You are on page 1of 23

Representation Theory

Michael Bushell
michael.bushell@student.manchester.ac.uk
January 16, 2014
Contents
1 Introduction 3
1.1 Matrix Representations . . . . . . . . . . . . . . . . . . . . . . 3
1.2 G-Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Correspondence to Matrix Representations . . . . . . . 3
1.3 Constructing Representations . . . . . . . . . . . . . . . . . . 4
2 Decomposition of G-Spaces 5
2.1 Decomposition of Vector Spaces . . . . . . . . . . . . . . . . . 5
2.2 Maschkes Theorem . . . . . . . . . . . . . . . . . . . . . . . . 6
3 Linear Representations 8
3.1 Cyclic Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Abelian Groups . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3 Non-abelian Groups . . . . . . . . . . . . . . . . . . . . . . . . 9
4 Schurs Lemma 10
5 Characters 11
5.1 Character Orthogonality . . . . . . . . . . . . . . . . . . . . . 13
5.1.1 Inner Product on Class Functions . . . . . . . . . . . . 13
5.1.2 The Fixed Point Space . . . . . . . . . . . . . . . . . . 13
5.1.3 Row Orthogonality . . . . . . . . . . . . . . . . . . . . 14
5.2 Unique Decomposition of G-Spaces . . . . . . . . . . . . . . . 15
5.3 Column Orthogonality . . . . . . . . . . . . . . . . . . . . . . 18
6 G-Sets 19
1
CONTENTS 2
7 Induced Representation 21
7.1 Frobenius Reciprocity . . . . . . . . . . . . . . . . . . . . . . . 23
1 INTRODUCTION 3
1 Introduction
1.1 Matrix Representations
Let G be a nite group, then a matrix representation of G is a group homo-
morphism : G GL
n
(C) said to be of dimension n.
Matrix representations
1
and
2
are equivalent if there exists T GL
n
C
such that

