On solutions of quadratic integral
equations
Ph.D Thesis
Written at
The Faculty of Mathematics and Computer Science,
Adam Mickiewicz University
Poznań, Poland, 2013
By
MOHAMED METWALI ATIA METWALI
Under the guidance of
Dr. hab. Mieczyslaw Cichoń
Department of Differential Equations,
Faculty of Mathematics and Computer Science,
Adam Mickiewicz University
In Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy
in
Mathematics
Acknowledgments
First of all, I am thankful to Allah for all the gifts has given me.
I would like to express my gratitude and thanks to my advisor and my professor,
Dr. hab. Mieczyslaw Cichoń, Department of Differential Equations, Faculty of
Mathematics and Computer Science, Adam Mickiewicz University for his help and
valuable advice in the preparation of this dissertation.
I am thankful to my family (my wife and my lovely sons ”Basem and Eyad”) for
their support, encouragement and standing beside me during my stay in Poland.
ii
Contents
1 Preliminaries
1.1 Introduction . . . . . . . . . . . . .
1.2 Notation and auxiliary facts . . . .
1.2.1 Lebesgue Spaces . . . . . .
1.2.2 Young and N-functions . . .
1.2.3 Orlicz spaces . . . . . . . .
1.3 Linear and nonlinear operators. . .
1.3.1 The superposition operators.
1.3.2 Fredholm integral operator.
1.3.3 Volterra integral operator.
1.3.4 Urysohn integral operator. .
1.3.5 The multiplication operator.
1.4 Monotone functions. . . . . . . . .
1.5 Measures of noncompactness. . . .
1.6 Fixed point theorems. . . . . . . .
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Monotonic integrable solutions for quadratic integral
a half line.
2.1 Motivations and historical background. . . . . . . . . .
2.2 Introduction. . . . . . . . . . . . . . . . . . . . . . . .
2.3 Main result . . . . . . . . . . . . . . . . . . . . . . . .
2.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
equations on
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 On some integrable solutions for quadratic functional integral equations
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Main result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1 The existence of L1 -solution . . . . . . . . . . . . . . . . . . .
3.2.2 The existence of Lp -solution p > 1 . . . . . . . . . . . . . . . .
3.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iii
4
4
4
5
5
6
7
7
9
10
11
12
14
17
19
23
23
25
26
29
31
31
33
33
38
42
4 Functional quadratic integral
half line
4.1 Introduction . . . . . . . . .
4.2 Main result . . . . . . . . .
4.3 Examples . . . . . . . . . .
equations with perturbations on a
45
. . . . . . . . . . . . . . . . . . . . . . . 45
. . . . . . . . . . . . . . . . . . . . . . . 46
. . . . . . . . . . . . . . . . . . . . . . . 54
5 On quadratic integral equations in Orlicz spaces
5.1 Introduction . . . . . . . . . . . . . . . . . . . . .
5.2 The case of operators with values in L∞ (I). . . .
5.2.1 The case of W1 = L∞ (I). . . . . . . . . .
5.2.2 The case of W2 = L∞ (I). . . . . . . . . .
5.3 The existence of Lp -solution. . . . . . . . . . . . .
5.3.1 Remarks and examples. . . . . . . . . . .
5.4 A general case of Orlicz spaces. . . . . . . . . . .
5.4.1 The case of N satisfying the ∆′ -condition.
5.4.2 The case of N satisfying the ∆3 -condition.
5.4.3 Remarks on classes of solutions. . . . . . .
5.5 Conclusions and fixed point theorems. . . . . . .
5.5.1 A fixed point theorem. . . . . . . . . . . .
iv
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
55
55
58
58
59
64
67
68
69
74
78
80
82
INTRODUCTION
Linear and nonlinear integral equations form an important class of problems
in mathematics. There are different motivations for their study. Some equations
describe mathematical models in physics, engineering or biology. There are also
such equations whose interest lies in other branch of pure mathematics.
Bearing in mind both mentioned aspects we are interested on a special class of
integral equations, namely on quadratic ones. In this case the unknown function
is treated by some operators, then a pointwise multiplication of such operators
is applied. The study of such a kind of problems was begun in early 60’s due to
mathematical modeling of radiative transfer (Chandrasekhar [44], Crum [49]). From
the mathematical point of view they are interesting because of lack of compactness
for considered operators. Thus some of the classical methods for proving existence
theorems are not allowed. There is one more reason which makes this type of
equations interesting. In contrast to the case of standard integral equations only
continuous solutions were considered. It seems to be strange from application point
of view (as will be described below) as well as it prevents a common treatment for
both quadratic and non-quadratic equations.
This dissertation is devoted to study quadratic integral equations and different
classes of their solutions. We concentrate on the aspect of possible discontinuity
of solutions and the best possible assumptions ensuring the existence of solutions.
This leads us to the Lebesgue spaces and some class of Orlicz spaces. Our approach
allow, for the first time, to consider simulatneously quadratic and ”classical” integral
equations. We stress on strongly nonlinear problems, which leads to mentioned
function spaces, but require a new method of the proof. We prove several results for
such a class of equations (existence, monotonicity) on finite and infinite intervals,
including functional-integral problems.
Let us begin a prototype for this theory, that is the Chandrasekhar integral
equation (Chandrasekhar [44, 43])
Z 1
t
ψ(s)x(s) ds.
x(t) = 1 + x(t)
0 t+s
It describe a scattering through a homogeneous semi-infinite plane atmosphere. In
particular, solutions for this equations need not to be continuous. Nevertheless, till
now only such a kind of solutions was investigated. Note, that for non-homogeneous
problems only approximated methods is known (Hollis and Kelley [74]). Different
types of quadratic integral equations will be described later (applicable in plasma
corners [40, 86], kinetic theory of gases, the theory of neutron transport, in the traffic
theory or in mathematical biology). We need only to stress, that the continuity of
1
considered solutions does not follow from applicability of problems (on the contrary),
but only from some mathematical unsolved questions (operators in functions spaces
and their properties, fixed point theorems). In particular, we solve a problem from
[40].
Quadratic integral equations was investigated by many authors. Initial studies
by Chandrasekhar ([44] in 1947, cf. also a book [43] in 1960) form only a beginning
for this theory, mainly made by astrophysicists. Then research was conducted by
mathematicians. They found some interesting open questions in this theory. Let
us mention some papers by Anichini, Conti [6], Cahlon, Eskin [42], Banaś, Argyros
[11], Caballero, Mingarelli, Sadarangani [40, 41], Nussbaum [96], Gripenberg [72],
Mullikin [94, 95], Rus [107], Shrikhant, Joshi [113], Schillings [111] and many others.
In the first Chapter we collect all necessary definitions and theorems. We present
some function spaces, linear and nonlinear operators and their properties are described. Among others, we present some new studies on a.e. monotonic functions
in Lebesgue and Orlicz spaces.
Our Chapter 2 is devoted to study quadratic integral equations on unbounded
intervals. We present our motivations for the study of presented equation and preceded by a historical background we investigate the quadratic integral equation on
a half-line. Here we are looking for a.e. monotonic locally integrable solutions. An
illustrative example completes this Chapter (and all others).
In the next Chapter 3, we study some functional integral equations of quadratic
type. This aspect of the theory is not sufficiently investigated due some restrictions
on functional part. Since we try to unify both quadratic and non-quadratic cases,
we need to investigate functional equations. Here we study a.e. monotonic L1 and
Lp solutions for the considered problem.
In the Chapter 4 we unify our research by considering quadratic functional integral equations on a unbounded intervals. Note, that in this Chapter we do not
assume, that the operators preserve monotonicity properties. A different method of
the proof is then used, which allow us to locate our results among earlier ones.
The last Chapter 5 contains our main theorems and conclusions. We consider
strongly nonlinear functions, which lead us to solutions in Orlicz spaces. This is
well-known for classical (non-quadratic) equations, but it is completely new in the
context of quadratic integral equations. We study the pointwise multiplication in
Orlicz spaces and in a class of such spaces we solve considered equations. By reducing
our problem to an operator equation we present an existence result for a large class
of function spaces. The idea of the proof is not only to prove our theorems but
also to fully cover the theory for classical equations, which was impossible in the
2
previous approach. We need to stress, that this allow us to prove new fixed point
theorem for product of two operators.
3
Chapter 1
Preliminaries
1.1
Introduction
This chapter is devoted to recall some notations and known results that will be
needed in the sequel. We start in section 1.2 by setting basic notations and definitions that are observed throughout the work. In section 1.3 we study some important
linear and nonlinear operators of various types and the multiplication of the operators. We discuss the monotonicity of the functions in section 1.4. Section 1.5 deals
with the strong and weak measure of noncompactness. We end this chapter by
section 1.6 in which we introduce some important fixed point theorems.
1.2
Notation and auxiliary facts
Let R be the field of real numbers, R+ be the interval [0, ∞) and by I = [a, b] denotes
an interval subset R.
Assume that (E, k · k) is an arbitrary Banach space with zero element θ. Denote
by Br (x) the closed ball centered at x and with radius r. The symbol Br stands
for the ball B(θ, r). When necessary we will also indicate the space by using the
notation Br (E). If X is a subset of E, then X̄ and convX denote the closure and
convex closure of X, respectively. We denote the standard algebraic operations on
sets by the symbols k · X and X + Y .
4
1.2.1
Lebesgue Spaces
Define Lp = Lp (I), 1 ≤ p < ∞ be the space of Lebesgue integrable functions
(equivalence classes of functions) on a measurable subset I of R, with the norm
||x||Lp (I) =
Z
p1
|x| dt .
p
I
For p = ∞, L∞ (I) denotes the Banach space of essentially bounded functions on I
with the norm
kxkL∞ = ess sup |x(t)| < ∞.
t∈I
Recall that the essential supermum is defined as
ess sup |x(t)| = inf{a :
t∈I
the set {t : |x(t)| > a} has measure 0}.
Let L1 (I) denote the space of Lebesgue integrable functions on the fixed interval
I ⊂ R, bounded or not.
Further, denote by BC(R+ ) the Banach space of all real functions defined, continuous and bounded on R+ . This space is furnished with the standard norm
kxk = sup{|x(t)| : t ∈ R+ }.
Let us fix a nonempty and bounded subset X of BC(R+ ) and a positive number
T . For x ∈ X and ε ≥ 0 let us denote by ω T (x, ε) the modulus of continuity of the
function x, on the closed and bounded interval [0, T ] (cf. [35]) defined by
ω T (x, ε) = sup{|x(t2 ) − x(t1 )| : t, s ∈ [0, T ], |t2 − t1 | ≤ ε}.
1.2.2
Young and N -functions
A function M : [0, +∞) → [0, +∞) is called a Young function if it has the form
Z u
M(u) =
a(s)du for
u ≥ 0,
0
where a : [0, +∞) → [0, +∞) is an increasing, left-continuous function which is
neither identically zero nor identically infinite on [0, +∞). In particular, if M is
finite-valued, where limu→0 Mu(u) = 0, limu→∞ Mu(u) = ∞ and M(u) > 0 if x > 0
(M(u) = 0 ⇐⇒ u = 0), then M is called an N−function.
5
The functions M and N are called complementary N-functions. If
N(x) = sup(xy − M(x)).
y≥0
Further, the N− function M satisfies the ∆2 -condition, i.e.
(∆2 )
there exist ω, t0 ≥ 0 such that for t ≥ t0 , we have M(2t) ≤ ωM(t).
Let us observe, that an N-function M(u) = exp u2 − 1 satisfies this condition,
while the function M(u) = exp |u| − |u| − 1 does not.
An N-function M is said to satisfy ∆′ -condition if there exist K, t0 ≥ 0 such
that for t, s ≥ t0 , we have M(ts) ≤ KM(t)M(s).
If the N-function M satisfies the ∆′ -condition, then it also satisfies ∆2 -condition.
α
Typical examples: M1 (u) = |u|α for α > 1, M2 (u) = (1 + |u|) ln (1 + |u|) − |u| or
√
M3 (u) = |u|α(| ln |u|| + 1) for α > 3 + 25 .
The last class of N-functions, interesting for us, consists of functions which
increase more rapidly than power functions.
An N-function M is said to satisfy ∆3 -condition if there exist K, t0 ≥ 0 such
that for t ≥ t0 , we have tM(t) ≤ M(Kt).
1.2.3
Orlicz spaces
The Orlicz class, denoted by OP , consists of measurable functions x : I → R for
which
Z
ρ(x; M) = M(x(t))dt < ∞.
I
We shall denote by LM (I) the Orlicz space of all measurable functions x : I → R
for which
Z
x(s)
M
ds ≤ 1 .
kxkM = inf
λ>0
λ
I
p
The N-function M(u) = |u|p , 1 < p < ∞ leads to the classical Lebesgue space
Lp (I) with the norm mention before.
Let EM (I) be the closure in LM (I) of the set of all bounded functions.
Note that EM ⊆ LM ⊆ OM . The inclusion LM ⊂ LP holds if, and only if, there
exists positive constants u0 and a such that P (u) ≤ aM(u) for u ≥ u0 .
An important property of EM spaces lies in the fact that this is a class of functions
from LM having absolutely continuous norms.
Moreover, we have EM = LM = OM if M satisfies the ∆2 -condition.
Sometimes, we will use more general concept of function spaces i.e. ideal spaces.
6
Definition 1.2.1. [118] A normed space (X, k·k) of (classes of ) measurable functions x : I → U (U is a normed space) is called pre-ideal if for each x ∈ X and
each measurable y : I → U the relation |y(s)| ≤ |x(s)| (for almost all s ∈ I) implies
y ∈ X and kyk ≤ kxk. If X is also complete, it is called an ideal space.
Ideal spaces are a very general class of normed spaces of measurable functions,
which includes Lebesgue, Orlicz, Lorentz, and Marcinkiewicz spaces as well as
weighted and combined forms of these spaces. Sometimes these spaces are also
called Banach function spaces or (normed) Köthe spaces.
1.3
Linear and nonlinear operators.
In this section we define and discuss some properties of the nonlinear superposition
operators and many integral operators that are needed throughout this dissertation
such as Fredholm, Volterra and Urysohn operators in Lp (I), p ≥ 1, L∞ (I) and
LM (I) spaces. We will distinguish between two different cases: when the operators take their values in Lebesgue (Orlicz) spaces Lp (I) (LM (I)) or in a space of
essentially bounded functions L∞ (I).
1.3.1
The superposition operators.
One of the most important operator studied in nonlinear functional analysis is the
so-called superposition operator [10].
Definition 1.3.1. Assume that a function f : I × R → R satisfies the Carathéodory
conditions i.e. it is measurable in t for any x ∈ R and continuous in x for almost
all t ∈ I. Then to every function x(t) being measurable on I we may assign the
function
Ff (x)(t) = f (t, x(t)), t ∈ I.
The operator Ff in such a way is called the superposition (Nemytskii) operator generated by the function f .
Furthermore, for every f ∈ L1 and every φ : I → I we define the superposition
operator generated by the functions f and φ, Fφ,f : L1 (I) → L1 (I) as
Fφ,f (t) = f (t, x(φ(t))) ,
t∈I
Lemma 1.3.1. ([10, Theorem 17.5]) Assume that a function f : I × R → R
satisfies Carathéodory conditions. Then the superposition operator F transforms
measurable functions into measurable functions.
7
Lemma 1.3.2. ([84, Lemma 17.5] in S and [102] in LM ) Assume that a function
f : I × R → R satisfies Carathéodory conditions. The superposition operator F
maps a sequence of functions convergent in measure into a sequences of functions
convergent in measure.
We will be interested in the case when F acts between some Lebesgue (Orlicz)
spaces.
In Lp (I) we have the ”automatic” continuity of the Nemytskii operator ([10, 81]):
Theorem 1.3.1. Let f satisfies the Carathéodory conditions. The superposition
operator F generated by the function f maps continuously the space Lp (I) into Lq (I)
(p, q ≥ 1) if and only if
p
(1.1)
|f (t, x)| ≤ a(t) + b · |x| q ,
for all t ∈ I and x ∈ R, where a ∈ Lq (I) and b ≥ 0.
This theorem was proved by Krasnoselskii [81] in the case when I is a bounded
interval. The generalization to the case of an unbounded interval I was given by
Appell and Zabrejko [10].
Remark 1.3.1. It should be also noted that the superposition operator F takes its
values in L∞ (I) iff the generating function f is independent of x (cf. [10, Theorem
3.17]).
Lemma 1.3.3. ([83, Theorem 17.5]) Assume that a function f : I × R → R
satisfies Carathéodory conditions. Then
M2 (f (s, x)) ≤ a(s) + bM1 (x),
where b ≥ 0 and a ∈ L1 (I), if and only if the superposition operator F acts from
LM1 (I) to LM2 (I).
In Orlicz spaces there is no automatic continuity of superposition operators like
in Lp spaces, but the following lemma is useful (remember, that the Orlicz space
LM is ideal and if M satisfies ∆2 condition it is also regular cf. [7, Theorem 1]):
Lemma 1.3.4. ([118, Theorem 5.2.1]) Let f be a Carathéodory function, X an ideal
space, and W a regular ideal space. Then the superposition operators F : X → W
is continuous.
Let us note, that in the case of functions of the form f (t, x) = g(t)h(x), the
superposition operator F is continuous from the space of continuous functions C(I)
into LM (I) even when M does not satisfies ∆2 condition ([7]). Since EM (I) is a
regular part of an Orlicz space LM (I) (cf. [119, p.72]), in the context of Orlicz
spaces, we will use the following (see also Lemma 1.3.3):
8
Lemma 1.3.5. Let f be a Carathéodory function. If the superposition operator F
acts from LM1 (I) into EM2 (I), then it is continuous.
The problem of boundedness of such a type of operators will be described in the
proofs of our main results.
Remark 1.3.2. Let us recall, that the acting condition from Lemma 1.3.3 is not
sufficient for taking EM1 (I) into EM2 (I) (cf. [10, p.95]), especially for the continuity
of this operator. For the case considered when M1 = M2 we can put, for example, f (t, x) = x to fulfil this requirement. But this is true also for an arbitrary
Carathéodory function f when M1 satisfies the ∆2 -condition. For a general result
of this type see [83, Th. 17.7].
Remark 1.3.3. Let X, Y be ideal spaces. A superposition operator F : X → Y is
called improving if it takes bounded subsets of X into the subsets of Y with equiabsolutely continuous norms.
The following two theorems ”Lusin and Dragoni” [56, 109], which explain the
structure of measurable functions and functions satisfying Carathéodory conditions,
where D c denotes the complement of D and the symbol meas(D) stands for the
Lebesgue measure of the set D.
Theorem 1.3.2. Let m : I → R be a measurable function. For any ε > 0 there
exists a closed subset Dε of the interval I such that meas(Dεc ) ≤ ε and m|Dε is
continuous.
Theorem 1.3.3. Let f : I×R → R be a function satisfying Carathéodory conditions.
Then for each ε > 0 there exists a closed subset Dε of the interval I such that
meas(Dεc ) ≤ ε and f |Dε ×R is continuous.
1.3.2
Fredholm integral operator.
Assume that k : I × I → R be measurable with respect to both variables. For an
arbitrary x ∈ Lp (I) let
Z
(K0 x)(t) = k(t, s)x(s) ds, t ∈ I.
(1.2)
I
This operator K is linear and is called Fredholm integral operator (cf. [84, 122]).
The next theorem gives a sufficient conditions which which ensure that K maps
from Lp into Lq and is continuous.
Theorem 1.3.4. [122] Let k : I × I → R be measurable with respect to both
variables. Let the linear integral operator K0 with kernel k(t, s) map Lp into Lq .
Then it is continuous.
9
Lemma 1.3.6. [78] Let k : I ×I → R be measurable with respect to both variables.
Let the linear integral operator K0 with kernel k(·, ·) maps Lp (I) into L∞ (I) i.e.
either
Z b
p1
p
<∞
essupt∈[a,b]
|K(t, s)| ds
a
or
Z
a
Then it is continuous.
b
essups∈[a,b] |K(t, s)|
p
p1
< ∞.
dt
The necessary results concerning the properties of such a kind of operators in
Orlicz spaces can be found in [83], let we mention Zaanen’s theorem [83] which shows
that the operator (1.2) acts between Orlicz spaces.
Let, the N-functions M1 and M2 are the complementary functions to the Nfunctions N1 and N2 respectively.
Lemma 1.3.7. Suppose the kernel k(x, y) satisfies either one of the following two
conditions:
(a) for almost all t ∈ I the kernel k(t, s), as a function of s, belongs to the space
LN1 , where the function ϕ(t) = kk(t, s)kN1 belongs to the space LM2 ,
(b) for almost all s ∈ I the kernel k(t, s), as a function of t, belongs to the space
LM2 , where the function Ψ(s) = kk(t, s)kM2 belongs to the space LN1 .
Then the operator (1.2) maps LM1 into LM2 and is continuous.
1.3.3
Volterra integral operator.
Suppose k : ∆ → R is a given function and measurable with respect to both
variables where ∆ = {(t, s) : 0 ≤ s ≤ t ≤ ∞}. For an arbitrary function x ∈ L1 (R+ )
define
Z t
(V x)(t) =
k(t, s)x(s) ds,
0
t ≥ 0.
The linear integral operator defined above is the well known linear Volterra integral
operator (cf. [84, 122]).
When we consider this operator on the space Lp ([a, b]), then it is a special case of
the Fredholm operator investigated in previous section, where
Z b
(V x)(t) =
χ[0,t] k(t, s)x(s) ds, t ≥ 0.
a
10
1.3.4
Urysohn integral operator.
The most important nonlinear integral operators are the Urysohn operators [122]:
Z
U(x)(t) =
u(t, s, x(s)) ds.
(1.3)
I
Here, the kernel u : I × I × R → R satisfies Carathéodory conditions i.e. it is
measurable in (t, s) for any x ∈ R and continuous in x for almost all (t, s) ∈ I × I.
Moreover, for arbitrary fixed s ∈ I and x ∈ R the function t → u(t, s, x(s)) is
integrable.
A particular case of a Urysohn operator (1.3) is the Hammerstein integral operator
H = K0 ◦ F :
H(x)(t) =
Z
k(t, s)f (s, x(s)) ds.
(1.4)
I
Note, that for Urysohn operators the continuity is not ”automatic” as in the case
of superposition operators (for Nemytskii operators see Theorem 1.3.1).
Let us recall an important sufficient condition:
Theorem 1.3.5. [84, Theorem 10.1.10] Let u : I ×I ×R → R satisfies Carathéodory
conditions i.e. it is measurable in (t, s) for any x ∈ R and continuous in x for almost
R
all (t, s) ∈ I × I. Assume that U(x)(t) = I u(t, s, x(s))ds maps Lp (I) into Lq (I)
(q < ∞) and for each h > 0 the function
Rh (t, s) = max |u(t, s, x)|
|x|≤h
is integrable on s for a.e. t ∈ I. If moreover for each h > 0 this operator satisfies
Z
lim
sup k
u(t, s, x(s))dskLq (I) = 0
meas(D)→0 |x|≤h
D
and for arbitrary non-negative z(t) ∈ Lp (I)
Z
lim sup k
u(t, s, x(s)) dskLq (I) = 0,
D→0 |x|≤z
D
then U is a continuous operator.
The first two conditions are satisfied when
R
I
Rh (t, s)ds ∈ Lq (I), for instance.
We will use also the majorant principle for Urysohn operators (cf. [84, Theorem
10.1.11]. The following theorem which is a particular case of much more general
result ([84, Theorem 10.1.16]), will be very useful in the proof of the main result for
operators in L∞ (I):
11
Theorem 1.3.6. [84] Let u : I × I × R → R satisfies Carathéodory conditions i.e.
it is measurable in (t, s) for any x ∈ R and continuous in x for almost all (t, s).
Assume that
|u(t, s, x)| ≤ k(t, s) · (a(s) + b · |x|),
where the nonnegative function k is measurable in (t, s), a is a positive integrable
function, b > 0 and such that the linear integral operator with the kernel k(t, s) maps
L1 (I) into L∞ (I). Then the operator U maps L1 (I) into L∞ (I). Moreover, if for
arbitrary h > 0
Z
lim k
max
|u(t, s, x1 ) − u(t, s, x2 )| dskL∞ (I) = 0,
δ→0
D |xi |≤h,|x1−x2 |≤δ
then U is a continuous operator.
We mention also that some particular conditions guaranteeing the continuity of
the operator U may be found in [116, 122].
1.3.5
The multiplication operator.
We need to describe the multiplication operator which is the key point of our work.
We will denote the pointwise multiplication operator by A(x)(t) of the form:
A(x)(t) = F (x)(t) · U(x)(t),
where U(x) is a Urysohn integral operator (1.3), in some chapters replaced by the
the Hammerstein integral operator H = K0 F , where K0 is the linear integral operator and F as in definition 1.3.1.
Generally speaking, the product of two functions x, y ∈ Lp (I)[LM (I)] is not in
Lp (I) [LM (I)]. However, if x and y belongs to some particular Lebesgue (Orlicz)
spaces, then the product x· y belong to a third Lebesgue (Orlicz) space. Let us note,
that one can find two functions belonging to Lebesgue (Orlicz) spaces: u ∈ Lp (LU )
and v ∈ Lp (LV ) such that the product uv does not belong to any Lebesgue (Orlicz)
space (this product is not integrable).
We will use the technique of factorization for some operators acting on Lebesgue
(Orlicz) spaces through another Lebesgue (Orlicz) spaces. We can mention, that by
using in this place different ideal spaces it is possible to obtain some extensions of
our results and then we try to facilitate this approach. To stress the connection of
our results with the growth condition we restrict ourselves to the case of Lebesgue
(Orlicz) spaces.
12
Remark 1.3.4. For the so-called pre-ideal spaces (cf. [118]) if x ∈ E and y ∈ L∞
implies that xy ∈ E and kxykE ≤ kxkE kykL∞ i.e. the elements from L∞ are
pointwise multipliers for E. For more details in L1 space see [46].
