Chapter 3

Representations

1
CHAPTER 3. REPRESENTATIONS 2

3.1 Basic definitions

The main ingredients that are necessary to develop representation theory are introduced in this section.
Some basic theorems about representations of groups are presented in the following section. Most of
the results will be quoted without a proof. The interested reader can refer to [Tun85].

3.1.1 Representations
Definition 3.1.1 A representation D of a group G is a homomorhpism from G into the set of linear
operators acting on a vector space V :

D: G −→ GL(V )
g −→ ĝ ,

where g ∈ G is an abstract element of the group G, while ĝ is an operator acting linearly on the
elements (vectors) of V .

Since D is a homorphism, the linear operators verify:

D(g)D(g ′ ) = D(gg ′ ) ; (3.1)

as usual, note that the product on the LHS is a product between linear operators acting on the vectors
in V , while the product inside the parenthesis on the RHS is the product of group elements in G.
The representation D(g) corresponds to an explicit realization of the group structure of G. If V is a
d–dimensional vector space, the representation is called d–dimensional. Having chosen a basis in the
vector space V , ĝ can be written as d × d matrix.

Faithful representation The representation is called faithful if the mapping G → GL(V ) is an

isomorphism.

Matrix representation Let {|ei i}i=1,...,d be a basis of V , complete and orthonormal, hei |ej i = δij .
Acting with ĝ on one of the vectors |ei i yields some other vector in V . Since the basis is complete, we
can write this new vector as a linear combination of the basis vectors. If we call D(g)ji its coordinates
in the {|ei i} basis, we obtain:
d
X
ĝ|ei i = |ej iD(g)ji . (3.2)
j=1

The set of all the D(g)ji as i varies from 1 to d defines a matrix representing the element g.

Exercise 3.1.1 Prove that the mapping D(g)ji defined above is a homomorphism from G into
the group of d × d matrices.
CHAPTER 3. REPRESENTATIONS 3

Unfaithful representation We can write down for every group G at least one trivial representa-
tion, known as the identity representation:

∀g ∈ G, D(g) = 1 . (3.3)

You can readily check that the equation above defines a homomorphism, which is obviously not a
one–to–one correspondence.

Equivalent representations
Definition 3.1.2 Two matrix representations {D(g)} , {D′ (g)} of the same dimension are called
equivalent if and only if there exist a matrix S, d × d, non–singular, such that:

∀g ∈ G, D′ (g) = SD(g)S −1 . (3.4)

Note that the same matrix S has to connect the matrices D(g) and D′ (g) for all the elements of
G. The two representations correspond to the same realization of the group structure seen in two
different bases. The matrix S is the matrix that implements the transformation from one basis to the
other.

Exercise 3.1.2 Let D, D′ be two equivalent representations, and let S be the transformation
that relates them. Prove that D′ (ab) = SD(ab)S −1 .

Unitary representations
Definition 3.1.3 A unitary representation of a group G is a representation such that the matrix
D(g) is unitary for all g ∈ G, i.e.
∀g ∈ G, D(g)D(g)† = 1 . (3.5)

Unitary representations are important in quantum mechanics since they preserve transition am-
plitudes.
Theorem 3.1.1 Any matrix representation of a finite group is equivalent to a unitary representation.

3.1.2 Reducible representations

We have seen that symmetry transformations naturally embody a group structure. It is important
to identify subsets of the vector space of physical states such that the states in these subsets are
transformed into each other by symmetry transformations. These structures are called invariant
subspaces.
Definition 3.1.4 Let us consider a representation of a group G, acting on a vector space V . If
Ls ⊂ V is a subspace of V , such that:

∀g ∈ G, ∀|φi ∈ Ls , ĝ|φi ∈ Ls , (3.6)

then Ls is called an invariant subspace, and the V is reducible under G.

CHAPTER 3. REPRESENTATIONS 4

The subspace Ls is also called closed under G.

The existence of invariant subspaces defines the reducible representations. Let {D(g), g ∈ G} be
a representation of G in V , and let dimV = d. If V has an invariant subspace Ls , dimLs = s < d,
then there exists a basis in V such that:
 (1) 
D (g) 0
∀g ∈ G, D(g) = , (3.7)
0 D(2) (g)

where D(1) is a s × s matrix, and D(2) is a (d − s) × (d − s) matrix.

Definition 3.1.5 In the case described above, the representation D is called a reducible representa-
tion. Conversely, if there is no invariant subspace in V , the representation is called irreducible.

