Lecture Notes on Mathematical Method of Physics I

Dr. A. N. Njah,
Department of Physics,
University of Agriculture,
Abeokuta.

PHS 471: Linear Algebra: Transformation in linear vector spaces and ma-
trix theory.
Functional analysis; Hilbert space, complete sets of orthogonal functions; Lin-
ear operations.
Special functions: Gamma, hypergometric, Legendre, Bessel, Hermite and
Laguerre functions. The Dirac delta function
Integral transform and Fourier series: Fourier series and Fourier transform;
Application of transform methods to the solution of elementary differential
equations in Physics and Engineering.

Suggested reading.

1. Advanced Engineering Mathematics by E. Kreyszig.

2. Further Engineering Mathematics by H.A. Stroud
3. Mathematical Physics by B.D. Gupta
4. Advanced Mathematics for Engineers and Scientist (Schaum series) by
Spiegel

1
Contents

1 LINEAR ALGEBRA 5
1.1 Vector Space or Linear Space . . . . . . . . . . . . . . . . . . 5
1.1.1 Algebraic Operations on Vectors . . . . . . . . . . . . . 6
1.1.2 Linearly Dependent and Independent sets of vectors . . 6
1.2 Matrix Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1 Determinant of a Matrix . . . . . . . . . . . . . . . . . 9
1.2.2 General properties of determinants . . . . . . . . . . . 10
1.2.3 Adjugate Matrix or Adjoint of a Matrix . . . . . . . . 10
1.2.4 Reciprocal Matrix or Inverse of a Matrix . . . . . . . . 11
1.2.5 The Transpose of a Matrix . . . . . . . . . . . . . . . . 11
1.2.6 Symmetric, Skew-symmetric and Orthogonal Matrices . 11
1.3 Complex Matrices . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.1 The Conjugate of a Matrix . . . . . . . . . . . . . . . . 12
1.3.2 The Conjugate transpose or Hermitian Conjugate of a
Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.3 Hermitian, Skew-Hermitian and Unitary Matrices . . . 13
1.4 Matrix Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4.1 Rank of a Matrix . . . . . . . . . . . . . . . . . . . . . 13
1.5 Consistency of equations . . . . . . . . . . . . . . . . . . . . . 15
1.5.1 Homogeneous and Non-Homogeneous Linear Equations 15
1.5.2 Uniqueness of Solutions . . . . . . . . . . . . . . . . . 16
1.6 Solution of sets of Equations . . . . . . . . . . . . . . . . . . . 17
1.6.1 Inverse method . . . . . . . . . . . . . . . . . . . . . . 17
1.6.2 Row Transformation method . . . . . . . . . . . . . . . 19
1.6.3 Gaussian elimination method . . . . . . . . . . . . . . 20
1.6.4 Triangular Decomposition method: LU-decomposition . 20
1.6.5 Cramer’s Rule . . . . . . . . . . . . . . . . . . . . . . . 20
1.7 Eigenvalues and Eigenvectors of a Matrix . . . . . . . . . . . . 20

2
 1.7.1 Nature of the eigenvalues and eigenvectors of special
types of matrices . . . . . . . . . . . . . . . . . . . . . 22
1.7.2 Diagonalisation of a matrix . . . . . . . . . . . . . . . 24
1.8 Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.8.1 Transformation . . . . . . . . . . . . . . . . . . . . . . 25
1.8.2 Resultant of two linear transformation . . . . . . . . . 26
1.8.3 Similarity transformation . . . . . . . . . . . . . . . . . 26
1.8.4 Unitary transformation . . . . . . . . . . . . . . . . . . 26
1.8.5 Orthogonal transformation . . . . . . . . . . . . . . . . 26
1.8.6 Orthogonal set . . . . . . . . . . . . . . . . . . . . . . 27
1.9 Bases and dimension . . . . . . . . . . . . . . . . . . . . . . . 27
1.9.1 Linear Maps . . . . . . . . . . . . . . . . . . . . . . . . 27

2 FUNCTIONAL ANALYSIS 28
2.1 Normed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.1.1 Cauchy Sequences . . . . . . . . . . . . . . . . . . . . . 28
2.1.2 Completeness . . . . . . . . . . . . . . . . . . . . . . . 28
2.1.3 Pre-Hilbert spaces . . . . . . . . . . . . . . . . . . . . 29
2.1.4 Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . 29
2.1.5 Geometry of Hilbert space . . . . . . . . . . . . . . . . 29

3 SPECIAL FUNCTIONS 30
3.1 The gamma and beta functions . . . . . . . . . . . . . . . . . 30
3.1.1 The gamma function Γ . . . . . . . . . . . . . . . . . . 30
3.1.2 The beta function, β . . . . . . . . . . . . . . . . . . . 32
3.1.3 Application of gamma and beta functions . . . . . . . . 32
3.2 Bessel’s Functions . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3 Legendre’s Polynomials . . . . . . . . . . . . . . . . . . . . . . 35
3.4 Hermite Polynomials . . . . . . . . . . . . . . . . . . . . . . . 36
3.5 Laguerre Polynomials . . . . . . . . . . . . . . . . . . . . . . . 38
3.5.1 Hypergeometric Function . . . . . . . . . . . . . . . . . 39

4 INTEGRAL TRANSFORM AND FOURIER SERIES 41

4.1 Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . 41
4.1.1 Inverse Transform . . . . . . . . . . . . . . . . . . . . . 44
4.1.2 Solution of Differential Equations by Laplace Transform 46
4.2 The Dirac Delta Function − the impulse function . . . . . . . 48

3
 4.2.1 Integration involving the impulse function . . . . . . . 49
4.2.2 Differential equations involving the impulse function . . 49
4.3 Fourier Series . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.3.1 Fourier series of functions of period 2π . . . . . . . . . 51
4.3.2 Half-range series . . . . . . . . . . . . . . . . . . . . . 54
4.3.3 Functions with arbitrary period T . . . . . . . . . . . . 54
4.3.4 Sum of a Fourier series at a point of finite discontinuity 55
4.4 Fourier Integrals . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.4.1 The Fourier integral . . . . . . . . . . . . . . . . . . . 55

4
Chapter 1

LINEAR ALGEBRA

1.1 Vector Space or Linear Space

A vector space V over a field F is a set of elements called vectors which may
be combined by two operations - addition and scalar multiplication; such that
a) if the vectors a and b belong to V , then
a + b also belongs to V . (This is known as closure property)
(i)
a + b = b + a (commutative law of addition)
(ii)
(iii)(a + b) + c = a + (b + c) = b + (c + a) (associative law of addition)
(iv) there exist an additive identity vector 0 known as the null vector
such that a + 0 = a
(v) to every vector a in V , there corresponds a vector −a known as the
additive inverse vector, such that a + (−a) = 0
b) if m, n (elements of F ) are any two scalars and a is a vector in V , then
(i) (m + n)a = ma + na (distributive law)
(ii) m(a + b) = ma + mb (distributive law)
(iii) m(na) = (mn)a = n(ma) (associative law of multiplication)
(iv) to every vector a in V , there corresponds a multiplicative identity
scalar 1, such that 1a = a
Conditions for a physical quantity to be representable by a vector:
• It must obey the parallelogram law of addition (
a + b
≤
a
+
b
)
• It must have magnitude as well as direction independent of any choice
of coordinate axes.

5
1.1.1 Algebraic Operations on Vectors
 and B
If A  are two vectors with components (a1, a2 , . . . , an ) and (b1, b2, . . . , bn)
respectively then
±B
(i) A  = (a1 ± b1, a2 ± b2, . . . , an ± bn )
 = (ka1 , ka2, . . . , kan), k a scalar
(ii) k A
·B
(iii) A  = a1 b1 + a2 b2 + . . . + an bn
 =1
(iv) A vector will be a unit vector if the magnitude |A|
 and B
(v) The vectors A  will be orthogonal if A
 ·B
 =0

1.1.2 Linearly Dependent and Independent sets of vectors

A set a vectors A  2, . . . A
 1, A  n is said to be linearly dependent if there exist
a set of n scalars k1, k2, . . . , kn (of which at least one is non-zero) such that
 1 + k2 A
k1 A  2 + . . . + kn A n = 0
In the case when k1 = k2 = . . . = kn = 0, in order that k1 A  1 + k2 A 2 + . . . +
 n = 0
kn A
the set of n vectors A  1, A 2, . . . , A
 n is said to be linearly independent
A vector A  is known as a linear combination of the set A  1, A 2, . . . A
 n if it is
expressable as A  = k1 A  1 + k2 A  2 + . . . + kn A
n
Examples:

1. The set of vectors (1,2,3), (2,-2,0) is linearly independent since k1(1, 2, 3)+
k2(2, −2, 0) = (0, 0, 0) is equivalent to the set of equations k1 + 2k2 =
0, 2k1 − 2k2 = 0 and 3k1 = 0 which gives k1 = k2 = 0.
2. The set of vectors (2,4,10), (3,6,15) is linearly dependent since k1 (2, 4, 10)+
k2(3, 6, 15) = (0, 0, 0) gives the system 2k1 + 3k2 = 0, 4k1 + 6k2 =
0, 10k1 + 15k2 = 0 ⇒ k1 = 3, k2 = −2

6
1.2 Matrix Theory
A set of numbers arranged in a rectangular array of m rows and n columns
such as ⎛ ⎞
a
⎜ 11
a12 . . . a1n ⎟
⎜ a21 a22 . . . a2n ⎟
⎜
⎟
⎜ ⎟
⎜ ... . . . . . . . . .
⎜ ⎟
⎟
⎝ ⎠
am1 am2 . . . amn
is called a matrix of order m × n or an m × n matrix. If m = n (i.e.
number of rows = number of columns) it is called a square matrix of order
n. aij , (i = 1, 2, . . . , m; j = 1, 2, . . . , n) are called its elements or constituents
or entries. aij represents the element in the ith row and j th column of the
matrix. The element aij (i = j) of a square matrix A lie on the main diagonal
or principal diagonal and are called its diagonal elements. The sum of the
diagonal elements is called the trace of A and is denoted by trA = ni=1 aii


A matrix can be represented by A or [aij ]

Null matrix: A matrix having all of its elements zero
Row matrix and column matrix: A row matrix is a matrix having only a single
row i.e. a 1 × n matrix. A column matrix is one having a single column, i.e.
m × 1 matrix
Equality of matrices: Two matrices A = [aij ] and B = [bij ] are equal if both
are of the same order (size) m × n and each element aij of A is equal to the
corresponding element bij of B.
Equivalent relation on matrices:
1. Reflexivity A = A
2. Symmetry A = B ⇒ B = A
3. Transitivity A = B and B = C ⇒ A = C
Addition and Subtraction of Matrices: Addition and subtraction are defined
only for matrices A and B of the same order or size and are done by adding
and subtracting corresponding entries. Addition of matrices is commutative,
associative, distributive by a scalar, has an identity and an inverse, obeys the
cancellation law (A + B = A + C ⇒ B = C)
Multiplication of a matrix by a scalar: The product of a matrix A by a scalar
c is obtained by multiplying each entry of A by c

7
Multiplication of a matrix by a matrix: For two matrices A = [aij ] and
B = [bij ] to be multiplied (i.e. for C=AB to be defined) the number of
columns of A must be equal the number of rows of B; i.e. if A is a p × n
matrix B must be an n × q matrix; C=AB is a p × q matrix whose elements
cij in the ith row and j th column is the algebraic sum of the products of the
elements in the ith row of A by the corresponding elements in the j th column
of B
cij = nk=1 aik bkj = ai1 b1j + ai2b2j + . . . + ain bnj


Matrix multiplication is:

• not commutative i.e. AB = BA
• distributive; A(B + C) = AB + AC
• associative; (AB)C = A(BC)
Matrix multiplication differ from multiplication of numbers in that:
1. it is not commutative i.e. AB = BA
2. AB = 0 does not necessarily imply A = 0 or B = 0 or BA = 0
3. AC = AD does not necessarily imply C = D
Upper and lower triangular Matrices: A square matrix U of order n is said to
be an upper triangular matrix if its elements uij = 0 for i > j. On the other
hand a square matrix L is said to be lower triangular matrix if its elements
lij = 0 for i < j
Examples
⎛ ⎞ ⎛ ⎞
⎜
u 11 u 12 u 13 ⎟ ⎜
l 11 0 0 ⎟
U=⎜ ⎜ 0
⎝
u 22 u ⎟
23 ⎠
⎟ L = ⎜
l
⎝ 21 22
⎜ l 0 ⎟
⎟
⎠
0 0 u33 l31 l23 l33
Diagonal Matrix: It is a square matrix D which is both upper and lower
triangular. If the diagonal elements are equal to a scalar quantity λ i.e.
d11 = d22 = . . . = dnn = λ then the matrix is called a scalar matrix S
(because multiplication of any square matrix A of the same size by S has the
same effect as the multiplication by a scalar λ, that is AS = SA = λA). In
particular, if λ = 1 we have a unit matrix or identity matrix I