Nevertheless, we have:
Lemma 1.3.8. ([83, Lemma 13.5]), [92, Theorem 10.2] Let ϕ1 , ϕ2 and ϕ are arbitrary N-functions. The following conditions are equivalent:
1. For every functions u ∈ Lϕ1 (I) and w ∈ Lϕ2 , u · w ∈ Lϕ (I).
2. There exists a constant k > 0 such that for all measurable u, w on I we have
kuwkϕ ≤ kkukϕ1 kwkϕ2 .
≤
3. There exists numbers C > 0, u0 ≥ 0 such that for all s, t ≥ u0 , ϕ st
C
ϕ1 (s) + ϕ2 (t).
4. lim supt→∞
−1
ϕ−1
1 (t)ϕ2 (t)
ϕ(t)
< ∞.
Let us recall the following simple sufficient condition for the above statements
hold true.
Lemma 1.3.9. ([83, p. 223]) If there exist complementary N-functions Q1 and Q2
such that the inequalities
Q1 (αu) < ϕ−1 [ϕ1 (u)]
Q2 (αu) < ϕ−1 [ϕ2 (u)]
hold, then for every functions u ∈ Lϕ1 (I) and w ∈ Lϕ2 , u · w ∈ Lϕ (I). If moreover
ϕ satisfies the ∆2 -condition, then it is sufficient that the inequalities
Q1 (αu) < ϕ1 [ϕ−1 (u)]
Q2 (αu) < ϕ2 [ϕ−1 (u)]
hold.
An interesting discussion about necessary and sufficient conditions for product
operators can be found in [83, 92].
Remark 1.3.5. An ideal space E is called regular if for every x ∈ E we have
limmesD→0 kx · χD kE = 0. A set of all elements x with this property is called a
regular part of E. Thus this a set of all x ∈ E with absolutely continuous norm. A
space is called perfect if the Fatou lemma holds for E.
13
1.4
Monotone functions.
Let S = S(I) denote the set of measurable (in Lebesgue sense) functions on I and
let meas stand for the Lebesgue measure in R. Identifying the functions equals
almost everywhere the set S furnished with the metric
d(x, y) = inf [a + meas{s : |x(s) − y(s)| ≥ a}]
a>0
becomes a complete space. Moreover, the space S with the topology convergence in
measure on I is a metric space, because the convergence in measure is equivalent to
convergence with respect to d (cf. Proposition 2.14 in [119]).
For σ-finite subsets of R we say that the sequence xn is convergent in finite
measure to x if it is convergent in measure on each set T of finite measure.
The compactness in such spaces we will call a ”compactness in measure” and
such sets have important properties when considered as subsets of some Orlicz spaces
(ideal spaces). Let us recall, in metric spaces the set U0 is compact if and only if
each sequence from U0 has a subsequence that converges in U0 (i.e. sequentially
compact).
In this dissertation, we need to investigate some properties of sets and operators
in such a class of spaces instead of the space S. Some of them are obvious, the rest
will be proved.
We are interested in finding of almost everywhere monotonic solutions for our
problems. We will need to specify this notion in considered solution spaces.
Let X be a bounded subset of measurable functions. Assume that there is a
family of subsets (Ωc )0≤c≤b−a of the interval I such that meas(Ωc ) = c for every
c ∈ [0, b − a], and for every x ∈ X, x(t1 ) ≥ x(t2 ), (t1 ∈ Ωc , t2 6∈ Ωc ).
It is clear, that by putting Ωc = [0, c) ∪ Z or Ωc = [0, c) \ Z, where Z is a set
with measure zero, this family contains nonincreasing functions (possibly except for
a set Z). We will call the functions from this family ”a.e. nonincreasing” functions.
This is the case, when we choose a measurable and nonincreasing function y and
all functions equal a.e. to y satisfies the above condition. This means that such
a notion can be also considered in the space S. Thus we can write, that elements
from L1 (I), LM (I) belong to this class of functions. Further, let Qr stand for the
subset of the ball Br consisting of all functions which are a.e. nonincreasing on I.
Functions a.e. nondecreasing are defined by similar way.
It is known, that such a family constitute a set which is compact in measure in S.
We are interested, if the set is still compact in measure as a subset of subspaces of
S. In general, it is not true, but for the case of Lebesgue spaces L1 (I), Lp (I), p > 1
and Orlicz spaces LM (I), we have the following:
14
Due to the compactness criterion in the space of measurable functions (with
the topology of the convergence in measure) (see Lemma 4.1 in [18]) we have a
desired theorem concerning the compactness in measure of a subset X of L1 (I) (cf.
Corollary 4.1 in [18] or Section III.2 in [60]).
Theorem 1.4.1. Let X be a bounded subset of L1 (I) consisting of functions which
are a.e. nonincreasing (or a.e. nondecreasing) on the interval I. Then X is compact
in measure in L1 (I).
In the following theorems, denote by E the spaces Lp (I), p ≥ 1 or LM (I) (cf.
[47, 48]).
We have a new characterization of compactness in measure for subspaces of S.
Lemma 1.4.1. Let X be a bounded subset of E consisting of functions which are
a.e. nondecreasing (or a.e. nonincreasing) on the interval I. Then X is compact in
measure in E.
Proof. Let r > 0 be such that X ⊂ Br ⊂ E. It is known (cf. [84, 18]), that X is
compact in measure as a subset of S. By taking an arbitrary sequence (xn ) in X we
obtain that there exists a subsequence (xnk ) convergent in measure to some x ∈ S.
Since Orlicz spaces are perfect (cf. [118]), the balls in E are closed in the topology
of convergence in measure. Thus x ∈ Br ⊂ E and then x ∈ X.
Remark 1.4.1. The above lemma remains true for subsets of arbitrary perfect ideal
spaces ([118]).
If we consider the set of indices c ≥ 0 in the definition of the family of a.e.
nonincreasing functions, we are able to extend this result for the space L1 (R+ ). For
simplicity, we will denote such a space by L1 . Due to some results of Väth we are
able to extend the desired result from the interval I = [a, b] into the σ-finite subsets
of R and the topology of the convergence in finite measure.
Theorem 1.4.2. Let X be a bounded subset of L1 (R+ ) consisting of functions which
are a.e. nonincreasing (or a.e. nondecreasing) on the half-line R+ . Then X is
compact in finite measure in L1 (R+ ).
Proof. If we consider the space L1 (T ) for σ-finite measure space T , then there is
some equivalent finite measure ν (ν(R+ ) < ∞) (Proposition 2.1. in [119] or Corollary
2.20 in [119]). Then the convergence of sequences in S are the same for the metric
d and for
dν (x, y) = inf [a + ν{s : |x(s) − y(s)| ≥ a}]
a>0
(Proposition 2.2 in [117]). Take an arbitrary bounded sequence (xn ) ⊂ X. As a
subset of a metric space X = (L1 (R+ ), dν ) the sequence is compact in this metric
15
space (Theorem 1.4.1). Then there exists a subsequence (xnk ) of (xn ) which is
convergent in the space X to some x i.e.
k→∞
dν (xnk , x) −→ 0.
As claimed above the two metrics have the same convergent sequences, then
k→∞
d(xnk , x) −→ 0.
This means that X is compact in L1 (R+ ).
We have also an important
Lemma 1.4.2. (Lemma 4.2 in [15]) Suppose the function t → f (t, x) is a.e. nondecreasing on a finite interval I for each x ∈ R and the function x → f (t, x) is a.e.
nondecreasing on R for any t ∈ I. Then the superposition operator F generated by
f transforms functions being a.e. nondecreasing on I into functions having the same
property.
We will use the fact, that the superposition operator takes the bounded sets
compact in measure into the sets with the same property.
Thus we can prove the following (cf. [46, Proposition 4.1]):
Proposition 1.4.1. Assume that a function f : I × R → R satisfies Carathéodory
conditions and the function t → f (t, x) is a.e. nondecreasing on a finite interval I
for each x ∈ R and the function x → f (t, x) is a.e. nondecreasing on R for any
t ∈ I. Assume, that F : LM (I) → EM (I). Then F (V ) is compact in measure for
arbitrary bounded and compact in measure subset V of LM (I).
Proof. Let V be a bounded and compact in measure subset of LM (I). By our
assumption F (V ) ⊂ EM (I). As a subset of S the set F (V ) is compact in measure (cf.
[18]). Since the topology of convergence in measure is metrizable, the compactness
of the set is equivalent with the sequential compactness. By taking an arbitrary
sequence (yn ) ⊂ F (V ) we get a sequence (xn ) in V such that yn = F (xn ). Since
(xn ) ⊂ V , as follows from Lemma 1.3.2 F transforms this sequence into the sequence
convergent in measure. Thus (yn ) is compact in measure, so is F (V ).
For the integral operator (1.2), we have the following theorem due to Krzyż ([85,
Theorem 6.2]):
Theorem 1.4.3.
The operator K0 preserve the monotonicity of functions iff
Z b
Z b
k(t1 , s) ds ≥
k(t2 , s) ds
0
0
for t1 < t2 , t1 , t2 ∈ I and for any b ∈ I.
16
1.5
Measures of noncompactness.
Now we present the concept of a regular measure of noncompactness (or of weak
noncompactness ), we denote by ME the family of all nonempty and bounded
subsets of E and by NE , NEW its subfamily consisting of all relatively compact and
weakly relatively compact sets, respectively. The symbol X̄ W stands for the weak
closure of a set X while X̄ denotes its closure.
Definition 1.5.1. [23] A mapping µ : ME → [0, ∞) is said to be a measure of
noncompactness in E if it satisfies the following conditions:
(1) the family kerµ = {X ∈ ME : µ(X) = 0} is nonempty and kerµ ⊂ NE ,
where kerµ is called the kernel of the measure µ.
(2) X ⊂ Y ⇒ µ(X) ≤ µ(Y ).
(3) µ (convX) = µ (X)
(4) µ [ λ X + (1 − λ) Y ] ≤ λ µ(X) + (1 − λ) µ(Y ), λ ∈ [0, 1].
(5) If Xn ∈ ME , Xn = X̄n and Xn+1 ⊂ Xn for n = 1, 2, . . . and if
lim µ(Xn ) = 0, then X∞ =
n→∞
∞
\
n=1
Xn 6= φ.
Definition 1.5.2. [29] A mapping γ : ME → [0, ∞) is said to be a measure of
weak noncompactness in E if it satisfies conditions (2)-(4) of definition 1.5.1 and
the following two conditions (being counterparts of (1) and (5)) hold:
(1’) the family kerγ = {X ∈ ME : γ(X) = 0} is nonempty and kerγ ⊂ NEW ,
where kerγ is called the kernel of the measure γ.
(5’) If Xn ∈ ME , Xn = X̄n
W
and Xn+1 ⊂ Xn for n = 1, 2, . . . and if
lim γ(Xn ) = 0, then X∞ =
n→∞
∞
\
n=1
Xn 6= φ.
In addition the measure of noncompactness µ (or of weak noncompactness γ)
is called
S
• Measure with maximum property if µ(X Y ) = max [ µ(X), µ(Y ) ].
• Homogeneous measure if µ (λX) = |λ| µ(X), λ ∈ R.
• Subadditive measure if µ (X + Y ) ≤ µ (X) + µ (Y ).
• Sublinear measure if it is homogeneous and subadditive.
• Complete (or full) if ker µ = NE (ker γ = NEW ) .
• Regular measure if it is full, sublinear and has a maximum property.
17
An classical example of measure of noncompactness is the following:
Definition 1.5.3. [23] Let X be a nonempty and bounded subset of E. The
Hausdorff measure of noncompactness βH (X) is defined as
βH (X) = inf{ǫ > 0 : Xcan be covered with a finite number of balls of a radii less than ǫ}
It is worthwhile to mention that the first important example of measure of weak
noncompactness has been defined by De Blasi [52] by:
β(X) = inf {r > 0 : there exists a weakly compact subset W of E such that x ⊂ W +Br }.
Both the Hausdorff measure βH and the De Blasi measure β are regular in the
sense of the above definitions.
Another regular measure of noncompactness was defined in the space L1 (I) (cf.
[28]). For any ε > 0, let c be a measure of equiintegrability of the set X (the so-called
Sadovskii functional [10, p. 39]) i.e.
Z
c(X) = lim sup{sup{sup[ |x(t)| dt, D ⊂ I, meas(D) ≤ ε]}}.
ε→0
x∈X
D
Restricted to the family compact in measure subsets of this space it forms a regular
measure of noncompactness (cf. [66]).
However, by considering this measure of noncompactness instead of usually considered ones based on Kolomogorov or Riesz criteria of compactness (cf. [23]) we
are able to examine by the same manner the case of Lp (I) spaces, where χD denotes
the characteristic function of D.
Let us also denote by c a measure of equiintegrability of the set X in an Orlicz
space LM (I) (cf. Definition 3.9 in [119] or [67, 66]):
c(X) = lim sup
ε→0
sup
sup kx · χD kLM (I) ,
meas D≤ε x∈X
where χD denotes the characteristic function of D.
Then we have the following theorem, which clarify the connections between different coefficients in Orlicz spaces. Since Orlicz spaces LM (I) are regular, when M
satisfies ∆2 condition, then Theorem 1 in [66] read as follows:
Proposition 1.5.1. Let X be a nonempty, bounded and compact in measure subset
of an ideal regular space Y . Then
βH (X) = c(X).
18
As a consequence, we obtain that bounded sets which are additionally compact
in measure are compact in LM (I) iff they are equiintegrable in this space (i.e. have
equiabsolutely continuous norms cf. [5]).
The contraction of the measure of weak non compactness is a bit more complicated when I is an unbounded interval. Let us, we recall the following criterion for
weak noncompactness due to Dieudonné [55, 60], which is of fundamental importance in our subsequent analysis.
Theorem 1.5.1. A bounded set X is relatively weakly compact in L1 (R+ ) if and
only if the following two conditions are satisfied:
R
(a) for any ε > 0 there exists δ > 0 such that if meas(D) < δ then D |x(t)|dt ≤ ε
for all x ∈ X,
R∞
(b) for any ε > 0 there is T > 0 such that T |x(t)|dt ≤ ε for any x ∈ X.
Now, for a nonempty and bounded subset X of the space L1 (R+ ) let us define:
Z
c(X) = lim{sup{sup [ |x(t)| dt, D ⊂ R+ , meas(D) ≤ ε]}},
(1.5)
ǫ→0 x∈X
and
D
Z
d(X) = lim {sup[
T →∞
Put
∞
T
|x(t)| dt : x ∈ X]}.
γ(X) = c(X) + d(X).
(1.6)
(1.7)
Then we have the following theorem, which clarify the connections between these
two measures βH (x) and γ(x) ([22]).
Theorem 1.5.2. Let X be a nonempty, bounded and compact in measure subset of
L1 (R+ ). Then
βH (x) ≤ γ(x) ≤ 2βH (x).
1.6
Fixed point theorems.
Fixed point theorems have always a major role in various fields, specially, in fields
of differential, integral and functional equations. Fixed point theorems constitute a
topological tool for the qualitative investigations of solution of linear and nonlinear
equations. The theory of fixed points is concerned with the conditions which guarantee that a map T : X → X of a topological space X into it self admits one or
more fixed points, that is, points x of X for which x = T x.
Here we give a brief history of fixed point theorems.
The following definition states some types of mapping in a metric space (X, ρ) [71].
19
Definition 1.6.1. Let (X, ρ) be a metric space. The mapping T : X → X is
called Lipschitz map, if there exist a number γ ≥ 0, such that
ρ(T x, T y) ≤ γ ρ(x, y), ∀ x, y ∈ X.
The mapping T is called contraction if γ
γ ≤ 1. Furthermore, T is contractive if
< 1, and is called non expensive, if
ρ(T x, T y) < γ ρ(x, y), ∀ x 6= y.
Problems concerning the existence of fixed point for Lipschitz map have been of
considerable interest in non linear operator theory. In 1922, the so-called Banach
contraction mapping principle was given to obtain solutions for several problems.
Theorem 1.6.1. (Banach contraction mapping principle, [71])
Let X be a complete metric space and let T : X → X be a contraction map.
Then T has a unique fixed point in X. Moreover, for any x0 ∈ X, the sequence
{T n (x0 )}∞
n=0 converges to the fixed point.
This theorem is perhaps the most useful fixed point theorem, which is involved in
many of the existence and uniqueness proofs in ordinary differential equations. The
mapping T is the Banach contraction mapping principle still has a unique fixed
point in any closed subset M of X. There are some conditions for a continuous
mapping T in X, that guarantee the existence of a unique fixed point, such as the
contraction of T n or if there exist a function φ : X → R+ , such that for all
x ∈ X, ρ (T x, T y) = φ(x) φ(T x).
In a normed space, the next fixed point theorem, is concerned with continuous
mapping and has an advantage over Banach Contraction Mapping Principle in that
is applied to a large class of functions.
Theorem 1.6.2. (Brouwer [71])
Let Q be a nonempty, convex, closed and bounded subset of a finite dimensional
Banach space En and let T : Q → Q be continuous. Then T has at least one fixed
point in the set Q.
A generalization of Brouwer’s result to any Banach space was due to Schauder.
Theorem 1.6.3. (Schauder, [71])
Let Q be a convex subset of a Banach space X, and T : Q → Q is compact,
continuous map. Then T has at least one fixed point in Q.
Next, we need the following definition.
20
Definition 1.6.2. [71] A mapping H : E → E is called completely continuous if
H is continuous and H(Y ) is relatively compact for every bounded subset of Y .
Theorem 1.6.4. (Schauder-Tychonoff, [71])
Let C be a nonempty, convex, closed and bounded subset of a Banach space E. Let
H : C → C be a completely continuous mapping. Then H has at least one fixed
point in C.
When the concept of measure of noncompactness appeared, some fixed point
theorems based on such measure were given. Among these is the Darbo fixed point
theorem. Such theorem is used for a contraction mapping with respect to the Hausdorff measure of non compactness, that is, there exist a constant α ∈ (0, 1), such
that χ(HX) ≤ α χ(X), for any nonempty bounded subset X of G.
An importance of such a kind of functions can be clarified by using the contraction property with respect to this measure instead of compactness in the Schauder
fixed point theorem. Namely, we have a theorem ([23]).
Theorem 1.6.5. (Darbo, [50])
Let Q be a nonempty, bounded, closed and convex subset of E and let V : Q → Q
be a continuous transformation which is a contraction with respect to the measure
of noncompactness µ, i.e. there exists k ∈ [0, 1) such that
µ(V (X)) ≤ kµ(X),
for any nonempty subset X of E. Then V has at least one fixed point in the set Q
and the set F ixV of all fixed points of V satisfies µ(F ixV ) = 0.
Emmanuele gives the corresponding version of Darbo fixed point theorem in the
weak sense.
Theorem 1.6.6. (Emmanuele, [65] )
Let Q be a nonempty, closed, convex and bounded subset of a Banach space E,
Assume that F : Q → Q be a weakly continuous operator having the proprty that,
there is a constant α ∈ (0, 1), such that β (F (X)) ≤ α β(X), for any nonempty
subset X of Q, where β(X) is the measure of noncompactness. Then F has at least
one fixed point in the set Q.
Theorem 1.6.7. [82]
Let M be a nonempty, closed, and convex subset of E. Suppose, that A, B be two
operators such that
i) A(M) + B(M) ⊆ M,
21
ii) A is a contraction mapping,
iii) B(M) is relatively compact and B is continuous. Then there exists a y ∈ M with
Ay + By = y.
Next we state a nonlinear alternative of Leray-Schauder type fixed point theorem
(cf. [51]).
Theorem 1.6.8. (the Leray-Schauder alternative)
Let C be an open subset of a convex set Q in a Banach space E. Assume 0 ∈ C and
the map T : C̄ → Q is continuous and compact. Then either
(i) T has a fixed point in C̄, or
(ii) there exist λ ∈ (0, 1) and u ∈ ∂C such that u = λT u, where ∂C is a boundary of
U.
The relative compactness for a subset in Lp (0, 1) can be proved by a several
methods, among these, Kolmogorov compactness criterion stated in the following
theorem [59].
Theorem 1.6.9. (the Kolmogorov compactness criterion)
Let Ω ⊆ Lp (0, 1), 1 ≤ p < ∞. If
(i) Ω is bounded in Lp (0, 1),
(ii) xh → x as h → 0 uniformly with respect to x ∈ Ω, then Ω is relatively compact
in Lp (0, 1), where
Z t+h
1
xh (t) =
x(s) ds.
h t
Theorem 1.6.10. (the Arzela-Ascoli theorem , [80])
Let E be a compact metric space and C(E) be the Banach space of real or complex
valued continuous functions normed by
kf k =
sup | f (t) |.
t ∈ E
If A = {fn } is a sequence in C(E) such that fn is uniformly bounded and equicontinuous, then Ā is compact.
Theorem 1.6.11. (the Lebesgue dominated convergence theorem, [80])
Let {fn } be a sequence of functions converging to a limit f on A, and suppose that
| fn (t) | ≤ φ(t),
t ∈ A, n = 1, 2, . . . ,
where φ is integrable on A. Then f is integrable on A and
Z
Z
lim
fn (t) dµ =
f (t) dµ.
n → ∞
A
A
22
Chapter 2
Monotonic integrable solutions for
quadratic integral equations on a
half line.
2.1
Motivations and historical background.
This Thesis is devoted to study so-called quadratic integral equations. This is a
kind of problems of the form
Z β
x(t) = g(t) + F (x)(t) ·
u(t, s, x(s)) ds,
α
where t ∈ I ⊂ R+ and F is an operator. Some generalizations for the presented
equations are also considered. Such a kind of problems is of mathematical and
practical interests and has a long history. A classical theory of Urysohn integral
equations does not include the above problem. Since for equations of this type
an approach via the Schauder fixed point theorem is not useful and the Banach
contraction principle is too restrictive in many applications, we need to investigate
such equations very carefully.
The first considered equation of this type is the Chandrasekhar equation
Z 1
t
ϕ(s)x(s) ds.
x(t) = 1 + x(t)
0 t+s
It is an important example, because it show some of our motivations. This equation
describe a radiative transfer through a homogeneous stellar atmosphere. It was investigated by many authors. The solutions was considered only in the space C(I) or
in Banach algebras (cf. [34]). However, such a class of solutions seems to be inadequate for integral problems and leads to several restrictions on functions. In order
to apply earlier results we have to impose an additional condition that the so-called
23
”characteristic” function ψ is a polynomial (as in the book of Chandrasekhar [43,
Chapter 5]) or at least continuous (cf. [40, Theorem 3.2]). This function is immediately related to the angular pattern for single scattering and then our results allow
to consider some peculiar states of the atmosphere. In astrophysical applications of
R1
the Chandrasekhar equation the only restriction, that 0 ψ(s) ds ≤ 1/2 is treated as
necessary (cf. [40, Chapter VIII; Corollary 2 p. 187] or [69]). The continuity assumption for ψ implies the continuity of solutions for the considered equation (cf. [40])
and then seems to be too restrictive even from the theoretical point of view. About
nonhomogeneous (discontinuous ”characteristic” ψ in the Chandrasekhar equation)
stellar atmosphere: it is only a discretization for the equation (Hollis and Kelley
1986 [74]) - till now there is no analytical methods (unless our results). An interesting discussion about the continuity of solutions for the Chandrasekhar equation
and the relation between the kernel of an integral operator can be found in [114,
Proposition 4.1, Theorem 4.3] - cf. also [69].
More general problem (motivated by some practical interests in plasma physics
(cf. Stuart [114]) was investigated in [86]
Z 1
J
2
2
K(t, s, x(t), x(s)) ds.
x (t) = t −
4π 0
Let us list some of considered previously particular cases of quadratic integral equations with their applications:
a) biology: model of spread of a disease (epidemic model) (Gripenberg [72] )
Z t
Z t
x(t) = k P −
A(t − s)x(s)ds ·
a(t − s)s(s)ds,
−∞
−∞
b) physics: kinetic theory of gases (Hu, Khavanin and Zhuang [75])
Z ∞
Z ∞
x(t) = a(t) + f (t, x(t)) +
g(t, s)x(s)ds ·
h(t, s)K(s, x(s))ds,
0
0
c) physics: statistical mechanics, the Percus-Yevick equation (Nussbaum [96],
Wertheim [121], Pimbley [101], Ramalho [104], Rus [107])
Z 1
x(t) = 1 + λ
x(s) · x(s − t)ds,
t
c) the Chandrasekhar equation: in astrophysics (Chandrasekhar, Fox, Argyros,
Crum, Cahlon, Rus, Shrikhant, Joshi, Schillings, Leonard and Mullikin [94, 95, 88],
Stuart [114] and many others):
Z 1
t
x(t) = 1 + x(t)
ψ(s)x(s) ds.
0 t+s
It is worthwhile to mention, that our equation cover as special cases among
others the following ones:
24
1. f1 (t, x) = g(t), f2 (t, x) = λ the functional Urysohn integral equation ([14, 15,
24]),
2. f1 (t, x) = g(t), f2 (t, x) = x, φ2 (t) = t the functional-integral equation ([91]),
3. f2 (t, x) = 0 the abstract functional equation ([15], for instance),
4. for continuous solutions with φ1 (t) = φ2 (t) = t and
u1 (t,s,x)
u(t, s, x) = Γ(α)·(t−τ
see [35, 61],
)1−α
5. f2 (t, x) = λ the functional integral equation (for continuous solutions see [3,
21, 54]),
6. f2 (t, x) = x the quadratic (functional) Urysohn integral equation ([27, 26], for
instance).
Note, that the choice of spaces allow us to consider less restrictive growth conditions, which will be clarified in next chapters.
2.2
Introduction.
