phdQusuayAlqifiarydiferencne PDF

UNIVERZITET U BEOGRADU
MATEMATICKI FAKULTET
Qusuay Hatim Eghaar Alqifiary

STABILNOST RESENJA DIFERENCIJALNIH

JEDNACINA U SMISLU HAJERSA I ULAMA
DOKTORSKA DISERTACIJA
BEOGRAD, 2015
UNIVERSITY OF BELGRADE
FACULTY OF MATHEMATICS
Qusuay Hatim Eghaar Alqifiary
HYERS-ULAM STABILITY OF THE

SOLUTIONS OF DIFFERENTIAL
EQUATIONS
DOCTORAL DISSERTATION
BELGRADE, 2015
Abstract
This thesis has been written under the supervision of my mentor Prof. dr. Julka
Knezevic-Miljanovic at the University of Belgrade in the academic year 2014-2015.
The aim of this study is to investigate Hyers-Ulam stability of some types of
dierential equations, and to study a generalized Hyers-Ulam stability and as
well as a special case of the Hyers-Ulam stability problem, which is called the
superstability. Therefore, when there is a dierential equation, we answer the
three main questions:
1- Does this equation have Hyers -Ulam stability?
2- What are the conditions under which the dierential equation has stability ?
3- What is a Hyers-Ulam constant of the dierential equation?
The thesis is divided into three chapters. Chapter 1 is divided into 3 sections. In
this chapter, we introduce some sucient( conditions ) under which each solution
of the linear dierential equation u (t) + 1 + (t) u(t) = 0 is bounded. Apart
from this we prove the Hyers-Ulam stability of it and the nonlinear dierential
equations of the form u (t) + F (t, u(t)) = 0, by using the Gronwall lemma and
we prove the Hyers-Ulam stability of the second-order linear dierential equations
with boundary conditions. In addition to that we establish the superstability
of linear dierential equations of second-order and higher order with continuous
coecients and with constant coecients, respectively. Chapter 2 is divided into
2 sections. In this chapter, by using the Laplace transform method, we prove

n1
that the linear dierential equation of the nth-order y (n) (t) + k y (k) (t) = f (t)
k=0
has the generalized Hyers-Ulam stability. And we prove also the Hyers-Ulam-
Rassias stability of the second-order linear dierential equations with initial and
boundary conditions, as well as linear dierential equations of higher order in the
form of y (n) (x) + (x)y(x) = 0, with initial conditions. Furthermore, we establish
the generalized superstability of dierential equations of nth-order with initial
conditions and investigate the generalized superstability of dierential equations
of second-order in the form of y (x)+p(x)y (x)+q(x)y(x) = 0. Chapter 3 is divided
into 2 sections. In this chapter, by applying the fixed point alternative method,
we give a necessary and sucient condition in order that the first order linear
Alqifiary Abstract ii
system of dierential equations z(t)

+ A(t)z(t) + B(t) = 0 has the Hyers-Ulam-
Rassias stability and find Hyers-Ulam stability constant under those conditions.
In addition to that, we apply this result to a second-order dierential equation
y(t) + f (t)y(t)
+ g(t)y(t) + h(t) = 0. Also, we apply it to dierential equations
with constant coecient in the same sense of proofs. And we give a sucient
condition in order that the first order nonlinear system of dierential equations
has Hyers-Ulam stability and Hyers-Ulam-Rassias stability. In addition, we present
the relation between practical stability and Hyers-Ulam stability and also Hyers-
Ulam-Rassias stability.
Scientific field (naucna oblast): Mathematics (matematika)(34A40, 34A12, 34D10)

Narrow scientific field (uza naucna oblast): Dierential Equations (Diferencijalne
jednacine)
UDC:517.937(043.3)
Podaci o mentoru i
clanovima
komisije:
MENTOR:
redovni profesor dr Julka Knezevic-Miljanovic
Matematicki fakultet,
Univerzitet u Beogradu
CLANOVI KOMISIJE :
redovni profesor dr Bosko Jovanovic
redovni profesor dr Miodrag Spalevic

Datum odbrane:
Acknowledgements
I would like to express my special appreciation and thanks to my supervisor

Professor Julka Knezevic-Miljanovic, for all the help she has given over the last
three years. Her advice on my research has been priceless.
I would like to acknowledge those who have shaped my mathematical education.

The sta of professors at the department of mathematics.
A special thanks to my family and friends, for their full support, patience and
understanding.
To each of the above, I express my deepest gratitude.
Qusuay Alqifiary
Introduction
This subject dates back to the talk given by the Polish-American mathematician
Ulam at the University of Wisconsin in 1940 (see [40]). In that talk, Ulam asked
whether an approximate solution of a functional equation must be near an exact
solution of that equation. This asking of Ulam is stated as follows:
Theorem 0.0.1. Let G1 be a group and let G2 be a metric group with a metric
d(., .). Given > 0, does there exist a > 0 such that if a function h : G1 G2
satisfies the inequality d(h(xy).h(x)h(y)) < for all x, y G1 , then there is a
homomorphism H : G1 G2 with d(h(x), H(x)) < for all x G1 ?
One year later, a partial answer to this question was given by D. H. Hyers [5] for
additive functions defined on Banach spaces:
Theorem 0.0.2. Let f : X1 X2 be a function between Banach spaces such that
f (x + y) f (x) f (y) ,
for some > 0 and for all x, y X1 . Then the limit
A(x) = limn 2n f (2n x)
exists for each x X1 , and A : X1 X2 is the unique additive function such that
f (x) A(x)
for every x X1 . Moreover, if f (tx) is continuous in t for each fixed x X1 , then

the function A is linear.
This result is called the Hyers-Ulam Stability of additive Cauchy equation g(x +
y) = g(x)+g(y). After Hyerss result, many mathematicians have extended Ulams
Alqifiary Introduction vi
problem to other functional equations and generalized Hyerss result in various

directions (see [6, 32, 38, 44]).
Ten years after the publication of Hyerss theorem, D. G. Bourgin extended the
theorem of Hyers and stated it in his paper [4] without proof. Unfortunately, it
seems that this result of Bourgin failed to receive attention from mathematicians
at that time. No one has made use of this result for a long time.
In 1978, Rassias [44] introduced a new functional inequality that we call Cauchy-
Rassias inequality and succeeded in extending the result of Hyers, by weakening
the condition for the Cauchy dierences to unbounded map as follows:
Theorem 0.0.3. Let f : X1 X2 be a function between Banach spaces. If f

satisfies the functional inequality
f (x + y) f (x) f (y) (xp + yp )
for some 0, p with 0 p < 1 and for any x, y X1 , then there exists a unique
additive function A : X1 X2 such that
2
f (x) A(x) xp
2 2p
for each x X1 . If, in addition, f (tx) is continuous in t for each fixed x X1 ,
then the function A is linear.
The stability phenomenon of this kind is called the Generalized Hyers-Ulam Sta-
bility (or Hyers-Ulam-Rassias stability). A generalization of Ulams problem was
recently proposed by replacing functional equations with dierential equations:
The dierential equation (f, y, y , . . . , y (n) ) = 0 has Hyers-Ulam stability if for
given > 0 and a function y such that |(f, y, y , . . . , y (n) )| , there exists
a solution ya of the dierential equation such that |y(t) ya (t)| K() and
lim0 K() = 0. If the preceding statement is also true when we replace and
K() by (t) and (t), where , are appropriate functions not depending on y
and ya explicitly, then we say that the corresponding dierential equation has the
generalized Hyers-Ulam stability (or Hyers-Ulam-Rassias stability).
Obloza seems to be the first author who has investigated the Hyers-Ulam stability
of linear dierential equations (see [21, 22]). Thereafter, Alsina and Ger published
their paper [3], which handles the Hyers-Ulam stability of the linear dierential
equation y (t) = y(t): If a dierentiable function y(t) is a solution of the inequality
|y (t) y(t)| for any t (a, ), then there exists a constant c such that
|y(t) cet | 3 for all t (a, ). Since then, this problem now known as
the problem of Hyers-Ulam stability - has been extensively investigated for the
algebraic, functional, dierential, integral, and operator equations.
Alqifiary Introduction vii
Those previous results were extended to the Hyers-Ulam stability of linear dier-
ential equations of first order [3337, 42, 43].
Rus investigated the Hyers-Ulam stability of dierential and integral equations
using the Gronwall lemma and the technique of weakly Picard operators (see [13,
14]). Recently, The results given in [36, 42, 45] have been generalized by Cimpean
and Popa [10] and by Popa and Rasa [8, 9] for the linear dierential equations.
In 1979, J.Baker, J. Lawrence and F. Zorzitto[16] proved a new type of stability
of the exponential equation f (x + y) = f (x)f (y). More precisely, they proved
that if a complex-valued mapping f defined on a normed vector space satisfies
the inequality |f (x + y) f (x)f (y)| for some given > 0 and for all x, y,
then either f is bounded or f is exponential. Such a phenomenon is called the
superstability of the exponential equation, which is a special kind of Hyers-Ulam
stability. It seems that the results of P. Gavruta, S. Jung and Y. Li [23] are the
earliest one concerning the superstability of dierential equations.
This thesis is about stability of some types of dierential equations, where we
introduce this thesis in three chapters.
Chapter one is titled by Hyers-Ulam stability of Dierential Equations. This
chapter consists of three sections. In section 1.1, we introduce some sucient
conditions under which each solution of the linear dierential equation (1.1.2) is
bounded. As well as we prove the Hyers-Ulam stability of the linear dierential
equations of the form (1.1.2). In section 1.2, we prove the Hyers-Ulam stability
of the nonlinear dierential equations of the form (1.2.1) by using the Gronwall
lemma. In section 1.3, we prove the Hyers-Ulam stability of the second-order linear
dierential equations with boundary conditions. Furthermore, the superstability
of linear dierential equations with constant coecients.
Chapter two is titled by Generlaized Hyers-Ulam stability of Dierential equa-
tions. This chapter consists of two sections. In section 2.1, by using the Laplace
transform method, we prove that the linear dierential equation of the nth-order

n1
y (n) (t) + k y (k) (t) = f (t)
k=0
has the generalized Hyers-Ulam stability, where k is a scalar, y and f are n times
continuously dierentiable and of exponential order, respectively. In section 2.2,
we establish the generalized superstability of dierential equations of nth-order
with initial conditions and investigate the generalized superstability of dieren-
tial equations of second-order in the form of y (x) + p(x)y (x) + q(x)y(x) = 0.
In additional, we prove the Hyers-Ulam-Rassias stability of the second-order lin-
ear dierential equations with initial and boundary conditions as well as linear
dierential equations of higher order in the form of
y (n) (x) + (x)y(x) = 0,
Alqifiary Introduction viii
with initial conditions

y(a) = y (a) = = y (n1) (a) = 0,
where n N+ , y C n [a, b], C 0 [a, b], < a < b < +.

Chapter three is titled by Hyers-Ulam stability of system of dierential equa-
tions. This chapter consists of two sections. In section 3.1, by applying the fixed
point alternative method, we give a necessary and sucient condition in order that
the first order linear system of dierential equations z(t)
+ A(t)z(t) + B(t) = 0
has the Hyers-Ulam-Rassias stability and find Hyers-Ulam stability constant un-
der those conditions. In addition to that, we apply this result to a second order
dierential equation y(t) + f (t)y(t)
+ g(t)y(t) + h(t) = 0. Also, we apply it to dif-
ferential equations with constant coecient in the same sense of proofs. In section
3.2, we give a sucient condition in order that the first order nonlinear system of
dierential equations has Hyers-Ulam stability and Hyers-Ulam-Rassias stability.
In addition, we present the relation between practical stability and Hyers-Ulam
stability and also Hyers-Ulam-Rassias stability.
Contents
Abstract i
Acknowledgements iv
Introduction v
1 Hyers-Ulam Stability of Dierential Equations 1

1.1 Hyers-Ulam Stability of Linear Dierential Equations . . . 1
1.1.1 Boundedness of Solutions of a Second Order Dierential
Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Hyers-Ulam Stability of Linear Dierential Equations of
Second Order . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Hyers-Ulam Stability of Nonlinear Dierential Equations
of Second Order . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Hyers-Ulam Stability of Dierential Equations with Bound-
ary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Generlaized Hyers-Ulam Stability of Dierential Equations 29

2.1 Generlaized Hyers-Ulam Stability of Linear
Dierential Equations . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Generlaized Hyers-Ulam Stability of Dierential Equations
with boundary Conditions . . . . . . . . . . . . . . . . . . . . . 40
2.2.1 Generalized Superstability of Dierential Equations with
Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.2 Hyers-Ulam-Rassias Stability of Linear Dierential Equa-
tions with Boundary Conditions . . . . . . . . . . . . . . 43
3 Hyers-Ulam Stability of System of Dierential Equations 47

Alqifiary Contents x
3.1 Hyers-Ulam Stability of Linear System of Dierential Equa-

tions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1.1 Preliminaries and Auxiliary Results . . . . . . . . . . . . 47
3.1.2 Hyers-Ulam Stability of First Order System of Dieren-
tial Equations . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.1.3 Hyers-Ulam Stability of Second Order Dierential Equa-
tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Hyers-Ulam Stability of Nonlinear System of Dierential
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2.1 Preliminaries and Auxiliary Results . . . . . . . . . . . . 57
3.2.2 Hyers-Ulam Stability of System of Dierential Equations 58
List of papers 62
Bibliography 63
Chapter 1
Hyers-Ulam Stability of
Dierential Equations
1.1 Hyers-Ulam Stability of Linear Dierential

Equations
1.1.1 Boundedness of Solutions of a Second Order Dierential Equation
In this subsection, we first introduce and prove a lemma which is a kind of the
Gronwall inequality.
Lemma 1.1.1. [28] Let u, v : [0, ) [0, ) be integrable functions, c > 0 be a

constant, and let t0 0 be given. If u satisfies the inequality
t
u(t) c + u( )v( )d (1.1.1)
t0
for all t t0 , then ( )

t
u(t) c exp v( )d
t0
for all t t0 .
Proof. It follows from (1.1.1) that
u(t)v(t)
t v(t)
c + t0 u( )v( )d
Alqifiary Hyers-Ulam Sability of The Solutions of Dierential Equations 2
for all t t0 . Integrating both sides of the last inequality from t0 to t, we obtain
( t ) t
ln c + u( )v( )d ln c v( )d
t0 t0
or t ( t )
c+ u( )v( )d c exp v( )d
t0 t0
for each t t0 , which together with (1.1.1) implies that

( t )
u(t) c exp v( )d
t0
for all t t0 .
In the following theorem, using Lemma 1.1.1, we investigate sucient conditions

under which every solution of the dierential equation
( )
u (t) + 1 + (t) u(t) = 0 (1.1.2)
is bounded.
Theorem 1.1.2. [28] Let : [0, ) R be a dierentiable function. Every solu-

tion u : [0, ) R of the linear dierential equation (1.1.2) is bounded provided

that 0 | (t)|dt < and (t) 0 as t .
Proof. First, we choose t0 large enough so that 1 + (t) 1/2 for all t t0 .
Multiplying (1.1.2) by u (t) and integrating it from t0 to t, we obtain
t
1 2 1
u (t) + u(t) + 2
( )u( )u ( )d = c1
2 2 t0
for all t t0 . Integrating by parts, this yields