8
Examples
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
d
⎜ 11
0 0 ⎟ ⎜
λ 0 0 ⎟ ⎜
1 0 0 ⎟
D=⎜ ⎜
⎝
0 d22 0 ⎟
⎟
⎠
, S=⎜
⎜
⎝
0 λ 0⎟⎟
⎠
, I=⎜
⎜
⎝
0 1 0 ⎟
⎟
⎠
0 0 d33 0 0 λ 0 0 1

1.2.1 Determinant of a Matrix

Determinants were originally introduced for solving linear systems (by Cramer’s
rule). They have important engineering application in eigenvalue problems,
differential equations, vector algebra, etc.
The determinant of a 2 × 2 matrix A is called a second order determinant
and is denoted and defined by
a a

|A| or detA =  11 12 

 = a11 a22 − a12 a21
a21 a22 
Similarly the determinant of a 3 × 3 matrix A is referred to as a third order
determinant and is defined by
a
 11 a12 a13 
 

a a a a a a
     
22 23 12 13 12 13
  
detA =  a21 a22 a23  = a11   − a21   + a31 
       
a a a a a a

32 33 32 33 22 23
    
a
31 a32 a33
 

= a11 M11 − a21 M21 + a31M31
Here we have expanded the determinant by the first column. Determinant
can be expanded by any row or column
Minor and Cofactors of
 a third  order determinant
a
 22 a23 
a
 12 a13 
a
 12 a13 
  
  
M11 =  , M21 =   and M31 = 
a32 a33  a
32 a33
a
22 a23

 
are called the minors of a11 , a21 and a31 respectively while their cofactors are
C11 = +M11 , C21 = −M21 and C31 = +M31 respectively. ⇒ the cofactor
of any one element is its minor together with its ”place sign”. The ”place
signs” in Cij form a checkerboard pattern as follows:
+ − +
− + −
+ − +
In general a determinant of n-order is a scalar associated with an n×n matrix
A = [aij ], which is written
 
a11 a12 . . . a1n






a 
a22 . . . a2n 
DetA = 21
 
 
... .........






am1 am2 . . . amn





9
 1.2.2 General properties of determinants
Theorem 1: (Behaviour of nth order determinant under elementary row op-
eration)
(a) Interchange of two rows multiplies the value of the determinant by −1.
(b) Addition of a multiple of a row to another row does not alter the value
of the determinant.
(c) Multiplication of a row by c multiplies the value of the determinant by
c
Theorem 2: (further properties of the nth -order determinant)
(a)-(c) in theorem 1 hold also for columns
(d) Transposition leaves the value of a determinent unaltered
(e) A zero row or column renders the value of a determinant zero.
(f) Proportional rows or columns render the value of a determinant zero.
In particular, a determinant with two identical rows or columns has the
value zero.
Singular and Non-singular matrices
A square matrix A is known as singular matrix if its determinant |A| = 0.
In case |A| =
 0 then A is known as non-singular matrix

1.2.3 Adjugate Matrix or Adjoint of a Matrix

Let A = [aij ] be a square matrix of order n and cij represents the cofactor
of the element aij in the determinant |A| then the transpose of the matrix of
cofactors (C = [cij ]) denoted by C′ = [cji] is called the adjugate or adjoint of
A and is denoted adjA.
Properties:
• A(adjA) = (adjA)A.............................∗
i.e. multiplication of A by adjA is commutative and their product is a
scalar matrix having every diagonal element as |A|.
• |adjA| = |A|n−1, where n is the order of the matrix.

10
 • If |A| = 0 then A(adjA) = (adjA)A = 0
• adj(AB) = adjBadjA
or adj(ABC) = adjCadjBadjA (prove!!)

1.2.4 Reciprocal Matrix or Inverse of a Matrix

If AB = BA = I (the identity matrix) then B is said to be the inverse of A
and vice versa and is denoted by A−1. That is AA−1 = A−1A = I
When the inverse of A exists, then A is said to be invertible. The necessary
and sufficient condition for a square matrix to be invertible is that it is non-
singular. From eq.(*) A−1 is given by
1
A−1 = adjA
|A|

1.2.5 The Transpose of a Matrix

If A is a matrix of order m × n the transpose of A denoted by A′ or AT , is
a matrix of order n × m obtained by interchanging its rows and columns
Examples ⎛ ⎞
⎛ ⎞ a11 a21
a a a
If A = ⎝ 11 12 13 ⎠ then A′ = ⎜
⎜ ⎟
⎜ a12 a22 ⎟
⎟
a21 a22 a23 ⎝ ⎠
a13 a23
Properties
• (A′)′ = A
• |A′ | = |A|
• (kA)′ = kA′ for k a scalar
• (A + B)′ = A′ + B′
• (AB)′ = B′A′

1.2.6 Symmetric, Skew-symmetric and Orthogonal Matrices

A square matrix A is said to be:
(i) symmetric if A′ = A i.e. if aij = aji for all i and j, i.e. if transposition
leaves it unchanged

11
(ii) skew-symmetric if A′ = −A i.e. if transposition gives the negative of A
(iii) orthogonal if A′ = A−1 i.e. if transposition gives the inverse of A.
Every square matrix A = P + Q
where P = 12 (A + A′ ) is a symmetric matrix
and Q = 12 (A − A′) is a skew-symmetric matrix

1.3 Complex Matrices

1.3.1 The Conjugate of a Matrix
If the elements of a matrix A are complex quantities, then the matrix ob-
tained from A, on replacing its elements by the corresponding conjugate
complex numbers, is said to be the conjugate matrix of A and is denoted by
Ā or A∗, with the following properties:
• (A∗)∗ = A
• (A + B)∗ = A∗ + B∗
• If α is a complex number and A a matrix of any order say m × n, then
(αA)∗ = α∗ A∗
• (AB)∗ = A∗B∗

1.3.2 The Conjugate transpose or Hermitian Conjugate of a Ma-

trix
The matrix, which is the conjugate of the transpose of a matrix A is said to
be the conjugate transpose of A and denoted by A† (called A dagger)
Example:
⎛ ⎞
−2 + 3i 3 − 4i i
A=⎝ ⎠
−5i −5 − 3i 4 + i
⎛ ⎞
−2 − 3i 3 + 4i −i
A∗ = ⎝ ⎠
5i −5 + 3i 4 − i
⎛ ⎞
−2 − 3i 5i
A† = ⎜
⎜ ⎟
⎜
⎝
3 + 4i −5 + 3i ⎟
⎟
⎠
−i 4−i

12
 Properties:

• (A†)† = A
• (A + B)† = A† + B†
• If α is a complex number and A a matrix, then (αA)† = α∗A†
• (AB)† = B†A† (prove!!)

1.3.3 Hermitian, Skew-Hermitian and Unitary Matrices

A matrix A is said to be
(i) Hermitian if A† = A
(ii) skew-Hermitian if A† = −A
(iii) unitary if A† = A−1
Examples:
√
⎛ ⎞
5 2 + 3i −i ⎛ ⎞ ⎛
1 1
⎞
⎜ ⎟ 3i 2+i 2i
√ 2 3
A= ⎜
2 − 3i 3 −3 − 4i ⎟,B
⎟
=⎝ ⎠ and C = ⎝ 1 1
⎠
−2 + i −i 3 2i
⎜
⎝ ⎠
i −3 + 4i 0 2

are Hermitian, skew-Hermitian, and unitary respectively.

Every square complex matrix A = P + Q
where P = 12 (A + A† ) is a Hermitian matrix
and Q = 21 (A − A†) is a skew-Hermitian matrix

1.4 Matrix Algebra

1.4.1 Rank of a Matrix
It is the order of the largest determinant that can be formed from the ele-
ments of the matrix. A matrix A is said to have rank r if it contains at least
one square submatrix of r rows with a non-zero determinant, while all square
submatrices of (r + 1) rows, or more, have zero determinants.
Examples:

13
 ⎛ ⎞  
4 2 4 2 

1. A = ⎝ is of rank 2 since |A| = = 18 i.e. not zero.

⎠ 
1 5 1 5 


⎛ ⎞  
6 3⎠  6 3 
 
2. B = ⎝ gives |B| =   = 0.
8 4  8 4 


Therefore B is not of rank 2. It is, however, of rank 1, since the subma-

trices (6),(3),(8),(4) are not zero. This implies that only a null matrix
is of rank zero.
⎛ ⎞ ⎛ ⎞
⎜
1 2 8 ⎟ ⎜
3 4 5⎟
3. Determine the ranks of (a) A = ⎜
⎜
⎝
4 7 6 ⎟
⎟
⎠
and (b) B = ⎜
⎜
⎝
1 2 3⎟⎟.
⎠
 
9 5 3 4 5 6



1 2 8 
Since |A| = 4 7 6  = −269 is not zero the rank of A is 3.






9 5 3 



3 4 5 
Since |B| = 1 2 3  = 0 the rank of B is not 3. We now try subma-




 4 5 6 
trices of order 2:
 3 4 
 
  = 2 = 0, therefore, rank of B is 2. We could equally well have
 1 2 
 

tested
               
 4 5   1 2   2 3   3 5   1 3   4 5   3 5   3 4 
               
 ,  ,  ,  ,  ,  ,  ,  . i.e. we
 2 3   4 5   5 6   1 3   4 6   5 6   4 6   4 5 
               

test all possible second order minors to find one that is not zero.
For a rectangular matrix of order m × n the rank is given by the order of the
largest square sub-matrix formed by the elements.
Example: ⎛ ⎞
⎜
2 2 3 1 ⎟
For a 3 × 4 matrix ⎜ ⎜ 0 8 2 4 ⎟ the largest square sub-matrix cannot be
⎝
⎟
⎠
1 7 3 2 
 2 2 3 
 
 
greater than order 3. We try  0 8 2  = 0
 1 7 3 
 

But we must
 also try other 3 × 3 sub-matrices, e.g,
 2 3 1 
 
 
 8 2 4  = 30 = 0, therefore, B is of rank 3
 
 
 7 3 2 
 

14
1.5 Consistency of equations
1.5.1 Homogeneous and Non-Homogeneous Linear Equations
A set of m simultaneous linear equations in n unknowns
a11 x1 + a12 x2 + . . . + a1nxn = b1
a21 x1 + a22 x2 + . . . + a2nxn = b2
..............................
am1 x1 + am2 x2 + . . . + amn xn = bm
can be written in the matrix form as follows
⎛ ⎞⎛ ⎞ ⎛ ⎞
⎜
a11 a12 . . . a1n ⎟⎜
x1 ⎟ ⎜
b1 ⎟
⎜
a21 a22 . . . a2n ⎟⎜
x2 ⎟ ⎜
b2 ⎟
=
⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜
⎜
⎝
... ......... ⎟⎜
⎟⎜
⎠⎝
... ⎟
⎟
⎠
⎜
⎜
⎝
... ⎟
⎟
⎠
am1 am2 . . . amn xn bm
i.e. Ax = b
This set of equations is homogeneous if b = 0, i.e. (b1, b2, . . . , bm ) = (0, 0, . . . , 0)
otherwise it is said to be non-homogeneous. This set of equations is said to
be consistent if solutions for x1, x2, . . . , xn exist and inconsistent if no such
solutions can be found.
The Augmented coefficient matrix Ab of A is
⎛ ⎞
⎜
a11 a12 . . . a1n b1 ⎟
⎜
a21 a22 . . . a2n b2 ⎟
Ab =
⎜ ⎟
⎜ ⎟
⎜
⎜
⎝
... ......... ... ⎟
⎟
⎠
am1 am2 . . . amn bm

which is an m × (n + 1) matrix formed by writing the constant terms as an

(n + 1)th column of the coefficient matrix A. Note:
• If the rank of the coefficient matrix A equal to the rank of the augmented
matrix Ab then the equations are consistent.
• If the rank of A is less than the rank of Ab then the equations are
inconsistent and have no solution.
• If the rank of the m × n matrix is r, then it has r linearly indepen-
dent column vectors and the remaining n − r column vectors is a linear
combination of these r column vectors.

15
Example:
⎛ ⎞⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 3 x
⎠⎝ 1
4 1 3 1 3 4
If ⎝ ⎠ =⎝ ⎠ then A = ⎝ ⎠ and Ab = ⎝ ⎠
2 6 x2 
5 2 6 2 6 5

1 3 
Rank of A: = 0 ⇒ rank of A = 1



 2 6  

 1 3 
 
Rank of Ab :   = 0 as before
 2 6 
 
 

3 4 
but = −9 ⇒ rank of Ab = 2 In this case rank of A is less than rank



6 5 
of Ab ⇒ no solution exists.

