Representation Theory

Lectured by S. Martin, Lent Term 2009

1 Group Actions 1
2 Linear Representations 2
3 Complete Reducibility and Maschke’s Theorem 6
4 Schur’s Lemma 8
5 Character Theory 10
6 Proofs and Orthogonality 14
7 Permutation Representations 17
8 Normal Subgroups and Lifting Characters 19
9 Dual Spaces and Tensor Products of Representations 21
10 Induction and Restriction 27
11 Frobenius Groups 31
12 Mackey Theory 33
13 Integrality 35
14 Burnside’s pa q b Theorem 36
15 Representations of Topological Groups 38

Examples Sheets
REPRESENTATION THEORY (D) 24 lectures, Lent term

Linear Algebra, and Groups, Rings and Modules are esssential.

Representations of finite groups

Representations of groups on vector spaces, matrix representations. Equivalence of represen-
tations. Invariant subspaces and submodules. Irreducibility and Schur’s Lemma. Complete
reducibility for finite groups. Irreducible representations of Abelian groups.

Characters
Determination of a representation by its character. The group algebra, conjugacy classes, and
or- thogonality relations. Regular representation. Induced representations and the Frobenius
reciprocity theorem. Mackey’s theorem. [12]

Arithmetic properties of characters

Divisibility of the order of the group by the degrees of its irreducible characters. Burnside’s
pa q b theorem. [2]

Tensor products
Tensor products of representations. The character ring. Tensor, symmetric and exterior
algebras. [3]

Representations of S 1 and SU2

The groups S 1 and SU2 , their irreducible representations, complete reducibility. The
Clebsch-Gordan formula. *Compact groups.* [4]

Further worked examples

The characters of one of GL2 (Fq ), Sn or the Heisenberg group. [3]

Appropriate books

J.L. Alperin and R.B. Bell Groups and representations. Springer 1995 (£37.50 paperback).
I.M. Isaacs Character theory of finite groups. Dover Publications 1994 (£12.95 paperback).
G.D. James and M.W. Liebeck Representations and characters of groups. Second Edition,
CUP 2001 (£24.99 paperback).
J-P. Serre Linear representations of finite groups. Springer-Verlag 1977 (£42.50 hardback).
M. Artin Algebra. Prentice Hall 1991 (£56.99 hardback).
 Representation Theory
This is the theory of how groups act as groups of transformations on vector spaces.

• group (usually) means finite group

• vector spaces are finite-dimensional and (usually) over C.

1. Group Actions
• F a field – usually F = C or R or Q : ordinary representation theory
– sometimes F = Fp or Fp (algebraic closure) : modular representation theory.

• V a vector space over F – always finite-dimensional over F

• GL(V ) = {θ : V → V , θ linear, invertible} – group operation is composition, identity is 1.

Basic linear algebra

If dimF V = n < ∞, choose a basis e1 , . . ., en over F so that we can identifyPit with F n .
Then θ ∈ GL(V ) corresponds to a matrix Aθ = (aij ) ∈ Fn×n where θ(ej ) = i aij ei , and
Aθ ∈ GLn (F ), the general linear group.

(1.1) GL(V ) ∼
= GLn (F ), θ 7→ Aθ . (A group isomorphism – check Aθ1 θ2 = Aθ1 Aθ2 , bijection.)

Choosing different bases gives different isomorphisms to GLn (F ), but:

(1.2) Matrices A1 , A2 represent the same element of GL(V ) with respect to different bases
iff they are conjugate/similar, viz. there exists X ∈ GLn (F ) such that A2 = XA1 X −1 .
P
Recall the trace of A, tr (A) = i aii where A = (aij ) ∈ Fn×n .

(1.3) tr (XAX −1 ) = tr (A), hence define tr (θ) = tr (A), independent of basis.

(1.4) Let α ∈ GL(V ) where V is finite-dimensional over C and α is idempotent, i.e.

αm = id, some m. Then α is diagonalisable. (Proof uses Jordan blocks – see Telemann p.4.)

Recall End(V ), the endomorphism algebra, is the set of all linear maps V → V with
natural addition of linear maps, and the composition as ‘multiplication’.

(1.5) Proposition. Take V finite-dimensional over C, α ∈ End(V ). Then α is diagonalis-

able iff there exists a polynomial f with distinct linear factors such that f (α) = 0.
Qm−1
Recall in (1.4), αm = id, so take f = X m − 1 = j=0 (X − ω j ), where ω = e2πi/n .

Proof of (1.5). f (X) = (X − λ1 ). . .(X − λk ).

(X − λ1 ). . .(X\ − λj ). . .(X − λk )
Let fj (X) = , where c means ‘remove’.
\
(λj − λ1 ). . .(λj − λj ). . .(λj − λk )
P
So 1 =
L fj (X). Put Vj = fj (α)V . The fj (α) are orthogonal projections, and V =
Vj with Vj ⊆ V (λj ) the λk -eigenspace. 2

(1.4*) In fact, a finite family of commuting separately diagonalisable automorphisms of a

C-space can be simultaneously diagonalised.

1
Basic group theory
(1.6) Symmetric group, Sn = Sym(Xn ) on the set Xn = {1, . . ., n}, is the set of all
permutations (bijections Xn → Xn ) of Xn . |Sn | = n!

Alternating group, An on Xn , is the set of products of an even number of transpositions

(ij) ∈ Sn . (‘An is mysterious. Results true for Sn usually fail for An !’)

(1.7) Cyclic group of order n, Cn = hx : xn = 1i. E.g., Z/nZ under +.

It’s also the group of rotations, centre 0, of the regular n-gon in R2 . And also the group of
nth roots of unity in C (living in GL1 (C)).

(1.8) Dihedral group, D2m of order 2m. D2m = hx, y : xm = y 2 = 1, yxy −1 = x−1 i.

Can think of this as the set of rotations and reflections preserving a regular m-gon (living in
GL2 (R)). E.g., D8 , of the square.

(1.9) Quaternion group, Q8 = hx, y : x4 = 1, y 2 = x2 , yxy −1 = x−1 i of order 8.

(‘Often used as a counterexample to dihedral results.’)

   
i 0 0 1
In GL2 (C), can put x = ,y= .
0 −i −1 0

(1.10) The conjugacy class of g ∈ G is CG (g) = {xgx−1 : x ∈ G}.

Then |CG (g)| = |G : CG (g)|, where CG (g) = {x ∈ G : xg = gx} is the centraliser of g ∈ G.

Definition. G a group, X a set. G acts on X if there exists a map ∗ : G × X → X,

(g, x) 7→ g ∗ x, written gx for g ∈ G, x ∈ X, such that:

1x = x for all x ∈ X
(gh)x = g(hx) for all g, h ∈ G, x ∈ X.

Given an action of G on X, we obtain a homomorphism θ : G → Sym(X) called the per-

mutation representation of G.

Proof. For g ∈ G, the function θg : X → X, x 7→ gx, is a permutation (inverse is θg−1 ).

Moreover, for all g1 , g2 ∈ G, θg1 g2 = θg1 θg2 since (g1 g2 )x = g1 (g2 x) for x ∈ X. 2

In this course, X is often a finite-dimensional vector space, and the action is linear, viz:
g(v1 + v2 ) = gv1 + gv2 , g(λv) = λgv for all v, v1 , v2 ∈ V = X, g ∈ G, λ ∈ F .

2. Linear Representations
G a finite group. F a field, usually C.

(2.1) Definition. Let V be a finite-dimensional vector space over F . A (linear) repre-

sentation of G on V is a homomorphism ρ = ρV : G → GL(V ).

We write ρg for ρV (g). So for each g ∈ G, ρg ∈ GL(V ) and ρg1 g2 = ρg1 ρg2 .

The dimension or degree of ρ is dimF V .

2
(2.2) Recall ker ρ
G and G/ ker ρ = ∼ ρ(G) 6 GL(V ). (The first isomorphism theorem.)
We say that ρ is faithful if ker ρ = 1.

Alternative (and equivalent) approach:

(2.3) G acts linearly on V if there exists a linear action G × V → V , viz:

action: (g1 g2 )v = g1 (g2 v), 1v = v, for all g1 , g2 ∈ G, v ∈ V

linearity: g(v1 + v2 ) = gv1 + gv2 , g(λv) = λgv, for all g ∈ G, v ∈ V , λ ∈ F .

So if G acts linearly on V , the map G → GL(V ), g 7→ ρg , with ρg : v 7→ gv, is a represen-

tation of V . And conversely, given a representation G → GL(V ), we have a linear action of
G on V via g.v = ρ(g)v, for all v ∈ V , g ∈ G.

(2.4) In (2.3) we also say

 Pthat V is a G-space or a G-module. In fact, if we define the
group algebra F G = g∈G αg g : αg ∈ F then V is actually an F G-module.

Closely related:

(2.5) R is a matrix representation of G of degree n if R is a homomorphism G → GLn (F ).

Given a linear representation ρ : G → GL(V ) with dimF V = n, fix basis B; get a matrix
representation G → GLn (F ), g 7→ [ρ(g)]B .

Conversely, given matrix representation R : G → GLn (F ), we get a linear representation

ρ : G → GL(V ), g 7→ ρg , via ρg (v) = Rg (v).

(2.6) Example. Given any group G, take V = F (the 1-dimensional space) and ρ : G →
GL(V ), g 7→ (id : F → F ). This is known as the trivial/principal representation. So
deg ρ = 1.

(2.7) Example. G = C4 = hx : x4 = 1i.

Let n = 2 and F = C. Then R : x 7→ X will determine all xj 7→ X j . We need X 4 = I.

We can take: X diagonal – any such with diagonal entries ∈ {±1, ±i} (16 choices)
Or: X not diagonal, then it will be isomorphic to some diagonal matrix, by (1.4)

(2.8) Definition. Fix G, F . Let V, V ′ be F -spaces and ρ : G → GL(V ), ρ′ : G → GL(V ′ )

be representations of G. The linear map φ : V → V ′ is a G-homomorphism if

(∗) φρ(g) = ρ′ (g)φ for all g ∈ G. ρg

V −→ V
φ ↓ ↓ φ
We say φ intertwines ρ, ρ′ .
V ′ −→ V ′
ρ′g
We write HomG (V, V ′ ) for the F -space of all of these.
The square commutes

We say that φ is a G-isomorphism if also φ is bijective; if such a φ exists we say that ρ, ρ′ are
isomorphic. If φ is a G-isomorphism, we write (∗) as ρ′ = φρφ−1 (meaning ρ′ (g) = φρ(g)φ−1
for all g ∈ G).

(2.9) The relation of being isomorphic is an equivalence relation on the set of all linear
representations of G (over F ).

3
Remark. The basic problem of representation theory is to classify all representations of
a given group G up to isomorphisms. Good theory exists for finite groups over C, and for
compact topological groups.

(2.10) If ρ, ρ′ are isomorphic representations, they have the same dimension. Converse is
false: in C4 there are four non-isomorphic 1-dimensional representations. If ω = e2πi/4 then
we have ρj (ω i ) = ω ij (0 6 i 6 3).

ρ
(2.11) Given G, V over F of dimension n and ρ : G → GL(V ). Fix V −→ V
a basis B for V ; we get a linear isomorphism φ : V → F n , v 7→ [v]B . φ ↓ ↓ φ
Get a representation ρ′ : G → GL(F n ) isomorphic to ρ. F n −→ F n
ρ′

(2.12) In terms of matrix representations, R : G → GLn (F ), R′ : G → GLn (F ) are G-

isomorphic if there exists a (non-singular) matrix X ∈ GLn (F ) with R′ (g) = XR(g)X −1
(for all g ∈ G).

In terms of G-actions, the actions of G on V , V ′ are G-isomorphic if there is an isomorphism

φ : V → V ′ such that g φ(v) = φ (gv) for all g ∈ G, v ∈ V .
|{z} |{z}
in V ′ in V

Subrepresentations
(2.13) Let ρ : G → GL(V ) be a representation of G. Say that W 6 V is a G-subspace if
it’s a subspace and is ρ(G)-invariant, i.e. ρg (W ) ⊆ W for all g ∈ G.

Say ρ is irreducible, or simple, if there is no proper G-subspace.

(2.14) Example. Any 1-dimensional representation of G is irreducible. (But not conversely:

e.g. D6 has a 2-dimensional C-irreducible representation.)

(2.15) In definition (2.13) if W is a G-subspace then the corresponding map G → GL(W ),

g 7→ ρ(g)|W is a representation of G, a subrepresentation.

(2.16) Lemma. ρ : G → GL(V ) a representation. If W is a G-subspace of V and if

B = {v1 , . . ., vn } is a basis of V containing the basis {v1 , . . ., vm } of W , then the matrix of
ρ(g) with respect to B is (with the top-left * being m × m)
 
* * (for each g ∈ G)
0 *

(2.17) Examples

(i) (2.10) revisited. The irreducible representations of C4 = hx : x4 = 1i are all 1-

dimensional, and four of these are x 7→ i, x 7→ −1, x 7→ −i, x 7→ 1.
(The two x 7→ ±i are faithful.)

In general, Cm = hx : xm = 1i has precisely m irreducible complex representations, all

of degree 1. Put ω = e2πi/m ∈ µm and define ρk by ρk : xj 7→ ω jk (0 6 j, k 6 m − 1).

It turns out that all irreducible complex representations of a finite abelian group are
1-dimensional: (1.4*) or see (4.4) below.

4
 (ii) G = D6 = hx, y : x3 = y 2 = 1, yxy −1 = x−1 i, the smallest non-abelian finite group.
G∼= S3 (generated by a 3-cycle and a 2-cycle).

G has the following irreducible complex representations:

2 of degree 1 : ρ1 : x 7→ 1, y 7→ 1
ρ2 : x 7→ 1, y 7→ −1
   
ω 0 0 1
1 of degree 2 : ρ3 : x 7→ , y →
7 , where ω = e2πi/3 ∈ µ3
0 ω −1 1 0

This follows easily later on. For now, by brute force. . .

Define u0 = 1 + x + x2 , v0 = u0 y,
u1 = 1 + ω 2 x + ωx2 , v1 = u1 y,
u2 = 1 + ωx + ω 2 x2 , v2 = u2 y.

Check easily xu1 = x + ω 2 x2 + ω = ωu1 , and in general xui = ω i ui (0 6 i 6 2). (I.e., in

the action of x, ui is an eigenvector, of eigenvalue ω i .) So hui i, hvi i are Chxi-modules.

Also: yu0 = v0 , yv0 = u0 , 
yu1 = v2 , yv1 = u2 , So hu0 , v0 i, hu1 , v2 i, hu2 , v1 i are Chyi-modules,
 and hence are all CG-submodules.
yu2 = v1 , yv2 = u1 .

Note, U3 = hu1 , v2 i, U4 = hu2 , v1 i are irreducible and hu0 , v0 i has U1 = hu0 + v0 i and
U2 = hu0 − v0 i as CG-submodules.

Moreover, CD6 = U1 ⊕ U2 ⊕ U3 ⊕ U4 .
| {z }
ր ↑ տ
trivial non-trivial isomorphic via u1 7→ v1 , v2 7→ u2

Challenge: repeat this for D8 . (See James & Liebeck, p.94.)

(2.18) Definition. We say that ρ : G → GL(V ) is decomposable if there are G-invariant

subspaces U, W with V = U ⊕ W . Say ρ is a direct sum ρU ⊕ ρW . If no such exists, we say
ρ is indecomposable.

(U, W must have G-actions on them, not just ordinary vector subspaces.)

(2.19) Lemma. Suppose ρ : G → GL(V ) is a decomposition with G-invariant decomposition

V = U ⊕ W . If B is a basis {u1 , . . ., uk , w1 , . . ., wl } consisting of a basis B1 of U and B2 of
W , then with respect to B,
   
* 0 [ρU (g)]B1 0
ρ(g)B = =
0 * 0 [ρW (g)]B2

(2.20) Definition. ρ : G → GL(V ), ρ′ : G → GL(V ′ ). The direct sum of ρ, ρ′ is

ρ ⊕ ρ′ : G → GL(V ⊕ V ′ ), (ρ ⊕ ρ′ )(g)(v1 + v2 ) = ρ(g)v1 + ρ′ (g)v2

– a block diagonal action.

For matrix representations, R : G → GLn (F ), R′ : G → GLn′ (F ), define

 
′ R(g) 0
R ⊕ R : G → GLn+n′ (F ), g 7→ , ∀g ∈ G.
0 R′ (g)

5
3. Complete Reducibility and Maschke’s Theorem
G, F as usual.

(3.1) Definition. The representation ρ : G → GL(V ) is completely reducible, or

semisimple, if it is a direct sum of irreducible representations.

Evidently, simple ⇒ completely reducible, but not conversely.

(3.2) Examples. Not all representations are completely reducible.

  
1 0
(i) G = : n ∈ Z , V = C2 , natural action (gv is matrix multiplication).
n 1
V is not completely reducible. (Note G not finite.)
 
1 0
(ii) G = Cp , F = Fp . xj 7→ (0 6 j 6 p − 1) defines a representation G → GL2 (F ).
j 1
V = (v1 , v2 ) where xj v1 = v1 , xj v2 = jv1 + v2 . Define W = (v1 ). W is an F Cp -module
but there is no X s.t. V = W ⊕ X. (Note F 6= R, C.)

(3.3) Theorem (Complete Reducibility Theorem). Every finite-dimensional represen-

tation of a finite group over a field of characteristic 0 is completely reducible.

Enough to prove:

(3.4) Theorem (Maschke’s Theorem). G finite, ρ : G → GL(V ) with V an F -space,

char F = 0. If W is a G-subspace of V then there exists a G-subspace U of V such
that V = W ⊕ U (a direct sum of G-subspaces).

Proof 1. Let W ′ be any vector space complement of W in V , i.e. V = W ⊕ W ′ . Let

q : V → W be the projection of V onto W along W ′ , i.e. if v = w + w′ then q(v) = w.

1 X
Define q : v 7→ ρ(g)q(ρ(g −1 )v), the ‘average of q over G’. Drop the ρs.
|G|
g∈G

Claim (i). q : V → W .

For v ∈ V , q(g −1 v) ∈ W and gW ⊆ W .

Claim (ii). q(w) = w for w ∈ W .

1 X 1 X 1 X
q(w) = gq(g −1 w ) = g(g −1 w) = w = w.
|G| | {z } |G| |G|
g∈G g∈G g∈G
∈W

So (i), (ii) ⇒ q projects V onto W .

Claim (iii). If h ∈ G then hq(v) = q(hv) (for v ∈ V ).

