See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332447091

Basic Concepts of integral equations

Chapter · April 2019

CITATIONS READS

0 6,450

3 authors, including:

M. A. Abdou
Alexandria University
667 PUBLICATIONS 12,919 CITATIONS

SEE PROFILE

All content following this page was uploaded by M. A. Abdou on 16 April 2019.

The user has requested enhancement of the downloaded file.

 Chapter 1
Basic Concepts of integral equations
By
M.A. Abdou; F. M. Al‫ ـ ـ‬Saedy
The purpose of this chapter is to state some definitions and notations for functions defined
in Hilbert space. In addition, we discuss the continuity and boundedness of the integral operator
in L2  space. Beside this, we discuss the uniqueness of the integral equation and state the
Fredholm theorems. Also, we state some important relations for special functions as Beta and
Gamma function, Bessel function, Legendre polynomials, Jacobi polynomials and Hyper
geometric polynomials.
§ 1.1 Definitions and Theorems
Definition 1.1(A linear space) A linear space over a field F is a collection X of elements with
two defined operations. The first is addition of elements in X and the second is multiplication
of elements in X by scalars in F. In addition, we stipulate the following conditions:
1.X forms a commutative group under the additive operation. That is, if f , g , h ,... belong to X,
then
(a) The operation is closed, so that ( f  g ) belongs to X for all f and g in X .

(b) The operation is associative: f  g   h  f  g  h 

(c) There exists an identity 0 for which f  0  f for all f in X.

(d) For every ƒ there exists an inverse element denoted by  f  such that, f   f   0 .

(e) The operation is commutative: f  g  g  f for all f , g in X.

2. Multiplication by scalars is closed that is,
(a) 1.f  f for all f in X.
(b)  f is in X for all f in X and  in F.

(c) for all  , in F and f in X,    f     f .

3. The following distributive laws hold.
(a)   f  g    f   g for all  in F and f , g in X.

(b)     f  f   f for all  ,  in F and f in X.

1
Definition 1.2 (Inner product)
A linear space X is said to be an inner product space if an inner product is defined on it. Such
an inner product assigns to every pair f and g in X a complex number denoted by  f , g .
By definition, such an inner product has the following properties:

(1)  f , g    g , f . (2)  f   g , h     f , h     g , h  .

(3)  f , f   0 and  f , f   0 if and only if f  0.

 f , f  is real ,since by property (1)  f , f    f , f  . Hence, the norm f is defined by:
f  f , f  .
1
2

Theorem 1.1 (Cauchy-Schwarz. Inequality): Let f and g belong to an inner product

space, then, f , g   f g .Equality is achieved if and only if f and g are linearly
dependent; that is for suitable scalars  and  , f   g  0 .
Proof: If in (1.1) , f  0 ,the proof is trivial, so for f  0 , we let
f , g 
f , g   f , g  ei  , i.e.  2
e i .
f

So that,   f  g ,  f  g   f  g
2 2 2
f f  0 . Using the properties of inner product, one

can expand the left side of this inequality to obtain

[ f , f     g , f    f , g    g , g ] f
2 2

  f , g    f , g   
2
=  
2 2 2
f g f

  f , g    f , g   g 
2
=  
2 2 2
f f  0.

Now, we use the particular value of  and  f , g

 f , g 
2
f , g 
2
f , g 
2

    g  f
2 2
 f , g 
2
 g
2
f
2
0
 f
2
f
2
f
2

 

f ,g 
2 2 2
Hence, we obtain, f g  . To complete the proof we note that if f and g are
linearly independent,  f  g  0. So that, we get a strict inequality. Equality will occur if

g  f for a suitable  . 

2
 Theorem 1.2: The norm f has the following three properties:

i. f  0 and f  0 if and only if f  0 .

ii. f  f . Iii. f g  f  g .

The last property is known as the triangle inequality.

Proof: The formulas (i) and (ii) are immediate consequences of the definition of an inner
product. To prove (iii) we use the Cauchy-Schwarz inequality as the following:

  f  g ,f  g  f  f ,g  f ,g  g,g 

2
f g ,f

f , g   f  g 
2 2 2 2 2
 f 2  g  f 2 f g  g

 f  g 
2
f g
2
Hence, we obtain , which yields, f g  f  g

One can also see that (iii) reduce to an equality only if f and g are linearly dependent. 
Definition 1.3: Let H be an inner product space and  f n  a Cauchy sequence in H. Such a
sequence has the property that for every   0 , we can find an   such that,

f n f m  for n, m    .

In other words, lim fn  fm  0

n, m  

The inner product H is said to be a Hilbert space if every Cauchy sequence converges to an
element in H.
It is easy to see that, if such a Cauchy sequence converges, it must converge to a unique
element.
For this, suppose lim f nk  g , lim f m k  h .
n k  m k 

That is, we have two subsequences of  f n  such that, each converges to a different element.
Then, g  h  g  f nk  f nk  f m k  f m k  h

 g  f nk  f nk  f m k  f m k  h

For nk and mk sufficiently large we have,

g  f nk   1 , f nk  f mk   2 , f mk  h   3 .