1 2 1 1 1 t
u (t) + u(t) + (t)u(t)
2 2
( )u( )2 d = c2 (1.1.3)
2 2 2 2 t0
for any t t0 . Then it follows from (1.1.3) that

1 1 1 1 1 1( )
u(t)2 u (t)2 + u(t)2 u (t)2 + 1 + (t) u(t)2
4 2 2 2 2 2
1 t
= c2 + ( )u( )2 d
2 t0
for all t t0 . Thus, it holds that

t t

u(t) 4c2 + 2
2
( )u( ) d 4|c2 | + 2
2
| ( )|u( )2 d (1.1.4)
t0 t0
for any t t0 .
In view of Lemma 1.1.1, (1.1.4) and our hypothesis, there exists a constant M1 > 0
such that ( t )

u(t) 4|c2 | exp
2
2| ( )|d < M12
t0
for all t t0 . On the other hand, since u is continuous, there exists a constant
M2 > 0 such that |u(t)| M2 for all 0 t t0 , which completes the proof.
Corollary 1.1.3. [28] Let : [0, ) R be a dierentiable function satisfying

(t) 1 as t . Every solution u : [0, ) R of the linear dierential
equation
u (t) + (t)u(t) = 0 (1.1.5)

is bounded provided 0
| (t)|dt < .
1.1.2 Hyers-Ulam Stability of Linear Dierential Equations of Second Or-
der
Given constants L > 0 and t0 0, let U (L; t0 ) denote the set of all functions
u : [t0 , ) R with the following properties:
(i) u is twice continuously dierentiable;
(ii) u(t0 ) = u (t0 ) = 0;

(iii) t0 |u ( )|d L.
We now prove the Hyers-Ulam stability of the linear dierential equation (1.1.2)
by using the Gronwall inequality.
Theorem 1.1.4. [28] Given constants L > 0 and t0 0, assume that

: [t0 , ) R is a dierentiable function with C := t0 | ( )|d < and
:= inf tt0 (t) > 1. If a function u U (L; t0 ) satisfies the inequality
( )
u (t) + 1 + (t) u(t) (1.1.6)
for all t t0 and for some 0, then there exist a solution u0 U (L; t0 ) of the
dierential equation (1.1.2) and a constant K > 0 such that

|u(t) u0 (t)| K (1.1.7)
for any t t0 , where

( )
2L C
K := exp .
1+ 2(1 + )
Proof. We multiply (1.1.6) with |u (t)| to get
|u (t)| u (t)u (t) + u(t)u (t) + (t)u(t)u (t) |u (t)|
for all t t0 . If we integrate each term of the last inequalities from t0 to t, then
it follows from (ii) that
t t t
1 2 1
|u ( )|d u (t) + u(t) + 2
( )u( )u ( )d |u ( )|d
t0 2 2 t0 t0
for any t t0 .
Integrating by parts and using (iii), we have
t
1 1 1 1
L u (t)2 + u(t)2 + (t)u(t)2 ( )u( )2 d L (1.1.8)
2 2 2 2 t0
for all t t0 .
Since 1 + > 0 holds for all t t0 , it follows from (1.1.8) that
1+ 1 1+ 1 1( )
u(t)2 u (t)2 + u(t)2 u (t)2 + 1 + (t) u(t)2
2 2 2 2 2
1 t
L + ( )u( )2 d
2 t0

1 t
L + | ( )|u( )2 d
2 t0
or t
2L 1
u(t)
2
+ | ( )|u( )2 d
1+ 1+ t0
for any t t0 .
Applying Lemma 1.1.1, we obtain
( t ) ( )
2L 1 2L C
u(t) 2
exp | ( )|d exp
1+ 1 + t0 1+ 1+
for all t t0 . Hence, it holds that

( )
C 2L
|u(t)| exp
2(1 + ) 1+
for any t t0 . Obviously, u0 (t) 0 satisfies the equation (1.1.2) and the conditions
(i), (ii), and (iii) such that

|u(t) u0 (t)| K
( )
for all t t0 , where K = 1+ 2L C
exp 2(1+) .
If we set (t) := 1+(t), then the following corollary is an immediate consequence

of Theorem 1.1.4.
Corollary 1.1.5. [28] Given constants L > 0 and t0 0, assume that

: [t0 , ) R is a dierentiable function with C := t0 | ( )|d < and
:= inf tt0 (t) > 0. If a function u U (L; t0 ) satisfies the inequality

u (t) + (t)u(t)
for all t t0 and for some 0, then there exist a solution u0 U (L; t0 ) of the
dierential equation (1.1.5) and a constant K > 0 such that

|u(t) u0 (t)| K
( C ) 2L
for any t t0 , where K := exp 2
.
Example 1.1.1. [28] Let : [0, ) R be a constant function defined by

(t) := a for all t 0 and for a constant a > 0. Then, we have

C = 0 | ( )|d = 0 and = inf t0 (t) = a. Assume that a twice continuously

dierentiable function u : [0, ) R satisfies u(0) = u (0) = 0, 0 |u ( )|d L,
and

u (t) + (t)u(t) = u (t) + au(t)
for all t 0 and for some 0 and L > 0. According to Corollary 1.1.5, there
exists a solution u0 : [0, ) R of the dierential equation, y (t) + ay(t) = 0,
such that

2L
|u(t) u0 (t)|
a
for any t 0.
Indeed, if we define a function u : [0, ) R by
2
u(t) := 2
cos at + 2
sin at ,
(t + 1) a(t + 1)

a
where we set = L, then u satisfies the conditions stated in the first part
a+ a+2
of this example, as we see in the following. It follows from the definition of u that
( ) ( )
2 2 a 4
u (t) = 2
3
cos at 2
+ 3
sin at
(t + 1) (t + 1) (t + 1) a(t + 1)
and hence, we get u(0) = u (0) = 0. Moreover, we obtain
( )
2 + a 4
u (t) + 2
(t + 1) 2 a (t + 1)3
and
( )

2+ a 4
u ( )d = d + 2 d
( + 1)2 a ( + 1)3
0 0
( 0
)
( ) 2
= 2+ a + 1
a
= L.

For any given > 0, if we choose the constant such that 0 <
a

a a+4a+2 a+12
,
then we can easily see that

u (t) + au(t)
( )
8 6
+ cos at
(t + 1) 3 (t + 1) 4
( )
4 a 1 12
+ + sin at a
(t + 1)3 a (t + 1)4
( ) ( )
8 6 4 a 1 12
+ + + a
(t + 1)3 (t + 1)4 (t + 1)3 a (t + 1)4

a a + 4a + 2 a + 12
=
a

for any t 0.
Theorem 1.1.6. [28] Given constants L > 0 and t0 0, assume that

: [t0 , ) (0, ) is a monotone increasing and dierentiable function. If a
function u U (L; t0 ) satisfies the inequality (1.1.6) for all t t0 and for some
> 0, then there exists a solution u0 U (L; t0 ) of the dierential equation (1.1.2)
such that

2L
|u(t) u0 (t)| (1.1.9)
(t0 )
for any t t0 .
|u (t)| u (t)u (t) + u(t)u (t) + (t)u(t)u (t) |u (t)|
t t t
1 2 1
|u ( )|d u (t) + u(t) + 2
( )u( )u ( )d |u ( )|d
t0 2 2 t0 t0
for any t t0 .
Integrating by parts, the last inequalities together with (iii) yield

1 2 1 1 1 t
L u (t) + u(t) + (t)u(t)
2 2
( )u( )2 d L
2 2 2 2 t0
for all t t0 . Then we have
t
1 1 t ( ) ( )
(t)u(t)
2
( )u( ) d + L L +
2
u( )2 d
2 2 t0 t0 ( ) 2
for any t t0 .
( )
1 t
( ) (t)
(t)u(t)2 L exp d = L
2 t0 ( ) (t0 )
for all t t0 , since : [t0 , ) (0, ) is a monotone increasing function. Hence,
it holds that
2L
|u(t)|
(t0 )
for any t t0 . Obviously, u0 (t) 0 satisfies the equation (1.1.2), u0 U (L; t0 ),
as well as the inequality (1.1.9) for all t t0 .

: [t0 , ) (1, ) is a monotone increasing and dierentiable function with
(t0 ) = 2. If a function u U (L; t0 ) satisfies the inequality

u (t) + (t)u(t)
for all t t0 and for some > 0, then there exists a solution u0 U (L; t0 ) of the
dierential equation (1.1.5) such that

|u(t) u0 (t)| 2L
for any t t0 .
If we set (t) := (t), then the following corollary is an immediate consequence

of Theorem 1.1.6.

: [t0 , ) (, 0) is a monotone decreasing and dierentiable function with
(t0 ) = 1. If a function u U (L; t0 ) satisfies the inequality
( )
u (t) + 1 (t) u(t)
for all t t0 and for some > 0, then there exists a solution u0 U (L; t0 ) of the
dierential equation
( )
u (t) + 1 (t) u(t) = 0
such that

|u(t) u0 (t)| 2L
for any t t0 .
Example 1.1.2. [28] Let : [0, ) (, 0) be a monotone decreasing function

defined by (t) := et 2 for all t 0. Then, we have (0) = 1. Assume that a
twice continuously dierentiable function u : [0, ) R satisfies u(0) = u (0) = 0,

0
|u ( )|d L, and
( ) ( )
u (t) + 1 (t) u(t) = u (t) + 3 et u(t)
for all t 0 and for some > 0 and L > 0. According to Corollary 1.1.8, there
exists a solution u0 : [0, ) R of the dierential equation,
( )
y (t) + 3 et y(t) = 0, such that

|u(t) u0 (t)| 2L
for any t 0.
Indeed, if we define a function u : [0, ) R by
1
u(t) := sin t + cos t ,
(t + 1)3 2 (t + 1)2 2
where is a real number with || 2
43
, then u satisfies the conditions stated
in the first part of this example, as we see in the following. It follows from the
definition of u that
3 1
u (t) = 4
sin t sin t
(t + 1) 2 (t + 1)2
and hence, we get u(0) = u (0) = 0. Moreover, we obtain

u (t) 3|| + 1 ||
(t + 1)4 2 (t + 1)2
and

3||
1 ||
u ( )d d + d =: L < .
0 0 ( + 1)4 0 2 ( + 1)2
We can see that
( )
u (t) + 3 et u(t)

12 3 ( )
5
sin t 4
cos t + 4 et sin t
(t + 1) (t + 1) (t + 1)3

2 et 3 et
+ cos t
2 (t + 1)2 2
12|| 3|| 4|| || 3
+ + + + ||
(t + 1)5 (t + 1)4 (t + 1)3 (t + 1)2 2
43
||
2

for any t 0.
1.2 Hyers-Ulam Stability of Nonlinear Dieren-

tial Equations of Second Order
In the following theorems, we investigate the Hyers-Ulam stability of the nonlinear
dierential equation
u (t) + F (t, u(t)) = 0. (1.2.1)
Theorem 1.2.1. [28] Given constants L > 0 and t0 0, assume that F : [t0 , )
R (0, ) is a function satisfying F (t, u(t))/F (t, u(t)) > 0 and F (t, 0) = 1 for
all t t0 and u U (L; t0 ). If a function u : [t0 , ) [0, ) satisfies u U (L; t0 )
and the inequality

u (t) + F (t, u(t)) (1.2.2)
for all t t0 and for some > 0, then there exists a solution u0 : [t0 , ) [0, )
of the dierential equation (1.2.2) such that
|u(t) u0 (t)| L
for any t t0 .
|u (t)| u (t)u (t) + F (t, u(t))u (t) |u (t)|
t t t
1 2
|u ( )|d u (t) + F (, u( ))u ( )d |u ( )|d
t0 2 t0 t0
for any t t0 .
Integrating by parts and using (iii), the last inequalities yield
t
1 2
L u (t) + F (t, u(t))u(t) F (, u( ))u( )d L
2 t0
for all t t0 . Then we have

t
F (t, u(t))u(t) L + F (, u( ))u( )d
t
0t
F (, u( ))
L + F (, u( ))u( )d
t0 F (, u( ))
for any t t0 .
( )
t
F (, u( ))
F (t, u(t))u(t) L exp d = LF (t, u(t))
t0 F (, u( ))
for all t t0 . Hence, it holds that |u(t)| L for any t t0 . Obviously, u0 (t) 0
satisfies the equation (1.2.1) and u0 U (L; t0 ) such that
|u(t) u0 (t)| L
for all t t0 .
In the following theorem, we investigate the Hyers-Ulam stability of the Emden-

Fowler nonlinear dierential equation of second order
u (t) + h(t)u(t) = 0 (1.2.3)
for the case where is a positive odd integer.
Theorem 1.2.2. [28] Given constants L > 0 and t0 0, assume that h : [t0 , )
(0, ) is a dierentiable function. Let be an odd integer larger than 0. If a
function u : [t0 , ) [0, ) satisfies u U (L; t0 ) and the inequality

u (t) + h(t)u(t) (1.2.4)
for all t t0 and for some > 0, then there exists a solution u0 : [t0 , ) [0, )
of the dierential equation (1.2.3) such that
( )1/
L
|u(t) u0 (t)|
h(t0 )
for any t t0 , where := + 1.
|u (t)| u (t)u (t) + h(t)u(t) u (t) |u (t)|
t t t
1 2
|u ( )|d u (t) + h( )u( ) u ( )d |u ( )|d
t0 2 t0 t0
for any t t0 .
Integrating by parts and using (iii), the last inequalities yield
t
1 2 u(t)+1 u( )+1
L u (t) + h(t) h ( ) d L
2 +1 t0 +1
for all t t0 . for all t t0 . Then we have

t
u(t)+1 u( )+1
h(t) L + h ( ) d
+1 t0 +1
t
h ( ) u( )+1
L + h( ) d
t0 h( ) +1
for any t t0 .
( )
u(t)+1 t
h ( ) h(t)
h(t) L exp d L
+1 t0 h( ) h(t0 )
for all t t0 , from which we have
( + 1)L
u(t)+1
h(t0 )

( )1/
L
|u(t)|
h(t0 )
for any t t0 , where we set = + 1. Obviously, u0 (t) 0 satisfies the equation

(1.2.3) and u0 U (L; t0 ). Moreover, we get
( )1/
L
|u(t) u0 (t)|
h(t0 )
for all t t0 .
Given constants L 0, M > 0, and t0 0, let U (L; M ; t0 ) denote the set of all
functions u : [t0 , ) R with the following properties:
(i ) u is twice continuously dierentiable;
(ii ) u(t0 ) = u (t0 ) = 0;

(iii ) |u(t)| L for all t t0 ;

