(Chapman & Hall_CRC Monographs and Research Notes in Mathematics) Svetlin G. Georgiev, Khaled Zennir - Multiple Fixed-Point Theorems and Applications in the Theory of ODEs, FDEs and PDEs-CRC Press (20

Multiple Fixed-Point
Theorems and Applications

in the Theory of ODEs,
FDEs and PDEs
Monographs and Research Notes in Mathema!cs
Series Editors:
John A. Burns, Thomas J. Tucker, Miklos Bona, Michael Ruzhansky
About the Series

This series is designed to capture new developments and summarize what is known over the
en!re field of mathematics, both pure and applied. It will include a broad range of monographs
and research notes on current and developing topics that will appeal to academics, graduate
students, and prac!!oners. Interdisciplinary books appealing not only to the mathema!cal
community, but also to engineers, physicists, and computer scien!sts are encouraged.
This series will maintain the highest editorial standards, publishing well-developed monographs
as well as research notes on new topics that are final, but not yet refined into a formal
monograph. The notes are meant to be a rapid means of publica!on for current material where
the style of exposi!on reflects a developing topic.
Glider Representa!ons
Frederik Caenepeel, Fred Van Oystaeyen
La"ce Point Iden!!es and Shannon-Type Sampling

Willi Freeden, M. Zuhair Nashed
Summable Spaces and Their Duals, Matrix Transforma!ons and Geometric Proper!es
Feyzi Basar, Hemen Du!a
Spectral Geometry of Par!al Differen!al Operators (Open Access)

Michael Ruzhansky, Makhmud Sadybekov, Durvudkhan Suragan
Linear Groups: The Accent on Infinite Dimensionality

Martyn Russel Dixon, Leonard A. Kurdachenko, Igor Yakov Subbo"n
Morrey Spaces: Introduc!on and Applica!ons to Integral Operators and PDE’s, Volume I
Yoshihiro Sawano, Giuseppe Di Fazio, Denny Ivanal Hakim
Morrey Spaces: Introduc!on and Applica!ons to Integral Operators and PDE’s, Volume II
Yoshihiro Sawano, Giuseppe Di Fazio, Denny Ivanal Hakim
Mul!ple Fixed-Point Theorems and Applica!ons in the Theory of ODEs, FDEs and PDEs
Svetlin G. Georgiev and Khaled Zennir
For more informa!on about this series please visit: h$ps://www.crcpress.com/Chapman--HallCRC-Monographs-and-Research-Notes-in-Mathema!cs/book-

series/CRCMONRESNOT
Multiple Fixed-Point
Theorems and Applications
in the Theory of ODEs,
FDEs and PDEs
Svetlin G. Georgiev
Khaled Zennir
First edition published 2020
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
and by CRC Press

2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

c 2020 Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers
have attempted to trace the copyright holders of all material reproduced in this publication and apologize to
copyright holders if permission to publish in this form has not been obtained. If any copyright material has not
been acknowledged please write and let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or
utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including
photocopying, microfilming, and recording, or in any information storage or retrieval system, without written
permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.com or contact
the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For
works that are not available on CCC please contact mpkbookspermissions@tandf.co.uk
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only
for identification and explanation without intent to infringe.
ISBN: 978-0-367-46432-5 (hbk)

ISBN: 978-1-003-02872-7 (ebk)
Typeset in Nimbus
by Nova Techset Private Limited, Bengaluru & Chennai, India
Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com
and the CRC Press Web site at

http://www.crcpress.com
Contents
Preface vii
1 Fixed-Point Index Theory. Multiple Fixed-Point Theorems 1