So that, g  h  ,   1   2   3

3
Since g and h are independent of nk and mk and  is arbitrary, we see that necessarily
g  h  0. So, we obtain g  h .

Definition 1.4 (Norm in L2  space)We say that   x  , is integrable function on a , b, if

b

the integral   2  x  dx exist and the class of integrable denoted by L2 a , b  .

a

b 2 2
The norm of L2  function   x  is defined by,           x  dx  .
a 
Definition 1.5 (Norm in L p ). The norm in L p  space, is defined as
b

    x  
p 1

dx p

Definition 1.6 (Norm in l p  space). Let p  1 be a fixed real number. We define each

  k  of numbers such that 1  2  

p p p
element in the space l by the sequence


  ,  p  1, fixed 
p
converges; then k
k 1


The norm in l p  space is given by     
p 1
p
k
k 1


If p  2 , we have the famous Hilbert sequence space l with the norm    
2 2
k
k 1

Definition 1.7 (The C a, b  space). The element of this space are all possible functions
l

defined on the interval a, b and having, on this interval, continuous derivatives up to the l
inclusive. The operation of addition and multiplication of functions by a number are defined in
the usual manner. We determine the norm of an element, a function,   x  C l a, b  from
l
the formula    max  k  x  ,  0  x  being equal to   x  .
a x b
k 0

Definition 1.8 (Norm C a , b   space). The space Ca , bcontains the continuous functions

with its derivatives . If  x  Ca , b then, we define the norm  by the equation
b

  max   x  or   max    x  dx .
a  x b a  x b
a

4
 Definition 1.9 (Norm in L2   C 0 , T  ).The norm in the space L2    C 0 , T  , where

 is the domain of integration of  x, t  , with respect to positions, x  xx1,, xn  while

1

 t
2
t  0,T  ,T  1 represent the time, is defined as   x , t   max     2  x , t  dx  dt
0 t T
0  

Definition 1.10 ( Holder inequality ). If  and  belong to L p  space, then

b b b

  x  x  dx    x  dx     q x  dx   1 1

1 1
p p q
 1
a a a
p q

As a special case, when p  q  2 , we have Cauchy‫ ـ‬Munkoviski inequality in L2 .

Definition 1.11. (Gamma and Beta function). We define the gamma and beta functions
 1
respectively by ;   x   and   x , y   t x 1 1  t 
y 1

0
e t t x 1 dt 
0
dt .The first integral is valid

only for x  0 , and the second only for x  0 and y  0 , because it is for just these values of

x and y that the above integrals are convergent.

 x    y 
The relation between beta and gamma functions is   x , y  
 x  y 

Definition 1.12. Let us define  r (the so‫ــ‬called Pochhammer symbol) by

   r 
 r     1   r 1  ,  0  1 (r a positive integer)
  

Then, we define the general hyper geometric function m Fn 1 , 2 ,, m ; 1 ,  2 ,,  n ; x  by

1 r  2 r  m r xr
Fn 1 ,  2 , ,  m ; 1 ,  2 , ,  n ; x       
m
r 0 1 r 2 r   n r r !

We shall confine ourselves to the two separate cases: m  n  1 (in which case the function will
be called the confluer hyper geometric function or Kummer function) and m  2, n  1 (in which
case we shall merely call it the hyper geometric function).

d 2y
       1 x 
dy
The differential equation x 1  x  2
  y  0 hyper geometric equation
dx dx
or Gauss's equation) has 2 F1  ,  ;  ; x  as solution. If  is not an integer a second independent

solution is given x 1 2 F1   1   ,   1   ;2   ; x  .

 1 x 
Theorem 1.3: i. Pn  x  2 F1   n, n  1,1;  , ( Pn is a Legendre polynomial ) and
 2 

5
 2 F1  ,  ; ; x  2 F1  , ; ; x 
Definition 1.13. Bessel functions are solutions of the differential equation:

x2
d 2y
dx 2
 x
dy
dx
 
 x 2  n 2 y  0. where n is a (possibly complex) constant, and is known as

Bessel differential equation of order n . If we replace the independent variable x by  x , the

resulting equation is;. x 2
d 2y
dx 2
 x
dy
dx
 
  2 x 2  n 2 y  0. This is another form of Bessel

differential equation.
Solution of Bessel differential equation:
One solution of Bessel differential equations is given by ;
 x2 x4 x6 
y1  a0 x n  1  2  4  6  
 2 n  1 2  2!n  1n  2 2  3!n  1n  2n  3 
1
Since a0 is at our disposal, we may take a0  . This will yield to the solution
2 n  1
n

denoted by J n  x  and given by;

x 2  x 2 x 2 x 2    1  x 2 
2 4 6 k n  2k

J n x    1    
  n  1  1! n  1 2! n  1 n  2  3! n  1 n  2  n  3  k  0 k !   n  k  1

where J n  x  is called the Bessel function of the first kind of order n and argument x .

