MATH 742 Character Theory Spring 2023

Character Theory
Alexander Duncan
February 8, 2023

Throughout this document, we make the standing assumptions

that G is a finite group and k is a field of characteristic coprime
to the order of G. This allows us to appeal to Maschke’s theorem: ev-
ery finite-dimensional representation is completely reducible. Thus, every
indecomposable representation is also irreducible.
Character theory is the core of finite group representations. Much of the
material below can be found in, for example, [AB95, §14–15], [DF04, §18.3],
[EGH+ 11, §4], [FH91, §2], [Lan02, §XVIII.1–5], or [Ser77, §2].

1 Representation Ring
Let Rk+ (G) be the set of isomorphism classes of finite-dimensional represen-
tations of G over k. Given representations V and W , let [V ] and [W ] denote
their images in Rk+ (G). We define addition on Rk+ (G) via

[V ] + [W ] := [V ⊕ W ]

and multiplication on Rk+ (G) via

[V ] · [W ] := [V ⊗ W ].

We write 0 for [0] and 1 for [k]. (One should check that these operations are
well-defined.)
For representations U, V, W , let u = [U ], v = [V ], and w = [W ]. Write
0 = [0] for the zero representation and 1 = [k] for the trivial representation.

Last Revised: February 8, 2023 1 of 15

MATH 742 Character Theory Spring 2023

Now, standard isomorphisms give rise to properties of Rk+ (G) as follows:

U ⊕V ∼ =V ⊕U u+v =v+u
U ⊕ (V ⊕ W ) ∼
= (U ⊕ V ) ⊕ W u + (v + w) = (u + v) + w
0⊕V =V∼ 0+v =v
V ⊕0∼ =V v+0=v
U ⊗V ∼ =V ⊗U uv = vu
∼
U ⊗ (V ⊗ W ) = (U ⊗ V ) ⊗ W u(vw) = (uv)w
k⊗V ∼ =V 1v = v
∼
V ⊗k =V v1 = v
0⊗V ∼ =0 0v = 0
V ⊗0∼ =0 v0 = 0
∼
U ⊗ (V ⊕ W ) = (U ⊗ V ) ⊕ (U ⊗ W ) u(v + w) = uv + uw
(V ⊕ W ) ⊗ U ∼
= (V ⊗ U ) ⊕ (V ⊗ U ) (v + w)u = vu + wu
We would like to conclude that Rk+ (G) is a ring, but we are missing addi-
tive inverses. Indeed, no non-zero representation has an additive inverse for
dimensional reasons. However, Rk+ (G) does have the structure of a commuta-
tive rig (“a riNg without Negatives”). In particular, Rk+ (G) is a commutative
monoid under addition, and the non-zero elements of Rk+ (G) form a commu-
tative monoid under multiplication.
Let Sk (G) be the set of isomorphism classes of irreducible representations
+
of G over k. From the Krull-Schmidt theorem, every element [V ] of RG (k)
can be written uniquely as a finite linear combination
[V ] = m1 [W1 ] + · · · + mr [Wr ]
where [W1 ], . . . , [Wr ] ∈ Sk (G) and m1 , . . . , mr are non-negative integers.
Thus there is an additive monoid isomorphism RG +
(k) ∼
= N⊕Sk (G) . (We
will see that Sk (G) is always finite, but we distinguish between the direct
sum and the direct product until we prove this).
+
However, while the additive structure on RG (k) is very nice, the multi-
plicative structure is considerably more subtle. Let X be an indexing set for
Sk (G) = {[Wi ]}i∈X . The multiplicative structure is completely determined
by the non-negative integers cij ℓ in the expressions
X ij
[Wi ] · [Wj ] = cℓ [Wℓ ]
ℓ∈X

Last Revised: February 8, 2023 2 of 15

MATH 742 Character Theory Spring 2023

where i, j vary independently over all values of X.

Remark 1.1. The numbers cij ℓ are called the Clebsch-Gordan numbers for the
group G. After selecting bases for the various Wi , we have explicit matri-
ces defining the isomorphism between Wi ⊗ Wj and the decomposition into
irreducible representations; the entries of this matrix are called the Clebsch-
Gordan coefficients. Analogs for infinite groups like the special orthogonal
group SO(3) are of special interest in quantum physics and quantum chem-
istry.

