedean local field of characteristic zero: F is a finite extension of
the p-adic numbers Qp , where p is a prime. In this case, G is a locally profinite
group.
F = Fq , a finite field. We say that G is a finite group of Lie type.

1. The Iwasawa decomposition

Let F = R or Qp and let G = SL2 (F ). Let

( ! ) ( ! )
a 0 1 x
A= |aF N= |xF .
0 a1 0 1
The group B = A n N is called a Borel subgroup of G. The subgroup A is a maximal F -split
torus in G. If F = R, set
( ! )
cos t sin t
K = SO2 (R) = { g G | t g g = I } = |tR .
sin t cos t
The subgroup K is a compact maximal torus in SL2 (R), as well as a maximal compact
subgroup of G. If F = Qp , let K = SL2 (Zp ) be the subgroup of G consisting of matrices
with entries in Zp . The subgroup K is a maximal compact (open) subgroup of SL2 (Qp ).

Lemma 1.1. (Iwasawa decomposition) G = KB = BK.

!
a b
Proof. Let g = , with ad bc = 1.
c d
69
70 5. REPRESENTATIONS OF REDUCTIVE GROUPS

Suppose that F = R. Given z C, define g z = (az + b)/(cz + d). If the imaginary part
=(z) of z is positive, then =(g z) > 0. Note that, for a R and x R,
! !
1 x a 0
1
i = x + a2 i,
0 1 0 a
so B i is the full upper half plane (complex numbers having positive imaginary part).
Hence, given g G, there exists g 0 B such that g i = g 0 i. Since g 01 g i = i and
K = { h G | h i = i }, we see that g BK.
Suppose that F = Qp . Let g be as above. If c = 0, then g B. If |c|p > |a|p , then
! !
0 1 c d
g= .
1 0 a b
!
0 1
Since belongs to K and | c|p > |a|p , it suffices to prove that g KB whenever
1 0
|a|p |c|p > 0. When a 6= 0,
! !
1 0 a b
g= ,
a1 c 1 0 d a1 bc
which, when |a|p |c|p , is clearly in KB. 

The Lie algebra of SL2 (R) also has an Iwasawa decomposition: g = sl2 (R) = k a n,
where
( ! ) ( ! ) ( ! )
0 b a 0 0 x
k= |bR , a= |aR and n = |xR
b 0 0 a 0 0
are the Lie algebras of K, A and N , respectively.
If G is a matrix Lie group (closed subgroup of GLn (C)), then G is stable under the
involution g 7 t g1 and the subgroup K of fixed points of the involution is a maximal
compact subgroup of G. Every maximal compact subgroup of G is conjugate to K. In this
setting, the Iwasawa decomposition of G takes the form G = P0 K = KP0 , where P0 is a
minimal parabolic subgroup of G. (Parabolic subgroups will be discussed in a later section.)
In general, a reductive p-adic group contains a finite number of conjugacy classes of
maximal compact subgroups. If G is a connected reductive p-adic group, there exists a
maximal compact (open) subgroup K of G such that G = P0 K = KP0 for any minimal
parabolic subgroup of G. Here, K is referred to as a good maximal compact subgroup of
 2. PRINCIPAL SERIES REPRESENTATIONS OF SL2 (F ) 71

G. (Note: The adjective connected refers to the underlying algebraic group. In the p-adic
topology, as discussed earlier in the course, the topological space G is totally disconnected.)
It is worth noting that K is not itself a reductive p-adic group. That is, K is not the F -
rational points of a reductive linear algebraic group defined over F . It is simply a particular
compact open subgroup of G. (This contrasts with the Lie group setting, where K is a
compact Lie group.)

2. Principal series representations of SL2 (F )

If G is a locally compact group, a quasicharacter of G is a continuous one-dimensional

representation of G. A unitary quasicharacter is referred to as a character.
Let G = SL2 (F ),! F = R or Qp , K, A, N , and B be as in Section 1. Given a F ,
a 0
let d(a) = . Let be a quasicharacter of F , which we may also view as a
0 a1
quasicharacter of A via the isomorphism given by a 7 d(a). The notation | |F will be used
for the given absolute value on F (that is, the usual one when F = R and the p-adic
absolute value when F = Qp ). Let

C() = { f C(G) | f (d(a)nk) = |a|F (a) f (k) | a F , n N, k K }.

We can see from the Iwasawa decomposition that a function in C() is determined by its
restriction to K. We define a norm on C() by
Z 1/2
2
kf k = |f (k)| dk , f C().
K

Here, dk denotes Haar measure on the compact group K. The completion H() of C() with
respect to this norm is a Hilbert space. Let I() = IBG () be the right regular representation
of G on H().
Note that the function d(a) 7 |a|F (a), which we will denote by | |F , is the internal
tensor product of the quasicharacters d(a) 7 |a|F and d(a) 7 (a) of A. Extending | |F
trivially across N , we obtain a quasicharacter of B = AN . Strictly speaking, the represen-
tation is induced from the quasicharacter | |F of B. We will see below that the factor | |F
is introduced so that I() will be unitary whenever is unitary. This version of induction
is called normalized induction.
Let B (d(a)n) = |a|2F , a F , n N . The following lemma is a consequence of the
Iwasawa decomposition and the fact that B | B K is trivial.
72 5. REPRESENTATIONS OF REDUCTIVE GROUPS

Lemma 2.1. If g = d(a0 )n0 k0 , a0 F , n0 N , k K, set B (d(a0 )n0 k) = B (d(a0 )).

Then B is a well-defined continuous function from G to the positive real numbers such that
B (d(a)ng) = B (d(a)) for all a F , n N and g G.

Since G is unimodular and B | B is the modular function of B, we can use the extension
B defined in the above lemma to produce a quasi-invariant measure on the coset space B\G.
For x and g G, set (Bg, x) = BB(gx)
(g)
. Then (see Theorem 5.11, Chapter 1), there exists a
quasi-invariant measure d g on B\G associated to the -function B , having the property:
Z Z
( g x) =
g ) d( (
g , x) (
g ) d
g, x G, Cc (B\G) = C(B\G).
B\G B\G

Lemma 2.2. Let V = { f C(G) | f (bg) = B (b) f (g) b B, g G }. For g G,

x G and f V , let (Rg f )(x) = f (xg). Then
(1) Given f V , the function g 7 B (g)1 f (g) factors to a continuous function on
B\G.
(2) The map : f 7 B\G (B1 f )( g defines a G-invariant element of V : (Rg f ) =
R
g ) d
(f ) for all g G and f V .