 1  x 2 
k n  2k

The second solution of Bessel differential equations is given by ; J  n  x    .
k 0 k !   n  k  1

For the values of n we consider, J n  x  and J  n  x  are convergent for all values of x expect

x  0 . Since J n x  contains negative powers of x while J n x  does not, near the origin
J n x  is unbounded while J n x  remains finite. Hence , when 2n is not an integer, J n x  and
J n x  are two linearly independent solutions of Bessel deferential equation. Then a general
solution of Bessel differential equation when 2n is not an integer is ; y  x   c1J n  x   c 2J  n  x 

Recurrence Relations for Bessel Functions: The Bessel function of the first kind J n  x  satisfies
the following relations:
d d
1. {x n J n  x }  x n J n 1  x  , 2. {x  n J n  x }  x  n J n 1  x  , 3.J n  x   J n 1  x   xn J n  x 
dx dx

4. J n  x   J n 1  x   xn J n  x  , 5.J n  x   1
2
{J n 1  x   J n 1  x }, 6. J n 1  x   J n 1  x   2n
x
J n x 

Definition 1.14. The differential equation of the form: 1  x 2  y   2xy   n  n  1 y  0

6
where n is non ‫ ــ‬negative integer, and called the Legendre differential equation. It is clear that the
only singularities are at x  1. In fact x  1 are regular singular points of the differential
equation.
The solution of Legendre equation is Pn  x  and call the Legendre polynomial of order n :

 2 
n
 2n  2r  ! n 2
n
for n even
Pn  x     1 n 2r
r
x where    n-1
r 0 2n r !  n - r  !  n  2r  ! 2  2 for n odd

This is the solution of Legendre equation which is finite for  1  x  1.

Recurrence relation for Legendre polynomial
1.  n  1 Pn 1  x    2n  1 xPn  x   nPn 1  x   0,
2. nPn  x   xPn  x   Pn1  x 

3.  2n  1 Pn  x   Pn1  x   Pn1  x  ,
4.  n  1 Pn  x   Pn1  x   xPn1  x  ,

 
5. 1  x 2 Pn  x   n [Pn 1  x   xPn  x ].

Definition 1.15. Jacobi polynomials satisfy the differential equation

1  x  y           1 x  y   n  n      1 y  0 ,
2

and denoted by Pn ,    x  . Their generating function is:

2   
  Pn
 , 
x  t n
  1  t  
 
n 0
1  2xt  t 2 1  t  1  2xt  t 2 1  2xt  t 2

Recurrence Relations for Jacobi polynomials

1
1. Pn
 , 
 x   1      n  Pn11,  1  x  ,
2

2.  x  1 Pn  x   nPn ,    x      n  Pn11,    x  .
 , 

3.  x  1 Pn  x  nP  ,   x    n P  ,  1 x ,
 , 
  n     n 1  
 1
4.Pn
 , 
x   {   n  Pn1   x     n  Pn1   x }.
 1,   ,  1

§ 1.2 Laplace Transformation: Let F t  be any function, then the Laplace transform of

F t  is denoted by L F t  and defined by L F t    e st F t  dt  f (s ), (s  0).
0

Transforms of elementary function: L  e k t   k  k , we have  

1 1
. If L e  kt  .
s k s k

In addition, we can obtain: L sin kt  

k
s2 k 2
; L cos kt   2
s
s k 2
n!
 
; L t n  n 1 ; L 1 
s
1
s

7
Transforms of derivatives: If F t  is continuous for t  0 and of exponential order as t   ,
 
and if F  t  is also continuous then L F  t   e st F  t  dt  [e st F t  0 F t  dt

 e
 st
 s
0 0

i .e . L F  t  =  F  0   s L F t . In addition, we have L F  t   s 2 L F t   sF  0  F   0

Theorem 1.4: If F t  is a function of continuous with its derivatives we have

L t F t    f  s  ;
d
ds

 
n
L t n F t    1
n d

ds n
f s 
Examples 1.1.
2
L sin 2t  = ;
s 42

d 2 4s
L t sin 2t    { 2 }  2 ;
ds s  4 s 4  
d  1 
 
L t e 2t     
1
ds  s  2   s  2 2


Theorem 1.5 ( Shift Theorem ) If f  s    e st F t  dt Then, we have
0

 
f s  a   e
at
. e st F t  dt  e
 st
[e at F t ] dt .
0 0

Inverse of Laplace:
 s  1   k   1 
L1  2 2 
 cos kt ; L1    1; L1  2 2 
 sin kt ; L1   e
k t

s  k  s  s  k  s  k 

Example 1.2 Find L1  

15

 s  4s  13 
2

 
Solution L1   1  
15 15
 L  
 s  4s  13   s  2  9 
2 2
 

 
 15  1  3 
Applying shift theorem: L1  2t
5e L  2  5e
2t
sin 3t
 s  2  9  s  9 
2
 

Example 1.3

1  
s 1  s  3  2 
   3t 1  s  2  1  s  1 1  4 
L1  2  L   e L  2
3t
  e {L  2   2 L  s 2  16 }.
  s  3  16 
 s  6s  25  
2
 s  16   s  16   


 s 1 
L1  2  e
3t
 cos 4t  1
sin 4t 
 s  6s  25 
2

8
 Theorem 1.6 (A convolution Theorem)
We now seek a formula for the inverse transform of a product of transforms.
If F t  and G t  are assumed to be functions of class A and if
  
f  s    e st F t  dt ; g  s    e st G t  dt , we have f  s  g  s   e
 st
F t  G t  dt
0 0 0

§ 1.3 Classification of Integral Equations

In this section, we will state the kinds of integral equation with respect to its formula and its
kernel.
(i) Integral equation with respect to its formula:

1. Fredholm integral equations: The Fredholm integral equations are written in the form,
b
h  x    x   f  x     k  x , y    y  dy . (1.1)
a

The functions f and k are known and called the free term and the kernel of integral equation

respectively, while  is an unknown function will be determined,  is a non-zero, real or

complex, parameter. The function h x  defined the kind of the integral equation.