1.1 Virtual Representations

A virtual representation X is an element of the free abelian group on Sk (G).
More concretely, there is an expression

X = m1 [W1 ] + m2 [W2 ] + · · · + mr [Wr ]

where [W1 ], . . . , [Wr ] are distinct isomorphism classes of irreducible repre-

sentations and m1 , . . . , mr are integers. If all the elements m1 , . . . , mr are
non-negative, then the virtual representation is an (isomorphism class of) a
representation via the evident interpretation in Rk+ (G).

Definition 1.2. The representation ring Rk (G) of G over k is the additive

group of virtual representations with multiplication obtained by extending
the multiplication of Rk+ (G) by linearity. If k = C, then we simply write
R(G).

Example 1.3. If G = 1 is the trivial group, then representations are just

vector spaces. In this case, Rk+ (1) ∼
= N where V 7→ dimk (V ). The ring of
virtual representations is Rk (1) ∼
= Z.

Remark 1.4. The theory of rings is far more powerful than that of rigs.
Thus, we “adjoined negatives” to make our rig Rk+ (G) into a ring Rk (G).
This procedure is an additive analog to forming the ring of fractions of a
ring. In general, one can similarly construct the Grothendieck group or group
completion G from a commutative monoid M . If the monoid M is a rig, then
the group G becomes a ring. However, we only get an embedding M ,→ G if
M is cancellative.

Last Revised: February 8, 2023 3 of 15

MATH 742 Character Theory Spring 2023

2 Characters
In this section, we assume throughout that k = C. Here for z ∈ C, the ex-
pression z denotes the complex conjugate. Observe that if z is a root of unity,
then z = z −1 . This small observation will have important consequences.

Definition 2.1. Let G be a finite group and (V, ρ) a representation of G.

We define the character of ρ as the function

χV : G → C

given by χV (g) = tr(ρg ) for all g ∈ G.

Characters satisfy numerous useful properties.

Proposition 2.2. Let (V, ρ) and (W, σ) be representations of a finite group

G. Then:

1. χV (1) = dim(V ) where 1 is identity of G,

2. χV (g −1 ) = χV (g) for any g ∈ G, and

3. χV (hgh−1 ) = χV (g), for all g, h ∈ G.

4. χV ⊕W (g) = χV (g) + χW (g) for g ∈ G.

5. χV ⊗W (g) = χV (g)χW (g) for g ∈ G.

6. χV ∨ (g) = χV (g) for g ∈ G.

7. χHomk (V,W ) (g) = χV (g)χW (g) for g ∈ G.

Proof. Fix an element g ∈ G. Since ρ(g) and σ(g) are complex matrices
of finite order, they are both diagonalizable with eigenvalues roots of unity.
Let e1 , . . . , en be a diagonal basis for ρ(g) with eigenvalues λ1 , . . . , λn , and
let f1 , . . . , fm be a diagonal basis for σ(g) with eigenvalues µ1 , . . . , µm . In
particular,

χV (g) = λ1 + · · · + λn and χW (g) = µ1 + · · · + µm .

We now prove each part of the statement

(1.) χV (1) is the n × n identity matrix where n = dim(V ).

Last Revised: February 8, 2023 4 of 15

MATH 742 Character Theory Spring 2023

(2.) Recall ζ −1 = ζ for any root of unity ζ. Thus

ρg−1 (ei ) = λ−1
i = λi = ρg (ei )

for each g ∈ G and i ∈ {1, . . . , n}. Since the trace is simply the sum of the
λi , the result follows.
(3.) Recall that tr(AB) = tr(BA) for any matrices A and B. Thus
tr(ABA−1 ) = tr(A−1 AB) = tr(B).
(4.) Follows from the fact that e1 , . . . , en , f1 , . . . , fm is a diagonal basis
for (ρ ⊕ σ)(g) and that
(λ1 + · · · + λn ) + (µ1 + · · · + µm ) = (λ1 + · · · + λn + µ1 + · · · + µm ) .
(5.) Note that {ei ⊗ fj } is a diagonal basis for (ρ ⊗ σ)(g). The result
follows from the observation that
n X
X m
(λ1 + · · · + λn )(µ1 + · · · + µm ) = λi µj .
i=1 j=1

(6.) Denote by e∨1 , . . . , e∨n the dual basis of e1 , . . . , en . Observe that

ρ∨ (g)(e∨i )(ej ) = e∨i (ρ(g −1 )(ej )) = λ−1
i δij = λi δij .

