Page 73

3.3.2 Hermitian matrices

Definition 3.1.3. The Hermitian conjugate of a matrix A is denoted by A∗ and is
defined as the transpose of the complex conjugate of A, i.e.,

A∗ = ĀT . (3.45)

Example 3.1.6. The Hermitian conjugate of the matrix

 
1+i 2−i 3
4 5 6

is  
1−i 4
 2 + i 5 .
3 6

Definition 3.1.4. A square matrix H is called a Hermitian matrix if H = H ∗ .

Note that a real symmetric matrix is a special case of a Hermitian matrix.

Example 3.1.7. Consider the matrices

 
2 1−i
A= ,
1+i 7
 
2 1+i
B= ,
1+i 7
 
2 1−i
C= .
1+i 7+i
A is Hermitian, whereas B and C are not.

Example 3.1.8. The matrices

     
0 1 0 −i 1 0
σ1 = , σ2 = , σ3 =
1 0 i 0 0 −1

are Hermitian matrices known as Pauli matrices.

Hermitian matrices are fundamental to the formulation of quantum mechanics

known as matrix mechanics. In general, eigenvalues of a matrix can be real or
complex. However, Hermitian matrices are special in that their eigenvalues are
always real.
 Mathematical Methods in Chemistry (MATH0042) Page 74

Proposition 3.1.1. The eigenvalues of a Hermitian matrix are all real.

Proof. Let H denote a Hermitian matrix. Let λ be an eigenvalue of H corresponding

to the eigenvector v, so that Hv = λv. Then

λ⟨v , v⟩ = λv̄ T v
= v̄ T (λv)
(3.46)
= v̄ T (Hv)
= v̄ T H̄ T v.

Now, recall that for any two matrices A and B (of appropriate dimensions), we have
B T AT = (AB)T . Hence,

v̄ T H̄ T v = (H̄ v̄)T v
= (Hv)T v
= (λv)T v (3.47)
= λ̄v̄ T v
= λ̄⟨v , v⟩.

Combining (3.46) and (3.47), we get

λ⟨v , v⟩ = λ̄⟨v , v⟩. (3.48)

Since v ̸= 0 (v is an eigenvector), this implies that ⟨v , v⟩ ̸= 0. Thus, dividing both

sides of (3.48) by ⟨v , v⟩ =
̸ 0, we conclude that

λ = λ̄,

i.e., λ is real.

3.3.3 Orthogonal matrices

Definition 3.1.5. A square matrix A with real entries is called an orthogonal matrix
if
AT A = AAT = I, (3.49)
i.e., its inverse is equal to its transpose:

A−1 = AT . (3.50)

Example 3.1.9. The matrix

 
cos θ sin θ 0
R = − sin θ cos θ 0
0 0 1
Page 75

is orthogonal for any θ ∈ [0, 2π]. One can check this as follows. We have
 
cos θ − sin θ 0
RT =  sin θ cos θ 0 .
0 0 1

Therefore,
    
cos θ sin θ 0 cos θ − sin θ 0 1 0 0
RRT = − sin θ cos θ 0  sin θ cos θ 0 = 0 1 0 = RT R.
0 0 1 0 0 1 0 0 1

3.3.4 Unitary matrices

Definition 3.1.6. A square complex matrix U is called a unitary matrix if

U ∗ U = U U ∗ = I, (3.51)

i.e., its inverse is equal to its Hermitian conjugate:

U −1 = U ∗ . (3.52)

Note that an orthogonal matrix is a special case of a unitary matrix, namely one
whose entries are all real.

Example 3.1.10. The matrix

 iθ 
e 0
U=
0 e−iθ

is unitary for all θ ∈ R. One can check this as follows. We have

 −iθ 
∗ e 0
U = .
0 eiθ

Hence  iθ   −iθ   
∗ e 0 e 0 1 0
UU = = .
0 e−iθ 0 eiθ 0 1

Proposition 3.1.2. The eigenvalues of a unitary matrix have modulus one.

Proof. Let U denote a unitary matrix. Let λ be an eigenvalue of U corresponding

 Mathematical Methods in Chemistry (MATH0042) Page 76

to the eigenvector v, so that U v = λv. Then we have

v̄ T v = v̄ T Iv
= v̄ T Ū T U v
= (Ū v̄)T (U v)
(3.53)
= (U v)T (U v)

T
= λv λv
= λ̄λv̄ T v.
Since v is nonzero, v̄ T v is nonzero as well. Hence, dividing by v̄ T v, we arrive at
1 = λ̄λ = |λ|2 .

3.3.5 Normal matrices

Definition 3.1.7. A complex square matrix N is called normal if
N N ∗ = N ∗N
(i.e., if N commutes with its Hermitian conjugate).
Note that real symmetric, orthogonal, Hermitian and unitary matrices are all
special cases of normal matrices (see Problem sheet 7).