(iv ) t0 |u ( )|d M for all t t0 .
We now investigate the Hyers-Ulam stability of the dierential equation of the

form
u (t) + u(t) + h(t)u(t) = 0, (1.2.5)
where is a positive odd integer.
Theorem 1.2.3. [28] Given constants L 0, M > 0 and t0 0, assume that

h : [t0 , ) [0, ) is a function satisfying C := t0 |h ( )|d < . Let be an
odd integer larger than 0. If a function u U (L; M ; t0 ) satisfies the inequality

u (t) + u(t) + h(t)u(t) (1.2.6)
for all t t0 and for some > 0, then there exists a solution u0 : [t0 , ) R of
the dierential equation (1.2.5) such that
( )
CL1
|u(t) u0 (t)| 2M exp
+1
for any t t0 .
|u (t)| u (t)u (t) + u(t)u (t) + h(t)u(t) u (t) |u (t)|
it follows from (ii ) that
t t t
1 2 1
|u ( )|d u (t) + u(t) + 2
h( )u( ) u ( )d |u ( )|d
t0 2 2 t0 t0
for any t t0 .
Integrating by parts and using (ii ) and (iv ), the last inequalities yield
t
1 2 1 1 1
M u (t) + u(t) + h(t)
2
u(t) +1
h ( )u( )+1 d M
2 2 +1 + 1 t0
for all t t0 . Then it follows from (iii ) that

t
1 1
u(t) M +
2
h ( )u( )+1 d
2 + 1 t0
t
2 1
M + u( )2 h ( )u( )1 d
+ 1 t0 2
t
2 1
M + u( )2 |h ( )||u( )|1 d
+ 1 t0 2

2L1 t 1
M + u( )2 h ( )d
+ 1 t0 2
for any t t0 .
( t 1 ) ( )
1 2L 2CL1
u(t) M exp
2
|h ( )|d M exp
2 t0 + 1 +1

( )
CL1
|u(t)| 2M exp
+1
for any t t0 . Obviously, u0 (t) 0 satisfies the equation (1.2.5) and u0

U (L; M ; t0 ). Furthermore, we get
( )
CL1
|u(t) u0 (t)| 2M exp
+1
for all t t0 .
1.3 Hyers-Ulam Stability of Dierential Equa-

tions with Boundary Conditions
Lemma 1.3.1. [29] Let I = [a, b] be a closed interval with < a < b < . If
y C 2 (I, R) and y(a) = 0 = y(b), then
(b a)2
max |y(x)| max |y (x)|.
xI 8 xI
Proof. Let M := max |y(x)|. Since y(a) = 0 = y(b), there exists x0 (a, b) such
xI
that |y(x0 )| = M . By the Taylors theorem, we have
y ()
y(a) = y(x0 ) + y (x0 )(a x0 ) + (a x0 )2 ,
2

y ()
y(b) = y(x0 ) + y (x0 )(b x0 ) + (b x0 )2
2
for some , [a, b]. Since y(a) = y(b) = 0 and y (x0 ) = 0, we get
2M 2M
|y ()| = , |y ()| = .
(a x0 )2 (b x0 )2
( ]
If x0 a, (a + b)/2 , then we have
2M 2M 8M
( )2 = .
(a x0 )2 ba (b a)2
2
[ )
If x0 (a + b)/2, b , then we have
2M 2M 8M
( ) = .
(b x0 )2 ba 2
(b a)2
2
Hence, we obtain
8M 8
max |y (x)| = max |y(x)|.
xI (b a)2 (b a)2 xI
Therefore,
(b a)2
max |y(x)| max |y (x)|,
xI 8 xI
which ends the proof.
Lemma 1.3.2. [29] Let I = [a, b] be a closed interval with < a < b < . If
y C 2 (I, R) and y(a) = 0 = y (a), then
(b a)2
xI 2 xI
Proof. By the Taylors theorem, we have

y ()
y(x) = y(a) + y (a)(x a) + (x a)2
2
for some [a, b]. Since y(a) = y (a) = 0 and (x a)2 (b a)2 , we get
|y ()|
|y(x)| (b a)2
2
for any x I. Thus, we obtain
(b a)2
max |y(x)| max |y (x)|,
xI 2 xI
which completes the proof.

In the following theorems, we prove the Hyers-Ulam stability of the following linear
dierential equation
y (x) + (x)y(x) = 0 (1.3.1)
with boundary conditions
y(a) = 0 = y(b) (1.3.2)
or with initial conditions
y(a) = 0 = y (a) (1.3.3)
where I = [a, b], y C 2 (I, R), C(I, R), and < a < b < .
Theorem 1.3.3. [29] Given a closed interval I = [a, b], let C(I, R) be a
function satisfying max |(x)| < 8/(b a)2 . If a function y C 2 (I, R) satisfies
xI
the inequality
|y (x) + (x)y(x)| , (1.3.4)
for all x I and for some 0, as well as the boundary conditions in (1.3.2),
then there exist a constant K > 0 and a solution y0 C 2 (I, R) of the dierential
equation (1.3.1) with the boundary conditions in (1.3.2) such that
|y(x) y0 (x)| K
for any x I.
Proof. By Lemma 1.3.1, we have
(b a)2
xI 8 xI
Thus, it follows from (1.3.4) that
(b a)2 (b a)2
max |y(x)| max |y (x) + (x)y(x)| + max |(x)| max |y(x)|
xI 8 xI 8 xI xI
(b a) 2
(b a) 2
+ max |(x)| max |y(x)|.
8 8 xI xI
2
Let C := (ba)
8
and K := 1C maxC
|(x)|
. Obviously, y0 0 is a solution of (1.3.1)
with the boundary conditions in (1.3.2) and
|y(x) y0 (x)| K
for any x I.
Theorem 1.3.4. [29] Given a closed interval I = [a, b], let : I R be a

function satisfying max |(x)| < 2/(b a)2 . If a function y C 2 (I, R) satisfies
xI
the inequality (1.3.4) for all x I and for some 0 as well as the initial
conditions in (1.3.3), then there exist a solution y0 C 2 (I, R) of the dierential
equation (1.3.1) with the initial conditions in (1.3.3) and a constant K > 0 such
that
|y(x) y0 (x)| K
for any x I.
Proof. On account of Lemma 1.3.2, we have
(b a)2
xI 2 xI
Thus, it follows from (1.3.4) that
(b a)2 (b a)2
max |y(x)| max |y (x) + (x)y(x)| + max |(x)| max |y(x)|
xI 2 xI 2 xI xI
(b a) 2
(b a) 2
+ max |(x)| max |y(x)|.
2 2 xI xI
2
Let C := (ba)
2
and K := 1C max C
|(x)|
. Obviously, y0 0 is a solution of (1.3.1)
with the initial conditions in (1.3.3) and
|y(x) y0 (x)| K
for all x I.
In the following theorems, we investigate the Hyers-Ulam stability of the dieren-

tial equation
y (x) + p(x)y (x) + q(x)y(x) = 0 (1.3.5)
y(a) = 0 = y(b) (1.3.6)
y(a) = 0 = y (a) (1.3.7)
where y C 2 (I, R), p C 1 (I, R), q C(I, R), and I = [a, b] with < a < b <
.
Let us define a function : I R by
1 1
(x) := q(x) p (x) p(x)2
2 4
for all x I.
Theorem 1.3.5. [29] Assume that there exists a constant L 0 with

x
L p( )d L (1.3.8)
a
for any x I and max |(x)| < 8/(b a)2 . If a function y C 2 (I, R) satisfies
xI
the inequality
|y (x) + p(x)y (x) + q(x)y(x)| (1.3.9)
for all x I and for some 0 as well as the boundary conditions in (1.3.6),
|y(x) y0 (x)| KeL
for any x I.
Proof. Suppose y C 2 (I, R) satisfies the inequality (1.3.9) for all x I. Let us
define
u(x) := y (x) + p(x)y (x) + q(x)y(x), (1.3.10)

( x )
1
z(x) := y(x) exp p( )d (1.3.11)
2 a
for all x I. By (1.3.10) and (1.3.11), we obtain

( ) ( x )
1 1 1
z (x) + q(x) p (x) p(x) z(x) = u(x) exp
2
p( )d
2 4 2 a
for all x I.
Now, it follows from (1.3.8) and (1.3.9) that
( ) ( x )

z (x) + q(x) 1 p (x) 1 p(x)2 z(x) = u(x) exp 1 p( )d eL/2 ,
2 4 2 a
that is,
|z (x) + (x)z(x)| eL/2
for any x I. Moreover, it follows from (1.3.11) that
z(a) = 0 = z(b).
In view of Theorem 1.3.3, there exists a constant K > 0 and a function z0

C 2 (I, R) such that
( )
1 1
z0 (x) + q(x) p (x) p(x) z0 (x) = 0,
2
(1.3.12)
2 4
z0 (a) = 0 = z0 (b)
and
|z(x) z0 (x)| KeL/2 (1.3.13)
for all x I.
We now set
( x )
1
y0 (x) := z0 (x) exp p( )d . (1.3.14)
2 a
Then, since
( x ) ( )
1 1 1 x
y0 (x) = z0 (x) exp p( )d p(x)z0 (x) exp p( )d ,
a 2 2 2 a
(1.3.15)
( x ) ( x )
1 1
y0 (x) = z0 (x) exp p( )d p(x)z0 (x) exp p( )d (1.3.16)
2 a 2 a
( )
1 1 x
p (x)z0 (x) exp p( )d
2 2 a
( )
1 1 x
+ p(x) z0 (x) exp
2
p( )d ,
4 2 a
it follows from (1.3.12), (1.3.14), (1.3.15), and (1.3.16) that
y0 (x) + p(x)y0 (x) + q(x)y0 (x)
( ( ) ) ( )
1 1 1 x
= z0 (x) + q(x) p (x) p(x) z0 (x) exp
2
p( )d
2 4 2 a
=0
for all x I. Hence, y0 satisfies (1.3.5) and the boundary conditions in (1.3.6).
Finally, it follows from (1.3.8) and (1.3.13) that
( ) ( )
1 x 1 x

|y(x) y0 (x)| = z(x) exp p( )d z0 (x) exp p( )d
2 a 2 a
( x )
1
= |z(x) z0 (x)| exp p( )d
2 a
( )
1 x
Ke exp
L/2
p( )d
2 a
KeL
for all x I.
Theorem 1.3.6. [29] Assume that there exists a constant L 0 such that (1.3.8)
holds for all x I. Assume moreover that max |(x)| < 2/(b a)2 . If a function
xI
y C 2 (I, R) satisfies the inequality (1.3.9) for all x I and for some 0 as
well as the initial conditions in (1.3.7), then there exist a constant K > 0 and a
solution y0 C 2 (I, R) of the dierential equation (1.3.5) with the initial conditions
in (1.3.7) such that
|y(x) y0 (x)| KeL

for any x I.
Proof. Suppose y C 2 (I, R) satisfies the inequality (1.3.9) for any x I. Let
us define u(x) and z(x) as in (1.3.10) and (1.3.11), respectively. By (1.3.10) and
(1.3.11), we obtain
( ) ( x )
1 1 1
2
p( )d
2 4 2 a
for all x I.
Now, it follows from (1.3.8) and (1.3.9) that
( ) ( x )
1 1 1
2
p( )d eL/2 ,
2 4 2 a
that is,
|z (x) + (x)z(x)| eL/2
for all x I. Furthermore, in view of (1.3.11), we have
z(a) = 0 = z (a).
By Theorem 1.3.4, there exists a constant K > 0 and a function z0 C 2 (I, R)

such that
( )
1 1
z0 (x) + q(x) p (x) p(x) z0 (x) = 0,
2
2 4
z0 (a) = 0 = z0 (a)
and
|z(x) z0 (x)| KeL/2
for any x I.
We now set
( x )
1
y0 (x) := z0 (x) exp p( )d .
2 a
Moreover, since
( x ) ( )
1 1 1 x
y0 (x) = z0 (x) exp p( )d p(x)z0 (x) exp p( )d
2 a 2 2 a
and
( x ) ( x )
1 1
y0 (x) = z0 (x) exp p( )d
p(x)z0 (x) exp p( )d
2 2
(
a
) a
1 1 x
p (x)z0 (x) exp p( )d
2 2 a
( )
1 1 x
+ p(x) z0 (x) exp
2
p( )d ,
4 2 a
we have
y0 (x) + p(x)y0 (x) + q(x)y0 (x)

( ( ) ) ( )
1 1 1 x
= z0 (x) + q(x) p (x) p(x) z0 (x) exp
2
p( )d
2 4 2 a
=0
for any x I. Hence, y0 satisfies (1.3.5) along with the initial conditions in (1.3.7).
Finally, it follows that
( x ) ( x )
1 1
|y(x) y0 (x)| = z(x) exp p( )d z0 (x) exp p( )d
2 a 2 a
( x )
1
= |z(x) z0 (x)| exp p( )d
2 a
( )
1 x
Ke exp
L/2
p( )d
2 a
KeL
for all x I.
In a similar way, we investigate the Hyers-Ulam stability of the dierential equation
k (x) l(x)
y (x) + y (x) + y(x) = 0 (1.3.17)
k(x) k(x)
y(a) = 0 = y(b) (1.3.18)
y(a) = 0 = y (a) (1.3.19)

where y C 2 (I, R), k C 1 (I, R\{0}), l C(I, R), and < a < b < .
Given a closed interval I = [a, b], we set
( )2
l(x) 1 d k (x) 1 k (x)
(x) :=
k(x) 2 dx k(x) 4 k(x)
for all x I.
Theorem 1.3.7. [29] Assume that there exists a constant L 0 with

x
k ( )
L d L (1.3.20)
a k( )
for any x I and max |(x)| < 8/(b a)2 . If a function y C 2 (I, R) satisfies
xI
the inequality

y (x) + k (x) y (x) + l(x) y(x) , (1.3.21)
k(x) k(x)
for all x I and some 0, as well as the boundary conditions in (1.3.18),
|y(x) y0 (x)| KeL
for any x I.
Proof. Suppose y C 2 (I, R) satisfies (1.3.21) for all x I. Let us define
k (x) l(x)
u(x) := y (x) + y (x) + y(x), (1.3.22)
k(x) k(x)
( x )
1 k ( )
z(x) := y(x) exp d (1.3.23)
2 a k( )
for all x I. By (1.3.22) and (1.3.23), we obtain
( ( )2 ) ( x )
l(x) 1 d k (x) 1 k (x) 1 k ( )
z (x) + z(x) = u(x) exp d .
k(x) 2 dx k(x) 4 k(x) 2 a k( )
Further, it follows from (1.3.20) and (1.3.21) that
( ( )2 ) ( x )
l(x) 1 d k
(x) 1 k (x) 1 k ( )
z (x) + z(x) = u(x) exp d
k(x) 2 dx k(x) 4 k(x) 2 k( )
( x a )
1 k ( )
exp d
2 a k( )
eL/2 ,
that is,
|z (x) + (x)z(x)| eL/2
for all x I. Moreover, it follows from (1.3.18) and (1.3.23) that
z(a) = 0 = z(b).
By Theorem 1.3.3, there exists a constant K > 0 and a function z0 C 2 (I, R)
such that
( ( )2 )
l(x) 1 d k (x) 1 k (x)
z0 (x) + z0 (x) = 0,
k(x) 2 dx k(x) 4 k(x)
z0 (a) = 0 = z0 (b)
and
|z(x) z0 (x)| KeL/2
for any x I.
We now set
( )
1 x
k ( )
y0 (x) := z0 (x) exp d .
2 a k( )
Then, since
( ) ( )
1 x
k ( ) 1 k (x) 1 x k ( )
y0 (x) = z0 (x) exp d z0 (x) exp d
2 a k( ) 2 k(x) 2 a k( )
and
( ) ( )
1 x k ( ) k (x) 1 x k ( )
y0 (x) = z0 (x) exp d z (x) exp d
2 a k( ) k(x) 0 2 a k( )
( ) ( )
1 k (x) 1 x k ( )
z0 (x) exp d
2 k(x) 2 a k( )
( )2 ( )
1 k (x) 1 x k ( )
+ z0 (x) exp d ,
4 k(x) 2 a k( )
we have
k (x) l(x)
y0 (x) + y0 (x) + y0 (x)
k(x) k(x)
( ( ( ) ( )2 ) ) ( )
l(x) 1 k (x) 1 k (x) 1 x k ( )
= z0 (x) + z0 (x) exp d
k(x) 2 k(x) 4 k(x) 2 a k( )
= 0.
Hence, y0 satisfies (1.3.17) along with the boundary conditions in (1.3.18).