1.5.2 Uniqueness of Solutions

1. With a set of n equations in n unknowns, the equations are consistent if
the coefficient matrix A and the augmented matrix Ab are each of rank
n. There is then a unique solution for the n equations.
2. If the rank of A and that of Ab is m, where m < n, then the matrix A is
singular, i.e. |A| = 0, and there will be an infinite number of solutions
for the equations.
3. If the rank of A is less than the rank of Ab, then no solution exists.
That is, with a set of n equations in n unknowns
(i) a unique solution exists if rank A = rank Ab = n
(ii) an infinite number of solution exists if rank A = rank Ab = m < n
(iii) no solution exists if rank A<rank Ab

Example: ⎛ ⎞⎛ ⎞ ⎛ ⎞
−4 5 ⎠ ⎝ x1 ⎠ −3 ⎠
Show that ⎝ = ⎝ has an infinite number of solutions.
⎛
−8⎞ 10 x2 ⎛ −6 ⎞  
−4 5 ⎠ −4 5 −3 ⎠ 
−4 5 
A= and Ab = Rank A: =   = 0 ⇒

⎝ ⎝ 
−8 10 
−8 10 
−6  
 −8 10 
 
 −4 5   5 −3   −4 −3 
     
rank A = 1 Rank Ab : =  
 = 0,
 
 = 0,
   =
 −8 10   10 −6   −8 −6 
  

0 ⇒ rank of Ab = 1. Therefore rank of A=rank of Ab = 1 < n = 2.

16
Therefore an infinite number of solutions exist.
For the homogeneous linear equations
⎛ ⎞⎛ ⎞ ⎛ ⎞
⎜
a11 a12 . . . a1n ⎟⎜
x1 ⎟ ⎜
0 ⎟
⎜
a21 a22 . . . a2n ⎟⎜
x2 ⎟ ⎜
0 ⎟
=
⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜
⎜
⎝
... ......... ⎟⎜
⎟⎜
⎠⎝
... ⎟
⎟
⎠
⎜
⎜
⎝
... ⎟
⎟
⎠
am1 am2 . . . amn xn 0

i.e. Ax = 0 (∗)
Let r be the rank of the matrix A of order m × n. We have the following
results:
1. A system of homogeneous linear equations always have one or more
solutions. The two cases are: r = n or r < n. For r = n the eq(*)
will have no linearly independent solutions, for in that case trivial (zero)
solution is the only solution, while in the case r < n there will be (n − r)
independent solutions and therefore the eq(*) will have more than one
solution.
2. The number of linearly independent solutions of Ax = 0 is (n − r) i.e.
if we assign arbitrary values to (n − r) of the variables, then the values
of the others can be uniquely determined.
Since the rank of A is r, it has r linearly independent columns.
3. If the number of equations is less than the number of variables, then the
solution is always other than x1 = x2 = . . . = xn = 0 (i.e. the solution
is always non-trivial solution)
4. If the number of equations is equal to the number of variables a necessary
and sufficient condition for solutions other than x1 = x2 = . . . = xn = 0
is that the determinant of the coefficients must be zero.

1.6 Solution of sets of Equations

1.6.1 Inverse method
To solve Ax = b we use x = A−1b. To find A−1 proceed as follows:
(i) evaluate |A|. If |A| = 0 then stop (no solution) else proceed.

17
(ii) Form C, the matrix of cofactors of A (the cofactor of any element is its
minor together with its ’place sign’)
(iii) Write C′ , the transpose of C
1
(iv) Then A−1 = |A| × C′
Example: To solve the system
3x1 + 2x2 − x3 = 4
2x1 − x2 + 2x3 = 10
x1 − 3x2 − 4x3 = 5
we rewrite it ⎞
⎛ in ⎛matrix ⎞ form ⎛ as ⎞
⎜
3 2 −1 ⎟ ⎜ 1 ⎟
x ⎜
4 ⎟
⎜ 2 −1 2 ⎟ ⎜ x ⎟ = ⎜ 10 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ ⎠ ⎝ 2 ⎠ ⎝ ⎠
1 −3 −4 x3 5
i.e. Ax
⎛ = b ⎞
⎜
3 2 −1 ⎟
A=⎜ ⎜ 2 −1
⎝
2 ⎟
⎟
⎠
1 −3 −4 
 3 2 −1
 

 
(i) |A| =  2 −1 2  = 55


 1 −3 −4 
 
⎛ ⎞
c
⎜ 11
c 12 c 13 ⎟
(ii) C = ⎜ ⎜
c c22 c23 ⎟
⎝ 21
⎟
⎠
c31

c 32 c 33    
 −1 2  2 2  2 −1 
     
where c11 =   = 10, c12 = −   = 10, c13 = 
 = −5,
 
    
 −3 −4   1 −4   1 −3 

     
 2 −1   3 −1   3 2 
   
c21 = −  
 = 11, c22 = 

 = −11, c23 = − 
 
 = 11,
 −3 −4   1 −4   1 −3 
     
 2 −1  3 −1   3 2
     
c31 =   = 3, c32 = −   = −8, c33 = 
 = −7
 
    
 −1 2  2 2  2 −1 

 
⎛ ⎞
⎜
10 10 −5 ⎟
So C = ⎜ ⎜
⎝
11 −11 11 ⎟ ⎟
⎠
3 −8 −7

18
 ⎛ ⎞
⎜
10 11 3 ⎟
′
(iii) C = ⎜ ⎜
⎝
10 −11 −8 ⎟ ⎟
⎠
(i.e. the adjoint of A)
−5 −11 ⎛
−7 ⎞
10 11 3
1 1 ⎜
⎜ ⎟
(iv) A−1 = |A| C′ = 55 ⎜ 10 −11 −8 ⎟
⎝
⎟
⎠
⎛ ⎞ ⎛ ⎞
−5 −11
⎛
−7 ⎞⎛ ⎞ ⎛ ⎞
x
⎜ 1 ⎟
b
⎜ 1 ⎟
10 11 3 4 40 + 110 + 15
1 ⎜ 1 ⎜
⎜ ⎟⎜ ⎟ ⎜ ⎟
−1
So ⎜
⎜
x ⎟=A ⎜
⎝ 2 ⎠
⎟ ⎜
b ⎟ = 55 ⎜
⎝ 2 ⎠
⎟
⎝
10 −11 −8 ⎟ ⎟⎜
⎠⎝
⎜ 10 ⎟ =
⎟
⎠
⎜ 40 − 110 − 40
55 ⎝
⎟
⎟
⎠
=
⎛
x⎞3 b3 −5 −11 −7 5 −20 + 10 − 35
⎜
3 ⎟
⎜ −2 ⎟
⎜ ⎟
⎝ ⎠
1
Therefore x1 = 3, x2 = −2, x3 = 1

1.6.2 Row Transformation method

Elementary row transformation are:
(a) interchange any two rows
(b) multiply (or divide) every element in a row by a non-zero scalar (con-
stant) k
(c) add to (or subtract from) all the elements of any row k times the corre-
sponding elements of any other row.
Equivalent Matrices: Two matrices A and B are said to be equivalent if B
can be obtained from A by a sequence of elementary transformations.
Theorem: Elementary operations do not change the rank of a matrix.
Corollary 1: Equivalent matrices have the same rank
Corollary 2: The rank of A′ is equal the rank of A
Solution of equations:
Example: ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
⎜
2 1 1 ⎟ ⎜
x 1 ⎟ ⎜
5 ⎟
The system ⎜ ⎜ 1
⎝
3 2 ⎟ ⎜ x2 ⎟ = ⎜ 1 ⎟
⎟ ⎜
⎠ ⎝
⎟
⎠
⎜
⎝
⎟
⎠
⎛
3 −2 4 ⎞ ⎛
x 3 ⎞ ⎛
4 ⎞ ⎛ ⎞
⎜
2 1 1 x
⎟ ⎜ 1 ⎟ ⎜
1 0 0 ⎟ ⎜
5 ⎟
is written as ⎜
⎜
⎝
1 3 2 ⎟ ⎜ x2 ⎟ = ⎜ 0 1 0 ⎟ ⎜ 1 ⎟
⎟ ⎜
⎠ ⎝
⎟
⎠
⎜
⎝
⎟ ⎜
⎠ ⎝
⎟
⎠
3 −2 4 x3 0 0 1 4

19
 8 −2 1
⎛ ⎞ ⎛ ⎞
2 1 1 1 0 0⎟ 1 0 0 17 17 17
−10 11 3
⎜ ⎜ ⎟
and ⎜
⎜
1 3 2 0 1 0⎟⎟
is transformed to ⎜
⎜
0 1 0 17 17 17
⎟.
⎟
11 −7 −5
⎝ ⎠ ⎝ ⎠
3 −2 ⎛ 4 0 0 1⎞ ⎛ ⎞ ⎛
0 0 1⎞ ⎛17 17
⎞ 17
1 0 0 ⎟ ⎜ x1 ⎟ 8 −2 1 ⎟ 5
1
⎜ ⎜ ⎜ ⎟
We now have ⎜⎜
⎝
0 1 0⎟⎟ ⎜
⎜ x ⎟ =
⎠ ⎝ 2 ⎠
⎟
17 ⎝ −10 11 3 ⎠
⎜
⎜
⎟
⎟
⎜
⎜
⎝
1 ⎟
⎟
⎠
0 0 1 x3 11 −7 5 −4
Which gives x1 = 2, x2 = −3, x3 = 4

1.6.3 Gaussian elimination method

Example: ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
⎜
2 −3 2 x
⎟ ⎜ 1 ⎟ ⎜
9⎟
For the system ⎜⎜
⎝
3 2 −1 ⎟ ⎜ x2 ⎟ = ⎜ 4 ⎟
⎟ ⎜
⎠ ⎝
⎟
⎠
⎜
⎝
⎟
⎠
1 −4 2 x3 ⎛ 6 ⎞
⎜
2 −3 2 : 9 ⎟
We write the augmented matrix as ⎝ 3 2 −1 : 4 ⎟
⎜
⎜ ⎟
⎠
1 4 2 :6
which is reduced to upper triangular matrix by elementary transformations
to
⎛ give ⎞
1 −4 2 : 6
2 3 ⎟
⎜ ⎟
⎜ 0 1 − : −
⎜
5 5 ⎠
⎟
⎝
0 0 1 : 4 ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
1 −4 2 x
⎟ ⎜ 1 ⎟
6
2 ⎟ ⎜ 3 ⎟
⎜ ⎜ ⎟
We now rewrite the system as ⎜⎜
⎝
0 1 − ⎟ ⎜ x2 ⎟ = ⎜ − ⎟
5 ⎠ ⎝
⎟
⎠
⎜
⎝ 5 ⎠
0 0 1 x3 4
By backward substitution x3 = 4, x2 = 1, x1 = 2

1.6.4 Triangular Decomposition method: LU-decomposition

1.6.5 Cramer’s Rule

1.7 Eigenvalues and Eigenvectors of a Matrix

A non-zero vector X is called an eigenvector or characteristic vector of a
matrix A, if there is a number λ called the eigenvalue or characteristic value
or characteristic root or latent root such that AX = λX
i.e. AX = λIX, where I is a unit matrix.

20
or (A − λI)X = 0
Since X = 0, the matrix (A − λI) is singular so that its determinant |A − λI|,
which is known as the characteristic determinant of A is zero. This leads to
the characteristics equation of A as
|A − λI)| = 0 (∗)
which follows that every characteristic root λ of a matrix A is a root of its
characteristic equation, eq(*)
Example: ⎛ ⎞
5 4
If A = ⎝ ⎠ find its characteristic roots and vectors.
1 2
The⎛ characteristic
⎞ ⎛equation⎞ of A is given by |A − λI| = 0
 5 4 1 0 ⎠  5 − λ 4 

i.e. 
⎝ ⎠ −λ ⎝ = =0
 1 2 0 1   1 2 − λ 
i.e. λ2 − 7λ + 6 = 0
or (λ − 1)(λ − 6) = 0
Therefore λ1 = 1, λ2 = 6 are ⎛ the⎞eigenvalues of A.