3.3.6 Diagonalisation of normal matrices

Lemma 3.1.1. Let N be a normal matrix. Then
N ∗ v = 0 ⇐⇒ N v = 0. (3.54)
Proof.
N ∗ v = 0 ⇐⇒ |N ∗ v|2 = 0. (3.55)
But
|N ∗ v|2 = (N ∗ v)∗ (N ∗ v)
= (N T v)T (N ∗ v)
= (v T N )(N ∗ v)
= vT N N ∗v
(3.56)
= vT N ∗N v
T
= (v T N )(N v)
= (N v)T (N v)
= |N v|2 .
It then follows from 3.55, that
N ∗ v = 0 ⇐⇒ N v = 0, (3.57)
as required.
Page 77

Lemma 3.1.2. Let N be a normal matrix. Then

N v = λv =⇒ N ∗ v = λv, (3.58)
i.e., if λ is an eigenvalue of N , then λ is an eigenvalue of N ∗ .
Proof.
N v = λv =⇒ (N − λI)v = 0. (3.59)
Let M = N − λI. Then
M ∗ = N ∗ − λI, (3.60)
and
M M ∗ = (N − λI)(N ∗ − λI)
= N N ∗ − λN − λN ∗ + |λ|2 I
(3.61)
= N ∗ N − λN − λN ∗ + |λ|2 I
= M ∗ M,
i.e., M is a normal matrix, too. But M v = 0. It then follows from Lemma 3.1.1,
that M ∗ v = 0, i.e.,
(N ∗ − λI)v = 0, (3.62)
from which follows the result.
Definition 3.1.8. Two complex vectors v and w are said to be mutually orthogonal
if
⟨v , w⟩ = 0. (3.63)
Proposition 3.1.3. Let N be a normal matrix. Then eigenvectors of N with
different eigenvalues are mutually orthogonal.
Proof. Suppose that v1 and v2 are eigenvectors of N with corresponding eigenvalues
λ1 and λ2 , respectively. Suppose furthermore that λ1 ̸= λ2 . We have
⟨N v1 , v2 ⟩ = ⟨λ1 v1 , v2 ⟩
(3.64)
= λ1 ⟨v1 , v2 ⟩.
But also
⟨N v1 , v2 ⟩ = (N v1 )∗ v2
= (N v1 )T v2
= (v1 T N ∗ )v2
(3.65)
= v1 T (N ∗ v2 )
= λ2 v1 T v2
= λ2 ⟨v1 , v2 ⟩,
where we have made use of Lemma 3.1.2 to get to the penultimate equality.
Hence, comparing (3.64) and (3.65), we get
(λ1 − λ2 )⟨v1 , v2 ⟩ = 0. (3.66)
But since λ1 ̸= λ2 , we conclude that ⟨v1 , v2 ⟩ = 0.
 Mathematical Methods in Chemistry (MATH0042) Page 78

Theorem 3.2. Given a normal matrix N , there exists a unitary matrix U such that
U ∗ N U = D, (3.67)
where D is a diagonal matrix whose entries along its main diagonal are the eigenval-
ues of N . (We say that U diagonalises N , and that N is unitarily diagonalisable.)
We will not give a proof of this result, except to mention that Proposition 3.1.3
can be used as part of such a proof. However, a matrix U that diagonalises an n × n
normal matrix H can be constructed as follows (note that U is not actually unique).

Step one. Find the eigenvalues λ1 , . . . λn of N and the corresponding eigenvec-

tors v1 , . . . , vn .

Step two. Find the normalised eigenvectors vb1 , . . . , vbn :

1
vbk = p vk , k = 1, . . . , n. (3.68)
⟨vk , vk ⟩

Step three. The matrix U is the matrix whose columns are the normalised
eigenvectors vb1 , . . . , vbn :

U = vb1 |b
v2 | . . . |b
vn . (3.69)
Note that we can arrange these columns in any order - the resulting matrix U will
always diagonalise N .

Example 3.2.1. Consider the matrix

 
1 i
H= .
−i 1
(H is a Hermitian matrix.) Let us find the unitary matrix U that diagonalises H.
To begin, let us compute the eigenvalues of H. We have
 
1−λ i
pH (λ) = det(H − λI) = det = λ2 − 2λ = λ(λ − 2).
−i 1 − λ
Hence the eigenvalues are λ1 = 0 and λ2 = 2.
Next, let us compute the corresponding eigenvectors.
ˆ λ = λ1 = 0. We have
    
1 i x1 0
(H − λ1 I) v1 = = ,
−i 1 x2 0
(
x1 + i x2 = 0,
=⇒
−i x1 + x2 = 0,
=⇒ x1 = −i x2 .
Hence, an eigenvector corresponding to the eigenvalue λ1 is
 