1 X 1 X 1 X

hq(v) = h gq(g −1 v) = hgq(g −1 v) = (hg)q (hg)−1 hv
|G| |G| |G|
g∈G g∈G g∈G
1 X

′ ′−1
= g q g (hv) = q(hv).
|G| ′
g ∈G

6
 Claim (iv). ker q is G-invariant.

If v ∈ ker q, h ∈ G, then hq(v) = 0 = q(hv), so hv ∈ ker q.

Then V = im q ⊕ ker q = W ⊕ ker q is a G-subspace decomposition. 2

Remark. Complements are not necessarily unique.

The second proof uses inner products, hence we need to take F = C (or R), and it can be
generalised to compact groups (chapter 15).

Recall for V a C-space, h , i is a C-inner product if

(a) hw, vi = hv, wi for all v, w

(b) linear in RHS
(c) hv, vi > 0 if v 6= 0

Additionally, h , i is G-invariant if

(d) hgv, gwi = hv, wi for all v, w ∈ V, g ∈ G

Note that if W is a G-subspace of V (with G-invariant inner product) then W ⊥ is also

G-invariant and V = W ⊕ W ⊥ .

Proof. Want: for all w ∈ W ⊥ , for all g ∈ G, we have gv ∈ W ⊥ .

Now, v ∈ W ⊥ ⇔ hv, wi = 0 for all w ∈ W . Thus hgv, gwi = 0 for all g ∈ G, w ∈ W .

Hence hgv, w′ i = 0 for all w′ ∈ W since we can take w = g −1 w′ by G-invariance of W .
Hence gv ∈ W ⊥ since g was arbitrary. 2

Hence if there is a G-invariant inner product on any complex G-space, we get:

(3.4*) (Weyl’s Unitary Trick). Let ρ be a complex representation of the finite group G
on the C-space V . There is a G-invariant inner product on V (whence ρ(G) is conjugate
to a subgroup of U (V ), the unitary group on V , i.e. ρ(g)∗ = ρ(g −1 )).

Proof. There is an inner product on V : take basis e1 , . . ., en , and define (ei , ej ) = δij ,
1 X
extended sesquilinearly. Now define hv, wi = (gv, gw).
|G|
g∈G

Claim. h , i is sesquilinear, positive definite, and G-invariant.

1 X
1 X ′
If h ∈ G, hhv, hwi = (gh)v, (gh)w = (g v, g ′ w) = hv, wi. 2
|G| |G| ′
g∈G g ∈G

(3.5) (The (left) regular representation of G.) Define the group algebra of G to be the
F -space F G = span{eg : g ∈ G}.
P P  P 
There is a G-linear action: for h ∈ G, h g ag e g = ag ehg = g′ ah−1 g′ eg′ .

ρreg is the corresponding representation – the regular representation of G.

This is faithful of dimension |G|.

7
It turns out that every irreducible representation of G is a subrepresentation of ρreg .

(3.6) Proposition. Let ρ be an irreducible representation of the finite group G over a field
of characteristic 0. Then ρ is isomorphic to a subrepresentation of ρreg .

Proof. Take ρ : G → GL(V ), irreducible, and let 0 6= v ∈ V .

P P
Let θ : F G → V , g ag eg 7→ ag gv (a G-homomorphism).
տ really ρ(g)

Now, V is irreducible and im θ = V (since im θ is a G-subspace). Then ker θ is a

G-subspace of F G. Let W be a G-complement of ker θ in F G (using (3.4)), so that
W < F G is a G-subspace and F G = ker θ ⊕ W .

Hence W ∼
= F G/ ker θ ∼
= im θ = V . 2
տ
G-isom.

More generally,

(3.7)
PLet G act on a set X. Let F X = span{ex : x ∈ X}, with G-action

P Definition.
g x∈X ax ex = ax egx .

So we have a G-space F X. The representation G → GL(V ) with V = F X is the correspond-

ing permutation representation.

4. Schur’s Lemma
(4.1) Theorem (‘Schur’s Lemma’). (a) Assume V, W are irreducible G-spaces (over a
field F ). Then any G-homomorphism θ : V → W is either 0 or is an isomorphism.

(b) Assume F is algebraically closed and let V be an irreducible G-space. Then any
G-endomorphism θ : V → V is a scalar multiple of the identity map idV (a homothety).

Proof. (a) Let θ : V → W be a G-homomorphism. Then ker θ is a G-subspace of V , and

since V is irreducible either ker θ = 0 or ker θ = V . And im θ is a G-subspace of W ,
so as W is irreducible, im θ is either 0 or W . Hence either θ = 0 or θ is injective and
surjective, so θ is an isomorphism.

(b) Since F is algebraically closed, θ has an eigenvalue λ. Then θ − λid is a singular

G-endomorphism on V , so must be 0, so θ = λid. 2

(4.2) Corollary. If V, W are irreducible complex G-spaces, then


1 if V, W are G-isomorphic
dimC HomG (V, W ) =
0 otherwise

Proof. If V, W are not isomorphic then the only G-homomorphism V → W is 0 by (4.1).

Assume V ∼
=G W and θ1 , θ2 ∈ HomG (V, W ), both 6= 0. Then θ2 is irreducible by (4.1)
and θ2−1 θ1 ∈ HomG (V, V ). So θ2−1 θ1 = λid for some λ ∈ C. Then θ1 = λθ2 . 2

(4.3) Corollary. If G has a faithful complex irreducible representation then Z(G) is cyclic.

Remark. The converse is false. (See examples sheet 1, question 11.)

8
Proof. Let ρ : G → GL(V ) be a faithful irreducible complex representation. Let z ∈ Z(G),
so zg = gz for all g ∈ G.

Consider the map φz : v 7→ zv for v ∈ V . This is a G-endomorphism on V , hence is

multiplication by a scalar µz , say (by Schur).

Then the map Z(G) → C× , z 7→ µz , is a representation of Z and is faithful (since ρ

is). Thus Z(G) is isomorphic to a finite subgroup of C× , hence is cyclic. 2

Applications to abelian groups

(4.4) Corollary. The irreducible complex representations of a finite abelian group G are all
1-dimensional.

Proof. Either (1.4*) to invoke simultaneous diagonalisation: if v is an eigenvector for each

g ∈ G and if V is irreducible, then V = (v).

Or let V be an irreducible complex representation. For g ∈ G, the map θg : V → V ′ ,

v 7→ gv, is a G-endomorphism of V and, as V is irreducible, θg = λg id for some λg ∈ C.
Thus gv = λg v for any g. Thus, as V is irreducible, V = (v) is 1-dimensional. 2

Remark. This fails on R. E.g., C3 has two irreducible real representations: one of dimension
1, one of dimension 2. (See sheet 1, question 12.)

Recall that any finite abelian group G is isomorphic to a product of cyclic groups, e.g.
C6 ∼
= C2 × C3 . In fact, it can be written as a product of Cpα for various primes p and α > 1,
and the factors are uniquely determined up to ordering.

(4.5) Proposition. The finite abelian group G = Cn1 ×. . .×Cnr has precisely |G| irreducible
complex representations, as described below.

Proof. Write G = hx1 i × . . . × hxr i where |xj | = nj . Suppose ρ is irreducible – so by (4.4)

it’s 1-dimensional, ρ : G → C× .
n
Let ρ(1, . . ., 1, xj , 1, . . ., 1) = λj ∈ C× . Then λj j = 1, so λj is an nth
j root of unity.

Now the values (λ1 , . . ., λr ) determine ρ, as ρ(xj11 , . . ., xjrr ) = λj11 . . .λjrr .

n
Thus ρ ↔ (λ1 , . . ., λr ) with λj j = 1 for all j. (And have n1 . . .nr such r-tuples, each
giving a 1-dimensional representation.

Examples. (a) G = C4 = hxi. 1 x x2 x3

ρ1 1 1 1 1
ρ2 1 i −1 i
ρ3 1 −1 1 −1
ρ4 1 −i −1 i

(b) G = V4 = hx1 i × hx2 i ∼

= C2 × C2 . 1 x1 x2 x1 x2
ρ1 1 1 1 1
ρ2 1 1 −1 −1
ρ3 1 −1 1 −1
ρ4 1 −1 −1 1

9
Warning. There is no ‘natural’ 1-1 correspondence between the elements of G and the
representations of G. If you choose an isomorphism G ∼= C1 × . . . × Cr , then you can identify
the two sets, but it depends on the choice of isomorphism.

** Non-examinable section **

Application to isotypical decompositions

(4.6) Proposition. Let V be a G-space over C, and assume V = U1 ⊕ . . . ⊕ Un = W1 ⊕ . . . ⊕
Wn , with all the Uj , Wk irreducible G-spaces. Let X be a fixed irreducible G-space.
Let U be the sum of all the Uj isomorphic to X, and W be the sum of all the Wj
isomorphic to X.

Then U = W , and is known as the isotypical component of V corresponding to X.

Hence:

# Uj isomorphic to X = # Wk isomorphic to X = (V : X) = multiplicity of X in V .

j i k π
Proof (sketch). Look at θjk : Uj −→ V −→ Wk , with Wk ∼ = X, where ij is inclusion and πk
is projection. If Uj ∼
= X then U j ⊂ W – all the projections to the other Wl are 0.

‘Then fiddle around with dimensions, then done.’ 2

(4.7) Proposition. (V : X) = dimC HomG (V, X) for X irreducible, V any G-space.

Proof. Prove HomG (W1 ⊕ W2 , X) = ∼ HomG (W1 , X) ⊕ HomG (W2 , X), then apply Schur.
(Or see James & Liebeck 11.6.)

(4.8) Proposition. dimC HomG (CG, X) = dimC X.

Proof. Let d = dim X, and take a basis {e1 , . . ., ed } of X. Define φi : CG → X, g 7→ gei

(1 6 i 6 d). Then φi ∈ HomG (CG, X) and {φ1 , . . ., φd } is a basis. (See James &
Liebeck 11.8.)

Remark. If V1 , . . ., Vr are all the distinct complex irreducible G-spaces then CG = n1 V1 ⊕

. . .⊕nr Vr where ni = dim Vi . Then |G| = n21 +. . .+n2r . (See (5.9), or James & Liebeck 11.2.)

Recall (2.17). G = D6 , CG = U1 ⊕ U3 ⊕ U3 ⊕ U4 , dim Hom(CG, U3 ) = 2. (Challenge: find a

basis for it.) U1 and U2 occur with multiplicity 1, and U3 occurs with multiplicity 2.

** End of non-examinable section **

5. Character Theory
We want to attach invariants to a representation ρ of a finite group G on V . Matrix coefficients
of ρ(g) are basis dependent, so not true invariants.

Take F = C, and G finite. ρ = ρV : G → GL(V ), a representation of G.

(5.1) Definition. The character χρ = χV = χ is defined as χ(g) = tr ρ(g) (= tr R(g),

where R(g) is any matrix representation of ρ(g) with respect to any basis). The degree of
χV is dim V .

10
Thus χ is a function G → C. χ is linear if dim V = 1, in which case χ is a homomorphism
G → C× .

• χ is irreducible if ρ is.
• χ is faithful if ρ is.
• χ is trivial (principal) if ρ is the trivial representation: write χ = 1G .

χ is a complete invariant in the sense that it determines ρ up to isomorphism – see (5.7).

(5.2) First properties.

(i) χV (1) = dim V

(ii) χV is a class function, viz it is conjugation invariant, i.e. χV (hgh−1 ) = χV (g)
for all g, h ∈ G.
Thus χV is constant on the conjugacy classes (ccls) of G.
(iii) χV (g −1 ) = χV (g).
(iv) For two representations, V, W , have χV ⊕W = χV + χW .

Proof. (i) tr (I) = n.

(ii) χ(hgh−1 ) = tr (Rh Rg Rh−1 ) = tr (Rg ) = χ(g).
(iii) g ∈ G has finite order, so by (1.4) can assume ρ(g) is represented by a diagonal ma-
! !
λ1 P λ1−1
trix . .. . Thus χ(g) = −1
λi . Now g is represented by . .. .
λn λ−1
n
P P P
and χ(g −1 ) = λ−1 i = λi = λi = χ(g).
(iv) Suppose V = V1 ⊕ V2 , ρi : G → GL(Vi ), ρ : G → GL(V ). Take basis B = B1 ∪ B2
of V , containing bases Bi of Vi .  
[ρ1 (g)]B1 0
With respect to B, ρ(g) has matrix .
0 [ρ2 (g)]B2
So χ(g) = tr (this) = tr ρ1 (g) + tr ρ2 (g) = χ1 (g) + χ2 (g). 2

Remark. We see later that χ1 , χ2 characters of G ⇒ χ1 χ2 also a character of G. This uses

tensor products – see (9.6).

(5.3) Lemma. Let ρ : G → GL(V ) be a complex representation affording the character χ.

Then |χ(g)| 6 χ(1), with equality iff ρ(g) = λid for some λ ∈ C, a root of unity.
Moreover, χ(g) = χ(1) ⇔ g ∈ ker ρ.
!
λ1
Proof. Fix g. W.r.t. a basis of V of eigenvectors of ρ(g), the matrix of ρ(g) is .. .
.
λn
P P P
Hence, |χ(g)| = | λi | 6 |λi | = 1 = dim V = χ(1), with equality iff all λj are
equal to λ, say. And if χ(g) = χ(1) then ρ(g) = λid.

Therefore, χ(g) = λχ(1), and so λ = 1 and g ∈ ker ρ. 2

(5.4) Lemma. If χ is a complex irreducible character of G, then so is χ, and so is εχ for

any linear character ε of G.

Proof. If R : G → GLn (C) is a complex (matrix) representation then so is R : G → GLn (C),

g 7→ R(g).

Similarly for R′ : g 7→ ε(g)R(g). Check the details. 2

11
(5.5) Definition. C(G) = {f : G → C : f (hgh−1 ) = f (g) ∀ h, g ∈ G}, the C-space of class
functions. (Where f1 + f2 : g 7→ f1 (g) + f2 (g), λf : g 7→ λf (g).)

List conjugacy classes as C1 (= {1}), C2, . . ., Ck . Choose g1 (= 1), g2 , . . ., gk as representatives

of the classes.

Note also that dimC C(G) = k, as the characteristic functions δj of the conjugacy classes for
a basis, where δj (g) = 1 if g ∈ Cj , and 0 otherwise.

Define Hermitian inner product on C(G) by

k k
1 X 1 X X 1
hf, f ′ i = f (g)f ′ (g) = |Cj | f (gj )f ′ (gj ) = f (gj )f ′ (gj )
|G| |G| j=1 j=1
|CG (g j )|
g∈G

using orbit-stabilier: |C| = |G : CG (x)|.

k
X 1 χ(gj−1 )χ′ (gj ) is a real symmetric form.
For characters, hχ, χ′ i =
j=1
|CG (gj )|

Main result follows.

(5.6) Big Theorem (completeness of characters). The C-irreducible characters of G

form an orthonormal basis of the space of class functions of G. Moreover,

(a) If ρ : G → GL(V ), ρ′ : G → GL(V ′ ) are irreducible representations of G affording

characters χ, χ′ then

χ χ ′ 1 if ρ, ρ′ are isomorphic
h , i=
0 otherwise
(b) Each class function of G can be expressed as a linear combinations of irreducible
characters of G.

Proof. In chapter 6.

(5.7) Corollary. Complex representations of finite groups are characterised by their char-
acters.

Proof. Have ρ : G → GL(V ) affording χ. (G finite, F = C.) Complete reducibility (3.3)

says ρ = m1 ρ1 ⊕ . . . ⊕ mk ρk , where ρj is irreducible and mj > 0.

Then mj = hχ, χj i where χj is afforded by ρj . Then χ = m1 χ1 + . . . + mk χk and

hχ, χj i = hm1 χ1 + . . . + mk χk , χj i = mj , by (5.6)(a). 2

(5.8) Corollary (irreducibility criterion) If ρ is a complex representation of G affording

χ then ρ irreducible ⇔ hχ, χi = 1.

Proof. (⇒) orthogonality.

P
χ χ χ mj χj , for χj irreducible, mj > 0. Then
P 2Assume h , i = 1. (3.3) says =
(⇐)
mj = 1, so χ = χj for some j. Therefore χ is irreducible.

(5.9) Theorem. If the irreducible

P complex representations of G, ρ1 , . . ., ρk , have dimensions
n1 , . . ., nk , then |G| = i n2i .

(Recall end of chapter 4.)

12
Proof. Recall from (3.5), ρreg : G → GL(CG), the regular representation of G, of dimension
|G|. Let πreg be its character.

Claim. πreg (1) = |G| and πreg (h) = 0 if h 6= 1.

Proof. Easy. Let G = {g1 , . . ., gn } and take h ∈ G, h 6= 1. For 1 6 i 6 n, hgi = gj ,

some j 6= i, so ith row of [ρreg (h)]B has 0s in every place, except column j – in
particular, the (i, i)th entry is 0 for all i. Hence πreg (h) = tr [ρreg (h)]B = 0.
P
By claim, πreg = nj χj with nj = χj (1) :

1 X 1
nj = hπreg , χj i = πreg (g)χj (g) = |G| χj (1) = χj (1)
|G| |G|
g∈G 2

(5.10) Corollary. The number of irreducible characters of G (up to equivalence) equals k,

the number of conjugacy classes.

(5.11) Corollary. Elements g1 , g2 ∈ G are conjugate iff χ(g1 ) = χ(g2 ) for all irreducible
characters of G.

Proof. (⇒) Characters are class functions.

(⇐) Let δ be the characteristic function of the the class of gi . Then δ is a class function,
so can be written as a linear combination of the irreducible characters of G, by (5.6)(b).
Hence δ(g2 ) = δ(g1 ) = 1. So g2 ∈ CG (g1 ). 2

Recall from (5.5) the inner product on C(G) and the real symmetric form h , i for characters.

(5.12) Definition. G finite, F = C. The character table of G is the k × k matrix X =

[χi (gj )] where χ1 (= 1), χ2 , . . ., χk are the irreducible characters of G, and C1 (= {1}), C2, . . ., Ck
are the conjugacy classes, with gj ∈ Cj .

I.e., the (i, j)th entry of X is χi (gj ).

Examples. C2 = hx : x2 = 1i C3 = hx : x3 = 1i

1 x 1 x x2
χ1 1 1 χ1 1 1 1
χ2 χ2 where ω = e2πi/3 ∈ µ3 .
1 −1 1 ω ω2
χ3 1 ω2 ω

G = D6 = ha, b : a3 = b2 = 1, bab−1 = a−1 i ∼

= S3 .