In this chapter we study the following functional integral equation
Z β
x(t) = g(t) + f t, x(t) ·
u(t, s, x(s)) ds .
(2.1)
α
The particular cases of our equation, were investigated for existence for both
continuous (cf. [6, 32, 35, 62] and integrable solutions ([20, 30, 31]). The existence
of different subclasses of solutions were proved (nonnegative functions, monotone,
having limit at infinity etc.).
Let us note, that the problem is investigated for finite or infinite intervals. We
extend the existing results dealing the monotonicity problem in a half-line for the
most complicated problem of the Urysohn operators. For continuous solutions such
a property was recently investigated in [62], for instance.
By applying Darbo fixed point theorem associated with the measure of noncompactness, we obtain the sufficient conditions for the existence of monotonic solutions
of equation (2.1), which are integrable. The results presented in this chapter are
motivated by the recent works of Banaś and Chlebowicz [20], Banaś and Rzepka
[32, 33] and extend these papers in many ways.
25
2.3
Main result
Denote by L1 for L1 (R+ ) and H the operator associated with the right hand side of
equation (2.1) which takes the form
x = Hx,
where
Z
(Hx)(t) = g(t) + f t, x(t) ·
β
u(t, s, x(s)) ds ,
α
t ≥ 0.
The operator H will be written as the product Hx(t) = g(t) + F Kx(t) where
F (x)(t) = f (t, x(t)),
Kx(t) = x(t) · U(x)(t)
and U(x) is the Urysohn integral operator of the form
Z β
(Ux)(t) =
u(t, s, x(s)) ds.
α
Thus equation (2.1) becomes
x(t) = g(t) + F Kx(t).
(2.2)
We shall treat the equation (2.1) under the following assumptions which are listed
below.
(i) g ∈ L1 (R+ ) and is a.e. nonincreasing on R+ .
(ii) f : R+ × R → R satisfies Carathéodory conditions and there are a positive
function a ∈ L1 and a constant b ≥ 0 such that
|f (t, x)| ≤ a(t) + b |x|,
for all t ∈ R+ and x ∈ R. Moreover, f (t, x) ≥ 0 for x ≥ 0 and f is
assumed to be nonincreasing with respect to both variable t and x separately.
(iii) u : R+ × R+ × R → R satisfies Carathéodory conditions i.e. it is measurable
in (t, s) for any x ∈ R and continuous in x for almost all (t, s). The function
u is nonincreasing with respect to each variable, separately. Moreover, for
arbitrary fixed s ∈ R+ and x ∈ R the function t → u(t, s, x(s)) is
integrable.
(iv) There exists a measurable function k such that:
|u(t, s, x)| ≤ k(t, s)
26
for all t, s ≥ 0 and x ∈ R. A measurable nonnegative function k : R+ → R+
is supposed to be nonincreasing with respect to each variable separately and
such that the linear integral operator K0 with kernel k(t, s) maps L1 into L∞ .
Moreover, for each non-negative z ∈ L1 let
Z
lim
sup k
u(t, s, x(s))dskL∞ = 0
meas(D)→0 |x|≤z
D
and assume that for arbitrary h > 0 (i = 1, 2)
Z
lim k
max
|u(t, s, x1 ) − u(t, s, x2 )| ds kL∞ = 0.
δ→0
D |xi |≤h ,|x1 −x2 |≤δ
(v) b · kK0 k∞ < 1.
Then we can prove the following theorem.
Theorem 2.3.1. Let the assumptions (i) - (v) be satisfied. Then the equation (2.1)
has at least one solution a.e. nonincreasing on R+ which is locally integrable.
Proof. First of all observe that by Assumption (ii) and Theorem 1.3.1 F is a continuous operator from L1 into itself. Moreover, by (iv) U is a continuous operator
from L1 into L∞ (see Theorem 1.3.6) and then by Hölder inequality the operator K
maps L1 into itself. Finally, for a given x ∈ L1 the function Hx belongs to L1 and
is continuous.
Using (2.2) together with assumptions (iii) and (iv), we get
kHxk
≤
≤
≤
≤
=
kgk + kF Kx(t)k
Z ∞
Z β
kgk +
[a(t) + b|x(t)|
|u(t, s, x(s))| ds ]dt
0
α
Z ∞
Z β
kgk + kak + b
|x(t)| [
k(t, s) ds ]dt
0
α
Z ∞
kgk + kak + b
[|x(t)| · kK0 (t)k∞ ]dt
0
kgk + kak + b · kK0 k∞ · kxk.
From the above estimate it follows, that there is a constant r > 0 such that H
maps the ball Br into itself. Indeed, by (v) we get
kHxk ≤ kgk + kak + b · kK0 k∞ · kxk
≤ kgk + kak + b · kK0 k∞ · r
and then we obtain that H(Br ) ⊂ Br , where
r =
kgk + kak
.
1 − bkK0 k∞
27
Further, let Qr stand for the subset of Br consisting of all functions which are a.e.
t
nonincreasing on R+ . This set is nonempty (x(t) = e r ∈ Br ∩ Qr , for instance),
bounded (by r), convex (direct calculation from the definition) and closed in L1 (R+ )
similarly as claimed in [17]. To prove the last property, let (yn ) be a sequence of
elements in Qr convergent in L1 to y. Then the sequence is convergent in finite
measure and as a consequence of the Vitali convergence theorem and of the characterization of convergence in measure (the Riesz theorem) we obtain the existence of
a subsequence (ynk ) of (yn ) which converges to y almost uniformly on R+ . Moreover,
y is still nonincreasing a.e. on R+ which means that y ∈ Qr and so the set Qr is
closed. Now, in view of Theorem 1.4.1 the set Qr is compact in measure. To see
this it suffices to put Ωc = [0, c] \ P for any c ≥ 0, where P denotes a suitable set of
with meas(P ) = 0.
Now, we will show, that H preserve the monotonicity of functions. Take x ∈ Qr ,
then x(t) is a.e. nonincreasing on R+ and consequently Kx(t) is also of the same
type in virtue of the assumption (iii) and Theorem 1.4.2. Further, F Kx(t) is a.e.
nonincreasing on R+ thanks for assumption (ii). Moreover, assumption (i) permits
us to deduce that Hx = g(t) + F Kx(t) is also a.e. nonincreasing on R+ . This
fact, together with the assertion H : Br → Br gives that H is also a self-mapping
of the set Qr . From the above considerations it follows that H maps continuously
Qr into Qr .
From now we will assume that X is a nonempty subset of Qr and the constant
ǫ > 0 is arbitrary, but fixed. Then for an arbitrary x ∈ X and for a set D ⊂ R+ ,
meas(D) ≤ ǫ we obtain
Z
D
|(Hx)(t)|dt
≤
=
Z
[ |g(t)| + a(t) + b · |x(t)| ·
Z
β
|u(t, s, x(s))| ds ] dt
Z β
kgkL1 (D) + kakL1 (D) + b · kxkL1 (D) · k
k(t, s) dskL∞
D
α
α
≤
kgkL1 (D) + kakL1 (D) + b · kK0 k∞ · kxkL1 (D) .
Hence, taking into account the obvious equality
Z
Z
lim {sup [
|g(t)| dt +
a(t) dt : D ⊂ R+ , meas(D) ≤ ǫ]} = 0
ǫ→0
D
D
and by the definition of c(X) (cf. Section 1.5) we get
c(HX) ≤ b · kK0 k∞ · c(X).
28
(2.3)
Furthermore, fixing T >
Z ∞
Z ∞
|(Hx)(t)|dt ≤
T
T
Z ∞
≤
ZT∞
≤
T
0 we arrive at the following estimate
Z β
[ |g(t)| + a(t) + b|x(t)|
|u(t, s, x(s))| ds ] dt
α
Z β
[ |g(t)| + a(t) + b|x(t)|
k(t, s) ds ] dt
α
Z ∞
Z ∞
|g(t)| dt +
a(t) dt + bkK0 k∞
|x(t)| dt.
T
T
As T → ∞, the above inequality yields
d(HX) ≤ b · kK0 k∞ · d(X),
(2.4)
where d(X) has been also defined in Section 1.5.
Hence, combining (2.3) and (2.4) we get
γ(HX) ≤ b · kK0 k∞ · γ(X),
where γ denotes our measure of noncompactness defined in Section 1.5.
The inequality obtained above together with the properties of the operator H and
the set Qr established before allow us to use Theorem 1.5.2 and as a consequence,
apply Theorem 1.6.5. This completes the proof.
Remark 2.3.1. If we assume that the functions g and t → u(t, s, x) are a.e. nondecreasing and negative then applying the same argumentation, we can show that there
exists a solution of our equation being a.e. negative and nondecreasing. Moreover,
let us remark, that the monotonicity conditions in the main theorem (and examples given below) seems to be restrictive, but they are necessary as claimed in ([32]
Example 2).
2.4
Examples
We need to show two examples of problems for which our main result is useful
and allow to extend the existing theorems. Let us recall, that we are looking for
monotonic solutions for the considered problems in a half-line.
Let us start with a classical Chandrasekhar integral equation.
R1
Example 2.4.1. In the case g(t) = 1 and f (t, x) = x(t) 0 t +t s φ(s)x(s) ds,
equation (2.1) takes the form
Z 1
t
x(t) = 1 + x(t)
φ(s)x(s) ds.
(2.5)
t + s
0
29
Equation (2.5) is the famous quadratic integral equation of Chandrasekhar which is
considered in many papers and monographs (cf. [11, 26, 43, 75] for instance).
In this case we have k(t, s) = t +t s φ(s) and as k(·, s) is increasing, we can put
m(s) = φ(s) and then for some sufficiently good functions φ our result applies
(φ(s) = e−s , for instance).
In order to illustrate the results proved in Theorem 2.3.1, let us consider the
following examples
Example 2.4.2. Let us consider the following equation
Z β
t
−t
ds.
x(t) = e + x(t)
2
2
t + s + (x(s))2
α
(2.6)
By putting g(t) = e−t , f (t, x) = x and u(t, s, x) = t2 +st2 +x2 it is easy to see, that
u is nonincreasing with respect to each variable separately and the integrability
condition is also satisfied (Assumptions (i),(ii) and (iii) are satisfied).
1
We have the following functions: k(t, s) = t2 +s
2 and since
Z
β
α
β
α
k(t, s) ds = arctan − arctan ,
t
t
⇒|
Z
β
α
k(t, s) ds| ≤ |β − α|.
Thus the expected property (Assumption (v)) for K0 holds (for sufficiently small
parameter b dependent on α and β).
Moreover, given arbitrary h > 0 and |x2 − x1 | ≤ δ we have
|u(t, s, x1 )−u(t, s, x2)| ≤ |
2htδ
t(x22 − x21 )
≤ 2
2
2
2
2
2
2
2
(t + s + x1 )(t + s + x2 )
(t + s + x21 )(t2 + s2 + x22 )
and the Assumption (iv) is satisfied.
Taking into account all the above observations and Theorem 2.3.1 we conclude
that the equation (2.6) has at least one solution x = x(t) defined, integrable and
a.e. nonincreasing on R+ .
30
Chapter 3
On some integrable solutions for
quadratic functional integral
equations
3.1
Introduction
The object of this chapter is to study the solvability of a nonlinear Urysohn functional integral equation
Z 1
x(t) = f1 (t, x(φ1 (t))) + f2 (t, x(t)) ·
u(t, s, x(φ2(t))) ds, t ∈ I.
(3.1)
0
Special cases for considered equation (quadratic integral equations) were investigated in connection with some applications of such a kind of problems in the
theories of radiative transfer, neutron transport and in the kinetic theory of gases
(cf. [12, 26, 40, 43]). More general problem (motivated by some practical interests
in plasma physics) was investigated in [86]. The existence of continuous solutions
for particular cases of the considered problem was investigated since many years
(see [33, 79] or a very recent paper [4]). On the other hand, different kind of integral equations (including quadratic integral equations) should be investigated in
different function spaces. This was remarked, for instance, in [86, Theorem 3.14] for
the case of Lp (I)-solutions, for the Hammerstein integral equation see also [79, 93]
for Lp -solutions or [13, 64, 108] for integrable solutions. A very interesting survey
about different classes of solutions (not only in C(I) or Lp (I), but also in Orlicz
spaces Lϕ (I) or even in ideal spaces) for a class of integral equations related to our
equation can be found in [8].
Next, let us recall that the equations involving the functional dependence have
still growing number of applications (cf. [73]). We try to cover the results of this
type. Let us mention, for example, the results from [14, 24].
31
We are interested in monotonic solutions of the above problem. The considered
problem can cover, for instance, as particular cases:
1. f1 (t, x) = g(t), f2 (t, x) = λ the functional Urysohn integral equation ([14, 15,
24]),
2. f1 (t, x) = g(t), f2 (t, x) = x, φ2 (t) = t the functional-integral equation ([91]),
3. f2 (t, x) = 0 the abstract functional equation ([15], for instance),
4. for continuous solutions with φ1 (t) = φ2 (t) = t and
u1 (t,s,x)
see [35, 61],
u(t, s, x) = Γ(α)·(t−τ
)1−α
5. f2 (t, x) = λ the functional integral equation (for continuous solutions see [3,
21, 54]),
6. f2 (t, x) = x the quadratic (functional) Urysohn integral equation ([27, 26], for
instance).
Our problem, as well as, the particular cases was investigated mainly in cases
when the solutions are elements of the space of continuous functions. Thus the
proofs are based on very special properties of this space (the compactness criterion,
in particular), cf. [35, 89].
On the other hand, by the practical interest it is worthwhile to consider discontinuous solutions. Here we are looking for integrable solutions. Thus the operators
F1 , F2 and U should take their values in the space L1 (I). Let us recall that we are
interested in finding monotonic solutions (a.e. monotonic in the case of integrable
solutions). In such a case discontinuous solutions are expected even in a simplest
case i.e. when
(
0
t is rational,
f1 (t, x) = h(t) =
t
t is irrational
An interesting example of discontinuous solutions for integral equations is taken
from [86, Example 3.5]:
χ[1/2,1] (t) · (2t − 1) · x(t) + χ[0,1/2] (t) · (1 − 2t) · (x(t) − 1)
Z
0
1
(1 − x(s)) ds = 0.
In contrast to the previous chapter, we extend the earlier result by considering functional integral equation in a more general form. Moreover, we prove the
existence of solutions in some subspaces of L1 (0, 1).
Let us add a few comments about functional dependence, i.e. functions ψ1 and
ψ2 . Our set of assumptions is based on the paper [24]. Functions of the form
ψi (t) = tα (α > 0) or ψi (t) = t − τ (t) with some set of assumptions for τ are most
32
important cases covered in our chapter. Let us note that functional equations with
state dependent delay are very useful in many mathematical models including the
population dynamics, the position control or the cell biology. A very interesting
survey about such a theory and their applications can be found in [73].
The last aspect of our results is to investigate the monotonicity property of
solutions. This is important property and there are many papers devoted to its
study. Let us note some recent ones [27, 28, 46, 61], for instance.
The results obtained in the current chapter create some extensions for several
known ones i.e. in addition to those mentioned previously also for the results from
earlier papers or books ([10, 15, 38, 51, 76, 97, 99, 122], for example).
3.2
Main result
Denote by H the operator associated with the right hand side of equation (3.1)
which takes the form
x = H(x),
where
H(x)(t) = f1 (t, x(φ1 (t))) + f2 (t, x(t)) ·
Z
1
u(t, s, x(φ2 (s)))ds.
(3.2)
0
This operator will be written as H(x) = Fφ1 ,f1 (x) + A(x),
A(x)(t) = Ff2 (x)(t) · U(x)(t) = Ff2 (x)(t) ·
Z
1
u(t, s, x(φ2(s))),
0
and the superposition operator F as in Definition 1.3.1. Thus equation (3.1) becomes
x(t) = Fφ1 ,f1 (x)(t) + A(x)(t).
3.2.1
The existence of L1-solution
We shall treat the equation (3.1) under the following assumptions listed below
(i) fi : I × R → R satisfies Carathéodory conditions and there are a positive
integrable on I functions ai and constants bi ≥ 0 such that
|fi (t, x)| ≤ ai (t) + bi |x| ,
i = 1, 2,
for all t ∈ [0, 1] and x ∈ R. Moreover, fi (t, x) ≥ 0 for x ≥ 0 and fi is
assumed to be nonincreasing with respect to both variable t and x separately
for i = 1, 2.
33
(ii) u : I × I × R → R satisfies Carathéodory conditions i.e. it is measurable in
(t, s) for any x ∈ R and continuous in x for almost all (t, s). The function
u is nonincreasing with respect to each variable, separately. Moreover, for
arbitrary fixed s ∈ I and x ∈ R the function t → u(t, s, x(s)) is integrable.
(iii) Assume that
|u(t, s, x)| ≤ k(t, s)(a3 (s) + b3 |x|),
for all t, s ≥ 0
and
x ∈ R,
where the function k is measurable in (t, s), a3 ∈ L1 (I) and a constant b3 > 0.
Assume that the linear integral operator K0 with the kernel k(t, s) maps L1 (I)
into L∞ (I). Moreover, assume that for arbitrary h > 0 (i = 1, 2)
Z
lim k
max
|u(t, s, x1 ) − u(t, s, x2 )| dskL∞ (I) = 0.
δ→0
D |xi |≤h,|x1−x2 |≤δ
(iv) φi : I → I are increasing, absolutely continuous functions (for i = 1, 2).
Moreover, there are constants Mi > 0 such that φ′i ≥ Mi a.e on (0, 1) (for
i = 1, 2).
Rb
Rb
(v) 0 k(t1 , s) ds ≥ 0 k(t2 , s) ds for t1 , t2 ∈ I with t1 < t2 and for any b ∈ [0, 1].
q
4b2 b3 ||K0 ||L∞ (I)
(ka1 k1 + kK0 kL∞ (I) ka2 k1 ka3 k1 ), where
(vi) let W >
M2
W =(
b1
b3
+
kK0 kL∞ (I) ka2 k1 + b2 kK0 kL∞ (I) ka3 k1 ) − 1
M1 M2
and let R denotes a positive solution of the quadratic equation
b2 b3 ||K0||L∞ (I) 2
·t
M2
b1
b3
+
kK0 kL∞ (I) ka2 k1 + b2 kK0 kL∞ (I) ka3 k1 )] · t
− [1 − (
M1 M2
+ (ka1 k1 + kK0 kL∞ (I) ka2 k1 ka3 k1 ) = 0.
Then we can prove the following theorem.
Theorem 3.2.1. Let the assumptions (i) - (vi) be satisfied. Put
L=[
b1
b3
+ b2 kK0 kL∞ (I) [ka3 k1 +
R]].
M1
M2
If L < 1, then the equation (3.1) has at least one integrable solution a.e. nonincreasing on I.
34
Proof. First of all observe that by the assumption (i) and Theorem (1.3.1) implies
that Fφ1 ,f1 and Ff2 are continuous mappings from L1 (I) into itself. By assumption
(iii) and Theorem 1.3.6 we can deduce that U maps L1 (I) into L∞ (I). From the
Hölder inequality the operator A maps L1 (I) into itself continuously. Finally, for a
given x ∈ L1 (I) the function H(x) belongs to L1 (I) and is continuous. Thus
kH(x)k1 ≤ kFφ1 ,f1 xk1 + kAxk1
Z 1
≤
[a1 (t) + b1 |x(φ1 (t))|]dt
0
Z 1
Z 1
+
[a2 (t) + b2 |x(t)|]
|u(t, s, x(φ2(s))| ds dt
0
0
Z 1
≤
[a1 (t) + b1 |x(φ1 (t))|]dt
0
Z 1
Z 1
+
[a2 (t) + b2 |x(t)|]
k(t, s)[a3 (s) + b3 |x(φ2 (s))|] ds dt
0
0
Z 1
b1
|x(φ1 (t))|φ′1 (t)dt
≤ ka1 k1 +
M1 0
Z 1Z 1
+
k(t, s)a2 (t)[a3 (s) + b3 |x(φ2 (s))|]dsdt
0
0
Z 1Z 1
+ b2
k(t, s)|x(t)|[a3 (s) + b3 |x(φ2 (s))|]dsdt
0
0
Z φ1 (1)
b1
≤ ka1 k1 +
|x(u)|du
M1 φ1 (0)
Z 1
Z 1
+
[a3 (s) + b3 |x(φ2 (s))|]
k(t, s)a2 (t)dtds
0
0
Z 1
Z 1
+ b2
[a3 (s) + b3 |x(φ2 (s))|]
k(t, s)|x(t)| dtds
0
0
Z 1
b1
|x(t)|dt
≤ ka1 k1 +
M1 0
Z 1
+kK0 kL∞ (I) ka2 k1
[a3 (s) + b3 |x(φ2 (s))|]ds
0
Z 1
+b2 kK0 kL∞ (I) kxk1
[a3 (s) + b3 |x(φ2 (s))|]ds
0
b1
kxk1
≤ ka1 k1 +
M1
1
b3
|x(φ2 (s))|φ′2 (s)]ds
M
2
0
Z 1
b3
+b2 kK0 kL∞ (I) kxk1
[a3 (s) +
|x(φ2 (s))|φ′2 (s)]ds
M
2
0
+kK0 kL∞ (I) ka2 k1
Z
[a3 (s) +
35
b3
b1
kxk1 + kK0 kL∞ (I) ka2 k1 [ka3 k1 +
kxk1 ]
M1
M2
b3
+b2 kK0 kL∞ (I) kxk1 [ka3 k1 +
kxk1 ]
M2
b3
b1
+
kK0 kL∞ (I) ka2 k1
= ka1 k1 + kK0 kL∞ (I) ka2 k1 ka3 k1 + [
M1 M2
b2 b3 ||K0 ||L∞ (I)
+b2 kK0 kL∞ (I) ka3 k1 ] · kxk1 +
· (kxk1 )2 .
M2
≤ ka1 k1 +
By our assumption (vi) , it follows that there exists a positive constant R being the
positive solution of the equation from the assumption (vi) and such that H maps
the ball BR into itself.
Further, let QR stand for the subset of BR consisting of all functions which are
a.e. nonincreasing on I. Similarly as claimed in [17] we are able to show that this set
is nonempty, bounded (by R), convex and closed in L1 (I). Only the last property
needs some comments. Let (yn ) be a sequence of elements in QR convergent in
L1 (I) to y. Then the sequence is convergent in measure and as a consequence of the
Vitali convergence theorem and of the characterization of convergence in measure
(the Riesz theorem) we obtain the existence of a subsequence (ynk ) of (yn ) which
converges to y almost uniformly on I. Moreover, y is nonincreasing a.e. on I which
means that y ∈ QR and so the set QR is closed. Now, in view of Theorem 1.4.1 the
set QR is compact in measure. To see this it suffices to put Ωc = [0, c] \ P for any
c ≥ 0, where P denotes a suitable set with meas(P ) = 0.
Now, we will show that H preserve the monotonicity of functions. Take x ∈ QR ,
then x(t) and x(φi (t)) are a.e. nonincreasing on I and consequently each fi is also
of the same type by virtue of the assumption (i) and Theorem 1.4.2. Further, Ux(t)
is a.e. nonincreasing on I due to assumption (ii). Moreover, Fφ1 ,f1 , A(x)(t) are
also of the same type. Thus we can deduce that H(x) = Fφ1 ,f1 + A(x) is also a.e.
nonincreasing on I. This fact, together with the assertion H : BR → BR gives that
H is also a self-mapping of the set QR . From the above considerations it follows
that H maps continuously QR into QR .
From now we will assume that X is a nonempty subset of QR and the constant
ε > 0 is arbitrary, but fixed. Then for an arbitrary x ∈ X and for a set D ⊂ I,
meas(D) ≤ ε we obtain
36
Z
D
|(H(x))(t)|dt ≤
Z
[a1 (t) + b1 |x(φ1 (t))|]dt
Z
Z 1
+ [a2 (t) + b2 |x(t)|]
|u(t, s, x(φ2 (s))| ds dt
D
0
Z
b1
|x(φ1 (t))|φ′1 (t)dt
≤ ka1 χD k1 +
M1 D
Z Z 1
+
k(t, s)a2 (t)[a3 (s) + b3 |x(φ2 (s))|]dsdt
D 0
Z Z 1
+ b2
k(t, s)|x(t)|[a3 (s) + b3 |x(φ2 (s))|]dsdt
D
D
≤ ka1 χD k1 +
0
b1
kxχD k1
M1
b3
R]
M2
b3
+ b2 kK0 kL∞ (I) kxχD k1 [ka3 k1 +
R].
M2
+ kK0 kL∞ (I) ka2 χD k1 [ka3 k1 +
Hence, taking into account the equalities
Z
lim{sup[ ai (t) dt : D ⊂ I, meas(D) ≤ ε]} = 0, i = 1, 2,
ε→0
D
and by the definition of c(X) (cf. Section 1.2) we get
b3
b1
+ b2 kK0 kL∞ (I) [ka3 k1 +
R] · c(X).
c(H(X)) ≤
M1
M2
(3.3)
Recall that L = Mb11 + b2 kK0 kL∞ (I) (ka3 k1 + Mb32 R) < 1 and then the inequality
obtained above together with the properties of the operator H and since the set
QR is compact in measure we are able to apply Theorem 1.6.5 which completes the
proof.
Remark. Let us recall that in the proof we utilize the following fact: U maps
L1 (I) into L∞ (I) and F2 maps L1 (I) into itself. This allows us to use the Hölder
inequality. In this situation, we prove the existence of a.e. monotonic solutions
which are integrable. Sometimes we need more information about the solution,
namely if a solution is in some subspace of L1 (I) (the space Lp , for instance). In
such a case we are able to use also the same type of inequality. Namely we need only
to modify the growth conditions and consequently the spaces in which our operators
act. As claimed in the introductory part of our chapter we can repeat our proof with
appropriate changes of domains for considered operators: F2 maps Lp (I) into Lq (I)
and U maps Lp (I) into Lr (I), where 1r + 1q = p1 . Whence we obtain an existence
result for Lp -solutions.