Proof. The first part of the lemma is a consequence of Lemma 2.1 and the definition
of V . Let be as defined in the statement of the lemma. Then
Z Z
1
(Rx f ) = B (g) f (gx) d
g= B (gx1 )1 f (g) d(
g x1 )
B\G B\G
Z Z
1 1 1
= B (gx ) f (g)(
g , x ) d
g= B (g)1 f (g) d
g = (f ).
B\G B\G

Lemma 2.3. Let be a character of F . Let V and be as in Lemma 2.2. Then

(1) If f1 , f2 C(), the function g 7 f1 (g)f2 (g) belongs to V .
(2) Set h f1 , f2 i = (f1 f2 ). This defines a G-invariant nondegenerate hermitian form
on C().

Proof. The lemma is a consequences of the definitions of V and C(), unitarity of ,

and Lemma 2.2. 
After verifying that, with appropriate normalizations of measures,
Z Z
1
(B f )(
g ) d
g= f (k) dk, f V,
B\G K
 3. PARABOLIC SUBGROUPS OF Sp4 (F ) 73

(we dont include the details here), we conclude that the Hilbert space H() is the completion
of C() with respect to h , i.

Proposition 2.4. If is a character of F , then I() is a unitary representation of

SL2 (F ).

Remark 2.5. For G = SL2 (Fq ), since all representations are unitary, we can use the
usual definition of induced representation to define principal
! series representations of G.
1 x
Define subgroups A = { d(a) | a F q }, N = { | x Fq } and B = AN . Given a
0 1
character of F
q , let I() be the right regular representation of G on the space of functions
f : G C such that f (d(a)ng) = (d(a))f (g) for all a F q , n N and g G.

3. Parabolic subgroups of Sp4 (F )

Suppose that G is the F -rational points of a connected reductive linear algebraic F -split
F -group G. In particular, G has a maximal F -torus A0 that splits over F : A0 = A0 (F ) is
isomorphic to F F . For example, G = GLn (F ), SLn (F ) and Sp2n (F ) are split
groups. For such groups, we can define Borel subgroups and other parabolic subgroups using
the root system = (G, A0 ) of A0 in G. We describe how this works for G = Sp4 (F ).

Remark 3.1. When we refer to a parabolic subgroup of G, we mean a subgroup P of

the form P(F ), where P is a parabolic subgroup of G and P is defined over the field F .

Remark 3.2. In general, noncompact reductive groups contain several conjugacy classes
of maximal tori. In particular, split reductive groups can contain compact maximal tori.
This is the case for SL2 (R) and SL2 (Qp ) (see Examples 6.1 and 6.2.

Let
0 0 0 1
0 0 1 0
J =


0 1 0 0
1 0 0 0
Then G = Sp4 (F ) = { g GL4 (F ) | t gJg = J} and g = {X M4 (F ) | t XJ + JX = 0}.
The group A0 := { d(a, b, b1 , a1 ) | a, b F } is (the F -rational points of) a maximal
F -split torus in G. Given integers i and j, set (i,j) (d(a, b, b1 , a1 )) = ai bj . Let =
(1,1) and = (0,2) . Consider the action of G on g by conjugation: Ad g(X) = gXg 1 ,
g G, X g. When we consider conjugation by elements of A0 , we obtain a root space
74 5. REPRESENTATIONS OF REDUCTIVE GROUPS

decomposition of g relative to the eigenspaces of Ad A0 . If (i, j) Z, (i,j) is a root of A0 if

there exists a nonzero X g such that Ad t(X) = (i,j) (t)X for all t A0 . In that case, the
corresponding root space is

g(i,j) = { X g | Ad t(X) = (i,j) (t) t A0 }.

Let = (G, A0 ) be the set of roots of A0 .. Then

= { , , ( + ), (2 + },

Because we often identify (i,j) with the pair (i, j), we write instead of 1 and +
instead of , etcetera.
The root spaces are easy to describe. If Eij is the matrix in M4 (F ) with a one in the
(i, j) entry and zeros elsewhere, then
g = SpanF (E12 E34 ), g = SpanF (E21 E43 ) g = SpanF E23
g+ = SpanF (E13 + E24 ), g2+ = SpanF E14 , etcetera.
For each root , there exists a root subgroup U of G. This group is of the form
U (F ) for an F -subgroup U of G, stable under conjugation by A0 and has Lie algebra g .
The group G is generated by A0 and the various root subgroups U , .
A base for a root system is a subset of having the property that each element of
P
has the form n , where either the n are all nonnegative integers, or the n are
all nonpositive integers. The elements of are called the simple roots. In this example,
:= {, } is a base of .
The Borel subgroup of G which corresponds to (or equivalently to + ) is the subgroup
B of upper triangular matrices in G. The subgroup N of upper triangular matrices in G
with ones on the diagonal is called the unipotent radical of B. The group N is generated by
the root groups U , + and the Lie algebra n of N is + g . Clearly, B = A0 n N . A
P

standard parabolic subgroup is a parabolic subgroup of G that contains the Borel subgroup
B. A parabolic subgroup of G is conjugate to some standard parabolic subgroup.
Each standard parabolic subgroup of G is attached to a subset of . Fix a subset I of
. Let I be the set of such that is an integral linear combination of elements of I.
Then I forms a root system. Let NI be the subgroup of N generated by the root groups
U such that + and / I . The Lie algebra of NI has the form

nI = + \+ g ,
I
 3. PARABOLIC SUBGROUPS OF Sp4 (F ) 75

0
where + +
I = I . Let AI = (I Ker ) , and MI = ZG (AI ) (the centralizer of AI in
G). (Here, (Ker )0 refers to the F -rational points of the identity component of the kernel
of the rational character of the algebraic torus A0 .) The set I coincides with the set of
roots in which are trivial on AI . The group MI is reductive and is generated by AI and
by the root groups U , I . The F -split torus AI is the identity component of the centre
of MI , and (MI , A0 ) = I . The Lie algebra of MI is equal to
M
mI = a0 g
I

Here, a0 is the Lie algebra of A0 . The group MI normalizes NI and PI = MI n NI . The

group MI is called a Levi factor of PI . It also called a Levi subgroup of G. If I and J are
subsets of then PI is conjugate to PJ if and only if I = J. However, it is possible for MI
to be conjugate to MJ when I 6= J.
When G = Sp4 (F ), apart from G = P and B = P , there are two standard parabolic
subgroups, P and P , attached to the subsets {} and {} of , respectively.
If I = {}, then I = { } and + \ + I = { , + , 2 + }. We can describe the
matrices in A , M and N :

A = Ker = { d(a, a, a1 a1 ) | a F },

X
! 0 !  