1
(g) = T
2
(g)T
1
for all g G. If
1
,
2
are viewed as linear transformations of C
n
, then they
are equivalent if there exists a change of basis that transforms one into the
other.
1.2 G-Spaces
A G-space is a C-vector space V with an action G V V , (g, v) gv,
satisfying
(i) for all g G the map v gv is a linear transformation,
(ii) for group identity e G, we have ev = v for all v V ,
(iii) for all g
1
, g
2
V and v V , we have (g
1
g
2
)(v) = g
1
(g
2
v).
A homomorphism of G-spaces is a linear map f : V
1
V
2
such that
f(gv) = gf(v) for all g G, v V . It is an isomorphism if it has an inverse
which is also a homomorphism of G-spaces.
Proposition 1.1. A homomorphism of G-spaces is an isomorphism of G-
spaces if and only if it is an isomorphism of vector spaces.
1.2.1 Correspondence to Matrix Representations
Proposition 1.2. There is a one-to-one correspondence
isomorphism classes of G-spaces equivalence classes of matrix representations
Proof. If : G GL
n
(C) is a matrix representation, then each g G
denes a linear map (g) : C
n
C
n
. Hence, dene the action of G on C
n
by (g, v) (g)v, this is a G-space.
Let be V a G-space and choose a basis for V , then each g G denes a
linear map by v gv, so let [g] be the matrix for this map with respect to
the chosen basis, then (g) = [g] denes a matrix representation.
One can show these are mutually inverse to each other, up to isomorphism
of G-spaces/equivalence of matrix representations.
1 INTRODUCTION 4
From now on we will go freely between the two viewpoints for represen-
tations of groups, the matrix representations and G-spaces, depending on
which is more convenient.
A representation is faithful if only the identity 1 G acts as the identity
transformation and it is trivial if all elements of G act as the identity.
1.3 Constructing Representations
There are various ways to construct representations.
(1) The regular representation of a group G is the group-algebra CG on
which G acts by left multiplication.
(2) If H G is a subgroup, then any G-space restricts to a H-space.
(3) Given a group homomorphism f : H G, we can inate a G-space to
a H-space by the action hv = f(h)v.
(4) If V is a G-space and W V a linear subspace such that gW W for
all g G, then W is also a G-space, called a G-subspace of V .
(5) If W V is a G-subspace, then the quotient space V/W is a G-space
under the action g(v +W) = g(v) +W.
(6) If V and W are two G-spaces, then their direct sum V W is a G-space
under the action g(v, w) = (gv, gw).
(7) If f : V W is a G-space homomorphism, then ker(f) is a G-subspace
of V and im(f) is a G-subspace of W.
(8) If V and W are two G-spaces, then Hom
C
(V, W) is a G-space under the
action gf = gfg
1
.
(9) Let V be a G-space, then V
G
= v V [ gv = v, g G is the xed
point space of V .
Proposition 1.3. The xed point space Hom
C
(V, W)
G
of the linear maps
V W is equal to Hom
G
(V, W) the set of G-space homomorphism V W.
Proof. f Hom
C
(V, W)
G
f = gfg
1
fg = gf f Hom
G
(V, W).
2 DECOMPOSITION OF G-SPACES 5
2 Decomposition of G-Spaces
A G-space is called irreducible if it has no proper, non-trivial G-subspaces;
and indecomposible if it cannot be written as a direct sum of proper, non-
trivial G-subspaces. Clearly irreducible G-spaces are indecomposible but the
converse isnt immediately obvious, as there may exist a G-subspace with no
direct complement.
Maschkes theorem tells us this is never the case and the two denitions
coincide. Moreover, this implies that every G-space can be decomposed as
a direct sum of irreducibles, so to classify G-spaces it is sucient to classify
the irreducibles.
2.1 Decomposition of Vector Spaces
Decompositions of vectors spaces (as direct sums) correspond to idempotents
in End
C
(V ) the endomorphism ring of C-linear maps V V .
Lemma 2.1. Let V be a vector space and such that V = M N, then
p : V V by p(m + n) = m the projection onto M is an idempotent.
Moreover, N = ker(p), M = im(p) and 1 p is the projection onto N.
Proof. We have p(m+n) = m and p(m) = m, so certainly p
2
= p and is an
idempotent, clearly N = ker(p) and M = im(p). Finally, (1 p)(m + n) =
m+n m = n, so 1 p is the projection onto N.
Conversely,
Lemma 2.2. If p End
C
(V ) is an idempotent, then V = ker(p) im(p).
Proof. Let y ker(p) im(p), then y = p(x) for some x V and
0 = p(y) = p(p(x)) = p
2
(x) = p(x) = y
hence ker(p) im(p) = 0.
Also y = (y p(x)) +y and
p(y p(x)) = p(y) p(p(x)) = p(y) p(y) 0
so y p(x) ker(p), hence ker(p) + im(p) = V , as required.
2 DECOMPOSITION OF G-SPACES 6
2.2 Maschkes Theorem
Theorem 2.3. Let W be a G-space and U W be a G-subspace, then there
exists a G-subspace V W such that W = U V .
Proof. We know there exists a vector subspace V W such that W = U V
as a direct sum of vector spaces; from this we construct the required G-
subspace.
Let p : W W be the projection onto V , then W = ker(p) im(p),
ker(p) = U and im(p) = V .
Let q : W W be dened by q =
1
|G|

gG
gpg
1
, then
(i) q is a G-space homomorphism it is a linear map since it is the
composition of linear maps and
qh =
1
[G[

gG
gpg
1
h
=
1
[G[

gG
h(h
1
g)p(h
1
g)
1
by associativity,
=
h
[G[

gG
gpg
1
as h
1
g runs through G as g does,
= hq
for all h G as required.
(ii) q is an idempotent rstly im(q) V as im(p) = V and V is a G-
space. If g G and v V , then p(gv) = gv (as gV V and p is the
projection onto V ), it follows that
q(v) =
1
[G[

gG
gpg
1
v =
1
[G[

gG
gg
1
v =
1
[G[

gG
v = v
for all v V , therefore im(q) = V .
If w W, then q(w) V , so q(q(w)) = q(w), thus q is an idempotent.
So W = ker(q)im(q) = UV with V is a G-space of W, as claimed.
Corollary 2.4. Every G-space V contains irreducible G-subspaces V
1
, . . . , V
r
such that V =

r
i=1
V
i
.
Proof. By induction on d = dim
C
(V ), if d = 1, then V is irreducible itself.
For d 1, if V is irreducible were done, otherwise it contains a proper,
non-trivial G-subspace U W and W = U V for a G-subspace V W
by Maschkes theorem, apply induction on U and V .
2 DECOMPOSITION OF G-SPACES 7
Example 2.5. Let U W be a G-subspace, then by Maschkes theorem
there exists a G-subspace V W such that W = U V .
Consider the quotient map q : W W/U, this is the canonical surjection
w w + U with ker(q) = U. For v
1
, v
2
V , we have q(v
1
) = q(v
2
)
v
1
+U = v
2
+U v
1
v
2
U but as W = U V we know U V = 0,
therefore v
1
v
2
= 0 and v
1
= v
2
, hence the restriction of q on V is a
bijection. Moreover, q is a G-space homomorphism, it is a linear map and
q(gw) = (gw)+U = g(w+U) by denition of the quotient G-space. Therefore
q restricts to an isomorphism and V