37
3.2.2
The existence of Lp -solution p > 1
It should be noted that in some papers, their authors consider the existence of solutions in Lp spaces simultaneously for p ≥ 1. As claimed above it cannot be done
for quadratic equations. Here is a version for p > 1. An interesting (and motivating) remark about the solutions in Lp spaces for integral equations (by using
similar method of the proof) can be found in [64, page 93]. However, by considering the measure of noncompactness c(X) = lim supmeas(D)→0 {supx∈X kxχD kLp (I) }
introduced by Appll and De Pascale [9] (cf. also Erzakova [66]) (restricted to the
family of sets compact in measure) instead of usually considered ones based on Kolomogorov or Riesz criteria of compactness (cf. [23]) we are able to examine by the
same manner the case of Lp (I) spaces.
Assume that p > 1 and p11 + p12 = 1p . Denote by q the value min(p1 , p2 ) and by r
the value max(p1 , p2 ). This implies, in particular, that q ≤ 2p. We shall treat the
equation (3.1) under the following set of assumptions presented below.
(i)’ Assume that functions fi : R+ × R → R satisfy Carathéodory conditions and
there are positive constants bi (i = 1, 2) and positive functions a1 ∈ Lp (I),
a2 ∈ Lq (I) such that
|f1 (t, x)| ≤ a1 (t) + b1 |x|,
p
|f2 (t, x)| ≤ a2 (t) + b2 |x| q ,
for all t ∈ I and x ∈ R. Moreover, fi (i = 1, 2) are assumed to be nonincreasing
with respect to both variable t and x separately.
(ii)’ u : R+ × R+ × R → R satisfies Carathéodory conditions. The function u
is nonincreasing with respect to each variable, separately. Suppose that for
arbitrary non-negative z(t) ∈ Lq (I)
Z
lim sup k
u(t, s, x(s)) dskLr (I) = 0
D→0 |x|≤z
D
and that
p
|u(t, s, x)| ≤ k(t, s)(a3 (s) + b3 |x| q ),
for all t, s ≥ 0 and
x ∈ R,
where the function k is measurable in (t, s), a3 ∈ Lq (I) and a constant b3 > 0.
Assume that the linear integral operator K0 with the kernel k(t, s) maps Lq (I)
into Lr (I).
(iii)’ φi : I → I are increasing, absolutely continuous functions (for i = 1, 2).
Moreover, there are constants Mi > 0 such that φ′i ≥ Mi a.e on (0, 1) (for
i = 1, 2).
38
(iv)’
Rb
0
k(t1 , s) ds ≥
Rb
0
k(t2 , s) ds for t1 , t2 ∈ I with t1 < t2 and for any b ∈ [0, 1].
(v)’ Assume, that the following equation
b2 b3 kK0 k
1/q
M2
2p
+ [b2 ka3 kLq )I) +
·tq
b3 ka2 kLq (I)
1/q
M2
p
]kK0 kt q + (
b1
1/p
M1
+ (ka1 kLp (I) + kK0 kka2 kLq (I) ka3 kLq (I) ) = 0
− 1) · t
has a solution in (0, 1] and denote by s its positive solution.
By L′ we will denote a number
b1
1/p
M1
+ b2 s
p
−1
q
kK0 k ka3 kLq (I) +
b3
1/p
M2
s
p
q
!
.
Theorem 3.2.2. Let the assumptions (i)’ - (v)’ be satisfied. If L′ < 1, then the
equation (3.1) has at least one Lp (I)-solution a.e. nonincreasing on I.
Proof. By assumption (i) and Theorem 1.3.1 implies that Fφ1 ,f1 maps Lp (I) into it
self continuously and Ff2 maps Lp (I) into Lq (I) and is continuous. By assumption
(ii) and Theorem 1.3.5 we can deduce that U is a continuous map from Lp (I) into
Lr (I). From the Hölder inequality the operator A maps Lp (I) into itself continuously. Finally, for a given x ∈ Lp (I) the function H(x) belongs to Lp (I) and is
continuous. Thus for 1q + q1′ = 1
kH(x)kp ≤ kFφ1 ,f1 x(t)kp + kAx(t)kp
≤ ka1 + b1 |x(φ1 )kp + kFf2 xkq kUxkr
Z 1
p1
≤ ka1 kp + b1
|x(φ1 (t))|p dt
0
Z 1
p
p
k(t, s)[a3 (s) + b3 |x(φ2 (s))| q ]kr
+ka2 + b2 |x| q kq · k
0
b1
≤ ka1 kp +
1
M1p
Z
0
1p
1
|x(φ1 (t))|p φ′1 (t)dt
p
p
+[ka2 kq + b2 kxkpq ] · kkk(t, ·)kq′ [ka3 kq + b3 k|x(φ2 )| q kq ]kr ,
where
p
q
kx kq =
Z 1
0
|x(s)|
p
q
q
ds
1q
p
= kxkpq and kK0 kr,q′ ≡ kt → kk(t, ·)kq′ kr .
39
kH(x)kp ≤ ka1 kp +
b1
Z
1
p
M1
φ1 (1)
|x(v)|p dv
φ1 (0)
! p1
p
q
+ [ka2 kq + b2 kxkp ] · kK0 kr,q′ [ka3 kq + b3
≤ ka1 kp +
b1
1
p
M1
Z
1
0
|x(v)|p dv
p1
+ [ka2 kq + b2 kxkp ] · kK0 kr,q′ [ka3 kq +
≤ ka1 kp +
b1
1
p
M1
+ [ka2 kq + b2 kxkp ] · kK0 kr,q′ [ka3 kq +
≤ ka1 kp +
1
p
M1
1
p
M1
1
0
1q
p q
′
|x(φ2 (s))| q φ2 (s)ds ]
Z
1
q
φ2 (1)
φ2 (0)
M2
|x(v)|
p
q
q
dv
! 1q
]
kxkp
+ [ka2 kq + b2 kxkp ] · kK0 kr,q′ [ka3 kq +
b1
1
M2q
b3
b3
p
q
≤ ka1 kp +
Z
q1
p q
|x(φ2 (s))| q ds ]
kxkp
p
q
b1
0
b3
p
q
1
Z
1
q
M2
p
q
Z 1
0
|x(v)|
p
q
q
kxkp + [ka2 kq + b2 kxkp ] · kK0 kr,q′ [ka3 kq +
≤ ka1 kp + kK0 kr,q′ ka2 kq ka3 kq +
+ kK0 kr,q′ [b2 ka3 kq +
b3 ka2 kq
1
M2q
b3 ka2 kq
1
q
1
M1p
p
b1
1
p
b3
1
q
M2
]
p
kxkpq ]
kxkp
] · kxkpq +
≤ ka1 kp + kK0 kr,q′ ka2 kq ka3 kq +
+ kK0 kr,q′ [b2 ka3 kq +
b1
dv
q1
b2 b3 kK0 kr,q′
1
M2q
2p
kxkpq
R′
M1
p
]R′ q +
M2
b2 b3 kK0 kr,q′
1
q
M2
R′
2p
q
≤ R′ .
From assumption (v)’ it follows there exits a positive solution R′ of the equation,
which implies that H maps the ball BR′ into it self.
As before in Theorem 3.2.1, we can construct a subset QR′ of BR′ is nonempty,
bounded, convex and closed in Lp (I) consisting of all functions which are a.e. nonincreasing on I. Such that H maps QR′ into it self.
From now we will assume that X is a nonempty subset of Q′R and the constant
ε > 0 is arbitrary, but fixed. Then for an arbitrary x ∈ X and for a set D ⊂ I,
40
meas(D) ≤ ε we obtain
kH(x)χD kp ≤ k[a1 + b1 |x(φ1 )]χD kp + kFf2 xχD kq kUxχD kr
b1
≤ ka1 χD kp + 1 kxχD kp
M1p
Z 1
p
p
q
+ [ka2 χD kq + b2 kxχD kq ] · k
k(t, s)[a3 (s) + b3 |x(φ2 (s))| q ]kr
0
≤ ka1 χD kp +
b1
1
M1p
kxχD kp
p
+
≤
+
≤
+
p
[ka2 χD kq + b2 kxχD kqq ]k · kk(t, ·)kq′ [ka3 kq + b3 k|x(φ2 )| q kq ]kr
b1
ka1 χD kp + 1 kxχD kp
M1p
p
p
b3
[ka2 χD kq + b2 kxχD kpq ] · kK0 kr,q′ [ka3 kq + 1 kxkpq ]
M2q
b1
ka1 χD kp + 1 kxχD kp
M1p
p
p
b3
−1
[ka2 χD kq + b2 R′ q kxχD kp ] · kK0 kr,q′ [ka3 kq + 1 R′ q ].
M2q
Since a1 ∈ Lp and a2 ∈ Lq ,
lim{sup[ka1 χD kp : D ⊂ I, meas(D) ≤ ε]} = 0
ε→0
and
lim{sup[ka2 χD kq : D ⊂ I, meas(D) ≤ ε]} = 0.
ε→0
Then by the definition of c(x) in Lp space, we have
c(H(X)) =
Recall that L′ = [
b1
1
M1p
b1
1
p
M1
+b2 s
p
+ b2 s q −1 kK0 kr,q′ (ka3 kq +
p
−1
q
kK0 kr,q′ (ka3 kq +
b3
1
M2q
p
q
b3
1
q
M2
p
s q ) · c(X).
s )] < 1 and then the inequality
obtained above together with the properties of the operator H and since the set QR′
is compact in measure we are able to apply Theorem 1.6.5 which completes the
proof.
Let us note, that in the assumption (v)’ we consider the equation of the type
p
2p
A+Bt+Ct q +Dt q = t. The case p = q leads to the quadratic equation (considered
in our first theorem). Although the case p < q seems to be more complicated, it
41
should be noted that since pq < 1 and 2p
< 2 this equation has a solution in (0, 1]. In
q
some papers the assumption of this type is described by using auxiliary functions.
In such a formulation the problem of existence of functions is unclear. Let us note,
that for arbitrary pair of spaces Lp (I) and Lq (I) we are able to solve our problem.
p
2p
≥ 1, then for t ∈ I we have A + Bt + Ct q + Dt q ≤ A + Bt + C + Dt
Indeed, if 2p
q
A+C
< 1. In the case 2p
< 1,
and our inequality has a solution in (0, 1] whenever 1−B−D
q
p
2p
we have the following estimation: A + Bt + Ct q + Dt q ≤ A + Bt + C + D and then
A+C+D
< 1 form a sufficient condition for the existence of solutions of our inequality
1−B
in (0, 1]. Thus the set of functions satisfying our assumptions is nonempty (cf. also
some interesting Examples in [16]). Let us remind that the first case is considered
in the chapter.
We would like to pay attention, that the condition (ii)’ implies that the kernels
k(t, s) are of Hille-Tamarkin classes i.e. kkk(t, ·)k′q kr and kkk(·, s)kq kr′ are finite
being at the same time the upper bounds for kK0 k, where q ′ and r ′ are conjugated
with q and r, respectively.
Moreover, it is worthwhile to note that by the same manner we can extend our
main result for other subspaces of L1 (I) for which we are able to check the required
properties of considered operators (some Orlicz spaces, for instance) cf. [47].
Remark 3.2.1. Till now, we are interested in finding monotonic solutions of our
problem. Assume that we have the decomposition of the interval I into the disjoint
subsets T1 and T2 with T1 ∪ T2 = I, such that fi (·, x) are a.e. nondecreasing on
T1 and a.e. nonincreasing on T2 . By an appropriate change of the monotonicity
assumptions we are able to prove the existence of solutions belonging to the class of
functions described above (similarly like in [15]). In such a case we need to consider
the operators preserving this property, too.
3.3
Examples
We need to show an example for which our main result is useful and allow to extend
the existing theorems. Let us recall that we are looking for monotonic solutions for
the considered problems in the interval I.
But first, let us recall that the quadratic equations have numerous applications
in the theories of radiative transfer, neutron transport and in the kinetic theory of
gases [12, 26, 40, 43]. In order to apply earlier results of the considered type, we
have to impose an additional condition that the so-called ”characteristic” function ψ
is continuous (cf. [40, Theorem 3.2]) or even Hölder continuous ([12]). In the theory
of radiative transfer this function is immediately related to the angular pattern
42
for single scattering and then our results allow to consider some peculiar states
of the atmosphere. In astrophysical applications of the Chandrasekhar equation
R1
R1 t
ψ(s)x(s) ds the only restriction that 0 ψ(s) ds ≤ 1/2 is
x(t) = 1 + x(t) 0 t+s
treated as necessary (cf. [40, Chapter VIII; Corollary 2 p. 187]. An interesting
discussion about this condition and the applicability of such equations can be found
in [40]. Recall that to ensure the existence of solutions normally one assumes that
ψ(t) is an even polynomial (as in the book of Chandrasekhar [43, Chapter 5]) or
continuous ([40]). The using of different solution spaces in the current chapter allow
us to remove this restriction and then we give a partial answer to the problem from
[40]. The continuity assumption for ψ implies the continuity of solutions for the
considered equation (cf. [40]) and then seems to be too restrictive even from the
theoretical point of view.
Let us consider now the following integral equation
2
− ln(1 + x2 ( 3t + t2 ))
(3.4)
x(t) = a(t) +
3+t
Z 1
1 + h(x)
1
x(s)
λ
+ arctan √
[√
] ds,
+
2
2
t+2
s + 1 1 + x2 (s)
0 t +s
where
a(t) =
(
0
t is rational,
, h(x) =
1 − t t is irrational
(
0
sin x
1+ex
for x ≤ 0
for x > 0.
It can be easily seen that equation (3.4) is a particular case of the equation (3.1),
where
2
− ln(1 + x2 ( 3t + t2 ))
1 + h(x)
f1 (t, x) = a(t) +
, f2 (t, x) = arctan √
3+t
t+2
and
u(t, s, x) =
t2
x(s)
1
λ
[√
].
+
2
+s
s + 1 1 + x2 (s)
√
< 1+h(x)
,
In view of the inequalities ln(1 + x ) ≤ x (x > 0) and arctan
t+2
the functions f1 , f2 and u are nonincreasing in each variable separately. Moreover,
1
+ 31 h(x) and
|f1 (t, x)| ≤ a(t) + 14 |x|, |f2 (t, x)| ≤ √t+2
2
|u(t, s, x)| ≤
t2
1+h(x)
√
t+2
1
λ
1
[√
+ |x|],
2
+s
s+1 2
1
with a1 (t) = a(t), a2 (t) = √t+2
, a3 (s) =
1
1
constants b1 = 4 , b2 = 3 and b3 = 21 .
√1
s+1
43
and k(t, s) =
λ
.
t2 +s2
Here we have the
R1 λ
1
Since 0 t2 +s
2 ds = λ arctan t , |k(t, s)| ≤ λ. Thus the expected property for the
operator K0 holds true. Moreover, given arbitrary h > 0 and |x2 − x1 | ≤ δ we have
x1 (1 + x22 ) − x2 (1 + x21 )
1
|
|
t2 + s2
(1 + x21 )(1 + x22 )
(x1 − x2 ) + x1 x2 (x2 − x1 )
1
|
|
= 2
2
t +s
(1 + x21 )(1 + x22 )
1
δ(1 + h2 )
≤ 2
.
t + s2 (1 + x21 )(1 + x22 )
|u(t, s, x1) − u(t, s, x2 )| =
Put φ1 (t) =
t
3
+
t2
2
and φ2 (t) = t, then
φ′1 (t) =
1
1
+ t > = M1
3
3
and φ′2 (t) = 1 >
1
= M2
2
.
Thus our assumptions (i)-(iv) are satisfied. Since 34 + 13 λ π2 (1 + R) < 1 for small
λ > 0, assumption (v) holds true for sufficiently small λ.
Taking into account all the above observations we are able to deduce from Theorem 3.2.1 that for sufficiently small λ the equation (3.4) has at least one integrable
solution x which is a.e. nonincreasing on I.
44
Chapter 4
Functional quadratic integral
equations with perturbations on a
half line
4.1
Introduction
We study the solvability of the following functional integral equation. Let t ∈ R+
Z t
x(t) = g(t, x(ϕ3 (t))) + f1 t, f2 (t, x(ϕ2 (t))) ·
u(t, s, x(ϕ1 (s))) ds . (4.1)
0
This equation has been studied for non quadratic integral equation in [19] with
g = 0, f2 = 1 using Schauder fixed point theorem and in [115] with a perturbation
term. In [24], it was checked the existence of monotonic solutions, where g(t, x(t)) =
h(t), f2 = 1, improved also by Emmanuele (cf. [63]). The authors used the general
Krasnoselskii fixed point theorem to obtain the existence result (cf. [57, 87, 115]).
In [124] the author studied a special case of our equation in a general Banach space
X by using the classical Krasnoselskii fixed point theorem.
In the case of the sum of two sufficiently regular operators the contraction condition is easily verified. Nevertheless, a construction for the set M make the above
theorem more restrictive. The presence of the perturbation term g(t, x(t)) in the
integral equation make the Schauder fixed point theorem unavailable. Given operators A and B, it may be possible to find the bounded domains MA and MB in such
a way that A : MA → MA and B : MB → MB , but it is often impossible to arrange
matters so that MA = MB = M and Ax + By ∈ MA for x, y ∈ M. Actually, the
Krasnoselskii fixed point theorem allow us to avoid these problems for obtaining the
result of the solution.
The results presented in this chapter are motivated by extending the recent results to the functional quadratic integral equation with a perturbation term by using
45
classical Krasnoselskii fixed point theorem and the measure of weak noncompactness.
Let us stress, that we will dispense the monotonicity assumptions that mentioned
in the previous chapters, so we need to use different method of the proof.
4.2
Main result
Equation (4.1) takes the following form
x = Ax + Bx,
where
Ax(t) = Fϕ3 ,g x(t) − g(t, 0),
(Bx)(t) =
=
f1 t, f2 (t, x(ϕ2 (t))) ·
Ff1 (Kx)(t) + g(t, 0),
Z
t
u(t, s, x(ϕ1 (s))) ds + g(t, 0)
0
Kx(t) = Fϕ2 ,f2 x · Ux(ϕ1 )(t) and Ux(t) =
Z
t
u(t, s, x(s)) ds.
0
We shall treat the equation (4.1) under the following assumptions listed below,
where L1 = L1 (R+ ).
(i) g, fi : R+ × R → R satisfies Carathéodory conditions and there are positive
functions ai ∈ L1 and constants bi ≥ 0 for i = 1, 2, 3. such that
|fi (t, x)| ≤ ai (t) + bi |x|, i = 1, 2,
and |g(t, 0)| ≤ a3 (t),
for all t ∈ R+ and x ∈ R. Moreover, the function g is assumed to satisfy
the Lipschitz condition with constant b3 for almost all t:
|g(t, x) − g(t, y)| ≤ b3 |x − y|.
(ii) u : R+ × R+ × R → R satisfies Carathéodory conditions i.e. it is measurable
in (t, s) for any x ∈ R and continuous in x for almost all (t, s). Moreover,
for arbitrary fixed s ∈ R+ and x ∈ R the function t → u(t, s, x(s)) is
integrable.
(iii) There exists a function k(t, s) = k : R+ ×R+ → R which satisfies Carathéodory
conditions such that:
|u(t, s, x)| ≤ k(t, s)
46
for all t, s ≥ 0 and x ∈ R, such that the linear integral operator K0 with kernel
k(t, s) maps L1 into L∞ . Moreover, assume that for arbitrary h > 0 (i = 1, 2)
Z
lim k
max
|u(t, s, x1 ) − u(t, s, x2 )| ds kL∞ = 0.
δ→0
D |xi |≤h ,|x1 −x2 |≤δ
(iv) ϕi : R+ → R+ is increasing absolutely continuous function and there are
positive constants Mi such that ϕ′i ≥ Mi a.e. on R+ for i = 1, 2, 3.
2 b1
kK
k
< 1,
(v) q = Mb33 + bM
0 ∞
2
(vi) p =
b3
M3
< 1.
Then we can prove the following theorem.
Theorem 4.2.1. Let the assumptions (i) - (vi) are satisfied, then equation (4.1)
has at least one integrable solution on R+ .
Proof. The proof will be given in six steps.
• Step 1. The operator A : L1 → L1 is a contraction mapping.
• Step 2. We will construct the ball Br such that A(Br ) + B(Br ) ⊆ Br , where
r will be determined later.
• Step 3. We will proof that µ(A(Q) + B(Q)) ≤ qµ(Q) for all bounded subset
Q of Br , where µ as in Definition 1.5.2.
• Step 4. We will construct a nonempty closed convex weakly compact set M in
on which we will apply fixed point theorem to prove the existence of solutions.
• Step 5. B(M) is relatively strongly compact in L1 .
• Step 6. We will check out the conditions needed in Krasonselskii’s fixed point
theorem are fulfilled.
Step 1. From assumption (i), we have
||g(t, x)| − |g(t, 0)||
|g(t, x)| − a3 (t)
≤
≤
|g(t, x) − g(t, 0)| ≤ b3 |x|
b3 |x| ⇒
|g(t, x)| ≤ a3 (t) + b3 |x|.
The inequality obtained above with Theorem 1.3.1 permits us to deduce that the
operator A maps L1 into itself.
47
Now,
Z
∞
0
|(Ax)(t) − (Ay)(t)|dt =
Z
∞
0
≤ b3
Z
0
|g(t, x(ϕ3(t))) − g(t, y(ϕ3(t)))|dt
∞
|x(ϕ3 (t)) − y(ϕ3 (t))|dt
Z ∞
b3
≤
|x(ϕ3 (t)) − y(ϕ3 (t))|ϕ′3 (t)dt
M3 0
Z ϕ3 (∞)
b3
|x(v) − y(v)|dv
=
M3 ϕ3 (0)
Z ∞
b3
≤
|x(v) − y(v)|dv,
M3 0
which implies that
kAx − AykL1 ≤
b3
kx − ykL1 .
M3
(4.2)
Assumption (vi) permits us to deduce that the operator A is a contraction mapping.
Step 2. Let x and y be arbitrary functions in Br ⊂ L1 (R+ ). In view of our
assumptions we get a priori estimation
Z ∞
≤
|g(t, x(ϕ3(t)))|dt
kAx + BykL1
0
Z ∞
Z t
+
|f1 t, f2 (t, y(ϕ2(t))) ·
u(t, s, y(ϕ1(s))) ds |dt
0
0
Z ∞
≤
[a3 (t) + b3 |x(ϕ3 (t))| ]dt
0
Z ∞
Z t
+
[a1 (t) + b1 · |f2 (t, y(ϕ2 (t)))|.
|u(t, s, y(ϕ1(s)))| dsdt
0
0
Z ∞
≤
[a3 (t) + b3 |x(ϕ3 (t))| ]dt
0
Z ∞
Z t
+
[a1 (t) + b1 · (a2 (t) + b2 |y(ϕ2(t))|).
k(t, s) dsdt
0
0
Z ∞
b3
|x(ϕ3 (t))|ϕ′3 (t)dt
≤
ka1 kL1 + ka3 kL1 +
M3 0
Z ∞
(a2 (t) + b2 |y(ϕ2(t))|)dt
+ b1 .kK0 kL∞
0
Z ϕ3 (∞)
b3
≤
ka1 kL1 + ka3 kL1 +
|x(v)|dv
M3 ϕ3 (0)
Z ∞
b2
|y(ϕ2(t))|ϕ′2 (t)dt ]
+ b1 .kK0 kL∞ [ka2 kL1 +
M2 0
48
Z ∞
b3
|x(t)|dt
≤
ka1 kL1 + ka3 kL1 +
M3 0
Z ∞
b2
+ b1 .kK0 kL∞ [ka2 kL1 +
|y(ϕ2 (t))|ϕ′2 (t)dt ]
M2 0
b3
≤
ka1 kL1 + ka3 kL1 +
kx(t)kL1
M3
b2
+ b1 ka2 kL1 .kK0 kL∞ +
kK0 kL∞ · kykL1
M2
b3
·r
≤
ka1 kL1 + ka3 kL1 +
M3
b1 b2
kK0 kL∞ · r ≤ r.
+ b1 ka2 kL1 .kK0 kL∞ +
M2
From the above estimate, we have that A(Br ) + B(Br ) ⊆ Br provided
r=
ka1 kL1 + ka3 kL1 + b1 ka2 kL1 .kK0 kL∞
> 0.
1 b2
)
1 − ( Mb33 + bM
kK
k
0
L
∞
2
Step 3. Take an arbitrary number ε > 0 and a set D ⊂ R+ such that meas(D) ≤
ε. For any x, y ∈ Q, we have
Z
Z
Z
|Ax(t) + By(t)|dt
≤
|Ax(t)|dt +
|By(t)|dt
D
D
D
Z
Z
=
|Fg,ϕ3 x(t)|dt +
|Ff1 Ky(t)|dt
D
D
Z
[a3 (t) + b3 |x(ϕ3 (t))|]dt
≤
D
Z
Z t
+
[a1 (t) + b1 |f2 (t, y(ϕ2(t)))|
|u(t, s, y(ϕ1(s)))| ds]dt
D
0
Z
≤
[a3 (t) + b3 |x(ϕ3 (t)))|]dt
D
Z
Z t
[a1 (t) + b1 · (a2 (t) + b2 |y(ϕ2(t)))|).