M = ZG (A ) = 0 1 t 1 0 1  X GL2 (F )

0
X
1 0 1 0

( ! ! )
I2 Y  y 1 y2
N = Y = , yj F

0 I2 y3 y1
When I = { }, I = { } and + \ +
I = { , + , 2 + }.

A = Ker = { d(a, 1, 1, a1 ) | a F }

d 0 0 0


0 c
11 c12 0 
M = ZG (A ) =  d F , c11 c22 c12 c21 = 1 ' SL2 (F ) F


0 c21 c22 0


0 d1

0 0
76 5. REPRESENTATIONS OF REDUCTIVE GROUPS

1 x y z


0
1 0 y 
N =  x, y, z F

0

0 1 x



0 0 0 1
When F = R, that is, when G is a Lie group, parabolic subgroups may be written in
a slightly different form. The Levi factor MI of a standard parabolic subgroup PI may be
expressed as a direct product of two groups MI0 and A0I . Here, A0I is the subgroup of AI
of the form (R ` 0
+ ) such that AI /AI is finite (and a product of groups of order two). For
example, for G = Sp4 (R), M0 is the set of elements in M such that det X = 1 (in the
notation above) and A0 = { d(a, a, a1 , a1 ) | a R 0
+ }. An element of M belongs to M if
and only if d = 1, and A0 = { d(a, 1, 1, a1 ) | a R+ }.

Remark 3.3. Let G = GLn (F ). The subgroup A0 of diagonal matrices in G is a maximal

split torus in G and the subgroup B of upper trangular matrices in G is a Borel subgroup
of G with respect to a particular choice of base of = (G, A0 ). The simple roots (that
is, the elements of ) are of the form d(a1 , . . . , an ) 7 aj a1
j+1 , 1 j n 1. If I , the
standard parabolic subgroup PI = MI n NI is the smallest subgroup of G that contains MI
and B. In particular, MI ' GLn1 (F ) GLnr (F ), where n1 + + nr = n and each
nj is a positive integer. More generally, as we saw above for G = Sp4 (F ), a Levi factor of
a parabolic subgroup of a reductive group is not necessarily a smaller rank version of the
same type of group: GL2 (F ) occurs as a Levi factor of a parabolic subgroup of Sp4 (F ) and
GL2 (F ) is not a symplectic group.

Although we have not used the Weyl group to define the standard parabolic subgroups,
we include some comments about the Weyl group of A0 for our example. The quotient group
W (A0 ) = NG (A0 )/A0 is called the Weyl group of A0 . The matrices

0 1 0 0 1 0 0 0
1 0 0 0 0 0 1 0
w = and w =


0 0 0 1 0 1 0 0
0 0 1 0 0 0 0 1
belong to NG (A0 ). The distinct elements of the Weyl group W (A0 ) are represented by the
matrices
{ I, w , w , w w , w w , w w w , w w w , (w w )2 }
 4. PARABOLIC SUBGROUPS AND INDUCTION 77

and W (A0 ) is isomorphic to the dihedral group of order 8.

Remark 3.4. The matrix w (resp. w ) belongs to NM (A0 ) (resp. NM (A0 )) and
represents the nontrivial element of the Weyl group of A0 in M (resp. M .)

4. Parabolic subgroups and induction

Let F = R or Qp (or a finite extension of Qp ). Let G be the F -rational points of a

connected reductive F -group and let P be a parabolic subgroup of G with Levi decomposition
P = M n N . The Iwawasa decomposition says that G = BK = KB for B a minimal
parabolic subgroup contained in P and some compact subgroup K of G. Since B P , we
have G = P K = KP .
The Levi factor M of P is the F -rational points of a connected reductive F -group. That
is, it is the same kind of group as G. (For example, when G = Sp4 (F ), then M is isomorphic
to one of F F , GL2 (F ), SL2 (F ) F and Sp4 (F ).)
Let g and n be the Lie algebras of G and N . Let g 7 Ad g : g g be the usual adjoint
representation of G: Ad g(X) = gXg 1 , g G, X g. If p P , then Ad p : n n. Set
1/2
P (p) = | det(Ad p) | n|F , p P . As in the principal series case, we use the function P to
normalize induction so that it takes unitary representations of M to unitary representations
of G. (The arguments are virtually identical to those used in the principal series case. We
simply replace the usual inner product on C with the appropriate inner product on the space
of the representation of M .)
Let (, W ) be a continuous representation of M in a Hilbert space W .

C() = { f C(G) | f (mnk) = P (mn) f (k) | m N n N, k K }.

We can see from the decomposition G = P K that a function in C() is determined by its
restriction to K. We define a norm on C() by
Z 1/2
2
kf k = kf (k)kW dk , f C().
K

Here, dk denotes Haar measure on the compact group K and k kW is the norm on W . The
completion H() of C() with respect to this norm is a Hilbert space. Let I() = IPG () be
the right regular representation of G on H().
Lemma 4.1. If g = m0 n0 k0 , m0 M , n0 N , k K, set P (m0 n0 k) = P (m0 ).
Then P is a well-defined continuous function from G to the positive real numbers such that
P (mng) = P (m) for all m M , n N and g G.
78 5. REPRESENTATIONS OF REDUCTIVE GROUPS

Since G is unimodular and P | P is the modular function of P , we can use the extension
P defined in the above lemma to produce a quasi-invariant measure on the coset space P \G.
For x and g G, set (P g, x) = PP(gx)
(g)
. Then (see Theorem 5.11, Chapter 1), there exists a
quasi-invariant measure d g on P \G associated to the -function P , having the property:
Z Z
( g x) =
g ) d( (
g , x) (
g ) d
g, x G, Cc (P \G) = C(P \G).
P \G P \G

Lemma 4.2. Let V = { f C(G) | f (pg) = P (p) f (g) p P, g G }. For g G,

x G and f V , let (Rg f )(x) = f (xg). Then
(1) Given f V , the function g 7 P (g)1 f (g) factors to a continuous function on
P \G.
(2) The map P : f 7 P \G (P1 f )( g defines a G-invariant element of V : P (Rg f ) =
R
g ) d
P (f ) for all g G and f V .

Lemma 4.3. Let (, W ) be a unitary representation of M , with M -invariant inner product

h , iW on W . Let V and P be as above. Then
(1) If f1 , f2 C(), the function (f1 ,f1 ) : g 7 h f1 (g), f2 (g) iW belongs to V .
(2) Set h f1 , f2 i = P ((f1 ,f2 ) ). This defines a G-invariant nondegenerate hermitian
form on C().