1.1 Measures of Noncompactness . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Brouwer Fixed-Point Theorem. The Schauder Fixed-Point Theorem 11
1.3 Fixed Points of Strict Set Contractions . . . . . . . . . . . . . . . . . . . 20
1.4 The Kronecker Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.5 The Brouwer Degree . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.5.1 Smooth Mappings . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.5.2 Homotopy Invariance . . . . . . . . . . . . . . . . . . . . . . . . 39
1.5.3 Continuous Mappings. Basic Properties . . . . . . . . . . . . . . 42
1.5.4 The Case f : ∂ D ⊂ Rn → Sn−1 . . . . . . . . . . . . . . . . . . . 49
1.6 The Leray-Schauder Degree . . . . . . . . . . . . . . . . . . . . . . . . 52
1.7 The Fixed-Point Index for Completely Continuous Mappings . . . . . . 56
1.8 The Fixed-Point Index for Strict Set Contractions . . . . . . . . . . . . . 57
1.9 Multiple Fixed-Point Theorems . . . . . . . . . . . . . . . . . . . . . . 64
2 Applications to ODEs 83
2.1 Periodic Solutions for First Order ODEs . . . . . . . . . . . . . . . . . . 83
2.2 BVPs for First Order ODEs . . . . . . . . . . . . . . . . . . . . . . . . 102
2.3 BVPs for Second Order ODEs . . . . . . . . . . . . . . . . . . . . . . . 119
2.4 BVPs with Impulses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3 Applications to FDEs 137

3.1 Global Existence for a Class of Fractional-Differential Equations . . . . 137
3.2 Multiple Solutions for a BVP of Nonlinear Riemann-Liouville Fractional
Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
3.3 Multiple Solutions for a BVP of Nonlinear Caputo Fractional Differential
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
4 Applications to Parabolic Equations 169

4.1 Differentiability of Classical Solutions with Respect to the Initial
Conditions of an IVP . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
v
vi Contents
4.2 Local Existence of Classical Solutions for an IBVP . . . . . . . . . . . . 182

4.3 Periodic Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
4.4 Multiple Solutions for an IBVP with Robin Boundary Conditions . . . . 213
5 Applications to Hyperbolic Equations 221

Conditions for an IVP for a Class of Hyperbolic Equations . . . . . . . . 221
5.2 Multiple Solutions for an IBVP with Robin Boundary Conditions . . . . 238
5.3 Periodic Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
6 Applications to Elliptic Equations 257

6.1 Multiple Solutions for a BVP with Robin Boundary Conditions . . . . . 257
6.2 Existence and Smoothness of Navier-Stokes Equations . . . . . . . . . . 264
Bibliography 293
Index 295
Preface
Many problems in science lead to nonlinear equations T x + Fx = x posed in some closed

convex subset of a Banach space. In particular, ordinary, fractional, partial differential equa-
tions and integral equations can be formulated like these abstract equations. It is the reason
for which it becomes desirable to develop fixed-point theorems for such equations. When
T is compact and F is a contraction, there are many classical tools to deal with such prob-
lems. The main aim of this monograph is to study the existence of multiple fixed points for
some operators that are of the form T + F, where T is an expansive operator and F is a
k-set contraction, and the monograph offers the reader an overview of recent developments
of multiple fixed-point theorems and their applications.
The book contains six chapters. In Chapter 1 are introduced measures of noncom-
pactness and some of their properties are derived. In Chapter 1 the main concepts for
fixed points of strict set contraction, the Kronecker index, the Brouwer degree, the Leray-
Schauder degree, the fixed-point index for completely continuous mappings, and strict
set contractions are introduced. Some multiple fixed-point theorems are formulated and
proved. Chapter 2 is devoted to some applications for ODEs such as existence of multi-
ple periodic solutions, existence of multiple positive solutions for some classes BVPs for
first and second order ODEs, and some BVPs with impulses. Chapter 3 deals with appli-
cations of some BVPs for nonlinear Riemann-Liouville and Caputo fractional differential
equations. In Chapter 4 are considered some applications of some classes of parabolic equa-
tions such as differentiability of classical solutions with respect to the initial conditions of
a class of IVPs for nonlinear parabolic equations, existence of multiple positive periodic
solutions for some classes of nonlinear parabolic equations, and existence of multiple pos-
itive solutions for some IVBPs with Robin boundary conditions. Chapter 5 is devoted to
some applications for some classes of nonlinear hyperbolic equations. In Chapter 6 are con-
sidered some applications of elliptic equations such as the existence of multiple solutions
for some classes of BVPs with Robin boundary conditions and the existence of classical
solutions for 3D Navier-Stokes equations.
The aim of this book is to present a clear and well-organized treatment of the concept
behind the development of mathematics and solution techniques. The text material in this
book is presented in a highly readable, mathematically solid format.
vii
viii Preface
The authors welcome any suggestions for the improvement of the text.
Svetlin G. Georgiev
Paris, France
Khaled Zennir
Al-Ras, KSA
1
Fixed-Point Index Theory. Multiple Fixed-Point
Theorems
1.1 Measures of Noncompactness