+
k(t, s) ds]dt
D
0
Z
Z
Z
b3
|x(ϕ3 (t))|ϕ′3 (t)dt
≤
a1 (t)dt +
a3 (t)dt +
M
3
D
D
Z D
+ b1 kK0 kL∞ [a2 (t) + b2 |y(ϕ2(t)))|]dt
D
Z
Z
Z
b3
≤
a1 (t)dt +
a3 (t)dt +
|x(v)|dv
M3 ϕ3 (D)
D
D
Z
b2
+ b1 kK0 kL∞ [a2 (t) +
|y(ϕ2(t))|ϕ′2 (t)]dt
M2
D
49
Z
b3
|x(v)|dv
≤
a1 (t)dt +
a3 (t)dt +
M3 ϕ3 (D)
D
D
Z
Z
b2
+ b1 kK0 kL∞ [ a2 (t)dt +
|y(v)|)dv]
M2 ϕ2 (D)
D
Z
Z
Z
b3
≤
a1 (t)dt +
a3 (t)dt +
|x(v)|dv
M3 ϕ3 (D)
D
D
Z
Z
b2
+ b1 kK0 kL∞ [ a2 (t)dt +
|y(v)|)dv],
M2 ϕ2 (D)
D
Z
Z
where the symbol kK0 kL∞ (D) denotes the norm of the operator K0 acting from
the space L1 (D) into L∞ (D).
Now, using the fact that
Z
lim sup[ ai (t)dt : D ⊂ R+ , m(D) ≤ ε] = 0, for i = 1, 2, 3.
ε→∞
D
From Definition 1.5 it follows that
c(A(Q) + B(Q)) ≤ [q = (
b3
b1 b2
+
kK0 kL∞ )]c(Q).
M3
M2
(4.3)
For T > 0 and any x, y ∈ Q,we have
Z ∞
Z ∞
Z ∞
Z ∞
b3
|x(v)|dv
|Ax(t) + By(t)|dt ≤
a1 (t)dt +
a3 (t)dt +
M3 ϕ3 (T )
T
T
T
Z ∞
Z ∞
b2
|y(v)|dv],
+ b1 kK0 kL∞ [
a2 (t)dt +
M2 ϕ2 (T )
T
where ϕi (T ) → ∞ as T → ∞ for i = 1, 2. Then as T → ∞ and by the Definition
1.6 we get
b1 b2
b3
+
kK0 kL∞ )]d(Q).
M3
M2
By combining equation 4.3 and 4.4 and Definition 1.7, we have
d(A(Q) + B(Q)) ≤ [q = (
µ(A(Q) + B(Q)) ≤ [q = (
(4.4)
b3
b1 b2
+
kK0 kL∞ )]µ(Q).
M3
M2
Step 4. Let Br1 = Conv(A(Br ) + B(Br )), where Br is defined in step 1, Br2 =
Conv(A(Br1 ) + B(Br1 )) and so on. We then get a decreasing sequence {Brn }, that is
Brn+1 ⊂ Brn for n = 1, 2, · · · Obviously all sets belonging to this sequence are closed
and convex, so weakly closed. By the fact proved in step 2.
That µ(A(Q) + B(Q)) ≤ qµ(Q) for all bounded subset Q of Br , we have
µ(Brn ) ≤ q n µ(Br ),
which yields that limn→∞ µ(Brn ) = 0.
n
Denote M = ∩∞
n=1 Br , and then µ(M) = 0. By the definition of the measure of weak
50
noncompactness we know that M is nonempty. From the definition of the operator
A, we can deduce that B(M) ⊂ M.
M is just nonempty closed convex weakly compact set which we need in the following steps.
Step 5. Let {xn } ⊂ M be arbitrary sequence. Since µ(M) = 0, ∃ T, ∀n, the
following inequality is satisfied:
Z
T
∞
ε
|xn (t)|dt ≤ .
4
(4.5)
Considering the function fi (t, x) on [0, T ], (i = 1, 2), u(t, s, x) on [0, T ]×[0, T ]×R,
and k(t, s) on [0, T ] × [0, T ], in view of Theorem 1.3.3 we can find a closed subset
Dε of the interval [0, T ], such that meas(Dεc ) ≤ ε, and such that fi |Dε ×R (i = 1, 2),
u |Dε ×Dε ×R , and k |Dε ×[0,T ] are continuous. Especially k |Dε ×[0,T ] is uniformly continuous.
Let us take arbitrary t1 , t2 ∈ Dε and assume t1 < t2 without loss of generality.
For an arbitrary fixed n ∈ N and denoting Hn (t) = (Fϕ2 ,f2 xn ).(Uxn ))(t) we obtain:
|Hn (t2 ) − Hn (t1 )| =
−
≤
+
−
≤
+
−
+
−
t2
Z
|f2 (t2 , xn (ϕ2 (t2 )))
u(t2 , s, xn (ϕ1 (s)))ds
0
Z t1
f2 (t1 , xn (ϕ2 (t1 )))
u(t1 , s, xn (ϕ1 (s)))ds|
0
Z t2
|f2 (t2 , xn (ϕ2 (t2 ))) − f2 (t1 , xn (ϕ2 (t1 )))|
|u(t2, s, xn (ϕ1 (s)))|ds
0
Z t2
|f2 (t1 , xn (ϕ2 (t1 )))
u(t2 , s, xn (ϕ1 (s)))ds
0
Z t1
u(t1 , s, xn (ϕ1 (s)))ds|
f2 (t1 , xn (ϕ2 (t1 ))) ·
0
Z t2
|u(t2, s, xn (ϕ1 (s)))|ds
|f2 (t2 , xn (ϕ2 (t2 ))) − f2 (t1 , xn (ϕ2 (t1 )))|
0
Z t2
|f2 (t1 , xn (ϕ2 (t1 )))
u(t2 , s, xn (ϕ1 (s)))ds
0
Z t1
f2 (t1 , xn (ϕ2 (t1 )))
u(t2 , s, xn (ϕ1 (s)))ds|
0
Z t1
u(t2 , s, xn (ϕ1 (s)))ds
|f2 (t1 , xn (ϕ2 (t1 )))
0
Z t1
f2 (t1 , xn (ϕ2 (t1 ))) ·
u(t1 , s, xn (ϕ1 (s)))ds|
0
51
Z
≤
t2
|f2 (t2 , xn (ϕ2 (t2 ))) − f2 (t1 , xn (ϕ2 (t1 )))|
k(t2 , s)ds
0
Z t2
|u(t2 , s, xn (ϕ1 (s)))|ds
+ |f2 (t1 , xn (ϕ2 (t1 )))|
t1
Z t1
+ |f2 (t1 , xn (ϕ2 (t1 )))|
|u(t2 , s, xn (ϕ1 (s))) − u(t1 , s, xn (ϕ1 (s)))|ds
0
Z t2
≤ |f2 (t2 , xn (ϕ2 (t2 ))) − f2 (t1 , xn (ϕ2 (t1 )))|
k(t2 , s)ds
0
Z t2
k(t2 , s)ds + [a2 (t1 ) + b2 |xn (ϕ2 (t1 ))|]
+ [a2 (t1 ) + b2 |xn (ϕ2 (t1 ))|]
t1
Z t1
×
|u(t2 , s, xn (ϕ1 (s))) − u(t1 , s, xn (ϕ1 (s)))|ds.
0
Then we have
|Hn (t2 ) − Hn (t1 )| ≤ ω T (f2 , |t2 − t1 |)T k̃ + [a2 (t1 ) + b2 |xn (ϕ2 (t1 ))|](t2 − t1 )k̃
+ [a2 (t1 ) + b2 |xn (ϕ2 (t1 ))|]T ω T (u, |t2 − t1 |),
(4.6)
where ω T (f2 , ·) and ω T (u, ·) denotes the modulus continuity of the functions f2
and u on the sets Dε × R and Dε × Dε × R respectively and k̃ = max{|k(t, s)| :
(t, s) ∈ Dε × [0, T ]}. The last inequality (4.6) is obtained since M ⊂ Br .
Taking into account the fact that µ({xn }) ≤ µ(M) = 0, we infer that the number
t2 − t1 is small enough, then the right hand side of (4.6) tends to zero independently
of xn as t2 − t1 tends to zero. We have {Hn } is equicontinuous in the space C(Dε ).
Moreover,
|Hn (t)| ≤ |f2 (t, xn (ϕ2 (t))| ·
Z
t
|u(t, s, xn (ϕ1 (s)))|ds
Z t
≤ [|a1 (t)| + b2 |xn (ϕ2 (t))|] ·
k(t, s)ds
0
0
≤ k̃T [d1 + b2 d2 ],
where |a1 (t)| ≤ d1 , |xn (ϕ2 (t))| ≤ d2 for t ∈ Dε . From the above, we have that {Hn }
is equibounded in the space C(Dε ).
Next, let us put
Y = sup{|Hn (t)| : t ∈ Dε , n ∈ N}.
Obviously Y is finite in view of the choice of Dε . Assumption (i) conclude that
the function f1 |Dε ×[−Y,Y ] is uniformly continuous. So {B(xn )} = {Ff1 Hn + g(t, 0)}
52
is equibounded and equicontinuous in the space C(Dε ). Hence, by Ascoli-Arzéla
theorem [80], we obtain that the sequence {B(xn )} forms a relatively compact set
in the space C(Dε ).
Further observe that the above reasoning does not depend on the choice of ε.
Thus we can construct a sequence Dl of closed subsets of the interval [0, T ] such
that meas(Dlc ) → 0 as l → ∞ and such that the sequence {B(xn )} is relatively
compact in every space C(Dl ). Passing to subsequences if necessary we can assume
that {B(xn )} is a Cauchy sequence in each space C(Dl ), for l = 1, 2, · · · .
In what follows, utilizing the fact that the set B(M) is weakly compact, let us choose
a number δ > 0 such that for each closed subset Dδ of the interval [0, T ] such that
meas(Dδc ) ≤ δ, we have
Z
ε
(4.7)
|Bx(t)|dt ≤ ,
4
Dδc
for any x ∈ M.
Keeping in mind the fact that the sequence {Bxn } is a Cauchy sequence in each
space C(Dl ) we can choose a natural number l0 such that meas(Dlc0 ) ≤ δ and
meas(Dl0 ) > 0, and for arbitrary natural numbers n, m ≥ l0 the following inequality
holds
ε
|(B(xn ))(t) − (B(xm ))(t)| ≤
(4.8)
4meas(Dl0 )
for any t ∈ Dl0 .
Now use the above facts together with (4.5), (4.7), (4.8) we obtain
Z
0
∞
|(Bxn )(t) − (Bxm )(t)|dt =
Z
T
∞
|(Bxn )(t) − (Bxm )(t)|dt +
− (Bxm )(t)|dt +
Z
Dlc
0
Z
Dl0
|(Bxn )(t)
|(Bxn )(t) − (Bxm )(t)|dt ≤ ε,
which means that {B(xn )} is a Cauchy sequence in the space L1 (R+ ). Hence we
conclude that the set B(M) is relatively strongly compact in the space L1 (R+ ).
Step 6. We now can show that:
(1) From step 4, we obtain that A(M) + B(M) ⊆ M, where M is the set
constructed in step 3.
(2) Step 1 allow us to know that A is a contraction mapping.
(3) By step 5, B(M) is relatively compact and by assumptions (i), (iii) B is
continuous.
53
We can apply Theorem 1.6.7, and have that equation (4.1) has at least one integrable
solution in R+ .
4.3
Examples
We need to show an example for which our main result is useful and allow to extend
the existing theorems.
Example 4.3.1. Consider the following quadratic integral equation. Let t ∈ R+
2
Z t
t cos(ts)
1 t t2
1
t2
1
−t
+ x( + )+arctan [e + x(t + )] ·
ds . (4.9)
x(t) =
2
1 + t2 4 3 2
6
2
0 1 + (x(s))
It can be seen that equation (4.9) is a particular case of equation (4.1), where
g(t, x) =
1
1
1
+ x, f2 (t, x) = e−t + x, f1 (t, x) = arctan x2 ≤ 2x
2
1+t
4
6
and
u(t, s, x) =
t cos(ts)
.
1 + (x(s))2
Let us note that
|u(t, s, x)| ≤ t cos(ts),
Rt
since 0 t cos(ts) ds = sin t2 , then | 0 k(t, s) ds| ≤ 1, which implies that kK0 kL∞ ≤ 1.
Rt
Moreover, given arbitrary h > 0 and |x2 − x1 | ≤ δ we have
x22 − x21
|
(1 + x21 )(1 + x22 )
2thδ
|.
≤ |
(1 + x21 )(1 + x22 )
|u(t, s, x1 ) − u(t, s, x2 )| ≤ |t cos(ts)||
In view of Theorem 4.2.1, we can deduce
• g, f1 , f2 satisfy assumption (i) with a1 (t) = 0, a2 (t) = e−t , a3 (t) =
with constants b1 = 2, b2 = 16 and b3 = 41 .
1
1+t2
and
• Assumptions (ii), (iii) are satisfied,
• ϕ1 = t, ϕ2 = (t +
M1 = M2 = 1, M3 = 21 ,
• p =
1
2
< 1, q =
1
2
+
t2
),
2
1∗2
6
ϕ3 = ( 2t +
=
5
6
t2
)
2
satisfied assumption (iv) with
< 1.
Thus all the assumptions of Theorem 4.2.1 are satisfied so the quadratic functional
integral (4.9) has at least one integrable solution in R+ .
54
Chapter 5
On quadratic integral equations in
Orlicz spaces
5.1
Introduction
The chapter is devoted to study the following quadratic integral equation
Z b
x(t) = g(t) + G(x)(t) · λ
K(t, s)f (s, x(s)) ds.
(5.1)
a
We will deal with problems in which either the growth of the function f or the
kernel K is not polynomial. This is motivated, for instance, by some mathematical
models in physics. An interesting discussion about such a kind of problems can
be found in [45] or [120]. The considered thermodynamical problem lead to the
R
integral equation x(t) + I k(t, s) exp x(s) ds = 0 and thus the integral equations
with exponential nonlinearities turn out important from an application point of
view. Let us also note, that such a kind of problems can be applied for integral
equations associated (making use of the Green kernel) for the operator −∆u + exp u
on a bounded regular subsets of R2 (see [37]) and that the solutions in Orlicz spaces
are also sometimes studied in partial differential equations ([36], for instance).
Our theorems allow to consider the cases of integral equations when the kernel
function K is more singular than in previously considered cases. Moreover, we
are able to consider strongly nonlinear functions f . Both extensions seems to be
important from the applications point of view (cf. [36, 37, 45, 120], for instance). Let
us note, that our results are motivated by the paper of Cheng and Kozak [45]. We
generalize some of their assumptions and we consider more complicated quadratic
integral equations.
An operator G is supposed to be continuous on a required space of solutions. The
problem is modeled on some quadratic integral equations (G is usually identity operator or the Nemytskii superposition operator), but our approach allows to include
55
also standard integral equations. For standard integral equations different classes
of solutions are considered and this is mainly dependent on growth restrictions for
f and K. Similar investigations for quadratic integral equations relate mainly to
continuous solutions. The key point is to ensure, that an operator of pointwise
multiplication is well-defined and has some compactness properties. We propose an
approach to this problem allowing us to consider a wide class of integral equations
with solutions in some spaces of discontinuous functions.
We are interested in a different class of solutions then the earlier papers. Such
a kind of integral equations was investigated in spaces of continuous or integrable
functions. Similar problems are also important when Lp -solutions are checked. The
currently considered case is less restrictive and include large class of real problems.
Whenever one has to deal with some problems involving strong nonlinearities (of
exponential growth, for instance), it is a useful device to look for solutions not in
Lebesgue spaces, but in Orlicz spaces.
In the literature, mostly solutions of integral equations are sought in C[0, 1] and
Lp [0, 1] with p > 1. The results obtained for Lp [0, 1] invariably assume a polynomial
growth (in x) on the nonlinearity f (t, x). On the other hand, seeking solutions in
other Orlicz spaces will lead to restrictions that are not of polynomial type, and hence
will allow us to consider new classes of equations. All very basic types of integral or
differential equations were satisfactory examined (cf. [83, 98, 100, 103, 112]). Some
additional properties of solutions (in a simplest case of the ∆2 -condition) are also
investigated (constant-sign solutions, for instance [1, 2]). An interesting discussion
about advantages of integral equations in Orlicz spaces can be found in [84, Section
40] (see also [7]).
Nevertheless, for the quadratic integral equations the operators generated by the
right-hand side of the equation are more complicated and was not investigated in
this case. Let us note, that in this case the methods based on properties of some
Banach algebras are usually applied (cf. [12, 34, 89, 90]). This approach seems to
be strictly related with continuous or Hölder continuous solutions (the product is
an inner operation in the Banach algebra of continuous functions and at the same
time is an operator used in the integral equation). Let us note, that this method
is dependent on some properties of C(0, 1) and cannot be easily applied to different
classes of spaces.
This suggest an operator oriented approach. In a class of Orlicz spaces we
consider spaces associated with growth conditions for G and f . For a moment
denote by X an Orlicz space of solutions for our problem and by F the Nemytskii
superposition operator generated by f . Thus we have G : X → W1 , F : X → U
and finally the linear integral operator H with a kernel K is acting from U into
56
W2 . The space U is depending on some growth assumptions of f - not necessarily
of polynomial type. In a typical case of quadratic problems the spaces W1 and W2
are supposed both to be the space of continuous functions and then some properties
of this Banach algebra allow to solve the problem. Unfortunately, this is really
restrictive assumption. We started to replace this assumption by considering X =
L1 (I) and W2 = L∞ (I) ([46]).
Here we present a complete theory for such problems. In general, allowing U be
an Orlicz space depending on f we consider the triples of Orlicz spaces (not: Banach
algebras) for which the pointwise multiplication take a pair of functions from W1
and W2 into X.
For a case of Lp -solutions we propose to use a factorization and we will assume
the Hammerstein integral operator which is multiplied by the function G(x) has
values in the space conjugated to the space of solutions. Such an idea was used
by Brézis and Browder for Hammerstein integral equations ([39]) by considering
conjugated Lebesgue Lp spaces. We extend this procedure for Orlicz (or: ideal)
spaces and for a triple of spaces (two for Hammerstein operator and one more for
a multiplication operator). This allows us to prove the existence theorems under
much more general conditions than previously considered ones. Our method leads
to extensions for both types of results (quadratic and non-quadratic ones).
We concentrate on the property of monotonicity of solutions for the equation
(5.9). This notion is broaden to some function spaces and the basic properties of
families of such functions are investigated.
Especially, the quadratic integral equation of Chandrasekhar type
Z 1
t
ϕ(s)x(s) ds
x(t) = 1 + x(t)
0 t+s
can be very often encountered in many applications (cf. [11, 12, 31, 26, 40, 43]).
The results of this chapter are divided into a few parts. This is because the
proofs are depending on the choice of considered spaces. We stress on the ”size” of
solution spaces and we try to relate the growth assumptions of functions and the
expected space of solutions.
Since for equations of this type an approach via the Schauder fixed point theorem is not useful and the Banach contraction principle is too restrictive in many
applications, we prefer to investigate the properties of operators with respect to the
topology of convergence in measure. We check this topology on considered Orlicz
spaces and then we use the Darbo fixed point theorem for proving main results. To
show a detailed theory we need to consider a few different cases (for different classes
of Orlicz spaces). The theorems proved by us extend, in particular, that presented
in [6, 11, 27, 33] considered in the space C(I) or in Banach algebras (cf. [34]). In the
57
context of non-quadratic integral equations in Orlicz spaces, which are also cover by
our theorems, let us mention the papers [1, 2, 98, 100, 103, 112].
5.2
The case of operators with values in L∞ (I).
This part will be devoted to present some new results, which are related to our
theorems from previous chapters. Nevertheless, we are looking for solutions in some
Orlicz spaces instead of Lebesgue ones. Functional dependence is not considered
here, but our current approach is presented in such a way to cover non-quadratic
integral equations as particular cases. Earlier results for quadratic integral equations
cannot be compared with standard ones.
Denote B the operator associated with the right-hand side of the equation (5.2)
Z b
x(t) = g(t) + λf1 (t, x(t))
K(t, s)f2 (s, x(s)) ds.
(5.2)
a
i.e. B(x) = g + U(x), where U(x)(t) = F1 (x)(t) · A(x)(t) such that
Z
F2 (x)(t) = f2 (t, x(t)), A = H ◦ F2 , H(x)(t) = λ K(t, s)x(s) ds, F1 = f1 (s, x(s)).
I
By considering different spaces of solutions (i.e. different growth conditions) we
need to investigate pair of spaces constituting the domain and the range for the
operator U. To facilitate the reading of (technical) assumptions we will consider
different pairs of spaces, including most typical ones. We stress on applicability of
our results. The general case will presented in an abstract form in the next Section
of this chapter.
Let F1 : V1 → W1 and A : V1 → W2 , where V1 , W1 and W2 are some function
spaces. Since U stands for the pointwise multiplication operator and the range of U
should be in a space of solutions W , we need to investigate pair of spaces (W1 , W2 )
such that x(t) · y(t) ∈ W for all t ∈ [a, b] and x ∈ W1 , y ∈ W2 .
We need to consider the case when either W1 or W2 is a space of all essentially
bounded functions L∞ [a, b].
5.2.1
The case of W1 = L∞ (I).
The case of W1 = L∞ (I) is trivial. The unique situation in which the Nemytskii
operator takes values in this space is when f1 is autonomous. Namely, we have:
Lemma 5.2.1. [10, Theorem 3.17] Let p ≥ 1. The superposition operator F1
generated by f1 maps Lp (I) into L∞ (I) if and only if
|f (s, u)| ≤ a(s) for u ∈ R
58
for some a ∈ L∞ (I), in this case, F1 is always bounded; F1 is continuous if and only
if F1 is constant (i.e. f1 does not depend on u).
Let us briefly discuss the case of Orlicz spaces. We have the following lemma
Lemma 5.2.2. Let LM (I) be an Orlicz space. Then F1 generated by a Carathéodory
function f1 is continuous from LM (I) into L∞ (I) if and only if F1 is constant (i.e.
f1 does not depend on u).
Proof. Since we consider a finite measure, an arbitrary function x from LM can
be obtained as a pointwise limit of a sequence of simple functions, a fortiori this
sequence is convergent in measure. Simple functions belong to EM and then LM is
a quasi-regular ideal space (cf. [10, Theorem 2.8, Theorem 3.17]).
On the other hand the space L∞ (I) is completely irregular i.e. the regular part
of the space consists only {θ}. Since both spaces are ideal, by [10, Theorem 2.8] F1
is continuous if and only if F1 is constant.
Note that a domain for F1 cannot be arbitrarily small. Some assumptions on
the domain of superposition operators are always expected (the identity on L∞ (I)
is continuous, for instance).
Thus, this is the standard non-quadratic case and can be easily reduced to known
theorems (cf. [100, 103, 112] for most interesting results).
5.2.2
The case of W2 = L∞ (I).
We will consider now the new case when W2 = L∞ (I). It is more interesting and we
stress on growth conditions for both f1 and f2 allowing to have nonlinear growth.
This is closely related to the choice of an intermediate space for A i.e. the domain
of H. It can be made because of the assumptions on f2 or the kernel K depending
on the practical problem described by our equation. Thus the choice of this space
allows to consider more general growth conditions for f2 or K. Theorem given below
is formulated for arbitrary Lebesgue space Lp (I) (p ≥ 1) taken as an intermediate
space.
Theorem 5.2.1.
Let
1
p
+
1
q
= 1 and assume, that ϕ is an N-function and that:
(C1) g ∈ Eϕ (I) is nondecreasing a.e. on I,
(C2) fi : I × R → R satisfies Carathéodory conditions and fi (t, x) is assumed to
be nondecreasing with respect to both variable t and x separately, for i=1, 2.
59
(C3) kF1 (x)kϕ ≤ G1 (kxkϕ ), and kF2 (x)k|p ≤ G2 (kxkϕ ), for all x ∈ Eϕ (I),
where Gi are positive, continuous and nondecreasing for i = 1, 2. Moreover
there exist γ > 0 such that G1 (u) ≤ γ|u| for |u| ≤ 1. Assume, that the
superposition operator F1 generated by f1 is acting from Eϕ (I) into itself.
(C4) Assume that the function K is measurable in (t, s) and assume that the linear
integral operator H with kernel K(t, s) maps Lp (I) into L∞ (I), s 7→ |K(·, s)| ∈
Lq (I), k(t) = |K(t, ·)| ∈ L∞ (I) and H is continuous with a norm
kK0 kL∞ := esssupt∈I
(K1)
R
K(t1 , s) ds ≥
I
R
I
Z
I
|K(t, s)|q ds
1q
.
K(t2 , s) ds for t1 , t2 ∈ I with t1 < t2 .
Assume, that there exists a positive number r ≤ 1 such that
kgkϕ + λkK0 kL∞ G1 (r) · G2 (r) ≤ r.
(5.3)
Then there exists a number ρ > 0 such that for all λ ∈ R with |λ| < ρ and for every
g ∈ Eϕ , there exists a solution x ∈ Eϕ (I) of (5.2) which is a.e. nondecreasing on
I.
Proof. We need to divide the proof into a few steps.
I. The operator B is well-defined from Lϕ (I) into itself. In particular, for operator H we need to check the properties of this operator.
II. We will construct an invariant ball Br for B in Lϕ (I).
III. We construct a subset Qr of this ball which contains a.e. nondecreasing functions and we investigate the properties Qr .
IV. We check the continuity and monotonicity properties of B in Qr , so
B : Qr → Qr .
V. We prove that B is a contraction with respect to a measure of noncompactness.
VI. We use the Darbo fixed point theorem to find a solution in Qr .
I. First of all observe that by the assumptions (C2) and (C3) and Lemma 1.3.5
implies that F2 is continuous mappings from Lϕ (I) into Lp . Note, that Lp can be
p
treated as an Orlicz space LMp for Mp (x) = |x|p . It is clear, that this space satisfies
the ∆2 -condition and therefore it is a regular space. Recall, that by Lemma 1.3.2 F2
60
is sequentially continuous with respect to convergence in measure and continuous
by Lemma 1.3.5. For the operator F1 the situation is described in Remark 1.3.2.