Finally, it follows that
( x ) ( x )
1 k ( ) 1 k ( )
|y(x) y0 (x)| = z(x) exp d z0 (x) exp d
2 a k( ) 2 a k( )
( x )
1 k ( )
= |z(x) z0 (x)| exp d
2 a k( )
( )
1 x k ( )
Ke exp
L/2
d
2 a k( )
KeL
for all x I.
By a similar method as we applied to the proof of Theorem 1.3.6, we can prove

the following theorem. Hence, we omit the proof.
Theorem 1.3.8. [29] Assume that max |(x)| < 2/(b a)2 and there exists a
xI
constant L 0 for which the inequality (1.3.20) holds for all x I. If a function
y C 2 (I, R) satisfies the inequality (1.3.21) for all x I and for some 0
as well as the boundary conditions in (1.3.19), then there exist a constant K > 0
and a solution y0 C 2 (I, R) of the dierential equation (1.3.17) with the boundary
conditions in (1.3.19) such that
|y(x) y0 (x)| KeL
for any x I.
Now, we give the definition of superstability with initial and boundary conditions.
Definition 1.3.9. [18] Assume that for any function y C n [a, b], if y satisfies
the dierential inequality
( )
f, y, y , . . . , y (n)
for all x [a, b] and for some 0 with initial(or boundary) conditions, then
either y is a solution of the dierential equation
( )
f, y, y , . . . , y (n) = 0 (1.3.24)
or |y(x)| K for any x [a, b], where K is a constant not depending on y ex-
plicitly. Then, we say that Eq.(1.3.24) has superstability with initial(or boundary)
conditions.
In the following theorem, we investigate the stability of dierential equation of
higher order in the form of
y (n) (x) + (x)y(x) = 0 (1.3.25)

y(a) = y (a) = = y (n1) (a) = 0, (1.3.26)
Theorem 1.3.10. [18] If max |(x)| < n!
(ba)n
. Then (1.3.25) has the superstability
with initial conditions (1.3.26).

Proof. For every > 0, y C 2 [a, b], if y (n) (x) + (x)y(x) and y(a) =

y (a) = = y (n1) (a) = 0. Similarly to the proof of Lemma 1.3.2,
y (n1) (a) y (n) ()
y(x) = y(a) + y (a)(x a) + + (x a)n1 + (x a)n .
(n 1)! n!
Thus (n)
y () (b a)n
|y(x)| = (x a) max y (n) (x)
n
n! n!
for every x [a, b]; so, we obtain
(b a)n (b a)n
max |y(x)| [max |y (x) + (x)y(x)|] +
(n)
max |(x)y(x)|
n! n!
(b a)n (b a)n
+ max |(x)| max |y(x)|.
n! n!
(ba)n (ba)n
Let = n!
max |(x)|, K = n!(1)
. It is easy to see that
|y(x)| K.
Hence (1.3.25) has superstability with initial condtions (1.3.26).
In the following theorems, we investigate the superstability of the dierential equa-

tion
y (n) (x) + an1 y (n1) (x) + + a1 y (x) + a0 y(x) = 0 (1.3.27)

y(a) = y (a) = = y (n1) (a) = 0, (1.3.28)
where y C n (I, C), ai R(i = 0, 1, , n 1), I = [a, b], < a < b < +.
Lemma 1.3.11. [19] Assume that y C 1 (I, C) and C {z C| |z| < 1

ba
}. If
|y (x) Cy(x)|
with y(a) = 0, then there exists a constant K > 0 such that
|y(x)| K.
Proof. Let y(x) = A(x)+iB(x), where i denotes imaginary unit and A(x), B(x)
C 1 (I, R). Since y(a) = 0, we have
A(a) = 0 and B(a) = 0;
By Taylor formula, we obtain
max |A(x)| (b a) max |A (x) CA(x)| + |C| (b a) max |A(x)|

(b a) max |y (x) Cy(x)| + |C| (b a) max |A(x)|
(b a) + |C| (b a) max |A(x)|
and
max |B(x)| (b a) + |C| (b a) max |B(x)| .
Since C {z C| |z| < 1
},
there exists a constant K such that
ba

max |y(x)| max |A(x)|2 + max |B(x)|2 K.
Theorem 1.3.12. [19] If all the roots of the characteristic equation are in the disc
{z C||z| < 1
ba
}, then (1.3.27) has superstability with initial conditions (1.3.28).
Proof. Assume that 1 , 2 , , n are the roots of the characteristic equation
n + an1 n1 + + a1 + a0 = 0.
Define g1 (x) = y (x) 1 y(x) and gi (x) = gi1

(x) i gi1 (x)(i = 2, 3, , n 1),
thus

gn1 (x) n gn1 (x) = y (n) (x) + an1 y (n1) (x) + + a1 y (x) + a0 y(x) ,
and gi (a) = 0 for every i = 1, 2, , n 1.

1
Since the absolute value of n < ba and gn1 (a) = 0, it follows from Lemma
1.3.11 that there exists a K1 > 0 such that
|gn1 (x)| K1 .

Recall gn1 (x) = gn2 (x) n1 gn2 (x), we have

gn2 (x) n1 gn2 (x) K1 .
By an argument similar to the above and by induction, we can show that there
exists a constant K > 0 such that
|y(x)| K.
This completes the proof of our theorem.

Chapter 2
Generlaized Hyers-Ulam Stability

of Dierential Equations
2.1 Generlaized Hyers-Ulam Stability of Linear

Throughout this section, F will denote either the real field R or the complex field
C. A function f : (0, ) F is said to be of exponential order if there are
constants A, B R such that
|f (t)| AetB
for all t > 0. For each function f : (0, ) F of exponential order, we define the
Laplace transform of f by

F (s) = f (t)est dt.
0
There exists a unique number < such that this integral converges
if (s) > and diverges if (s) < , where (s) denotes the real part of the
(complex) number s. The number is called the abscissa of convergence and
denoted by f . It is well known that |F (s)| 0 as (s) . Furthermore, f is
analytic on the open right half plane {s C : (s) > } and we have

d
F (s) = test f (t)dt ((s) > ).
ds 0
The Laplace transform of f is sometimes denoted by L(f ). It is well known that

L is linear and one-to-one.
Conversely, let f (t) be a continuous function whose Laplace transform F (s) has
the abscissa of convergence f , then the formula for the inverse Laplace transforms
yields
+iT
1 st 1
f (t) = lim F (s)e ds = e(+iy)t F ( + iy)dy
2i T iT 2
for any real constant > f , where the first integral is taken along the vertical line
(s) = and converges as an improper Riemann integral and the second integral
is used as an alternative notation for the first integral (see [1]). Hence, we have

L(f )(s) = f (t)est dt ((s) > f )
0

1 1
L (F )(t) = e(+iy)t F ( + iy)dy ( > f ).
2
The convolution of two integrable functions f, g : (0, ) F is defined by
t
(f g)(t) := f (t x)g(x)dx.
0
Then L(f g) = L(f )L(g).

n
m
Lemma 2.1.1. [12] Let P (s) = k sk and Q(s) = k sk , where m, n are
k=0 k=0
nonnegative integers with m < n and k , k are scalars. Then there exists an
infinitely dierentiable function g : (0, ) F such that
Q(s)
L(g) = ((s) > P )
P (s)
and {
0 (for i {0, 1, . . . , n m 2}),
g (i) (0) =
m /n (for i = n m 1)
where P = max{(s) : P (s) = 0}.
Lemma 2.1.2. [12] Given an integer n > 1, let f : (0, ) F be a continuous

function and let P (s) be a complex polynomial of degree n. Then there exists an n
times continuously dierentiable function h : (0, ) F such that
L(f )
L(h) = ((s) > max{P , f }),
P (s)
where P = max{(s) : P (s) = 0} and f is the abscissa of convergence for f . In
particular, it holds that h(i) (0) = 0 for every i {0, 1, . . . , n 1}.
Let F denote either R or C. In the following theorem, using the Laplace trans-
form method, we investigate the generalized Hyers-Ulam stability of the linear
dierential equation of first order
y (t) + y(t) = f (t). (2.1.1)
Theorem 2.1.3. [30] Let be a constant in F and let : (0, ) (0, ) be

an integrable function. If a continuously dierentiable function y : (0, ) F
satisfies the inequality
|y (t) + y(t) f (t)| (t) (2.1.2)
for all t > 0, then there exists a solution y : (0, ) F of the dierential
equation (2.1.1) such that
t
()t
|y(t) y (t)| e e()x (x)dx
0
for any t > 0.

Proof. If we define a function z : (0, ) F by z(t) := y (t) + y(t) f (t) for
each t > 0, then
y(0) + L(f ) L(z)
L(y) = . (2.1.3)
s+ s+
If we set y (t) := y(0)et + (E f )(t), where E (t) = et , then y (0) = y(0)
and
y(0) + L(f ) y (0) + L(f )
L(y ) = = . (2.1.4)
s+ s+
Hence, we get
( )
L y (t) + y (t) = sL(y ) y (0) + L(y ) = L(f ).
Since L is a one-to-one operator, it holds that
y (t) + y (t) = f (t).
Thus, y is a solution of (2.1.1).

Moreover, by (2.1.3) and (2.1.4), we obtain L(y) L(y ) = L(E z). Therefore,
we have
y(t) y (t) = (E z)(t). (2.1.5)
In view of (2.1.2), it holds that
|z(t)| (t) (2.1.6)

for all t > 0, and it follows from the definition of convolution, (2.1.5), and (2.1.6)
that
|y(t) y (t)| = |(E z)(t)|
t

= E (t x)z(x)dx

0
t
(tx)
e (x)dx
0
t
()t
e e()x (x)dx
0
t
for all t > 0. (We remark that 0
e()x (x)dx exists for each t > 0 provided is
an integrable function.)
Corollary 2.1.4. [30] Let be a constant in F and let : (0, ) (0, ) be an

integrable function such that
t
e()(xt) (x)dx K(t) (2.1.7)
0
for all t > 0 and for some positive real constant K. If a continuously dierentiable
function y : (0, ) F satisfies the inequality (2.1.2) for all t > 0, then there
exists a solution y : (0, ) F of the dierential equation (2.1.1) such that
|y(t) y (t)| K(t)
for any t > 0.
In the following remark, we show that there exists an integrable function :

(0, ) (0, ) satisfying the condition (2.1.7).
Remark 2.1.5. [30] Let be a constant in F with () > 1. If we define

(t) = Aet for all t > 0 and for some A > 0, then we have
t t
()(xt)
e (x)dx = e()(xt) Aex dx
0 0
1 ( t )
= Ae Ae()t
1 + ()
1
(t)
1 + ()
for each t > 0.
Now, we apply the Laplace transform method to the proof of the generalized
Hyers-Ulam stability of the linear dierential equation of second order
y (t) + y (t) + y(t) = f (t). (2.1.8)
Theorem 2.1.6. [30] Let and be constants in F such that there exist a, b F
with a + b = , ab = , and a = b. Assume that : (0, ) (0, ) is an
integrable function. If a twice continuously dierentiable function y : (0, ) F
satisfies the inequality
|y (t) + y (t) + y(t) f (t)| (t) (2.1.9)
for all t > 0, then there exists a solution yc : (0, ) F of the dierential equation
(2.1.8) such that

e(a)t t
(a)x e(b)t t
|y(t) yc (t)| e (x)dx + e(b)x (x)dx
|a b| 0 |a b| 0
for all t > 0.

Proof. If we define a function z : (0, ) F by z(t) := y (t)+y (t)+y(t)f (t)
for each t > 0, then we have
( )
L(z) = s2 + s + L(y) [sy(0) + y(0) + y (0)] L(f ). (2.1.10)
In view of (2.1.10), a function y0 : (0, ) F is a solution of (2.1.8) if and only if
( 2 )
s + s + L(y0 ) sy0 (0) [y0 (0) + y0 (0)] = L(f ). (2.1.11)
Now, since s2 + s + = (s a)(s b), (2.1.10) implies that
sy(0) + [y(0) + y (0)] + L(f ) L(z)
L(y) = . (2.1.12)
(s a)(s b) (s a)(s b)
If we set
aeat bebt
yc (t) := y(0) + [y(0) + y (0)]Ea,b (t) + (Ea,b f )(t), (2.1.13)
ab
eat ebt
where Ea,b (t) := ab
, then yc (0) = y(0). Moreover, since
a2 eat b2 ebt aeat bebt d

yc (t)
= y(0)
+ [y(0) + y (0)] + (Ea,b f )(t),
ab ab dt
t bt t
eat e
(Ea,b f )(t) = eax f (x)dx ebx f (x)dx,
ab 0 ab 0
we have
a 2 b2 ab
yc (0) = y(0) + [y(0) + y (0)]
ab ab

= (a + b)y(0) + y(0) + y (0)
= y (0).
It follows from (2.1.13) that
syc (0) + [yc (0) + yc (0)] + L(f )

L(yc ) = . (2.1.14)
(s a)(s b)
Now, (2.1.11) and (2.1.14) imply that yc is a solution of (2.1.8). Applying (2.1.12)
and (2.1.14) and considering the facts that
L(z)
yc (0) = y(0), yc (0) = y (0), and L(Ea,b z) = (sa)(sb) , we obtain
L(y) L(yc ) = L(Ea,b z) or equivalently, y(t) yc (t) = (Ea,b z)(t).
In view of (2.1.9), it holds that |z(t)| (t), and it follows from the definition of
the convolution that
|y(t) yc (t)| = |(Ea,b z)(t)|

t t
e(a)t (a)x e(b)t
e (x)dx + e(b)x (x)dx
|a b| 0 |a b| 0
t t
for any t > 0. We remark that 0 e(a)x (x)dx and 0 e(b)x (x)dx exist for any
t > 0 provided is an integrable function.
Corollary 2.1.7. [30] Let and be constants in F such that there exist a, b F
with a + b = , ab = , and a = b. Assume that : (0, ) (0, ) is an
integrable function for which there exists a positive real constant K with
t
( (a)(tx) )
e + e(b)(tx) (x)dx K(t) (2.1.15)
0
for all t > 0. If a twice continuously dierentiable function y : (0, ) F satisfies

the inequality (2.1.9) for all t > 0, then there exists a solution yc : (0, ) F of
the dierential equation (2.1.8) such that
K
|y(t) yc (t)| (t)
|a b|
for all t > 0.
We now show that there exists an integrable function : (0, ) (0, ) which
satisfies the condition (2.1.15).
Remark 2.1.8. [30] Let and be constants in F such that there exist a, b F
with a + b = , ab = , (a) < 1, (b) < 1, and a = b. If we define (t) = Aet
for all t > 0 and for some A > 0, then we get
t
( (a)(tx) )
e + e(b)(tx) (x)dx
0
t
( (a)(tx) )
= e + e(b)(tx) Aex dx
0
A ( t ) A ( t )
= e e(a)t + e e(b)t
1 (a) 1 (b)
( )
1 1
+ (t)
1 (a) 1 (b)
for all t > 0.
Similarly, we apply the Laplace transform method to investigate the generalized

Hyers-Ulam stability of the linear dierential equation of nth order

n1
(n)
y (t) + k y (k) (t) = f (t) (2.1.16)
k=0
Theorem 2.1.9. [30] Let 0 , 1 , . . . , n be scalars in F with n = 1, where n

is an integer larger than 1. Assume that : (0, ) (0, ) is an integrable
function of exponential order. If an n times continuously dierentiable function
y : (0, ) F satisfies the inequality

n1
(n)
y (t) + k y (k) (t) f (t) (t) (2.1.17)

k=0
for all t > 0, then there exist real constants M > 0 and g and a solution yc :
(0, ) F of the dierential equation (2.1.16) such that
t
|y(t) yc (t)| M e(tx) (x)dx
0
for all t > 0 and > g .