(a)-For the Fredholm integral equation of the first kind, hx   0 , thus,
b
f  x     k  x , y    y  dy  0 (1.2)
a

(b)- For the Fredholm integral equation of the second kind, assume
h  x     constant  0 , to have
b
  x   f  x     k  x , y   y  dy . (1.3)
a

(c)-If hx   0 then, the formula (1.1) is called the Fredholm integral equation of the third kind.
(d)-The homogeneous Fredholm integral equation of the second kind is special case of (b), when
f  x   0 ; i.e. we have
b
  x     k  x , y    y  dy (1.4)
a

2. Volterra integral equations: Volterra integral equations of the first, second, third and
homogeneous kinds respectively are defined precisely as above except that b  x is the
variable upper limit of integration namely,

9
 x
f  x     k  x , y    y  dy  0 , (first kind)
a

x
  x   f  x     k  x , y    y  dy , (second kind)
a

x
h  x    x   f  x     k  x , y    y  dy , (third kind)
a

and
x
  x     k  x , y    y  dy . (homogeneous formula) (1.5)
a

Nonlinear integral equations: The nonlinear integral equation of the first kind take the form
b
f  x     k  x , y ,   y   dy  0 (first kind )
a

b
  x   f  x     k  x , y ,   y   dy (sec ond kind )
a

b
h  x    x   f  x     k  x , y ,   y   dy (third kind ). (1.6)
a

In addition, we have nonlinear Volterra equations of the first kind,

x
f  x     k  x , y ,   y   dy  0
a

x
  x   f  x     k  x , y ,   y   dy
a

x
h  x    x   f  x     k  x , y ,   y   dy (1.7)
a

The integral equations h  x    x     k  x  y   y  dy  f  x  , , are called the Wiener- Hopf

a

integral equation of the first kind when h  x   0 ,of the second kind when h  x     0 , and of the

third kind when hx  is variable .

x

The integral equations h  x    x     k  x  y    y  dy  f  x  are called Renwal equations, and

a

its kind depends on the values of h x  .

x
y 
The Abel's integral equations: h  x    x      x  y  dy  f  x  , ( 0    1)
a

11
 b

The Hammerstein integral equations h  x    x     k  x , y  F  y ,   y   dy  f  x 

a

The Hammerstein-Volterra integral equations h  x    x     k  x , y  F  y ,   y   dy  f  x 

a

b
x   x   y 
The Chandrasekhar- integral equation 1    dy  f  x 
a
x y

.Integral equation with respect to its kernel:

If the kernel k  x , y  is continuous in a , b   a , b  , or at least the discontinuity of the kernel be
b b 2

such that the double integral,  k  x , y  dx dy , has a finite value, then the integral equation is
a a

called an equation of Fredholm type. Otherwise, it is likely to be singular. i.e,

 H x , y 
 0  1
If the kernel has the form, k  x , y    x  y H x , y  is continuous

where,

H  x , y  ln x  y

and differentiable function, then the integral equation is said to have a weak sigularity or,it may
H x , y 
called the integral equation with Carleman kernel k  x , y   
, or logarithmic kernel
x y

k  x , y   H  x , y  ln x  y

The integral equation will be regarded as being of singular type, when the kernel under
A x , y 
consideration is of the form, k  x , y   , where the numerator, A  x , y  is a differentiable
x y
b b
A x , y 
function of x and y . In this case, the integral  k x
a
, y    y  dy  
a
x y
  y  dy ; .

Appearing, in the integral equation, diverges in general. A kernel is said to have a strong
B x , y 
singularity, if we write it in the form, k  x , y   where B  x , y  is a differentiable
x  y 
2

function of x and y .
A kernel is said to have a supper strong singularity, if we write it in the form,
C x , y 
k x , y   ( m  2), where C x , y  is a differentiable function of x and y .
x  y 
m