After verifying that, with appropriate normalizations of measures,

Z Z
1
(P f )(
g ) d
g= f (k) dk, f V,
P \G K

we see that the Hilbert space H() is the completion of C() with respect to h , i.

Proposition 4.4. If (, W ) is a unitary representation of M , then I() is a unitary

representation of G.

5. Discrete series representationsgeneral properties

Let (, V ) be a topologically irreducible unitary representation of a locally compact group

G. If every matrix coefficient of belongs to L2 (G), we say that is square-integrable or
belongs to the discrete series.
Let Z be the centre of G. If z Z, then (zg) = (gz) for all g G. That is,
(z) HomG (, ). Since is topologically irreducible and unitary, according to Schurs
Lemma, there exists a (nonzero) scalar (z) such that (z)1 (z) is the identity operator
 5. DISCRETE SERIES REPRESENTATIONSGENERAL PROPERTIES 79

on V . By continuity of , the map z 7 (z) is a quasicharacter of Z. Since is unitary,

is unitary that is, is a character of the locally compact abelian group Z. The character
is called the central character of . For v, w V ,

h (zg)v, w i = (z)h (g)v, w i, g G, z Z.

In particular, the function g 7 |h (g)v, wi| is left (and right) Z-invariant.

Lemma 5.1. If G is a locally compact group with noncompact centre, then a nonzero
matrix coefficient of a topologically irreducible unitary representation of G cannot belong to
L2 (G).

Suppose that G is unimodular. (This is the case when G is the F -rational points of
a reductive algebraic group over a locally compact field F , such as F = R or F = Qp ).
The centre Z of G is a locally compact abelian group, so Z is unimodular. Because G is
also unimodular, by Theorem 5.8 of Chapter 1, there exists a G-invariant measure on the
coset space G/Z. Fix a character of Z. We say that a function f : G C is compactly
supported modulo Z if there exists a compact set C in G such that the support of f is a
subset of Z C = C Z. Let
Cc (Z\G, ) = { g C(G) | f (zg) = (z)f (g) z Z, g G,
such that f is compactly supported modulo Z }.

Note that if f1 , f2 Cc (Z\G, ), the function f1 f2 belongs to the space Cc (Z\G) of con-
tinuous compactly supported functions on Z\G. We define an inner product on Cc (Z\G, )
by Z
h f1 , f 2 i = f1 (
g )f2 (
g ) d
g, f1 , f2 Cc (Z\G, ).
Z\G

Let L2 (Z\G, ) be the completion of Cc (Z\G, ) with respect to the norm associated to
this inner product. Recall that Proposition 4.1 of Chapter 2 shows that the left regular
representation of G on the space L2 (G/Z) is a continuous unitary representation of G. Of
course, the right regular representation of G on L2 (Z\G) is also unitary. A similar argument
can be used to see that the right regular representation of G on L2 (Z\G, ) is continuous
and unitary.

Theorem 5.2. Let G be a unimodular locally compact group with centre Z. Let (, V )
be a topologically irreducible unitary representation of G with central character . Then the
following are equivalent
80 5. REPRESENTATIONS OF REDUCTIVE GROUPS

(1) is equivalent to a subrepresentation of ( , L2 (Z\G, )).

(2) There exists a nonzero matrix coefficient of that belongs to L2 (Z\G, ).
(3) Every matrix coefficient of belongs to L2 (Z\G, ).

We say that a representation (, V ) is essentially square integrable (or square integrable

modulo Z) if the conditions of the theorem are satisfied. We may generalize our definition
of discrete series representation to the square integrable setting.

Remark 5.3. Even though the matrix coefficients of an essentially square integrable
representation with central character belong to L2 (G, ) and are continuous functions
on G, they do not generally belong to Cc (Z\G, ). As discussed below, there are certain
families of essentially square-integrable representations of reductive p-adic groups whose
matrix coefficients do belong to Cc (Z\G, ).

Theorem 5.4. Let G be a unimodular locally compact group. Let (, V ) and (, W ) be

topologically irreducible essentially square integrable representations with the same central
character . of G.
R
(1) If f and f are matrix coefficients of and , respectively, then Z\G f ( g )f (
g ) d
g=
0.
(2) Suppose that v, v 0 , u, u0 V and h , i is a G-invariant inner product on V . Then
there exists a positive real number d(), called the formal degree of , such that
Z
h (g)v, v 0 ih (g)u, u0 i d
g = d()1 hh v, u ih v 0 , u0 i
Z\G

Remark 5.5. The formal degree d() depends on the choice of normalization of the
G-invariant measure on G/Z.

6. Maximal Tori

We will see that (relative) discrete series of reductive Lie groups and certain kinds of
discrete series representations of reductive p-adic groups are associated with characters of
compact-mod-centre maximal tori. In this section, we describe examples of such tori.
A maximal torus T of G is said to be compact-mod-centre if T /Z is compact, where Z is
the centre of G. (When F = R, a compact-mod-centre maximal torus in G, or its identity
component, is often referred to as a relatively compact Cartan subgroup of G.)
 6. MAXIMAL TORI 81

Example 6.1. If G = SL2 (R), the group K = SO2 (R) is the fixed points of the Cartan
involution and it is also a maximal torus in G. If G = SL3 (R), then the group K = SO3 (R)
is the fixed points of the Cartan involution of G. The group

cos t sin t 0


T = sin t cos t 0 | t R


0 0 1

is a maximal torus in SO3 (R). But T is not a maximal torus in SL3 (R). In fact SL3 (R) has
rank two and
r cos t r sin t 0


0
T := r sin t r cos t 0 | r R , t R


0 0 r1

is a (noncompact) maximal torus of SL2 (R) (of dimension 2) containing T . (Note that T 0
is a maximal torus in the proper Levi subgroup SL2 (R) R of SL3 (R).) In fact, SL3 (R)
does not contain any compact maximal tori (see Remark 6.3).

Because p-adic fields have many extensions of varying degrees, there are plenty of maximal
tori in reductive p-adic groups that are compact modulo the centre of the group.