Now the eigenvector X  1 = ⎝ x1 ⎠ of A corresponding to λ1 = 1 is given by

x2
(A − 1I)X 1 = 0
⎛⎡ ⎤ ⎡ ⎤⎞ ⎛ ⎞ ⎛ ⎞
5 4 1 0 x
⎦⎠ ⎝ 1 ⎠ = ⎝
0
i.e. ⎝⎣ ⎦ −1⎣ ⎠
⎛
1 2
⎞⎛ ⎞
0 ⎛
1 ⎞
x 2 0
4 4 x 1 0
or ⎝ ⎠⎝ ⎠ =⎝ ⎠
1 1 x2 0
i.e. 4x1 + 4x2 = 0
and x1 + x2 = 0
which yield x1 = −x2 .
If we take x1 = 1, ⎛ then⎞ x2 ⎛ = −1 ⎞
Therefore X  1 = ⎝ x1 ⎠ = ⎝ 1 ⎠
x2 −1
⎛ ⎞
2 = ⎝ x 1
Again the eigenvector X ⎠ of A corresponding to λ = 6 is given by
2
x2
(A −  = 0
⎛⎡6I)X2 ⎤ ⎡ ⎤⎞ ⎛ ⎞ ⎛ ⎞
5 4 1 0 x 1 0
i.e. ⎝⎣ ⎦ −6⎣ ⎦⎠ ⎝ ⎠ =⎝ ⎠
⎛
1 2 ⎞⎛
0 ⎞
1 ⎛
x
⎞2
0
−1 4 ⎠ ⎝ x1 ⎠ ⎝ 0 ⎠
or ⎝ =
1 −4 x2 0

21
i.e. −x1 + 4x2 = 0
and x1 − 4x2 = 0
which yield x1 = 4x2.
If we take x1 = 4,
⎛ then
⎞ x2 ⎛
= 1⎞
Therefore X 2 = ⎝ x1 ⎠ = ⎝ 4 ⎠
x2 1⎛ ⎞ ⎛ ⎞

In general we write X  1 = C1 ⎝ 1 ⎠ and X  2 = C2 ⎝ 4 ⎠, where C1 and C2

−1 1
are non-zero scalars called the normalization constants.
Normalization: The constants C1 and C2 can be fixed by normalization as
follows: ⎛ ⎞
 1† X
 1 = 1 i.e. C1 1 −1 C1 ⎝ 1 ⎠ = 1 or 2C 2 = 1 ⇒ C1 = √12 ,

X 1
⎛ ⎞
−1
hence X  1 = √1 ⎝ 1 ⎠
2 −1 ⎛ ⎞
 2X
† 4
 2 = 1 i.e. C2 4 1 C2 ⎝ ⎠ = 1 or 17C 2 = 1 ⇒ C2 = √1 ,

X 2
⎛ ⎞
1 17

hence X  2 = √1 ⎝ 4 ⎠
17 1

1.7.1 Nature of the eigenvalues and eigenvectors of special types

of matrices
Theorem 1: The eigenvalues of a Hermitian matrix are all real.
Proof: Let λ be an eigenvalue of a Hermitian matrix A. By definition there
exists a vector X  = 0, such that
AX  = λX 
⇒ X  † AX =X  † λX
 = λX  = λX
 †X  † IX


†

so that λ = XX
†AIXX


But X  † AX =X  † AX
 which implies X † AX
 is a real number and therefore
X † IX
 is also a real number. Hence λ is real.
Corollary: The eigenvalues of a real symmetric matrix are all real.
Theorem 2: The eigenvalues of a skew-Hermitian matrix are purely imagi-
nary or zero.
Corollary: The eigenvalues of a real skew-symmetric matrix are either zero
or purely imaginary.

22
Theorem 3: The modulus of each eigenvalue of a unitary matrix is unity, i.e.
the eigenvalues of a unitary form have absolute value 1
Corollary: The eigenvalues of an orthogonal matrix have the absolute value
unity and are real, or complex conjugate in pairs.
Theorem 4: Any two eigenvectors corresponding to two distinct eigenvalues
of a Hermitian matrix are orthogonal.
Proof: Let X  1 and X  2 be two eigenvectors corresponding to two distinct
eigenvalues λ1 and λ2 of a Hermitian matrix A; then
AX  1 = λ1 X  1 ...............................(1)
AX  2 = λ2 X  2 ...............................(2)
From theorem 1 λ1 and λ2 are real. Premultiplying (1) and (2) by X  2† and
X 1† respectively
X 2† AX  1 = λ1 X  2† X
 1 ...............................(3)
X 1 AX
†  2 = λ2 X  1X
†  2 ...............................(4)
But (X  2† AX  1 )† = X  1† AX 2
Therefore for a Hermitian matrix A† = A and also (X  2† )† = X
 2 ; therefore we
have, from (3) and (4),
(λ1X  2† X
 1 )† = λ2 X  1† X
2
or λ1 X  1† X 2 = λ2 X  1† X
2
or (λ1 − λ2 )X  1† X2 = 0
Since λ1 − λ2 = 0 for distinct roots we have X 2 = 0 ⇒ X
 1† X  1 and X 2 are
orthogonal
Corollary: Any two eigenvectors corresponding to two distinct eigenvalues of
a real symmetric matrix are orthogonal.
Theorem 5: Any two eigenvectors corresponding to two distinct eigenvalues
of a unitary matrix are orthogonal.
Theorem 6: The eigenvectors corresponding to distinct eigenvalues of a ma-
trix are linearly independent.
Theorem 7: The characteristic polynomial and hence the eigenvalues of sim-
ilar matrices are the same. Also if X  is an eigenvector of A corresponding
to the eigenvalue λ, then P−1X  is an eigenvector of B corresponding to the
eigenvalue λ where B = P−1AP
Proof: Let A and B be two similar matrices. Then there exists an invertible
matrix P such that B = P−1AP. Consider
B − λI = P−1AP − λI
= P−1(A − λI)P

23
since P−1(λI)P = λP−1P = λI
Therefore

|B − λI| = |P−1|.|A − λI|.|P|

= |P−1||P|.|A − λI| since scalar quantities commute
= |P−1P|.|A − λI| where |CD| = |C||D|
= |A − λI| where |P−1 P| = |I| = 1

which follows that A and B have the same characteristic polynomial and so
they have the same eigenvalues.
Corollary 1: The eigenvalues of a matrix are invariant under similarity trans-
formation.
Corollary 2: If A is similar to a diagonal matrix D then the diagonal elements
of D are the eigenvalues of A.

1.7.2 Diagonalisation of a matrix

Modal matrix: If the eigenvectors of a matrix A are arranged as columns
of a square matrix, the modal matrix of A denoted by M, is formed.
Example: ⎛ ⎞
⎜
1 0 4 ⎟
For A = ⎝ 0 2 0 ⎟
⎜
⎜ ⎟ , λ1 = 2, λ2 = 3, λ3 = −5 and the corresponding
⎠
3 1 −3 ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
4 2 2
1 = ⎜ 2 = ⎜ 3 = ⎜
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
eigenvectors are X ⎜ −7 ⎟, X
⎝
⎟
⎠
⎜ 0 ⎟, X
⎝
⎟
⎠
⎜ 0 ⎟
⎝
⎟
⎠
1 ⎛ 1 ⎞ −3
⎜
4 2 2 ⎟
Then the modal matrix M = ⎜ ⎜
⎝
−7 0 0 ⎟ ⎟
⎠
1 1 −3
Spectral matrix: Also we define the spectral matrix of A, i.e. S, as a
diagonal
⎛ matrix ⎞with the eigenvalues only on the main diagonal
⎜
2 0 0 ⎟
S=⎜ ⎜
⎝
0 3 0 ⎟ ⎟
⎠
0 0 −5
Note that the eigenvalues of S and A are the same as indicated in corollary
2 above.

24
Show that AM = MS
⇒ M−1AM = S (i.e. similarity transformation of A to S which implies
that S and A are similar matrices)
Note
1. M−1AM transforms the square matrix A into a diagonal matrix S
2. A square matrix A of order n can be so transformed if the matrix has n
independent eigenvectors.
3. A matrix A always has n linearly independent eigenvectors if it has n
distinct eigenvalues or if it is a symmetric matrix.
4. If the matrix has repeated eigenvalues, it may or may not have linearly
independent eigenvectors

1.8 Transformation
n
Linear form: An expression of the form j=1 aij xj is said to be linear form
of the variable xj .

1.8.1 Transformation
If aij are the given constants and xj the variables then the set of equation
yi = nj=1 aij xj (for i = 1, 2, 3 . . . , n) (1)


is called a linear transformation connecting the variables xj and the variables

yi . The square matrix
⎛ ⎞
⎜
a11 a12 . . . a1n ⎟
⎜
a21 a22 . . . a2n ⎟
A=
⎜ ⎟
⎜ ⎟
⎜
⎜
⎝
... ......... ⎟
⎟
⎠
an1 an2 . . . ann
is said to be the matrix of transformations. The determinant of the matrix
|A| is said to be the determinant or modulus of transformation. For short
eq.(1) is written as y = Ax
When |A| = 0 the transformation is called singular and when |A| =  0 the
transformation is said to be non-singular. For non-singular matrices the
transformation may be expressed as x = A−1y

25
If A is an identity matrix I then we have the identical transformation and
its determinant is unity. In this case y1 = x1, y2 = x2, y3 = x3, . . . , yn = xn

1.8.2 Resultant of two linear transformation

Let two successive transformations be y = Px and z = Qy , then the resultant
transformation is z = QPx

1.8.3 Similarity transformation

If A, B are two non-singular matrices and there exist two non-singular ma-
trices P and Q such that B = QAP with Q = P−1 so that
B = P−1AP (2)
then the transformation of A into B is termed similarity transformation and
the matrices A and B are known as similar matrices.
A matrix equation Ax = B preserves its structure (or form) under similarity
transformation, P−1(Ax)P = P−1 BP
⇒ P−1APP−1xP = P−1BP as PP−1 = I
⇒ (P−1AP)(P−1xP) = P−1 BP
⇒ Cy = D, (where C = P−1AP, y = P−1xP and D = P−1BP). By eq.(2)
this is of the form Ax = B
The trace of a matrix is invariant under similarity transformation.

1.8.4 Unitary transformation

Let A be a unitary matrix of order n × n and y , x are column vectors of
order n × 1, then the linear transformation y = Ax is known as unitary
transformation. Since A† A = AA† = I
Therefore y†y = (Ax)†(Ax) = x† A†Ax = x†x
⇒ the norm of vectors is invariant under similarity transformation. In
eq.(2), if P is unitary, i.e., P† P = PP† = I, i.e. P† = P−1 then the
transformation B = P−1AP is also unitary.

1.8.5 Orthogonal transformation

Any transformation y = Ax that transforms y 2 into x2 is said to be an or-
 

thogonal transformation and the matrix A is known as an orthogonal matrix.

The necessary and sufficient condition for a square matrix to be orthogonal

26
is AA′ = I. In eq.(2) if P is orthogonal, then P−1 = P′ and eq.(2) is an
orthogonal transformation. The product of two orthogonal transformations
is an orthogonal transformation. Two n-vectors x and y are orthogonal to
each other if x.y = x, y = 0 i.e. if x†y = 0.

1.8.6 Orthogonal set

A set of vectors is said to be orthonormal if:
• each vector of the set is a normal vector
• any two vectors of the set are orthogonal

1.9 Bases and dimension

Let X be a linear space over K. Every subset of a linearly independent subset
of X is automatically linearly independent. X possesses a maximal linearly
independent subset called a basis of X. The cardinality of a basis is an invari-
ant of X, since any two bases possess the same cardinality. This invariant is
called the dimension of X. If the dimension of X is finite, then X is called a
finite dimensional linear space, otherwise X is an infinite dimensional linear
space.

1.9.1 Linear Maps

Let X, Y be two linear spaces over K and D a subspace of X. A transfor-
mation T : D → Y is called a linear map or a linear operator if
T (αx1 + βx2) = αT (x1) + βT (x2)
where x1 , x2 ∈ D and α, β ∈ K. The set D is called the domain of T and
is often written as D(T ). The set R(T ) = {T x : x ∈ D(T )} is the range or
codomain of T . R(T ) is automatically a linear space over K

27
Chapter 2

FUNCTIONAL ANALYSIS

2.1 Normed Spaces

A normed space over a field K is a pair (X,
·
) consisting of a linear space
X over K and a map
·
: X → ℜ with the following properties:
(i)
x
≥ 0∀x ∈ X i.e.
·
is non-negative
(ii)
x
= 0 implies x = 0 (i.e. ker(
·
) = {0})
(iii)
λx
= |λ|
x
for all x ∈ X, λ ∈ K (i.e.
·
is positively
homogeneous)
(iv)
x +y
≤
x
+
y
for all x, y ∈ X (i.e.
·
satisfies the triangular
inequality).
For x ∈ X, the number
x
is called the norm of x

2.1.1 Cauchy Sequences

Let (X,
·
) be normed space and xn a sequence of number of X. Then xn
is a Cauchy sequence if given any ε > 0, there is natural number N (ε), such
that
xn − xm
< ε whenever n, m > N (ε)

2.1.2 Completeness
A normed space in which every Cauchy sequence has a limit is said to be
complete.

28
2.1.3 Pre-Hilbert spaces
An inner product space or a pre-Hilbert space is a pair (H, ·, · ) consisting
of a linear space H over K and a functional ·, · : H × H → K, called the
inner product of H, with the following properties:
(i) x + y, z = x, z + y, z , ∀x, y, z ∈ H
(ii) αx, y = αx, y , ∀x ∈ H, α ∈ K
(iii) x, y = y, x , ∀x, y ∈ H
(iv) x, x ≥ 0, ∀x ∈ H and x, x = 0 iff x = 0.
Remark
1. For x, y ∈ H the number x, y is called the inner product of x and y.