Note that D(1) and D(2) are also representations of G, of dimension d and d−s respectively. Moreover,
the invariant subspace Ls may contain further subspaces that are left invariant by the matrices D(1) (g).
In this case the representation D(1) can be further reduced, i.e. each matrix D(1) (g) can be written into
two diagonal sub-blocks. When all the invariant subspaces have been mapped out, the representation
D is said to have been fully decomposed, each matrix D(g) can be written in a block diagonal form,
which we shall write as:  (1) 
Γ (g) 0 ...
∀g ∈ G, D(g) =  0
 Γ(2) (g) . . . , (3.8)
.. .. ..
. . .

where Γ(i) represent irreducible representation of G. Note that the same Γ(i) can appear more than
once in the decomposition of a given reducible representation D. In general we denote the decompo-
sition of D by:
D = m1 Γ(1) ⊕ m2 Γ(2) ⊕ . . . ⊕ mn Γ(n) , (3.9)
where mi are positive integers.
A representation D is called simply reducible if the coefficients are all mi = 0, 1.
CHAPTER 3. REPRESENTATIONS 5

3.2 Schur’s lemma

Schur’s lemma plays a fundamental role in the classification of group representations. In the first part
this section, we state the theorem and prove it. The rest of the section is devoted to the discussion
of some of the major consequences of Schur’s lemma.
Theorem 3.2.1 Let G be a group, Γ an irrep (irreducible representation) on V , and let V be a
complex vector space. If there exist a matrix T ∈ GL(V ) such that

T Γ(g) = Γ(g)T, ∀g ∈ G , (3.10)

then necessarily
T = λ1 . (3.11)
(Schur’s lemma)

A matrix that commutes with all the elements of an irrep must be a multiple of the identity matrix.
Given the importance of Schur’s lemma, we shall present here its proof for the interested reader.

Proof of Schur’s lemma First we need to prove that, under the assumptions above, the kernel of
T,
ker T = {|v0 i ∈ V : T |v0 i = |0i} (3.12)
is closed under G. Let us consider |v0 i ∈ ker T , and g ∈ G, then:

T Γ(g)|v0 i = Γ(g)T |v0 i = |0i , (3.13)

and therefore Γ(g)|v0 i ∈ ker T .

Now we can use the fact that Γ is an irrep, and therefore there can not be non–trivial invariant
subspaces. The only two options are either ker T = V , or ker T = {|0i}. In the first case we have
proved Schur’s lemma, since T = λ1, with λ = 0.
If ker T = {|0i}, then λ = 0 is not an eigenvalue of T . Let λ 6= 0 be an eigenvalue of T , with
eigenvector |v1 i, and let us define the operator Tλ = T − λ1. From the definition of Tλ you can easily
check that:
Tλ Γ(g) = Γ(g)Tλ , ∀g ∈ G , (3.14)
but now we know that ker Tλ 6= {|0i}, because by definition |v1 i ∈ ker Tλ . Since Γ is an irrep, the
only other possibility is:
ker Tλ = V ⇐⇒ T = λ1 . (3.15)

Exercise 3.2.1 Prove the following corollary of Schur’s lemma. All the irreps of an Abelian
group G are 1–dimensional.

Using Schur’s lemma, we can prove the following

CHAPTER 3. REPRESENTATIONS 6

Theorem 3.2.2 Let Γ(i) , Γ(j) be inequivalent irreps of dimension di , dj respectively, if there exists a
(di × dj ) matrix M such that:
Γ(i) (g)M = M Γ(j) (g), ∀g ∈ G , (3.16)
then M = 0.
(Schur’s second lemma)
We shall not prove explicitly this theorem. However it plays a central role in establishing the
so-called orthogonality theorem.
Theorem 3.2.3 Let Γ(i) , Γ(j) be irreps of a finite group G, |G| = g, and let us assume that the two
representations are inequivalent for i 6= j, and identical for i = j, then
X g
Γ(i) (a)pq Γ(j) (a)∗rs = δij δpr δqs , (3.17)
di
a∈G

where di = dim Γ(i) .

3.3 Characters of a representation

Let us now introduce some more mathematical tools that are useful to study the representations of a
group.
Definition 3.3.1 The characters of a matrix representation {D(g)} of a group G are defined as:
X
χ(g) = TrD(g) = D(g)kk , ∀g ∈ G . (3.18)
k

The set of all characters is called the character system. For a finite group, |G| = n, χ(g) is a vector
with n components.