1
v1 = .
i
Page 79

ˆ λ = λ2 = 2. We have
    
−1 i x1 0
(H − λ2 I) v2 = = ,
−i −1 x2 0
(
−x1 + i x2 = 0,
=⇒
−i x1 − x2 = 0,
=⇒ x1 = i x2 .
Hence, an eigenvector corresponding to the eigenvalue λ2 is
 
i
v2 = .
1

In order to find the normalised eigenvectors, we need to compute the moduli of

v1 and v2 : p √
|v1 | = ⟨v1 , v1 ⟩ = 2,
p √
|v2 | = ⟨v2 , v2 ⟩ = 2,
Therefore, we obtain
   
1 1 1 i
vb1 = √ , vb2 = √ .
2 i 2 1

The matrix U is thus given by

 
1 1 i
U=√ .
2 i 1

It is straightforward to check that U is indeed unitary. In fact,

    
∗ 1 1 i 1 −i 1 0
UU = = .
2 i 1 −i 1 0 1

Furthermore, the matrix U diagonalises H:

     
∗ 1 1 −i 1 i 1 i 0 0
U HU = = .
2 −i 1 −i 1 i 1 0 2

Notice that the diagonal elements of this last diagonal matrix are the eigenvalues of
H.
 Part III

Symmetries

80
Chapter 4

Group theory

In this final part of the course, we shall present a mathematical formulation that
can be used to study (among other things) molecular symmetries. We begin in this
chapter by learning about group theory.

4.1 Definition of a group

Let G be a set, i.e., G = {g1 , g2 , ...}, for some g1 , g2 , ..., which are known as the
elements of G. Note that G might contain a finite number of elements, or an
infinite number of elements. Furthermore, g1 , g2 , ... do not necessarily have to be
numbers. For example, we might have G = {1, −1}, or G = R (i.e., the set of all
real numbers), or G might be the set of all real, invertible matrices.
An operation ⋆ on G is a rule for combining two elements g1 and g2 of G to obtain
a third element denoted by g1 ⋆ g2 . For example, ⋆ might represent multiplication
or addition, or matrix multiplication (if the elements of G are matrices).
Definition 4.0.1. Consider a set G and an operation ⋆ defined on G. We denote
this pair by (G, ⋆). (G, ⋆) is said to be a group if it satisfies the following four axioms
(referred to as the group axioms):

1. Closure: If g1 and g2 are contained in G, then so is g1 ⋆ g2 .

2. Associativity: For g1 , g2 , g3 ∈ G, we have g1 ⋆ (g2 ⋆ g3 ) = (g1 ⋆ g2 ) ⋆ g3 .
3. Identity: There exists an element I ∈ G satisfying I ⋆ g = g ⋆ I = g for all
g ∈ G. I is called the identity element of the group.
4. Inverse: For every g ∈ G, there is an element g −1 ∈ G which satisfies g⋆g −1 =
g −1 ⋆ g = I. g −1 is called the inverse of g.

Example 4.0.1. Some examples of groups are:

ˆ (R, +), i.e., the real numbers together with the operation of addition. One can
check that the group axioms 1–4 hold in this case, as follows:

1. If g1 , g2 ∈ R, then g1 + g2 ∈ R.
2. For g1 , g2 , g3 ∈ R, we have g1 + (g2 + g3 ) = (g1 + g2 ) + g3 .

81
 Mathematical Methods in Chemistry (MATH0042) Page 82

3. For all g ∈ R, we have g + 0 = 0 + g = g, i.e., the identity element for

this group is 0.
4. For every g ∈ R, we have g + (−g) = (−g) + g = 0, i.e., g −1 = −g.

ˆ GL(n, R). For each integer n ≥ 2, this denotes the group that consists of the
set of all n×n real, invertible matrices, together with the operation of ordinary
matrix multiplication. GL(n, R) is called the general linear group. One can
check that the group axioms 1–4 hold in this case, as follows:

1. If g1 , g2 are n × n real, invertible matrices, then so is g1 g2 (in particular,

note that the inverse of g1 g2 is g2−1 g1−1 ).
2. If g1 , g2 , g3 are n × n real, invertible matrices, then g1 (g2 g3 ) = (g1 g2 )g3
(note - more generally - that matrix multiplication is always associative).
3. The identity element for this group is In , i.e., the n × n identity matrix
(this is clearly itself an n × n real, invertible matrix).
4. For every n × n real, invertible matrix g, obviously (by definition) g −1
exists, but this is also an n × n real, invertible matrix.