Example 6.2. Let p be an odd prime. There exists Zp such that the image of in
Zp /pZp ' Fp is a nonsquare. We can show that is a nonsquare in Qp . It is easy to see that
p is not a square in Qp . (A nonzero element x Qp satisfies |x|p = p` for some integer ` and

|p|p = p1 .) We have quadratic extensions E = Qp ( ) and L = Qp ( p) of Qp . Let
( ! )
a b
T = | a, b Qp , a2 b2 = 1
b a
( ! )
a bp
T0 = | a, b Qp , a2 b2 = 1 ,
bp1 a
( ! )
a bp
Tp = | a, b Qp , a2 b2 p = 1 .
b a
The subgroups T , T0 and Tp are examples of compact maximal tori in SL2 (Qp ). No two of
these subgroups are conjugate. (Note that T is a closed subgroup of the maximal compact
subgroup SL2 (Zp ) of SL2 (Qp ). The group T0 s a closed
! subgroup of the maximal compact
p 0
subgroup d(p)SL2 (Zp )d(p)1 , where d(p) = GL2 (Qp ). The group Tp is a closed
0 1
82 5. REPRESENTATIONS OF REDUCTIVE GROUPS

subgroup of the intersection of those two maximal compact subgroups.) Let be the non-
trivial element of the Galois group Gal(E/Qp ) Then T and T0 are isomorphic to the kernel
of the norm map NE/Qp : E Q
p : NE/Qp (x) = x(x), x E . Similarly, Tp is isomorphic
to the kernel of the norm map NL/Qp . (There exist compact maximal tori in SL2 (Qp ) which
are not conjugate to one of T , T0 and Tp .)
Now suppose that n is an integer and n 3. Suppose that E is an extension of Qp of
degree n. There are various ways to produce such extensions. For example, we can show
that there exists a root of unity Qp of order prime to p such that Qp () is an extension
of Qp of degree n. This is an example of an unramified extension of Qp . Or, we can see that
xn p is irreducible over Qp . A root of this polynomial in Qp generates a degree n extension
of Qp . This is an example of a totally ramified extension of Qp . Let be a basis for E over
Qp (viewing E as an n-dimensional vector space over Qp ). Given x E , let x be the
matrix of the invertible F -linear transformation `x of E given by multiplication by x. Then
x 7 x is an injection of E into GLn (Qp ). The intersection of the image of this map with
SLn (Qp ) is isomorphic to the kernel of the norm map NE/Qp : E Q p and is a compact
maximal torus in SLn (Qp ).
If T is a compact-mod-centre maximal torus in a reductive group over a local field, then
T cannot lie inside any proper Levi subgroup of G. (Note that a maximal torus of a proper
Levi subgroup of G contains the centre of the Levi subgroup and the centre of the Levi
subgroup contains an F -split torus that is not compact modulo the centre of G.) We often
refer to a maximal torus that is compact-mod-center as an elliptic maximal torus.
Remark 6.3. Of course, the maximal torus SO2 (R) of SL2 (R) is obtained from the
quadratic extension C of R in the same way that the maximal tori T , T0 and Tp of SL2 (Qp )

are obtained from quadratic extensions Qp ( ) and Qp ( p) of Qp . The lack of degree n
extensions of R for n 3 is what prevents SLn (R) from having compact maximal tori.
We may also consider tori in reductive groups over finite fields. In this context, we say
that a maximal torus is elliptic if it does not lie an any proper Levi subgroup of G. Elliptic
maximal tori always exist for GLn (Fq ) and SLn (Fq ), this is connected with the fact that,
given any positive integer n, there exists a degree n extension of Fq .

7. Discrete series representations of reductive Lie groups

Let G be a real reductive Lie group. In this case, G has (relative) discrete series repre-
sentations if and only if G has a compact-mod-centre maximal torus. subgroup). Suppose
 7. DISCRETE SERIES REPRESENTATIONS OF REDUCTIVE LIE GROUPS 83

that such a torus T exists. Then T is unique up to conjugacy. Let Tb be the group of unitary
characters of T . The complex Weyl group W (see below) acts on Tb. Let

Tb0 = { Tb | w 6= w W, w 6= 1 }.

The finite group WG (T ) := NG (T )/T also acts on Tb and is a subgroup of the complex Weyl
group W . (When G is compact WG (T ) is equal to the complex Weyl group.)
Let gC , kC and tC be the complexifications of the Lie algebras g, k and t of G, K and
T , respectively. By assumption, tC is a Cartan subalgebra (Lie algebra of a maximal torus)
of gC and of kC . Let G = (gC , tC ) and K = (kC , tC ) be the root systems of gC and kC
(respectively) with respect to tC . The complex Weyl group is the Weyl group of the root
system G .

Example 7.1. Let G = Sp4 (R) (realized as in Section 3). Then

cos t1 0 0 sin t1


0
cos t2 sin t2 0
T := | t1 , t 2 R




0 sin t2 cos t2 0


sin t1

0 0 cos t2
is a compact maximal torus in the group K of fixed points of the Cartan involution of G,
as well as a compact maximal torus in G. Hence G has discrete series representations. In
this example, K is isomorphic to U (2) and is realized as K = { g G | t gg = I }. The Lie
algebra k of K is realized as follows:


0 a b 1 b 2


a
t 1 0 b 3 b 1

k = { X M4 (R) | X = X and J XJ = X } = | a, bj R



b1 b3 0 a


b2 b1 a 0

The Lie algebra t of T has the form

0 0 0 t1


0
0 t2 0
t= | tj R .





0 t 2 0 0


t1 0 0 0

The root system G = (sp4 (C), tC ) contains eight roots and the complex Weyl group
is a dihedral group of order eight (see Section 3 for information on the Weyl group of
84 5. REPRESENTATIONS OF REDUCTIVE GROUPS

Sp4 (C)). The root system K contains two roots and has a Weyl group of order two.
Viewing characters of T = SO2 (R) SO2 (R) as tensor products 1 2 where each j is a
character of SO2 (R),
Tb0 = { 1 2 | 2j 6= 1, 1 6= 1
2 }