= W/U, hence W

= U W/U.
3 LINEAR REPRESENTATIONS 8
3 Linear Representations
A 1-dimensional representation : G GL
1
(C)

= C

is called a linear
representation. If [G[ = n, then for all g G there exists an m 1 such that
g
m
= 1, therefore (g)
m
= 1 and (g) is a primitive mth root of unity and
an nth root of unity. It follows that : G C

is a group homomorphism
into the roots of unity.
3.1 Cyclic Groups
Theorem 3.1. An irreducible representation of a cyclic group is 1-dimensional.
Proof. Let G be a cyclic group of order n with generator g, so
G = g, g
2
, . . . , g
n1
, g
n
= 1
Suppose G has matrix representation : G GL
n
(C), then (g) has an
eigenvalue C, since C is an algebraically closed eld.
Let v V a corresponding eigenvector and V the G-space, then the
line Cv = cv [ c C V on which g acts by multiplication by , is a
G-subspace as
g
i
(cv) = cg
i
v = c
i
v Cv
Therefore, if V is irreducible, then V = Cv and dim
C
(V ) = 1.
Corollary 3.2. A cyclic group of order n has n linear representations.
Proof. There are n choices of (g) = [
i
], as
i
must be an nth root of unity,
and this denes the representation.
.
Corollary 3.3. Let V be a representation of a cyclic group G with generator
g G, then there exists a basis in which
(g) =
_
_
_
_
_

1
0 0
0
2
0 0
0
.
.
.
.
.
.
0
0 0
n
_
_
_
_
_
for some
1
, . . . ,
n
C.
3 LINEAR REPRESENTATIONS 9
Proof. By Maschkes and the above theorem, we can decompose V into a
direct sum of 1-dimensional irreducibles. It follows that the corresponding
matrix (g) (which is a linear transformation V V ) is diagonalizable, as
claimed.
Furthermore, each
i
is an nth root of unity where n = [G[ is the order of
the group. As all h G can be written h = g
i
, it follows that every matrix
(h) is simultaneously diagonalizable.
3.2 Abelian Groups
Theorem 3.4. An irreducible representation of an abelian group is 1-dimensional.
Proof. Let V be an irreducible G-space, we proceed by induction on [G[.
The result holds for [G[ = 1, as then G is cyclic and the previous theorem is
applicable.
For [G[ 1 we can write G = a H, for some non-zero a G, by the
fundamental theorem of abelian groups.
If H = 0, then G is cyclic, otherwise we can consider V as a H-space.
Let be an eigenvalue for (a) and V

be its eigenspace, then V

is a G-
subspace. Let v V

and g G, then agv = gav = gv = gv, so gv V

as required. As V is irreducible, then V = V

and the action of a on V is


scalar multiplication.
It follows that any H-subspace is also a G-subspace, so it must be 0 or
V as V is irreducible as a G-space hence V is irreducible as a H-space.
Since 1 [H[ < [G[, we can use induction and conclude dim
C
(V ) = 1.
Corollary 3.5. Let V be a representation of an abelian group G, then there
exists a basis in which, for g G, we can write
(g) =
_
_
_
_
_

1
0 0
0
2
0 0
0
.
.
.
.
.
.
0
0 0
n
_
_
_
_
_
for some
1
, . . . ,
n
; and we can do so simultaneously for every g G.
3.3 Non-abelian Groups
If G is a group, then [G, G] the commutator subgroup of G is such that
G/[G, G] is an abelian group; it is the largest abelian quotient group of
G. Any irreducible (and necessarily linear) representation of G/[G, G] gives
an irreducible linear representation of G, by ination via the quotient map
4 SCHURS LEMMA 10
q : G G/[G, G]. Conversely, if : G C

is a linear representation, then


(ghg
1
h
1
) = 1, so [G, G] ker() and is a representation of G/[G, G].
Hence, all the linear representations of G come from those of G/[G, G].
We have completely classied the representations of abelian groups (they
are all linear) and the linear representations of non-abelian groups; but a non-
abelian group may have an irreducible representation of higher dimension.
4 Schurs Lemma
A useful result is the following known as Schurs lemma.
Lemma 4.1. Let V and W be irreducible G-spaces and f : V W a G-space
homomorphism, then
(1) If V ,