In this section we will present the concept of the Kuratowski measure of noncompactness,
which will be useful in the sequel sections and chapters. Suppose that E is a real Banach
space. Assume that S is a bounded set in E. Let
n
[
α(S) = inf δ > 0 : S = S j : S j ⊂ E, diam(S j ) < δ ,
j=1

j = 1, 2, . . . , n, n∈N .
Because S is bounded, we have

0 ≤ α(S) < ∞.
Definition 1.1.1 α(S) is said to be the Kuratowski measure of noncompactness and it is

called the noncompactness measure for short.
Note that α(S) = 0 if and only if S is a relatively compact subset of E and if T ⊂ E such
that S ⊆ T , then
α(S) ≤ α(T ).
Below we will list some of the properties of the Kuratowski measure.
Theorem 1.1.2 Let S be a bounded subset of E. Then
α(S) = α(S).
Proof 1.1.3 Since S ⊆ S, we get

α(S) ≤ α(S). (1.1)
1
2 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
Let, ε > 0 be arbitrarily chosen and fixed. Then there exists a partition
m
[
S= Sj
j=1
such that
Sj ⊆ E,
diam(S j ) < α(S) + ε, j ∈ {1, . . . , m}.

Now, using that
m
[
S= Sj
j=1
and
diam(S j ) = diam(S j )
< α(S) + ε, j ∈ {1, . . . , m},

we obtain
α(S) ≤ α(S) + ε.
Because ε > 0 was arbitrarily chosen, we obtain
α(S) ≤ α(S).
Hence, and considering (1.1), we get
α(S) = α(S).
This completes the proof.
Theorem 1.1.4 Let S and T be bounded subsets of E. Then
[
α(S T ) = max{α(S), α(T )}.
Proof 1.1.5 Let
η = max{α(S), α(T )}.
Then [
η ≤ α(S T ). (1.2)
Take ε > 0 arbitrarily. Then there are partitions
m
[ k
[
S= S j, T= Tl ,
j=1 l=1
Measures of Noncompactness 3
such that
Sj ⊆ E,
Tl ⊆ E,
diam(S j ) ≤ α(S) + ε
≤ η + ε,
diam(Tl ) ≤ α(T ) + ε
≤ η + ε, j ∈ {1, . . . , m}, l ∈ {1, . . . , k}.
Because ! !
[ m
[ [ k
[
S T= Sj Tl ,
j=1 l=1
we get [
α(S T ) ≤ η + ε.
Since ε > 0 was arbitrarily chosen, we go to
[
α(S T ) ≤ η.
Hence, and considering (1.2), we arrive at

[
α(S T ) = η.
α(S + T ) ≤ α(S) + α(T ),
where
S + T = {y + z : y ∈ S, z ∈ T }.
Proof 1.1.7 Take ε > 0 arbitrarily. Then there are partitions

m
[ k
[
S= S j, T= Tl ,
j=1 l=1
such that
Sj ⊆ E,
Tl ⊆ E,
diam(S j ) ≤ α(S) + ε,
diam(Tl ) ≤ α(T ) + ε, j ∈ {1, . . . , m}, l ∈ {1, . . . , k}.

Denote
V jl = {y + z : y ∈ S j , z ∈ Tl }, j ∈ {1, . . . , m}, l ∈ {1, . . . , k}.
We have
!
m
[ k
[
S+T = V jl
j=1 l=1
!
k
[ m
[
= V jl
l=1 j=1
and
diam(V jl ) ≤ diam(S j ) + diam(Tl )
≤ α(S) + ε + α(T ) + ε
= α(S) + α(T ) + 2ε, j ∈ {1, . . . , m}, l ∈ {1, . . . , k}.