For convenience, we will assume that G is semisimple. The discrete series representa-
tions of G are parametrized by Tb0 . If , 0 Tb0 , then the corresponding discrete series
representations are equivalent if and only if there exists w WG (T ) such that 0 = w.
There is a connection with the irreducible representations of K. If we have a unitary
representation (, V ) of G we can restrict the representation to K. Then we can con-
sider the space of K-finite vectors in V : the vectors belonging to finite-dimensional K-
subrepresentations of V (see Definition 7.4 of Chapter 3). The space of K-finite vectors
is dense in V (Corollary 7.5 of Chapter 3). A K-type of is defined to be an irreducible
unitary representation of K that occurs in the space of K-finite vectors. Equivalently, the
space V of -isotypic vectors in V is nonzero (see Definition 9.8 of Chapter 3). According
to the theory of finite-dimensional irreducible representations of compact Lie groups (see
Theorem 4.13, Chapter 4), each K-type of has a highest weight. The parametrization of
discrete series representations is expressed in terms of the highest weights of the K-types of
the representations.
For convenience, we assume that G is semisimple. Take Tb0 . As in Chapter 4, via the
isomorphism X + K 7 e2iX from tR /K 7 T , where tR = it, corresponds to an element
of tR which is trivial on K, that is, an element of the weight lattice Y (see Definition 2.2,
Chapter 4). Perhaps is not dominant. Since the WG (T )-orbit of contains exactly one
dominant weight, after replacing by a suitable element in this orbit, we may assume that
is dominant. In the current setting, it may be convenient to view as an element of (tC )
(recall that tC = tR itR ). Each root space gC , G either lies inside kC or has trivial
intersection with kC . This means that the root system K is a subset of G . In defining
dominant elements of the weight lattice Y d , we fixed a choice +
K of positive roots in K . As
in Definition 4.1 of Chapter 4, for each G , let H tR be such that (X) = K(H , X)
for all X tR . (Here, K is the Killing form.) Since is dominant, we have (H ) > 0 for
+ b0
K . The fact that T implies that (H ) 6= 0 for all G . There is a unique
choice of positive roots + +
G such that (H ) > 0 for all G . Let G and K be the
half-sums of the positive roots in + +
G and K , respectively.
For Y d , let be the irreducible representation of K with highest weight .
 8. DISCRETE SERIES REPRESENTATIONS OF REDUCTIVE p-ADIC GROUPS 85

A proof of the following theorem can be found in [K1] (see Theorems 9.20 and 12.21).

Theorem 7.2. (Harish-Chandra) Assume that G is semisimple and G Y. Let be

the discrete series representation corresponding to . Let = + G 2 K . Then
(1) The isotypic subspace V is nonzero and the multiplicity of in V equals one.
(2) If Y d occurs as the highest weight of a K-type of , then = + + n
P
G
for nonnegative integers n .
Two such representations are equivalent if and only if their parameters are conjugate
under WG (T ).

There is an additional property of the discrete series , which, together with the two
properties stated above in the theorem, characterizes the representation. By a version of
Schurs Lemma, the centre Z(g) of the universal enveloping algebra of g acts on the space
K-finite vectors in the space of via scalar operators. This results in a C-algebra homomor-
phism from Z(g) to C, which is associated to a Weyl group orbit in (tC ) . This is called the
infinitesimal character of . The additional property of is that the infinitesimal character
of is the Weyl group orbit of .
If (, V ) is a continuous representation of G in a Hilbert space V , for each f Cc (G), we
R
define a linear operator (f ) on V by (f )v = G f (g)(g)v dg, v V . Assume that (f ) is
of trace class for all f Cc (G) (for example, this is the case if is irreducible and unitary).
Then the map f 7 (f ) is a distribution on G (that is, a continuous linear functional on
Cc (G)). This distribution is called the (global) character of . It is realized by a locally
integrable function on G:
Z
tr (f ) = f (g) (g) dg, f Cc (G).
G

Harish-Chandra parametrized the discrete series representations and constructed their

characters as locally integrable functions. On the compact Cartan subgroup T , the character
of the discrete series representation has a formula that is analogous to the Weyl
character formula for compact Lie groups. (The character also has nonzero values on the
noncompact Cartan subgroups.)

8. Discrete series representations of reductive p-adic groups

Let G be a (nonabelian) connected reductive p-adic group. In this case, G has sev-
eral conjugacy classes of compact-mod-centre maximal tori and there are discrete series
86 5. REPRESENTATIONS OF REDUCTIVE GROUPS

representations associated with characters of such tori. In addition, there are discrete series
representations that are not associated with characters of compact-mod-centre maximal tori.
Let (, V ) be a continuous representation of G. We say that a vector v V is smooth if
there exists a compact open subgroup K of G such that (k)v = v for every k K. This
is equivalent to saying that StabG (v) := { g G | (g)v = v } is open. (Note that here K
is any compact open subgroup of G it is not the compact open subgroup in the Iwasawa
decomposition of G.) We can show that V is a subspace of V and V is dense in V . Recall
that (see Proposition 3.1 of Chapter 3) that V = V whenever is finite-dimensional.
For the next example, we need the definition of cuspidal representation of a finite group
of Lie type. Let H = H(Fq ) be a finite group of Lie type. An irreducible representation
(, V) of H is cuspidal if

V N := { v V | (n)v = v n N } = {0}

for any unipotent radical of a proper parabolic subgroup of H. (Recall that a parabolic
subgroup P is a semidirect product of the form P = M n N .) It can be shown that is
cuspidal if it does not occur as an irreducible consituent of any representation of the form
IndGP , where P = M n N is a proper parabolic subgroup of H and is a finite-dimensional
representation of M .
Example 8.1. Let K = SL2 (Zp ). The maximal normal pro-p-subgroup of K consists of
the set of matrices k K such that all entries of k I belong to pZp . The factor group K/K u
is isomorphic to SL2 (Fp ). Let (, V) be an irreducible cuspidal representation of SL2 (Fp ).
Let
V G = { f : G V | f (kg) = (k) f (g), k K, g G, suppf j Kgj ,
for a finite subset {gj } of G }.

Let indG G G
K be the right regular representation of G on V . Then indK is an irreducible
smooth representation of G. Let be a matrix coefficient of . Define 0 on G by

(g) if g K

0 (g) =
0, if g
/K

Let v V and V be such that (k) = ((k)v) for k K. Define fv : G V by

(g)v if g K
fv (g) =
0, if g
/K
 8. DISCRETE SERIES REPRESENTATIONS OF REDUCTIVE p-ADIC GROUPS 87

Then fv V G . Define V G by (f ) = (f (1)), f V G . Then we can show that

(indG 0 0 G 0
K (g)fv ) = (g), g G. Thus is a matrix coefficient of indK . By definition, is
a compactly supported continuous function. In particular, 0 L2 (G). The representation
indG
K is an example of a supercuspidal representation of G.
Since G and K are unimodular (G is reductive and K is compact), we have a G-invariant
measure d g on K\G. The representation is unitary. Let h , i be a K-invariant inner
product on V. The function g 7 h f1 (g), f2 (g) i factors to a function in Cc (K\G). Set
Z
h f1 , f 2 i = h f1 ( g ) i d
g ), f2 ( g, f1 , f2 V G .
K\G

Using G-invariance the measure, we can see that the inner product is G-invariant. Hence
the completion of V G with respect to this inner product is a unitary representation of G.

A smooth representation of G is supercuspidal if every matrix coefficient of is com-

pactly supported modulo the centre Z of G. It follows that a unitarizable supercuspidal
representation of G has the property that its matrix coefficients belong to L2 (G).

Lemma 8.2. Let is an irreducible smooth representation of G, then is supercuspidal

if and only if some nonzero matrix coefficient of is compactly supported modulo Z.