= W, then f = 0,
(2) If V = W, then f is multiplication by a scalar.
Proof. (1) We know ker(f) V and im(f) W are G-subspaces, so by
irreducibility are 0 or the whole space. We cannot have ker(f) = 0 and
im(f) = W, otherwise f is a bijection and thus a G-space isomorphism,
then V

= W contrary to our assumption. So either ker = V or im = 0,
hence f = 0 as claimed.
(2) If V = W, then f must have an eigenvalue C, in which case the
eigenspace V

is a non-trivial G-subspace of V , but by the irreducibility


of V this means V = V

and f is multiplication by .
5 CHARACTERS 11
5 Characters
The character of a representation : G GL
n
C is the function

: G C,

(g) = tr((g))
We will use characters to study representations, as tr is a class function
(dened up to conjugacy class)

is also class function and is dened up to


the equivalence class of the representation .
Lemma 5.1. Let be the character of an n-dimensional representation, then
(1) (1) = n,
(2) (g
1
) = (g),
(3) (hgh
1
) = (g).
Proof.
(1) (1) = tr(Id
n
) = n,
(2) For g G we can choose a basis in which (g) = diag(
1
, . . . ,
n
), where

i
are roots of unity. As (g
1
) = (g)
1
= diag(
1
1
, . . . ,
1
n
) and

1
i
=
i
for such roots of unity, the result follows.
(3) (hgh
1
) = tr((hgh
1
)) = tr((h)(g)(h)
1
) and tr is a class function
on matrices.
We can now give a result that tells us characters of representations can
be added an multiplied to produce other characters, this is important, as
it does not required knowledge of the underlying representation to produce
another character.
Lemma 5.2. If V and W are G-spaces, then
(1)
V W
=
V
+
W
,
(2)
Hom
C
(V,W)
=
V

W
,
(3)
Hom
C
(V

,W)
=
V

W
.
Proof. (1) We have

V W
(g) =
_

V
(g) 0
0
W
(g)
_
so tr(
V W
) = tr(
V
(g)) + tr(
W
(g)), as required.
5 CHARACTERS 12
(2) If g : X X is a linear map and X has an eigenbasis, then g is diagonal
with respect to this basis and tr(g) is the sum of the eigenvalues (with
multiplicity). We will apply this with X = Hom
C
(V, W).
Given g G we can choose a basis v
1
, . . . , v
m
V in which
V
(g) =
diag(
1
, . . . ,
m
) and a basis w
1
, . . . , w
n
W such that
W
(g) =
diag(
1
, . . . ,
n
).
For Hom
C
(V, W) choose the basis H
i,j
where H
i,j
is a zero nm-matrix
except the (i, j)-entry is 1.
Then the action for g G on Hom
C
(V, W) is
gH
i,j
=
W
(g) H
i,j

V
(g
1
) =
i

j
H
i,j
so H
i,j
forms an eigenbasis for g with eigenvalues
i

j
, the trace is

Hom
C
(V,W)
(g) = tr(
Hom
C
(V,W)
(g)) =

i,j

i
=
V
(g)
W
(g)
as claimed.
(3) As
V
=
Hom
C
(V,C)
=
V

C
=
V
since
C
= 1 for the trivial represen-
tation C. Hence
Hom
C
(V

,W)
=
V

W
=
V

W
as claimed.
5 CHARACTERS 13
5.1 Character Orthogonality
5.1.1 Inner Product on Class Functions
An hermitian inner product on a vector space V is a map V V C
satsifying
(i) ((a +b), c) = (a, c) +(b, c),
(ii) (a, b) = (b, a),
(iii) (a, a) 0 and (a, a) = 0 a = 0.
for all a, b V , C.
We can dene a positive denite hermitian inner product of class functions
on a group G (which includes characters of its representations) by
(, ) =
1
[G[