2. For x ∈ H, define
x
by
x
= x, x , x ∈ H Then,
·
is a
norm on H, whence (H,
·
) is a normed space.
·
is called the norm
induced by the inner product ·, ·
3. With
·
as in 2., one can show that

x + y
2 +
x − y
2 = 2
x
2 +2
y
2 for all x, y ∈ H.
This result is called the parallelogram law and is a characterizing property
of prer-Hilbert spaces, i.e. if a norm does not satisfy the parallelogram
law, then it is not induced by an inner product

2.1.4 Hilbert Spaces

A Hilbert space is a pre-Hilbert space (H, ·, · ) such that the pair (H,
·
),
where
·
is the norm induced by ·, · , is a complete normed space
Example of Hilbert space: Define ·, · : K n × K n → K by x, y = nj=1 x̄j yj


where x = (x1, x2, . . . , xn), y = (y1, y2, . . . , yn) Then (K n, ·, · ) is a Hilbert
space finite dimension.

2.1.5 Geometry of Hilbert space

29
Chapter 3

SPECIAL FUNCTIONS

3.1 The gamma and beta functions

3.1.1 The gamma function Γ
The gamma function Γ(x) is defined by the integral
Γ(x) = 0∞ tx−1 e−t dt (1)


and is convergent for x > 0. It follows from eq.(1) that

Γ(x + 1) = 0∞ tx e−t dt


Integrating by parts

∞
x e−t
Γ(x + 1) = t ( −1 ) + x 0∞ tx−1 e−t dt

0
Γ(x + 1) = xΓ(x) (2)
This is a fundamental recurrence relation for gamma functions. It can also
be written as Γ(x) = (x − 1)Γ(x − 1).
A number of other results can be derived from this as follows:
If x = n, a positive integer, i.e. if n ≥ 1, then

Γ(n + 1) = nΓ(n).
= n(n − 1)Γ(n − 1) since Γ(n) = (n − 1)Γ(n − 1)
= n(n − 1)(n − 2)Γ(n − 2) since Γ(n − 1) = (n − 2)Γ(n − 2)
= ....................................................
= n(n − 1)(n − 2)(n − 3) . . . 1Γ(1)
= n!Γ(1)
∞
But Γ(1) = 0∞ t0 e−t dt = [−e−t ]0 = 1


⇒ Γ(n + 1) = n! (3)
Examples:

30
Γ(7) = 6! = 720, Γ(8) = 7! = 5040, Γ(9) = 40320
We can also use the recurrence relation in reverse
Γ(x + 1) = xΓ(x) ⇒ Γ(x) = Γ(x+1) x
Example:
If Γ(7) = 720 then Γ(6) = Γ(6+1) 6 = Γ(7) 720
6 √= 6 = 120
If x = 21 it can be shown that Γ( 12 ) = π (F.E.M. 150)

Using the recurrence relation Γ(x + 1)√= xΓ(x) we can obtain the following:
√
Γ( 23 ) = 21 Γ( 12 ) = 21 ( √ π) ⇒ Γ( 32 ) = √2π
Γ( 52 ) = 23 Γ( 32 ) = 23 ( 2π ) ⇒ Γ( 52 ) = 3 4 π
Negative values of x
Since Γ(x) = Γ(x+1) x , then as x → 0, Γ(x) → ∞ ⇒ Γ(0) = ∞
The same result occurs for all negative integral values of x
Examples:
At x = −1, Γ(−1) = Γ(0) −1 = ∞
Γ(−1)
At x = −2, Γ(−2) = −2 = ∞ etc.
Γ( 1 ) √
Also at x = − 12 , Γ(− 12 ) = −21 = −2 π
2
Γ(− 21 ) √
3
and at x = − 2 , Γ(− 2 ) = − 3 = 34 π
3
2
Gragh of y = Γ(x)

Examples:
1. Evaluate 0∞ x7e−xdx


Γ(x) = 0∞ tx−1 e−t dt



Let Γ(v) = 0∞ xv−1e−xdx ⇒v=8


 ∞ 7 −x
i.e. 0 x e dx = Γ(8) = 7! = 5040
2. Evaluate 0∞ x3e−4xdx


Since Γ(v) = 0∞ xv−1e−xdx we use the substitution y = 4x ⇒ dy = 4dx



⇒ I = 414 0∞ y 3 e−y dy = 414 Γ(v) where v = 4 ⇒ I = 128

3

1 2
3. Evaluate 0∞ x 2 e−x dx


Use y = x2 therefore dy = 2xdx. Limits x = 0, y = 0 x = ∞, y = ∞

31
 1 1 1
x = y 2 ⇒ x2 = y 4
1
4 e−y  ∞ y 41 e−y 1  ∞ − 41 −y
I = 0∞ y 2x dy = 1 dy = 2 0 y e dy = 1  ∞ v−1 −y
2 0 y e dy where

0
2y 2
v = 43 ⇒ I = 12 Γ( 34 )
From tables, Γ(0.75) = 1.2254 ⇒ I = 0.613

3.1.2 The beta function, β

The beta function β(m, n) is defined by β(m, n) = 01 xm−1(1 − x)n−1dx


It can be shown that the beta function and the gamma function are related
as β(m, n) = Γ(m)Γ(n) (m−1)!(n−1)!
Γ(m+n) = (m+n−1)!

3.1.3 Application of gamma and beta functions

Examples:
1. Evaluate I = 01 x5(1 − x)4dx


Comparing this with β(m, n) = 01 xm−1(1 − x)n−1dx



then m − 1 = 5 ⇒ m = 6 and n − 1 = 4 ⇒ n = 5
I = β(6, 5) = 5!4!
10! = √1
 1 41260
2.Evaluate I = 0 x 1 − x2dx
Comparing this with β(m, n) = 01 xm−1(1 − x)n−1dx


we see that we have x2 in the root, instead of a single x. Therefore, put

1 1
x2 = y ⇒ x = y 2 and dx = 12 y − 2 dy
The limits remain unchanged.
1 1 3 1
I = 01 y 2 (1 − y) 2 12 y − 2 dy = 12 01 y 2 (1 − y) 2 dy
 

m − 1 = 32 ⇒ m = 52 and n − 1 = 21 ⇒ n = 32
3√ 1√
Γ( 5 )Γ( 3 ) 1 ( 4 π)( 2 π)
Therefore, I = 12 β( 52 , 23 ) = 21 Γ(25 + 32) = 2 3! = π
32
2 2
 3 x3 dx
3. Evaluate I = 0 3−x √ (F.E.M. 170)

3.2 Bessel’s Functions

Bessel’s functions are solutions of the Bessel’s differential equation
2
2dy dy
x 2
+ x + (x2 − v 2)y = 0 (1)
dx dx
where v is a real constant.
By the Frobenius method we assume a series solution of the form

32
 ∞
c 2 3 r c 
y = x (a0 + a1 x + a2 x + a3 x + . . . + ar x + . . .) or y = x ar xr
r=0
∞
i.e. y = a0 xc + a1 xc+1 + a2 xc+2 + . . . + ar xc+r + . . . or y = ar xc+r

(2)
r=0
where c, a0 , a1 , a2, . . . , ar are constants. a0 is the first non-zero coefficient. c
is called the indicial constant.
dy
= a0 cxc−1 + a1 (c + 1)xc + a2 (c + 2)xc+1 + . . . + ar (c + r)xc+r−1 + . . . (3)
dx
d2 y
2
= a0 c(c−1)xc−2+a1 c(c+1)xc−1+a2 (c+1)(c+2)xc+. . .+ar (c+r−1)(c+r)xc+r−2+. . .
dx
(4)
Substituting eqs.(2),(3) and (4) into (1) and equating coefficients of equal
powers of x, we have c = ±v and a1 = 0.
ar−2
The recurrence relation is ar = v2 −(c+r)2 for r ≥ 2.

It follows that a1 = a3 = a5 = a7 = . . . = 0 so that when c = +v

a2 = 22−a 0
(v+1)
a0
a4 = 24×2!(v+1)(v+2)
−a0
a6 = 26×3!(v+1)(v+2)(v+3)
r
(−1) 2 a0
ar = 2r × r2 !(v+1)(v+2)...(v+ 2r )
for r even. The resulting solution is
x2 x4 x6
⎧ ⎫
⎨ ⎬
v
y1 = a0 x ⎩1 − 2 + 4 − 6 + . . .⎭
2 (v + 1) 2 × 2!(v + 1)(v + 2) 2 × 3!(v + 1)(v + 2)(v + 3)

This is valid provided v is not a negative integer.

Similarly, when c = −v
x2 x4 x6
⎧ ⎫
−v
⎨ ⎬
y2 = a0 x 1 + + + + . . .
⎩ 22(v − 1) 24 × 2!(v − 1)(v − 2) 26 × 3!(v − 1)(v − 2)(v − 3) ⎭

This is valid provided v is not a positive integer. The complete solution is

y = Ay1 + By2

with the two arbitrary constants A and B.

Besssel’s functions:

33
 1
Let a0 = 2v Γ(v+1) then the solution y1 gives for c = v = n (where n is a
positive integer) Bessel’s functions of the first kind of order n denoted by
Jn (x) where
x2 x4
⎧ ⎫
x ⎨ 1
Jn(x) = ( )n ⎩
⎬
− 2 + 4 − . . .⎭
2 Γ(n + 1) 2 (1!)Γ(n + 2) 2 (2!)Γ(n + 3)
x  ∞ (−1)k x2k
= ( )n
2 k=0 22k (k!)Γ(n + k + 1)
x  ∞ (−1)k x2k
= ( )n
2 k=0 22k (k!)(n + k)!
Similarly for c = −v = −n (a negative integer)
x2 x4
⎧ ⎫
x 1
( )−n ⎩
⎨ ⎬
J−n(x) = − 2 + 4 − . . .⎭
2 Γ(1 − n) 2 (1!)Γ(2 − n) 2 (2!)Γ(3 − n)
x ∞ (−1)k x2k
( )−n

= 2k
2 k=0 2 (k!)Γ(k − n + 1)
x  ∞ (−1)k x2k
= (−1)n( )n (for details see F.E.M. 247)
2 k=0 22k (k!)(n + k)!
= (−1)nJn(x)

⇒ The two solutions Jn(x) and J−n(x) dependent on each other. Further
more the series for Jn(x) is
⎧ ⎫
x ⎨1 1 x 1 x
Jn (x) = ( )n ⎩ − ( )2 + ( )4 − . . .⎭
⎬

2 n! (n + 1)! 2 (2!)(n + 2)! 2

From this we obtain two commonly used functions
1 x 2 1 x 4 1 x 6
J0(x) = 1 − 2
( ) + 2
( ) − ( ) + ...
(1!) 2 (2!) 2 (3!)2 2
⎧ ⎫
x⎨ 1 x 1 x
( )2 + ( )4 + . . .⎭
⎬
J1(x) = ⎩1 −
2 (1!)(2!) 2 (2!)(3!) 2
Graphs of Bessel’s functions J0 (x) and J1(x)

34
Remark: Note that J0(x) and J1(x) are similar to cos x and sin x respec-
tively.
Generating function: If we want to study a certain sequence {f (x)} and
can find a function G(t, x) = ∞ n
n=0 fn (x)t we may obtain the properties of


{fn(x)} from those of G which ”generates” this sequence and is called a gen-
erating function of it.
x 1
The generating function for Jn (x) is e 2 (t− t ) = ∞
−∞ Jn (x)t
n


Recurrence formula: Jn(x) can also be obtained from the recurrence for-
mula → Jn+1(x) = 2n x [Jn (x) − Jn−1 (x)]
For (0 < x < 1) Jn (x) are orthogonal

3.3 Legendre’s Polynomials

These are solutions of the Legendre’s differential equation

2 d2 y dy
(1 − x ) 2 − 2x + k(k + 1)y = 0
dx dx
where k is a real constant. Solving it by the Frobenius method as before we
obtain c = 0 andc = 1 and the corresponding solutions  are
k(k+1) 2 k(k−2)(k+1)(k+3) 4
a) c = 1 : y = a0 1 − 2! x + 4! x −...
 