Properties Characters satisfy the following properties:

1. for a 1-dim representation, the character system is the representation itself:
D(g) = λ(g) ∈ C, TrD(g) = λ(g) . (3.19)

2. characters of conjugate elements are the same;

if ∃g ∈ G, b = gag −1 , then
χ(a) = Tr[D(g)D(b)D(g −1 )] = χ(b) ;
i.e. all elements of a class have the same character. The character χ is called a class function,
and will sometimes be denotede as χk , where the subscript k identifies the class Ck .
3. if two representations D(1) and D(2) are equivalent, they have the same character system:
D(1) (g) = SD(2) (g)S −1 , ∀g ∈ G,
⇒ TrD(1) (g) = Tr[SD(2) (g)S −1 ] = TrD(2) (g) ;
even though we are not going to prove it explicitly, the implication in the opposite direction
is also true. Two representations are equivalent if and only if they have the same character
system.
CHAPTER 3. REPRESENTATIONS 7

First orthogonality relation Starting from the orthogonality theorem, setting p = q, r = s, and
summing over p and r yields:
XX X ∗ g
Γ(i) (a)pp Γ(j) (a)rr = δij di , (3.20)
p r
di
a∈G

i.e. X
χ(i) (a)χ(j) (a)∗ = gδij , (3.21)
a∈G
where g is the order of the group G. Using the fact that the character is a class function, and collecting
the elements a ∈ G in classes Ck , with ck elements in each class Ck , we can rewrite the above relation
as:
Nc  ∗
(i) (j)
X
c k χk χk = gδij , (3.22)
k=1
where Nc indicates the number of classes.

Exercise 3.3.1 Prove the following, useful result. For the trivial representation:
X
χ(0) (a) = g ; (3.23)
a∈G

while for any other irreducible representation:

X
χ(i) (a) = 0 . (3.24)
a∈G

3.4 Reduction of reducible representations

We have seen that a reducble representation can be decomposed as:
M
D(a) = mi Γ(i) (a), a ∈ G , (3.25)
i

where Γ(i) are irreps and mi are integers, that indicate how many times the representation Γ(i) appears
in the decomposition of D.
The matrix D is a block diagonal matrix, with the matrices Γ(i) appearing in the diagonal blocks.
Taking the trace of D, we find: X
χ(a) = mi χ(i) (a) . (3.26)
i
We can use the first orthogonality relation to determine the coefficients mi ; if we multiply Eq. (3.26)
by χ(j) (a)∗ , and sum over a ∈ G, we obtain:
X XX
χ(a)χ(j) (a)∗ = χ(i) (a)χ(j) (a)∗ = gmi (3.27)
a∈G i a∈G
CHAPTER 3. REPRESENTATIONS 8

i.e.
1X
mj = χ(a)χ(j) (a)∗ , (3.28)
g
a∈G

where g is the order of the finite group G.

Reducibility criterion The previous result can be used to establish a useful criterion to identify
an irreducible representation. Let D be a generic representation of a group G, with character system
χ(a), and of order g, then:
1X X
χ(a)χ(a)∗ = m2i > 0 . (3.29)
g i
a∈G

The only way this sum can be equal to one, is if all the mi are equal to zero, except for one whcih we
denote mk . Then the representation D corresponds to Γ(k) and is irreducible. Therefore:
X
D is irreducible ⇐⇒ χ(a)χ(a)∗ = g . (3.30)
a∈G

The following relation is useful to identify the irreps of finite group.

Theorem 3.4.1 Let G be a finite group of order g, and let Ni be the number of irreps of G, then:
Ni
X
d2i = g , (3.31)
i=1

i.e. the sum of the squares of the dimensions of all the irreps of G is equal to the order of the group.

Second orthogonality relation The second orthogonality relation is a completeness relation for
the characters:
Ni
X (i) (j)∗ g
χk χl = δkl , (3.32)
i=1
c k

where k, l are indices that identify classes of G, while Γ(i) are the irreps as usual.
Finally we state one more theorem, which you can use without worrying about its proof.
Theorem 3.4.2 For a finite group G the number of inequivalent irreps is equal to the number of
classes.

Character tables The character system of a group is conveniently summarized in the character
table.
(i)
Definition 3.4.1 A character table is a square matrix of characters χk , where the rows of the matrix
are labelled by the distinct irreps of G, i = 1, . . . , Ni , and the columns are labelled by the distinct
classes of G, k = 1, . . . , Nc .