gG
(g)(g)
We will nd that the characters of irreducible representations (irrreducible
characters) have special properties with regards to this inner product.
5.1.2 The Fixed Point Space
If V is a representation of a group G, then the xed point space is
V
G
= v V [ gv = v, g G
Lemma 5.3. Let V be a G-space and dene
: V V, v
1
[G[

gG
gv
then
(1) (v) V
G
,
(2) If v V
G
, then (v) = v,
(3) V = V
G
ker().
Proof. This is similar to the proof of Maschkes theorem (1) and (2)
are obvious and tell us im() = V
G
and is an idempotent, so we have
a vector space decomposition V = V
G
ker(), but is a G-space
homomorphism so ker() is also a G-space, establishing (3).
5 CHARACTERS 14
Lemma 5.4. dim
C
(V
G
) =
1
|G|

gG

V
(g).
Proof. We rst make the observation that if V = U
1
U
2
and p is projection
onto U
1
with kernal U
2
, then dim
C
(U
1
) = tr(p). From this, it follows that
dim
C
(V
G
) = tr() =
1
[G[

gG
tr((g)) =
1
[G[

gG

V
(g)
as claimed.
Corollary 5.5. dim
C
(V
G
) = (1,
V
) = (
v
, 1).
Proof. Here 1 denotes the trivial representation C. The result is immediate
from the previous lemma and denition of the inner product.
5.1.3 Row Orthogonality
Lemma 5.6. (, ) = (1, ).
Theorem 5.7. dim(Hom
G
(V, W)) = (
V
,
W
).
Proof.
dim
C
(Hom
G
(V, W)) = dim
C
(Hom
C
(V, W)
G
)
= (1,
Hom
C
(V,W)
)
= (1,
V

W
)
= (
V
,
W
)
Corollary 5.8. If V and W are irreducible representations, then
(
V
,
W
) =
_
1 V

= W
0 V ,

= W
Proof. From the above theorem and Schurs lemma.
This result is known as row orthogonality, the names refers to the char-
acter table we dene later.
Corollary 5.9. The number of irreducible G-spaces is the number of con-
jugacy classes.
Proof. The above corollary says the characters of irreducible representations
are orthogonal in the space of class functions on G. They are therefore
linearly independent and cant be greater in number than the dimension of
the space, which is easily seen to be the number of conjugacy classes.
In fact we will prove later that equality holds and irreducible characters
are an orthonormal basis of the space of class functions on a group G.
5 CHARACTERS 15
5.2 Unique Decomposition of G-Spaces
Theorem 5.10. Let V =

n
i=1
V
i
be a G-space written as a sum of irre-
ducibles. If W is another irreducible G-space, then the number of V
i
isomor-
phic to W is equal to (
V
,
W
).
Proof.
(
V
,
W
) = (

n
i=1
V
i
,
W
) =
n

i=1
(
V
i
,
W
)
Now each term in the sum is either 1 or 0 depending on whether V
i

= W or
not, respectively, so the result follows.
Corollary 5.11.
(1) The decomposition of a representation into irreducibles is essentially
unique, i.e. up to isomorphism and reordering of direct summands.
(2) Any two representations with the same character must be isomorphic.
(3) If
1
, . . . ,
n
are all the irreducible characters, then
V
=

n
i=1
(
V
,
i
)
i
for any arbitrary character
V
.
(4) A G-space V is irreducible if and only if (
V
,
V
) = 1.
Proof. If V is an irreducible G-space, then (
V
,
V
) = 1 by the above results.
Conversely, if (
V
,
V
) = 1, write V =

n
i=1
V
i
as a sum of irreducibles, then
1 = (
V
,
V
) =
n

i=1
(
V
,
V
i
)
Each term is a non-negative integer, so all but say the jth term must be 0,
then (
V
,
V
j
) = 1 implies V

= V
j
, hence V is irreducible.
Proposition 5.12. The regular representation has decomposition
CG