b) c = 0 : y = a1 x − (k−1)(k−2)
3! x3 + (k−1)(k−3)(k+2)(k+4)
5! x5 − . . .
where a0 and a1 are the usual arbitrary constants. When k is an integer
n, one of the solution series terminates after a finite number of terms. The
resulting polynomial in x denoted by Pn (x) is called Legendre polynomial with
a0 and a1 being chosen so that the polynomial has unit value when x = 1.
(−1 < x < 1) orthogonality
e.g. P0 (x) = a0 {1 − 0 + 0 − . . .} = a0 . We choose a0 = 1 so that P0 (x) = 1
P1 (x) = a1 {x − 0 + 0 − . . .} = a1 x
a1 is then chosen
 to make P1 (x) =1 when x = 1 ⇒ a1 = 1 ⇒ P1 (x) = x
2x3 2
p2(x) = a0 1 − 2! x + 0 + 0 + . . . = a0 {1 − 3x2}
If P2 (x) = 1 when x = 1 then a0 = − 12 ⇒ P2 (x) = 12 (3x2 − 1)
Using the same procedure obtain:

35
P3 (x) = 12 (5x3 − 3x)
P4 (x) = 81 (35x4 − 30x2 + 3)
P5 (x) = 81 (63x5 − 70x3 + 15x) etc.
Legendre polynomials can also be expressed by Rodrigue’s formula given by
1 dn 2
Pn (x) = n n
(x − 1)n
2 n! dx
(Use this formula to obtain P0 (x), P1(x), P2(x), P3(x), etc)
The generating function is
1 ∞
√ Pn (x)tn

2
=
1 − 2xt + t n=0
1
To show this, start from the binomial expansion of √1−v where v = 2xt − t2,
multiply the powers of 2xt − t2 out, collect all the terms involving tn and
verify that the sum of these terms is Pn (x)tn.
The recurrence formula for Legendre polynomials is
2n + 1 n
Pn+1(x) = xPn (x) − Pn−1 (x)
n+1 n+1

This means that if we know Pn−1(x) and Pn (x) we can calculate Pn+1 (x),
e.g. given that P0 (x) = 1 and P1 (x) = x we can calculate P2 (x) using the
recurrence formula by taking Pn−1 = P0 , Pn = P1 and Pn+1 = P2 ⇒ n = 1.
Substituting these in the formula,
P2 (x) = 2×1+1 1 1 2
1+1 xP1 (x) − 1+1 P0 (x) = 2 (3x − 1)
Similarly to find P3 (x) we set Pn−1 = P1 , Pn = P2 and Pn+1 = P3 where
n = 2. Substituting these in the formula we have
P3 (x) = 2×2+1 2
2+1 xP2 (x) − 2+1 P1 (x)
= 35 x × 21 (3x2 − 1) − 32 x
1 3
= 2 (5x − 3x)
(Using the recurrence formula obtain P4 (x) and P5 (x))

3.4 Hermite Polynomials

They are solutions of the Hermite differential equation
d2 y dy
− 2x + 2vy = 0 (∗)
dx2 dx
36
where v is a parameter. Using the Frobenius method the solution is
∞ c+r
y = r=0 ar x , where a0 = 0
dy ∞ c+r−1
dx
= r=0 ar (c + r)x and
2
d y ∞ c+r−2
dx2 = r=0 ar (c + r)(c + r − 1)x


Substituting these in eq.(*) and equating coefficients of like terms we have

a0 c(c − 1) = 0 ⇒ c = 0, or c = 1
2(c+r−v)ar
and ar+2 = (c+r+2)(c+r+1) ⇒ a1 = a3 = a5 = . . . = 0
When c= 0 (M.P. by Gupta 8.94), 
2v 2 22 v(v−2) 4 (−1)r 2r v(v−2)...(v−2r+2) 2r
y1 = a0 1 − 2! x + 4! x − . . . + 2r! x + ...
(where a1 = 0). When c=1, 
2(v−1) 2 22 (v−1)(v−3) 4 (−1)r 2r (v−1)(v−3)...(v−2r+1)
y2 = a0 x 1 − 2! x + 4!
x −...+ (2r+1)!
+...
The complete

solution of eq.(*) isthen given 
by y = Ay1 + By2 i.e. 
2v 2 22 v(v−2) 4 2(v−1) 2 22 (v−1)(v−3) 4
y = A 1 − 2! x + 4! x − . . . + Bx 1 − 3! x + 5! x −...
where A and B are arbitrary constants. When v = n, an integer, the series
terminates after a few terms. The resulting polynomials Hn (x) are called
Hermite polynomials. The first 5 of them are:
H0 (x) = 1
H1 (x) = 2x
H2 (x) = 4x2 − 2
H3 (x) = 8x3 − 12x
H4 (x) = 16x4 − 48x2 + 12
H5 (x) = 32x5 − 160x3 + 120x
They can also be given by a corresponding Rodrigue’s formula
dn −x2
x2 n
Hn (x) = e (−1) (e )
dxn
The generating function is given by
2
∞ Hn (x) n
e2tx−t =

t
n=0 n!
This can be proved using the formula for the coefficients of a Maclaurin series
and noting that tx − 12 t2 = 12 x2 − 12 (x − t)2
Hermite polynomials satisfy the recursion formula
Hn+1(x) = 2xHn(x) − 2nHn−1(x)
(Given that H0 = 1 and H1 = 2x use this formula to obtain H2 , H3, H4 and
H5 ).

37
3.5 Laguerre Polynomials
They are solutions of the Laguerre differential equation
d2 y dy
x dx 2 + (1 − x) dx + vy = 0 (∗1)
Using the Frobenius method again we have
∞ c+r
y = r=0 ar x , where a0 = 0
dy ∞ c+r−1
dx = r=0 ar (c + r)x and
2
d y ∞ c+r−2
dx2 = r=0 ar (c + r)(c + r − 1)x


Substituting these in eq.(*1) and equating coefficients of like terms we have

c2 = 0 ⇒ c = 0
c+r−v r−v
and ar+1 = (c+r+1) 2 ar = (r+1)2 ar
 
y = a0 1 − vx + v(v−1) 2
(2!)2 x − . . . + (−1)
r v(v−1)...(v−r+1) r
(r!)2 x + ... (∗2)
In case v = n (a positive integer) and a0 = n! the solution eq.(*2) is said to
be the Laguerre polynomial of degree n and is denoted by Ln(x) i.e.
n2 n2 (n − 1)2 n−2
⎧ ⎫

Ln(x) = (−1)n ⎩xn − xn−1 + + . . . + (−1)nn!⎭ (∗3)

⎨ ⎬
x
1! 2!
Then the solution of Laguerre equation for v to be a positive integer is
y = ALn(x)
From eq.(*3) it is easy to show that
L0(x) = 1
L1(x) = 1 − x
L2(x) = x2 − 4x + 2
L3(x) = −x3 + 9x2 − 18x + 6
L4(x) = x4 − 16x3 + 72x2 − 96x + 48
They can also be given by the Rodrigue’s formula
n
x d
Ln (x) = e n
(xne−x)
dx
Their generating function is
xt ∞ Ln (x) n
e− (1−t) =

t
n=0 n!
They satisfy the recursion formula
Ln+1 = (2n + 1 − x)Ln(x) − n2 Ln−1(x)
They are orthogonal for 0 < x < ∞

38
 3.5.1 Hypergeometric Function
The solutions of Gauss’s hypergeometric differential equation

x(1 − x)y ′′ + [γ − (α + β + 1)x]y ′ − aby = 0 (1)

where α, β and γ are parametric constants, and by choosing specific values

we can obtain a large number of elementary and higher functions as solutions
of (1). This accounts for the practical importance of (1)
a) Hypergeometric series
Using the Frobenius method the indicial equation of (1) has roots c = 0 and
c = 1 − γ; ar+1 = (c+r+β)(c+r+α)
(c+r+1)(c+r+γ) ar
(α+r)(β+r)
For c1 = 0, ar+1 = (r+1)(r+γ) ar and the Frobenius method gives
 
αβ
y1 (x) = a0 1 + 1!γ x + α(α+1)β(β+1)
+ α(α+1)(α+2)β(β+1)(β+2)
2!γ(γ+1) x
2
3!γ(γ+1)(γ+2) x3 + . . . (2)
where γ = 0, −1, −2, . . .. This series is called the hypergeometric series.
Its sum y1 (x) is denoted by F (α, β, γ; x) and is called the hypergeometric
function.
b) Special cases: It can be shown that
1
(i) 1−x = F (1, 1, 1; x) = F (1, β, β; x) = F (α, 1, α; x)
(ii) (1 + x)n = F (−n, β, β; −x) or F (−n, 1, 1; −x)
(iii) (1 − x)n = 1 − nxF (1 − n, 1, 2; x)
(iv) arctan x = xF ( 21 , 1, 32 ; −x2)
(v) arcsin x = xF ( 12 , 12 , 23 ; x2)
(vi) ex =lim F (n, 1, 1; nx )
n→∞

(vii) ln(1 + x) = xF (1, 1, 2; −x)

1+x
(viii) ln 1−x = 2xF ( 12 , 1, 23 ; x2) etc
c) Second solution
For c2 = 1 − γ, ar+1 = (α+r+1−γ)(β+r+1−γ)
(2−γ+r)(r+1) and the Frobenius method yields
the following solution
 (where γ = 2, 3, 4, . . .) 
1−γ (α−γ+1)(β−γ+1) (α−γ+1)(α−γ+2)(β−γ+1)(β−γ+2) 2
y2 (x) = a0 x 1+ 1!(−γ+2) x+ 2!(−γ+2)(−γ+3) x +...

39
which gives y2 (x) = x1−γ F (α − γ + 1, β − γ + 1, 2 − γ; x). The complete
solution is

y = Ay1(x) + By2(x) = AF (α, β, γ; x) + Bx1−γ F (α − γ + 1, β − γ + 1, 2 − γ; x)

40
Chapter 4

INTEGRAL TRANSFORM AND

FOURIER SERIES

4.1 Laplace Transform

LT is used in solving ordinary differential equations (ode). It has the following
advantages:
• Solution of the ode is obtained by algebraic processes.
• the initial conditions are involved from the early stages so that the de-
termination of the particular solution is considerably shortened.
• the method enables us to deal with situations where the function is
discontinuous.
The LT of a function f (t) is denoted by L{f (t)} or F (s) and is defined by
the integral 0∞ f (t)e−stdt


i.e. L{f (t)} or F (s) = 0∞ f (t)e−stdt



where s is a positive constant such that f (t)e−st converges as t → ∞

Examples
1. To find the LT of a constant function f (t) = a
−st ∞
 
∞
L{a} = 0 ae−st dt= a e−s = − as [e−st ]0 = − as [0 − 1] = a
∞
0 s

a
 
 
⇒ 
 L{a} = 
 (1)
 s 

1
e.g. for a = 1, L{1} = s

41
2. If f (t) = eat  ∞
e−(s−a)t 1 1
L{eat }= 0∞ eat e−st dt=
 ∞ −(s−a)t
0 e dt= −(s−a) 0 = − s−a [0 − 1] =

s−a

1
 
 
⇒ 
 L{eat } = 
 (2)
 s−a 
1
 
 
Similarly 
 L{e−at } = 
 (3)
 s+a 

3. If f (t) = sin at

eiat − e−iat ⎠ −st

⎛ ⎞
 ∞  ∞
L{sin at} = sin(at)e−stdt = ⎝ e dt
0 0 2i
1  ∞ −(s−ia)t  ∞ 
= e dt − e−(s+ia)t dt
2i  0 0
1 1 1
= −
2i s − ia s + ia
1 1 1
= +
2 a + is a − is
1 a − is + a + is
=
2 (a + is)(a − is)
a
=
s + a2
2

a
 
 
⇒ 
L{sin at} = 2 
(4)
s + a2
 
 

2
e.g. L{sin 2t} = s2 +4

Similarly (show that):

4. if f (t) = cos at
s
 
 
⇒ 
L{cos at} = 2 
(5)
s + a2
 
 
s
e.g. L{cos 4t} = s2 +16

5. if f (t) = tn
n!
 
 
⇒ 
L{tn } = 
(6)
sn+1
 
 

3! 6
e.g. L{t3 } = s3+1 = s4 .