Proof. Applying the integration by parts repeatedly, we derive
( )
k
L y (k)
= s L(y)
k
skj y (j1) (0)
j=1
for any integer k > 0. Using this formula, we can prove that a function y0 :
(0, ) F is a solution of (2.1.16) if and only if

n
n
k
(j1)
L(f ) = k sk L(y0 ) k skj y0 (0)
k=0 k=1 j=1
n
n n
(j1)
= k sk L(y0 ) k skj y0 (0)
k=0 j=1 k=j

n
(j1)
= Pn,0 (s)L(y0 ) Pn,j (s)y0 (0), (2.1.18)
j=1

n
where Pn,j (s) := k skj for j {0, 1, . . . , n}.
k=j
Let us define a function z : (0, ) F by

n1
z(t) := y (n)
(t) + k y (k) (t) f (t) (2.1.19)
k=0
for all t > 0. Then, similarly as in (2.1.18), we obtain

n
L(z) = Pn,0 (s)L(y) Pn,j (s)y (j1) (0) L(f ).
j=1
Hence, we get
( )
1
n
L(z)
L(y) Pn,j (s)y (j1) (0) + L(f ) = . (2.1.20)
Pn,0 (s) j=1
Pn,0 (s)
Let f be the abscissa of convergence for f , let s1 , s2 , . . . , sn be the roots of the

polynomial Pn,0 (s), and let P = max{(sk ) : k {1, 2, . . . , n}}. For any s with
(s) > max{f , P }, we set
( n )
1
G(s) := Pn,j (s)y (j1)
(0) + L(f ) . (2.1.21)
Pn,0 (s) j=1
By Lemma 2.1.2, there exists an n times continuously dierentiable function f0

such that
L(f )
L(f0 ) = (2.1.22)
Pn,0 (s)
for all s with (s) > max{f , P } and
(i)
f0 (0) = 0 (2.1.23)
for any i {0, 1, . . . , n 1}.

For j {1, 2, . . . , n}, we note that

j1
k sk
Pn,j (s) 1 k=0
= j j (2.1.24)
Pn,0 (s) s s Pn,0 (s)
for every s with (s) > max{0, P }. Applying Lemma 2.1.1 for the case of

j1
Q(s) = k sk and P (s) = sj Pn,0 (s), we can find an infinitely dierentiable
k=0
function gj such that

j1
k sk
k=0
L(gj ) = (2.1.25)
sj P n,0 (s)
(k)
and gj (0) = 0 for k {0, 1, . . . , n 1}.
Let
tj1
fj (t) := gj (t) (2.1.26)
(j 1)!
for j {1, 2, . . . , n}. Then we have
{
(i) 0 (for i {0, 1, . . . , j 2, j, j + 1, . . . , n 1}),
fj (0) = (2.1.27)
1 (for i = j 1).
If we define

n
yc (t) := y (j1) (0)fj (t) + f0 (t),
j=1
then the conditions (2.1.23) and (2.1.27) imply that
yc(i) (0) = y (i) (0) (2.1.28)

for every i {0, 1, . . . , n 1}. Moreover, it follows from (2.1.21) to (2.1.28) that

n
L(yc ) = y (j1) (0)L(fj ) + L(f0 )
j=1
n ( )
1 L(f )
= y (0) j L(gj ) +
(j1)
j=1
s Pn,0 (s)
( n )
1
= Pn,j (s)y (j1) (0) + L(f ) (2.1.29)
Pn,0 (s) j=1
for each s with (s) > max{0, f , P }.

Now, (2.1.18) implies that yc is a solution of (2.1.16). Moreover, by (2.1.20) and
(2.1.29), we have
L(z)
L(y) L(yc ) = . (2.1.30)
Pn,0 (s)
Applying Lemma 2.1.1 for the case of Q(s) = 1 and P (s) = Pn,0 (s), we find an
infinitely dierentiable function g : (0, ) F such that
1
L(g) = (2.1.31)
Pn,0 (s)
which implies that

( )
1 1 1 1
g(t) = L = e(+iy)t dy
Pn,0 (s) 2 Pn,0 ( + iy)
for any real constant > g .

Moreover, it holds that

1 (+iy)(tx) 1
|g(t x)| e dy
2 |Pn,0 ( + iy)|

1 1
e(tx) dy
2 |Pn,0 ( + iy)|

1 (tx) 1
e dy
2 |Pn,0 ( + iy)|
M e(tx) (2.1.32)
for all > g , where

1 1
M= dy < ,
2 |Pn,0 ( + iy)|
because n is an integer larger than 1. By (2.1.17) and (2.1.19), it also holds that
|z(t)| (t) for all t > 0.
In view of (2.1.30), (2.1.31), and (2.1.32), we get
L(y) L(yc ) = L(g)L(z) = L(g z).
Consequently, we have y(t) yc (t) = (g z)(t) for any t > 0. Hence, it follows
from (2.1.17), (2.1.19), and (2.1.32) that
t t
|y(t) yc (t)| = |(g z)(t)| |g(t x)||z(x)|dx M e(tx) (x)dx
0 0
for all t > 0 and for any real constant > g , which completes the proof.
Corollary 2.1.10. [30] Let 0 , 1 , . . . , n be scalars in F with n = 1, where n is

an integer larger than 1. Assume that there exist real constants and K > 0 such
that a function : (0, ) (0, ) satisfies
t
e(tx) (x)dx K(t)
0
for all t > 0. Moreover, assume that the constant g given in Theorem 2.1.9 is less
than . If an n times continuously dierentiable function y : (0, ) F satisfies
the inequality (2.1.17) for all t > 0, then there exist a real constants M > 0 and a
solution yc : (0, ) F of the dierential equation (2.1.16) such that
|y(t) yc (t)| KM (t)
for all t > 0.
Remark 2.1.11. [30] Assume that < 1. If we define (t) = Aet for all t > 0
and for some A > 0, then we get
t t
A ( t ) 1
e(tx)
(x)dx = e(tx) Aex dx = e et (t)
0 0 1 1
for all t > 0.
2.2 Generlaized Hyers-Ulam Stability of Dier-

ential Equations with boundary Conditions
2.2.1 Generalized Superstability of Dierential Equations with Initial Con-
ditions
In this subsection, we investigate the generalized superstability of linear dierential

equation of nth-order in the form of
y (n) (x) + (x)y(x) = 0, (2.2.1)

y(a) = y (a) = = y (n1) (a) = 0, (2.2.2)

In addition to that we investigate the generalized superstability of dierential
equations of second order in the form of y (x) + p(x)y (x) + q(x)y(x) = 0 and the
superstability of linear dierential equations with constant coecients.
First of all, we give the definition of generalized superstability with initial and
boundary conditions.
Definition 2.2.1. [19],[25] Assume that for any function y C n [a, b], if y satisfies
the dierential inequality
( )
f, y, y , . . . , y (n) (x)
for all x [a, b] and for some function : [a, b] [0, ) with initial(or boundary)
conditions, then either y is a solution of the dierential equation
( )
f, y, y , . . . , y (n) = 0 (2.2.3)
or |y(x)| (x) for any x [a, b], where : I [0, ) is a function not de-
pending on y explicitly. Then, we say that Eq.(2.2.3) has generalized superstability
with initial(or boundary) conditions.
In this subsection, given the closed interval I = [a, b], we assume that : I
[0, ) and let M (p(x)) denote max [a,x] |p( )| for every p C(I, R).
Theorem 2.2.2. [19] If |(x)| < n!/(b a)n for every x I, then Eq.(2.2.1) has
generalized superstability with initial conditions (2.2.2).
Proof. Suppose that a function y C n (I, R) satisfies the inequality
(n)
y (x) + (x)y(x) (x), (2.2.4)
for all x I,
By Taylor formula, we have
y (n1) (a) y (n) ()
y(x) = y(a) + y (a)(x a) + + (x a)n1 + (x a)n .
(n 1)! n!
Therefore, (n)
y () (x a)n

|y(x)| = n
(x a) M(y (n) (x))
n! n!
for every x [a, b]. Then,
( )
(x a)n
M (y(x)) M M(y (x)) (n)
n!
( )
( ) (x a)n
M M(y (x)) M
(n)
n!
(x a) n
= M(y (n) (x))
n!
Thus
(x a)n
M(y(x)) M(y (n) (x))
n!
(x a)n ( (n) ) (x a)n
M y (x) + (x)y(x) + M |(x)| M(y(x))
n! n!
(x a)n (b a)n
M ((x)) + max |(x)|M (y(x)) .
n! n!
(ba)n
Let C1 = 1 n!
max |(x)|. It easy to see that
(x a)n
M (y(x)) M ((x)) .
n!C1
Moreover, |y(x)| M (y(x)), which completes the proof of our theorem.
In the following theorem, we investigate the generalized superstability of the dif-

ferential equation

y (x) + p(x)y (x) + q(x)y(x) = 0 (2.2.5)

y(a) = 0 = y (a), (2.2.6)
where y C 2 [a, b], p C[a, b], q C 0 [a, b], < a < b < +.
2
Theorem 2.2.3. [19] If max{q(x) 12 p (x) p 4(x) } < 2/(b a)2 , then (2.2.5) has
generalized superstability with initial conditions (2.2.6).
Proof. Suppose that y C 2 [a, b] satisfies the inequality

y (x) + p(x)y (x) + q(x)y(x) (x). (2.2.7)
Let

u(x) = y (x) + p(x)y (x) + q(x)y(x), (2.2.8)
for all x [a, b], and define z(x) by
x
1
y(x) = z(x)exp( p( )d ). (2.2.9)
2
a
By a substitution (2.2.9) in (2.2.8), we obtain

x
( 1 p2 (x) ) 1
z (x) + q(x) p (x) z(x) = u(x)exp( p( )d ).
2 4 2
a
Then it follows from inequality (2.2.7) that

x
( 2 )
z (x) + q(x) 1 p (x) p (x) z(x) = |u(x)| exp( 1 p( )d )
2 4 2
a
x
1
(x)exp( p( )d )
2
a
From (2.2.6) and (2.2.9) we have
z(a) = 0 = z(b). (2.2.10)
It follows from Theorem 2.2.2 that there exists a constant C1 > 0 such that

x
(x a)n
1
|z(x)| M (x)exp( p( )d ) .
n!C1 2
a
From (2.2.9) we have

x x
(x a)n
1 1
|y(x)| M (x)exp( p( )d ) exp( p( )d ).
n!C1 2 2
a a
Thus (2.2.5) has generalized superstability with initial conditions (2.2.6).
2.2.2 Hyers-Ulam-Rassias Stability of Linear Dierential Equations with
Boundary Conditions
Lemma 2.2.4. [24] Let y C n [a, b] and satisfies the initial conditions

y(a) = y (a) = = y (n1) (a) = 0, (2.2.11)

max y (n) (x).
n
then max |y(x)| (ba)
n!
Proof. By Taylor formula, we have

y (n1) (a) y (n) ()
y(x) = y(a) + y (a)(x a) + + (x a)n1 + (x a)n .
(n 1)! n!
We have (x a)n (b a)n . Therefore,
(n)
y () (n) (b a)n

|y(x)| = (x a) max y (x)
n
n! n!
for every x [a, b].
In the following theorems, we prove the Hyers-Ulam-Rassias stability of the fol-
lowing linear dierential equation

y (x) + (x)y(x) = 0 (2.2.12)
y(a) = 0 = y(b) (2.2.13)

y(a) = 0 = y (a) (2.2.14)
where I = [a, b], y C 2 (I, R), C(I, R), < a < b < +.
Given the closed interval I and a function : I R, define a function f : I R

by

f (x) = y (x) + (x)y(x), (2.2.15)
for all x I.
Theorem 2.2.5. [24] Given the closed interval I and a function : I R,

assume that : I [0, ) is a dicreasing (increasing) function and max |(x)| <
8/(b a)2 . If a function y C 2 (I, R) satisfies the inequality

y (x) + (x)y(x) (x), (2.2.16)
for all x I, with boundary conditions (2.2.13) such that the function (2.2.15) is
increasing (decreasing) function then there exist a constant K > 0 and a solution
y0 C 2 (I, R) of the dierential equation (2.2.12) with boundary conditions (2.2.13)
such that
|y(x) y0 (x)| K(x) (2.2.17)
for any x I.
Proof. We have by Lemma 1.3.1 that
(b a)2

max |y(x)| max y (x) .
8
Thus
(b a)2 (b a)2

max |y(x)| max y (x) + (x)y(x) + max |(x)| max |y(x)|
8 8
(b a)2 (b a)2
(x) + max |(x)| max |y(x)| .
8 8
2
Let C = (ba)
8
C
, K = 1C max|(x)| . Obviously, y0 (x) = 0 is a solution of (2.2.12)
with boundary conditions (2.2.13) and
|y(x) y0 (x)| K (x).


assume that : I [0, ) is a decreasing (increasing) function and max |(x)| <
2/(b a)2 . If a function y C 2 (I, R) satisfies the inequality (2.2.16) for all
x I, with initial conditions (2.2.14) such that the function (2.2.15) is increasing
(decreasing) then there exist a solution y0 C 2 (I, R) of the dierential equation
(2.2.12) and a constant K > 0 such that
|y(x) y0 (x)| K(x) (2.2.18)
for any x I.
(b a)2

max |y(x)| max y (x) .
2
Thus
(b a)2 (b a)2

max |y(x)| max y (x) + (x)y(x) + max |(x)| max |y(x)|
2 2
(b a)2 (b a)2
(x) + max |(x)| max |y(x)| .
2 2
2
Let C = (ba)2
C
, K = 1C max|(x)| . Obviously, y0 (x) = 0 is a solution of (2.2.12)
with initial conditions (2.2.14) and |y(x) y0 (x)| K (x).
In the following theorems investigate the Hyers-Ulam-Rassias stability of linear
dierential equation of nth - order
y (n) (x) + (x)y(x) = 0, (2.2.19)
with initial conditions (2.2.11).
Given the closed interval I = [a, b] and a function : I R, define a function

f : I R by
f (x) = y n (x) + (x)y(x), (2.2.20)
for all x I.