11
(ii) Mixed integral equations:
The Fredholm-Volterra integral equation: The integral equations

b t
h  x    x , t     k  x , y    y , t  dy    F t ,     x ,   d   f  x , t  (1.8)
a 0

are called Fredholm-Volterra integral equations, where the Fredholm integral term is considered
with respect to position, while the Volterra integral term is measured with respect to time, ( see
Abdou [16] and [17] ) . The kinds of (1.73) depend on the value of h x  .
The Volterra-Fredholm integral equation: The integral equations
t b
h  x    x , t      k  x , t  F t ,     y ,   dy d   f  x , t  (1.9)
0 a

are called the Volterra-Fredholm integral equations and its kinds depend on the value of hx  ,(
see Abdou [20] and [19] ) .
The integral equation
t
  x , y , t    k  x   , y      , , t  d  d    F t ,    x , y ,  d   f  x , y , t  (1.10)
 0

is called Fredholm-Volterra integral equation of the second kind and its value depends on the
formula of the kernel of position and the domain of integration with respect to position, see Abdou
and Salama [20] and Abdou and Moustafa [21] .
§ 1.4 Fredholm theorems for continuous kernel
In this section, we will state the Fredholm theorem for continuous kernel without proof. The
proof can be found in Hochstadt [3], Tricomi [5] and Kanwal [7] .
Theorem 1.7: Let k  x , y  be a continuous kernel for all 0  x, y  1 .The integral equation,
b
  x   f  x     k  x , y   y  dy
a

has a unique continuous solution given by ,

b
D x , y ;  
 x   f x    D  
f  y  dy
a


 
where, D  x , y ;    C n  x , y  n ! , C 0  x , y   1
n


n 0

 b
 
   n
C n 1    n  1 c n k  x , y    k  x , z C n  z , y  dz  , D     c n n!
 a  n 0

12
 b b
x1 x 2 xn 
c 0  1, c n    k  x  dx 1 dx n
a a 1 x2 xn 

Both of these functions are entire functions of .

Theorem 1.8. The following relations are satisfied ,
b
D    b
 D       D  x , x ;  dx ;    n 1  k n  x , x dx
a D   n 1 a

The equation hold for all  sufficiently small ;that is for all   0 where 0 is a zero of

D    closest to the origin.

Theorem 1.9: Every zero of D    is an eigenvalue of the integral operator K ,and in turn every

eigenvalue of K is a zero of D    .

Theorem 1.10: If   0 is a root of D     0 of order q ,then the homogenous equation,

b
  x   0  k  x , y    y  dy
a

has q linearly independent solutions in term of which every other solution is expressible linearly
and homogeneously. Such a system of q independent solutions is given by,

x1 x i 1 , x , x i 1 , ,xq 
 
D 0 
y , ,y q 
  x    1   1, ,q
x1 , ,x q 
 
D 0 
y , ,y q 
 1

Theorem 1.11: If 0 is a characteristic constant of k  x , y  of index q then, the equation,

b
  x   f  x   0  k  x , y    y  dy
a

has in general, no continuous solution. In order that a continuous solution exist, it is necessary
that
b

 f x 
a
  x  dx  0, (  1, , q ).

Where the  x  are complete set of fundamental functions for the associated homogeneous
equation

13
 b
  x   0  k  y , x    y  dy
a

If these conditions are satisfied, then there are a q-fold infinitude of solutions of the equation
b
  x   f  x   0  k  x , y   y  dy
a

given by
b
  x   f  x    H  x , y    y  dy  c11  x   c 22  x    c q q  x 
a

where
 x , x 1, , x q 
 
D 0 
y,y , ,y 
H x , y    1 q 
 1
x , , x q 
 
D 0 
y , , y 
 1 q 

§ 1.5 Integral operator

In the study of integral equations, the notation of an operator and specifically an integral
operator is fundamental. Such an operator assigns to an element f in H a new element, say Kf
in H . If the operator satisfies the condition,
K  f   g    Kf   Kg

we say that, K is a linear operator. An example of a linear operator is,

1
Kf   k  x , y  f  y  dy (1.11)
0

where, f  L2 0,1 and k  x , y  is continuous in x and y .Such an operator may or may not be
defined on the whole space H .
Example 1.4
Consider the space L2 0,1, and the kernel k x, y   sin xy .If f  x  is an element (
function ) in the space then,
1
Kf   sin xy f  y  dy
0

is also in the space. Since,

14
 1

1 1
2
2
Kf   Kf , Kf     sin  xy  f  y  dy dx 
1
2

0 
 0

Using the Cauchy-Schwarz inequality, to have
1 2
1 1 1
2 1 
Kf     sin 2  xy  dy  f y  dy dx     f  y 
2 2
dy   f
0 0 0  0 

It follows that for f in L2 0,1 , Kf is also in L2 0,1 .

Example 1.5: Consider the operator
1
sin  xy 
Kf   f  y  dy . (1.12)
0 y2

For an arbitrary element f  y  in L2 0,1 , the formula (1.94) may not be defined. It will certainly

f y 
2
1

be well defined for all f  y  such that, 

0
y
dy   .

In this case of (1.12) K is not defined on all of L2 0,1 but on a subset of L2 0,1 . This subset is

said to be the domain of K . For linear operators the domain will be a subspace of the Hilbert
space.
*Operators that do not satisfy (1.11) are said to be nonlinear. A general nonlinear integral operator
is of the form,
1
Kf   k  x , y , f  y   dy . (1.13)
0

1 1

Example of such operators are Kf   k  x , y  f

2
 y  dy , Kf   sin  xf  y   dy .
0 0

Definition 1.16 An operator K is said to be bounded, if for some constant M , we have

Kf  M f , . for all f in a Hilbert space H .The greatest lower bound of all M for which

holds, is called the norm of K , and denoted by K .