In (2.17) we found a complete set of non-isomorphic irreducible CG-modules: U1 , U2 , U3 .

Let χi = χUi , (1 6 i 6 3).

1 {a, a2 } {b, ab, a2b} ← gj Orthogonality:

χ1 1 1 1
χ2 1 1 −1 1 × 2 (−1)(1)
χ3 + = 0 (rows 2 & 3)
2 −1 0 6 3
տ
12 +12 +22 =6

6 3 2 ← |CG (gj )| 22 (−1)2

+ = 1 (row 3)
1 2 3 ← |C(gj )| 6 3

13
6. Proofs and Orthogonality
We want to prove (5.6), the Big Theorem. We’ll do this in two ways.

Proof 1 of (5.6)(a). Fix bases of V and V ′ . Write R(g), R′ (g) for the matrices of ρ(g),
ρ′ (g) with respect to these, respectively.
1 X χ′ −1 χ 1 X
hχ′ , χi = (g ) (g) = R′ (g −1 )ii R(g)jj
|G| |G|
g∈G
16i6n′
16j6n

1 X ′ −1
Let φ : V → V ′ be linear, and define φaverage = φ̃ : V → V ′ , v 7→ ρ (g )φρ(g)v.
|G|
g∈G

Then φ̃ is a G-homomorphism. For if h ∈ G,

1 X ′

1 X ′ ′−1
ρ′ (h−1 )φ̃ρ(h)(v) = ρ (gh)−1 φ ρ(gh) (v) = ρ (g )φρ(g ′ )(v) = φ̃(v).
|G| |G| ′
g∈G g ∈G

Assume first that ρ, ρ′ are not isomorphic. Schur’s Lemma says φ̃ = 0 for any linear
φ : V → V ′.

Let φ = εαβ having matrix Eαβ (with respect to our basis), namely 0 everywhere except
1 in the (α, β)th place.

1 X ′ −1

Then ε̃αβ = 0, so R (g )Eαβ R(g) ij = 0.
|G|
g∈G

1 X ′ −1
Thus R (g )iα R(g)βj = 0 for all i, j.
|G|
g∈G

1 X ′ −1
With α = i, β = j, R (g )ii R(g)jj = 0. Sum over i, j and conclude hχ′ , χi = 0.
|G|
g∈G

Now assume that ρ, ρ′ are isomorphic, so χ = χ′ . Take V = V ′ , ρ = ρ′ . If φ : V → V

is linear, then φ̃ ∈ HomG (V, V ).

1 X
1 X
Now tr φ = tr φ̃, as tr φ̃ = tr ρ(g −1 )φρ(g) = tr φ = tr φ.
|G| |G|

1
By Schur, φ̃ = λid for some λ ∈ C (depending on φ). Now λ = tr φ.
n
1 1 X
Let φ = εαβ , so tr φ = δαβ . Hence ε̃αβ = δαβ id = ρ(g −1 )εαβ ρ(g).
n |G| g

1 X 1
In terms of matrices, take the (i, j)th entry: R(g −1 )iα R(g)βj = δαβ δij ,
|G| g n

1 X 1
and put α = i, β = j to get R(g −1 )ii R(g)jj = δij .
|G| g n

Finally sum over i, j: hχ, χi = 1. 2

14
Before proving (b), let’s prove column orthogonality, assuming (5.10).

k
X
(6.1) Theorem (column orthogonality). χi (gj )χi (gl ) = δjl |CG (gj )|.
i=1

This has an easy corollary:

k
X
(6.2) Corollary. |G| = χ2i (1).
i=1

X 1
Proof of (6.1). δij = hχi , χj i = χi (gl )χj (gl ).
|Cg (gl )|
l

Consider the character table X = (χi (gj )).

!
|CG (g1 )|
Then XD −1 t
X = Ik×k , where D = .. .
.
|CG (gk )|

t t
As X is a square matrix, it follows that D−1 X is the inverse of X. So X X = D. 2

Proof of (5.6)(b). List all the irreducible characters χ1 , . . ., χl of G. It’s enough to show
that the orthogonal complement of span{χ1 , . . ., χl } in C(G) is 0.

To see this, assume f ∈ C(G) with hf, χj i = 0 for all irreducible χj .

Let ρ : G → GL(V ) be irreducible affording χ ∈ {χ1 , . . ., χl }. Then hf, χi = 0.

1 X
Consider f (g)ρ(g) : V → V . This is a G-homomorphism, so as ρ is irreducible
|G|
it must be λid for some λ ∈ C (by Schur).

1 X 1 X
Now, nλ = tr f (g)ρ(g) = f (g)χ(g) = 0 = hf, χi.
|G| |G|
P
So λ = 0. Hence f (g)ρ(g) = 0, the zero endomorphism on V , for all representations
ρ. Take ρ = ρreg , where ρreg (g) : e1 7→ eg (g ∈ G), the regular representation.
P P P
So g f (g)ρreg (g) : e1 7→ g f (g)eg . It follows that f (g)eg = 0.

Therefore f (g) = 0 for all g ∈ G. And so f = 0. 2

Various important corollaries follow from this:

• # irreducibles of G = # conjugacy classes (5.10)

• column orthogonality (6.1)
P
• |G| = χ2i (1) (6.2)
• irreducible χ is constant on conjugacy classes (5.11)
• if g ∈ G, then g, g −1 are G-conjugate ⇔ χ(g) ∈ R for all irreducible χ.

15
Example. 6 3 2 ←|CG (gj )|
1 a b ← gj
χ1 1 1 1
χ2 1 1 −1
χ3 2 −1 0 ← coming from operations on equilateral triangle

3
X
Column orthogonality: χi (gr )χi (gs ).
i=1

r = 1, s = 2 : 1.1 + 1.1 + 2(−1) = 0 r= 6 s

r = 1, s = 3 : 1.1 + 1(−1) + 2.0 = 0 r 6= s
r = 2, s = 2 : 1.1 + 1.1 + (−1)(−1) = 3 r = s, weight by |CG (gr )|

** Non-examinable section **

Proof 2 of (5.6)(a). (Uses starred material at the end of chapter 4.)

Lm
X irreducible G-space, V any G-space. V = i=1 Ui , with Ui irreducible.

Then #Uj isomorphic to X is independent of the decomposition. We wrote (V : X)

for this number, and in (4.7) we observed (V : X) = dimC HomG (V, X) (∗).

Let ρ : G → GL(U ) have character χ Write U G = {u ∈ U : ρ(g)u = u ∀ g ∈ G}, the

G-invariants of U .
1 X
Consider the map π : U → U , u 7→ ρ(g)u.
|G| g

This is a projection onto U G (because it’s a G-homomorphism, andPwhen restricted

1
to U G it acts as the identity there). Verify dim U G = tr π = |G| χ
g (g) (∗∗) (by
decomposing U and looking at bases).

Now choose U =
HomC (V, V ′ ) with V, V ′ being G-spaces. G acts on U via g.θ(v) =
ρV (g) θρV ′ (g −1 )v for θ ∈ U .

G 1 Xχ
But HomG (V, V ′ ) = HomC (V, V ′ ) , so by (∗∗), dimC HomG (V, V ′ ) = V (g).
|G| g

Finally, show χV (g) = χV ′ (g −1 )χV (g) – see section on tensor products in chapter 9.

The orthogonality of the irreducible characters now follows from (∗). 2

** End of non-examinable section **

16
7. Permutation Representations
Preview was given in (3.7). Recall:

• G finite, acting on finite set X = {x1 , . . ., xn }.

P
• CX = C-space, basis {ex1 , . . ., exn } of dimension |X|. CX = j aj e x j : aj ∈ C .
• corresponding permutation representation,PρX : G → GL(CX),
P g 7→ ρ(g), where ρ(g) :
exj 7→ egxj , extended linearly. So ρX (g) : x∈X ax ex 7→ x∈X ax egx .
• ρX is the permutation representation corresponding to the action of G on X.
• matrices representing ρX (g) with respect to the basis {ex }x∈X are permutation ma-
trices: 0 everywhere except one 1 in each row and column, and (ρ(g))ij = 1 precisely
when gxj = xi .

(7.1) Permutation character πX is πX (g) = |fixX (g)| = |{x ∈ X : gx = x}|.

(7.2) πX always contains 1G . P Px1 + . . . + exn ) is a trivial G-subspace of CX with

For: span(e
G-invariant complement span( ax ex : ax = 0).

(7.3) ‘Burnside’s Lemma’ (Cauchy, Frobenius). hπX , 1i = # orbits of G on X.

Proof. If X = X1 ∪ . . . ∪ Xl , a disjoint union of orbits, then πX = πX1 + . . . + πXl with πXj

the permutation character of G on Xj . So to prove the claim, it’s enough to show that
if G is transitive on X then hπX , 1i = 1.

So, assume G is transitive on X. Then

1 X
hπX , 1i = πX (g)
|G|
g∈G
1 
= (g, x) ∈ G × X : gx = x
|G|
1 X
= |Gx |
|G|
x∈X
1 1
= |X| |Gx | = |G| = 1
|G| |G| 2

(7.4) Let G act on the sets X1 , X2 . Then G acts on X1 × X2 via g(x1 , x2 ) = (gx1 , gx2 ).
The character πX1 ×X2 = πX1 πX2 and so hπX1 , πX2 i = # orbits of G on X1 × X2 .

Proof. hπX1 , πX2 i = hπX1 πX2 , 1i = hπX1 ×X2 , 1i = # orbits of G on X1 × X2 (by (7.3)). 2

(7.5) Let G act on X, |X| > 2. Then G is 2-transitive on X if G has just two orbits on
X × X, namely {(x, x) : x ∈ X} and {(x1 , x2 ) : xi ∈ X, x1 6= x2 }.

(7.6) Lemma. Let G act on X, |X| > 2. Then πX = 1 + χ with χ irreducible ⇔ G is

2-transitive on X.

Proof. πX = m1 1 + m2 χ2 + . . . + ml χl with 1, χ2 , . . ., χl distinct irreducibles and mi ∈ Z>0 .

P
Then hπX , πX i = li=1 m2i . Hence G is 2-transitive on X iff l = 2, m1 = m2 = 1. 2

(7.7) Sn acting on Xn (see 1.6) is 2-transitive. Hence πXn = 1 + χ with χ irreducible of

degree n − 1. Similarly for An (n > 3)

17
(7.8) Example. G = S4 .

Conjugacy classes correspond to different cycle types.

1 3 8 6 6 ← sizes
1 (12)(34) (123) (1234) (12) ← ccl reps
→
trivial χ1 1 1 1 1 1 o
two linear characters
→
sign χ2 1 1 1 −1 −1 since S4 /S4′ = C2
πX4 − 1 → χ3 3 −1 0 −1 1
πX3 × πX2 → χ4 3 −1 0 1 −1
χ5 d x y z w

Know: 24 = 1 + 1 + 9 + 9 + d2 ⇒ d = 2.

Column orthogonality: 1 + 1 − 3 − 3 + 2x = 0 ⇒ x=2

1 + 1 + 2y = 0 ⇒ y = −1
1 − 1 − 3 + 3 + 2z = 0 ⇒ z=0
1 − 1 + 3 − 3 + 2w = 0 ⇒ w=0

Or: χreg = χ1 + χ2 + 3χ3 + 3χ4 + 2χ5 ⇒ χ5 = 12 (χreg − χ1 − χ2 − 3χ3 − 3χ4 ).

Or: can obtain χ5 by observing S4 /V4 ∼

= S3 and ‘lifting’ characters – see chapter 8.

(7.9) Example. G = S5 .
1 15 20 24 10 20 30 ←|Cj |
1 (12)(34) (123) (12345) (12) (123)(45) (1234) ←gj
→
trivial χ1 1 1 1 1 1 1 1
→
sign χ2 1 1 1 1 −1 −1 −1
πX5 − 1 → χ3 4 0 1 −1 2 −1 0
πX3 ×πX2 → χ4 4 0 1 −1 −2 1 0
χ5 5 1 −1 0 −1 −1 1
χ6 5 1 −1 0 1 1 −1
χ7 6 −2 0 1 0 0 0

There are various methods to get χ5 , χ6 of degree 5.

One way is to note that if X = Syl5 (G) then |X| = 6 and one checks that hχX , χX i = 2.
Therefore πX − 1 is irreducible.
P
For χ7 , first d2i = 120 gives degχ7 = 6, and orthogonality for the remaining entries.


5
Or: let S5 act on the set of 2 unordered pairs of elements of {1, 2, 3, 4, 5}.

π(5) : 10 2 1 0 4 1 0
2


hχ(5) , χ(5) i = 3 
2 2 
hχ(5) , 1i = 1 ⇒ χ(5) = 1 + χ3 + ψ
2 
 2
hχ(5) , χ3 i = 1
2

ψ has degree 5 (and is actually χ6 in the table).

18
(7.10) Alternating groups.

Let g ∈ An . Then |CSn (g)| = |Sn : CSn (g)|

⊆
↑ An index 2 in Sn
|CAn (g)| = |An : CAn (g)|

but not necessarily equal: e.g., if σ = (123), then CAn (σ) = {σ}, but CSn (σ) = {σ, σ −1 }.

We know |Sn : An | = 2 and in fact:

(7.11) If g ∈ An then CSn (g) = CAn (g) precisely when g commutes with some odd permuta-
tion; otherwise it breaks up into two classes of equal size. (In the latter case, precisely
when the disjoint cycle decomposition of g is a product of odd cycles of distinct lengths.)

Proof. See James & Liebeck 12.17

(7.12) G = A4 .
1 3 4 4 ←|Cj |
1 (12)(34) (123) (123)−1 ←gj
1G → χ1 1 1 1 1
πX −1→ χ2 3 −1 0 0
χ3 1 1 ω ω2
χ4 1 1 ω2 ω
↑
d2i =12
P

Final two linear characters are found via G/G′ = G/V4 = C3 , by lifting – see chapter 9.

For A5 see Telemann chapter 11, or James & Liebeck 20.14.

8. Normal Subgroups and Lifting Characters

(8.1) Lemma. Let N
G, let ρ̃ : G/N → GL(V ) be a representation of G/N . Then
q ρ̃
ρ : G −→ G/N −→ GL(V ) is a representation of G, where ρ(g) = ρ̃(gN ) (and q is the
natural homomorphism). Moreover, ρ is irreducible if ρ̃ is.

The corresponding characters satisfy χ(g) = χ̃(gN ) for g ∈ G, and degχ = degχ̃. We
say that χ̃ lifts to χ. The lifting sending χ̃ 7→ χ is a bijection between

{irreducibles of G/N } ←→ {irreducibles of G with N in the kernel}

Proof. (See examples sheet 1, question 4.)

Note: χ(g) = tr (ρ(g)) = tr (ρ̃(gN )) for all g, and χ(1) = χ̃(N ), so degχ = degχ̃.

Bijection. If χ̃ is a character of G/N and χ if a lift to G then χ̃(N ) = χ(1). Also, if

k ∈ N then χ(k) = χ̃(kN ) = χ̃(N ) = χ(1). So N 6 ker χ.

Now let χ be a character of G with N 6 ker χ. Suppose ρ : G → GL(V ) affords χ.

Define ρ̃ : G/N → GL(V ), gN 7→ ρ(g) for g ∈ G. This is well-defined (as N 6 ker χ)
and ρ̃ is a homomorphism, hence a representation of G/N . If χ̃ is the character of ρ̃
then χ̃(gN ) = χ(g) for all g ∈ G.

Finally, check irreducibility is preserved. 2

19
Definition. The derived subgroup of G is G′ = h[a, b] : a, b ∈ Gi, where [a, b] = aba−1 b−1
is the commutator of a and b. (G′ is a crude measure of how abelian a group is.)

(8.2) Lemma. G′ is the unique minimal normal subgroup of G such that G/G′ is abelian.
(I.e., G/N abelian ⇒ G′ 6 N , and G/G′ is abelian.)

G has precisely l = |G/G′ | representations of degree 1, all with kernel containing G′

and obtained by lifting from G/G′ .

Proof. G′
G – easy exercise.

Let N
G. Let g, h ∈ G. Then g −1 h−1 gh ∈ N ⇔ ghN = hgN ⇔ (gN )(hN ) =

(hN )(gN ). So G′ 6 N ⇔ G/N abelian. Since G′
G, G/G′ is an abelian group.

By (4.5), G/G′ has exactly l irreducible characters, χ1 , . . ., χl , all of degree 1. The

lifts of these to G also have degree 1 and by (8.1) these are precisely the irreducible
characters χi of G such that G′ 6 ker χi .

But any linear character χ of G is a homomorphism χ : G → C× , hence χ(ghg −1 h−1 ) =

χ(g)χ(h)χ(g −1 )χ(h−1 ) = 1.

Therefore G′ 6 ker χ, so the χ1 , . . ., χl are all irreducible characters of G. 2

Examples. (i) Let G = Sn . Show G′ = An . Thus G/G′ ∼

= C2 . So Sn must have exactly
two linear characters.

(ii) G = A4 .

1 (12)(34) (123) (123)−1 1 x x2

1G 1 1 1 1 χ1 1 1 1
χ2 (ω = e2πi/3 )
χ2 1 1 ω ω2 1 ω ω2
k
χ3 1 1 ω2 ω χ3 1 ω2 ω
g
χ4
c
3 −1 0 0

∼ ∼
Let N = {1, (12)(34), (13)(24), (14)(23)} 6 G. In fact, N = V4 , N
G, and G/N = C3 .

Also, G′ = V4 , so G/G′ ∼
= C3 .

(8.3) Lemma. G is not simple iff χ(g) = χ(1) for some irreducible character χ 6= 1G and
1 6= g ∈ G. Any normal\subgroup of G is the intersection of kernels of some of the
irreducibles of G, N = ker χi .
χi irred

Proof. If χ(g) = χ(1) for some non-principal character χ (afforded by ρ, say), then g ∈ ker ρ
(by (5.3)). Therefore if g 6= 1 then 1 6= ker ρ
G.

If 1 6= N
G, take an irreducible χ̃ of G/N (χ̃ 6= 1G/N ). Lift to get an irreducible χ

afforded by ρ of G, then N 6 ker ρ
G. Therefore χ(g) = χ(1) for g ∈ N .