By assumption (C4) the operator A maps Eϕ (I) into L∞ (I) and is continuous. F1
maps Eϕ (I) into itself continuously (thanks to Assumption (C3) and Lemma 1.3.5).
Thus U is a continuous mapping from Eϕ (I) into itself. Finally, Assumption (C1)
permits us to deduce that the operator B maps Eϕ (I) into itself and is continuous.
II. If ϕ satisfies the ∆2 -condition then F1 is bounded in Lϕ (I). In a general
case we need restrict this operator to the (regular) space Eϕ (I). We will prove the
boundedness of the operator B on this space, namely we will construct the invariant
set for this operator.
Let Γ be a set of all positive numbers λ such that
kgkϕ + λkK0 kL∞ G1 (r) · G2 (r) ≤ r.
Put ρ = min sup Γ, γkK0 kL1 G2 (1) and fix λ with |λ| < ρ.
∞
Let r be a positive number r ≤ 1 such that
kgkϕ + |λ|kK0kL∞ · G1 (r) · G2 (r) ≤ r.
(5.4)
Since we provided a continuity of operators F1 and B only in Eϕ (I), as a domain
for the operator B we will consider the ball Br (Eϕ (I)) in this space.
Recall, that Lϕ (I) is an ideal space, F1 (x) ∈ Lϕ (I) and A(x) ∈ L∞ (I), so
U(x) ∈ Lϕ (I) and kU(x)kϕ ≤ kF1 (x)kϕ kA(x)k∞ .
We have
kB(x)kϕ ≤ kgkϕ + kU(x)kϕ
= kgkϕ + kF1 (x) · A(x)kϕ
≤ kgkϕ + |λ| · kF1 (x)kϕ · k
Z
I
K(t, s)|f2 (s, x(s))|ds k∞ .
But by (C4) H is continuous and then the norm of H(x) is estimated by (cf. [78,
Theorem XI.1.6])
kH(x)k∞ ≤ |λ|kK0kL∞ kxkp .
Thus for x ∈ Eϕ (I) with kxkϕ ≤ r
kB(x)kϕ ≤ kgkϕ + kF1 (x)kϕ · |λ|kK0kL∞ · kF2 (x)kp
≤ kgkϕ + |λ|kK0 kL∞ G1 (kxkϕ ) · G2 (kxkϕ )
≤ kgkϕ + |λ|kK0 kL∞ G1 (r) · G2 (r) ≤ r.
Then we have B : Br (Eϕ (I)) → Br (Eϕ (I)). Moreover, B is continuous on Br (Eϕ (I))
(see the part I of the proof).
61
III. Let Qr stand for the subset of Br (Eϕ (I)) consisting of all functions which
are a.e. nondecreasing on I. Similarly as claimed in [17] this set is nonempty,
bounded (by r), convex (direct calculation from the definition) and closed in Lϕ (I).
To prove the last property, let (yn ) be a sequence of elements in Qr convergent
in Lϕ (I) to y. Then the sequence is convergent in measure and as a consequence
of the Vitali convergence theorem for Orlicz spaces and of the characterization of
convergence in measure (the Riesz theorem) we obtain the existence of a subsequence
(ynk ) of (yn ) which converges to y almost uniformly on I (cf. [100]). Moreover, y is
still nondecreasing a.e. on I which means that y ∈ Qr and so the set Qr is closed.
Now, in view of Lemma 1.4.1 the set Qr is compact in measure.
IV. Now, we will show, that B preserve the monotonicity of functions. Take
x ∈ Qr , then x is a.e. nondecreasing on I and consequently F1 and F2 are also
of the same type in virtue of the assumption (C2) and Theorem 1.4.2. Further,
A(x) = H ◦ F2 (x) is a.e. nondecreasing on I (thanks for the assumption (K1)).
Since the pointwise product of a.e. monotone functions is still of the same type, the
operator U is a.e. nondecreasing on I. Moreover, the assumption (C1) permits us
to deduce that B(x)(t) = g(t) + U(x)(t) is also a.e. nondecreasing on I. This
fact, together with the assertion that B : Br (Eϕ (I)) → Br (Eϕ (I)) gives us that B
is also a self-mapping of the set Qr .
V. We will prove that B is a contraction with respect to a measure of strong
noncompactness. Since
Z
Z
|A(x)(t)| = |λ K(t, s)f2 (s, x(s))ds| ≤ |λ| |K(t, s)||f2(s, x(s))|ds
IZ
I
≤ |λ| · |K(t, s)||F2(x)(s)|ds
I
≤ |λ| · kK(t, ·)kq · |F2 (x)|p
≤ |λ| · kK(t, ·)kq · G2 (kxkϕ )
for a.e. t ∈ I, whence kA(x)χD kL∞ ≤ |λ|kK0 kL∞ · G2 (kxkϕ ).
Assume that X is a nonempty subset of Qr and let the fixed constant ε > 0 be
arbitrary. Then for an arbitrary x ∈ X and for a set D ⊂ I, meas(D) ≤ ε
and t ∈ D we have F1 (x)(t) = F1 (xχD )(t) and for t 6∈ D F1 (x)(t) · χD (t) = 0 and
F1 (xχD )(t) = f1 (t, 0). Whence
|F1 (x)(t)χD (t)| = |f1 (t, x(t))| = |f1 (t, x(t)) − f1 (t, 0) + f1 (t, 0)|
≤ |f1 (t, x(t)) − f1 (t, 0)| + |f1 (t, 0)|
≤ |F1 (xχD )(t) − F1 (0)(t)χD (t)| + |F1 (0)(t)χD (t)|.
62
Thus for all t ∈ I we have
|F1 (x)(t)χD (t)| ≤ |F1 (xχD )(t) − F1 (0)(t)χD (t)| + |F1 (0)(t)χD (t)|.
Finally
kF1 (x) · χD kϕ ≤ kF1 (xχD ) − F1 (0)kϕ + kF1 (0) · χD kϕ ≤ kF1 (xχD )kϕ + 2kF1 (0) · χD kϕ .
For the operator B we get the following estimation
kB(x) · χD kϕ ≤ kg · χD kϕ + kU(x) · χD kϕ
= kg · χD kϕ + k F1 (x) · A(x) · χD kϕ
= kg · χD kϕ + k (F1 (x) · χD ) · A(x)kϕ
≤ kg · χD kϕ + k F1 (x) · χD kϕ · kA(x)kL∞
≤ kg · χD kϕ + (kF1 (xχD )kϕ + 2kF1 (0) · χD kϕ )|λ|kK0 kL∞ G2 (kxkϕ )
≤ kg · χD kϕ + (G1 (kxχD kϕ ) + 2kF1 (0) · χD kϕ ) |λ|kK0kL∞ G2 (kxkϕ )
≤ kg · χD kϕ + (γ · kxχD kϕ + 2kF1 (0) · χD kϕ ) |λ|kK0 kL∞ G2 (kxkϕ ).
Since g ∈ Eϕ and F1 (0) ∈ Eϕ ,
lim
sup
ε→0 meas(D)≤ε
lim
sup
ε→0 meas(D)≤ε
kg · χD kϕ = 0,
kF1 (0) · χD kϕ = 0.
Thus by definition of c(x) and by taking the supremum over all x ∈ X and all
measurable subsets D with meas(D) ≤ ε we get
c(B(X)) ≤ γ|λ|kK0kL∞ · G2 (r) · c(X).
Since X ⊂ Qr is a nonempty, bounded and compact in measure subset of an ideal
regular space Eϕ (I), we can use Proposition 5.5.1 and get
βH (B(X)) ≤ γ · |λ| · kK0 kL∞ · G2 (1) · βH (X).
The constant in the above inequality is smaller that 1, so the properties of the
operator B and assumption (C4) allow us to apply the Darbo Fixed Point Theorem
1.6.5, which completes the proof.
Remark 5.2.1. In this chapter we consider continuous linear operators of the form
H acting on Lp (I) with values in the space L∞ (I). It is known, that the continuity
63
property is depending on the kernel K. In a particular case of Riemann-Liouville
1
(t − s)α−1 χ[a,t] (s)
fractional integral operators i.e. for K(t, s) = Γ(α)
1
Jα x(s) =
Γ(α)
Z
a
s
(s − t)α−1 x(t)dt
s ∈ [a, b]
,
is not continuous from Lp (I) into L∞ (I) when p =
continuous for p < α1 (cf. also [7]).
1
α
([70, Remark 4.1.2]), but
Remark 5.2.2. Let us add some comments about assumption (C3). Our acting
condition from Lemma 1.3.3 has the form |f2 (t, x)|p ≤ a2 (t)+b2 ϕ(x). This pointwise
estimation implies our assumption. Indeed, this implies that for x ∈ Lϕ (I) with
kxkϕ ≤ 1
Z
Z
p
|f2 (t, x(t))| dt ≤ ka2 k1 + b2 ϕ(x(t))dt
I
I
kF2 (x)kpp ≤ ka2 k1 + b2 kxkϕ
1
and then kF2 (x)kp ≤ (ka2 k1 + b2 kxkϕ ) p , so our assumption holds true for a special
1
case of G2 (t) = (ka2 k1 + b2 · t) p which is used in [1], for instance. The growth
restrictions for G1 result from necessary and sufficient conditions for continuity of
F1 (see [10]) and since L∞ (I) is completely irregular i.e. θ is the only singleton with
absolutely continuous norm.
The boundedness of the Nemytskii operators on ”small” balls was firstly proved
by Shragin in [110] and then used to investigate the Hammerstein integral equations
in Orlicz spaces by Vainberg and Shragin.
5.3
The existence of Lp-solution.
Let us present a special case of Lp -solutions. This will be still more general result
than the earlier ones.
Theorem 5.3.1.
Assume, that p ≥ 1 and
1
p
+
1
q
= 1.
(C1’) g ∈ Lp (I) is nondecreasing a.e. on I,
(C2’) fi : I × R → R satisfies Carathéodory conditions and fi (t, x) is assumed to
be nondecreasing with respect to both variable t and x separately, for i=1, 2.
p
(C3’) |f1 (t, x)| ≤ a1 (t) + b1 |x|, and |f2 (t, x)| ≤ a2 (t) + b2 |x| q , for all t ∈ I and
x ∈ R, where a1 ∈ Lp (I), a2 ∈ Lq (I) and some constants bi ≥ 0 for i = 1, 2.
64
(C4’) Assume that the function K is measurable in (t, s) and that the linear integral
operator K0 with kernel K(·, ·) maps Lq (I) into L∞ (I),
essupt∈[a,b]
b
Z
q
a
|K(t, s)| ds
q1
<∞
and is continuous.
Rb
Rb
(K1) a K(t1 , s) ds ≥ a K(t2 , s) ds for t1 , t2 ∈ [a, b] with t1 < t2 .
Assume, that there exists a positive number r < 1 such that
kgkp + |λ|kK0 kL∞ · [ ka1 kL1 + b1 · r ][ ka2 kL1 + b2 · r p/q ] ≤ r
and choose λ in such a way to get kK0 kL∞ <
1
p
|λ|·b1 b2 ·r q
(5.5)
.
Then there exists a number ρ > 0 such that for all λ ∈ R with |λ| < ρ and for every
g ∈ Lp (I), there exists a solution x ∈ Lp (I) of (5.2) which is a.e. nondecreasing
on I.
Proof. I. First of all observe that by assumptions (C2’), (C3’) and Theorem 1.3.1 we
have that F1 maps Lp (I) into itself and F2 maps Lp (I) into Lq (I) continuously. By
assumption (C4’) we can deduce that A maps Lp (I) into L∞ (I) and is continuous.
From the Hölder inequality the operator U maps Lp (I) into itself continuously.
Finally, assumption (C1’) permits us to deduce that B maps Lp (I) into itself and is
continuous.
To prove step II. Put
ρ=
1
b1 b2 kK0′ kL∞ ka1 kLp
· ka2 kLq
.
Fix λ with |λ| < ρ.
Choose a positive number R in such a way that
p
kgkLp + |λ| · [ k a1 kLp + b1 · R ] · kK0′ kL∞ [ ka2 kLq + b2 · R q ] ≤ R.
As a domain for the operator B we will consider the ball BR (Lp (I)).
kB(x)kLp ≤ kgkLp + kUxkLp
= kgkLp + kF1 (x) · A(x)kLp
≤ kgkLp + kF1 (x)kLp · kA(x)kL∞
Z
≤ kgkLp + |λ|kf1(t, x)kLp k K(t, s)|f2 (s, x(s))|dskL∞
I
65
(5.6)
≤ kgkLp + |λ| · k a1 + b1 |x(s)|kLp
Z
p
× k K(t, s) [ a2 (s) + b2 |x(s)| q ]ds kL∞
I
p
≤ kgkLp + |λ|[ka1 kLp + b1 kxkLp ]kK0′ kL∞ [ka2 kLq + b2 kxkLq p ],
where
p
p
kx q kLq = kxkLq p .
p
kB(x)kLp ≤ kgkLp + |λ| · [ka1 kLp + b1 · R] · kK0′ kL∞ [ka2 kLq + b2 · R q ] ≤ R.
Then we have B : BR (Lp (I)) → BR (Lp (I)). Moreover, B is continuous on BR (Lp (I))
(see the part I of the proof).
Step III and Step IV are similar as in Theorem 5.2.1.
V. We will prove that B is a contraction with respect to a measure of strong
noncompactness. Assume that X is a nonempty subset of QR and let the fixed
constant ǫ > 0 be arbitrary . Then for an arbitrary x ∈ X and for a set D ⊂ I,
meas(D) ≤ ǫ we obtain
kB(x) · χD kLp ≤ kg · χD kLp + kUx · χD kLp
≤ kg · χD kLp + k F1 (x) · χD kLp · kA(x) · χD kL∞
Z
K(t, s)f2 (s, x(s))dskL∞
= kg · χD kLp + |λ| · kf1 (t, x) · χD kLp k
D
≤ kg · χD kLp + |λ| · k [ a1 (s) + b1 |x(s)| ] · χD kLp
Z
p
× k
K(t, s)[ a2 (s) + b2 |x(s)| q ]ds kL∞
D
≤ kg · χD kLp + |λ| · k [ ka1 · χD kLp + b1 kx · χD kLp ]
Z
Z
p
× k
K(t, s) a2 (s) ds + b2 K(t, s)|x(s)| q ds kL∞
D
I
≤ kg · χD kLp + |λ| · k [ ka1 · χD kLp + b1 kx · χD kLp ]
p
×
[ kK0′ kL∞ ka2 · χD kLp + b2 kK0′ kL∞ kx(s)kLq p ]
×
kK0′ kL∞ [ ka2 · χD kLp + b2 R q ].
≤ kg · χD kLp + |λ| · k [ ka1 · χD kLp + b1 kx · χD kLp ]
p
Since g ∈ Lp , a1 ∈ Lp and a2 ∈ Lq , then we have
lim { sup [sup{kg · χD kLp }]} = 0, lim { sup [sup{ka1 · χD kLp }]} = 0
ε→0
mes D≤ε x∈X
ε→0
66
mes D≤ε x∈X
and
lim { sup [sup{ka2 · χD kLq }]} = 0.
ε→0
mes D≤ε x∈X
Thus by definition of c(x),
p
c(B(X)) ≤ b1 b2 |λ| · kK0′ kL∞ R q · c(X).
Since X ⊂ Qr is a nonempty, bounded and compact in measure subset of an ideal
regular space Lp , we can use Proposition 5.5.1 and get
p
βH (B(X)) ≤ b1 b2 |λ| · kK0′ kL∞ R q · βH (X).
The inequality obtained above together with the properties of the operator B
and the set QR established before and assumption (C’4) allow us to apply the Darbo
fixed point theorem (see [23]), which completes the proof.
5.3.1
Remarks and examples.
We need to stress on some aspects of our results. First of all we can observe, that our
solutions are not necessarily continuous as in almost all previously investigated cases.
In particular, we need not to assume, that the Hammerstein operator transforms
the space C(I) into itself. Our treatment allows to solve problems when function
f1 doesn’t have sublinear growth. In this case it is sufficient to take a little bit
”nicer” kernel K, but this is still weaker assumption than in the case of continuous
solutions.
We need to stress, that quadratic equations are also strictly related to problems
of the type
′
x(t) − g(t)
= f2 (t, x(t)) x(0) = 0,
f1 (t, x)
where f1 : I × R → R \ {0}.
It is well-known that under typical assumptions this problem is equivalent to the
integral equation
Z b
g(0)
x(t) = g(t) + f1 (t, x(t)) ·
χ[0,t] (s)f2 (s, x(s)) ds −
.
f1 (0, 0)
a
Nevertheless, when we are looking for continuous solutions for integral equation,
we obtain classical solutions for differential one i.e. x is continuously differentiable.
This seems to be too restrictive in many applications. In our case we investigate
Caratéodory solutions for the Cauchy problem.
Another typical example is a boundary value problem (f1 : I × R → R \ {0}, f2
and g satisfy some regularity conditions)
′′
x(t) − g(t)
= f2 (t, x(t)) x(0) = 0 , x′ (0) = 0.
(5.7)
f1 (t, x)
67
In this case we can consider an equivalent integral problem with a kernel G (an
appropriate Green function)
Z b
g(0)
(5.8)
x(t) = g(t) +
f1 (t, x(t)) ·
G(t, s)f2 (s, x(s)) ds −
f1 (0, 0)
a
!!
1
(0,
0)
g(0) · ∂f
g ′(0)
∂t
+ t·
−t·
,
f1 (0, 0)
(f1 (0, 0))2
where f1 : I × R → R \ {0}.
In this case when we are looking for continuous solutions for the quadratic integral equations, the solutions for the above differential problems are classical. Our
approach allows to investigate weaker types of solutions (in Orlicz-Sobolev spaces).
5.4
A general case of Orlicz spaces.
Here we will present a detailed theory of quadratic integral equations of the form:
Z b
x(t) = g(t) + G(x)(t) · λ
K(t, s)f (s, x(s)) ds
(5.9)
a
in Orlicz spaces.
Denote by B the operator associated with the right-hand side of the equation
(5.9) i.e. B(x) = g + U(x), where
U(x)(t) = G(x)(t) · λ
Z
b
K(t, s)f (s, x(s)) ds.
a
Rb
Thus B = g + G · A = g + G · H ◦ F , where H(x)(t) = λ a K(t, s)x(s) ds and
F = f (t, x(t)).
We will try to choose the domains of operators defined above in such a way
to obtain the existence of solutions in a desired Orlicz space Lϕ (I). We stress on
conditions allowing us to consider strongly nonlinear operators.
Let us note, that our assumptions on G are referred to the case of quadratic
integral equations (i.e. G(x)(t) = q(t) · x(t)).
We need to distinguish two different cases. This allow us to obtain more general
growth conditions on f (cf. [100, 103, 112, 98] for non-quadratic equations). In
every case we need to describe some assumptions on ”intermediate” spaces being
the images of Lϕ (I) for G and F (Lϕ1 (I) and LM (I), respectively) and the range
for H (i.e. Lϕ2 (I)). This approach is based on a classical (non-quadratic) case as in
[100, 103, 112]) and seems to be important in view of optimality of assumptions for
every considered case.
68
The case of N satisfying the ∆′-condition.
5.4.1
Theorem 5.4.1. Assume, that ϕ, ϕ1 , ϕ2 are N-functions and that M and N are
complementary N-functions. Moreover, put the following set of assumptions:
(N1) there exists a constant k1 > 0 such that for every u ∈ Lϕ1 (I) and w ∈ Lϕ2 (I)
we have kuwkϕ ≤ k1 kukϕ1 kwkϕ2 ,
(C1) g ∈ Eϕ (I) is nondecreasing a.e. on I,
(C2) f : I × R → R satisfies Carathéodory conditions and f (t, x) is assumed to
be nondecreasing with respect to both variable t and x separately,
(C3) |f (t, x)| ≤ b(t) + R(|x|) for t ∈ I and x ∈ R, where b ∈ EN (I) and R is
nonnegative, nondecreasing, continuous function defined on R+ ,
(C4) Let N satisfies the ∆′ -condition and suppose that there exist ω, γ, u0 ≥ 0 for
which
N(ω(R(u))) ≤ γϕ2 (u) ≤ γM(u) for u ≥ u0 ,
(G1) G : Lϕ (I) → Lϕ1 (I), takes continuously Eϕ (I) into Eϕ1 (I) and there exists a
constant G0 > 0 such that kG(x)kϕ1 ≤ G0 kxkϕ and that G takes the set of all
a.e. nondecreasing functions into itself,
(K1) s → K(t, s) ∈ LM (I) for a.e. t ∈ I,
(K2) K ∈ EM (I 2 ) and t → K(t, s) ∈ Eϕ2 (I) for a.e. s ∈ I with
kKkM <
(K3)
,
Rb
a
K(t1 , s) ds ≥
Rb
a
1
2k1 · |λ| · G0 · R(1)
K(t2 , s) ds for t1 , t2 ∈ [a, b] with t1 < t2 .
Then there exists a number ρ > 0 such that for all λ ∈ R with |λ| < ρ and for all g
with kgkϕ < 1 there exists a solution x ∈ Eϕ (I) of (5.9) which is a.e. nondecreasing
on I.
Proof. We need to divide the proof into a few steps.
I. The operator B is well-defined from Lϕ (I) into itself and continuous on a
domain depending on the considered case.
II. We will construct an invariant ball Br for B in Lϕ (I).
III. We construct a subset Qr of this ball which contains a.e. nondecreasing functions and we investigate the properties Qr .
69
IV. We check the continuity and monotonicity properties of B in Qr , so U : Qr →
Qr .
V. We prove that B is a contraction with respect to a measure of noncompactness.
VI. We use the Darbo fixed point theorem to find a solution in Qr .
I. First of all observe that under the assumptions (C2) and (C3) by Lemma
1.3.3 the superposition operator F acts from Lϕ (I) to LN (I).
In this case we will prove, that U is a continuous mapping from the unit ball in
Eϕ (I) into the space Eϕ (I).
Let us recall, that x ∈ Eϕ (I) iff for arbitrary ε > 0 there exists δ > 0 such that
kxχT kϕ < ε for every measurable subset T of I with the Lebesgue measure smaller
that δ (i.e. x has absolutely continuous norm). First, let us observe that in view of
Lemma 1.3.8, it is sufficient to check this property for the operator A = H ◦ F .
Since N is an N-function satisfying ∆′ -condition and by (C3), we are able to
use [83, Lemma 19.1]. From this there exists a constant C (not depending on the
kernel) such that for any measurable subset T of I and x ∈ Lϕ (I), kxkϕ ≤ 1 we
have
kA(x)χT kϕ2 ≤ CkKχT ×I kM .
(5.10)
Now, by the Hölder inequality and the assumption (C2) we get
|K(t, s)f (s, x(s))| ≤ kK(t, s)k · |f (s, x(s))| ≤ kK(t, s)k · | (b(s) + R(|x(s)|)) |
for t, s ∈ I. Put k(t) = 2kK(t, ·)kM for t ∈ I. As K ∈ EM (I 2 ) this function
is integrable on I. By the assumptions (K1) and (K2) about the kernel K of the
operator H (cf. [112]) we obtain that
kA(x)(t)k ≤ k(t) · (kbkN + kR(|x(·)|)kN ) for a.e. t ∈ I.
Whence for arbitrary measurable subset T of I and x ∈ Eϕ (I)
kA(x)χT kϕ2 ≤ kkχT kϕ2 · (kbkN + kR(|x(·)|kN ) .
Finally if t is such that K(t, ·) ∈ EM (I) and x ∈ Eϕ (I) we have
Z
kK(t, s)f (s, x(s))k ds ≤ 2kK(t, ·)χT kM · (kbkN + kR(|x(·)|)kN ) for a.e. t ∈ I.
T
From this it follows that A maps B1 (Eϕ (I)) into Eϕ2 (I).
We are in a position to prove the continuity of A as a mapping from the unit ball
B1 (Eϕ (I)) into the space Eϕ2 (I). Let xn , x0 ∈ B1 (Eϕ (I)) be such that kxn −x0 kϕ →
0 as n tends to ∞. Suppose, contrary to our claim, that A is not continuous and
70
the kA(xn ) − A(x0 )kϕ2 does not converge to zero. Then there exists ε > 0 and a
subsequence (xnk ) such that
kA(xnk ) − A(x0 )kϕ2 > ε for k = 1, 2, ...
(5.11)
and the subsequence is a.e. convergent to x0 . Since (xn ) is a subset of the ball the
Rb
sequence ( a ϕ(|xn (t)|)dt) is bounded. As the space Eϕ2 (I) is regular the balls are
Rb
norm-closed in L1 (I) so the sequence ( a |xn (t)|dt) is also bounded.
Moreover, by (C3) and (C4) there exist ω, γ, u0 > 0, s.t. (cf. [83, p. 196])
1
kωR(|x(·)|)kN
ω
Z
1
≤
inf
N(ωR(|x(t)|)/r)dt ≤ 1
ω r>0
Z b
1
1+
N(ωR(|x(t)|))dt
≤
ω
a
Z b
1
1 + N(ωR(u0 )) · (b − a) + γ
ϕ2 (|x(t)|)dt ,
≤
ω
a
kR(|x(·)|)kN =
whenever x ∈ Lϕ (I) with kxkϕ ≤ 1.
Thus
Z
kK(t, s)f (s, xn (s))k ds ≤ 2kK(t, ·)χT kM · (kbkN + kR(|xn (·)|)kN )
T
≤ 2kK(t, ·)χT kM · (kbkN
Z b
1
1 + N(ωR(u0 )) · (b − a) + γ
ϕ2 (|xn (t)|)dt )
+
ω
a
and then the sequence (kK(t, s)f (s, xn (s))k) is equiintegrable on I for a.e. t ∈ I.