=
n

i=1
V
(dim
C
(V
i
))
i
where V
1
, . . . , V
n
are all the irreducible representations.
Proof. For irreducible character
V
i
we have
(
CG
,
V
i
) =
1
[G[

gG

CG
(g)
V
i
(g) =
V
i
(1) = dim
C
(V
i
)
since
CG
is zero except on 1 where it is [G[ = dim
C
(V
i
).
5 CHARACTERS 16
Corollary 5.13. [G[ =

n
i=1
(dim
C
(V
i
))
2
.
Proof. From the previous proposition
(
CG
,
CG
) =
_

CG
,
n

i=1
dim
C
(V
i
)
V
i
_
=
n

i=1
dim
C
(V
i
) (
CG
, dim
C
(V
i
))
=
n

i=1
dim
C
(V
i
)
2
but by denition of the inner product we also have
(
CG
,
CG
) =
1
[G[

gG

CG
(g)
CG
(g) = [G[
as required.
We now come to an important result that has many consequences and
gives the theory of characters its power.
5 CHARACTERS 17
Theorem 5.14. The number of irreducible characters of a group G is equal
to dim
C
(Cl (G)) the number of conjugacy classes.
Proof. Let
1
, . . . ,
n
be a complete set of irreducible characters. Weve
already seen they are linearly independent in the space of class functions
Cl (G), hence if the theorem is false then they cannot span this space.
Take h Cl (G) not in the span of the irreducibles, by the Gram-Schmidt
process, we can create a function orthogonal to each of the irreducibles.
h f = h (h,
1
)
1
(h,
2
)
2
(h,
n
)
n
Given a G-space V and corresponding matrix representation let

f
=

gG
f(g)(g)
which is a linear transformation V V by v

gG
f(g)gv.
Now
f
is a G-space homomorphism, since
(h)
f
(h
1
) =

gG
f(g)(h)(g)(h
1
) =

gG
f(g)(hgh
1
)
=

gG
f(hgh
1
)(hgh
1
) =

gG
f(g)(g) =
f
using the fact that f is a class function.
If V is irreducible, then by Schurs lemma
f
is multiplication by some
scalar C, hence
f
=

gG
f(g)(g) = Id
n
; taking traces gives
dim
C
(V ) =

gG
f(g)
V
(g), hence
=
1

V
(1)

gG
f(g)
V
(g) =
[G[

V
(1)
(f,
V
)
Now let V = CG be the regular representation, which we recall is a direct
sum of irreducible characters and so (f,
V
) = 0 hence = 0 and
0 = 1 =
f
(1) =

gG
f(g)g 1 =

gG
f(g)g
but this is a linear dependence on the basis elements of V , thus f = 0.
It follows that the irreducible characters must span Cl (G).
5 CHARACTERS 18
5.3 Column Orthogonality
Theorem 5.15. Let
1
, . . . ,
n
be a complete set of irreducible characters of
a group G and g
1
, . . . , g
n
conjugacy class representatives, then
n

k=1

k
(g
i
)
k
(g
j
) = [C
G
(g
i
)[
i,j
for 1 i, j n.
Proof. Write the character table of G as a matrix A, then the inner product
on rows (characters) i, j has been dened
(
i
,
j
) = (r
i
, r
j
) =
1
[G[
n

k=1
c
k
r
i
k
r
j
k
where c
k
is the size of the conjugacy class represented by column k.
Let C = diag(c
1
, . . . , c
n
), then row orthogonality can be expression as
1
[G[
ACA
T
= Id
n
hence C = [G[A
1
(A
T
)
1
and taking inverses gives [G[C
1
= A
T
A, this tells
us(?) the columns of T are orthogonal under the usual dot-product.
Now C
i
[C
G
(g
i
)[ = [G[ (where C
G
(g
i
) is the centralizer of g
i
in G), so if
D = diag([C
G
(g
1
)[, . . . , [C
G
(g
n
)[)
then we have D = A
T
A, that is, for 1 i, j n we have
n