(7.) Follows from the isomorphism Homk (V, W ) ∼

= V ∨ ⊗ W.
Exercise 2.3. Using notation as the previous proposition.
1
χ(g)2 + χ(g 2 ) for g ∈ G.


1. χSym2 (V ) (g) = χS 2 (V ) (g) =
2
1
χ(g)2 − χ(g 2 ) for g ∈ G.


2. χAlt2 (V ) (g) = χΛ2 (V ) (g) =
2
The proposition above proves that all characters are class functions:
Definition 2.4. A function f : G → C is called a class function if f (hgh−1 ) =
f (g) for all g, h ∈ G. The set of all class functions is denoted C(G).
Equivalently, C(G) consists of exactly those functions f : G → C such
that f (g) = f (h) whenever g, h belong to the same conjugacy class in G.
The set of all functions f : G → C is a C-algebra via pointwise addition
and multiplication. Alternatively, the ring is the same as the usual ring
structure on the Cartesian power CG . The subset of class function C(G) are
a subalgebra of this ring. If c(G) denotes the number of conjugacy classes of
G, then C(G) is naturally isomorphic to Cc(G) .

Last Revised: February 8, 2023 5 of 15

MATH 742 Character Theory Spring 2023

Example 2.5. Let V be the regular representation. Then

(
|G| for g = 1,
χV (g) =
0 otherwise.

This follow since heg = eg only when h = 1 in the natural basis {eg }g∈G .

The properties proven above about characters imply the following nice
corollary:

Corollary 2.6. The assignment of a representation to its character induces

a ring homomorphism
χ• : R(G) → C(G)
from the representation ring to the algebra of class functions on G.

We will see shortly that χ• is injective. An element in the image of the

map χ• above is called a virtual character. Note that R(G) is only a ring,
while C(G) is a C-algebra. They are never isomorphic, but we will see that
R(G) ⊗Z C ∼ = C(G) as C-algebras.
Since the set of isomorphism classes of irreducible representations of G
form a Z-basis for R(G) and the set of conjugacy classes of G induce a C-
basis for C(G), the map χ• is totally determined by pairing these two bases.
The resulting matrix is called the character table of G.

Example 2.7. It turns out the symmetric group S4 has exactly 5 irreducible
representations; label the irreducible characters as χ1 , . . . , χ5 . The character
table is as follows:
e (12) (12)(34) (123) (1234)
χ1 1 1 1 1 1
χ2 1 −1 1 1 −1
χ3 2 0 2 −1 0
χ4 3 1 −1 0 −1
χ5 3 −1 −1 0 1

There are some common conventions regarding character tables. Almost

universally, the trivial representation is the first row and the class of the
identity is the first column. Another convention is that the irreducible repre-
sentations are listed in increasing order by dimension. Finally, a less common

Last Revised: February 8, 2023 6 of 15

MATH 742 Character Theory Spring 2023

convention is that the conjugacy classes are listed in increasing order by the
order of any representing element.
Computer algebra packages can produce (or look up) character tables
for finite groups. Warning: the orderings of the rows and the columns of
the table may not be consistent even if exactly the same program is run on
exactly the same data!

3 Orthogonality Relations
We begin by pointing out that class functions give rise to G-equivariant
endomorphisms.

Lemma 3.1. Let (V, ρ) be a complex representation of a finite group G, let

f : G → C be a function, and let F ∈ Endk (V ) be the endomorphism defined
by X
F (v) = f (g)ρg (v).
g∈G

Then F is G-equivariant if f is a class function.

Proof. Exercise. This is similar to the reindexing argument from Maschke’s

Theorem.
Recall that if (V, ρ) is a linear representation of a finite group G, then V G
is the invariant subspace of vectors v ∈ V such that ρg (v) = v for all g ∈ G.