assume that : I [0, ) is a decreasing (increasing) function and max |(x)| <
n!/(b a)n . If a function y0 C n [a, b] satisfies the inequality
|y n (x) + (x)y(x)| (x), (2.2.21)
for all x I, with initial conditions (2.2.11) such that the function (2.2.20) is
increasing (decreasing) then there exist a solution y0 C n (I, R) of the dierential
equation (2.2.19) and a constant K > 0 such that
|y(x) y0 (x)| K(x) (2.2.22)
for any x I.
(b a)n
max |y(x)| max |y n (x)| .
n!
Thus
(b a)n (b a)n
max |y(x)| max |y n (x) + (x)y(x)| + max |(x)| max |y(x)|
n! n!
(b a)n (b a)n
(x) + max |(x)| max |y(x)| .
n! n!
n
Let C1 = (ba)
n!
C1
, K = 1C1 max|(x)| . Obviously, y0 (x) = 0 is a solution of (2.2.19)
with initial conditions (2.2.11) and |y(x) y0 (x)| K (x).
Chapter 3
Hyers-Ulam Stability of System

of Dierential Equations
3.1 Hyers-Ulam Stability of Linear System of Dif-

ferential Equations
In this section, by applying the fixed point alternative method, we give a necessary
and sucient condition in order that the first order linear system of dierential
equations z(t)
+ A(t)z(t) + B(t) = 0 has the Hyers-Ulam-Rassias stability and
find Hyers-Ulam stability constant under those conditions. In addition to that, we
apply this result to a second order dierential equation
y(t) + f (t)y(t)
+ g(t)y(t) + h(t) = 0.
Also, we apply to dierential equations with constant coecient in the same sense
of proofs.
3.1.1 Preliminaries and Auxiliary Results
Definition 3.1.1. ([7],[2],[47]) Let I be any interval, let z : I Rn , A : I

Rnn , B : I Rn , then
z(t)
+ A(t)z(t) + B(t) = 0 (3.1.1)
is Hyers-Ulam-Rassias stable with respect to : I [0, ), with, z(t) =

n
|zi (t)|, if there exists a real constant K > 0 such that for each solution s
i=1
C 1 (I, Rn ) of inequality
z(t)
+ A(t)z(t) + B(t) (t)
there exists a solution z C 1 (I, Rn ) of equation (3.1.1) with
s(t) z(t) K(t)

, t I.
Definition 3.1.2. For a nonempty set X, a function d : X X [0, ] is called

a generalized metric on X if and only if d satisfies :
(i) d(x, y) = 0 if and only if x = y;
(ii) d(x, y) = d(y, x) for all x,y X ;
(iii) d(x, z) d(x, y) + d(y, z) for all x,y,z X.
Theorem 3.1.3. (The fixed point alternative) [15] Let (X, d) be a generalized
complete metric space. Assume that : X X is a strictly contractive operator
with Lipschitz constant L < 1. If there exists a nonnegative integer k such that
d(k+1 x, k x) < for some x X, then the followings are true :
(a) The sequence {n x} convergens to a fixed point x of ;
(b) x is the unique fixed point of in
{ }
X = y X/d(k x, y) < ;
(c) If y X , then d(y, x ) 1

1L
d(y, y).
Lemma 3.1.4. [26] For given real numbers a and b with a < b, let I = [a, b] be a
closed interval and let be
X = {f : I Rn , is continuous f unction}, and d : X X [0,] a function is
defined as follows :
d(f, g) = inf{c [0, ]/ f (t) g(t) c(t) t I}

where (t) : I (0, ) is a continuous function, then d is a generalized metric
on X.
Proof. By definition of a function d, then for all f, g X
(1) d(f, g) = 0 f (t) = g(t) for all t I;

(2) d(f, g) = d(g, f ).
To prove that d(f, g) d(f, h) + d(h, g) for all f, g, h X.
Assume that d(f, g) > d(f, h) + d(h, g) for some f, h, g X. Then, by definition
of d, there exists t0 I with
f (t0 ) g(t0 ) > {d(f, h) d(h, g)} (t0 )

= d(f, h)(t0 ) d(h, g)(t0 )
f (t0 ) h(t0 ) + h(t0 ) g(t0 )
and this is contradiction .
Lemma 3.1.5. [26] For given real numbers a and b with a < b, let I = [a, b] be a
closed interval and let
X = {f : I Rn , is continuous function }, consider a generalized metric function
on X, d : X X [0,] which is defined as follows:
d(f, g) = inf{c [0, ]/ f (t) g(t) c(t) t I} (3.1.2)

where (t) : I (0, ) is a continuous function, then (X, d) is a complete metric
space.
Proof. Let {hn } be a Cauchy sequence in (X, d). Then > 0 there exist N N
such that d(hm , hn ) m, n N . That means that with equation (3.1.2)
> 0 N N : m, n N , t I hm (t) hn (t) (t). (3.1.3)
If t is fixed, equation (3.1.3) implies that {hn (t)} is a cauchy sequence in Rn .
Since Rn is complete, {hn (t)} converge for each t I. Thus, we can define a
function h : I Rn by h(t) = lim hn (t).
n
If we let m , it then follows from (3.1.3) that
> 0 N N : n N , t I h(t) hn (t) (t), (3.1.4)
that is, since is bounded on I, {hn (t)} converges uniformly to h. Hence, h is
continuous and h X.
If we consider equation (3.1.2) and (3.1.4), then we may conclude that
> 0 N N : n N d(h, hn )
that is, the cauchy sequence {hn (t)} converge to h in (X, d). Hence, (X, d) is
complete.
3.1.2 Hyers-Ulam Stability of First Order System of Dierential Equa-
tions
We will prove the Hyers-Ulam-Rassias stability for the equation (3.1.1) on the
intervals I = [a, b), where < a < b .
Theorem 3.1.6. [26] Let A : I Rnn and B : I Rnn be continuous matrices

functions and let for a positive constant N , such that A(t) N for all t in I.
Assume that : I [o, ) is an integrable function with the property that there
exists P in (0, 1) such that
t
A(t1 ) (t1 )dt1 P (t) (3.1.5)
a
for all t I. If a continuously dierential function z : I Rnn verifies the

relation :
z(t)
+ A(t)z(t) + B(t) (t) (3.1.6)
for all t in I. Then there exists a unique solution s : I Rnn of the equation
(3.1.1) which verifies the following relation:
P
z(t) s(t) (t) (3.1.7)
N NP
for all t I and s(a) = z(a).
Proof. Let us consider the set
= {h : I Rn /h is continuous and h(a) = z(a)}
with a function d : [0, ] defined on as
d(h1 , h2 ) = d (h1 , h2 )
= inf{K > 0, h1 (t) h2 (t) K(t), t I}
By lemma (3.1.4) and Lemma (3.1.5), the (, d) is generalized complete metric

space. We define the operator T : ,
t
T h(t) = z(a) (A(t1 )h(t1 ) + B(t1 ))dt1 , t I
a
for all h . Indeed T h is a continuously dierentable function on I, since A and

B are continuous functions and T h(a) = z(a). Now, let h1 , h2 . Then we have
t

T h1 (t) T h2 (t) = A(t1 )(h1 (t1 ) h2 (t1 )dt1 )

a
t
A(t1 ) h1 (t1 ) h2 (t1 )dt1

a
t
d(h1 , h2 ) A(t1 )(t1 )dt1
a
P (t)d(h1 , h2 )t I.
Therefore,
d(T h1 (t) T h2 (t)) P d(h1 , h2 ) (3.1.8)
Thus, the operator T is a contraction with the constant P .
By integrating the both sides of the relation (3.1.6) on [a, t] we obtain

t

z(t) z(a) + (A(t1 )z(t1 ) + B(t1 ))dt1 P (t) f or all t I. (3.1.9)
N

a
which means d(z, T z) N P

(t) < . By the fixed point alternative theorem
(3.1.3) there exists an element s = lim T n z and s is a unique fixed point of T in
n
the set = {h /d(T n0 z, h) < } .
It may be proved that = {h /d(z, h) < }. Therefore, the set is inde-
pendent of n0 . To prove that the function s is a solution to the equation (3.1.1),
we derive, with respect to t, the both sides of the following relation:
s(t) = T s(t) t I
Thus,
= A(t)s(t) B(t)
s(t)
for all t I, which implies that the function s is a solution of the equation (3.1.1)
and verifies the relation s(a) = z(a). Applying the fixed point alternative theorem
again, we obtain d(h, s) 1P 1
d(h, T h) for all h .
Since z , we have
d(z, S) 1P
1
d(z, T z) N (1P
P
)
.
Hence, z(t) s(t) N (1P ) (t) for all t I.
P
This inequality proves the relation (3.1.1) .

In the same manner it is possible to proved the following theorem of the Hyers-
Ulam-Rassias stability of the equation (3.1.1) on the interval J = (b, a] , where
b < a < .
Theorem 3.1.7. [26] Let A : J Rnn and B : J Rnn be continuous ma-

trices functions and let for a positive constant N , such that A(t) N for all
t J.
Assume that : J [o, ) is an integrable function with the property that there
exists P (0, 1)such that
t
A(t1 ) (t1 )dt1 P (t) (3.1.10)
a
for all t J. If a continuously dierential function z : J Rnn verifies the

relation :
z(t)
+ A(t)z(t) + B(t) (t) (3.1.11)
for all t J, then there exists a unique solution s : J Rnn of the equation
P
z(t) s(t) (t) (3.1.12)
N NP
for all t J and s(a) = z(a).
The Hyers-Ulam-Rassias stability equation (3.1.1) on R will be proved by Theorem
(3.1.6) and Theorem (3.1.7).
Corollary 3.1.8. [26] Let A : R Rnn and B : R Rnn be continuous ma-

trices functions and let for a positive constant N , such that A(t) N for all
t R. Assume that : R [o, ) is an integrable function with the property that
there exists P (0, 1) such that
t

A(t1 ) (t1 )dt1 P (t) (3.1.13)

0
for all t R. If a continuously dierential function z : R Rnn verifies the

relation :
z(t)
+ A(t)z(t) + B(t) (t) (3.1.14)
for all t R, then there exists a unique solution s : R Rnn of the equation
P
z(t) s(t) (t) (3.1.15)
N NP
for all t R and s(0) = z(0).
Proof. By relation (3.1.13) we have
t
A(t1 ) (t1 )dt1 P (t) (3.1.16)
0
for all t 0. Applying Theorem (3.1.6), there exists a solution of equation (3.1.1),
s1 : [0, ) Rnn , which verifies relations (3.1.7) and s1 (o) = z(o) . From
relation(3.1.13) we also obtain
o
A(t1 ) (t1 )dt1 P (t) (3.1.17)
t
for all t 0. Applying Theorem (3.1.7), there exists a solution of equation (3.1.1),
s2 : (, 0] Rnn which verifies relation (3.1.12) and s2 (o) = z(o). It is easy
to check if the function {
s1 (t), t 0
s(t) = (3.1.18)
s2 (t), t < 0
is a continuously dierentiable function on R, a solution of equation (3.1.1) on R
and it verifies relation (3.1.15).
Corollary 3.1.9. [26] Let A = 0 be n n constant matrix and B : R Rnn

be n 1 a continuous matrix function (n 1 constant matrix ). Assume that
: R [o, ) is an integrable function with the property that there exists
P (0, 1) such that
t

(t1 ) P (t) (3.1.19)
A

0
for all t R. If a continuously dierential function z : R Rnn verifies the

relation :
z(t)
+ Az(t) + B(t) (t) (z(t)
+ Az(t) + B (t)) (3.1.20)
for all t R, then there exists a unique solution s : R Rnn of the equation
z(t)
+ Az(t) + B(t) = 0 (z(t)
+ Az(t) + B = 0)
which verifies the following relation:

P
z(t) s(t) (t) (3.1.21)
A(1 P )
for all t R and s(0) = z(0).

Proof. By Corollary (3.1.8) and let N = A .
3.1.3 Hyers-Ulam Stability of Second Order Dierential Equation
In this section we will prove the Hyers-Ulam-Rassias stability for the following
scallar equation
y(t) + f (t)y(t)
+ g(t)y(t) + h(t) = 0 (3.1.22)
In the same manner, at first we will prove the Hyers-Ulam-Rassias stability for the
equation (3.1.22) on the intervals I = [a, b) , where < a < b .
Theorem 3.1.10. [26] Let f, g, h : I R be continuous functions and let for a

positive constant 0 < N < 1,
|f (t)| + |g(t)| N for all t I. Assume that : I [o, ) is an integrable
function with property that there exists P (0, 1) such that
t
(1 + |f (t1 )| + |g(t1 )|)(t1 )dt1 P (t) (3.1.23)
a
for all t I. If a function y C 2 (I, R) verifies the relation
| + g(t)y(t) + h(t)| (t)

y (t) + f (t)y(t) (3.1.24)
for all t I, then there exists a unique solution s1 : I R of equation (3.1.22),

which verifies the following relation
P
|y(t) s1 (t)| (t) (3.1.25)
N (1 P )
for all t I, and s1 (a) = y(a).

Proof. Let [ z1 (t) ]= y(t), z2 (t) ] = f (t)z

[ = z1 (t), z2 (t) [ 2 (t) ] g(t)z1 (t).
z1 (t) 0 1 0
Let z(t) = , A(t) = , H(t) = , then equation(3.1.22)
z2 (t) g(t) f (t) h(t)
transfer into
z(t)
+ A(t)z(t) + H(t) = 0 (3.1.26)
for all t I. By hypotheses ,since |f (t)| + |g(t)| N , then
A(t) = 1 + |f (t)| + |g(t)| N and by relation (3.1.23) we obtain
t
A(t) (t1 )dt1 P (t).
a
Now, let[ a function
] [y satisfy
] the relation (3.1.24), since
z1 (t) y(t)
z(t) = = and
z2 (t) y(t)

[ ] [ ][ ] [ ]
y(t)
0 1 y(t) 0
z(t)
+ A(t)z(t) + H(t) = + +
y(t) g(t) f (t) y(t)
h(t)
[ ]
y(t)
y(t)
=
y(t) + f (t)y(t)
+ g(t)y(t) + h(t)
Therefore,
z(t)
+ A(t)z(t) + H(t) = | y (t) + f (t)y(t)
+ g(t)y(t)
[ + h(t)| ] (t). Hence, by
s1 (t)
Theorem (3.1.6), there exists solution such s(t) = of equation (3.1.26)
s2 (t)
and s(a) = z(a) such that z(t) s(t) N (1P P
)
(t) for all t I.
Therefore, there exists s1 (t) satisfying equation (3.1.22) and s1 (a) = y(a) such
that
|y(t) s1 (t)| N (1P
P
)
(t) for all t I.
In the same manner and by Theorem (3.1.7) we can prove the Hyers-Ulam-Rassias
stability for the equation (3.1.22) on the interval J = (b, a], where b < a <
.
Theorem 3.1.11. [26] Let f, g, h : J R be continuous functions and let for a

positive constant 0 < N < 1,
|f (t)| + |g(t)| N for all t J. Assume that : J [o, ) is an integrable
function with property that there exists P (0, 1) such that
a
(1 + |f (t1 )| + |g(t1 )|)(t1 )dt1 P (t) (3.1.27)
t
for all t J. If a function y C 2 (J, R) verifies the relation
| + g(t)y(t) + h(t)| (t)

y (t) + f (t)y(t) (3.1.28)
for all t J. then there exists a unique solution s1 : J R of equation (3.1.22),

P
|y(t) s1 (t)| (t) (3.1.29)
N (1 P )
for all t J, and s1 (a) = y(a).