Consequently
α(S + T ) ≤ α(S) + α(T ) + 2ε.
Because ε > 0 was arbitrarily chosen, we get
α(S + T ) ≤ α(S) + α(T ).
Theorem 1.1.8 Let S be a bounded subset of E and a ∈ R. Then
α(aS) = |a|α(S),
where
aS = {ay : y ∈ S}.
Proof 1.1.9 1. Let a = 0. Then

aS = {0}.
Hence,
0 = α(aS)
= |a|α(S).
2. Let a 6= 0. Take ε > 0 arbitrarily. Then there is a partition

m
[
S= S j,
j=1
such that
Sj ⊆ E,
diam(S j ) ≤ α(S) + ε, j ∈ {1, . . . , m}.

We have
m
[
aS = (aS j )
j=1
and
diam(aS j ) = |a|diam(S j )
≤ |a|(α(S) + ε).
Consequently
α(aS) ≤ |a|α(S) + |a|ε.
Because ε > 0 was arbitrarily chosen, by the last inequality we get
α(aS) ≤ |a|α(S). (1.3)
On the other hand, using that
S = a−1 (aS),
as in the above, we obtain
α(S) = α a−1 (aS)

≤ |a|−1 α(aS),
whereupon
|a|α(S) ≤ α(aS).
Hence, and applying (1.3), we get
α(aS) = |a|α(S).
Theorem 1.1.10 Let S be a bounded set of E. Then
α(coS) = α(S).
Proof 1.1.11 Note that S ⊆ coS. Therefore
α(S) ⊆ α(coS). (1.4)

m
[
Note that for any a1 > α(S) there is a partition S = S j such that
j=1
Sj ⊆ E,
diam(S j ) < a1 , j ∈ {1, . . . , m}.
Take a2 ∈ (0, a1 ) such that
diam(S j ) < a2 , j ∈ {1, . . . , m}.
Then
diam(coS j ) = diam(S j )
< a2 , j ∈ {1, . . . , m}.
Let
m
D = {λ : λ = (λ1 , . . . , λm ), λ j ≥ 0, j ∈ {1, . . . , m}, ∑ λ j = 1}.
j=1
m
Note that D is a closed bounded subset of R . For λ ∈ D, define the set
( )
m
X(λ ) = x:x= ∑ λ jx j, x j ∈ coS j , j ∈ {1, . . . , m} .
j=1
Take x, y ∈ X(λ ),
m
x = ∑ λ jx j,
j=1
m
y = ∑ λ jy j,
j=1
where x j , y j ∈ coS j , j ∈ {1, . . . , m}. Then

m m
kx − yk = ∑ λ j x j − ∑ λ j y j

j=1 j=1

m
= ∑ λ j (x j − y j )

j=1
m
≤ ∑ λ j kx j − y j k
j=1
m
≤ ∑ λ j diam(S j )
j=1
m
< a2 ∑ λ j
j=1
= a2 .
Consequently
diam(X(λ )) ≤ a2
for any λ ∈ D. Let [
X(D) = X(λ ).
λ ∈D
We have
S ⊆ X(D) ⊆ coS.
Let x, y ∈ X(D) and
m m
x= ∑ λ jx j, y= ∑ µ jy j, x j , y j ∈ coS j , j ∈ {1, . . . , m},
j=1 j=1
and λ = (λ1 , . . . , λm ), µ = (µ1 , . . . , µm ) ∈ D. Letting
z = hx + (1 − h)y, h ∈ (0, 1),

we have
m m
z = h ∑ λ j x j + (1 − h) ∑ µ j y j
j=1 j=1
m m
= ∑ hλ j x j + ∑ (1 − h)µ j y j
j=1 j=1
m
hλ j (1 − h)µ j
= ∑ (hλ j + (1 − h)µ j ) hλ j + (1 − h)µ j x j + hλ j + (1 − h)µ j y j .
j=1
Let
vj = hλ j + (1 − h)µ j ,
hλ j (1 − h)µ j
zj = xj + y j, j ∈ {1, . . . , m}.
hλ j + (1 − h)µ j hλ j + (1 − h)µ j
Then
m
z= ∑ v jz j.
j=1
We have
v = (v1 , . . . , vm ) ∈ D,
zj ∈ coS j , j ∈ {1, . . . , m}.
Thus z ∈ X(D) and X(D) is convex. Therefore
X(D) = coS.
Let
1
η = (a1 − a2 ).
2
For λ ∈ D, define
Xη (λ ) = {x ∈ E : inf kx − zk < η}.
z∈X(λ )
Since S is bounded, we have that coS and X(D) are bounded. Therefore there exists a δ > 0
such that
X(µ) ⊂ Xη (λ )
for any µ, λ ∈ D such that
kµ − λ k < δ .
n
[
Because D is a compact set in Rm , there exists a partition D = D j such that
j=1
diam(D j ) < δ , j ∈ {1, . . . , n}.