We shall always label C1 the class of the identity, [e] = {e}.

CHAPTER 3. REPRESENTATIONS 9

(i)
Exercise 3.4.1 Prove that χ1 = di , where di is the dimension of the irreducible representation
Γ(i) . In solving the exercise, check that you remember the definition of a class, of an irreducible
representation, and of the dimension of a representation. Remember that the character is a class
function. Deduce that the first column of the character tables is made of the dimensions of the
irreps.

Exercise 3.4.2 Consider the trivial representation (1-dimensional), Γ(1) (g) = 1, ∀g ∈ G, prove
(1)
that χk = 1, ∀k = 1, . . . , Nc . Deduce that the first row of thecharacter table is made of 1’s.

Character table for D3 Let us consider the symmetry group of the equilateral triangle, D3 , and
let us construct the character table step by step.
D3 is a finite group of order g = 6, and we will denote the 6 elements G = {e, a, b, c, d, f }, where
e is the identity, a, b are rotations by 2π/3, and 4π/3, and c, d, f are reflexions.

Exercise 3.4.3 Show by explicit calculations that there are three classes in D3 . They are:
C1 = [e] = {e}, C2 = [a] = {a, b}, C3 = [c] = {c, d, f }.

Since the number of classes equals the number of irreps, we know that there are three irreps of
D3 , which we denote by Γ(1) , Γ(2) , Γ(3) .
The relation between the dimension of the irreps and the order of the group yields:
Ni
X
d2i = 6 ; (3.33)
i=1

remember that the coefficients di are integers – they are the dimension of the irreps – and therefore
Eq. (3.33) has a unique solution: d1 = 1, d2 = 1, d3 = 2.
We shall identify Γ(1) with the trivial representation, which associates 1 to every element of D3 .
A second inequivalent (and one–dimensional) irrep of D3 of dimension 1 is provided by

Γ(2) (e) = 1, Γ(2) (a) = 1, Γ(2) (b) = 1, Γ(2) (c) = −1, Γ(2) (d) = −1, Γ(2) (f ) = −1 . (3.34)

Exercise 3.4.4 Check that the above is a genuine representation of D3 .

CHAPTER 3. REPRESENTATIONS 10

Finally the 2-dimensional irreducible representation Γ(3) is given by the matrices that implement
the geometrical transformations in two–dimensional Euclidean space:
   √ 
1 0 −1/2 − 3/2
Γ(3) (e) = Γ(3) (a) = √ (3.35)
0 1 3/2 −1/2
   
x x −1 0
Γ(3) (b) = Γ(3) (c) = (3.36)
x x 0 1
   
x x x x
Γ(3) (d) = Γ(3) (f ) = (3.37)
x x x x

We now have all the ingredients for the following exercise.

Exercise 3.4.5 Write the character table for D3 . First write the first column, then complete the
first two lines. Finally complete the remaining entries by using the orthogonality relation. You
can check the entries in the last line against the traces of the matrices Γ(3) .

3.5 Direct product of representations

Let G be a group, {D(p) }, {D(q) } two representations of dimensions p, q. For a generic a ∈ G, D(p) (a)
is a p × p matrix, and D(q) (a) is a q × q matrix. We can define a (pq) × (pq) matrix as follows:
 (p)
D (a)11 D(q) (a) D(p) (a)12 D(q) (a) . . .


(p) (q) .. 
D(a) =  D (a)21 D (a) .  (3.38)
..

.

where each block represents a q × q matrix.

The set of matrices constructed in this way define the direct product of the representations D(p)
and D(q) :
D = D(p) ⊗ D(q) . (3.39)
Clearly the representation D has dimension pq.
In dealing with direct products of representations, it is useful to introduce collective indices,
denoted by capital latin letters A, B, . . ., that represent the pair of indices (i, r), where i = 1, . . . , p,
r = 1, . . . , q. The action of D(a) on a vector can be written as:
X X
D(a)ABv
B
= D(a)ir
js v
js
(3.40)
B j,s

= D(p) (a)ij D(q) (a)rs v js . (3.41)

CHAPTER 3. REPRESENTATIONS 11

Clebsch–Gordan series The product of two irreps yields a larger representation, which in general
will not be irreducible. Such larger representation can be decomposed in a series of irreps according
to: M ij
Γ(i) ⊗ Γ(j) = ck Γ(k) , (3.42)
k

where the cij

kare non-negative integers that count how many times the irrep Γ(k) appears in the direct
product. They can be computed as discussed inthe previous section using characters.
Note that the character of the direct product representation is given by:
X
χ(D(a)) = Tr(Γ(i) (a) ⊗ Γ(j) (a)) = D(a)A A (3.43)
A
X
= Γ(i) (a)kk Γ(j) (a)rr (3.44)
k,r