Kf
Another way of defining K is K  sup
f 0 f

Theorem 1.12 Consider L2 a, b  where, the interval a, b  may be infinite. If,
b b 2

k  x , y  dx dy  M 2  , Then the operator Kf   k  x , y f  y  dy is bounded

b


a a a

Proof: By Cauchy-Schwarz inequality,

15
 1

b b
2
Kf    k  x , y   f y 
2 2
dy dy 
a a 
This leads us to
1

b b b
2
   k x , y   f y 
2 2
Kf dx dy  dy 
a a a 
1 1

b b 2  b 2
    k  x , y  dx dy  .   f  y   
1
2 2
dy   M 2
f M f
2

a a  a 

Hence, we have Kf  M f .

For the continuity of integral operator, we assume two function f1 x , f 2 x  such that,
b b
Kf 1   k  x , y  f 1  y  dy , Kf 2   k  x , y  f 2  y  dy
a a

Hence, we have
b
Kf 1  Kf 2   k  x , y  [f 1  y   f 2  y ] dy
a

For this, we get

1

b b b
2
   k x , y   f1 y f 2 y 
2 2
Kf 1  Kf 2 dx dy  dy 
a a a 
1 1

b b 2  b 2
    k  x , y  dx dy     f 1  y  - f 2  y 
2 2
dy   M f 1  f 2 .
a a  a 

§ 1.6 Compact operator: Let K be a bounded, linear operator on a Hilbert space H , and

f n  be an infinite uniformly bounded sequence in H ; that is for some M , we have f n  M for

all n , K will be said to be a compact operator if from the sequence Kf n  , one can extract a

subsequence Kf n , that is a Cauchy sequence. The sequence Kf n

k k
 converges, of course, since

H is a Hilbert space.

Theorem 1.13: Let H be L2 a, b  , and K a degenerate integral operator,

m b
Kf   ai  x   bi  y  f  y  dy
i 1 a

K is a compact operator, if ai  x  and bi  x  belong to L2 a, b  for all i.

Proof: Consider a uniformly bounded set of functions f n  , hence f n  M .

For this, the above formula takes the form

16
 m b
Kf n   ai  x   bi  y  f n  y  dy .
i 1 a

Clearly,  b  y  f  y  dy
a
i n  bi  y  . f n  M b i  y 

 
b

so that the sequence   bi  y  f n  y  dy  is a uniformly bounded sequence of complex numbers. It

 a 
must therefore have a point of accumulation, and a suitable subsequence will converge to that
 b 
point of a accumulation. Let  
f n 1 be a subsequence of f n  such that,  b1  y  f 1  y  dy 
 a 

converges to some complex number say c1 .By the same argument we can now extract a

 b 
 
subsequence f n   2  
of f n   1 such that   b2  y  f n    y  dy  converges to, say
 a
2


c2 .By extracting

successive subsequence we finally arrive at f n m  with property that,  

b
lim  bi  y  f n m  y  dy  c i i  1, 2, ,m
n m 
a

Where we have the inclusion, f n   f n     f n    

1 2  
 f n m  .

Finally we note that the sequence Kf n   m  is a subsequence of Kf n  , and is also a Cauchy
m
sequence, and lim

Kf n     c i ai  x  .So that K is a compact operator.
m
m
n  i 1

Theorem 1.14: Let K be a compact operator on the Hilbert space, H , and L a bounded
operator. Then both KL and LK are compact operators.
Proof:
Let f n  be a uniformly bounded sequence in H . Since K is compact it contains a

subsequence f n   such that Kf n   is a Cauchy sequence. Since L is bounded, we see that,

LKf n   KLf m   L . Kf n   Kf m  .

Therefore, that LKf n   is also a Cauchy sequence and LK is compact.

Consider the operator KL . If f n  is uniformly bounded, and L is bounded, then Lf n  is again

uniformly bounded. Since K is compact there exists a subsequence such that KLf n   is a Cauchy

sequence. So that KL is also compact. 

17
Theorem 1.15: Let K and L be two compact operators on a Hilbert space H . Then K  L is
again compact, and so is  K for any scalar  .
Proof: Let f n  be a uniformly bounded sequence in H . It certainly contains a subsequence

f n   such that Kf n   is a Cauchy sequence. Now, if  is any scalar then  Kf n   will again be a

Cauchy sequence so that,  K is compact. Now, we can extract from f n   a subsequence f n  

such that both Kf n   and Lf n   are Cauchy sequence.In that case  K  L  f n   is a Cauchy

sequence. So that K  L is also compact. 

Corollary 1.1
Let K k  be a finite sequence of n compact operators on a Hilbert space H , and  k  a set
n
of scalars. Then 
k 1
k K k is compact.