42
 6. If f (t) = sinh at
eat − e−at ⎠ −st
⎛ ⎞
 ∞  ∞
L{sinh at} = sinh(at)e−st dt = ⎝ e dt
0 0 2
1  ∞ −(s−a)t  ∞ 
= e dt − e−(s+a)t dt
2 0 0
1 1 1
= −
2 s−a s+a
a
= 2
s − a2
a
 
 
⇒ 
L{sinh at} = 2 
(7)
s − a2
 
 

e.g. L{sinh 2t} = s22−4

Similarly (show that)
7. if f (t) = cosh at = 12 (eat + e−at )
s
 
 
⇒ 
L{cosh at} = 2 
(8)
s − a2
 
 

s 4s
e.g. L{4 cosh 3t} = 4 s2 −3 2 = s2 −9

Existence theorem for Laplace Transforms

Let f (t) be a function that is piecewise continuous on every finite interval
in the range t ≥ 0 and satisfies |f (t)| ≤ Me−kt for all t ≥ 0 and for some
constants k and M. Then the LT of f (t) exists for all s > k
A function f (t) is said to be piecewise continuous in an interval (a, b) if
(i) the interval can be divided into a finite number of subintervals in each
of which f (t) is continuous.
(ii) the limits of f (t) as t approaches the endpoints of each subinterval are
finite.
In other words a piecewise continuous function is one that has a finite number
of finite discontinuities. e.g see fig.

43
4.1.1 Inverse Transform
Given a LT, F (s) one can find the function f (t) by inverse transform
f (t) = L!−1 {F"(s)} where L−1 indicates inverse transform.
1
e.g. L−1 s−2 = e2t
L−1 ! 4s ="4
! "

s
L−1 s2 +25 = cos 5t
3s + 1 1 2
# $ # $
−1 −1
L = L + (by partial fractions)
s2 − s − 6 s + 2 s − 3
1 2
# $ # $
−1 −1
= L +L
s+2 s−3
−1
(Note: L, L are linear operators. Prove it)
= e−2t + 2e3t

Rules of Partial Fractions

1. The numerator must be of lower degree than the denominator. If it is
not then we first divide out
2. Factorize the denominator into its prime factors. These determine the
shapes of the partial fractions.
A
3. A linear factor (s + a) gives a partial fraction s+a where A is a constant
to be determined.
A B
4. A repeated factor (s + a)2 gives s+a + (s+a)2

A B C
5. Similarly (s + a)3 gives s+a + (s+a)2 + (s+a)3

As+B
6. Quadratic factor (s2 + ps + q) gives s2 +ps+q
As+B Cs+D
7. repeated quadratic factor (s2 + ps + q)2 gives s2 +ps+q + (s2 +ps+q)2

Examples
s2 −15s+41 3 2 1
1. (s+2)(s−3) 2 = s+2 − s−3 + (s−3)2

4s2 −5s+6
2. L−1 (s+1)(s 2 +4)

4s2 −5s+6 A 3 3 6
but (s+1)(s2 +4) ≡ s+1 + Bs+C
s2 +4 = s+1 + s−6
s2 +4 = s+1 + s
s2 +4 − s2 +4

44
  
−1 4s2 −5s+6 −1 3 s 6
! "
⇒ f (t) = L =L
(s+1)(s2 +4) s+1 + s2 +4 − s2 +4
−t
⇒ f (t) = 3e + cos 2t − 3 sin 2t

PROPERTIES OF LAPLACE TRANSFORM

Linearity: L{af (t) + bg(t)} = aL{f (t)} + bL{g(t)} (Prove!)
(1) The first shift theorem (or s-shifting):
It states that if L{f (t)} = F (s) then
⇒ | L{e−at f (t)} = F (s + a) | (9)
i.e. L{e−at f (t)} is the same as L{f (t)} with s replaced by (s + a)
Examples
1. If L{sin 2t} = s22+4 then L{e−3t sin 2t} = (s+3)2 2 +4 = s2 +6s+13
2

2
2. If L{t2 } = s23 then L{t2 e4t } = (s−4)3

(2) Theorem 2: Multiplying by t (or derivative of LT):

If L{f (t)} = F (s) then
d
 
 
⇒ 
 L{tf (t)} = − F (s) 
 (10)
 ds 

e.g. if L{sin 2t} = s22+4 ⇒ L{t sin 2t} = − ds

d 2 4s

s2 +4 = (s2 +4)2
(3) Theorem 3: Dividing by t: If L{f (t)} = F (s) then
 ⎧ ⎫ 
 ⎨ f (t) ⎬  ∞ 
⇒ L⎩ = F (s)ds (11)
 
 
t s
 ⎭ 
 

If limit! of f (t)t as t → 0! exists, we use l’Hopital’s rule to find out if it does

e.g. L sint at ; here lim sin at
= 00 (undefined).
" "
t→0 t
By l’Hopital’s rule, we differentiate top and bottom separately and substitute
t =!0 in "the result to "ascertain the limit of the new function.
sin at a cos at
!
lim
t→0 t =lim
t→0 1 = a,
i.e. the limit exists. The theorem can, therefore, be applied.
a
Since L{sin at} = s2 +a 2
&∞
sin at
= s s2 +a2 ds = arctan as s = π2 − arctan as = arctan as
! " ∞ a %  
then L t
(4) Transform of derivative
2
Let dfdt(t) = f ′(t) and d dtf (t) 2 = f ′′(t)
Then L{f ′(t)} = 0∞ e−st f ′ (t)dt by definition.


Integrating by parts,

45
L{f ′ (t)} = [e−st f (t)]∞ ∞
0 − 0 f (t){−se
−st
}dt


′
i.e. L{f (t)} = −f (0) + sL{f (t)}

⇒ | L{f ′(t)} = sF (s) − f (0) | (12)

Similarly L{f ′′(t)} = −f ′(0) + sL{f ′ (t)} = −f ′ (0) + s[−f (0) + sL{f (t)}]

⇒ | L{f ′′ (t)} = s2 F (s) − sf (0) − f ′(0) | (13)

Similarly ⇒ | L{f ′′′(t)} = s3 F (s) − s2 f (0) − sf ′ (0) − f ′′ (0) | (14)

Alternative notation:
Let x = f (t), f (0) = x0 , f ′ (0) = x1, f ′′(0) = x2, . . . , f (n) (0) = xn and
x = L{x} = L{f (t)} = F (s) we now have
L{x} = x̄
L{ẋ} = sx̄ − x0
L{ẍ} = s2 x̄ − sx0 − x1
...
L{x} = s3x̄ − s2 x0 − sx1 − x2

4.1.2 Solution of Differential Equations by Laplace Transform

Procedure:
a) Rewrite the equation in terms of LT.
b) Insert the given initial conditions.
c) Rearrange the equation algebraically to give the transform of the solu-
tion.
d) Determine the inverse transform to obtain the particular solution
Solution of first order differential equations
Example
Solve the equation dx
dt
− 2x = 4, given that at t = 0, x = 1.
We go through the four stages as follows:
a) L{x} = x̄, L{ẋ} = sx̄ − x0, L{4} = 4s
4
Then the equation becomes (sx̄ − x0) − 2x̄ = s

b) Insert the initial condition that at t = 0, x = 1, i.e., x0 = 1

⇒ sx̄ − 1 − 2x̄ = 4s

46
 c) Now we rearrange this to give an expression for x̄:
s+4
i.e. x̄ = s(s−2)

d) finally, we take inverse transform to obtain x:

s+4 A B
s(s−2) = s + s−2 ⇒ s + 4 = A(s − 2) + Bs
(i) Put (s − 2) = 0, i.e., s = 2 ⇒ 6 = 2B or B = 3
(ii) Put s = 0 ⇒ s = −2A or A = −2
3
s+4
x̄ = s(s−2) = s−2 − 2s
⇒ x = 3e2t − 2

Solve the following equations:

1. dx 3t
dt + 2x = 10e , given that at t = 0, x = 6
2. dx 2t 4t
dt − 4x = 2e + e , given that at t = 0, x = 0
Solution of second order differential equation
Example
2
Solve the equation ddtx2 − 3 dx
dt
+ 2x = 2e3t , given that at t = 0, x = 5 and
dx
dt = 7

a) L{x} = x̄
L{ẋ} = sx̄ − x0
L{ẍ} = s2 x̄ − sx0 − x1
2
The equation becomes (s2x̄ − sx0 − x1) − 3(sx̄ − x0) + 2x̄ = s−3

b) Insert the initial conditions. In this case x0 = 5 and x1 = 7

2
(s2x̄ − 5s − 7) − 3(sx̄ − 5) + 2x̄ = s−3
5s2 −23s+26
c) Rearrange to obtain x̄ as x̄ = (s−1)(s−2)(s−3)

d) Now for partial fractions

5s2 −23s+26 A B C
(s−1)(s−2)(s−3) = s−1 + s−2 + s−3
⇒ 5s2 − 23s + 26 = A(s − 2)(s − 3) + B(s − 1)(s − 3) + C(s − 1)(s − 2)
⇒ A = 4, B = 0, C = 1
4 1
⇒ x̄ = s−1 + s−3
⇒ x = 4et + e3t
2
Solve ddtx2 − 4x = 24 cos 2t, given that at t = 0, x = 3 and dx
dt =4
Solution of simultaneous differential equations
Example

47
Solve the pair of simultaneous equations
ẏ − x = et
ẋ + y = e−t
given that at t = 0, x = 0 and y = 0
1
a) (sȳ − y0 ) − x̄ = s−1
1
(sx̄ − x0) + ȳ = s+1
b) Insert the initial conditions x0 = 0 and y0 = 0
1
sȳ − x̄ = s−1
1
sx̄ + ȳ = s+1
c) Eliminating = ȳ we have
1
sȳ − x̄ = s−1
s
sȳ + s2x̄ = s+1
s2 −2s−1 1 1 1 1 s 1
⇒ x̄ = (s−1)(s+1)(s 2 +1) = − 2 . s−1 − 2 . s+1 + s2 +1 + s2 +1

d) x = 12 et − 12 e−t + cos t + sin t= sin t + cos t − cosh t

Eliminating x̄ in (b) we have
s2 +2s−1 1 1 1 1 s 1
⇒ ȳ = (s−1)(s+1)(s 2 +1) = 2 . s−1 + 2 . s+1 − s2 +1 + s2 +1

⇒ y = 21 et + 12 e−t − cos t + sin t= sin t − cos t + cosh t

So the results are:
x = sin t + cos t − cosh t
y = sin t − cos t + cosh t

4.2 The Dirac Delta Function − the impulse function

It represents an extremely large force acting for a minutely small interval of
time. Consider a single rectangular pulse of width b and height 1b ocurring at
t = a, as shown in Figs. 1(a) and (b).
If we reduce the width of the pulse to 2b and keep the area of the pulse con-
stant (1 unit) the height of the pulse will be 2b . If we continue reducing the
width of the pulse while maintaining an area of unity, then as b → 0, the
height 1b → ∞ and we have the Dirac delta function. It is denoted by δ(t−a).
Graphically it is represented by a rectangular pulse of zero width and infinite
height, Fig. 2.

48
If the Dirac delta function is at the origin, a = 0 and so it is denoted by
δ(t)

4.2.1 Integration involving the impulse function

From the definition of⎧δ(t − a)
⎪
⎪
⎪
⎨
(i) p < t < a, δ(t − a) = 0
q
p δ(t − a)dt = 1 for ⎪ (ii) t = a area of pulse = 1
⎩ (iii) a < t < q, δ(t − a) = 0
⎪
⎪

Now consider pq f (t)δ(t − a)dt since f (t)δ(t − a) is zero for all values of t


within the interval [p, q] except at the point t = a, f (t) may be regarded as
a constant f (a), so that pq f (t)δ(t − a)dt = f (a) pq δ(t − a)dt = f (a)
 

Examples
Evaluate 13 (t2 + 4) · δ(t − 2)dt. Here a = 2 f (t) = t2 + 4 ⇒ f (a) = f (2) =


22 + 4 = 8
Evaluate
6
1. 0 5 · δ(t − 3)dt
 5 −2t
2. 2 e · δ(t − 4)dt
Laplace transform of δ(t − a)
Recall that pq f (t) · δ(t − a)dt = f (a), p < a < q


⇒ if p = 0 and q = ∞ then 0∞ f (t) · δ(t − a)dt = f (a)



Hence, if f (t) = e−st , this becomes

 ∞ −st
0 e · δ(t − a)dt = L{δt − a} = e−as
Similarly L{f (t) · δ(t − a)} = 0∞ e−st · f (t) · δt − adt = f (a)e−as


4.2.2 Differential equations involving the impulse function

Example
Solve the equation ẍ + 4ẋ + 13x = 2δ(t) where, at t = 0, x = 2 and ẋ = 0

a) Expressing in LT, we have (s2x̄ − sx0 − x1) + 4(sx̄ − x0) + 13x̄ = 2 × 1

2s+10
b) Inserting the initial conditions and simplifying we have x̄ = s2 +4s+13

49
 c) Rearranging the denominator by completing the square, this can be
2(s+2) 6
written as x̄ = (s+2)2 +9 + (s+2)2 +9

d) The inverse LT is x = 2e−2t cos 3t + 2e−2t sin 3t = 2e−2t(cos 3t + sin 3t)