For fixed λ j ∈ D j , j ∈ {1, . . . , n}, we have
X(λ ) ⊆ Xη (λ j )
for any λ ∈ D j . Let [
X(D j ) = X(λ ).
λ ∈D j
We have
n
[
coS = X(D j ),
j=1
and
diam(X(D j )) ≤ diam(Xη (λ j ))
≤ diam(X(λ j )) + 2η
< a2 + 2η
= a1 , j ∈ {1, . . . , n}.
Therefore
α(coS) ≤ a1 .
Because a1 > α(S) was arbitrarily chosen, we obtain
α(coS) ≤ α(S).
α(S) = α(coS).
|α(S) − α(T )| ≤ 2dh (S, T ),
where dh (S, T ) denotes the Hausdorf distance between the sets S and T , i.e.,
( )
dh (S, T ) = max sup d(x, T ), sup d(y, S) ,
x∈S y∈T
where d(·, ·) denotes the distance from an element of E to a subset of E.

m
[
Proof 1.1.13 Take ε > 0 arbitrarily. Then there exists a partition S = S j such that
j=1
Sj ⊆ E,
diam(S j ) < α(S) + ε, j ∈ {1, . . . , m}.
Let
η = dh (S, T ) + ε
and define
T j = {y ∈ T : ∃x ∈ S j : kx − yk < η},
m
[
j ∈ {1, . . . , m}. Since dh (S, T ) < η, we have that T = T j . Let j ∈ {1, . . . , m} and y1 ,
j=1
y2 ∈ T j be arbitrarily chosen. Then there exist x1 , x2 ∈ S j such that
kx1 − y1 k < η,
kx2 − y2 k < η.
Hence,
ky1 − y2 k = ky1 − x1 + x1 − x2 + x2 − y2 k
≤ ky1 − x1 k + kx1 − x2 k + kx2 − y2 k
< 2η + diam(S j ).
Therefore
diam(T j ) ≤ 2η + diam(S j )
< 2dh (S, T ) + 2ε + α(S) + ε
= 2dh (S, T ) + α(S) + 3ε
and
α(T ) ≤ 2dh (S, T ) + α(S) + 3ε.
As in the above, one can prove
α(S) ≤ 2dh (S, T ) + α(T ) + 3ε.

The Brouwer Fixed-Point Theorem. The Schauder Fixed-Point Theorem 11
Consequently
|α(S) − α(T )| ≤ 2dh (S, T ) + 3ε.
Because ε > 0 was arbitrarily chosen, we obtain
|α(S) − α(T )| ≤ 2dh (S, T ).
1.2 The Brouwer Fixed-Point Theorem. The Schauder Fixed-Point

Theorem
Let Rn have the usual inner product h·, ·i and let k · k be the norm. Let also,
Bn = {x ∈ Rn : kxk < 1}
be the open unit ball,

n
B = {x ∈ Rn : kxk ≤ 1}
be the closed unit ball, and
Sn−1 = {x ∈ Rn : kxk = 1}
be the unit sphere in Rn .
n
Lemma 1.2.1 There is no C 1 -map f : B → Sn−1 such that f (x) = x for any x ∈ Sn−1 .
n
Proof 1.2.2 Assume the contrary, i.e., assume that there is a C 1 -map f : B → Sn−1 such
that f (x) = x for any x ∈ Sn−1 . Let
n
g(x) = f (x) − x, x∈B .
For t ∈ [0, 1], define

n
ft (x) = (1 − t)x + t f (x), x∈B .
Then
ft (x) = x + t( f (x) − x)
n
= x + tg(x), x∈B .
Note that
k ft (x)k = k(1 − t)x + t f (x)k
≤ (1 − t)kxk + tk f (x)k
≤ 1−t +t
n
= 1, t ∈ [0, 1], x∈B .
Therefore
n n
ft : B → B .
Also, for x ∈ Sn−1 , we have
ft (x) = (1 − t)x + tx
= x, t ∈ [0, 1].
Because f is C , we have that g is C . Therefore, there exists a constant C > 0 such that
1 1
kg(x2 ) − g(x1 )k ≤ Ckx2 − x1 k