Example 4.0.2. The following are not groups:

ˆ (R, ×), i.e., the set of all real numbers together with the operation of (scalar)
multiplication. For this, the above group axioms 1–3 all hold - in particular,
the identity element is 1 - but, axiom 4 does not: 0 has no inverse (there is
no real number, g, such that 0 × g = 1). Note, however, that (R \ {0}, ×) -
where R \ {0} denotes the set of all real numbers but excluding 0 - is a group
(Exercise: check this). Note that R \ {0} is sometimes denoted by R∗ .

ˆ (Z∗ , ×) (where Z∗ = Z \ {0}). For this, the above group axioms 1–3 all hold,
but axiom 4 does not: for any non-zero integer g ̸= ±1, the inverse of g is not
an integer.

Lemma 4.0.1. The identity element I of a group (G, ⋆) is unique.

Proof. Assume that there are two identity elements I and I ′ . Then, by definition,
they satisfy
g ⋆ I = I ⋆ g = g, for all g ∈ G, (4.1)
g ⋆ I ′ = I ′ ⋆ g = g, for all g ∈ G. (4.2)
But then, choosing g = I ′ in (4.1) gives
I ′ ⋆ I = I ⋆ I ′ = I ′.
On the other hand, choosing g = I in (4.2) gives
I ⋆ I ′ = I ′ ⋆ I = I.
Page 83

Thus we conclude that I = I ′ .

Lemma 4.0.2. Suppose that (G, ⋆) is a group and that g ∈ G. Then the inverse of
g is unique.
Proof. Suppose that x and y are both inverses of g. Then, x, y ∈ G and by definition,
x⋆g =g⋆x=I (4.3)
and
y ⋆ g = g ⋆ y = I. (4.4)
Then we have
(x ⋆ g) ⋆ y = I ⋆ y = y, (4.5)
where the first equality follows from (4.3). But by associativity, we can also write:
(x ⋆ g) ⋆ y = x ⋆ (g ⋆ y) = x ⋆ I = x, (4.6)
where the second equality follows from (4.4). Comparing (4.5) and (4.6), we conclude
that
y = x. (4.7)

Definition 4.0.2. A group (G, ⋆) is called a finite group if G contains a finite

number of elements. Then, the number of elements contained in G is called the order
of the group and is denoted by |G|. If G contains an infinite number of elements,
then (G, ⋆) is said to be an infinite group, or, of infinite order.
For a finite set G that contains, say, n elements g1 , . . . , gn , where n is not too
large (if n is too large, the following is not practical to implement), together with an
operation ⋆, one can write down what is known as a composition table, also known
as a Cayley table. For 1 ≤ i, j ≤ n, the i-th row and j-th column of this table are
labelled by gi and gj , respectively, and the entry that lies in the intersection of these
is gi ⋆ gj . One can use this table to check whether (G, ⋆) is a group or not. If (G, ⋆)
is a group, this table is also referred to as its group table.

Example 4.0.3. Consider

G = {1, −1} (4.8)
together with the operation of multiplication ×. Then the corresponding composi-
tion table is
1 −1
1 1 −1
−1 −1 1

It is straightforward to check that in this case, (G, ×) is a group, and of course is

finite.
Now consider the same set G but together with the operation of addition. Then
the corresponding composition table is
 Mathematical Methods in Chemistry (MATH0042) Page 84

1 −1
1 2 0
−1 0 −2

In this case, (G, +) is clearly not a group (since ±2, 0 are not contained in G).

Example 4.0.4. Consider

G = {1, i, −1, −i} (4.9)
together with the operation of multiplication ×. Then the corresponding composi-
tion table is
1 i −1 −i
1 1 i −1 −i
i i −1 −i 1
−1 −1 −i 1 i
−i −i 1 i −1

One can check that in this case, (G, ×) is a group, and of course is finite.

Some further definitions:

Definition 4.0.3. A group (G, ⋆) is said to be Abelian (or commutative) if g1 ⋆g2 =
g2 ⋆ g1 for all g1 , g2 ∈ G.

Example 4.0.5. (R, +) is an Abelian group, but GL(n, R) is not. The groups
({1, −1}, ×) and ({1, i, −1, −i}, ×) considered in Examples 4.0.3 and 4.0.4 are also
Abelian groups.

Definition 4.0.4. A subgroup of a group (G, ⋆) is a group (H, ⋆) consisting of a

subset H of G together with the same operation ⋆.

Example 4.0.6. (Z, +) is a subgroup of (R, +). ({1, −1}, ×) is a subgroup of

({1, i, −1, −i}, ×).

Some important subgroups of the general linear group, GL(n, R), are as follows:
ˆ The special linear group SL(n, R), defined by
SL(n, R) = {A ∈ GL(n, R) | det A = 1}.

ˆ The orthogonal group O(n), defined by

O(n) = {A ∈ GL(n, R) | A is orthogonal}.

ˆ The special orthogonal group SO(n), defined by

SO(n) = {A ∈ GL(n, R) | det A = 1 and A ∈ O(n)}.