In fact, if 1 6= N
G then N is the intersection of the kernels of the lifts of all irreducibles
of G/N . 6 is clear. For >: if g ∈ G \ N then gN 6= N , so χ̃(gN ) 6= χ̃(N ) for some
irreducible χ̃ of G/N , and then lifting χ̃ to χ we have χ(g) 6= χ(1). 2

20
9. Dual Spaces and Tensor Products of Representations
Recall (5.5), (5.6): C(G) = C-space of class functions of G, dimC C(G) = k, basis χ1 , . . .χk
the irreducible characters of G.

• (f1 + f2 )(g) = f1 (g) + f2 (g)

• (f1 f2 )(g) = f1 (g)f2 (g)
• ∃ involution (homomorphism of order 2) f 7→ f ∗ where f ∗ (g) = f (g −1 )
• ∃ inner product h , i

Duality

(9.1) Lemma. Let ρ : G → GL(V ) be a representation over F and let V ∗ = HomF (V, F ),
the dual space of V .

Then V ∗ is a G-space under ρ∗ (g)φ(v) = φ(ρ(g −1 )v), the dual representation of ρ.

Its character is χρ∗ (g) = χρ (g −1 ).





Proof. ρ∗ (g1 ) ρ∗ (g2 )φ (v) = ρ∗ (g2 )φ ρ(g1−1 )v = φ ρ(g2−1 )ρ(g1−1 )v



= φ ρ(g1 g2 )−1 (v) = ρ∗ (g1 g2 )φ (v)

Character. Fix g ∈ G and let e1 , . . ., en be a basis of V of eigenvectors of ρ(g), say

ρ(g)ej = λj ej . let ε1 , . . ., εn be the dual basis.



Then ρ∗ (g)εj = λ−1 ∗
j εj , for ρ (g)εj (ei ) = εj ρ(g
−1
)ei = εj λ−1 −1
j ei = λj εj ei for all i.

P
Hence χρ∗ (g) = λ−1 χ −1 ).
j = ρ (g 2

(9.2) Definition. ρ : G → GL(V ) is self-dual if V ∼

= V ∗ (as an isomorphism of G-spaces).
Over F = C, this holds iff ρ (g) = ρ (g ), and since this = χρ (g), it holds iff χρ (g) ∈ R for
χ χ −1

all g.

Example. All irreducible representations of Sn are self-dual: the conjugacy classes are
determined by cycle types, so g, g −1 are always Sn -conjugate. Not always true for An : it’s
okay for A5 , but not for A7 – see sheet 2, question 8.

Tensor Products
V and W , F -spaces, dim V = m, dim W = n. Fix bases v1 , . . ., vm and w1 , . . ., wn of V, W ,
respectively. The tensor product space V ⊗ W (or V ⊗F W ) is an mn-dimensional F -space
with basis {vi ⊗ wj : 1 6 i 6 m, 1 6 j 6 n}. Thus:
( )
X
(a) V ⊗W = λij vi ⊗wj : λij ∈ F , with ‘obvious’ addition and scalar multiplication.
16i6m
16j6n

P P P
(b) if v = αi vi ∈ V , w = βj wj ∈ W , define v ⊗ w = i,j αi βj (vi ⊗ wj ).

Note: not all elements of V ⊗W are of this form. Some are combinations, e.g. v1 ⊗w1 +v2 ⊗w2 ,
which cannot be further simplified.

21
(9.3) Lemma. (i) For v ∈ V , w ∈ W , λ ∈ F , have (λv) ⊗ w = λ(v ⊗ w) = v ⊗ (λw)
(ii) If x, x1 , x2 ∈ V and y, y1 , y2 ∈ W , then (x1 + x2 ) ⊗ y = (x1 ⊗ y) + (x2 ⊗ y) and
x ⊗ (y1 + y2 ) = (x ⊗ y1 ) + (x ⊗ y2 ).
P P P
Proof. (i) v = αi vi , v = βj wj , then (λv) ⊗ w = i,j (λαi )βj vi ⊗ wj
P
λ(v ⊗ w) = λ i,j αi βj vi ⊗ wj
P
v ⊗ (λw) = i,j αi (λβj )vi ⊗ wj

All three are equal. (ii) is similar. 2

(9.4) Lemma. If {e1 , . . ., em } is a basis of V and {f1 , . . ., fn } is a basis of W , then {ei ⊗ fj :

1 6 i 6 m, 1 6 j 6 n} is a basis of V ⊗ W .
P P P
Proof. Writing vk = i αik ei , wl = jl fj , we have vk ⊗ wl = i,j αik βjl ei ⊗ fj , hence
{ei ⊗ fj } spans V ⊗ W and since there are mn of them, they are a basis. 2

(9.5) Digression. (Tensor products of endomorphisms.) If α : V → V , β : W → W are

linear endomorphisms, define α ⊗ β : V ⊗ W → V ⊗ W , v ⊗ w 7→ α(v) ⊗ β(w), and extend
linearly on a basis.

Example. Given bases A = {e1 , . . ., em } of V , and B = {f1 , . . ., fn } of W , if [α]A = A

and [β]B = B, then ordering the basis A ⊗ B lexicographically (i.e., e1 ⊗ f1 , e1 ⊗ f2 ,. . .,
e1 ⊗ fn ,e2 ⊗ f1 ,. . . ), we have
 
[a11 B] [a12 B] . . .
 
[α ⊗ β]A⊗B = [a21 B] [a22 B] . . .
.. .. ..
. . .

(9.6) Proposition. Let ρ : G → GL(V ), ρ′ : G → GL(V ′ ) be representations of G. Define

ρ ⊗ ρ′ : G → GL(V ⊗ V ′ ) by
X X
(ρ ⊗ ρ′ )(g) : λij vi ⊗ wj 7→ λij ρ(g)vi ⊗ ρ′ (g)wj

Then ρ ⊗ ρ′ is a representation of G, with character χρ⊗ρ′ (g) = χρ (g)χρ′ (g) for all
g ∈ G.

Hence the product of two characters of G is also a character of G. Note: example sheet 2,
question 2, says that if ρ is irreducible and ρ is degree 1, then ρ ⊗ ρ′ is irreducible. if ρ′ is
not of degree 1, then this is usually false, since ρ ⊗ ρ′ is usually reducible.

Proof. Clear that (ρ ⊗ ρ′ )(g) ∈ GL(V ⊗ V ′ ) for all g, and so ρ ⊗ ρ′ is a homomorphism

G → GL(V ⊗ V ′ ).

Let g ∈ G. Let v1 , . . ., vm be a basis of V of eigenvectors of ρ(g), and w1 , . . ., wn be a

basis of V ′ of eigenvectors of ρ′ (g). So ρ(g)vj = λj vj , ρ′ (g)wj = µj wj .

Then (ρ ⊗ ρ′ )(g)(vi ⊗ wj ) = ρ(g)vi ⊗ ρ′ (g)wj = λi vi ⊗ µj wj = (λi µj )(vi ⊗ wj ).

X m
X n
X
So χρ⊗ρ′ (g) = λi µj = λi µj = χρ (g)χρ′ (g). 2
i,j i=1 j=1

P P
Take V = V ′ and define V ⊗2 = V ⊗ V . Let τ : λij vi ⊗ vj 7→ λij vj ⊗ vi , a linear
G-endomorphism of V ⊗2 such that τ 2 = 1.

22
(9.7) Definition. S 2 V = {x ∈ V ⊗2 : τ (x) = x} – symmetric square of V
Λ2 V = {x ∈ V ⊗2 : τ (x) = −x} – exterior square of V

(9.8) S 2 V and Λ2 V are G-subspaces of V ⊗2 , and V ⊗2 = S 2 V ⊕ Λ2 V .

S 2 V has a basis {vi vj := vi ⊗ vj + vj ⊗ vi , 1 6 i 6 j 6 n}, so dim S 2 V = 12 n(n + 1)

Λ2 V has a basis {vi ∧ vj := vi ⊗ vj − vj ⊗ vi , 1 6 i 6 j 6 n}, so dim Λ2 V = 12 n(n − 1)

Proof. Elementary linear algebra.

To show V ⊗2 is reducible, write x ∈ V ⊗2 as x = 21 (x + τ (x)) + 21 (x − τ (x)). 2

| {z } | {z }
∈ S2 ∈ Λ2

(9.9) Lemma. If ρ : G → GL(V ) is a representation affording character χ, then χ2 =

χS + χΛ where χS (=S 2 χ) is the character of G on the subrepresentation on S 2 V , and
χΛ (=Λ2 χ) is the character of G on the subrepresentation on Λ2 V .



Moreover, for g ∈ G, χS (g) = 1
2
χ2 (g) + χ(g 2 ) and χΛ (g) = 1
2
χ2 (g) − χ(g 2 ) .

Proof. Compute the characters χS , χΛ . Fix g ∈ G. Let v1 , . . ., vm be a basis of V of

eigenvectors of ρ(g), say ρ(g)vi = λi vi .

Then g vi vj = λi λj vi vj and g vi ∧ vj = λi λj vi ∧ vj .
X X
Hence χS (g) = λi λj and χΛ (g) = λi λj .
16i6j6n 16i<j6n

2  X 2 X 2 X
Now χ(g) = λi = λi + 2 λi λj = χ(g 2 ) + 2χΛ (g).
i<j



So, χΛ (g) = 1
2
χ2 (g) − χ(g 2 ) , and so χS (g) = 1
2
χ2 (g) + χ(g 2 ) , as χ2 = χS + χΛ . 2

‘Usual trick to find characters: diagonalise and hope for the best!’

Example. G = S5 (again)
1 15 20 24 10 20 30 ←|Cj |
1 (12)(34) (123) (12345) (12) (123)(45) (1234) ←gj
1G = χ1 1 1 1 1 1 1 1
sign = χ2 1 1 1 1 −1 −1 −1
χ = |fix[1,5] (g)|−1 = χ3 4 0 1 −1 2 −1 0
χ3 χ2 = χ4 4 0 1 −1 −2 1 0
S − 1 − χ3 =
2χ χ5 5 1 −1 0 −1 −1 1
χ5 χ2 = χ6 5 1 −1 0 1 1 −1
Λ2 χ = χ7 6 −2 0 1 0 0 0

We use (9.9) on χ2 χ3 .
1 (12)(34) (123) (12345) (12) (123)(45) (1234)
χ2 (g) 16 0 1 1 4 1 0
χ(g 2 ) 4 4 1 −1 4 1 0
χS (g) 10 2 1 0 4 1 0
χΛ (g) 6 −2 0 1 0 0 0
We have seen χS already as π(5) . Check inner product = 3; contains 1, χ3 .
2

23
Characters of G × H (cf. (4.5) for abelian groups)
(9.10) Proposition. If G, H are finite groups, with irreducible characters χ1 , . . ., χk and
ψ1 , . . ., ψl respectively, then the irreducible characters of the direct product G × H are
precisely {χi ψj : 1 6 i 6 k, 1 6 j 6 l} where χi ψj (g, h) = χi (g)ψj (h).

Proof. If ρ : G → GL(V ) affording χ and ρ′ : H → GL(W ) affording ψ, then ρ ⊗ ρ′ :

G × H → GL(V ⊗ W ), (g, h) 7→ ρ(g) ⊗ ρ′ (h) is a representation of G × H on V ⊗ W
by (9.6). And χρ⊗ρ′ = χψ, also by (9.6).

Claim: χi ψj are distinct and irreducible, for:

1 X
hχi ψj , χr ψs iG×H = χi ψj (g, h) χr ψs (g, h)
|G × H|
(g,h)
! !
1 Xχ χ 1 X
= i (g) r (g) ψj (h) ψs (h)
|G| g |H|
h

= δir δjs
P P P
Complete set: i,j χi ψj (1)2 = i χ2i (1) j ψj2 (1) = |G| |H| = |G × H|. 2

Exercise. D6 × D6 has 9 characters.

Digression: a general approach to tensor products

V, W , F -spaces (general F , even a non-commutative ring).

(9.11) Definition. V ⊗ W is the F -space with a bilinear map t : V × W → T , (v, w) 7→

v ⊗ w =: t(v, w), such that any bilinear f : V × W → X (X any F -space) can be
‘factored through’ it:
t
V × W −→ T
I.e., there exists linear f ′ : T → X such that f ′ ◦ t = f . fց ւ ∃f ′
X
The triangle commutes

This is the universal property of the tensor product.

Claim. Such T exists and is unique up to isomorphism.

Existence. Take space M with basis {(v, w) : v ∈ V, w ∈ W }. Factor out the subspace N
generated by ‘all the things you want to be zero’, i.e. by

(v1 + v2 , w) − (v1 , w) − (v2 , w)

(v, w1 + w2 ) − (v, w1 ) − (v, w2 ) for all v, v1 , v2 ∈ V , w, w1 , w2 ∈ W , λ ∈ F .
(λv, w) − λ(v, w), (v, λw) − λ(v, w)

Define t to be the map embedding V × W → M followed by the natural quotient map

t
V × W −→ M/N
f ց ւ ∃f ′
X

Check t is bilinear (we’ve quotiented out the relevant properties to make it so). f ′ is

24
 defined on our basis of M , (v, w) 7→ f (v, w), extended linearly. f ′ = 0 on all elements
of N , hence well-defined on M/N .

Uniqueness. V × W −→ T Apply universal property with respect to T, T ′ .

ց ւ ր Linear maps give isomorphism.
T′

Henceforth, we think of V ⊗ W as being generated by elements v ⊗ w (v ∈ V, w ∈ W ) and

satisfying
(v1 + v2 ) ⊗ w = v1 ⊗ w + v2 ⊗ w
v ⊗ (w1 + w2 ) = v ⊗ w1 + v ⊗ w2
λ(v ⊗ w) = λv ⊗ w = v ⊗ λw

(9.12) Lemma. If e1 , . . ., em and f1 , . . ., fn are bases of V , W respectively then {ei ⊗ fj :

1 6 i 6 m, 1 6 j 6 n} is a basis of V ⊗ W .

Proof. (Span.)
P Any vP⊗ w can be expressed
P (hence so can any element of V ⊗ W ) as
v = i αi ei , w = j βj fj ⇒ v ⊗ w = i,j αi βj ei ⊗ fj .

(Independence.) Find a linear functional φ sending ei ⊗ fj to 1 and all the rest to 0.

For, take dual basis {εi }, {φj } to the above. Define φ(v ⊗ w) = εi (v)φj (w) and check
φ(ei ⊗ fj ) = 1, other = 0. 2

(9.13) Lemma. There is a ‘natural’ (basis independent) isomorphism in each of the follow-
ing.
(i) V ⊗ W ∼=W ⊗V
(ii) U ⊗ (V ⊗ W ) ∼
= (U ⊗ V ) ⊗ W
(iii) (U ⊕ V ) ⊗ W ∼
= (U ⊗ W ) ⊕ (V ⊗ W )

Proof. (i) v ⊗ w 7→ w ⊗ v and extend linearly. It’s well-defined: (v, w) 7→ w ⊗ v is a

bilinear map V × W → W ⊗ V . So by the universal property v ⊗ w 7→ w ⊗ v gives
a well-defined linear map.
(ii) u ⊗ (v ⊗ w) 7→ (u ⊗ v) ⊗ w and extend linearly. It’s well-defined: fix u ∈ U , then
(v, w) 7→ (u ⊗ v) ⊗ w is bilinear, so get v ⊗ w 7→ (u ⊗ v) ⊗ w.
Varying u, (u, v ⊗ w) 7→ (u ⊗ v) ⊗ w is a well-defined bilinear map U × (V ⊗ W ) →
(U ⊗ V ) ⊗ W . Hence, get linear map u ⊗ (v ⊗ w) 7→ (u ⊗ v) ⊗ w.
(iii) Similar. (See Telemann, chapter 6.) 2

(9.14) Lemma. Let dim V, dim W < ∞. Then Hom(V, W ) ∼

= V ∗ ⊗W naturally as G-spaces,
if V, W are both G-spaces.

Proof. The natural map V ∗ × W → Hom(V, W ), (α, w) 7→ (φ : v 7→ α(v)w) is bilinear, so

α ⊗ w 7→ φ, extended linearly, is a linear map, V ∗ ⊗ W → Hom(V, W ).

It’s bijective as it takes basis to basis: εi ⊗ fj 7→ (Eji : ei 7→ fj ). 2

Returning to the proof of orthogonality at the end of chapter 6: the missing link was to
observe that U = Hom(V ′ , V ) ∼
= (V ′ )∗ ⊗ V , hence χn (g) = χ(V ′ )∗ ⊗V (g) = χV ′ g −1 χV (g).

25
Symmetric and exterior powers
V an F -space, dim V = d, basis {e1 , . . ., ed }, n ∈ N. Then V ⊗n = V ⊗ . . . ⊗ V (n times), of
dimension dn .

Note, Sn acts on V ⊗n : for σ ∈ Sn , σ(v1 ⊗ . . . ⊗ vn ) = vσ(1) ⊗ . . . ⊗ vσ(n) , and extend linearly

(‘place permutations’).

The Sn -action commutes with any G-action on V .

(9.15) Definition.
The symmetric powers, S n V = {x ∈ V ⊗n : σ(x) = x for all σ ∈ Sn }.
The exterior powers, Λn V = {x ∈ V ⊗n : σ(x) = sgn(σ)x for all σ ∈ Sn }.

These are G-subspaces of V ⊗n , but if n > 2 then there are others obtained from the Sn -action.
 
n 1 X
Exercises. Basis for S V is viσ(1) ⊗ . . . ⊗ viσ(n) : 1 6 i1 6 . . . 6 in 6 d .
n!
σ∈Sn

 
n 1 X
Basis for Λ V is sgn(σ)viσ(1) ⊗ . . . ⊗ viσ(n) : 1 6 i1 < . . . < in 6 d .
n!
σ∈Sn

   
d+n−1 d
So dim S n V = and dim Λn V = .
d n

(9.16) Definition. Let T n V = V ⊗n = V ⊗ . . . ⊗ V .

M
The tensor algebra of V is T (V ) = T n V , where T 0 V = {0} – an F -space with
n>0
obvious addition and scalar multiplication.

There is a product: for x ∈ T n V , y ∈ T m V , get x.y := x ⊗ y ∈ T n+m V , thus giving a

graded algebra (with product T n V ⊗ T m V → T n+m V ).

Finally, define:

S(V ) = T (V )/(ideal generated by u ⊗ v − v ⊗ u) – the symmetric algebra,

Λ(V ) = T (V )/(ideal generated by v ⊗ v) – the exterior algebra.

Character ring
C(G) is a ring, so the sum and product of characters are class functions. This chapter verified
that they are in fact characters afforded by the sum and tensor product of their corresponding
representations.