Lemma 3.2. If (V, ρ) be a linear representation of a finite group G, then

1 X
dim(V G ) = χV (g).
|G| g∈G

Proof. By the same argument as in the proof of Maschke’s Theorem, the

linear transformation π : V → V defined by
1 X
π(v) = ρg (v)
|G| g∈G

is a G-equivariant projection onto V G . Choosing an appropriate basis for V ,

the map π corresponds to a matrix with dim(V G ) ones on the diagonal and

Last Revised: February 8, 2023 7 of 15

MATH 742 Character Theory Spring 2023

zeroes elsewhere. The computation

!
1 X 1 X 1 X
χV (g) = tr(ρ(g)) = tr ρ(g) = tr(π) = dim(V G )
|G| g∈G |G| g∈G |G| g∈G

finishes the argument.

Definition 3.3. Define a pairing

1 X
(ϕ, ψ) := ϕ(g)ψ(g)
|G| g∈G

for any two functions ϕ, ψ : G → C.

Proposition 3.4. The product (−, −) makes C(G) an inner product space.

Proof. The pairing (−, −) is conjugate linear in the first variable and linear
in the second. Also, (ψ, ψ) is a sum of non-negative real numbers which is
zero if and only if ψ(g) = 0 for every g ∈ G.
The inner product has an important interpretation:

Lemma 3.5. (χV , χW ) = dimk HomG

k (V, W ) .

Proof. If τ is the character of Homk (V, W ), then

1 X 1 X
(χV , χW ) = χV (g)χW (g) = τ (g).
|G| g∈G |G| g∈G

Thus:
 
(χV , χW ) = dimk (Homk (V, W ))G = dimk (HomG
k (V, W ))

as desired.
A character is irreducible if its corresponding representation is. The main
result of this section is as follows:

Theorem 3.6. The irreducible characters are an orthonormal basis for the
space of class functions C(G).

Last Revised: February 8, 2023 8 of 15

MATH 742 Character Theory Spring 2023

Proof. Suppose V and W are irreducible. Then Schur’s Lemma gives us

(
1 V ∼ =W
(χV , χW ) = dimk HomG
k (V, W ) = .
0 V ≁= W
Thus, the irreducible characters are an orthonormal set. It remains to show
they span C(G).
Let f ∈ C(G) be orthogonal to every irreducible character (thus every
character). For any representation (V, ρ), the endomorphism
X
Fρ = f (g)ρg
g∈G

is equivariant by Lemma 3.1. Taking the trace, we find

X
tr(Fρ ) = f (g)χV (g) = |G|(f, χV ),
g∈G

which is 0 by the orthogonality condition. When V is irreducible, Fρ is scalar

multiplication by Schur’s Lemma; moreover, the trace being zero implies that
Fρ = 0. If W is an irreducible subrepresentation of V , then Fρ (W ) ⊆ W
since Fρ is a linear combination of functions satisfying that property. Thus,
we conclude Fρ = 0 for all representations ρ.
When V is the regular representation with basis {eg }g∈G , we obtain
X
0 = Fρ (e1 ) = f (g)eg .
g∈G

We must conclude f (g) = 0 for all g ∈ G. Thus no class function is orthogonal

to every irreducible character.
This theorem is enormously powerful. In particular, it tells us that the
number of irreducible representations is not only finite, but exactly how many
there are:
Corollary 3.7. The number of irreducible representations is equal to the
number of conjugacy classes in G.
The character table can be viewed as the “change of basis” matrix between
irreducible representations of R(G) and characteristic functions of conjugacy
classes in C(G). While there exists a bijection between conjugacy classes
and irreducible representations, we do not explicitly construct one. Indeed,
there are good reasons to not expect an explicit “combinatorial bijection” in
general.

Last Revised: February 8, 2023 9 of 15

MATH 742 Character Theory Spring 2023

Corollary 3.8. If V is a representation with isotypic decomposition

V ∼
= W1⊕m1 ⊕ · · · ⊕ Wr⊕mr
for irreducible representations W1 , . . . , Wr , then we may recover the multi-
plicities via
mi = (χV , χWi ) .
Proof. Exercise.
Corollary 3.9. A character χ is irreducible if and only if (χ, χ) = 1.
Proof. Exercise.
Corollary 3.10. Suppose G has irreducible representations V1 , . . . , Vr with
dimensions n1 , . . . , nr . Then the isotypic decomposition of the regular repre-
sentation is
V1n1 ⊕ · · · ⊕ Vrnr .
In particular, |G| = n21 + n22 + · · · + n2r .
Proof. Let W be the regular representation. Using Example 2.5, we compute
1 X 1
(χVi , χW ) = χVi (g)χW (g) = χV (1)|G| = dim(Vi ) = ni .
|G| g∈G |G| i

The second statement follows by comparing the dimensions in the regular

representation with its isotypic decomposition.