By the continuity of f (t, ·) we get limk→∞ K(t, s)f (s, xnk (s)) = K(t, s)f (s, x0 (s))
for a.e. s ∈ I. Now, applying the Vitali convergence theorem we obtain that
lim A(xnk )(t) = A(x0 )(t) for a.e. t ∈ I.
k→∞
But the equation (5.10) implies that A(xnk ) is a subset of Eϕ2 (I) and then
lim A(xnk )(t) = A(x0 )(t)
k→∞
which contradicts the inequality (5.11). Since A is continuous between indicated
spaces, By our assumption (G1) the operator G is continuous from B1 (Eϕ (I)) into
Eϕ1 (I) and then by (N1) the operator U has the same property and then U is a continuous mapping from B1 (Eϕ (I)) into the space Eϕ (I). Finally, by the assumption
(C1) B maps B1 (Eϕ (I)) into Eϕ (I) continuously.
71
II. We will prove the boundedness of the operator U, namely we will construct
the invariant ball for this operator. By B we will denote the right-hand side of our
integral equation i.e. B = g + U.
Set r ≤ 1 and let
ρ=
1 − kgkϕ
.
2k1 · C · G0 · kKkM
Let x be an arbitrary element from B1 (Eϕ (I)). Then by using the above consideration, the assumption (C3), the formula (5.10) and Proposition 1.3.8 for sufficiently
small λ (i.e. |λ| < ρ) we obtain
kB(x)kϕ ≤ kgkϕ + kUxkϕ
= kgkϕ + kG(x) · A(x)kϕ
≤ kgkϕ + k1 kG(x)kϕ1 · kA(x)kϕ2
Z b
= kgkϕ + k1 |λ| · G0 · kxkϕ · k
K(·, s)f (s, x(s)) dskϕ2
a
≤ kgkϕ + 2k1 · |λ| · C · G0 · kxkϕ · kKkM
≤ kgkϕ + 2k1 · |λ| · C · G0 · r · kKkM
≤ kgkϕ + 2k1 · ρ · C · G0 · kKkM ≤ r
whenever kxkϕ ≤ r.
Then we have B : Br (Eϕ (I)) → Br (Eϕ (I)). Moreover, B is continuous on
Br (Eϕ (I)) (see the part I of the proof).
III. Let Qr stand for the subset of Br (Eϕ (I)) consisting of all functions which are
a.e. nondecreasing on I. Similarly as claimed in [17] this set is nonempty, bounded
(by r) and convex (direct calculation from the definition). It is also a closed set in
Lϕ (I).
Indeed, let (yn ) be a sequence of elements in Qr convergent in Lϕ (I) to y. Then
the sequence is convergent in measure and as a consequence of the Vitali convergence
theorem for Orlicz spaces and of the characterization of convergence in measure
(the Riesz theorem) we obtain the existence of a subsequence (ynk ) of (yn ) which
converges to y almost uniformly on I (cf. [100]). Moreover, y is still nondecreasing
a.e. on I which means that y ∈ Qr and so the set Qr is closed. Now, in view of
Lemma 1.4.1 the set Qr is compact in measure.
IV. Now, we will show, that B preserve the monotonicity of functions. Take
x ∈ Qr , then x is a.e. nondecreasing on I and consequently F (x) is also of the same
type in virtue of the assumption (C2) and Lemma 1.4.2. Further, A(x) = H ◦F (x) is
a.e. nondecreasing on I thanks for the assumption (K3). Since the pointwise product
of a.e. monotone functions is still of the same type and by (G1), the operator U is
a.e. nondecreasing on I.
72
Moreover, the assumption (C1) permits us to deduce that Bx(t) = g(t) + U(x)(t)
is also a.e. nondecreasing on I. This fact, together with the assertion that B :
Br (Eϕ (I)) → Br (Eϕ (I)) gives us that B is also a self-mapping of the set Qr . From
the above considerations it follows that B maps continuously Qr into Qr .
V. We will prove that B is a contraction with respect to the measure of noncompactness µ. Assume that X is a nonempty subset of Qr and let the fixed constant
> 0 be arbitrary. Then for an arbitrary x ∈ X and for a set D ⊂ I, meas(D) ≤ ε
we obtain
kB(x) · χD kϕ ≤ kgχD kϕ + kU(x) · χD kϕ
= kgχD kϕ + kG(x) · A(x)χD kϕ
≤ kgχD kϕ + k1 kG(x)χD kϕ1 · kA(x) · χD kϕ2
Z
= kgχD kϕ + k1 · |λ| · kG(x)χD kϕ1 · k
K(·, s)f (s, x(s)) dskϕ2
D
Z
≤ kgχD kϕ + k1 |λ|G0kxχD kϕ k
|K(·, s)|(b(s) + R(|x(s)|)) dskϕ2
D
≤ kgχD kϕ + k1 · |λ| · G0 · kxχD kϕ · 2kKkM k[bχD + R(r))]kN
≤ kgχD kϕ + 2k1 · |λ| · G0 · kxχD kϕ · kKkM [ kbχD kN + R(1)].
Hence, taking into account that g ∈ Eϕ , b ∈ EN
lim { sup [sup{kgχD kϕ }]} = 0 and lim { sup [sup{kbχD kN }]} = 0.
ε→0
ε→0
mes D≤ε x∈X
mes D≤ε x∈X
Thus by definition of c(x) and by taking the supremum over all x ∈ X and all
measurable subsets D with meas(D) ≤ ε we get
c(B(X)) ≤ 2k1 · |λ| · G0 · kKkM · R(1) · c(X).
Since X ⊂ Qr is a nonempty, bounded and compact in measure subset of an ideal
regular space Eϕ , we can use Proposition 5.5.1 and get
βH (B(X)) ≤ 2k1 · |λ| · G0 · kKkM · R(1) · βH (X).
The inequality obtained above together with the properties of the operator B and
the set Qr established before and the inequality from the Assumption (K2) allow us
to apply the Darbo Fixed Point Theorem 1.6.5, which completes the proof.
73
The case of N satisfying the ∆3-condition.
5.4.2
Let us consider the case of N-functions with the growth essentially more rapid than
a polynomial. In fact, we will consider N-functions satisfying ∆3 -condition. This
is very large and important class, especially from an application point of view (cf.
[36, 37, 106, 120]). An extensive description of this class can be found in [106,
Section 2.5]. Recall, that an N-function M determines the properties of the Orlicz
space LM (I) and then the less restrictive rate of the growth of this function implies
the ”worser” properties of the space. By ϑ we will denote the norm of the identity
operator from Lϕ (I) into L1 (I) i.e. sup{kxk1 : x ∈ B1 (Lϕ (I))}. For the discussion
about the existence of ϕ which satisfies our conditions see [84, p. 61].
Theorem 5.4.2. Assume, that ϕ, ϕ1 , ϕ2 are N-functions and that M and N are
complementary N-functions and that (N1), (C1), (C2), (C3), (G1), (K1) and (K3)
hold true. Moreover, put the following assumptions:
(C5)
1. N satisfies the ∆3 -condition,
2. K ∈ EM (I 2 ) and t → K(t, s) ∈ Eϕ2 (I) for a.e. s ∈ I,
3. There exist β, u0 > 0 such that
R(u) ≤ β
M(u)
,
u
for
u ≥ u0 ,
(K4) ϕ2 is an N-function satisfying
ZZ
ϕ2 (M(|K(t, s)|)) dtds < ∞
I2
and
2k1 · (2 + (b − a)(1 + ϕ2 (1))) · |λ| · G0 · kKkϕ2 ◦M · R(r0 ) < 1,
where
ω
1
− kbkN .
r0 =
ϑ 2|λ| · k1 · G0 · (2 + (b − a)(1 + ϕ2 (1))) · kKkϕ2 ◦M
Then there exist a number ρ > 0 and a number ̟ > 0 such that for all λ ∈ R
with |λ| < ρ and for all g ∈ Eϕ (I) with kgkϕ < ̟ there exists a solution x ∈ Eϕ (I)
of (5.9) which is a.e. nondecreasing on I.
Proof. We will indicate only the points of the proof if they differ from the previous
case.
I. In this case the operator B can be considered as continuous when acting on
the whole Eϕ (I).
74
By [84, Lemma 15.1 and Theorem 19.2] and the assumption (K4):
kA(x)χT kϕ2 ≤ 2·(2+(b−a)(1+ϕ2 (1)))·kK·χT ×I kϕ2 ◦M (kbkN + kR(|x(·)|)kN ) (5.12)
for arbitrary x ∈ Lϕ (I) and arbitrary measurable subset T of I.
Let us note, that the assumption (C5) 3. implies that there exist constants
ω, u0 > 0 and η > 1 such that N(ωR(u)) ≤ ηu for u ≥ u0 .
Thus for x ∈ Lϕ (I)
Z
1
1 + N(ωR(|x(s)|) ds
kR(|x(·)|)kN ≤
ω
I
Z
1
≤
1 + ηu0 (b − a) + η |x(s)| ds .
ω
I
The remaining estimations can be derived as in the first main theorem and then we
obtain, that A : Eϕ (I) → Eϕ2 (I), so by the properties of G we get B : Eϕ (I) →
Eϕ (I).
II. Put
ρ=
1
2 · k1 · G0 · (2 + (b − a)(1 + ϕ2 (1))) · kKkϕ2 ◦M · kbkN +
1
ω
(1 + ηu0(b − a))
Fix λ with |λ| < ρ.
Choose a positive number r in such a way that
kgkϕ
+
.
2rk1 · (2 + (b − a)(1 + ϕ2 (1))) · G0 · |λ| · kKkϕ2 ◦M (kbkN
1
+
(1 + ηu0(b − a) + ηϑr) ≤ r.
(5.13)
ω
As a domain for the operator B we will consider the ball Br (Eϕ (I)).
Let us remark, that the above inequality is of the form a + (b + vr)cr ≤ r with
a, b, c, v > 0. Then vc > 0 and if we assume that bc−1 < 0 and that the discriminant
is positive, then Viète’s formulas imply that the quadratic equation has two positive
solutions r1 < r2 for sufficiently small λ. By the definition of ρ it is clear, that our
assumptions guarantee the above requirements, so there exists a positive number r
satisfying this inequality.
Put C = (2 + (b − a)(1 + ϕ(1))). Let us note, in view of the above considerations,
that the assumption about the discriminant which implies the existence of solutions
for the above problem is of the form:
1
1
]2
(1 + ηu0 (b − a)) −
ω
2|λ| · k1 · G0 · C · kKkϕ2 ◦M
2kgkϕ ηϑ
× |λ| · k1 · G0 · C · kKkϕ2 ◦M >
ω
[kbkN +
75
i.e.
2
1
1
̟ = kbkN + (1 + ηu0(b − a)) −
ω
2|λ| · k1 · G0 · C · kKkϕ2 ◦M
|λ| · k1 · G0 · C · ωkKkϕ2 ◦M
×
.
2ηϑ
For x ∈ Br (Eϕ (I)) we have the following estimation:
kB(x)kϕ
≤
kgkϕ + kUxkϕ
kgkϕ + kG(x) · A(x)kϕ
=
kgkϕ + k1 kG(x)kϕ1 · kA(x)kϕ2
Z b
kgkϕ + k1 |λ|kG(x)kϕ1 · k
K(·, s)f (s, x(s)) dskϕ2
≤
=
a
≤
+
≤
+
≤
+
=
+
kgkϕ + 2k1 · C · G0 · |λ| · kxkϕ kKkϕ2 ◦M [kbkN
Z
1
(1 + N(ωR(u0 )) · (b − a) + η |x(s)| ds)
ω
I
kgkϕ + 2k1 · C · G0 · |λ| · kxkϕ kKkϕ2 ◦M [kbkN
1
(1 + N(ωR(u0 )) · (b − a) + ηkxk1 )
ω
kgkϕ + 2k1 · C · G0 · |λ| · kxkϕ kKkϕ2 ◦M [kbkN
1
(1 + N(ωR(u0 )) · (b − a) + ηϑkxkϕ )
ω
kgkϕ + 2rk1 · C · G0 · |λ| · kKkϕ2 ◦M (kbkN
1
(1 + ηu0 (b − a) + ηϑr) ≤ r.
ω
Then B : Br (Eϕ (I)) → Br (Eϕ (I)).
Note, that the parts III. and IV. of the previous proof are similar to those from
the first theorem, so we omit the details.
V. We will prove that B is a contraction with respect to a measure of noncompactness. Assume that X is a nonempty subset of Qr and let the fixed constant ε > 0
be arbitrary. Then for an arbitrary x ∈ X and for a set D ⊂ I, meas(D) ≤ ε
76
we obtain
kB(x) · χD kϕ
≤
=
≤
kgχD kϕ + kU(x) · χD kϕ
kgχD kϕ + kG(x) · A(x)χD kϕ
kgχD kϕ + k1 kG(x)χD kϕ1 · kA(x) · χD kϕ2
Z
=
kgχD kϕ + k1 · |λ| · G0 · kxχD kϕ · k
K(·, s)f (s, x(s)) dskϕ2
D
Z
≤
kgχD kϕ + k1 |λ|G0 kxχD kϕ k
|K(·, s)|(b(s) + R(|x(s)|)) dskϕ2
D
Z
≤
kgχD kϕ + k1 · |λ| · G0 · kxχD kϕ · k
|K(·, s)|b(s) dskϕ2
D
Z
+ k
|K(·, s)|R(|x(s)|)) dskϕ2
D
≤
kgχD kϕ + 2 · C · k1 · G0 · |λ| · kxχD kϕ · ϑ · kKkϕ2 ◦M kbχD kN
Z
+ 2 · C · k1 · G0 · kxχD kϕ · k
|K(·, s)|R(|x(s)|) dskϕ2
D
≤
≤
kgχD kϕ + 2Ck1 G0 |λ| · kxχD kϕ kKkϕ2 ◦M [kbχD kN + R(r)]
kgχD kϕ + 2Ck1 G0 |λ| · kxχD kϕ kKkϕ2 ◦M [kbχD kN + R(r0 )],
where
ω
1
− kbkN .
r0 =
ϑ 2|λ| · k1 · G0 · (2 + (b − a)(1 + ϕ2 )) · kKkϕ2 ◦M
Let us note, that r0 is an upper bound for solutions of (5.13).
Similarly as in the previous theorem we get
βH (B(X)) ≤ 2 · k1 · C · G0 · |λ|kKkϕ2 ◦M · R(r0 ) · βH (X).
The inequality obtained above together with the properties of the operator B and
the set Qr established before and then the assumption (K4) allow us to apply the
Theorem 1.6.5, which completes the proof.
We need to stress on some aspects of our results. First of all we can observe,
that our solutions are not necessarily continuous as in previously investigated cases.
In particular, we need not to assume, that the Hammerstein operator transforms
the space C(I) into itself. For the examples and conditions related to Hammerstein
operators in Orlicz spaces we refer the readers to [106, Chapter VI.6.1., Corollary 6
and Example 7].
We have two more information about the set of solutions: it is included in Eϕ (I)
and in view of Theorem 1.6.5 it can be proved, that this set is compact as a subset
of Lϕ (I).
77
5.4.3
Remarks on classes of solutions.
There is one more interesting question related to the case of continuous solutions.
This is the question if we are able to put in our results the same Orlicz space (as in
the case of C(I)). In other words, the case of Banach-Orlicz algebras. Note, that the
presented case in this Chapter seems to be more general we need to mention, that as
claimed by Kalton [77] this property is true for ϕ(x) = x(1 + x)−1 or ϕ(x) = log+ x.
These spaces are not of big interest in the theory of integral equations, then let us
present some remarks on operators satisfying our assumptions.
Let X, Y be ideal spaces. A superposition operator F : X → Y is called improving (cf. remark 1.3.3). The applications of such operators are based on the
observation that large classes of linear integral operators
Z
Hy(t) = λ
k(t, s)y(s)ds,
D
although not being compact, map sets with equiabsolutely continuous norms into
precompact sets. In contrast to the classical (non-quadratic) case, for quadratic
integral equations even such a nice assumption is not sufficient for using the Schauder
fixed point theorem.
Moreover, we assume in our main theorem, that G maps sets with equiabsolutely
continuous norms into the same family, but we don’t need to assume, that it is an
improving operator. Conversely, any improving operator G can be considered in our
results. Let us note, that for operators from Lebesgue spaces Lp (I) into Lr (I) (i.e.
Orlicz spaces with p(x) = xp and r(x) = xr , respectively), the characterization of
improving operators is known ([123]): a superposition operator F : Lp (I) → Lr (I)
is improving if and only if there exists a continuous and even function M satisfying
limu→∞ Mu(u) = ∞ and such that G(x)(t) = M(f (t, x(t))) is also an operator from
Lp (I) into Lr (I) (for an appropriate growth condition of f see [123]).
The aspect of applicability of our results deal also with the technique of Orlicz spaces for partial differential equations, so for an appropriate class of integral
equations. In this context one can consider more singular equations than in a classical case. Motivated by previously considered equations (see [36, 37, 45, 120] or
[100, 112]) we extend this method to the case of quadratic integral equations.
It should be recalled that our method of the proof can be also adapted to classical
equations considered in [83, 36, 45, 120]. For more information we refer the readers
to the Chapter IX ”Nonlinear PDEs and Orlicz spaces” in [105].
Finally, let us remark, that our results can be applied also for Lebesgue spaces
Lp (I) (p ≥ 1) (cf. [89, 90]). As mentioned above this class of spaces is also included
into the class of Orlicz spaces. But even in this case we allow for f or K to be strongly
78
nonlinear. The simplest case is that when F : L1 (I) → LN (I), H : LN (I) → Lp (I),
G : L1 (I) → Lq (I) ( 1p + 1q = 1). Thus we will have integrable solutions (in L1 (I)),
but f or K can be strongly nonlinear. Of course, strong nonlinearity of one of them
implies that we need to consider the weak one for another (cf. [83, Chapter IV.19]).
Let us present an example of such spaces. By N1 , N2 we denote complementary
functions for M1 , M2 , respectively. Put M1 (u) = exp |u| − |u| − 1 and M2 (u) =
u2
= N2 (u). Note, that M1 satisfies the ∆3 -condition. In this case N1 (u) = (1 +
2
|u|) · ln (1 + |u|) − |u|. If we define an N-function either as Ψ(u) = M2 [N1 (u)] or
Ψ(u) = N1 [M2 (u)], then by choosing arbitrary kernel K from the space LΨ (I) we
are able to apply [83, Theorem 15.4]. Thus H : LM1 (I) → LM2 (I) is continuous
and we may apply our result (Theorem 5.4.2) for operators G : L1 (I) → Lq (I) and
F : L1 (I) → LM1 (I) (with natural growth conditions, see Lemma 1.3.3).
Let us also to pay attention to the particular case of our problem with G(x) =
a(t)x(t):
Z 1
x(t) = g(t) + λa(t) · x(t)
K(t, s)f (s, x(s))ds.
0
Since we are motivated by some study on quadratic integral equations, this is of our
particular interest. Note, that a full description for acting and continuity conditions
for G(x) = a(t)x(t) can be found in [83, Theorem 18.2].
79
5.5
Conclusions and fixed point theorems.
If we are looking for the proofs of our main results, in contrast to earlier theorems
we stress on some properties of spaces rather on continuity of solutions and the
properties of this particular space C(I). We will present a general approach for
differential and integral problems by presenting a new fixed point theorem specialized
to quadratic equations.
For completness, we need to recall some necessary facts. In this section some
properties of function spaces play a major role. We need to consider the triples
of spaces with the following property: for a triple of spaces E, E1 , E2 there exists
a constant k such that for arbitrary x ∈ E1 and y ∈ E2 a product (pointwise
multiplication) x · y ∈ E and
kx · ykE ≤ k · kxkE1 · kykE2 .
Let us recall some special cases. Most known is the case of Banach algebras i.e.
the space of continuous functions. In this case E = E1 = E2 = C(I, X) (k = 1).
Moreover, some subalgebras of this space can be interesting. If we try to consider
”bigger” spaces we need to go outside the class of Banach algebras.
For discontinuous functions, let us recall the Hölder inequality for Lebesgue
spaces: kx · ykL1 ≤ kxkLp · kykLq whenever 1p + 1q = 1. Thus, a triple (L1 , Lp , Lq ) is
good enough.
Third important example (and most important) is for Orlicz spaces (for definitions see [83, 92], for instance). Generally speaking, the product of two functions
x, y ∈ LM (I) is not in LM (I). However, if x and y belongs to some particular Orlicz
spaces, then the product x · y belong to a third Orlicz space. Let us note, that
one can find two functions belonging to Orlicz spaces: u ∈ LU (I) and v ∈ LV (I)
such that the product uv does not belong to any Orlicz space (this product is not
integrable). Nevertheless, Lemma 1.3.8 give us an interesting characterization for
such a triple of spaces. An interesting discussion about necessary and sufficient
conditions for product operators can be found in [83, 92]. Note, that since Lp = LM
p
for M(t) = tp the case of Lebesgue spaces Lp is included in the mentioned Lemma.
Finally we have a special case for E2 = L∞ and some function spaces for which
(E = E1 ) kx·ykE ≤ kxkE ·kykL∞ . The class of spaces with this property is known as
preideal∗ spaces (cf. [119, p. 66] or [118]) or Köthe function spaces. Although this
case seems to be general it has one weakness from our point of view: the measure
of noncompactness in L∞ seems to be inapplicable and we do not discuss it in this
Thesis.
It is possible to check this property for a given triple of spaces. An open question
is if is possible to characterize all such spaces?
80
Recall, that for any ε > 0, let c be a measure of equiintegrability of the set X
in a ideal space E (introduced in [9], cf. also [119, Definition 3.9], [67, 66]):
c(X) = lim sup
ε→0
sup sup kx · χD kE ,
mesD≤ε x∈X
where χD denotes the characteristic function of D. To distinguish between measures
of noncompactness µ (or : c) in different spaces we will indicate an appropriate space
as an index i.e. µE , cE1 , µE2 etc.
The following theorem clarify the connections between the two coefficients in E.
Proposition 5.5.1. ([66, Theorem 1]) Let X be a nonempty, bounded and compact
in measure subset of an ideal regular space E. Then
βH (X) = c(X).
As a consequence, we obtain that bounded sets which are additionally compact
in measure are compact in E iff they are equiintegrable in this space (i.e. have
equiabsolutely continuous norms, in particular when X is a subset for a regular part
of E).
In contrast to the case considered in [25] we will need the following property:
Lemma 5.5.1. Assume, that for a triple of regular ideal function spaces E, E1 , E2
there exists a constant k such that for arbitrary x ∈ E1 , y ∈ E2 and t ∈ I a product
x · y ∈ E and kx · ykE ≤ k · kxkE1 · kykE2 . Then for any set X ⊂ E1 , Y ⊂ E2 we
have X · Y ⊂ E and
cE (X · Y ) ≤ k · cE1 (X) · cE2 (Y ),
where cV stands for a measure of equiintegrability in the space V for V = E, E1 or
E2 , respectively.
Proof. By the properties of spaces we obtain, that X · Y ⊂ E. Take arbitrary x ∈ X
and y ∈ Y and arbitrary measurable subset D of I. Then
k(x · y) · χD kE ≤ k · kx · χD kE1 · ky · χD kE2 .
Then by the properties of supremum
sup sup k(x · y) · χD kE ≤ k · sup sup kx · χD kE1 · ky · χD kE2
x∈X y∈Y
x∈X y∈Y
and by the property of lim supmeas(D)→0
lim sup sup sup k(x·y)·χD kE ≤ k · lim sup sup kx·χD kE1 · lim sup sup ky ·χD kE2 ,
meas(D)→0 x∈X y∈Y
meas(D)→0 x∈X
meas(D)→0 y∈Y
which ends the proof.
The last property is obvious but useful:
Lemma 5.5.2. Let E be a regular ideal space. For any bounded subset X of E we
have c(X) ≤ kXkE .
81
5.5.1
A fixed point theorem.
Since we are interested on a fixed points of some product operators we will assume,
that our operators have values in some intermediate spaces and then the product
will again turn the values into the original space.
First, let us apply our approach for the most applicable theorem of this type.
Consider an arbitrary (in the sense of Definition 1.5.1) measure of noncompactness
µ in C(I, E). An interesting fixed point theorem in Banach algebras was proved by
Banaś and Lecko (cf. [25]). Let us consider different spaces of continuous functions
with a suitable choice of measures of noncompactness µE on E = C(I), µE1 on E1
and µE2 on E2 . By using our approach we are able to present the following extension
of the mentioned theorem:
Theorem 5.5.1. Let E, E1 , E2 be regular ideal function spaces. Assume that T
is nonempty, bounded, closed, and convex subset of the Banach space E, and the
operators A : E → E1 and B : E → E2 . Moreover, assume:
1. (A1) A transform continuously the set T into T1 ⊂ E1 and A(T ) is bounded
in E1 ,
2. (A2) there exists a constant k1 > 0 such that A satisfies an inequality:
µE (A(U)) ≤ k1 · µE1 (U)
for arbitrary bounded subset U of E,
3. (B1) B transform continuously the set T into T2 ⊂ E2 and B(T ) is bounded
in E2 ,
4. (B2) there exists a constant k2 > 0 such that B satisfies an inequality:
µE (B(U)) ≤ k2 · µE2 (U)
for arbitrary bounded subset U of E,
5. (E1) for a triple of spaces E, E1 , E2 there exists a constant k such that for
arbitrary x ∈ E1 , y ∈ E2 and t ∈ I a product x · y ∈ E and kx · ykE ≤
k · kxkE1 · kykE2 ,
6. (E2) for every x ∈ T1 and y ∈ T2 one has x · y ∈ T ,
7. (C) kA(T )kE1 · k2 + kB(T )kE2 · k1 < 1.
Then there exists at least one fixed point for the operator K = A · B in the set T
and that the set of all fixed points of K belongs to the kernel kerµE .