4.3 Fourier Series

You have seen from Maclaurin’s and Taylor’s series that an infinitely differen-
tiable function can be expressed in the form of an infinite series in x. Fourier
series on the other hand, enables us to represent a periodic function as an
infinite trigonometrical series in sine and cosine terms. We can use Fourier
series to represent a function containing discontinuities unlike Maclaurin’s
and Taylor’s series.
Periodic function: A function f (t) is periodic iff
f (t) = f (t + nT ), n = 0, ±1, ±2 . . .
T is called the period. For sine and cosine the period T = 2π so that
sin t = sin(t + 2πn) and cos t = cos(t + 2πn)
Analytical description of a periodic function
Many periodic functions are non-sinusoidal
Examples

1. f (t) = 3 0 < t < 4

f (t) = 0 4 < t < 6
f (t) = f (t + 6) i.e. the period is 6
2. f (t) = 58 t 0 < t < 8
f (t) = f (t + 8)
Sketch the following periodic functions
1. f (t) = 4 0 < t < 5
f (t) = 0 5 < t < 8
f (t) = f (t + 8)
2. f (t) = 3t − t2 0 < t < 3
f (t) = f (t + 3)

50
4.3.1 Fourier series of functions of period 2π
Any periodic function f (x) = f (x + 2πn) can be written in Fourier series as
1 ∞
f (x) = a0 + (an cos nx + bn sin nx)
2 n=1
1
= a0 + a1 cos x + a2 cos 2x + . . . + b1 sin x + b2 sin 2x + . . .
2
(where a0 , an , bn, n = 1, 2, 3 . . . are Fourier coefficients) or as
1
f (x) = a0 + c1 sin(x + α1 ) + c2 sin(2x + α2 ) + . . .
2

where ci = a2i + b2i and αi = arctan( abii ).
c1 sin(x + α1 ) is the first harmonic or fundamental
c2 sin(2x + α2 ) is the second harmonic
cn sin(nx + αn ) is the nth harmonic.
For the Fourier series to accurately represent f (x) it should be such that if
we put x = x1 in the series the answer should be approximately equal to
the value of f (x1) i.e. the value should converge to f (x1) as more and more
terms of the series are evaluated. For this to happen f (x) must satisfy the
following Dirichlet conditions:
a) f (x) must be defined and single-valued.
b) f (x) must be continuous or have a finite number of discontinuities within
a periodic interval.
c) f (x) and f ′(x) must be piecewise continuous in the periodic interval.
If these conditions are met the series converges fairly quickly to f (x1) if
x = x1, and only the first few terms are required to give a good approxima-
tion of the function f (x)
Fourier coefficients: The Fourier coefficients above are given by
a0 = π1 −π
π
f (x)dx
1 π
an = π −π f (x) cos nxdx
bn = π1 −π
π
f (x) sin nxdx
Odd and even functions
a) Even functions: A function f (x) is said to be even if f (−x) = f (x). The
graph of an even function is, therefore, symmetrical about the y-axis. e.g.

51
f (x) = x2 f (x) = cos x

b) Odd functions: A function f (x) is said to be odd if f (−x) = −f (x);

the graph of an odd function is thus symmetrical about the origin. e.g.
f (x) = x3 f (x) = sin x

Products of odd and even functions

(even)×(even)=even
(odd)×(odd)=even
(neither)×(odd)=neither
(neither)×(even)=neither
Theorem 1: If f (x) is defined over the internal −π < x < π and f (x) is
even, then the Fourier series for f (x) contains cosine terms only. Here a0 is
included.
Example
f (x) = 0 − π < x < − π2
f (x) = 4 − π2 < x < π2
f (x) = 0 π2 < x < π
f (x) = f (x + 2π)
The waveform is symmetrical about the y-axis, therefore, it is even.
⇒ f (x) = 12 a0 + ∞ n=1 an cos nx

π π
1 π 2 π 2 2 2
a) a0 = π −π f (x)dx= π 0 f (x)dx= π 0 4dx = π [4x]02 = 4

π &π
b) an = π1 2 π 2 2 8 sin nx 2
π %
−π f (x) cos nxdx = π 0 f (x) cos nxdx = π 0 4 cos nxdx = π n 0
=
8 nπ
πn sin 2 ⎧
0 for n even
⎪
⎪
⎪
⎨
nπ
But sin = ⎪ 1 for n = 1, 5, 9 . . .
2
⎩ −1 for n = 3, 7, 11 . . .
⎪
⎪

8 1 1 1
# $
⇒ f (x) = 2 + cos x − cos 3x + cos 5x − cos 7x + . . .
π 3 5 7
Theorem 2: If f (x) is an odd function defined over the interval −π < x < π,
then the Fourier series for f (x) contains sine terms only. Here a0 = an = 0

52
Example
f (x) = −6 − π < x < 0
f (x) = 6 0 < x < π
f (x) = f (x + 2π)
This is an odd function so f (x) contains only the sine terms
i.e. f (x) = ∞ n=1 bn sin nx


1 π
and bn = π −π f (x) sin nxdx
f (x) sin nx is even since it is a product of two odd functions.
2 π 2 π 12 −cosnx π 12
% &
⇒ bn = π 0 f (x) sin nxdx= π 0 6 sin nxdx= π n 0
= πn (1 − cos nπ)
24 1 1
# $
f (x) = sin x + sin 3x + sin 5x + . . .
π 3 5
If f (x) is neither even nor odd we must obtain expressions for a0 , an and bn
in full
Examples
Determine the Fourier series of the function shown.
f (x) = 2x
π 0<x<π
f (x) = 2 π < x < 2π
f (x) = f (x + 2π)
This is neither odd nor even,
1 ∞
⇒ f (x) = a0 + {an cos nx + bn sin nx}
2 n=1
1  2π 1 π 2  2π
! "
a) a0 = 0 f (x)dx = 0 π xdx + π 2dx
π  π  π
1 x2 1
= π π + [2x]π2π = π {π + 4π − 2π} = 3
0
⇒ a0 = 3
b) an =! π1 02π f (x) cos nxdx


= π1 0π 2x
 2π "
cos nxdx + 2 cos nxdx

! % π π
2 1 x sin nx π 1 π  2π
& "
=π π n 0
− πn 0 sin nxdx + π cos nxdx
1 cos nx π sin nx 2π
 & 
= π2 1
% & %
(π sin nπx) + +
! πn πn n 0 n π
2 1 1 1
"
= π ! n sin nπx + πn2 (cos πnx − 1) +" n (sin 2πnx − sin nπx)
= π2 πn1 2 (cos πnx − 1) + n1 sin 2nπx
an = 0 (n even); an = π−4 2 n2 (n odd)

c) bn =! π1 02π f (x) sin nxdx



= π1 0π 2x
 2π "
sin nxdx + 2 sin nxdx

π π

53
 1 −x cos nx π
= π2 1 π
! % &  2π "
− πn 0
cos nxdx + π sin nxdx
π n 0
sin nx π nx 2π
& 
= π πn (−π cos nπx) + πnx n 0 + − cos
2 1 1
% & %
n π
2 1 1
! "
= π !− n cos nπx +" (0 − 0) − n (cos 2πnx − cos nπx)
= π2 − n1 cos 2nπx = − πn 2
cos 2nπx
2
But cos 2nπ = 1 ⇒ bn = − πn

3 4 1 1
# $
f (x) = − 2 cos x + cos 3x + cos 5x + . . .
2 #π 9 25
2 1 1 1
$
− sin x + sin 2x + sin 3x + sin 4x . . .
π 2 3 4

4.3.2 Half-range series

Sometimes a function of period 2π is defined over the range 0 to π instead of
the normal −π to π, or 0 to 2π. In this case one can choose to obtain a half-
range cosine series by assuming that the function is part of an even function
or a sine series by assuming that the function is part of an odd function.
Example
f (x) = 2x 0 < x < π
f (x) = f (x + 2π)
from fig. 1

To obtain a half-range cosine series we assume an even function as in fig.

2. % &π
a0 = π2 0π f (x)dx = π2 0π = π2 x2 0 = 2π
 
&π
an = π2 0π 2x cos nxdx = π4 x sinn nx 0 − n1 0π sin nxdx
 !%  "

−8
Simplifying, a!n = 0 for n even and an = πn 2 for n odd. In this b0 = 0 and so
8 1 1
"
f (x) = π − π cos x + 9 cos 3x + 25 cos 5x + . . .
Obtain a half-range sine series for f (x).

4.3.3 Functions with arbitrary period T

i.e. f (t) = f (t + T ), frequency f = T1 and angular frequency ω = 2πf
⇒ ω = 2π T
and T = 2π ω
. The angle x = ωt and the Fourier series is f (t) =
1 ∞
2 a0 + n=1 {an cos nωt + bn sin nωt}


54
 = 12 a0 + ∞ 2πnt 2πnt
n=1 {an cos T + bn sin T }


where
 2π
a0 = T2 0T f (t)dt = ωπ 0ω f (t)dt

2π
2 T ω  ω
an = T 0 f (t) cos nωtdt = π 0 f (t) cos nωtdt
2π
2 T ω ω
bn = T 0 f (t) sin nωtdt = π 0 f (t) sin nωtdt
Example
Determine the Fourier series for a periodic function defined by
f (t) = 2(1 + t) − 1 < t < 0
f (t) = 0 0 < t < 1
f (t) = f (t + 2) 0 < t < 1
Answer:
1 4 1 1
# $
f (t) = + 2 cos ωt + cos 3ωt + cos 5ωt . . .
2 #ω 9 25
2 1 1 1
$
− sin ωt + sin 2ωt + sin 3ωt + sin 4ωt
ω 2 3 4

4.3.4 Sum of a Fourier series at a point of finite discontinuity

At x = x1 the series converges to the value
f (x1) as the number of terms included
increases to infinity. But if there is a
’jump’ at x1 as shown in Fig.
f (x1 − 0) = y1 (approaching x1 from below) f (x1 + 0) = y2 (approaching x1
from above)
If we sustitute x = x1 in the Fourier series for f (x), it can be shown that
the series converges to the value 21 {f (x1 − 0) + f (x1 + 0)} i.e. 21 (y1 + y2), the
average of y1 and y2 .

4.4 Fourier Integrals

4.4.1 The Fourier integral
: While Fourier series is for periodic functions Fourier integral is for non-
periodic function. If a non-periodic f (x) (i) satisfies the Dirichlet conditions
in every finite interval (−a, a) and (ii) is absolutely integrable in(−∞, ∞), i.e.
∞
−∞ |f (x)|dx converges, then f (x) can be represented by a Fourier’s integral

55
as follows:  ∞
f (x) = {A(k) cos kx + B(k) sin kx}dx (1)
0
∞
where A(k) = −∞ f (x) cos kxdx (2)

∞
B(k) = −∞ f (x) sin kx}dx 
(3)

f (x+0)+f (x−0)
If x is a point of discontinuity, then f (x) must be replaced by 2
as in the case of Fourier series. This can, in other words, be expressed by the
following theorem.
Theorem 1: If f (x) is piecewise continuous in every finite interval and has a
∞
right-hand derivative and a left-hand derivative at every point and if −∞ |f (x)|dx
exists, then f (x) can be represented by a Fourier integral. At a point where
f (x) is discontinuous the value of the Fourier integral equals the average of
the left- and right-hand limits of f (x) at that point.
Examples
Find the Fourier integral representation of the function ⎧ in fig. below
⎨ 1 if |x| < 1
f (x) = ⎩
0 if |x| > 1
Solution: From (2) and (3) we have
1
A(k) = π1 −∞ f (x) cos kxdx = π1 −1 f (x) cos kxdx = sinπkkx −1 = sin k
∞ 1
πk
B(k) = π1 −1
1
sin kxdx = 0
and (1) gives the answer
2  ∞ cos kx sin k
f (x) = dk (4)
π 0 k
The average of the left- right-hand limits of f (x) at x = 1 is equal to (1+0)/2,
that is, 1/2.
Furthermore, from (4) and Theorem 1 we obtain
⎧
π
⎪
⎨ 2
if 0 ≤ x < 1
cos kx sin k π
 ∞ ⎪
⎪
π
= f (x) = 4 if x = 1
0 k 2 ⎪
⎩ 0 if x > 1
⎪
⎪

This integral is called Dirichlet’s discontinuous factor. If x = 0, then

sin k ∞ π
dk = .
0 k 2
This integral is the limit of the so-called sine integral
 u
sin k
Si(u) = dk
0 k
56
as u → ∞
In the case of a Fourier series the graghs of the partial sums are approximation
curves of the periodic function represented by the series. Similarly, in the
case of the Fourier integral, approximations are obtained by replacing ∞ by
numbers a. Hence the integral
 a cos kx sin k
dk
0 k
approximates the integral in (4) and therefore f (x)

57