3.1 Computing inner products

In practice, it is inefficient to compute inner products by summing over every
element in the group. We already know that the class functions will give
the same output for each element of the conjugacy class. Therefore, it is
convenient to rewrite the inner product to take advantage of this.
Proposition 3.11. Let K1 , . . . , Kr be the conjugacy classes of G and let
g1 , . . . , gr be representatives in each class. Then
r
X |Ki |
(ψ, ϕ) = ψ(gi )ϕ(gi )
i=1
|G|

for all class functions ψ, ϕ ∈ C(G).

Last Revised: February 8, 2023 10 of 15

MATH 742 Character Theory Spring 2023

Recall that an orthogonal matrix has an orthogonal transpose. The for-

mula above gives an orthonormal pairing for the rows of the character table
(when appropriately weighted by the number of elements in each conjugacy
class). We also have orthogonality of the columns via the following:

Exercise 3.12. Let χ1 , . . . , χr be the irreducible characters. Let K(g) denote

the number of elements in the conjugacy class of g ∈ G. We have
r
(
|G|
X if g is conjugate to h
χi (g)χi (h) = K(g)
i=1
0 otherwise

for g, h ∈ G.

3.2 Explicit Decomposition

If we know the characters, then we can explicitly compute the isotypic de-
composition of linear representation.

Theorem 3.13. Let W1 , . . . , Wr be the irreducible representations of G with

corresponding characters χ1 , . . . , χr . Let (V, ρ) be a representation and let Vi
be the isotypic component of V associated to Wi . Then the map πi : V → V
given by
χi (1) X
πi (v) := χi (g)ρg (v)
|G| g∈G

is the projection onto the isotypic component Vi .

Proof. From Lemma 3.1, we see that πi is equivariant. Thus πi (U ) ⊆ U for

every subrepresentation U of V . By Schur’s Lemma, πi |U is scalar multi-
plication whenever U is irreducible. If ψ is the character of an irreducible
subrepresentation U of V , then

tr(πi |U ) = χi (1)(χi , ψ).

We conclude that
χi (1)
πi |U = (χi , ψ) .
ψ(1)
Thus πi |U = 0 if Wi ∼
̸= U and πi |U = id if Wi ∼ = U.

Last Revised: February 8, 2023 11 of 15

MATH 742 Character Theory Spring 2023

Note that the theorem above generalizes the projection constructed in

the proof of Lemma 3.2.
There are also formulas that can explicitly decompose an isotypic repre-
sentation into irreducible summands. Such a decomposition obviously cannot
be canonical. See §2.7 of [Ser77].

4 Examples
Example 4.1. Let G = Z/nZ. Let ζ = e2πi/n be the principal primitive nth
root of unity. The conjugacy classes of G are just the elements of G. The
irreducible characters are just the group homomorphisms G → C× . They
are precisely the maps
χi (j) = ζ ij
for i ∈ 1, . . . , n and j ∈ G. The character table is as follows:
0 1 2 ··· n−1
χ0 1 1 1 ··· 1
χ1 1 ζ ζ2 ··· ζ −1
χ2 1 ζ2 ζ4 ··· ζ −2
.. .. .. .. .. ..
. . . . . .
χn−1 1 ζ −1 ζ −2 ··· ζ

This is exactly the discrete Fourier transform.