Theorem 1.16: Let K be a compact operator on a Hilbert space H , and let f n  be a linearly

independent sequence of eigenfunctions corresponding to some nonzero eigenvalues  , that is

Kf n   f n , for all n . Then f n  contains a finite number of elements.
Proof: We shall replace the set fn by equivalent set g n , where all g n are again
eigenfunctions but orthonormal so that,
 g n , g m   0, n  m ;  1, n  m

By using Gram-Schmidt process, let,

f1 f 2  f 2 , g1  g1
g1  ; g2 
f1 f 2  f 2 , g1  g1

n 1
f n   f n , g k  g k
gn  k 1
n 1
f n   f n , g k  g k
k 1

The sets f 1 , f 2 , , f n  and g 1 , g 2 , , g n  are subsets and also linearly independent of f n  and

g n  respectively. So both sets f 1 , f 2 , , f n  and g 1 , g 2 , , g n  span the same n dimensional

subspace of H . Let us suppose that f n  not linearly independent.
Now,

gn  gm  gn  g n , g m   g m , g n   g m 2
2 2 2

since  f n  is uniformly bounded, then, g n  is uniformly bounded.

18
If g n  infinite we could select a subsequence g n  such that Kg n  is a Cauchy sequence .
Then, Kg n   Kg m    g n    g m    2.

Since   0 the above cannot be a Cauchy sequence. Kg n   Kg m  could be smaller than  2

for sufficiently large n , m  .

It follows that the number of linearly independent eigenfunctions must be finite. 
Theorem 1.17: Let K n  be a sequence of compact operators on a Hilbert space H . Such that

for some operator K we have, nlim K  K n  0. Then K is also compact.



Proof: Let f n  be any infinite, uniformly bounded sequence. We can select a subsequence f n    1

such that K 1f n    1  is a Cauchy sequence. From the sequence f n     1 we now extract a subsequence

f  n 
2 
such that K 2 f n   2  is a Cauchy sequence. Proceeding in this fashion, we obtain a succession
of subsequences, f n   f n     f n    
1 2  
 f n k   ,such that K k f n i   is a Cauchy

sequence for k  1, 2, , n .

    expect possibly for a

Now, if we take the sequence f n   , which is a subsequence of every f n  
n k

finite number of terms. So that K f  will be a Cauchy sequence for all k .

k n 
n

To prove the operator K is compact, we must show that the sequence Kf  is also a Cauchy n 
n

sequence. Hence, we have

Kf n  n   Kf m  m    Kf n 
n  
 K k f n  n   K k f n n   K k f m m   K k f m m   Kf m m  
  K  K k  f n   K k f n
n m  K k f m m    K k  K  f m m 

 K Kk f n n   K k f n n   K k f m m   K k  K f m m 

if f n    M and k  k   , n   and m  m  are larger than N   so that,

n
n

Kf n n   Kf m m   M     M    2M  1  ; {k  k   , n   ,m   N  }.
n m

Therefore, the sequence Kf n    n  is a Cauchy sequence. Then K is compact. 

Theorem 1.18: Let k  x , y  be continuous for all 0  x , y  1 .The associated integral operator,
1
Kf   k  x , y  f  y  dy is a compact operator on L 2 0,1 .
0

Proof: Since, k  x , y   M ; then Kf  M f i,e. the operator is bounded. We assume

19
that K  1 . Now, we take any uniformly bounded sequence  f n  such that, f n 1

Consider the set of functions g n  x  where g n  Kf n . Now we shall show that there is a Cauchy

subsequence g n   x  that converges to a function in L2 0,1 . For this we select a dense countable

set of points on 0,1 . So, we shall denote these by rk . Now consider the set g n  r1  . Hence,
we have, g n  x  K nf  M f 
n .MIn particular we have, g n  r1   M .So that, the set

g  r 
n 1 has a point of accumulation in  M , M  . Then, we can select a subsequence g n    r1   1 
that converges to b1 .

 
Next, we consider the sequence g n    x  ,which at x  r1 it converges to b1 . Now we examine
1

g n 
1  r2  and select a subsequence g n 
2  r2  that converges to b 2 . We repeat this process at all

rationales. Finally we consider the set of function g n    x  . This sequence converges at all
n

rational. For convenience we shall reined the sequence g n    x  , so that we call it g n  x  .

n

Define a function g  x  by: nlim


g n  rk   g  rk  . So that g  x  is only defined for rational x .
Now, let x1 , x2 be only two rational numbers then,
1
g  x 1   g  x 2   lim g n  x 1   g n  x 2   lim   k  x 1 , y   k  x 2 , y   f n  y  dy
n  n 
0
1

 1
2
 k x 1, y   k x 2 , y 
2
 lim  dy  fn
n 
 0 
1

 1
2
 k x 1, y   k x 2 , y 
2
 dy 
 0 

Moreover, since k x, y  is continuous so it g  x  .

Now we can define g  x  for all 0  x  1. Let x be arbitrary and xk a sequence of
rational numbers such that lim xk  x then, g  x   klim

g x k 
k 

The above g  x  is well defined. To see this, suppose that x k  is a second sequence of rational

converging to x ( klim xk  x ) and suppose that, lim g  x k   g  x 

 k 

Then,
g x   g x   g x   g x k   g x k   g x k   g x k   g x 
 g  x   g  x k   g  x k   g  x k   g  x k   g  x   3 , for k  k   

21
Which leads us to the result, g  x   g  x  .

Now it follows that g  x  is well defined as a continuous function for all 0  x  1 . Being

continuous on closed and bounded interval it is also uniformly continuous and this certainly
belongs to L2 0,1 . It follows that the operator K is indeed compact. 

k x, y  be such that,  

1 1
k x , y 
2
Theorem 1.19: Let dx dy  .
0 0

Then the operator, Kf   k  x , y  f  y  dy is a compact operator on L2 0,1 .