Now in the same manner Corollary (3.1.8) and by Theorems (3.1.10) and (3.1.11),
we obtain the following Corollary
Corollary 3.1.12. [26] Let f, g, h : R R be continuous functions and let for

a positive constant 0 < N < 1, |f (t)| + |g(t)| N for all t R. Assume that
: R [o, ) is an integrable function with property that there exists P (0, 1)
such that t

(1 + |f (t1 )| + |g(t1 )|)(t1 )dt1 P (t) (3.1.30)

0
for all t in R. If a function y C 2 (R, R) verifies the relation
| + g(t)y(t) + h(t)| (t)

y (t) + f (t)y(t) (3.1.31)
for all t R, then there exists a unique solution s1 : R R of equation (3.1.22),

P
|y(t) s1 (t)| (t) (3.1.32)
N (1 P )
for all t R, and s1 (0) = y(0).

Remark 3.1.13. [26] The results can be applied to all dierential equations of
higher order by transferring it to system of first order.
3.2 Hyers-Ulam Stability of Nonlinear System of

In 1961, the notion of practical stability was discussed in the monograph by Lasalle
and Lefschetz [17]. In which they point out that stability investigations may not as-
sure practical stability and vice versa. For example an aircraft may oscillate around
a mathematically unstable path, yet its performance may be acceptable. Motivated
by this, Weiss and Infante introduced the concept of finite time stability[20]. There
are many studies about the relation between types of stability, Lyapunov stability
and practical stability (see [17],[31],[2]). With these results in mind, in this sec-
tion, we give a sucient condition in order that the first order nonlinear system of
dierential equations has Hyers-Ulam stability and Hyers-Ulam-Rassias stability.
In addition, we present the relation between practical stability and Hyers-Ulam
stability and also Hyers-Ulam-Rassias stability.
3.2.1 Preliminaries and Auxiliary Results
Let (B, .) be a Banach space (real or complex), and let J = [t0 , t0 + T ) for some
T > 0, t0 0. We consider two systems: a system
x = f (t, x) , t J, (3.2.1)
where f is defined and continuous on J B. The equilibrium state is at the origin
: f (t, 0) = 0, t J.
A system that depends on parameter (0, 0 ],(0 (0, )) which is said to be
perturbed system
x = f (t, x) + p(t, x). (3.2.2)
Let P be the set of all perturbations p satisfying p(t, x) = 0 for all t J
and all x, let Q be a closed and bounded set of B containing the origin and let Q0
be a subset of Q .
Definition 3.2.1. Practical stability[17]

Let x (t, x0 , t0 ) be the solution of (3.2.2) satisfying x (t0 , x0 , t0 ) = x0 . If for each
p P ,i.e. (0, 0 ], x0 in Q0 and each t0 0 , x (t, x0 , t0 ) in Q for all t J,
then the origin is said to be (Q0 , Q, 0 )-practically stable .
The slutions which start initially in Q0 remain thereafter in Q .
Definition 3.2.2. [39],[41] Let be a positive real number. We consider the system
(3.2.1) with following dierential inequality
y(t)
f (t, y(t)) , t J. (3.2.3)
The equation (3.2.1) is generalized Hyers-Ulam stability (GHUs) if for each

(0, 0 ] and for each solution y(t, t0 , x0 ) C 1 (J, B) of (3.2.3) there exists a solution
x C 1 (J, B) of (3.2.1) with
|y(t) x(t)| K(),

where K() is an expression of with lim K() = 0 .

0
Definition 3.2.3. [7],[44]

We consider the system (3.2.1) with following dierential inequality
y(t)
f (t, y(t)) (t) t J, (3.2.4)
where : J [0, ) is a continuous function. The equation (3.2.1) is generalized

Hyers-Ulam-Rassias stability (GHURs) with respect to if there exists K > 0
such that for each solution y(t, t0 , x0 ) C 1 (J, B) of (3.2.4) there exists a solution
x C 1 (J, B) of (3.2.1) with
|y(t) x(t)| K(t), t J
Definition 3.2.4. [31] We say that V : J B R is a Lyapunov function if

V (t, x) is continuous in (t, x), bounded on bounded subset of B.
3.2.2 Hyers-Ulam Stability of System of Dierential Equations
Lemma 3.2.5. [27] Consider the following dierential equation
x = f (t, x(t)) , t J (3.2.5)
with initial condition

x0 = x(t0 ) Q0 , (3.2.6)
where f is defined and continuous on J B, and equilibrium state is at the origin:
f (t, 0) = 0 , t J. The system (3.2.5),(3.2.6) to be (Q0 , Q, 0 )-practically stable
it is sucient that there exists a continuous non increasing on the system (3.2.5)
solutions Lyapunov function V (t, x) such that
{x B : V (t, x) 1} Q , t J (3.2.7)
Q0 {x B : V (t0 , x) 1} (3.2.8)
Proof. We will prove by contradiction. Suppose that conditions (3.2.7) ,(3.2.8)
are satisfied but there are J and x0 Q0 such that the solution x(t) =
x(t, x0 , t0 ) of (3.2.5) leaves the set Q. From (3.2.7) follows inequality
V (, x( )) > 1which contradicts the condition (3.2.8). Therefore our assumption
is false and the equilibrium of system (3.2.5),(3.2.6) is (Q0 , Q, 0 )-practically stable
.
Theorem 3.2.6. [27] Consider two systems: the system of dierential equation
(3.2.5), (3.2.6) and the system (3.2.2) . If equilibrium of (3.2.5) (at the origin) is
(Q0 , Q, 0 )-practically stable then the system (3.2.5), (3.2.6) is generalized Hyers-
Ulam stability.
Proof. Since Q is closed and bounded set then there exists real number M > 0
such that Q = {x : x M }.
Now, let x = f (t, x0 , t0 ) satisfying (3.2.3) for arbitrary (0, 0 ], then x satisfies
(3.2.2). Since the equilibrium of (3.2.5) is (Q0 , Q, 0 )-practically stable then x in
Q, that means that x M. Since M > 0 , > 0 then there exists s > 0 such
that M = s .
Hence, x s for all t J ,lim K() = lim s = 0 . Obviously ,w(t) = 0
0 0
satisfies the equation (3.2.5) and
x (t) w(t) s , t J.
Hence, the equation (3.2.5) with initial condition (3.2.6) has generalized Hyers-
Ulam stability.
Corollary 3.2.7. [27] For the system (3.2.5), (3.2.6) to be generalized Hyers-
Ulam stability it sucient that there exists a continuous non increasing on the
system (3.2.5) solutions Lyapunov function V (t, x) such that satisfies the condi-
tions (3.2.7) and (3.2.8).
Proof. Suppose that conditions (3.2.7), (3.2.8) are satisfied, then by lemma 3.2.5
the system (3.2.5),(3.2.6) is (Q0 , Q, 0 )-practically stable. Hence, by theorem 3.2.6
the system has generalized Hyers-Ulam stability.
Theorem 3.2.8. [27] Consider the following dierential equation
x = f (t, x(t)) , t J (3.2.9)

x0 = x(t0 ) Q0 , (3.2.10)
where f is defined and continuous on J B, and equilibrium state is at the origin

:f (t, 0) = 0 , t J.
If equilibrium is (Q0 , Q, 0 )-practically stable and there exists 1 > 0 such that
1 (t) t J then the system (3.2.9), (3.2.10) is GHURs with respect to .
Proof. Since Q is closed and bounded set then there exists real number M > 0
such that Q = {x : x M }.
Now, let x = f (t, x0 , t0 ) satisfying (3.2.9), since (t) then x satisfies (3.2.2).
Since the equilibrium of (3.2.9) is (Q0 , Q, 0 )-practically stable then x in Q, that
mean that x M . Since M > 0, 1 > 0 then there exists K > 0 such that
M = K1 .
Then, x K1 for all t J ,hence x K(t) for all t J. Obviously,
w(t) = 0 satisfies the equation (3.2.9) and
x (t) w(t) K(t) , t J.
Hence, the equation (3.2.9) with initial condition (3.2.10) has generalized Hyers-
Ulam-Rassias stability.
Corollary 3.2.9. [27] For the system (3.2.9), (3.2.10) to be generalized Hyers-
Ulam stability it sucient that there exsist a continuous nonincreasing on the sys-
tem (3.2.9) solutions Lyapunov function V (t, x) such that satisfies the conditions
(3.2.7) and (3.2.8).
Proof. Suppose that conditions (3.2.7), (3.2.8) are satisfied, then by lemma 3.2.5
the system (3.2.9), (3.2.10) is (Q0 , Q, 0 )-practically stable. Hence, by theorem
3.2.8 the system has generalized Hyers-Ulam-Rassias stability.
Theorem 3.2.10. [27] Let (B, .) be a Banach space (real or complex), and let
J = [t0 , t0 + T ) for some T > 0, t0 0. Consider two systems : a system
x = f (t, x) , t J, (3.2.11)

x(t0 ) = 0 Q0 , (3.2.12)
for a set Q0 , where f is defined, continuous on J B and satisfies Lipschitz con-

dition. The equilibrium state is at the origin : f (t, 0) = 0, t J.
A system that depends on parameter (0, 0 ], (0 (0, )) which is said to be
perturbed system
x = f (t, x) + p(t, x). (3.2.13)
Let P be the set of all perturbations p satisfying p(t, x) = 0 for all t J

and all x .
If the system of dierential equation (3.2.11), (3.2.12) has Hyers-Ulam stability

with Hyers-Ulam constant K then the origin is (Q0 , Q, 0 )-practically stable, where
Q = {x : x K0 }, contains the origin.

Proof. Let > 0, (0, 0 ] and let x = f (t, x0 , t0 ), x0 Q0 be a so-
.
lution of (3.2.13), i.e. x f (t, x ) . Since the system (3.2.11), (3.2.12)
has Hyers-Ulam stability with constant
K > 0 then there exists y a solution

of (3.2.11), (3.2.12) with x y K. By uniqueness of solution then y=0.

Hence x K K0 . Thus the equilibrium of (3.2.11), (3.2.12) is (Q0 , Q, 0 )-
practically stable.
Remark 3.2.11. [20] In case Q0 Q then we have expansive stability. If Q Q0

then we have contractive stability.
Remark 3.2.12. If we have a dierential equation of n-order in a Banach space B1

then we reduce it to a dierential equation of first order in Banach space B = Bn1 .
List of Papers
Published papers
(1) Qusuay.H. Alqifiary, S.M. Jung, On the Hyers-Ulam stability of dierential

equations of second order, Hindawi Publ. Corp. J. Abstract and Applied
Analysis , Article ID 483707(2014),8 pages.
(2) Qusuay.H. Alqifiary, S.M. Jung, Hyers-Ulam stability of second-order linear

dierential equations with boundary conditions, SYLWAN. , 158(5),(2014),289-
301 pages.
(3) Qusuay.H. Alqifiary, S.M. Jung, Laplace transform and generalized Hyers-
Ulam stability of linear dierential equations, Electronic Journal of Dieren-
tial Equations. 2014 (2014), no. 80, 1-11.
(4) Jinghao Huang,Qusuay.H. Alqifiary, Yongjin Li, Superstability of dierential

equations with boundary conditions, Electronic Journal of Dierential Equa-
tions. 2014 (2014), no. 215, 1-8.
(5) Qusuay.H. Alqifiary, Some properties of second order dierential equations,

Mathematica Moravica, Vol. 17-1 (2013), pp. 89-94.
(6) Qusuay.H. Alqifiary, Note on the stability for linear systems of dierential
equations, International Journal of Applied Mathematical Research, 3 (1)
(2014) pp.15-22.
(7) Qusuay.H. Alqifiary, J. Kneevi-Miljanovi, Note on the stability of system of

dierential equations x = f(t; x(t)), Gen. Math. Notes.,Vol. 20, No. 1,
January 2014, pp. 27-33.
Alqifiary List of papers 63
Accepted papers
(1) Jinghao Huang,Qusuay.H. Alqifiary, Yongjin Li, On the generalized supersta-

bility of nth-order linear dierential equations with initial conditions, Publi-
cations de lInstitut Mathematique.
(2) Qusuay.H. Alqifiary, On Hyers-Ulam-Rassias stability of linear dierential

equations with boundary conditions, Southeast Asian Bulletin of Mathemat-
ics.
Bibliography
[1] B. Davies, Integral Transforms and Their Apllications, Springer, New York,
2001.
[2] B.N. Chakip, B.M. Walid and G. Germain, Finite time stability by solving
LMIs problem: Application on four tanks system, Int. Conference on Control,
Engineering and Information Technology, 1(2013), 1-6.
[3] C.Alsina and R.Ger, On some inequalities and stability results related to the
exponential function, J.of Inequal and Appl.. 2, (1998),373-380.
[4] D. G. Bourgin, Classes of transformations and bordering transformations,

Bull. Amer. Math. Soc. 57 (1951), 223237.
[5] D.H. Hyers, On the stability of the linear functional equation, Proc. Natl.
Soc. USA 27, pp. 222224, 1941.
[6] D.H. Hyers, G. Isac and Th.M. Rassias, Stability of Functional Equations in
Several Variables, Birkhauser, Boston, 1998.
[7] D. Popa, Hyers-Ulam-Rassias stability of a linear recurrence, J. Math. Anal.