= χ(i) (a)χ(j) (a) . (3.45)

From Eq. (3.42) we obtain: X

χ(i) (a)χ(j) (a) = cij
kχ
(k)
(a) . (3.46)
k

Finally using the orthogonality relation:

1 X (k) ∗ (i)
cij
k = χ (a) χ (a)χ(j) (a) , (3.47)
g
a∈G

or, equivalently:
N
c
1X (k) (i) (j)
cij
k = cl (χl )∗ χl χl , (3.48)
g
l=1

where l runs over the classes of the group G, and cl is the number of elements in the l-th class. The
coefficients cij
k are called Clebsch–Gordan coefficients.

Exercise 3.5.1 Write the character table for D3 . Deduce the character for the product repre-
sentation Γ(3) ⊗ Γ(3) , where Γ(3) is the two–dimensional irrep of D3 .

3.6 Restriction to a subgroup

Let H ⊂ G be a subgroup, and Γ(i) an irrep of G, then
n o
Γ(i) (a), a ∈ H (3.49)

defines a representation of H. However this representation is not necessarily irreducible; since there
are less elements in H than there are in G, there could be subspaces of V that are not invariant under
the full group G, but are invariant under the smaller set of transformations H.
CHAPTER 3. REPRESENTATIONS 12

When this happens, the induced representation can be reduced:

Γ(i) = ⊕j mij γ (j) , (3.50)
where γ (j) are irreps of H.

Exercise 3.6.1 Let us consider Z3 ⊂ D3 . Write down the character table for the irreps of Z3 .
Write down the restriction of the irreps of D3 to Z3 (in doing so note that elements that are
conjugate to each other in D3 , and therefore belong to the same class in the character table
of D3 , are no longer conjugate in Z3 . They define distinct classes, which appear as additional
columns in the character table.
Decompose the induced representation of Z3 in terms of irreps of Z3 .

3.7 Representations of the product group

Consider the product group G = H ⊗K, whose elements can be written as g = hk, where the elements
of H and K are assumed to commute with each other.
The irreps of the product group are fully classified by the following theorem.
Theorem 3.7.1 If ΓH , ΓK are representations of H and K respectively, then ΓH ⊗ ΓK is a represen-
tation of G.
(i) (j) (i) (j)
Moreover if ΓH , ΓK are irreps, then ΓH ⊗ ΓK is also an irrep of G. Such irreps exhaust all the
irreps of G.

(i) (j)
Exercise 3.7.1 Using the reducibility criterion, prove that the representations ΓH ⊗ ΓK are
indeed irreps.
By computing the sum of their dimensions squared, prove that they do exhaust all possible irreps
of G.

3.8 Representations in Quantum Mechanics

We shall now discuss how representations appear in the study of physical systems. Let us consider a
quantum system, whose states are described by vectors (kets) in a complex vector space with a scalar
product (Hilbert space).
Any state |ψi can be expanded on an orthonormal complete basis {|ψn i}:
hψn |ψm i = δnm , (3.51)
X
|ψn ihψn | = 1, (3.52)
n
X X
|ψi = |ψn ihψn |ψi = cn |ψn i , (3.53)
n n
CHAPTER 3. REPRESENTATIONS 13

where the coefficients cn represent the probability amplitude for |ψi to be in a state |ψn i.
Physical observables are described by hermitean operators:

O† = O (real eigenvalues) . (3.54)

The energy eigenstates are defined as:

H|un i = En |un i , (3.55)

and can be expanded in the {|ψn i} basis:

X
|um i = |ψn ihψn |um i . (3.56)
n

Let us now consider a group G, acting on the states of the Hilbert space:

a ∈ G, |ψi 7→ â|ψi , (3.57)

X
â|ψn i = |ψm iΓ(a)m
n . (3.58)
m

Let us see how H transforms under G. Remember that H is the generator of time translations:

i∂t |ψi = H|ψi , (3.59)

and that under the action of G:

|ψi 7→ |ψ ′ i = â|ψi . (3.60)
By looking at the time evolution of the transformed state:

i∂t |ψ ′ i = i∂t â|ψi = âH|ψi (3.61)

= (âH↠)â|ψi = (âH↠)|ψ ′ i , (3.62)

we deduce the transformed Hamiltonian:

H ′ = âH↠. (3.63)

If the group describes a symmetry of the system, then H ′ = H, i.e.