Example 4.2. More generally, if G is a finite abelian group, then the char-
acter table is |G| × |G| since every conjugacy class has only one element.
Once again, the irreducible characters are group homomorphisms G → C× .
For example, in the case of the Klein 4-group we have:
(0, 0) (1, 0) (0, 1) (1, 1)
χ(0,0) 1 1 1 1
χ(1,0) 1 −1 1 −1
χ(0,1) 1 1 −1 −1
χ(1,1) 1 −1 −1 1

Example 4.3. Let G be the dihedral group of order 2n:

D2n = ⟨s, r | s2 , rn , (sr)2 ⟩

Last Revised: February 8, 2023 12 of 15

MATH 742 Character Theory Spring 2023

where n ≥ 3.
Let ζ be a primitive nth root of unity. Aside from the trivial representa-
tion ρ0 , there is always a 1-dimensional representation ρ1 where ρ1 (s) = (−1)
and ρ1 (r) = (1). For each integer 1 ≤ i ≤ n − 1 we may produce a two-
dimensional representation σi as follows:
   i 
0 1 ζ 0
σi (s) = σi (r) = .
1 0 0 ζ −i

However, notice that σi is isomorphic to σn−i and, in the case where n is

even, we see that σn/2 is not irreducible. Thus, when n is even, we also have
two additional 1-dimensional representations χ2 and χ3 .
Recall that ζ i + ζ −i = 2 cos( 2πi
n
). It’s natural to consider the character
tables in the even and odd cases separately. When n is odd we have:
1 sri r±1 ··· r±(n−1)/2
χ0 1 1 1 ··· 1
χ1 1 −1 1 ··· 1 
(n−1)π
2 cos 2π


σ1 2 0 n
··· 2 cos n
.. .. .. .. .. ..
. . . .
.  . 2 
σ(n−1)/2 2 0 2 cos (n−1)π
n
··· 2 cos (n−1)
n
π

When n is even we have:

1 sr2i sr2i+1 r±1 ··· rn/2
χ0 1 1 1 1 ··· 1
χ1 1 −1 −1 1 ··· 1
χ2 1 −1 1 −1 ··· (−1)n/2
χ3 1 1 −1 −1
··· (−1)n/2
σ1 2 0 0 2 cos 2πn
··· −2
.. .. .. .. .. .. ..
. . . . .
.  .
(n−2)π
σn/2−1 2 0 0 2 cos n
··· −2

We can check that these are all irreducible by checking that their norms are
1 with respect to the inner product. By summing the squares of the values
on 1, we see that we have all of them.
Example 4.4. The character table for the alternating group on 4 letters is:

Last Revised: February 8, 2023 13 of 15

MATH 742 Character Theory Spring 2023

() (1 2)(3 4) (1 2 3) (1 3 2)
χ1 1 1 1 1
χ2 1 1 ζ ζ −1
χ3 1 1 ζ −1 ζ
χ4 3 −1 0 0
where ζ is a primitive cube root of unity.
Example 4.5. The character table for the alternating group on 5 letters is:
() (1 2)(3 4) (1 2 3) (1 2 3 4 5) (1 3 5 2 4)
χ1 1 1 1 1 1
χ2 3 −1 0 φ −1/φ
χ3 3 −1 0 −1/φ φ
χ4 4 0 1 −1 −1
χ5 5 1 −1 0 0
√
where φ is the golden ratio 12 (1 + 5).

References
[AB95] J. L. Alperin and Rowen B. Bell. Groups and representations,
volume 162 of Graduate Texts in Mathematics. Springer-Verlag,
New York, 1995.
[DF04] David S. Dummit and Richard M. Foote. Abstract algebra. John
Wiley & Sons, Inc., Hoboken, NJ, third edition, 2004.
[EGH+ 11] Pavel Etingof, Oleg Golberg, Sebastian Hensel, Tiankai Liu, Alex
Schwendner, Dmitry Vaintrob, and Elena Yudovina. Introduc-
tion to representation theory, volume 59 of Student Mathematical
Library. American Mathematical Society, Providence, RI, 2011.
With historical interludes by Slava Gerovitch.
[FH91] William Fulton and Joe Harris. Representation theory, volume 129
of Graduate Texts in Mathematics. Springer-Verlag, New York,
1991. A first course, Readings in Mathematics.
[Lan02] Serge Lang. Algebra, volume 211 of Graduate Texts in Mathemat-
ics. Springer-Verlag, New York, third edition, 2002.

Last Revised: February 8, 2023 14 of 15

MATH 742 Character Theory Spring 2023

[Ser77] Jean-Pierre Serre. Linear representations of finite groups. Grad-

uate Texts in Mathematics, Vol. 42. Springer-Verlag, New York-
Heidelberg, 1977. Translated from the second French edition by
Leonard L. Scott.

Last Revised: February 8, 2023 15 of 15