(9.17) The Z-submodule of C(G) spanned by the irreducible characters of G is the character
ring of G, written R(G).
 
 difference  X
Elements of R(G) are called generalised characters. φ ∈ R(G) : ψ = nχ χ, nχ ∈ Z.
 
virtual χ irred

R(G) is a ring, and any generalised character is a difference of two characters. (ψ = α − β,

26
 X X
α, β characters, where α = n χ χ, β = − nχ χ.)
nχ >0 nχ <0

The {χi } form a Z-basis for R(G), as free Z-module.

Henceforth we don’t distinguish between a character and its negative, and we often study
generalised characters of norm 1 (hα, αi = 1) rather than just irreducible characters.

10. Induction and Restriction

Throughout, H 6 G.

(10.1) Definition (Restriction). Let ρ : G → GL(V ) be a representation affording χ.

Can think of V as a H-space by restricting attention to h ∈ H.

Get ResG
H ρ : H → GL(V ), the restriction of ρ to H. (Also written ρ|H or ρH .)

It affords the character ResG

H
χ = χ|H = χH .

(10.2) Lemma. If ψ is any non-zero character of H, then there exists an irreducible char-
acter χ of G such that

• ψ ⊂ ResG H
χ 
• ψ is a constituent of ResG
H
χ 3 ways of saying the same thing

• hResG χ
H , ψi =6 0

Proof. List the irreducible characters of G: χ1 , . . .χk . Recall χreg from (5.9).

|G| X
0 6= ψ(1) = hχreg |H , ψiH = degχi hχi |H , ψiH
|H|

Therefore hχi |H , ψi =
6 0 for some i. 2
P
(10.3) Lemma. Let χ be an irreducible character of G, and let ResG
H
χ= χ
i ci i with χi
irreducible characters of H, where ci ∈ Z>0 .
X
Then c2i 6 |G : H|, with equality iff χ(g) = 0 for all g ∈ G \ H.

X 1 X χ
Proof. c2i = hResG χ G χ
H , ResH iH = | (h)|2 .
|H|
h∈H

1 X χ
But 1 = hχ, χiG = | (g)|2
|G|
g∈G
X X 
1 χ 2 χ 2
= | (h)| + | (g)|
|G|
h∈H g∈G\H

|H| X 2 1 X χ
= ci + | (g)|2
|G| |G|
g∈G\H
| {z }
> 0, and = 0 ⇔ χ(g) = 0 ∀ g ∈ G \ H
P
Therefore c2i 6 |G : H|, with equality iff χ(g) = 0 for all g ∈ G \ H. 2

27
Example. G = S5 , H = A5 , ψi = ResG χ
H i.

6=0 somewhere outside H χ(g)=0 ∀ g∈S5 \A5

ց ւ
degχi 1 1 4 4 5 5 6
\/ \/ \/ /\
deg ψi 1 4 5 3 3
|{z}
general fact about normal subgroups:
splits into constituents of equal degree
(Clifford’s Theorem)

(10.4) Definition (Induction). If ψ is a class function of H, define

1 X ◦ −1
ψ G = IndG
H ψ(g) = ψ (x gx)
|H|
x∈G

◦ ψ(y) y ∈ H
where ψ =
0 y∈/H

(10.5) Lemma. If ψ is a class function of H, then IndG

H ψ is a class function of G, and
IndG
H ψ(1) = |G : H| ψ(1).

1 X ◦
Proof. Clear, noting that IndG
H ψ(1) = ψ (1) = |G : H| ψ(1). 2
|H|
x∈G

Let n = |G : H|. Let t1 = 1, t2 , . . ., tn be a left transversal of H in G (i.e., a complete set

of coset representatives), so that t1 H = H, t2 H, . . ., tn H are precisely the left cosets of H in
G.
n
X ◦
(10.6) Lemma. Given a transversal as above, IndG
H ψ(g) = ψ (t−1
i gti ).
i=1

◦
◦
Proof. For h ∈ H, ψ (ti h)−1 g(ti h) = ψ (t−1
i gti ), as ψ is a class function of H. 2

(10.7) Theorem (Frobenius Reciprocity). ψ a class function on H, φ a class function

on G. Then
G
hResGH φ, ψiH = hφ, IndH ψiG .

1 X
Proof. hφ, ψ G iG = φ(g)ψ G (g)
|G|
g∈G
1 X ◦
= φ(g) ψ (x−1 gx)
|G| |H| g,x
1 X ◦
= φ(y) ψ (y) (put y = x−1 gx)
|G| |H| x,y
1 X ◦
= φ(y) ψ (y) (independent of x)
|H|
y∈H
1 X
= φ(y)ψ(y)
|H|
y∈H
= hφH , ψiH 2

28
(10.8) Corollary. If ψ is a character of H then IndG
H ψ is a character of G.

Proof. Let χ be an irreducible character of G.

By (10.7), hIndG χ G χ G χ
H ψ, iG = hψ, ResH i ∈ Z>0 , since ψ, ResH are characters.

Hence IndGH ψ is a linear combination of irreducible characters, with positive coefficients,

hence a character. 2

(10.9) Lemma. Let ψ be a character (or even a class function) of H and let g ∈ G. Let
CG (g) ∩ H = ∪mi=1 CH (xi ) (disjoint union), where xi are representatives of the m H-
conjugacy classes of elements of H conjugate to g.

Xm
ψ(xi )
Then IndG
H ψ(g) = |CG (g)| .
i=1
|CH (xi )|

1 X ◦ −1
Proof. IndG
H ψ(g) = ψ (x gx)
|H|
x∈G
1 X ◦
= ψ |CG (g)| x−1 gx (as x runs through G)
|H| will hit xi precisely |CG (g)|
y∈CG (g)∩H
Xm times; there are |H : CH (xi )|
1
= |CG (g)| |H : CH (xi )| ψ(xi ) H-conjugates of xi in H
|H| i=1
X ψ(xi )
= |CG (g)|
|CH (xi )|

Note that this holds even if m = 0: then no elements of CG (g) lie in H, in which case
IndG
H ψ(g) = 0. 2

(10.10) Lemma. If ψ = 1H , the principal character of H, then IndGH 1H = πX , the permu-

tation character of G on the set X of left cosets of H in G.
X ◦
Proof. IndG
H 1H (g) = 1H (t−1
i gti )

= {i : t−1
i gti ∈ H}

= {i : g ∈ ti Ht−1
i } ← stabiliser in G of the point ti H ∈X

= |fixX (g)| = πX (see (7.1)) 2

Remark. Recalling (7.3), hπX , 1G iG = hIndG

H 1H , 1G iG = 1.
(10.10) (10.7)

Examples. (a) Recall (7.9), G = S5 acting on X = Syl5 (G). πX = IndG

H 1H , where
H = h(12345), (2354)i.

H − ccls 1 (12345) (2354) (2453) (25)(34)

size 1 4 5 5 5

G − ccls 1 (12)(34) (123) (12345) (12) (123)(45) (1234)

size 1 15 20 24 10 20 30

πX (2354) = 4( 41 + 41 ) = 2
πX ((25)(34)) = 8( 14 ) = 2.

29
 (b) Recall (2.17) and (7.8). H = C4 = h(1234)i 6 G = S4 , index 6.
Character of induced representation IndSC44 (α), where α is faithful 1-dimensional
representation of C4 . If α((1234)) = i then character of α is:

1 (1234) (13)(24) (1432)

χα 1 i −1 −i

Induced representations:

size 1 6 8 3 6
ccls 1 (12) (123) (12)(34) (1234)
IndSC44 (α) 6 0 0 −2 0

For (12)(34), only one of 3 elements in S4 that it’s conjugate to lies in H. So

IndG 1
H (α) = 8(− 4 ) = −2.

(1234) is conjugate to 6 elements of S4 , of which 2 are in C4 (viz. (1234), (1432)).

So IndG i i
H (α) = 4( 4 − 4 ) = 0.

Induced modules
H 6 G, index n. t1 = 1, t2 , . . ., tn a transversal. W a H-space.

(10.11) Definition. Let V = W ⊕ t2 ⊗ W ⊕ . . . ⊕ tn ⊗ W , where ti ⊗ W = {ti ⊗ w : w ∈ W }.

(‘Essentially tensored group algebra with W ’)

So dim V = n dim W and we write V = IndG

H W.

G-action. g ∈ G, for all i, there exists a unique j with t−1

j gti ∈ H (namely tj H is the unique
coset which contains gti ).


Define g(ti w) = tj (t−1
j gti )w . (Drop the ⊗s, so ti w := ti ⊗ w.) Check this is a G-action:


g1 (g2 ti w) = g1 tj (t−1
j g2 ti )w
| {z }
(∃ unique j s.t. g2 t1 H = tj H)


= tl (t−1 g1 tj )(t−1 g2 ti )w
| l{z } j
(∃ unique l s.t. g1 tj H ∈ t ellH)


= tl t−1
l (g1 g2 )ti w
= (g1 g2 )(ti w)

l is unique with (g1 g2 )ti H ∈ tl H.

It has the right character (still dropping the ⊗) g : ti w 7→ tj (t−1

j gti )w
| {z }
∈W
so the contribution to the character is 0 unless j = i, i.e. unless t−1
i gti ∈ H, then it
−1 G P ◦ −1
contributes ψ(ti gti ), i.e. IndH ψ(g) = ψ (ti gti ).

Remarks (non-examinable). (1) There is also a ‘Frobenius reciprocity’ for modules: for
W a H-space, V a G-space, HomH (W, ResG ∼ G
H V ) = HomG (IndH W, V ) naturally,
as vector spaces.
This is an example of a ‘Nakayama relation’. See Telemann 15.9 – works over
general fields.

30
 (2) Tensor products of modules over rings. In (10.11), V = F G ⊗F H W .
Replace F G by R, F H by S, and try to generalise. In general, given rings R, S,
and modules U an (R, S)-bimodule and W a left S-module, then U ⊗ W is a left
R-module with balanced map t : U × W → U ⊗ W such that any balanced map
f : U × W → X, any left R-module X can be factored through t.
t
U × W −→ U ⊗ W
fց ւ ∃ unique module homomorphismf ′
X
‘Balanced’ means f (u1 + u2 , w) = f (u1 , w) + f (u2 , w)
f (u, w1 + w2 ) = f (u, w1 ) + f (u, w2 )
f (λu, w) = f (u, λw) (for all λ ∈ S)
Then IndG
H W = F G ⊗ W is now a well-defined F G-module, since W is a left
F H-module, F G is (F G, F H)-bimodule. (Alperin-Bell.)

11. Frobenius Groups

(11.1) Frobenius Theorem (1891). G a transitive permutation group on a set X, |X| =
n. Assume that each non-identity element of G fixes at most one element of X. Then

K = {1} ∪ {g ∈ G : gα 6= α for all α ∈ X}

is a normal subgroup of G of degree n.

Proof. (Suzuki, Collins (book).) Required to prove K  G.

Let H = Gα (stabiliser of α ∈ X), so conjugates of H are the stabilisers of single

elements of X. No two conjugates can share a non-identity element (hypothesis).

So H has n distinct conjugates and G has n(|H| − 1) elements that fix exactly one
element of X. But |G| = |X| |H| = n |H|. (X and G/H are isomorphic G-sets), hence
|K| = |G| − n(|H| − 1) = n.

Let 1 6= h ∈ H. Suppose h = gh′ g −1 , some g ∈ G, h′ ∈ H. Then h lies in H = Gα and

gHg −1 = Ggα . By hypothesis, gα = α, hence g ∈ H. So conjugacy class in G of h is
precisely the conjugacy class in H of h.

Similarly, if g ∈ CG (h) then h = ghg −1 ∈ Ggα hence g ∈ H, i.e. CG (h) = CH (h).

Every element of G lies either in K or in one of the n stabilisers, each of which is

conjugate to H. So every element of G \ K is conjugate with a non-1 element of H. So
{1, h2 , . . ., ht , y1 , . . ., yu }
| {z } | {z }
reps of H-ccls reps of ccls of G comprising K \ {1}

is a set of conjugacy class representatives for G.

Problem. To show K 6 G.
Take θ = 1G , {1H = ψ1 , ψ2 , . . ., ψt } irreducible characters of H. Fix some 1 6 i 6 t.
Then if g ∈ G,

 |G : H| ψi (1) = nψi (1) g = 1
IndG
H ψi (g) = ψi (hj ) g = hj (2 6 j 6 t)

 0 g = yk (1 6 k 6 u)

CG (hj ) = CH (hj ) & (10.9)

31
 Fix some 2 6 i 6 t and put θi = ψiG − ψi (1)ψ1G + ψi (1)θ1 ∈ R(G).

Values for 2 6 j 6 t, 1 6 k 6 u:

1 hj yk
ψiG nψi (1) ψi (hj ) 0
ψi (1)ψ1G nψi (1) ψi (1) 0
ψi (1)θ1 ψ(1) ψi (1) ψi (1)
θi ψi (1) ψi (hj ) ψi (1)

1 X
hθi , θi i = |θi (g)|2
|G|
g∈G
1 X X X 
= |θi (g)|2 + |θi (g)|2
|G|
g∈K α∈X 16=g∈Gα
1  2 X 
= nψi (1) + n |θi (h)|2
|G|
16=h∈H
1 X
= |ψi (h)|2
|H|
= hψi , ψi i
= 1 (row orthogonality)

P either θi or −θi is an irreducible characterP

By (9.17) of G, since θi (1) > 0, it is θi . Let
θ = ti=1 θi (1)θi . Column P
orthogonality ⇒ θ(h) = ti=1 ψi (1)ψi (h) = 0 (1 6= h ∈ H)
and for any y ∈ K, θ(y) = ψi2 (1) = |H|.

|H| if g ∈ K
So θ(g) =
0 if g ∈/K

Therefore K = {g ∈ G : θ(g) = θ(1)}  G. 2

(5.3)

(11.2) Definition. A Frobenius group is a group G having a subgroup H such that

H ∩ H g = 1 for all g ∈ H. H is a Frobenius complement.

(11.3) Any finite Frobenius group satisfies the hypothesis of (11.1). The normal subgroup
K is the Frobenius kernel of G.

If G is Frobenius and H a complement then the action of G on G/H is faithful and tran-
sitive. If 1 6= g ∈ G fixes xH and yH then g ∈ xHx−1 ∩ yHy −1 , which implies that
H ∩ (y −1 x)H(y −1 x)−1 6= 1, and so xH = yH.

Remarks. (i) Thompson (thesis, 1959) worked on the structure of Frobenius groups – e.g.
showed that K is nilpotent (i.e., K is the direct product of its Sylow subgroups).
(ii) There is no proof of (11.1) known in which character theory is not used.
(iii*) Show that G = K ⋊ H, semi-direct product.

32
12. Mackey Theory
This describes restriction to a subgroup K 6 G of an induced representation ∈ W . K, H are
unrelated but usually we take K = H, in which case we can tell when IndGH W is irreducible.

Special case: W = 1 (trivial H-space, dim 1). Then by (10.10) IndG H 1 = permutation
representation of G on X = G/H (coset action on the set of left cosets of H in G).

Recall. If G is transitive on a set X and H = Gα (α ∈ X) then the action of G on X is

isomorphic to the action on G/H, viz:

(12.1) g.α ←→ gH is a well-defined bijection and commutes with G-actions.

|{z} |{z}
∈X ∈G/H

I.e., x(gα) = (xg)α ←→ x(gH) = (xg)H.

Consider the action of G on G/H and restriction to some K 6 G. G/H splits into K-orbits;
these correspond to double cosets KgH = {kgh : k ∈ K, h ∈ H}. The K-orbit containing
gH contains precisely all kgH (k ∈ K).

(12.2) Definition. K \ G/H is the set of double cosets KgH.

Note |K \ G/H| = hπG/K , πG/H i – see (7.4). Clearly GgH = gHg −1 . Therefore KgH =
gHg −1 ∩ K. So by (12.1) the action of K on the orbit containing gH is isomorphic to the
action of K on K/(gHg −1 ∩ K).
M
G
(12.3) Proposition. ResG
K IndH 1 = IndK
gHg−1 ∩K 1,
g∈K\G/H

summed over set of representatives of double cosets.

S
Now choose g1 , . . ., gr such that G = Kgi H. Write Hg = gHg −1 ∩ K 6 K. Let W be
an H-space, and write Wg for the Hg -space with the same underlying vector space as W of
vectors, but with Hg -action from ρg (x) = ρ(g −1 xg ) for x ∈ gHg −1 .
| {z }
∈H

We will prove:
M
G
(12.4) Theorem (Mackey’s Restriction Formula). ResG
K IndH W = IndK
Hg Wg .
g∈K\G/H

In terms of characters:
X
G
(12.5) Theorem. If ψ ∈ C(H), then ResG
K IndH ψ = IndK
Hg ψg , where ψg is the class
g∈K\G/H
−1
function on Hg given by ψg (x) = ψ(xg ).

The most useful form for applications is:

(12.6) Corollary (Mackey’s Irreducibility Criterion). H 6 G, W and H-space. Then

V = IndG
H W is irreducible iff

(i) W is irreducible, and

(ii) for each g ∈ G \ H, the two (gHg −1 ∩ H)-spaces Wg and ResH Hg W have no irre-
ducible constituents in common. (We say they are disjoint.)

33
Proof of Corollary. Take K = H in (12.4), so Hg = gHg −1 ∩ H. Assume W is irreducible
with character ψ.
hIndG G
H ψ, IndH ψi = hψ, ResG G
H IndH ψi
(F.R)
X
= hψ, IndH
Hg ψg iH
(12.5) g∈H\G/H
X
= hResH
Hg ψ, ψg iHg
(F.R) g∈H\G/H
X
= 1+ dg where dg = hResH
Hg ψ, ψg iHg
g∈H\G/H
g∈H
/

So to get irreducibility we need all the dg = 0. 2

(12.7) Corollary. If H  G, assume ψ is an irreducible character of H. Then IndG H ψ is

irreducible iff ψ is distinct from all its conjugates ψg for g ∈ G \ H, where ψg (h) =
−1
ψ(hg ) = ψ(g −1 hg).