[H, â] = 0, ∀a ∈ G , (3.64)

and therefore:
Hâ|uα i = âH|uα i = Eα (â|uα i) , (3.65)
i.e. the transformation â connects states with the same energy.
The set of eigenvectors {uαn } associated to each eigenvalue of the energy Eα defines a basis of an
vector subspace which is invariant under G.
The existence of invariant subspaces means that the representation Γ is reducible:

Γ = ⊕α mα Γ(α) . (3.66)

Each eigenvalue of H, Eα defines an irrep of G. The degeneracy of the eigenvalue Eα is dα , the

dimension of the irrep.
CHAPTER 3. REPRESENTATIONS 14

Distinguishing degenerate states Consider now a second operator O, which commutes with H,
i.e. H and O can be diagonalized simultaneously. There exist a set of common eigenvectors:

O|uα α α
i i = λi |ui i . (3.67)

If O is invariant under G, then by Schur’s lemma, O must be proportional to the identity in each
subspace spanned by {uα n } and associated to a given eigenvalue of the Hamiltonian. The matrix
associated to O in the basis of the energy eigenstates is block–diagonal, with blocks of size dα :
 
λα 1α
O=
 λβ 1β .

(3.68)
..
.

On the other hand, if O is invariant under a subgroup H ∈ G, then Schur’s lemma does not apply,
and the eigenvalues λα
i of O can be used to distiguish degenerate energy eigenstates. The blocks in
the matrix representation of O are no longer proportional to the identity:
 α 
λ1

 λα
2 
  β  
 λ1

O= β . (3.69)
 

 λ2


β

 λ3 

..
.

We will see an application of this property when we study angular momentum.

Symmetry breaking Another interesting case is when there is a perturbation in the system which
breaks the symmetry of the unperturbed Hamiltonian:

H = H0 + ǫV (3.70)

where H0 is invariant under G, but V is only invariant under a subgroup H ∈ G, and ǫ is a small
parameter.
In this case the perturbation V splits the degenerate energy levels of H0 , according to the decom-
position of the irreps Γ(α) of G into irreps γ (β) of H:

Γ(α) = ⊕β mβ γ (β) ; (3.71)

where Γ(α) and γ (β) are irreps of G and H repsectively.

Exercise 3.8.1 To illustrate this property, let us consider the crystalline splitting of atomic
energy levels, i.e. the splitting of atomic levels that appears when atoms are arranged in a crystal.
Free atoms are symmetric under rotations, i.e. under the group SO(3) of orthogonal trans-
formations with unit determinant acting on a three–dimensional real vector space. According to
CHAPTER 3. REPRESENTATIONS 15

the discussion in the previous sections, the enrgy levels of the free atoms appear in multiplets
with degeneracies that correspond to the dimensions of the irreps of SO(3). We shal study these
representations in more detail later. For the time being we can assume that the irreps are clas-
sified by one integer quantum number, l ∈ N, related to the angular momentum of the state.
By convention the irreps with l = 0, 1, 2 are called the s, p, d representations respectively. The
characters in the l representation are given by:

sin(l + 1/2)θ
χ(l) (θ) = , (3.72)
sin θ/2

where θ is the angle of the rotation. Note that the character does not depend on the axis of
rotation. Rotations by the same angle around different axes are conjugate elements of SO(3), and
therefore have the same character.
When θ = 0, the rotation reduces to the identity, and its character yields the dimension of
the representation:
lim χ(l) (θ) = dl = 2l + 1 . (3.73)
θ→0

According to the previous discussion, the energy levels arrange in multiplets according to
the dimension of the irreps of SO(3). The s–wave corresponds to a 1–fold degenerate state, the
p–wave to a 3–fold degenerate level, and finally the d–wave yields a 5–fold degenerate level.
When the atoms are arranged in a crystalline structure the symmetry of the system is reduced
to the symmetry of the crystal. We will consider as an example a crystal with tetrahedral
symmetry, described by the group T . Using the character table constructed in Problem 3.9.4,
show that:
1. the 1–fold degeneracy of the s–wave is not lifted;
2. the 3–fold degeneracy of the p–wave is not lifted;
3. the 5–fold degeneracy of the p–wave is lifted , and we obtain three different eigenvalues,
which are 1–fold, 1–fold, and 3–fold degenerate respectively.