Let f : G C be a locally constant function such that the support of f is compact

R
modulo Z. Then f is a cusp form if N f (gn) dn = 0 for all g G and all unipotent radicals
of proper parabolic subgroups of G. (Here, dn denotes a Haar measure on N .)

Lemma 8.3. If is an irreducible supercuspidal representation of G, then all matrix

coefficients of are cusp forms.

Remark 8.4. We may also define cusp forms on finite groups of Lie type. It is easy to
see that a matrix coefficient of a cuspidal representation of a finite group of Lie type is a
cusp form. For this reason, we may view supercuspidal representations of reductive p-adic
groups as analogues of cuspidal representations of finite groups of Lie type.

As we saw above, some supercuspidal representations are obtained via induction from
inflations of cuspidal representations of finite groups of Lie type. However, there are many
supercuspidal representation that are induced from representations of compact-mod-centre
subgroups where the inducing representations are not related to cuspidal representations of
finite groups of Lie type. There exist families of supercuspidal representations which arise
from characters of elliptic (compact-mod-centre) maximal tori.
88 5. REPRESENTATIONS OF REDUCTIVE GROUPS

Example 8.5. Let T = T be as in Section 6. Let be a character ot T such that

| T K u is nontrivial, where K u is as in the previous example. There is a way to de-
fine a compact open subgroup K of SL2 (Zp ) that is normalized by T and an irreducible
smooth representation of the compact open subgroup T K such that the representation
indGT K is an irreducible smooth representation of SL2 (Qp ). This induced representation
is supercuspidal.

As seen above, we may construct unitary discrete series representations from unitarizable
supercuspidal representations. Although we dont take the time to discuss examples, there
are some discrete series representations of reductive p-adic groups which are not obtained
from supercuspidal representations (and are not associated with characters of maximal tori).

9. Maximal tori and representations of finite groups of Lie type

In this section, q is a power of a prime p, G is a connected reductive Fq -group, and

G = G(Fq ). If T is a maximal Fq -torus in G we will refer to T := T(Fq ). as a maximal
torus in G. The so-called Deligne-Lusztig construction associates class functions on G to
characters of maximal tori in G. Certain of these class functions are, up to sign, equal to
characters of irreducible representations of G. The book of Carter ([Car]) is a good basic
reference for this (and for other information about the representation theory of reductive
groups over finite fields).
Let ` 6= p be a prime. The `-adic cohomology groups with compact support play a
role in the Deligne-Lusztig construction. Suppose that X is an algebraic variety over the
algebraic closure Fq of Fq . Each automorphism of X induces a nonsingular linear map
Hci (X, Q` ) Hci (X, Q` ). This makes Hci (X, Q` ) a module for the group of automorphisms
of X. If g is an automorphism of X of finite order, the Lefschetz number of g on X L(g, X)
is defined to be i (1)i trace(g, Hci (X, Q` )). It is known that L(g, X) is an integer that is
P

independent of `.
Let T = T(Fq ) be a maximal torus in G and let Let B be a Borel subgroup (minimal
parabolic subgroup) of G that contains T. Note that B is not necessarily defined over Fq .
If N is the unipotent radical of B, then B = T n N. Let Fr be the Frobenius element
of Gal(Fq /Fq ). (That is, Fr(x) = xq for every x Fq .) The notation Fr will also be
used for the action of Fr on G, T, etc. Define L(g) = g 1 Fr(g), g G. This map is
called Langs map. Note that G = GFr = L1 (1). The set X e = L1 (N) is an affine
algebraic variety. The group G acts on X e by left multiplication: if g G and x X,
e then
 10. HARISH-CHANDRAS PHILOSOPHY OF CUSP FORMS 89

L(gx) = x1 g 1 Fr(g)Fr(x) = x1 Fr(x) = L(x) N, so gx X. e The group T acts on X e by

e and t T , then L(xt) = t1 x1 Fr(x) t t1 Nt = N. These
right multiplication: if x X
actions commute with each other. Thus Hci (X,e Q` ) is a left G-module and a right T -module
such that (gv)t = g(vt) for g G, t T and v Hci (X,
e Q` ).
Let Tb be the set of characters (one-dimensional representations) of the maximal torus
T . Let Tb. If n is the order of T and t T , then (t) is an nth root of unity, so
(t) is an algebraic integer. Because Q` contains the algebraic numbers, we can view as a

homomorphism from T to Q` . Let Hci (X, e Q` ) be the T -submodule of H i (X,
c
e Q` ) on which
G
T acts by the character . Define RT, = RT, : G Q` by
X
RT, (g) = (1)i trace(g, Hci (X,
e Q` ) ), g G.
i0

We say that the character is in general position if w 6= for all nontrivial elements
w of the Weyl group NG (T )/T of T in G.
Properties of RT, :
(1) RT, is a generalized character (an integral linear combination of characters of
irreducible representations of G).
(2) RT, (g) = |T | tT (t1 ) L((g, t), X),
P e g G.
(3) RT, is independent of the choice of Borel subgroup B that has T as Levi factor.
(4) If is in general position, then, up to sign, RT, is the character of an irreducible
representation of G.
(5) If T is elliptic (that is, T does not lie in the Levi factor of a proper Fq -parabolic
subgroup of G) and is in general position, then, up to sign, RT, is the character
of an irreducible cuspidal representation of G.

Remark 9.1. In general, there exist irreducible representations of G whose characters

are not of the form RT, for some T and , with in general position.

10. Harish-Chandras Philosophy of Cusp Forms

Harish-Chandras philosophy of cusp forms describes similarities between the following

four theories:
Representation theory of finite groups of Lie type
Representation theory of reductive Lie groups
Representation theory of reductive p-adic groups
90 5. REPRESENTATIONS OF REDUCTIVE GROUPS