Proof. Take K = H, so Hg = gHg −1 ∩ H = H for all g (since H  G). ψg is the character

of H conjugate to ψ, so ResH
Hg ψ = ψ and the ψg are just the conjugates of ψ. 2

Proof of (12.4). Write V = IndG H W . Fix g ∈ G, so KgH ∈ K \ G/H. Observe V is a

direct sum of images of the form xW (officially x ⊗ W , recall), with x running over
representatives of left cosets of H in G (see (10.11)).
M Collect together the images xW
with x ∈ KgH (as in (12.3)) and define V (g) = xW .
x∈KgH

M
Now V (g) is a K-space and ResG
KV = V (g).
g reps of
K\G/H

We have to prove V (g) = IndK

Hg Wg , as K-spaces. The subgroup of K consisting of the
M
elements x with xgW = gW is Hg = gHg −1 ∩K (see (12.2)), and V (g) = x(gW ).
x∈K\Hg

Hence V (g) ∼
= IndK
Hg (gW ).

Finally Wg ∼
= gW as K-spaces, as the map w 7→ gw is an isomorphism. Hence the
assertion. 2

Examples. (a) Give a direct proof of (12.3).

[
Hint. Write G = Kgi H (1 6 i 6 r).
gi reps of
K\G/H
Let Hgi have transversal ki r1 , . . ., ki ri in K. Then {kij gi : 1 6 i 6 r, 1 6 j 6 ri }
is a transversal of K in G. Then compute IndG H ψ(k).
(b) (Examples sheet 3, question 4.) Cn
D2n = hx, y : xn = y 2 = 1, y −1 xy = x−1 i.
Mackey says that for any 1-dimensional representation α of Cn , the 2-dimensional
representation IndD Cn α is irreducible iff α is not isomorphic to αg .
2n

Now y xy = x , so this says that if α(x) = ζ i (ζ ∈ µn ), αg is the representation

−1 −1

αg (x) = ζ −i . So for 0 < i < n/2 we get a 2-dimensional irreducible representation

of D2n this way.

34
13. Integrality
(13.1) Definition. a ∈ C is an algebraic integer if it is a root of a monic polynomial
in Z[X]. Equivalently, the subring Z[a] = {f (a) : f (x) ∈ Z[X]} of C is a finitely-generated
Z-module.

Fact 1. The algebraic integers form a subring of C. (James & Liebeck 22.3)

Fact 2. If a ∈ C is both an algebraic integer and a rational number then a ∈ Z. (James &
Liebeck 22.3)

Fact 3. Any subring S of C which is finitely generated as a Z-module consists of algebraic

integers. (Show a is the root of a characteristic polynomial of a matrix.)

(13.2) Proposition. If χ is a character of G and g ∈ G then χ(g) is an algebraic integer.

Corollary. There are no entries in the character table of any finite group which are rational
but not integers. (Fact 2.)

Proof of (13.2). χ(g) is the sum of nth roots of 1 (n = |g|). Each root of unity is an
algebraic integer, and any sum of algebraic integers is an algebraic integer. (Fact 1.) 2
P
Recall from (2.4) the group algebra CG = { αg g : αg ∈ C} of a finite group G, the C-space
with basis the elements of G. It is also a ring.
X
List C1 = {1}, C2 , . . ., Ck , the G-conjugacy classes. Define the class sums, Cj = g ∈ CG.
g∈Cj

Z(CG) is the centre of CG (not the same as CZ(G)).

(13.3) Proposition. C1 , . . ., Ck isPa basis of Z(CG). There exist non-negative integers aijl
(1 6 i, j, l 6 k) with Ci Cj = aijl Cl . These are the structure constants for Z(CG).

E.g., 1, (12) + (13) + (23), (123) + (132) form a basis of Z(CS3 ).

Proof. gCj g −1 = Cj , so Cj ∈ Z(CG). Clearly the Cj are linearly independent (because the
conjugacy classes are pairwise disjoint).
P
Now suppose z ∈ Z(CG), z = g∈G αg g. Then for all h ∈ G we have αh−1 gh = αg ,
so the function
P g 7→ αg is constant on G-conjugacy classes. Writing αg = αi (g ∈ Ci ),
then z = αj Cj .

Finally Z(CG) is a C-algebra (‘vector space over C with ring multiplication’), so Ci Cj =

Pk
l=1 aijl Cl , as the Cj span. We claim that aijl ∈ Z>0 .

For: fix gl ∈ Cl , then aijl = #{(x, y) ∈ Ci × Cj : xy = gl } ∈ Z>0 . 2

Definition. Let ρ : G → GL(V ) be an irreducible representation over C affording χ.

Extend by linearity to ρ : CG → End V , an algebra homomorphism. Such a homomorphism
of algebras, CG = A → End V is a representation of A.

Let z ∈ Z(CG). Then ρ(z) commutes with all ρ(g) (g ∈ G), so by Schur’s Lemma ρ(g) = λz I
for some λz ∈ C. Consider the algebra homomorphism wχ = w : Z(CG) → C, z 7→ λz .

35
 P
Then ρ(Ci ) = w(Ci )I, so χ(1)w(Ci ) = g∈Ci
χ(g) = |Ci | χ(gi ) (gi a representative of Ci ).

χ(gi )
Therefore wχ (Ci ) = χ |Ci |.
(1)

χ(gi )
(13.5) Lemma. The values of wχ (Ci ) = χ |Ci | are algebraic integers.
(1)
P
Proof. Since w is an algebra homomorphism, have wχ (Ci )wχ (Cj ) = kl=1 aijl wχ (Cl ), with
aijl ∈ Z>0 . Thus the span {w(Ci ) : 1 6 i 6 k} is a subring of C, so by Fact 3 consists
of algebraic integers. 2

Example. Show that aijl = #{(x, y) ∈ Ci × Cj : xy = gl } can be obtained from the character
table. In fact,
Xk χ
s (gi )χs (gj )χs (gl )
−1
|G|
aijl = χs (1) .
|CG (gi )| |CG (gj )| s=1

Hint: use column orthogonality. (See James & Liebeck 30.4.)

(13.6) Theorem. The degree of any irreducible character of G divides |G|.

I.e., χi (1) |G| (1 6 i 6 k).

Proof. Given irreducible χ. (‘Standard trick: show |G|/χ(1) ∈ N.’)

|G| 1 X χ χ −1
χ(1) = χ(1) (g) (g )
g∈G
k
1 X
= χ(1) |Ci |χ(gi )χ(gi−1 )
i=1
Xk
|Ci |χ(gi ) χ −1
= χ(1) | (g{zi })
i=1 | {z } sum of roots of 1, so algebraic integer
algebraic integer by (13.5)

is an algebraic integer, and since it’s clearly rational, it is an integer. 2

Examples. (a) If G is a p-group then χ(1) is a p-power (χ irreducible). If |G| = p2 then

χ(1) = 1 (hence G is abelian).
(b) No simple group has an irreducible character of degree 2 (see James & Liebeck
22.13).
(c*) In fact, if χ is irreducible then χ(1) divides |G|/|Z| (Burnside).

14. Burnside’s pa q b Theorem

(14.1) Theorem (Burnside, 1904). p, q primes. Let |G| = pa q b where a, b ∈ Z>0 , with
a + b > 2. Then G is not simple.

Remarks. (1) In fact, even more is true: G is soluble.

(2) The result is best possible: A5 is simple, and 60 = 22 .3.5.
(3) If either a or b is 0 then |G| = p-power and we know Z(G) 6= 1. Then there is
g ∈ Z, |g| = p and hgi
G, with hgi =6 1 or G.

36
(14.2) Proposition. χ an irreducible C-character of G, C a G-conjugacy class, g ∈ G such
that (χ(1), |C|) = 1. Then |χ(g)| = χ(1) or 0.
bχ(g) χ(g)
Proof. There are a, b ∈ Z>0 such that aχ(1)+b|C| = 1. Define α = aχ(g)+ χ |C| = χ .
(1) (1)

Then α is an algebraic integer, so the assertion follows from:

m
1 X
(14.3) Lemma. Assume α = λi is an algebraic integer with λnj = 1 for all j, some n.
m i=1
Then |α| = 1.

For (14.2), we take n = |g|, m = χ(1).

6 0. Now α ∈ F = Q(ε) where ε = e2πi/n and

Proof (non-examinable). Assume |α| =
λj ∈ F for all j.

Let G = Gal(F/Q). Observe {β ∈ F : β σ = β for all σ ∈ G} = F G = Q. (Result from

Galois Theory.)

Consider the norm N (α) of α, namely the product of all the Galois conjugates ασ
(σ ∈ G). The norm ∈ Q because it’s fixed by all of G. It’s an algebraic integer (all
Galois group conjugates of an algebraic integer are algebraic integers). Hence N (α) ∈ Z.
Y P
σ roots of 1
But N (α) = α is a product of expressions ∈ C if absolute value 6 1.
m
σ∈G

Hence the norm must be ±1, hence |α| = 1. 2

(14.4) Theorem. If in a finite group G the number of elements in a conjugacy class C =

6 {1}
is a p-power, then G is not non-abelian simple.

Remark. This implies (14.1). Assume a > 0, b > 0. Let Q ∈ Sylq (G). Then Z(Q) 6= 1, so
choose 1 6= g ∈ Z(Q). So CG (g) ⊇ Q. Therefore |Ci (g)| = |G : CG (g)| = pr (some r).

Hence if pr = 1 then g ∈ Z(G). Therefore Z(G) 6= 1 (so not simple). If pr then G is

not simple (by (14.4)).

Proof of (14.4). Assume that G is non-abelian simple, and let 1 6= g ∈ G with |CG (g)| = pr .
X
By column orthogonality, 0 = χ(1)χ(g) – (∗)
χ irred
of G

G is non-abelian simple, so |χ(g)| =

6 χ(1) for any irreducible χ 6= 1. By (14.2), for any
irreducible character 6= 1 of G, we have p|χ(1) or χ(g) = 0.
χ

X χ(1)
Deleting zero terms in (∗), 0 = 1 + p χ(g).
χ irred
p
p|χ(1)

Thus 1/p is an algebraic integer, since 1/p ∈ Q, hence 1/p ∈ Z. Contradiction. 2

Remarks. (a) In 1911, Burnside conjectured that if |G| is odd then G is not non-abelian
simple. Only proved in 1963 by Feit & Thompson, a result which began the
Classification of Finite Simple Groups. The Classification only ended in 2005.
(b) A group-theoretic proof given only in 1972 (H. Bender)

37
15. Representations of Topological Groups
(15.1) A topological group is a group which is also a topological space such that the group
operations G × G → G, (h, g) 7→ hg and G → G, g 7→ g −1 are continuous. It is compact if
it is so as a topological space.
2 2
Basic examples. (a) GLn (R), GLn (C) are open subspaces of Rn or Cn .
(b) G finite, discrete topological. Also compact.
(c) G = S 1 = U (1) = {g ∈ C : |g| = 1}.
(d) O(n) = {A ∈ GLn (R) : AAt = I} – orthogonal group.
Compact: set of orthonormal bases for Rn = {(e1 , . . ., en ) ∈ Rn × . . . × Rn :
hei , ej i = δij }.
t
U (n) = {A ∈ GLn (C) : AA = I} – unitary group.
Compact: A ∈ U (n) iff its columns are orthonormal.
(e) SU (n) = {A ∈ U (n) : det A = 1} = SLn (C) ∩ U (n).
  
z1 z2
E.g., SU (2) = : zi ∈ C, |z1 |2 + |z2 |2 = 1 .
−z2 z1
∼
= S 3 = {z ∈ C2 : ||z|| = 1} ֒→ C2 ∼
= R4
SO(n) = {A ∈ O(n) : det A = 1} = SLn (R) ∩ O(n).
E.g., SO(2) ∼
= U (1), rotation of θ 7→ eiθ
SO(3), rotations about various axes in R3 .

SO(n), SU (n), U (n), O(n) are groups of isometries of geometric objects – known as compact
Lie groups. Theory is done by H. Weyl, ‘Classical Groups’.

(15.2) Definition. A representation of a topological group on a finite-dimensional vector

space V is a continuous group homomorphism ρ : G → GL(V ) with the topology of GL(V )
inherited from the space End V .

(There exist extensions when V is infinite-dimensional – see Telemann, remark 19.2.)

Here, continuous ρ : G → GL(V ) ∼
= GLn (C) means each g 7→ ρ(g) ij is continuous for i, j.

The compact group U(1)

(15.3) Theorem. The continuous homomorphisms C 1 → GL1 (C) = C× (i.e. the 1-dim.
representations of S 1 ) are precisely the representations z 7→ z n (some n ∈ Z).

The proof is closely tied with Fourier Series. We need a. . .

(15.4) Lemma. Consider (R, +). If ψ : R → R is a continuous homomorphism then ψ is

multiplication by a scalar.

Proof. Put c = ψ(1). Then ψ(n) = nc (n ∈ Z). Also mψ(1/m) = c, so ψ(1/m) = c/m
(m ∈ Z). Hence ψ(n/m) = cn/m. Thus ψ(x) = cx (x ∈ Q), but Q is dense in R and ψ
is continuous, so ψ(x) = cx for all x ∈ R.

(15.5) Lemma. If φ : R+ → U (1) is a continuous homomorphism then there exists c ∈ R

with φ(x) = eicx for all x ∈ R.

38
Proof. Claim. There is a unique continuous homomorphism α : R → R such that φ(x) =
eiα(x) (so we deduce (15.5) from (15.4)).

Recall that the exponential map ε : R+ → U (1), x 7→ eix , maps the real line around
the unit circle with period 2π.

R+ 0r r
2π φ
1r
R 
@@
complete the triangle!
-
ε  U (1)

For any continous φ : R+ → U (1) such that φ(0) = 1, there exists a unique continuous
lifting α of this function to the real line such that α(0)P= 0 – i.e., there exists a unique
continuous α : R → R such that α(0) = 0 and φ(x) = (α(x)) for all x.

(Lifting is constructed starting with condition α(0) = 0 and then extending it a small
interval at a time. See Telemann, section 21. Non-examinable!)

Claim. If φ is a homomorphism then its lift α is also a homomorphism.

We tensor φ(a + b) = φ(a)φ(b), hence ε α(a + b) − α(a) − α(b) = 1. Hence α(a + b) −
α(a) − α(b) = 2πm for some m ∈ Z depending only on a, b. Varying a, b continuously,
m = constant; setting a = b = 0 shows m = 0. 2

Proof of (15.3). Given a representation ρ : S 1 → C× , it has a compact, hence bounded,

image. This image lies on the unit circle (integral powers of any other complex number
would form an unbounded sequence). Thus ρ : S 1 → S 1 is a continuous homomorphism.

Thus we get a homomorphism R → S 1 , x 7→ ρ(eix ), so by (15.5), there exists c ∈ R

with ρ(eix ) = eicx .

Finally, 1 = ρ(ei2π ) = ei2πc , thus c ∈ Z. Putting n = c we have ρ(z) = z n . 2

So ρn : U (1) → C× , z 7→ z n , (n ∈ Z) give the complete list of irreducible representations of

U (1).

Schur’s Lemma applies – all irreducibles are 1-dimensional (cf. (4.4.)). Clearly their charac-
ters are linearly independent; in fact they are orthonormal under the inner product
Z 2π
1
hφ, ψi = φ(θ)ψ(θ) dθ (∗)
2π 0

where z = eiθ . I.e., ‘averaging

Pn over U (1)’. Finite linear combinations of these ρn are the
Fourier polynomials = m=−n am z m ; the ρn are the Fourier modes.

U (1) is abelian, hence coincides with the space of conjugacy classes

(15.6) Theorem. (i) The functions ρn form a complete list of the irreducible representa-
tions of U (1).
(ii) Every finite-dimensional representation V of U (1) is isomorphic to a sum of the
ρn . Its character χV is a Fourier polynomial. The multiplicity of ρn in V equals
hρn χV i (as in (∗)).

39
Remark. Complete reducibility of a finite-dimensional representation requires invoking
Weyl’s Unitary Trick (3.4) to average over a given inner product using integration on U (1) –
so before moving on to SU (2), let’s consider. . .

General theory of compact groups

The main tools for studying representations of finite groups are:

• Schur’s Lemma – holds here too

• Maschke’s Theorem. The relevant proof used Weyl’s trick of averaging over G. Need
to replace summation by integration over compact group G.
Z
Namely, for each continuous function f on G, we have f (g) dg ∈ C such that:
G
Z
• is a non-trivial functional
G
Z Z Z Z
• is left/right-invariant, i.e. f (g) dg = f (hg) dg = f (gh) dg (h ∈ G)
G G G G
Z
• G has total volume 1, i.e. dg = 1
G

A (difficult) theorem of Haar asserts that these constraints determine existence and unique-
ness for any compact G. We’ll assume it, but for our Lie groups of interest (U (1), SU (2),
etc) there are easier proofs of existence.
Z
1 X
Examples. (a) G finite. f (g) dg = f (g).
G |G|
g∈G
Z Z 2π
1
(b) G = S 1 . f (g) dg = f (eiθ dθ.
G 2π 0
(c) G = SU (2), 2 × 2 C-matrices preserving complex inner product and det = 1.
  
u v 2 2
I.e, SU (2) = : |u| + |v| = 1 .
−v u
Identify G with the unit 3-sphere S 3 ⊆ C2 ∼ = R4 in such a way that left/right
translation by elements of G give isometries on the sphere. With this identification,
translation-invariant integration on G can be taken to be integration over S 3 with
usual Euclidean measure ×1/2π 2 (to normalise).
  
z1 z2
(d) Embed SU (2) ⊆ H = : zi ∈ C , the quaternion algebra.
−z2 z1
(Actually, it’s a division algebra, so that every non-zero element has an inverse.)
√
H is a 4-dimensional Euclidean space: ||A|| = det A = (x21 + x22 + x23 + x24 )1/2
with SU (2) as the unit sphere in this normed space.
Multiplication (from left or right) by an element of SU (2) is an isometry of H, viz:

(AX, AY ) = det AX = det A det X = det X = (X, X) = (XA, Y A).

Once we have found our translation-invariant integration on the set of continuous functions
on our compact group G, a lot can be proved about the representation theory of G in parallel
with finite groups.

40
Representations (continuous, finite-dimensional) ; Characters (continuous functions → C).

Complete reducibility ; Weyl’s Unitary Trick of averaging over G replaced by integration.

Z
Character inner product: hχ, χ′ i = χ(g)χ′ (g) dg (†)
G

χ irreducible iff hχ, χi = 1.