k=1
r
i
k
r
j
k
= [C
G
(g
i
)[
i,j
where r
l
k
=
k
(g
l
) as required.
6 G-SETS 19
6 G-Sets
A group G acts on a nite set X if there is a function GX X, (g, x) gx
such that ex = x and (g
1
g
2
)x = g
1
(g
2
)x for all x X, g
1
, g
2
G. We say X
is a G-set. The disjoint union X
1
. X
2
and product X
1
X
2
of two G-sets
are also G-sets.
An equivalence relation on elements of a G-set X dened by x y if and
only if y = gx for some g G gives a partition of X into orbits. If X has a
single orbit it is called transitive (each orbit can be considered as a transitive
G-set by itself).
A map of G-sets : X Y is a function such that (gx) = g(x) for
all x X, g G. An isomorphism of G-sets is a map of G-sets with an
inverse that is also a map of G-sets (we can show a bijective map of G-sets
is necessarily an isomorphism).
If H G is a subgroup, then the cosets G/H (not necessarily a group)
form a G-set under the action g(xH) = (gx)H. Moreover, it is transitive
since yx
1
(xH) = yH for all x, y G.
The following result tells us these are all the transitive G-sets (up to
isomorphsm).
Proposition 6.1. Every transitive G-set X is isomorphic to G/H for some
subgroup H G.
Proof. Choose x X and let H = Stab
G
(x) = g G [ gx = x. Dene
: G/H X by (gH) = gx, we claim is a G-set isomorphism.
Now is well-dened, if g
1
H = g
2
H, then g
1
g
1
2
H = Stab
G
(x) so
g
1
g
1
2
x = x, thus g
1
x = g
2
x and (g
1
H) = (g
2
H).
And is a map of G-sets since
((g
1
g
2
)H) = (g
1
g
2
)x = g
1
(g
2
x) = g
1
(x)
as required.
It is injective
(g
1
H) = (g
2
H) g
1
x = g
2
x g
1
2
g
1
x = x g
1
H = g
2
H
It is surjective, if y X, then as X is transitive there exists g G such that
gx = y, that is (gH) = y, as required.
As is a bijective map of G-sets, it is therefore a G-set isomorphism and
X

= G/H as claimed.
Corollary 6.2. If X is transitive, then [X[[Stab
G
(x)[ = [G[.
For example, G acts on itself by conjugation h ghg
1
and for g G
we have Stab
G
(g) = C
G
(g), the centralizer of g in G, so [C[[C
G
(g)[ = [G[.
6 G-SETS 20
To every G-set X we can associate a G-space CX wth basis x X. We
have C(X . Y )

= CX CY .
Proposition 6.3.
CX
(g) = the number of xed points xed by g in X.
Proof. Let g G, then
CX
(g) is a permutation matrix (the denition of a
G-set requires g permutes the elements of X) with a 1 in the (x, x)-position
if and only if gx = x, otherwise it is 0. Hence tr(
CX
(g) is the number of
xed points.
Proposition 6.4. The G-space CX contains the trivial representation, so
CX

= C V .
Proof. Let s =

xX
x, then gs = s for all g G and Cs is a G-subspace
on which G acts trivially, thus Cs

= C as G-spaces.
Alternatively, we have
(
CX
, 1) =
1
[G[

gG

CX
(g) =
1
[G[
([X[ + ) > 0
where the rst term is [X[ since 1 G xes every element of X and each
additional term represents a number of xed points, so is non-negative.
Corollary 6.5. For any group G, we know
GG
1 is a character of G.
7 INDUCED REPRESENTATION 21
7 Induced Representation
Let H G be a subgroup, a set of left coset representatives is a subset
x
1
, . . . , x
n
G, with exactly one element from each coset of G/H, whence
G =

x
i
H, and every element g G can be written uniquely as g = x
i
h for
some 1 i n and h H.
Now G-space C(G/H) has basis x
1
H, . . . , x
n
H on which g G acts as
g(x
i
H) = (gx
i
)H and permutes the basis elements; we have a permutation

g
S
n
on the basis index.
Hence, we can write g(x
i
) = x

g
(i)
h
g,i
uniquely for some h
g,i
H, for the
action of G on the coset representatives.
We could write C(G/H) as x
1
C x
n
C and g G acts on x
i
c by
g(x
i
c) = x

g
(i)
h
g,i
c, then assuming H G acts trivially on the elements of
C, we have g(x
i
c) = x

g
(i)
c. We write Ind
G
H
C =

n
i=1
x
i
C

= C(G/H) for
this G-space.
Generalizing this procedure, suppose V is a H-space, then dene G-space
Ind
G
H
V the induced representation as having the vector space

n
i=1
x
i
V
(where again x
1
, . . . , x
n
are a set of left coset representatives), on which G
acts by g(x
i
v) = x

g
(i)
h
g,i
v (noting that h
g,i
v V as V is a H-space).
We must prove Ind
G
H
V is indeed a G-space. Each g G acts linearly (this
follows by construction) and 1 g clearly acts as the identity. Let f, g G,
then
(fg)x
i
= x