Appl., 309(2005), 591-597.
[8] D. Popa and I. Rasa, On the Hyers-Ulam stability of the linear dierential
equation, J. Math. Anal. Appl. 381, pp. 530537, 2011.
[9] D. Popa and I. Rasa, Hyers-Ulam stability of the linear dierential operator
with non-constant coecients, Appl. Math. Comput. 219, pp. 15621568,
2012.
[10] D.S. Cimpean and D. Popa, On the stability of the linear dierential equation
of higher order with constant coecients, Appl. Math. Comput. 217, pp.
41414146, 2010.
Alqifiary Bibliography 65
[11] G. Wang, M. Zhou and L. Sun, Hyers-Ulam stability of linear dierential

equations of first order, Appl. Math. Lett. 21, pp. 10241028, 2008.
[12] H. Rezaei, S.-M. Jung and Th.M. Rassias, Laplace transform and Hyers-Ulam
stability of linear dierential equations, J. Math. Anal. Appl. 403, pp. 244
251, 2013.
[13] I.A. Rus, Remarks on Ulam stability of the operatorial equations, Fixed Point
Theory 10, pp. 305320, 2009.
[14] I.A. Rus, Ulam stability of ordinary dierential equations, Stud. Univ. Babes-
Bolyai Math. 54, pp. 125134, 2009.
[15] J.B.Diaz and B.Margolis, A fixed point theorem of the alternative for contrac-
tions on a generalized complete metric space,Bull.Amer.Math.Soc., 74, (1968),
305-309.
[16] J. Baker, J. Lawrence and F. Zorzitto, The stability of the equation f (x+y) =
f (x)f (y), Proc. Amer. Math. Soc. 74 (1979), 242-246.
[17] J.P. Lasalle and S. Lefschetz, Stability by Liapunovs Direct Method with Ap-
plications, Academic Press, New York, (1961).
[18] Jinghao Huang,Qusuay.H. Alqifiary, Yongjin Li, Superstability of dierential

equations with boundary conditions, Electronic Journal of Dierential Equa-
tions. 2014 (2014), no. 215, 1-8.
[19] Jinghao Huang,Qusuay.H. Alqifiary, Yongjin Li, On the generalized supersta-

bility of nth-order linear dierential equations with initial conditions, Publi-
cations de lInstitut Mathematique. ((accepted))
[20] L. Weiss and E.F. Infante, On the stability of systems defined over a finite
time interval, Proc. Nat. Academic Sci, U.S.A., 54(1965), 44-48.
[21] M. Obloza, Hyers stability of the linear dierential equation, Rocznik Nauk.-
Dydakt. Prace Mat. 13, pp. 259270, 1993.
[22] M. Obloza, Connections between Hyers and Lyapunov stability of the ordinary
dierential equations, Rocznik Nauk.-Dydakt. Prace Mat. 14, pp. 141146,
1997.
[23] P. Gavruta, S.-M. Jung and Y. Li, Hyers-Ulam stability for second-order linear
dierential equations with boundary conditions, Electronic J. Dierential Equ.
2011 (2011), 15.
[24] Qusuay.H. Alqifiary, On Hyers-Ulam-Rassias Stability of Linear Dierential

Equations with Boundary Conditions, Southeast Asian Bulletin of Mathemat-
ics. ((accepted))
[25] Qusuay.H. Alqifiary, Some Properties of Second Order Dierential Equations,

Mathematica Moravica, Vol. 17-1 (2013), pp. 89-94.
[26] Qusuay.H. Alqifiary, Note on the stability for linear systems of dierential
equations, International Journal of Applied Mathematical Research, 3 (1)
(2014) pp.15-22.
[27] Qusuay.H. Alqifiary, J. Kneevi-Miljanovi, Note on the Stability of System of

Dierential Equations x = f(t; x(t)), Gen. Math. Notes.,Vol. 20, No. 1, Jan-
uary 2014, pp. 27-33.
[28] Qusuay.H. Alqifiary, S.M. Jung, On the Hyers-Ulam Stability of Dierential

Equations of Second Order, Hindawi Publ. Corp. J. Abstract and Applied
Analysis , Article ID 483707(2014),8 pages.
[29] Qusuay.H. Alqifiary, S.M. Jung, Hyers-Ulam Stability of Second-Order

Linear Dierential Equations with Boundary Conditions, SYLWAN. ,
158(5),(2014),289-301 pages.
[30] Qusuay.H. Alqifiary, S.M. Jung, Laplace Transform and Generalized Hyers-
Ulam Stability of Linear Dierential Equations, Electronic Journal of Dier-
ential Equations. 2014 (2014), no. 80, 1-11.
[31] R. Stephen and V. Lakshmikanthan, Practical stability and Lyapunov func-

tions, Tohoku Math.Journal, 32(1980), 607-613.
[32] S. Czerwik, Functional Equations and Inequalities in Several Variables, World

Scientific, Singapore, 2002.
[33] S.-E. Takahasi, T. Miura and S. Miyajima, On the Hyers-Ulam stability of

the Banach space-valued dierential equation y = y, Bull. Korean Math.
Soc. 39, pp. 309315, 2002.
[34] S.-E. Takahasi, H. Takagi, T. Miura and S. Miyajima, The Hyers-Ulam stabil-
ity constants of first order linear dierential operators, J. Math. Anal. Appl.
296, pp. 403409, 2004.
[35] S.-M. Jung, Hyers-Ulam stability of linear dierential equations of first order,
Appl. Math. Lett. 17, pp. 11351140, 2004.
III, J. Math. Anal. Appl. 311, pp. 139146, 2005.
II, Appl. Math. Lett. 19, pp. 854858, 2006.
[38] S.-M. Jung, Hyers-Ulam-Rassias Stability of Functional Equations in Nonlin-

ear Analysis, Springer, New York, 2011.
[39] S.-M. Jung, Hyers-Ulam stability of first order linear dierential equations
with constant coecients ,Math.Anal.Appl., 320 (2006),549-561.
[40] S.M. Ulam, Problems in Modern Mathematics, Chapter VI, Scince Editors,
Wiley, New York, 1960.
.. .
[41] S.M. Jung, Hyers-Ulam stability of dierential equation y +2x y 2ny =
0, Hindawi Publ. Corp. J. of Inequalities and Applications, Article ID
793197(2010),12 pages.
[42] T. Miura, S. Miyajima and S.-E. Takahasi, A characterization of Hyers-Ulam

stability of first order linear dierential operators, J. Math. Anal. Appl. 286,
pp. 136146, 2003.
[43] T. Miura, S. Miyajima and S.-E. Takahasi, Hyers-Ulam stability of linear

dierential operator with constant coecients, Math. Nachr. 258, pp. 9096,
2003.
[44] Th.M. Rassias, On the stability of the linear mapping in Banach spaces, Proc.
Amer. Math. Soc. 72, pp. 297300, 1978.
[45] Y. Li and Y. Shen, Hyers-Ulam stability of linear dierential equations of

second order, Appl. Math. Lett. 23, pp. 306309, 2010.
[46] Y. Li and Y. Shen, Hyers-Ulam stability of nonhomogeneous linear dierential

equation of second order, Int. J. Math. Mathl. Sci.,2009(2009), Article ID
576852,7 pages.
[47] Y.-H. Lee and K.-W. Jun, A generalization of Hyers-Ulam-Rassias stability

of Jensens equation, J. Math. Anal. Appl. 238 (1999), 305315.
Prilog 1.
Izjava o autorstvu
Potpisani Qusuay Hatim Eghaar Alqifiary
broj indeksa 2001/2011
Izjavljujem
da je doktorska disertacija pod naslovom
HYERS-ULAM STABILITY OF THE SOLUTIONS OF

DIFFERENTIAL EQUATIONS
rezultat sopstvenog istra

zivackog rada,
da predlo
zena disertacija u celini ni u delovima nije bila predlozena
za dobijanje bilo koje diplome prema studijskim programima
drugih visokoskolskih ustanova,
da su rezultati korektno navedeni i
da nisam kr
sio autorska prava i koristio intelektualnu svojinu
drugih lica.
Potpis doktoranda
U Beogradu,
Prilog 2.
Izjava o istovetnosti
stampane i elektronske
verzije doktorskog rada
Ime i prezime autora: Qusuay Hatim Eghaar Alqifiary

Broj indeksa: 2001/2011
Studijski program:
Naslov rada:

Mentor: redovni Prof. dr. Julka Knezevic-Miljanovic
Potpisani Qusuay Hatim Eghaar Alqifiary
Izjavljujem da je stampana verzija mog doktorskog rada istovetna

elektronskoj verziji koju sam predao za objavljivanje na portalu Dig-
italnog repozitorijuma Univerziteta u Beogradu.
Dozvoljavam da se objave moji licni podaci vezani za dobijanje akadem-

skog zvanja doktora nauka, kao sto su ime i prezime, godina i mesto
rodjenja i datum odbrane rada.
Ovi licni podaci mogu se objaviti na mreznim stranicama digitalne
biblioteke, u elektronskom katalogu i u publikacijama Univerziteta u
Beogradu.
Potpis doktoranda
U Beogradu,
Prilog 3.
Izjava o kori
s
cenju
Ovlascujem Univerzitetsku biblioteku Svetozar Markovic da u Dig-

italni repozitorijum Univerziteta u Beogradu unese moju doktorsku
disertaciju pod naslovom:

koja je moje autorsko delo.
Disertaciju sa svim prilozima predao sam u elektronskom formatu

pogodnom za trajno arhiviranje.
Moju doktorsku disertaciju pohranjenu u Digitalni repozitorijum Uni-

verziteta u Beogradu mogu da koriste svi koji postuju odredbe sadrzane
u odabranom tipu licence Kreativne zajednice (Creative Commons)
za koju sam se odlucio.
1. Autorstvo
2. Autorstvo - nekomercijalno
3. Autorstvo - nekomercijalno - bez prerade
4. Autorstvo - nekomercijalno - deliti pod istim uslovima
5. Autorstvo - bez prerade
6. Autorstvo - deliti pod istim uslovima
(Molimo da zaokruzite samo jednu od sest ponudjenih licenci, kratak

opis licenci dat je na poledjini lista).
Potpis doktoranda
U Beogradu,
1. Autorstvo - Dozvoljavate umnozavanje, distribuciju i javno saopstavanje

dela, i prerade, ako se navede ime autora na nacin odredjen od strane
autora ili davaoca licence, cak i u komercijalne svrhe. Ovo je najslo-
bodnija od svih licenci.
2. Autorstvo - nekomercijalno. Dozvoljavate umnozavanje, distribu-

ciju i javno saopstavanje dela, i prerade, ako se navede ime autora
na nacin odredjen od strane autora ili davaoca licence. Ova licenca
ne dozvoljava komercijalnu upotrebu dela.
3. Autorstvo - nekomercijalno - bez prerade. Dozvoljavate umnozavanje,

distribuciju i javno saopstavanje dela, bez promena, preoblikovanja
ili upotrebe dela u svom delu, ako se navede ime autora na nacin
odredjen od strane autora ili davaoca licence. Ova licenca ne dozvol-
java komercijalnu upotrebu dela. U odnosu na sve ostale licence,
ovom licencom se ogranicava najveci obim prava koriscenja dela.
4. Autorstvo - nekomercijalno - deliti pod istim uslovima. Dozvolja-

vate umnozavanje, distribuciju i javno saopstavanje dela, i prerade,
ako se navede ime autora na nacin odredjen od strane autora ili
davaoca licence i ako se prerada distribuira pod istom ili slicnom
licencom. Ova licenca ne dozvoljava komercijalnu upotrebu dela i
prerada.
5. Autorstvo - bez prerade. Dozvoljavate umnozavanje, distribuciju

i javno saopstavanje dela, bez promena, preoblikovanja ili upotrebe
dela u svom delu, ako se navede ime autora na nacin odredjen od
strane autora ili davaoca licence. Ova licenca dozvoljava komerci-
jalnu upotrebu dela.
6. Autorstvo - deliti pod istim uslovima. Dozvoljavate umnozavanje,

distribuciju i javno saopstavanje dela, i prerade, ako se navede ime
autora na nacin odredjen od strane autora ili davaoca licence i ako
se prerada distribuira pod istom ili slicnom licencom. Ova licenca
dozvoljava komercijalnu upotrebu dela i prerada. Slicna je softver-
skim licencama, odnosno licencama otvorenog koda.

phdQusuayAlqifiarydiferencne PDF

Uploaded by

Copyright:

Available Formats

phdQusuayAlqifiarydiferencne PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

phdQusuayAlqifiarydiferencne PDF

Uploaded by

Copyright:

Available Formats

UNIVERZITET U BEOGRADU

Qusuay Hatim Eghaar Alqifiary

Qusuay Hatim Eghaar Alqifiary

HYERS-ULAM STABILITY OF THE

system of dierential equations z(t)

Scientific field (naucna oblast): Mathematics (matematika)(34A40, 34A12, 34D10)

redovni profesor dr Miodrag Spalevic

I would like to express my special appreciation and thanks to my supervisor

I would like to acknowledge those who have shaped my mathematical education.

To each of the above, I express my deepest gratitude.

Theorem 0.0.2. Let f : X1 X2 be a function between Banach spaces such that

for some > 0 and for all x, y X1 . Then the limit

A(x) = limn 2n f (2n x)

for every x X1 . Moreover, if f (tx) is continuous in t for each fixed x X1 , then

problem to other functional equations and generalized Hyerss result in various

Theorem 0.0.3. Let f : X1 X2 be a function between Banach spaces. If f

f (x + y) f (x) f (y) (xp + yp )

with initial conditions

where n N+ , y C n [a, b], C 0 [a, b], < a < b < +.

1 Hyers-Ulam Stability of Dierential Equations 1

2 Generlaized Hyers-Ulam Stability of Dierential Equations 29

3 Hyers-Ulam Stability of System of Dierential Equations 47

3.1 Hyers-Ulam Stability of Linear System of Dierential Equa-

1.1 Hyers-Ulam Stability of Linear Dierential

1.1.1 Boundedness of Solutions of a Second Order Dierential Equation

Lemma 1.1.1. [28] Let u, v : [0, ) [0, ) be integrable functions, c > 0 be a

for all t t0 , then ( )

Proof. It follows from (1.1.1) that

for each t t0 , which together with (1.1.1) implies that

In the following theorem, using Lemma 1.1.1, we investigate sucient conditions

Theorem 1.1.2. [28] Let : [0, ) R be a dierentiable function. Every solu-

for all t t0 . Integrating by parts, this yields

for any t t0 . Then it follows from (1.1.3) that

for all t t0 . Thus, it holds that

Corollary 1.1.3. [28] Let : [0, ) R be a dierentiable function satisfying

u (t) + (t)u(t) = 0 (1.1.5)

1.1.2 Hyers-Ulam Stability of Linear Dierential Equations of Second Or-

(ii) u(t0 ) = u (t0 ) = 0;

Theorem 1.1.4. [28] Given constants L > 0 and t0 0, assume that

for any t t0 , where

|u (t)| u (t)u (t) + u(t)u (t) + (t)u(t)u (t) |u (t)|

for all t t0 . Hence, it holds that

If we set (t) := 1+(t), then the following corollary is an immediate consequence

Example 1.1.1. [28] Let : [0, ) R be a constant function defined by

Theorem 1.1.6. [28] Given constants L > 0 and t0 0, assume that

Corollary 1.1.7. [28] Given constants L > 0 and t0 0, assume that

If we set (t) := (t), then the following corollary is an immediate consequence

Corollary 1.1.8. [28] Given constants L > 0 and t0 0, assume that

Example 1.1.2. [28] Let : [0, ) (, 0) be a monotone decreasing function

1.2 Hyers-Ulam Stability of Nonlinear Dieren-

for all t t0 . Then we have

In the following theorem, we investigate the Hyers-Ulam stability of the Emden-

u (t) + h(t)u(t) = 0 (1.2.3)

for the case where is a positive odd integer.

|u (t)| u (t)u (t) + h(t)u(t) u (t) |u (t)|

for all t t0 . for all t t0 . Then we have

for all t t0 , from which we have

for all t t0 . Hence, it holds that

for any t t0 , where we set = + 1. Obviously, u0 (t) 0 satisfies the equation