Selection rules Transformation properties under a symmetry group also determine transition am-
plitudes. Consider a physical system, whose states are classified by the representations of a symmetry
group G. We shall denote these states as ψiα , where α labels the irrep, and i labels the vectors that
form a basis of the invariant subspace on which the representation α is acting.
The transition amplitude due to some interaction V is given by:

|hφ|V |ψi|2
P (ψ → φ) = . (3.74)
hφ|φihψ|ψi
CHAPTER 3. REPRESENTATIONS 16

The symmetry of the system implies: [V, â] = 0, ∀a ∈ G; therefore we have:

1X α †
hψiα |V |φβj i = hψi |â aV |φβj i
g
a∈G
1X α †
= hψi |â V a|φβj i
g
a∈G
1 X X  (α) k ∗  (β) l  α
= Γ (a)i Γ (a)j hψk |V |φβl i
g
a∈G k,l
1 X
= δαβ δij δkl hψkα |V |φα
l i
dα
k,l
1 X
= δαβ δij hψlα |V |φα
l i
dα
l
= δαβ δij hψ α ||V ||ψ α i . (3.75)
α α
hψ ||V ||ψ i is called a reduce matrix element, and only depends on the energy level α, and not on
the particular state we consider belonging to that multiplet.
The matrix V can be written as:
 α 
v



 .. 
dα × dα matrix
.

  

α
 
v
  
  β  
V =
 v 
 .

(3.76)
  . . 
d × d matrix



 . 

β β 

β 

 v 

..
.
This is called a selection rule; in particular we have:
hψiα |V |ψjβ i = 0, if α 6= β, i 6= j . (3.77)

Tensor operators
Definition 3.8.1 Tensor operators are multiplets of operators, which transform amongst themselves
according to some representations of a group G:
(α) (α)
X (α)
Ti 7→ âTi ↠= Tj Γ(α) (a)ji . (3.78)
j

Note that an invariant operator is a special case of a tensor operator, which transforms under the
trivial representation Γ(1) .
Let us now consider tensor operators T (α) transforming as an irrep Γ(α) , and focus on the trans-
formation properties of the state T (α) |ψjβ i:
(α) (α) †
âTi |ψjβ i = âTi â â|ψjβ i
(α)
X
= Γ(alpha) (a)ki Γ(β) (a)lj Tk |ψlβ i (3.79)
k,l
CHAPTER 3. REPRESENTATIONS 17

(α)
i.e. the state Ti |ψjβ i transforms as the product representation Γ(α) ⊗ Γ(β) .
The product representation can be decomposed in a Clebsch–Gordan series:
Γ(α) ⊗ Γ(β) = ⊕k cαβ
γ Γ
(γ)
, (3.80)
where Γ(γ) are irreps, and therefore:
(α)
hψkδ |Ti |ψjβ i = 0 , (3.81)
unless the irrep Γ(δ) appears in the CG series of Γ(α) ⊗ Γ(β) , i.e. cαβ
δ 6= 0.

Exercise 3.8.2 Consider the transitions between the enrgy levels of an atom in a crystal with
D3 symmetry. The transitions are induced by the operator T = eR · E. Here R is an operator
acting on the physical states, while E is an external electric field, and e is the electriccharge of
the electron. The generic matrix element for the dipole transition can be written:

hψ|T |φi = eE · hψ|R|φi . (3.82)

In order to use the above result, we need to find out the transformation properties of the states:
R|ψ (α) i.
The operators R are a set of tensor operators, which transform as a vector under SO(3), i.e. they
transform according to the irrep with l = 1. Using the formula above we can write the character
for a rotation by an angle θ as:
χT (θ) = 1 + 2 cos θ . (3.83)
We can use the formula above to deduce the character of the transformation in D3 .
On the other hand, the eigenstates of H for a system with a D3 symmetry are organized in
multiplets that correspond to the irreps of the symmetry group. Remember that there are three
irreps for D3 , with dimensions 1,1, and 2. Therefore we must have two non–degenerate states,
corresponding to α = 1, 2, and a two–fold degenerate state for α = 3.
We obtain the following character table:
D3 e 2 C3 3 C2
Γ(1) 1 1 1
Γ(2) 1 1 -1
Γ(3) 2 -1 0
ΓT 3 0 -1
ΓT ⊗ Γ(1) 3 0 -1
ΓT ⊗ Γ(2) 3 0 1
ΓT ⊗ Γ(3) 6 0 0
From the character table we can deduce the CG series:

ΓT ⊗ Γ(1) = Γ(3) ⊕ Γ(2) (3.84)

(2) (3) (1)
ΓT ⊗ Γ = Γ ⊕Γ (3.85)
ΓT ⊗ Γ(3) = Γ(1) ⊕ Γ(2) ⊕ 2Γ(3) (3.86)

Therefore the allowed transitions can be summarized in the following table:

CHAPTER 3. REPRESENTATIONS 18

T |ψ (1) i |ψ (2) i |ψ (3) i

|ψ (1) i ✘ ✔ ✔
|ψ (2) i ✔ ✘ ✔
|ψ (3) i ✔ ✔ ✔
CHAPTER 3. REPRESENTATIONS 19

3.9 Problems
3.9.1 General properties
(a) Show that, in any irreducible representation of a finite group, all the elements of the centre of the
group are represented by multiples of the unit matrix.
(b) If Γ is a representation of a finite group, G, show that either Γ and Γ∗ are both reducible or they
are both irreducible.

3.9.2 Three–dimensional representation of D3

The elements of D3 can be seen as geometrical transformations in a three–dimensional Euclidean
space, where the third axis is chosen to come out of the plane of the equilateral triangle. Write down
the 3 × 3 matrices that correspond to the elements of D3 in this representation. Show that such a
representation is reducible, and identify explicitly the invariant subspaces.

3.9.3 Character tables

Obtain the character tables and the irreducible representations of

1. the cyclic group of order 3, Z3 ;

2. the cyclic group of order 4, Z4 ;

3. the non-cyclic group of order 4, V4 ;

4. the cyclic group of order n, Zn .

3.9.4 The tetrahedral group T

The tetrahedral group, T , consists of the set of proper symmetry operations on a regular tetrahedron
and is of order 12. Its elements fall into four classes: {e}, {3C2 }, {4C3+ }, {4C3− }, where Cp denotes
rotations of magnitude 2π/p and opposite senses of rotation are indicated by ±.
Find the dimensions of the irreps of T , and complete the character table:

T {e} {3C2 } {4C3+ } {4C3− }

Γ(1) 1 1 1 1
Γ(2) x 1 ω ω∗
Γ(3) x 1 ω∗ ω
Γ(4) x x x x

where ω = exp[i2π/3].
Using the character table for the tetrahedral group T derived above,
(a) reduce all the direct products Γ(i) ×Γ(j) , i, j = 1, . . . , 4 to direct sums of irreducible representations.
(b) restrict each Γ(i) to the subgroup V4 , reducing it to a direct sum of irreducible representations
γ (i) of V4 . Do the same thing for the subgroup Z3 .
CHAPTER 3. REPRESENTATIONS 20

3.9.5 Direct products

Obtain the direct products of all the irreducible representations of D3 with each other and reduce
them into direct sums of irreducible representations of D3 .
Using the character table for the tetrahedral group T derived in Q3.6
(a) reduce all the direct products Γ(i) ×Γ(j) , i, j = 1, . . . , 4 to direct sums of irreducible representations.
(b) restrict each Γ(i) to the subgroup V4 , reducing it to a direct sum of irreducible representations
γ (i) of V4 . Do the same thing for the subgroup Z3 .

3.9.6 More direct products

Let Γ and Γ(j) be two inequivalent irreducible representations of a finite group, G. Show that the
(i)

direct product representation Γ(i) ⊗ Γ(j)∗ does not contain the identity representation.
Show also that the direct product of an irreducible representation with its own complex conjugate
representation contains the identity representation once and only once.
Illustrate these results by applying them to the irreducible representations of Z4 .
Bibliography

[Arm88] M.A. Armstrong. Groups and Symmetry. Springer Verlag, New York, USA, 1988.
[Gol80] H. Goldstein. Classical Mechanics, 2nd Ed. Addison-Wesley, Reading MA, USA, 1980.
[LL76] L.D. Landau and E.M. Lifshitz. Mechanics, 3rd Ed. Course of Theoretical Physics, Vol.I.
Pergamon Press, Oxford, 1976.
[Sak64] J.J. Sakurai. Invariance principles and elementary particles. Princeton University Press,
Princeton NJ, USA, 1964.
[Tun85] Wu-Ki Tung. Group theory in physics. World Scientific, Singapore, 1985.

21