Moreover,

(15.8) Theorem. (a) Every finite-dimensional representation is a direct sum of irreducible

representations (so completely reducible).
(b) Schur’s Lemma applies: if ρ, ρ′ are irreducible representations of G then

C if ρ is isomorphic to ρ′
Hom(ρ, ρ′ ) =
0 otherwise
(c) The characters of irreducible representations form an orthonormal set with respect
to the inner product (†) above. (The set is infinite, and it is not a basis for the
space of all continuous class functions.)
Even showing completeness of characters is hard – needs Peter-Weyl Theorem.
(d) If the characters of ρ, ρ′ are equal then ρ ∼
= ρ′ .
(e) If χ is a character with hχ, χi = 1 then χ is irreducible.
(f) If G is abelian then all irreducible representations are 1-dimensional.

The group SU(2)

  
a b 2 2
Recall G = SU (2) = : a, b ∈ C, |a| + |b| = 1 .
−b a
 
a b
G → S 3 ֒→ C2 = R4 , 7→ (a1 , a2 , b1 , b2 ). (Homeomorphism, i.e. continuous inverse.)
−b a
  
a 0
The centre is Z(G) = {±I}. Define the maximal torus T = : |a| = 1 = S 1 .
2
0 a

Conjugacy

(15.9) Proposition. (a) Every conjugacy class C of SU (2) meets T , i.e. C ∩ T 6= ∅.


{x, x−1 } if C =
6 {±I}
(b) In fact, C ∩ T =
C if C = {±I}
(c) The normalised trace, 12 tr : SU (2) → C, gives a bijection of the set of G-conjugacy
classes with the interval [−1, 1], namely
 
1 1 λ 0
g ∈ C 7→ tr = (λ + λ−1 ) if g ∼
2 2 0 λ−1

Ir
'$
Picture of ccls: 1
g 7→ 21 tr
-
r
&%
2-dim spheres of constant latitude
on unit sphere, plus the two poles
−1
−I

41
  
0 1
Proof. Let S = ∈ G, S 2 = −I.
−1 0

(a) Every unitary matrix has an orthonormal basis of eigenvectors, hence is conjugate
t
in U (2) to T , say QXQ ∈ T . We seek Q with det Q = 1 (so that Q ∈ SU (2)).
t
Let δ = det Q. Since QQ = I, |δ| = 1. If ε is a square of δ then Q1 = εQ ∈ SU (2),
t
hence Q1 XQ1 ∈ T .
(b) Let g ∈ SU (2) and suppose g ∈ CG . If g= ±I then C ∩ T = {g}.
 Otherwise g has
−1 λ 0 −1
distinct eigenvalues λ, λ and C = h h :h∈G .
0 λ−1
   −1     −1 
λ 0 λ 0 λ λ
Thus C ∩ T = , , by noting S S= .
0 λ−1 0 λ λ−1 λ
 
µ
Further, if ∈ C then {µ, µ−1 } = {λ, λ−1 }, i.e. the eigenvalues are
µ−1
preserved under conjugacy.
(c) Consider 21 tr : {ccls} → [−1, 1]. By (b) matrices are conjugate in G iff their
eigenvalues agree up to order. Now
 
1 λ 1
tr = (λ + λ−1 ) = Re(λ) = cos θ (λ = eiθ )
2 λ−1 2
hence the map is surjective onto [−1, 1].
It’s injective: 21 tr (g) = 21 tr (g ′ ) then g, g ′ have the same characteristic polynomial,
viz X 2 − tr (g)X + 1, hence the same eigenvalues, hence are conjugate. 2

Thus we write Ct = {g ∈ SU (2) : 12 tr (g) = t}.

Representations
Let Vn be the space of all homogeneous polynomials of degree n in the variables x, y. I.e., Vn =
{r0 xn + r1 xn−1 y + . . .+ rn y n }, and (n + 1)-dimensional C-space, with basis xn , xn−1 y, . . ., y n .

(15.10) GL2 (C) acts on Vn .

 
a b
First, define ρn : GL2 (C) → GL(Vn ). Let g = .
c d


ρn (g)f (x, y) = f (ax + cy, bx + dy) = f (x, y).g (i.e., matrix product)
Pn
I.e., for f = j=0 rj xn−j y j , ρ(g)f = r0 (ax + cy)n + r1 (ax + cy)n−1 (bx + dy) + . . . +
rn (bx + dy)n .

Check that this defines a representation.

E.g. (a) n = 0, ρ0 = trivial

   
a b a b
(b) n = 1, natural 2-dimensional representation. ρ1 has matrix with
c d c d
respect to the standard basis: x 7→ ax + cy, y 7→ bx + dy.
 2 
  a cb b2
a b
(c) n = 2, ρ2 = has matrix 2ac ad + bc 2bd with respect to the stan-
c d
c2 cd d2
dard basis.

42
Characters
 

z
χVn (g) = tr ρn (g) , g ∼ ∈ T.
z −1
 
z
ρn xi y i = (zx)i (z −1 y)j = z i−j xi y j .
z −1
 n 
z
   z n−2 
z  
So ρn has matrix  ..  with respect to the standard basis.
z −1  . 
z −n
   
z z n+1 − z −(n+1)
Hence, χn = χVn n
= z +z n−2
+ . . .+ z −n
= unless z = ±1.
z −1 z − z −1

(15.11) Theorem. The representations ρn : SU (2) → GL(Vn ) of dimension n + 1 are

irreducible for n ∈ Z>0 .

Proof. Telemann (21.1) shows hχn , χn i = 1 (implying χn irreducible). We will use combi-
natorics. Assume 0 6= W 6 Vn , G-invariant.
P
Claim. If w = j rj xn−j y j ∈ W with some rj 6= 0, then xn−j y j ∈ W .

Proof of claim. We argue by induction on the number of non-zero rj . If a unique

rj 6= 0 then it’s clear (multiply be its inverse), so we’ll assume more than one and
choose one.

Pick z ∈ C with z n , z n−2 , . . ., z −n distinct in C.

 
z P
Now, ρn w= rj z n−2j xn−j y j ∈ W (G-space).
z
 
z
Define wi = ρn w − z n−2i w ∈ W .
z −1
P ′ n−j j
Then wi = j rj x y and rj′ 6= 0 ⇔ (rj 6= 0 and j 6= i). By induction
hypothesis, we have xn−j y j ∈ W for all j with (rj 6= 0 and j 6= i).
P
Finally, xn−i y i = ri−1 (w − rj xn−j y j ) ∈ W , so the claim is proved.

Now let 0 6= w ∈ W . Wlog, w = xn−j y j . It is now easy to find matrices in SU (2), the
action of which will give all the xn−i y i ∈ W . E.g.,
 
1 1 −1 1
√ : xn−i y i 7→ √ (x + y)n−j (−x + y)j → xn ∈ W
2 1 1 2
 
a −b
: xn 7→ (ax + by)n → all xn−i y i ∈ W
b a ↑
(a, b 6= 0) all coefficients in here 6= 0

So all basis elements are in W . So W = Vn . 2

Next we show that all irreducibles of SU (2) are of the form in (15.11).

43
  X
n 
−1 m
Notation. Write N0 [z, z ]= am z : am ∈ N0 .
m=−n

And N0 [z, z −1]ev = {even Laurent polynomials, i.e. am = a−m for all odd m}.

Let χ = χV be thecharacter of some representation ρ : G → GL( V ). If g ∈ G = SU (2)

z
then g ∼G for some z ∈ C. So χV is determined by its restriction to T , hence
z −1    −1 
χV ∈ N0 [z, z −1] by (†). Actually χV ∈ N0 [z, z −1 ]ev since χV z χ
= V
z
,
z −1 z
   −1   
z z 1
because ∼G via S = .
z −1 z −1

(15.12) Theorem. Every (finite-dimensional, continuous) irreducible representation of G

is one of the ρn : G → GL(Vn ) above (n > 0).

Proof. Assume ρ : G → GL(V ) is an irreducible representation affording the character χ.

The characters characterise representations (15.8), so it’s enough to show χ = χn for
some n.

χ0 = 1, χ1 = z + z −1 , χ2 = z 2 + 1 + z −2 , . . . form a basis of Q[z, z −1]ev , hence

NowP
χ = an χn , a finite sum with an ∈ Q.

Clear the denominators and move all summands with negative coefficients to the LHS:
X X
mχ + m i χi = n j χj
i∈I j∈J

with I, J disjoint finite subsets of N, and m, mi , nj ∈ N.

The left and right hand sides are characters of representations of SU (2):
M M
mV ⊕ mi Vi ∼= nj Vj .
I J

Since V is irreducible we must have V ∼

= Vn , for some n ∈ J. 2

So far we have found all irreducible representations of G; they are ρn : G → GL(Vn ) (n 6= 0)

with Vn the (n + 1)-dimensional space of homogeneous polynomials of degree n in x, y. The
characters of ρn are given by (†).

To compute representations we ‘just’ work with characters: as an example we derive a famous

rule for decomposing tensor products.

Tensor product of representations

Reclal from section 9: if V, W are G-spaces we have V ⊗ W afforidng χV ⊗W = χV χW .

Examples. V1 ⊗ V1 = V2 ⊕ V0 . Character = (z + z −1)2 = z 2 + 2 + z −2 = (z| 2 + 1{z+ z −2}) + 1

V2 V0

V1 ⊗V2 = V3 ⊕V1 . Character = (z +z −1 )(z 2 +1+z −2 ) = (z 3 +z +z −1 +z −3 )+(z +z −1 )

44
(15.13) Theorem (Clebsch-Gordan). Vn ⊗ Vm = Vn+m ⊕ Vn+m−2 ⊕ . . . ⊕ V|n−m|

Proof. Just check that the characters work.

Wlog n > m and prove χn χm = χn+m + χn+m−2 + . . . + χn−m .

χn (g)χm (g) = z n+1 − z −n−1 m

−1
z + z m−2 + . . . + z −m
z−z
Xm
z n+m+1−2j − z 2j−n−m−1
=
j=0
z − z −1
m
X
= χn+m−2j
j=0

(The n > m ensures no cancellations in the sum.) 2

Some SU(2)-related groups

Check (see Telemann 22.1, and Examples Sheet 4 Question 6):

• SO(3) ∼
= SU (2)/{±I}
(∗)
• SO(4) ∼
= SU (2) × SU (2)/{±(I, I)}

(Isomorphisms, but actually homeomorphisms.)

So continuous representations of these groups are the same as continuous representations of

SU (2) and SU (2) × SU (2), respectively, which send −I and (−I, −I) to the identity matrix.

(15.14) Corollary. The irreducible representations of SO(3) are precisely ρ2m : SO(3) →
GL(V2m ) (m > 0).

Remarks. (a) We get precisely those Vn with −id in the kernel of the action, and −id acts
on Vn as  
(−1)n
 (−1)n−2 
 
 ..  = (−1)n id
 . 
(−1)−n

(b) V2 is the standard 3-dimensional representation of SO(3). (The only 3-dimensional

representation in the list.)
(c*) For SO(4) the complete list is ρm ⊗ ρm (m, n > 0, m ≡ n(2)) (see Telemann 22.7).
For U (2) the list is det⊗m ⊗ρn (m, n ∈ Z, n > 0) where det : U (2) → U (1) is
1-dimensional (see Telemann 22.9).

Sketch proof of (∗) Recall from (15.7)(d) that SU (2) ⊆ H ∼ = R4 can be viewed as the space
of unit norm quaternions. We also saw that multiplication from the left (and right) by
elements of SU (2) gives isometries of H. The left/right multiplication action of SU (2)
fives a homomorphism φ : SU (2) × SU (2) → SO(4), (g, h) 7→ {θ : q 7→ gqh−1 }.

Kernel. (g, h) sends 1 ∈ H to gh−1 , so (g, h) fixes the identity iff g = h, i.e. G =
{(g, g) : g ∈ SU (2)} = stabSU(2)×SU(2) (1).

Now (g, g) fixes every other quaternion iff g ∈ Z(SU (2)), i.e. g = ±id. Thus ker φ =
{±(I, I)}.

45
 Surjective and homeomorphic (i.e. inverse map is continuous). Restricting the
left/right action to G (the diagonal embedding of SU (2)) give the conjugation action
of SU (2) on the space of ‘pure quaternions’, hi, j, kiR (the trace 0 skew-Hermitian 2 × 2
matrices). So get a 3-dimensional Euclidean space on which G acts, and φ(G) 6 SO(3).

φ(G) = SO(3). Rotations in (i, j)-plane implemented by a + bk, similarly with any
permutations of i, j, k, and these rotations generate SO(3) (see some Geometry course).
So we have a surjective homomorphism SU (2) → SO(3), and we know that ker =
{±id}. The result follows.

Homeomorphism. Prove it directly or ‘recall’ the fact that a continuous bijection

from a compact space to a Hausdorff space is a homeomorphism (Sutherland 5.9.1)

Further worked example

Sn , GL2 (Fq ), Hp .

We consider Heisenberg groups. For p prime, the abelian groups of order p3 are Cp3 , Cp2 ×Cp ,
Cp × Cp × Cp , and their character tables can be constructed using (4.5).

Suppose G is any non-abelian group of order p3 . Let Z = Z(G), then it’s well-known that
Z 6= 1 and G/Z is non-cyclic, i.e. G/Z ∼
= Cp × Cp and Z = Cp .
  
 1 * * 
Take G = Hp = 0 1 * : * ∈ Fp , the modular Heisenberg group.
 
0 0 1

We take p odd (else G = D8 or Q8 ).

 
1 0 1
Have Z = hgi, z = 0 1 0.
0 0 1
   
1 1 0 1 0 0
With a = 0 1 0 and b = 0 1 1, [a, b] = z and G′ = Z.
0 0 1 0 0 1

There are p2 linear characters (of degree 1) (recall G/G′ = Cp × Cp ), and (p − 1) characters
of degree p, induced from the 1-dimensional characters of the abelian subgroup
 
 1 * * 
ha, zi = 0 1 0 
 
0 0 1

of order p2 .

Conjugacy classes
p conjugacy classes of size 1. The rest have size p and there are p2 − 1 such classes.

We’ll show that the character table of Hp looks like

46
 ←− p central ccls −→ ← p2 −1 ccls each of size p →
1 z ... z p−1 a ab ... a−1 b−1
1 1 ... 1
p2 linear 1 1 ... 1 char. table of
characters .. .. .. .. Cp × Cp lifted
. . . .
1 1 ... 1
p pω ... ...
p − 1 characters p
of degree p ..
.
p × char.
table of Cp 0
p

More formally,

• Z = hzi gives p conjugacy classes of size 1: {1}, {z}, . . ., {z p−1 }.

• G/Z = haZ, bZi = {ai bj Z : 0 6 i 6 p − 1, 0 6 j 6 p − 1}.
So, in particular, every element of G is of the form ai bj z k , 0 6 i, j, k, 6 p − 1.
• the p2 −1 conjugacy classes of size p are C(ai bj ) = {ai bj z k : 0 6 k 6 p−1, (i, j) 6= (0, 0)}.
For aba−1 b−1 = z : aba−1 = zb (= bz as z central)
bab−1 = az −1
⇒ aai bj a−1 = ai (aba−1 )j = ai bj z j .
bai bj b−1 = (bab−1 )i bj = ai bj z −i
I.e., any conjugate of ai bj is some ai bj z k , as above.

Irreducible characters

(15.15) Theorem. As above, let G = {ai bj z k : 0 6 i, j, k 6 p − 1} be a non-abelian group

of order p3 . Write ω = e2πi/p ∈ µp . Then the irreducible characters of G are:

χu,v (0 6 u, v 6 p − 1) (p2 of degree 1)

φu (1 6 u 6 p − 1) (p2 − 1 of degree p

where for all i, j, k,

χu,v (ai bj z k ) = wiu+jv

pω uk if i = j = 0
φu (ai bj z k ) =
0 otherwise

Proof. First, the p2 linear characters.

The irreducible characters of Gab = G/G′ = G/Z = Cp × Cp are ψu,v (ai bj Z) = ω iu+jv
(0 6 u, v 6 p − 1).

The lift to G of ψu,v is precisely χu,v .

Next, the p − 1 character of degree p.

  
 1 * * 
Now, H = 0 1 0 : * ∈ Fp ∼
= ha, zi is a normal abelian subgroup of index p.
 
0 0 1

47
 Let ψu be a character of H defined as ψu (ai z k ) = ω uk (0 6 k 6 p − 1), and calculate
ψuG .

Choose transversal {1, b, . . ., bp−1 } of H in G.

ψuG (ai z k ) = ψu (ai ) + ψu (ai z) + . . . + ψu (ai z p−1 )

p−1
X
= ψu (ai ) ψu (z r ) (as homomorphic)
r=0
p−1
X
= ψu (ai ) ω ur = 0
r=0
X ◦
ψuG (z k ) = ψ u (bj z k b−j ) = p ψu (z k ) = pω uk , and ψu (g) = 0 for all g ∈
/ H.
j

Thus ψuG = φu . Finally,

1 X 1 X
hφn , φn i = φu (g)φu (g) = φu (g)φu (g)
p3 p3
g∈G g∈Z
p−1
X
1 1 X 2
= φu (z k )φu (z k ) = p = 1
p3 p3 z
k=0

Remarks. 1. Alternative is to apply Mackey (12.6).

2. Typically for p-groups: any irreducible representation is induced from a 1-dimen-
sional representation of some subgroup (Telemann, chapter 17).
3. For p odd, in fact there are two non-abelian groups of order p3 :
2
G1 = ha, b : ap = bp = 1, b−1 ab = ap+1 i with Z = hap i
G2 = ha, b, z : ap = bp = z p = 1, az = za, bz = zb, b−1 ab = azi with Z = hzi

Notes typeset in LATEX by Gareth Taylor

Please let me know of any corrections: [email protected]

48
 PART II REPRESENTATION THEORY
SHEET 1

Unless otherwise stated, all groups here are finite, and all vector spaces are finite-dimensional
over a field F of characteristic zero, usually C.

1 Let ρ be a representation of the group G.

(a) Show that δ : g 7→ det ρ(g) is a 1-dimensional representation of G.
(b) Prove that G/ ker δ is abelian.
(c) Assume that δ(g) = −1 for some g ∈ G. Show that G has a normal subgroup of
index 2.

2 Let θ : G → F × be a 1-dimensional representation of the group G, and let ρ : G →

GL(V ) be another representation. Show that θ⊗ρ : G →GL(V ) given by θ⊗ρ : g 7→ θ(g)·ρ(g)
is a representation of G, and that it is irreducible if and only if ρ is irreducible.

3 Given any prime p. Find an example of a representation of some finite group over
some field of characteristic p, which is not completely reducible. Find an example of such a
representation in characteristic 0 for an infinite group. [Thus Maschke’s Theorem can fail if
F is not R or C or if G is not finite.]