fg
(i)
h
fg,i
f(gx
i
) = f(x

g
(i)
h
g,i
)
= x

f
(
g
(i))
h
f,
g
(i)
h
g,i
it follows that
fg
(i) =
f
(
g
(i)) and h
fg,i
= h
f,
g
(i)
h
g,i
by uniqueness. Thus,
by similar calculations, (fg)x
i
v = f(gx
i
v) for i = 1, . . . , n, and this extends
linearly, as required.
We now look for the matrix representation of G-space Ind
G
H
V of an induced
representation of H-space V .
Let be the matrix representation of V , then dene (g) =
_
(g) if g H
0 otherwise
,
and P(g) the n n-block matrix with (i, j) block (x
1
i
gx
j
).
Proposition 7.1. P is the matrix representation of Ind
G
H
V (with the obvious
choice of basis).
7 INDUCED REPRESENTATION 22
Proof. It is enough to check on vectors of the form x
i
v for v V , since they
span Ind
G
H
V .
As a column vector x
i
v = [0, . . . , 0, v, 0, . . . , 0]
T
with v in the ith position,
so g(x
i
v) should correspond to P(g)(x
i
v) = [ (x
1
1
gx
i
)v, . . . , (x
1
n
gx
i
)v]
T
,
the ith column of P(g) times v. But we know g(x
i
v) = x

g
(i)
h
g,i
v hence this
action, in vector form, results in [0, . . . , 0, (h
g,i
)v, 0, . . . , 0]
T
with (h
g,i
) in
the
g
(i)th position.
Now (x
1
j
gx
i
)v ,= 0 if and only if x
1
j
gx
i
H, that is, if and only
if x
1
j
gx
i
= h for some h H, i.e. gx
i
= x
j
h but by uniqueness then
x
j
= x

g
(i)
and h = h
g,i
so (x
1
j
gx
i
) = (h
g,i
) as required.
Theorem 7.2. The induced representation Ind
G
H
V has character

Ind
G
H
(g) =
1
[H[

yG

V
(y
1
gy)
where

V
(f) =
_

V
(f) if f H
0 otherwise
which can also be written

Ind
G
H
V
(g) =
n

i=1

V
(x
1
i
gx
i
)
Proof. The second form follows from the formula for P(g), as there is a non-
zero block in position (i, i) if and only if x
1
i
gx
i
H, in which case it is
(x
1
i
gx).
Since G =

x
i
H we have
1
[H[

yG

V
(y
1
gy) =
1
[H[

x
i
hH

V
((x
i
h)
1
g(x
i
h))
=
1
[H[

x
i
hH

V
(x
1
i
gx
i
)
=
n

i=1

V
(x
1
i
gx
i
)
using the fact that
V
is a class function on H.
If H G is a subgroup we can dene induction on class functions by
(Ind
G
H
)(g) =
1
[H[

yG

(y
1
gy)
where was a class function on H, now Ind
G
H
() is a class function on G.
7 INDUCED REPRESENTATION 23
7.1 Frobenius Reciprocity
Theorem 7.3. If is a class function on H G and is a class function
on G, then
(Ind
G
H
, )
G
= (, Res
G
H
)
H
Proof.
(Ind
G
H
, )
G
=
1
[G[

gG
(Ind
G
H
)(g)(g)
=
1
[G[

gG
_
1
[H[

yG

(y
1
gy)
_
(g)
=
1
[G[

yG
_
1
[H[

gG

(y
1
gy)(g)
_
=
1
[G[

yG
_
1
[H[

fG

(f)(yfy
1
)
_
=
1
[G[

yG
_
1
[H[

fG

(f)(f)
_
=
1
[G[

yG
_
1
[H[

hH
(h)(h)
_
=
1
[G[

yG
(, Res
G
H
)
H
= (, Res
G
H
)
H
Aside from denitions we have used the facts that is a class function on
G, so (yfy
1
) = (y) holds, and also

(f) = 0 for f GH.
Example 7.4. Let H G be a subgroup, then
(Ind
G
H
(1
H
), 1
G
)
G
= (1
H
, 1
H
)
H
= 1
where Ind
G
H
(1
H
) is the character of Ind
G
H
C

= C[G/H].
If X is a transitive G-set, then it is isomorphic to G/H for some H G,
therefore (
CX
, 1) = 1. If X is not transitive, then X =

X
i
for orbits
X
1
, . . . , X
n
and
(
C[x]
, 1) = (

C[X
i
]
, 1) =
n

i=1
(
C[X
i
]
, 1) = n
that is (
C[x]
, 1) equals the number of orbits of the G-set X.

You might also like