Theory of automorphic forms

The philosophy says that there are certain basic representations from which all other
representations are constructed. In the first three cases, the basic representations are cus-
pidal, supercuspidal, and discrete series representations, respectively, and the construction
is parabolic induction. In the fourth case, the basic representations are cuspidal automor-
phic representations and the construction is Eisenstein series. (We will not be discussing
the fourth case here.) Much of the time, though not always, the construction produces
irreducible representations. When the construction produces a representation which is not
irreducible, some work may be involved in determining the irreducible subquotients of the
representation.
An irreducible representation of a finite group G of Lie type is cuspidal or is a constituent
of a representation that is parabolically induced from a cuspidal representation of a proper
Levi subgroup of G.
When working with representations of reductive Lie groups and reductive p-adic groups,
most results are valid in the context of admissible representations.
A smooth representation (, V ) of a reductive p-adic group is admissible if the space V K
of K-fixed vectors in V is finite-dimensional for every compact open subgroup K of G. An
irreducible smooth representation of a reductive p-adic group is admissible. An irreducible
smooth representation of a reductive p-adic group G is supercuspidal or occurs as a sub-
quotient of a representation that is parabolically induced from an irreducible supercuspidal
representation of a proper Levi subgroup of G.
Let G be a reductive Lie group and let K be group of fixed points of the Cartan in-
volution of G. A continuous representation (, V ) of G in a Hilbert space V is admissible
if for every irreducible continuous representation of K, the dimension of the -isotypic
subspace V is finite-dimensional. (That is, the representation occurs finitely many times
in the decomposition of the restriction of to K.) Any irreducible unitary representation of
G is admissible. We say that a parabolic subgroup P = M n N is cuspidal if M has relative
discrete series representations. Roughly speaking, the Langlands classification of irreducible
admissible representations of G is stated as follows. An irreducible admissible representation
is a quotient of a representation obtained via parabolic induction from a representation of
the form , where , resp. , is a relative discrete series representation, resp. quasichar-
acter, of a Levi factor of a cuspidal parabolic subgroup of G. For a precise statement, see
 10. HARISH-CHANDRAS PHILOSOPHY OF CUSP FORMS 91

Theorem 14.92 of [K1]. (In fact we also have to allow some other representations, known as
limits of discrete series on the Levi subgroups.)
 Bibliography

Introductions to representation theory

[Bo] A. Borel, Properties and linear representations of Chevalley groups, in Seminar on Algebraic Groups
and Related Finite Groups, Springer Lecture Notes in Math. 131 (1970), 155.
[BD] T. Brocker and T. tom Dieck, Representations of Compact Lie Groups, Graduate Texts in Mathe-
matics, 98, Springer-Verlag, 1985.
[BH] C.J. Bushnell and G. Henniart, The local Langlands conjecture for GL(2), A Series of Compre-
hensive Studies in Mathematics 335 Springer-Verlag 2006. (Chapter 1: Introduction to smooth
representations of locally profinite groups; Chapters 24: Representations of GL(2) over finite and
p-adic fields.)
[C] W. Casselman, Introduction to the theory of admissible representations of p-adic reductive groups,
Unpublished manuscript, 1974. http://www.math.ubc.ca/people/faculty/cass/research.html.
[FH] W. Fulton and J. Harris, Representation Theory: A First Course, Graduate Texts in Mathemat-
ics, 129, Springer-Verlag, New York, 1991. (Representations of finite groups; Finite-dimensional
representations of Lie groups and Lie algebras.)
[Ha] Brian C. Hall, Lie Groups, Lie Algebras, and Representations, An Elementary Introduction, Grad-
uate Texts in Mathematics 222, 2003, Springer.
[H] Introduction to Lie Algebras and Representation Theory, James E. Humphreys, Graduate Texts in
Mathematics, 9, Springer-Verlag.
[K1] Anthony W. Knapp, Representation Theory of Semisimple Groups: An Overview Based on Exam-
ples, Princeton University Press, 1986.
[K2] Anthony W. Knapp, Lie Groups: Beyond an Introduction, Birkhauser 1996.
[M] F. Murnaghan, Mat 1197 Course Notes: Representations of reductive p-adic groups, available at
www.math.toronto.edu/ murnaghan.
[R] A. Robert, Introduction to the Representation Theory of Compact and Locally Compact Groups,
London Mathematical Society Lecture Note Series 80, Cambridge University Press, 1983. (Main
reference for the first part of the course.)
[Se] M. Sepanski, Compact Lie Groups, Graduate Texts in Mathematics, 235, Springer-Verlag, 2007.
[Si] B. Simon, Representations of finite and compact groups, Graduate Studies in Mathematics, 10,
American Mathematical Society, 1996.

93
94 BIBLIOGRAPHY

Topological groups and measure theory

[B] N. Bourbaki, Integration I, Chapters 16, Translated from the original French edition (19591965),
Springer-Verlag, 2004.
[Hal] P. Halmos, Measure Theory, D. Van Nostrand Company, Inc. New York, 1950.
[HR] E. Hewitt and K. Ross, Abstract Harmonic Analysis. Vol. I, Structure of topological groups, in-
tegration, theory, group representations, Grundlehren der Mathematischen Wissenshaften, 115,
Springer-Verlag, 1979.
[Hu] T. Husain, Introduction to Topological Groups, W.B. Saunders Co., 1966.
[L] L. Loomis, An Introduction to Abstract Harmonic Analysis, D. Van Nostrand Company, Inc., 1953.
[Ma] N. Markley, Toopological groups: an introduction, Wiley, 2010.
[P] L.S. Pontryagin, Topological Groups, second edition, Gordon and Breach, 1966.

Profinite groups and local fields

[Ba] G. Bachman, Introduction to p-adic numbers and valuation theory, Academic Press, London, 1964.
[Cas] J.W.S. Cassels, Local Fields, London Mathematical Society Student Texts. 3, Cambridge University
Press, 1986.
[S] S. Shatz, Profinite groups, arithmetic, and geometry, Annals of Mathematics Studies 67 Princeton
University Press, 1972.

Representations of reductive p-adic groups

[BZ1] I.N. Bernstein and A.V. Zelevinsky, Representations of the group GL(n, F ), where F is a non-
archimedean local field, Russian Math. Surveys 31:3 (1976), 168.
[BZ2] I.N. Bernstein and A.V. Zelevinsky, Induced representations of reductive p-adic groups. I, Ann.
Norm. Sup. 4e serie, 10 (1977), 441472.
scient. Ec.
[Z] A. Zelevinsky, Induced representations of p-adic reductive groups II: Irreducible representations of
GL(n), Ann. Sci. Ec. Norm. Sup. 3 (1980), 165210.

Other references

[Car] R. Carter, Finite groups of Lie type: conjugacy classes and complex characters, Pure and Applied
Mathematics, Wiley, 1985.
[Ka] R. Kallman, A characterization of uniformly continuous unitary representations of connected locally
compact groups, Michigan Math. J. 16 (1969), 257263.
[KT] E. Kaniuth and K. Taylor, Induced representations of locally compact groups, Cambridge Tracts in
Mathematics, Cambridge University Press, 2013.
[K] S. Kantorovitz, Introduction to modern analysis, Oxford Graduate Texts in Mathematics, 8, Oxford
University Press, 2003.
[Sg] I.M. Singer, Uniformly continuous representations of Lie groups, Annals of Math. 56 (1952), 242
247.