82
This theorem was proved by Banaś in a special case of Banach algebras E =
E1 = E2 = C(I, R) (k = 1) (cf. also Dhage and Kumpulainen [53], for instance).
We do not require, that the values of all operators are from the same space, but
by using a property of considered spaces we are able to repeat the proof, so we omit
the details. We will present a proof for a more general case. In the first theorem the
Ascoli criterion of compactness in spaces of continuous functions simplify the proof,
because the convergence of sequences is directly related with pointwise convergence.
Now, we will consider the case of functions spaces without such a nice property.
We will consider some subspaces of a space of L0 (I) of measurable functions, bigger
than C(I). This allows us to apply the fixed point theorem for the problems with
discontinuous solutions. This proof will be based on different compactness criterion
(the Dunford-Pettis theorem and the Erzakova theorem).
Theorem 5.5.2. Assume that T is nonempty, bounded, closed, convex and compact
in measure subset of a regular ideal function space E, and the operators A : E → E1
and B : E → E2 . Put the following set of assumptions:
1. (A1) A transform continuously the set T into T1 ⊂ E1 and A(T ) is bounded
in E1 ,
2. (A2) there exists a constant k1 > 0 such that A satisfies an inequality:
c(A(U)) ≤ k1 · cE1 (U)
for arbitrary bounded subset U of E,
3. (B1) B transform continuously the set T into T2 ⊂ E2 and B(T ) is bounded
in E2 ,
4. (B2) there exists a constant k2 > 0 such that B satisfies an inequality:
c(B(U)) ≤ k2 · cE2 (U)
for arbitrary bounded subset U of E,
5. (E1) for a triple of regular ideal spaces E, E1 , E2 there exists a constant k
such that for arbitrary x ∈ E1 and y ∈ E2 a product x · y ∈ E and kx · ykE ≤
k · kxkE1 · kykE2 ,
6. (E2) for every x ∈ T1 and y ∈ T2 one has x · y ∈ T ,
7. (C1) k · k1 · kB(T )kE2 < 1,
8. (C2) k · kA(T )kE1 · k2 < 1.
83
Assume, that (A1), (B1), (E1), (E2) and either (A2) and (C1) or (B2) and (C2)
are satisfied. Then there exists at least one fixed point for the operator K = A · B
in the set T and the set of all fixed points F ixK is relatively compact in E.
Proof. Let us present the proof, when (A2) and (C1) are satisfied. The second case
is similar.
It is obvious that the operator K is well-defined on T and by (E2) it acts between
T into itself.
Denote M1 = supt∈T kA(t)kE1 and M2 = supt∈T kB(t)kE2 . Let (xn ) be an arbitrary sequence in T tending to x ∈ T . Then
kK(xn ) − K(x)kE
=
≤
≤
≤
kA(xn ) · B(xn ) − A(x) · B(x)kE
kA(xn ) · B(xn ) − A(x) · B(x) − A(x) · B(xn )
+ A(x) · B(xn )kE
k(A(xn ) − A(x)) · B(xn )kE + k(B(xn ) − B(x)) · A(x)kE
k · kA(xn ) − A(x)kE1 · M2 + k · kB(xn ) − B(x)kE2 · M1 .
From our assumptions it follows that K is continuous from T into E.
Now, we will investigate the contraction property for a measure c(X).
Assume that X is a nonempty subset of T and let the fixed constant ε > 0 be
arbitrary. Then for an arbitrary x ∈ X and for a set D ⊂ I, meas(D) ≤ ε we obtain
kK(x) · χD kE ≤ kkA(x) · χD kE1 · kB(x)kE2 .
Since for any non-negative real-valued functions f and g we have supI (f · g) ≤
supI f · supI g, by definition of c(x) and by taking the supremum over all x ∈ X and
all measurable subsets D with meas(D) ≤ ε we get
c(K(X)) ≤ k · k1 · kB(T )kE2 · c(X).
Because X ⊂ T is a nonempty, bounded and compact in measure subset of an ideal
regular space E, we can use Proposition 5.5.1 and get
βH (K(X)) ≤ k · k1 · kB(T )kE2 · βH (X).
The inequality obtained above together with the properties of the operator K and
the set T established before, allow us to apply the classical Darbo fixed point theorem
for βH . If we suppose, that βH (F ixK) 6= 0, then K = F ixK implies βH (F ixK) =
βH (K) < βH (K), a contradiction, which completes the proof.
Remark 5.5.1. We need to remark, that one of our assumptions can be easily
relaxed. We assume, that the space is regular. Denote by 0 a regular part of E. It is
84
sufficient to assume, that K : T ∩ E0 → T ∩ E0 . This seems to be important for the
case of so-called improving operators (taking bounded subsets of E into the sets with
equiabsolutely continuous norms i.e. into E0 ). A detailed theory of compactness in
regular ideal spaces can be found in [118]. If an ideal space E has nontrivial regular
part, then our result applies for any operator which is measure-compact (see [68]).
85
Bibliography
[1] R.P Agarwal, D. O’Regan, P. Wong, Constant-sign solutions of a system of
Volterra integral equations in Orlicz spaces, J. Integral Equations Appl. 20
(2008), 337–378.
[2] R.P Agarwal, D. O’Regan, P. Wong, Solutions of a system of integral equations
in Orlicz spaces, J. Integral Equations Appl. 21 (2009), 469–498.
[3] A. Aghajani, Y. Jalilian, Existence and global attractivity of solutions of a nonlinear functional integral equation, Commun. Nonlinear Sci. Numer. Simulat.
15 (2010), 3306–3312.
[4] A. Aghajani, J. Banaś and Y. Jalilian, Existence of solutions for a class of
nonlinear Volterra singular integral equation, Comp. Math. Appl. 61 (2011),
1215–1227.
[5] J. Alexopoulos, De La Vallée Poussin’s theorem and weakly compact sets in
Orlicz spaces, Quaestiones Mathematicae 17 (1994), 231–248.
[6] G. Anichini, G. Conti, Existence of solutions of some quadratic integral equations, Opuscula Mathematica 28 (2008), 433–440.
[7] J. Appell, The importance of being Orlicz, Banach Center Publ. 64 (2004),
21–28.
[8] J. Appell, O.W. Diallo and P.P. Zabrejko, On linear integro-differential equations of Barabashin type in spaces of continuous and measurable functions, Jour.
Integral Equat. Appl. 1 (1988), 227–247.
[9] J. Appell, E. De Pascale, Su alcuni parametri connessi con la misura di non
compattezza di Hausdorff in spazi di funzioni misurabili, Boll. Un. Mat. Ital. 6,
3-B (1984), 497-515.
[10] J. Appell, P.P. Zabrejko, Nonlinear Superposition Operators, Cambridge Tracts
in Mathematics 95, Cambridge University Press, Cambridge, 1990.
86
[11] I.K. Argyros, Quadratic equations and applications to Chandrasekhar and related equations, Bull. Austral. Math. Soc. 32 (1985), 275-292.
[12] I.K. Argyros, On a class of quadratic integral equations with perturbations,
Functiones et Approximatio 20 (1992), 51–63.
[13] K. Balachandran, S. Ilamaran, A note on integrable solutions of Hammerstein
integral equations, Proc. Indian Acad. Sci. (Math. Sci.) 105 (1995), 99–103.
[14] J. Banaś, An existence theorem for nonlinear Volterra integral equation with
deviating argument, Rend. Circ. Mat. Palermo 25 (1986), 82–89.
[15] J. Banaś, Applications of measure of weak noncompactness and some classes of
operators in the theory of functional equations in the Lebesgue sense, Nonlin.
Anal. 30 (1997), 3283–3293.
[16] J. Banaś, Existence results for Volterra-Stjeltjes quadratic integral equations on
an unbounded interval, Math. Scand. 98 (2006), 143–160.
[17] J. Banaś, Integrable solutions of Hammerstein and Urysohn integral equations,
J. Austral. Math. Soc. 46 (1989), 61–68.
[18] J. Banaś, On the superposition operator and integrable solutions of some functional equations, Nonlinear Anal. 12 (1988), 777–784.
[19] J. Banaś, A. Chlebowicz, On existence of integrable solutions of a functional
integral equation under Carathéodory conditions, Nonlin. Anal. 70 (2009), 3172–
3179.
[20] J. Banaś, A. Chlebowicz On integrable solutions of a nonlinear Volterra integral
equation under Carathéodory conditions, Bull. London Math. Soc. 41 (2009),
1073–1084.
[21] J. Banaś, B.C. Dhage, Global asymptotic stability of solutions of a functional
integral equation, Nonlin. Anal. 69 (2008), 949–952.
[22] J. Banaś, W.G. El-Sayed, Measures of noncompactness and solvability of an
integral equation in the class of functions of locally bounded variation, J. Math.
Anal. Appl., 167 (1992), 133–151.
[23] J. Banaś, K. Goebel, Measures of Noncompactness in Banach Spaces, Lect.
Notes in Math. 60, M. Dekker, New York - Basel, 1980.
[24] J. Banaś, Z. Knap, Integrable solutions of a functional-integral equation, Rev.
Mat. Univ. Compl. Madrid 2 (1989), 31–38.
87
[25] J. Banaś, M. Lecko, Fixed points of the product of operators in Banach algebra,
Panamerican Mathematical Journal 12 (2002), 101–109.
[26] J. Banaś, M. Lecko and W.G. El-Sayed, Existence theorems for some quadratic
integral equations, Jour. Math. Anal. Appl. 222 (1998), 276–285.
[27] J. Banaś, A. Martinon, Monotonic solutions of a quadratic integral equation of
Volterra type, Comp. Math. Appl. 47 (2004), 271–279 .
[28] J. Banaś, L. Olszowy, Measures of noncompactness related to monotonicity,
Comment. Math. Prace Matem. 41 (2001), 13–23.
[29] J. Banaś, J. Rivero, On measures of weak noncompactness, Ann. Mat. Pura
Appl. 151 (1988), 213–224.
[30] J. Banaś, J. Rocha, K.B. Sadarangani Solvability of a nonlinear integral equation of Volterra type, Jour. Comp. Appl. Math. 157 (2003), 31–48.
[31] J. Banaś, B. Rzepka, On existence and asymptotic stability of a nonlinear integral equation, J. Math. Anal. Appl. 284 (2003), 165–173.
[32] J. Banaś, B. Rzepka, Monotonic solutions of a quadratic integral equation of
fractional order, J. Math. Anal. Appl. 332 (2007) 1371–1379.
[33] J. Banaś, B. Rzepka, Nondecreasing solutions of a quadratic singular Volterra
integral equation, Math. Comp. Model. 49 (2009) 488–496.
[34] J. Banaś, K. Sadarangani, Solutions od some functional-integral equations in
Banach algebra, Math. Comp. Model. 38 (2003), 245–250.
[35] J. Banaś, T. Zaja̧c, Solvability of a functional integral equation of fractional
order in the class of functions having limits at infinity, Nonlin. Anal. 71 (2009),
5491–5500.
[36] A. Benkirane, A. Elmahi, An existence theorem for a strongly nonlinear elliptic
problem in Orlicz spaces, Nonlin. Anal. 36 (1999), 11–24.
[37] J. Berger, J. Robert, Strongly nonlinear equations of Hammerstein type, J.
London Math. Soc. 15 (1977), 277–287.
[38] P.J. Bushell, On the class of Volterra and Fredholm nonlinear integral equations,
Math Proc. Cambridge Philos. Soc. 79 (1976), 329–335.
[39] H. Brézis, F. Browder, Existence theorems for nonlinear integral equations of
Hammerstein type, Bull. Amer. Math. Soc. 81 (1975), 73–78.
88
[40] J. Caballero, A.B. Mingarelli and K. Sadarangani, Existence of solutions of
an integral equation of Chandrasekhar type in the theory of radiative transfer,
Electr. Jour. Differ. Equat. 57 (2006), 1–11.
[41] J. Caballero, D. O’Regan, K. Sadarangani, On monotonic solutions of some
integral equations, Archivum Math (Brno) 41 (2005), 325–338.
[42] B. Cahlon, M. Eskin, Existence theorems for an integral equation of the Chandrasekhar H-equation with perturbation, Jour. Math. Anal. Appl. 83 (1981),
159–171.
[43] S. Chandrasekhar, Radiative Transfer, Dover Publications, New York, 1960.
[44] S. Chandrasekhar, The transfer of radiation in stellar atmospheres, Bull. Amer.
Math. Soc. 53 (1947), 641–711.
[45] I.-Y. S. Cheng, J.J. Kozak, Application of the theory of Orlicz spaces to statistical mechanics. I. Integral equations, J. Math. Phys. 13 (1972), 51–58.
[46] M. Cichoń, M. Metwali, Monotonic solutions for quadratic integral equations,
Discuss. Math. Diff. Incl. 31 (2011), 157–171.
[47] M. Cichoń, M. Metwali, On quadratic integral equations in Orlicz spaces, Jour.
Math. Anal. Appl. 387 (2012), 419–432.
[48] M. Cichoń, M. Metwali, On monotonic integrable solutions for quadratic functional integral equations, Mediterranean Jour. Math. 10 (2013), 909–926.
[49] M.M. Crum, On an integral equation of Chandrasekhar, Quarterly Jour. Math.
18 (1947), 244–252.
[50] G. Darbo, Punti uniti in trasformazioni a codominio noncompatto, Rend. Sem.
Mat. Univ. Padova 24 (1955), 84–92.
[51] K. Deimling, Nonlinear Functional Analysis, Springer, Berlin, 1985.
[52] F.S. De Blasi, On a property of the unit sphere in Banach spaces, Bull. Math.
Soc. Sci. Math. R. S. Roumanie 21 (1977), 259–262.
[53] B.C. Dhage, M. Kumpulainen, Nonlinear functional boundary value problems
involving the product of two nonlinearities, Appl. Math. Letters 21 (2008), 537–
544.
[54] M. Diana Julie, K. Balachandran, On solutions of a Volterra integral equation
with deviating arguments, Electr. Jour. Diff. Equat. 57 (2009), 1–8.
89
[55] J. Dieudonné, Sur les espaces de Köthe, J. Anal. Math. (1951), 81–115.
[56] D. Dinculeanu, Vector Measures, Pergamon Press, 1967.
[57] S. Djebali, Z. Sahnoun, Nonlinear alternatives of Schauder and Krasnoselskij
types with applications to Hammerstein integral equations in L1 spaces, Jour.
Diff. Equat. 249 (2010), 2061–2075.
[58] H. Domke, An equivalence theorem for Chandrasekhar’s H-functions and its
application for accelerating convergence, J. Quant. Spectrosc. Radiat. Transfer
39 (1988), 283–286.
[59] J. Dugundji, A. Granas, Fixed Point Theory, Monografie Mathematyczne,
PWN, Warszawa, 1963.
[60] N. Dunford, J.T. Schwartz, Linear Operators. Part I: General Theory, Interscience Publishers, New York, 1958.
[61] M.M. El Borai, W.G. El-Sayed and M.I. Abbas, Monotonic solutions of a class
of quadratic singular integral equations of Volterra type, Int. J. Contemp. Math.
Sci. 2 (2007), 89–102.
[62] W.G. El-Sayed, B. Rzepka, Nondecreasing solutions of a quadratic integral equation of Urysohn type, Comp. Math. Appl. 51 (2006), 1065–1074.
[63] G. Emmanuele, About the existence of integrable solutions of functional-integral
equation, Rev. Mat. Univ. Complut.Madrid 4 (1991), 65–69.
[64] G. Emmanuele, Integrable solutions of a functional-integral equation, J. Integr.
Equat. Appl. 4 (1992), 89–94.
[65] G. Emmanuele, Measure of noncompactness and fixed point theorem , Bull.
Math. Soc. Math. R.S. Roum. 25 (1981), 253–285.
[66] N. Erzakova, Compactness in measure and measure of noncompactness,
Siberian Math. J. 38 (1997), 926–928.
[67] N. Erzakova, On measures of non-compactness in regular spaces, Z. Anal. Anwendungen 15 (1996), 299–307.
[68] N. Erzakova, On measure-compact operators, Russian Mathematics 55 (2011),
37–42.
[69] Ch. Fox, A solution of Chandrasekhar’s integral equation, Trans. Amer. Math.
Soc. 99 (1961), 285–291.
90
[70] R. Gorenflo, S. Vessela, Abel Integral Equations, Lect. Notes Math. 1461,
Springer, Berlin-Heidelberg, 1991.
[71] K. Goebel, W. A. Kirk, Topics in Metric Fixed Point Theory, Cambridge University Press, Cambridge, 1990.
[72] G. Gripenberg, Periodic solutions of an epidemic model, Jour. Mathematical
Biology 10 (1980), 271–280.
[73] F. Hartung, T. Krisztin, H.-O. Walther and J. Wu, Functional differential equations with state-dependent delays: theory and applications, in: Handbook of
Differential Equations. Ordinary Differential Equations vol. 3, A. Canada, P.
Drábek, A Fonda Eds., Elsevier, 2006.
[74] S.L. Hollis, C.T. Kelley, Vector algorithms for H-equations arising in radiative transfer through inhomogeneous media, Transport Theory and Statistical
Physics 15 (1986), 33–48.
[75] S. Hu, M. Khavani, W. Zhuang, Integral equations arising in the kinetic theory
of gases, Appl. Analysis 34 (1989), 261–266.
[76] G. Infante, J.R.L. Webb, Nonzero solutions of Hammerstein integral equations
with discontinuous kernels, J. Math. Anal. Appl. 272 (2002), 30–42.
[77] N.J. Kalton, Subalgebras of Orlicz spaces and related algebras of analytic functions, Arkiv för Matematik 18 (1980), 223–254.
[78] L.V. Kantorovich, G.P. Akilov, Functional Analysis, Pergamon Press, Oxford,
1982.
[79] A. Karoui, A. Jawahdou, Existence and approximate Lp and continuous solutions of nonlinear integral equations of the Hammerstein and Volterra types,
Appl. Math. Comput. 216 (2010), 2077–2091.
[80] A.N. Kolmogorov, S.V. Fomin, Introduction for Real Analysis, Prentice-Hall
Inc., 1970.
[81] M.A. Krasnoselskii, On the continuity of the operator F u(x) = f (x, u(x)), Dokl.
Akad. Nauk 77 (1951), 185–188.
[82] M.A. Krasnoselskii, Two remarks on the method of successive approximations,
Uspehi. Mat. Nauk. 10 (1955), 123–127.
[83] M.A. Krasnoselskii, Yu. Rutitskii, Convex Functions and Orlicz Spaces,
Gröningen, 1961.
91
[84] M.A. Krasnoselskii, P.P. Zabrejko, E.I. Pustylnik and P.E. Sobolevskii, Integral
Operators in Spaces of Summable Functions, Nauka, Moscow, 1966 (English
translation: Noordhoff, Leyden, 1976.
[85] J. Krzyż, On monotonicity-preserving transformations, Ann. UMCS 6 (1952),
91–111.
[86] M. Kunze, On a special class of nonlinear integral equations, Jour. Integr.
Equat. Appl. 7 (1995), 329–350.
[87] K. Latrach, M. A. Taoudi, Existence results for a generalized nonlinear Hammerstein equation on L1 spaces, Nonlin. Anal. 66 (2007), 2325–2333.
[88] A. Leonard, T.W. Mullikin, Integral equations with difference kernels on finite
intervals, Trans. Amer. Math. Soc. 116 (1965), 465–473.
[89] K. Maleknejad, R. Mollapourasl and K. Nouri, Study on existence of solutions
for some nonlinear functional-integral equations, Nonlin. Anal. 69 (2008), 2582–
2588.
[90] K. Maleknejad, K. Nouri, R. Mollapourasl, Existence of solutions for some nonlinear integral equations, Commun. Nonlinear Sci. Numer. Simulat. 14 (2009),
2559–2564.
[91] K. Maleknejad, K. Nouri, R. Mollapourasl, Investigation on the existence of
solutions for some nonlinear functional-integral equations, Nonlin. Anal. 71
(2009), 1575–1578.
[92] L. Maligranda, Orlicz Spaces and Interpolation, Campinas SP Brazil: Departamento de Matemática, Universidade Estadual de Campinas, 1989.
[93] M. Meehan, D. O’Regan, Positive Lp solutions of Hammerstein integral equations, Arch. Math. 76 (2001), 366–376.
[94] T.W. Mullikin, Radiative transfer in finite homogeneous atmospheres with
anisotropic scattering. I. Linear singular equations, The Astrophysical Journal 139 (1964), 379–395.
[95] T.W. Mullikin, A nonlinear integrodifferential equation in radiative transfer,
Journal of the Society for Industrial and Applied Mathematics 13 (1965), 388–
410.
[96] R. Nussbaum, A quadratic integral equation, Annali della Scuola Normale Superiore di Pisa-Classe di Scienze 7 (1980), 375–480.
92
[97] W. Okrasiński, Nonlinear Volterra integral equations and physical applications,
Extracta Math. 4 (1989), 51–80.
[98] D. O’Regan, Solutions in Orlicz spaces to Urysohn integral equations, Proceedings of the Royal Irish Academy, Section A 96 (1996), 67–78.
[99] D. O’Regan, M. Meehan, Existence Theory for Nonlinear Integral and Integrodifferential Equations, Kluwer Academic Publishers, Dordrecht, 1998.
[100] W. Orlicz, S. Szufla, On some classes of nonlinear Volterra integral equations
in Banach spaces, Bull. Acad. Polon. Sci. Sér. Sci. Math. 30 (1982), 239–250.
[101] G.H. Pimbley, Positive solutions of a quadratic integral equation, Archive for
Rational Mechanics and Analysis 24 (1967), 107–127.
[102] R. Pluciennik, The superposition operator in Musielak-Orlicz spaces of vectorvalued functions, Proceedings of the 14th winter school on abstract analysis
(Srnı́, 1986), Rend. Circ. Mat. Palermo (2) Suppl. No. 14 (1987), 411–417.
[103] R. Pluciennik, S. Szufla, Nonlinear Volterra integral equations in Orlicz spaces,
Demonstratio Math. 17 (1984), 515–532.
[104] R. Ramalho, Existence and uniqueness theorems for a nonlinear integral equation, Mathematische Annalen 221 (1976), 35–44.
[105] M.M. Rao, Z.D. Ren, Applications Of Orlicz Spaces, Marcel Dekker, New
York, 2002.
[106] M.M. Rao, Z.D. Ren, Theory of Orlicz Spaces, Marcel Dekker, New York,
1991.
[107] I.A. Rus, An abstract point of view for some integral equations from applied
mathematics, in: Proc. od the International Conference: Analysis and Numerical Computation of Solutions of Nonlinear Systems Modelling Physical Phenomena, Especially: Nonlinear Optics, Inverse Problems, Mathematical Material
Sciences and Theoretical Fluid Mechanics, Timişoara, 1997. pp. 256-270.
[108] N. Salhi, M.A. Taoudi, Existence of integrable solutions of an integral equation
of Hammerstein type on an unbounded interval, Mediterr. J. Math. (2011) DOI
10.1007/s00009-011-0147-3.
[109] G. Scorza Dragoni, Un teorema sulle funzioni continue rispetto ad une e misarubili rispetto ad un altra variable, Rend. Sem. Mat. Univ. Padova 17 (1948),
102–106.
93
[110] I.V. Shragin, The boundedness of the Nemytskii operator in Orlicz spaces,
Kišinev. Gos. Univ. Učen. Zap. 50 (1962), 119–122.
[111] M.B. Schillings, Existence theorems for generalized Hammerstein equations, J.
Functional Analysis 23 (1976), 177–194.
[112] A. Soltysiak, S. Szufla, Existence theorems for Lϕ -solutions of the Hammerstein integral equation in Banach spaces, Comment. Math. Prace Mat. 30
(1990), 177–190.
[113] P.N. Srikanth, M.C. Joshi, Existence theorems for generalized Hammerstein
equations, Proc. Amer. Math. Soc. 78 (1980), 369–374.
[114] Ch.A. Stuart, Existence theorems for a class of non-linear integral equations,
Mathematische Zeitschrift 137 (1974), 49–66.
[115] M. A. Taoudi, Integrable solutions of a nonlinear functional integral equation
on an unbounded interval, Nonlin. Anal. 71 (2009), 4131–4136.
[116] M. Väth, A general theorem on continuity and compactness of the Urysohn
operator, J. Integral Equations Appl. 8 (1996), 379–389.
[117] M. Väth, Continuity of single- and multivalued superposition operators in generalized ideal spaces of measurable functions, Nonlin. Funct. Anal. Appl. 11
(2006), 607–646.
[118] M. Väth, Ideal Spaces, Lect. Notes Math. 1664, Springer, Berlin-Heidelberg,
1997.
[119] M. Väth, Volterra and Integral Equations of Vector Functions, Marcel Dekker,
New York-Basel, 2000.
[120] J.D. Weeks, S.A. Rice, J.J. Kozak, Analytic approach to the theory of phase
transitions, J. Chem. Phys. 52 (1970), 2416–2426.
[121] M. Wertheim, Exact solution of the mean spherical model for fluids of hard
spheres with permanent electric dipole moments, The Journal of Chemical
Physics 55 (1971), 4291–4298.
[122] P.P. Zabrejko, A.I. Koshlev, M.A. Krasnoselskii, S.G. Mikhlin, L.S.
Rakovshchik, V.J. Stecenko, Integral Equations, Noordhoff, Leyden, 1975.
[123] P.P. Zabrejko, E.I. Pustyl’nik, On continuity and complete continuity of nonlinear integral operators in Lp spaces, Uspekhi Mat. Nauk 19 (116) (1964),
204–205.
94
[124] H. Zhu, On a Nonlinear Integral Equation with Contractive Perturbation, Advances in Difference Equations, Vol. (2011), Article ID 154742, 10 pages.
95