0

Proof: We consider L 2 kernel k  x , y  as a limit of continuous kernel such that,

1 1 2

lim   k  x , y   k n  x , y  dx dy  0
n 
0 0

1 1 2 1 1

 k  x , y  dx dy  lim   k n x , y 
2
and defined, dx dy .
n 
0 0 0 0

Now,
1
1 1


1
2

K  K n f   k  x , y   k n  x , y   f  y  dy   k x , y   k n x , y   f y 
2 2
  dy dy 
0 0 
 0 

1



1 1
2

 k x , y   k n x , y 
2
 dx dy  f

 0 0 

1



1 1
2

 k x , y   k n x , y 
2
Then, K Kn  dx dy  .

 0 0 


Hence, we have lim K  K n  0. Then, K is compact. 

n 

References

[1] M. Bernkopf, The development of function space to with particular reference to their
origins in integral equation theory. Arch. Hist. Exact. Sci. Vol. 3 (1966) 1- 96.
[2] C.D. Green, Integral Equation Methods, New York, 1969.
[3] H. Hochstadt, Integral Equations, Awelye Inter Science Publication, New York, 1971.
[4] M. A. Golberg. ed. Solution Methods for Integral Equations, New York, 1979.
[5] F.G. Tricomi, Integral Equations, Dover, New York, 1985.
[6] T.A. Burton, Volterra Integral and Differential Equations, London, New York, 1983.

21
[7] R. P. Knawel, Linear Integral Equations Theory and Technique, Boston, 1996.
[8] P. Schiavone, C. Constanda and A. Mioduchowski, Integral Methods in Science and
Engineering, Birkhauser Boston, 2002.
[9] N.I. Muskhelishvili, Singular Integral Equations, Noordhoff, Groningen, The Netherland,
1953.
[10] Peter Linz, Analytic and Numerical Methods for Volterra Equations, SIAM, Philadelphia,
1985.
[11] K.E. Atkinson, A Survey of Numerical Method for the Solution of Fredholm Integral
Equation of the Second Kind, Philadelphia, 1976.
[12] K.E. Atkinson, The Numerical Solution of Integral Equation of the Second Kind, Cambridge
University, Cambridge, 1997.
[13] Christopher T.H. Baker, Treatment of Integral Equations by Numerical Methods , Academic
Press, 1982.
[14] L.M. Delves and J.L. Mohamed, Computational Methods for Integral Equations, New York,
London, 1985.
[15] M.A.Golberg.ed, Numerical Solution for Integral Equations, New York, 1990.
[16] M.A. Abdou, Fredholm‫ــ‬Volterra integral equation of the first kind and contact problem, J.
Appl. Math. Compute. 125 (2002) 177‫ــ‬193.
[17] M.A. Abdou, Fredholm‫ــ‬Volterra integral equation and generalized potential kernel, J. Appl.
Math. Compute. 131 (2002) 81‫ــ‬94.
[18] M.A. Abdou, On asymptotic methods for Fredholm‫ــ‬Volterra integral equation of the second
kind in contact problem, J. Comp. Appl. Math. 154 (2003) 431‫ــ‬446.
[19] M.A. Abdou, Fredholm‫ــ‬Volterra integral equation with singular kernel, J. Appl. Math.
Compute. 137 (2003) 231‫ــ‬243.
[20] M.A. Abdou, F.A. Salama, Volterra‫ــ‬Fredholm integral equation of the first kind and spectral
relationships, Appl. Math. Compute. 153 (2004) 141‫ــ‬153.
[21] M.A. Abdou, O.L. Mustafa, Fredholm‫ــ‬Volterra integral equation in contact problem, J.
Appl. Math. Compute. 138 (2003) 199‫ــ‬215.
[22] M.A. Abdou, A.A. Nasr, On the numerical treatment of the singular integral eqution of the
second kind, J. Appl. Math. Compute. 146 (2003) 373‫ــ‬380.
[23] M.A. Abdou, Fredholm integral equation with potential kernel and its structure resolvent,
Appl. Math. Compute . 107 (2000) 169 ‫ ـ‬180 .

22
 [24] I.S. Gradshteyn and I.M. Ryzhik, Table of Integrals, Series and Products, Academic Press,
New York, 1980 .
[25] H. Bateman, A. Erdely, Higher Transcendental Functions, Vol.2, Nauka Moscow 1973 .
[26] S.M. Mkhitarian, M.A .Abdou, On different methods for solving the integral eqution of the
first kind with logarithmic kernel, Dokl. Acad. Hauk. Armenia 90 (1990) 1‫ــ‬10.
[27] E.V. Covalence, Some approximate methods of solving integral equations of mixed problem,
Appl. Math. Mech. 53 (1) (1989) 85‫ ـ‬92 .

23

View publication stats