4 Let N be a normal subgroup of the group G. Given a representation of the quotient

G/N , use it to obtain a representation of G. Which representations of G do you get this way?
Recall that the derived subgroup G0 of G is the unique smallest normal subgroup of G
such that G/G0 is abelian. Show that the 1-dimensional complex representations of G are
precisely those obtained from the abelianisation G/G0 .

5 Describe Weyl’s unitary trick.

Let G be a finite group acting on a complex vector space V , and let h , i be an alternating
bilinear form from V × V to C (so h P y, x i = −h x, y i for x, y in V ).
1
Show that the form (x, y) = |G| hgx, gyi, where the sum is over all elements g ∈ G, is
a G-invariant alternating form.
Does this imply that every finite subgroup of GL2m (C) is conjugate to a subgroup of the
symplectic group Sp2m (C)?

6 Let G be a cyclic group of order n. Decompose the regular representation of G explicitly

as a direct sum of 1-dimensional representations, by giving the matrix of change of coordinates
from the natural basis {eg }g∈G to a basis where the group action is diagonal.

7 Let G be the dihedral group D10 of order 10, with presentation

D10 = hx, y : x5 = 1 = y 2 , yxy −1 = x−1 i.
Show that G has precisely two 1-dimensional representations. By considering the effect of y
on an eigenvector of x show that any complex irreducible representation of G of dimension
at least 2 is isomorphic to one of two representations of dimension 2. Show that all these
representations can be realised over R.

1
2 PART II REPRESENTATION THEORY SHEET 1

8 Let G be the quaternion group with presentation

Q8 = hx, y | x4 = 1, y 2 = x2 , yxy −1 = x−1 i.
By considering the effect of y on an eigenvector of x show that any complex irreducible
representation of G of dimension at least 2 is isomorphic to the standard representation of
Q8 of dimension 2.
Show that this 2-dimensional representation cannot be realised over R; that is, Q8 is not
a subgroup of GL2 (R).

9 State Maschke’s theorem.

Show that any irreducible complex representation of the finite group G is isomorphic to
a subrepresentation of the regular representation of G.

10 State Schur’s lemma.

Show that if G is a finite group with trivial centre and H is a subgroup of G with
non-trivial centre, then any faithful representation of G is reducible on restriction to H.

11 Let G be a subgroup of order 18 of the symmetric group S6 given by

G = h(123), (456), (23)(56)i.
Show that G has a normal subgroup of order 9 and four normal subgroups of order 3. By
considering quotients, show that G has two representations of degree 1 and four inequivalent
irreducible representations of degree 2. Deduce that G has no faithful irreducible representa-
tions.

12 Work over F = R. Show that the cyclic group C3 = Z/3 has up to equivalence only
one non-trivial irreducible representation over R. If (ρ, V ) is this representation, show that
dimR HomG (V, V ) = 2. Comment.

13 Show that if ρ is a homomorphism from the finite group G to GLn (R), then there is a
matrix P ∈GLn (R) such that P ρ(g)P −1 is an orthogonal matrix for each g ∈ G. (Recall that
the real matrix A is orthogonal if At A = I.)
Determine all finite groups which have a faithful 2-dimensional representation over R.

14 Prove that for every finite simple group G, there exists a faithful irreducible complex
representation. (Hint: recall that the regular representation is faithful).

SM, Lent Term 2009

Comments and corrections on this sheet may be emailed to [email protected]
 PART II REPRESENTATION THEORY
SHEET 2

Unless otherwise stated, all groups here are finite, and all vector spaces are finite-dimensional
over a field F of characteristic zero, usually C.

1 Let ρ : G →GL(V ) be a representation of G of dimension d, and affording character χ.

Show that ker ρ = {g ∈ G | χ(g) = d}. Show further that |χ(g)| 6 d for all g ∈ G, with
equality only if ρ(g) = λI, a scalar multiple of the identity, for some root of unity λ.

2 Let χ be the character of a representation V of G and let g be an element of G. If g has

order 2, show that χ(g) is an integer and χ(g) ≡ χ(1) mod 2. If G is simple (but not C2 ),
show that in fact χ(g) ≡ χ(1) mod 4. (Hint: consider the determinant of g acting on V .) If
g has order 3 and is conjugate to g −1 , show that χ(g) ≡ χ(1) mod 3.

3 Construct the character table of the dihedral group D8 and of the quaternion group Q8 .
You should notice something interesting.

4 Construct the character table of the dihedral group D10 .

Each irreducible representation of D10 may be regarded as a representation of the cyclic
subgroup C5 . Determine how each irreducible representation of D10 decomposes into irre-
ducible representations of C5 .
Repeat for D12 ∼= S3 × C2 and the cyclic subgroup C6 of D12 .

5 Construct the character tables of A4 , S4 , S5 , and A5 .

The group Sn acts by conjugation on the set of elements of An . This induces an action
on the set of conjugacy classes and on the set of irreducible characters of An . Describe the
actions in the cases where n = 4 and n = 5.

6 A certain group of order 720 has 11 conjugacy classes. Two representations of this group
are known and have corresponding characters α and β. The table below gives the sizes of the
conjugacy classes and the values which α and β take on them.
1 15 40 90 45 120 144 120 90 15 40
α 6 2 0 0 2 2 1 1 0 −2 3
β 21 1 −3 −1 1 1 1 0 −1 −3 0
Prove that the group has an irreducible representation of degree 16 and write down the
corresponding character on the conjugacy classes.

7 The table below is a part of the character table of a certain finite group, with some
of the rows √
missing. The columns√ are labelled by the sizes of the conjugacy classes, and
γ = (−1 + i 7)/2, ζ = (−1 + i 3)/2. Complete the character table. Describe the group in
terms of generators and relations.
1 3 3 7 7
χ1 1 1 1 ζ ζ̄
χ2 3 γ γ̄ 0 0
χ3 3 γ̄ γ 0 0

1
2 PART II REPRESENTATION THEORY SHEET 2

8 Let x be an element of order n in a finite group G. Say, without detailed proof, why
(a) if χ is a character of G, then χ(x) is a sum of nth roots of unity;
(b) τ (x) is real for every character τ of G if and only if x is conjugate to x−1 ;
(c) x and x−1 have the same number of conjugates in G.
(d) Prove that the number of irreducible characters of G which take only real values
(so-called real characters) is equal to the number of self-inverse conjugacy classes (so-called
real classes).
A group of order 168 has 6 conjugacy classes. Three representations of this group are
known and have corresponding characters α, β and γ. The table below gives the sizes of the
conjugacy classes and the values α, β and γ take on them.
1 21 42 56 24 24
α 14 2 0 −1 0 0
β 15 −1 −1 0 1 1
γ 16 0 0 −2 2 2
Construct the character table of the group.√
[You may assume, if needed, the fact that 7 is not in the field Q(ζ), where ζ is a primitive
7th root of unity.]

9 Let a finite group G act on itself by conjugation. Find the character of the corresponding
permutation representation.

10 Let G have conjugacy class representatives g1 , . . . , gk and character table Z. Show that
det Z is either real or purely imaginary, and that
k
Y
2
| det Z| = |CG (gi )|.
i=1

Compute ±(det Z) when G ∼

= C3 .
 PART II REPRESENTATION THEORY SHEET 2 3

11 The character table obtained in Question 8 is in fact the character table of the group
G =PSL2 (7) of 2×2 matrices with determinant 1 over the field F7 (of seven elements) modulo
the two scalar matrices.
Deduce directly from the character table which you have obtained that G is simple.
[Comment: it is known that there are precisely five non-abelian simple groups of order
less than 1000. The smallest of these is A5 ∼=PSL2 (5), while G is the second smallest. It is
also known that for p > 5, PSL2 (p) is simple.]
Identify the columns corresponding to the elements x and y where x is an element of
order 7 (eg the unitriangular matrix with 1 above the diagonal) and y is an element of order
3 (eg the diagonal matrix with entries 4 and 2).
The group G acts as a permutation group of degree 8 on the set of Sylow 7-subgroups
(or the set of 1-dimensional subspaces of the vector space (F7 )2 ). Obtain the permutation
character of this action and decompose it into irreducible characters.
Show that the group G is generated by an element of order 2 and an element of order 3
whose product has order 7.
[Hint: for the last part use the formula that the number of pairs
P of elements−1conjugate to x
and y respectively, whose product is conjugate to t, equals c χ(x)χ(y)χ(t )/χ(1), where
the sum runs over all the irreducible characters of G, and c = |G|2 (|CG (x)||CG (y)||CG (t)|)−1 .]

SM, Lent Term 2009

Comments and corrections on this sheet may be emailed to [email protected]
 PART II REPRESENTATION THEORY
SHEET 3

Unless otherwise stated, all groups here are finite, and all vector spaces are finite-dimensional
over a field F of characteristic zero, usually C.

1 Recall the character table of S4 from Sheet 2. Find all the characters of S5 induced from
the irreducible characters of S4 . Hence find the complete character table of S5 .
Repeat, replacing S4 by the subgroup h(12345), (2354)i of order 20 in S5 .

2 Recall the construction of the character table of the dihedral group D10 of order 10 from
Sheet 2.
(a) Use induction from the subgroup D10 of A5 to A5 to obtain the character table of
A5 .
(b) Let G be the subgroup of SL2 (F5 ) consisting of upper triangular matrices. Compute
the character table of G. Hint: bear in mind that there is an isomorphism G/Z → D10 )

3 Let H be a subgroup of the group G. Show that for every irreducible representation
ρ for G there is an irreducible representation ρ0 for H with ρ a component of the induced
representation IndG 0
H ρ.
Prove that if A is an abelian subgroup of G then every irreducible representation of G
has dimension at most |G : A|.

4 Obtain the character table of the dihedral group D2m of order 2m, by using induction
from the cyclic subgroup Cm . Note that it matters whether m is odd or even.

5 Calculate χΛ2 ρ and χS 2 ρ , where ρ is the irreducible representation of dimension 2 of D8 ;

repeat this for Q8 . Which of these characters contains the trivial character in the two cases?

6 Let ρ : G →GL(V ) be a representation of G of dimension d.

(a) Compute the dimension of S n V and Λn V for all n.
(b) Let g ∈ G and let λ1 , . . . , λd be the eigenvalues of g on V . What are the eigenvalues
of g on S n V and Λn V ?
(c) Let f (x) = det(g − xI) be the characteristic polynomial of g on V . Describe how to
obtain the trace χΛn V (g) from the coefficients of f (x).
(d)* Find a relation between χS n V (g) and the polynomial f (x). [Hint: do the case where
V has dimension 1 first.]

7 Let G be the symmetric group Sn acting naturally on the set X = {1, . . . , n}. For any
integer r 6 n2 , write Xr for the set of all r-element subsets of X, and
let πr be the permutation
n
character of the action of G on Xr . Observe πr (1) = |Xr | = r . If 0 6 ` 6 k 6 n/2, show
that
hπk , π` i = ` + 1.
Let m = n/2 if n is even, and m = (n − 1)/2 if n is odd. Deduce that Sn has distinct
irreducible characters χ(n) = 1G , χ(n−1,1) , χ(n−2,2) , . . . , χ(n−m,m) such that for all r 6 m,
πr = χ(n) + χ(n−1,1) + χ(n−2,2) + · · · + χ(n−r,r) .
In particular the class functions πr − πr−1 are irreducible characters of Sn for 1 6 r 6 n/2
and equal to χ(n−r,r) .
1
2 PART II REPRESENTATION THEORY SHEET 3

8 Given any complex representation V of the cyclic group Z/2, write down the projections
to the two isotypical summands of V , directly from the action of G on V . Show that your
formulae give a decomposition of V as a direct sum of two subspaces even if V is an infinite-
dimensional representation of Z/2.
More generally, given any complex representation V of any finite cyclic group Z/n, write
down the projections to the n isotypical summands of V , directly from the action of G on V .

9 If ρ : G →GL(V ) is an irreducible complex representation for G affording character χ,

find the characters of the representation spaces V ⊗ V , Sym2 (V ) and Λ2 (V ).
Define the Frobenius-Schur indicator ιχ of χ by
1 X
ιχ = χ(x2 )
|G| x∈G
and show that 
0, if χ is not real-valued
ιχ =
±1, if χ is real-valued.
[Remark. The sign +, resp. −, indicates whether ρ(G) preserves an orthogonal, respectively,
symplectic form on V , and whether or not the representation can be realised over the reals.
You can read about it in Isaacs or in James and Liebeck.]

10 The group G × G acts on G by (g, h)(x) = gxh−1 . In this way, the regular representation
space CG becomes a G × G-space. (So far, we only considered CG as a representation space
of the group G × {1} ≤ G × G.)
Determine the character π of G × G in this action. For each irreducible character χψ of
G × G, determine its multiplicity in π. Compare π to the character of the subgroup G × {1}
in this action.

11 If θ is a faithful character of the group G, which takes r distinct values on G, prove that
each irreducible character of G is a constituent of θ to power i for some i < r.
[Hint: assume that hχ, θi i = 0 for all i < r; use the fact that the Vandermonde r × r matrix
involving the row of the distinct values a1 , ..., ar of θ is nonsingular to obtain a contradiction.]

12 Construct the character table of the symmetric group S6 . Identify which of your char-
acters are equal to the characters χ(6) , χ(5,1) , χ(4,2) , χ(3,3) constructed in question 7.

SM, Lent Term 2009

Comments and corrections on this sheet may be emailed to [email protected]
 PART II REPRESENTATION THEORY
SHEET 4

Unless otherwise stated, all vector spaces are finite-dimensional over C. In the first eight
questions we let G = SU(2). The last four questions are roughly of Tripos standard.

1 (a) Let Vn be the vector space of complex homogeneous polynomials of degree n in the
variables x and y. Describe a representation ρn of G on Vn and show that it is irreducible.
Describe the character χn of ρn .
(b) Decompose V4 ⊗ V3 into irreducible G-spaces (that is, find a direct sum of irreducible
representations which is isomorphic to V4 ⊗ V3 . In this and the following questions, you are
not being asked to find such an isomorphism explicitly.)
(c) Decompose also V3⊗2 , Λ2 V3 and S 2 V3 .
(d) Show that Vn is isomorphic to its dual Vn∗ .

2 Decompose V1⊗n into irreducibles.

3 Determine the character of S n V1 for n ≥ 1.

Decompose S 2 Vn and Λ2 Vn for n ≥ 1.
Decompose S 3 V2 into irreducibles.

4 Let G =SU(2) act on the space M3 (C) of 3 × 3 complex matrices, by

A : X 7→ A1 XA−1
1 ,
where A1 is the 3 × 3 block diagonal matrix with block diagonal entries A, 1. Show that this
gives a representation of G and decompose it into irreducibles.

5 Let χn be the character of the irreducible representation ρn of G on Vn .

Show that Z 2π
1
K(z)χn χm dθ = δnm ,
2π 0
where z = eiθ and K(z) = 21 (z − z −1 )(z −1 − z).
[ Note
R 2π that all you need to know about integrating on the circle is orthogonality of characters:
1 n
2π 0
z dθ = δn,0 . This is really a question about Laurent polynomials. ]

6 (a) Let G be a compact group. Show that there is a continuous group homomorphism
ρ : G →O(n) if and only if G has an n-dimensional representation over R. Here O(n) denotes
the subgroup of GLn (R) preserving the standard (positive definite) symmetric bilinear form.
(b) Explicitly construct such a representation ρ : SU(2) → SO(3) by showing that SU(2) acts
on the vector space of matrices of the form
   
a b t
A= ∈ M2 (C) : A + A = 0
c −a
by conjugation. Show that this subspace is isomorphic to R3 , that (A, B) 7→ −tr(AB) is a
positive definite non-degenerate invariant bilinear form, and that ρ is surjective with kernel
{±I}.

1
2 PART II REPRESENTATION THEORY SHEET 4

7 Check that the usual formula for integrating functions defined on S 3 ⊆ R4 defines an
SU(2)-invariant inner product on
  
a b
SU(2) = : aā + bb̄ = 1 ,
−b̄ ā
and normalize it so that the integral over the group is one.

8 Compute the character of the representation S n V2 of G for any n > 0. Calculate

dimC (S n V2 )G (by which we mean the subspace of S n V2 where G acts trivially).
Deduce that the ring of complex polynomials in three variables x, y, z which are invariant
under the action of SO(3) is a polynomial ring. Find a generator for this polynomial ring.

9 The Heisenberg group of order p3 is the (non-abelian) group

  
 1 a x 
G=  0 1 b  : a, b, x ∈ Fp .
0 0 1
 

of 3 × 3 upper unitriangular matrices over the finite field Fp of p elements (p prime).

Show that G has p conjugacy classes of size 1, and p2 − 1 conjugacy classes of size p.
Find p2 characters of degree 1.
Let H be the subgroup of G comprising matrices with a = 0. Let ψ : Fp → C× be a non-
trivial 1-dimensional representation of the cyclic group Fp = Z/p, and define a 1-dimensional
representation ρ of H by  
1 0 x
ρ  0 1 b  = ψ(x).
0 0 1
Check that Vψ = IndG H ρ is irreducible.
Now list all the irreducible representations of G, explaining why your list is complete.

10 Recall that, up to isomorphism, there are precisely two non-abelian groups of order p3 .
When p = 2 they are D8 and Q8 . Suppose p = 3 and let H be the group of order 27 which
is given by:
H = ha, b, z : a3 = b3 = z 3 = 1, az = za, bz = zb, b−1 ab = azi.
List the conjugacy classes of H, and use Theorem 16.1 to write down the character table of
H.
 PART II REPRESENTATION THEORY SHEET 4 3

11 Recall Sheet 3, q.7 where we used inner products to construct some irreducible characters
χ(n−r,r) for Sn . Let n ∈ N, and let Ω be the set of all ordered pairs (i, j) with i, j ∈ {1, 2, . . . , n}
and i 6= j. Let G = Sn act on Ω in the obvious manner (namely, σ(i, j) = (σi, σj) for σ ∈ Sn ).
Let’s write π (n−2,1,1) for the permutation character of Sn in this action.
Prove that
π (n−2,1,1) = 1 + 2χ(n−1,1) + χ(n−2,2) + ψ,
where ψ is an irreducible character. Writing ψ = χ(n−2,1,1) , calculate the degree of χ(n−2,1,1) .
Find its value on any transposition and on any 3-cycle. Returning to the character table of
S6 calculated on Sheet 3, identify the character χ(4,1,1) .

SM, Lent Term 2009

Comments and corrections on this sheet may be emailed to [email protected]