n n
for any x1 , x2 ∈ B . Now, assume that there are two points x1 , x2 ∈ B , x1 6= x2 , such that
ft (x1 ) = ft (x2 ).
Then
x1 + tg(x1 ) = x2 + tg(x2 )
or
x1 − x2 = t(g(x2 ) − g(x1 )),
and
kx1 − x2 k = kt(g(x2 ) − g(x1 ))k
= tkg(x2 ) − g(x1 )k
≤ Ctkx2 − x1 k.
Hence,
1 ≤ Ct.
1 n n
Thus, when t < , the function ft : B → B is an injection. Let
C
Gt = ft (Bn ).
We have that ft0 : Rn → Rn and
ft0 (x) = I + tg0 (x), x ∈ Rn .
Here, I is the identity map. Since g is C 1 , there is a point t0 > 0 such that
det ft0 (x) > 0
for any t ∈ [0,t0 ]. Then, by the inverse function theorem, it follows that Gt is an open set for
all t ∈ [0,t0 ]. We take t0 > 0 small enough so that ft is\injective for all t ∈ [0,t0 ]. Assume
that Gt 6= Bn for any t ∈ [0,t0 ]. Then there is y0 ∈ ∂ Gt Bn , where ∂ Gt is the boundary of
Gt . Hence, there is a sequence {xn }n∈N ⊆ Bn so that
lim ft (xn ) = y0 .
n→∞
Next, there is a subsequence, relabeled as {xn }n∈N , such that
lim xn = x0 .
n→∞
Hence, by the continuity of f , it follows that
ft (x0 ) = y0 , t ∈ [0, 1].
But Gt = ft (Bn ) and Gt is an open set for t ∈ [0,t0 ], and every open set is disjoint near
n
its boundary. Then y0 ∈ / Gt , t ∈ [0,t0 ], and x0 ∈ B \Bn = Sn−1 . From here, f (x0 ) = x0
and ft (x0 ) = x0 , t ∈ [0,t0 ]. Therefore y0 ∈ S . This is a contradiction because y0 ∈ Bn .
n−1
n n
Consequently ft : B → B is a bijection for any t ∈ [0,t0 ]. Define the function F : [0, 1] → R
as follows.
Z
F(t) = n
det ft0 (x)dx
B
Z
= n
det(I + tg0 (x))dx.
B
n n n
Since ft (B ) = B for t ∈ [0,t0 ], we have that F(t) is the volume of B for t ∈ [0,t0 ], or
n
F(t) = volume(B ), t ∈ [0,t0 ].
Note that F is a polynomial in t, t ∈ [0, 1], and since a polynomial that is a constant on an
interval, is constant everywhere, we get
n
F(t) = volume(B ), t ∈ [0, 1].
In particular,
n
F(1) = volume(B ).
n
Because f1 (x) = f (x) ∈ Sn−1 , x ∈ B , we obtain
h f1 (x), f1 (x)i = k f1 (x)k2
n
= 1, x∈B .
Now, for any v ∈ Rn , we have

d
2h f10 (x)v, f1 (x)i = h f1 (x + tv), f1 (x + tv)i
dt
d
= 1
dt
n
= 0, x∈B .
Therefore the range of f10 (x) is contained in f (x)⊥ , the orthogonal compliment of f (x).
Then
n
rank f10 (x) ≤ n − 1, x ∈ B ,
and then
n
det f10 (x) = 0, x∈B .
Hence,
n
volume(B ) = F(1)
Z
= n
det f10 (x)dx
B
= 0.
This is a contradiction. This completes the proof.

n n
Theorem 1.2.3 (Brouwer Fixed-Point Theorem) Every continuous map f : B → B has
a fixed point.
Proof 1.2.4 By the Stone-Weierstrass Theorem, it follows that there is a sequence { fm }m∈N
of C 1 -functions such that
n
fm : B → Rn , m ∈ N,
and
1 n
k f (x) − fm (x)k ≤ , x∈B .
m
Then
k fm (x)k = k fm (x) − f (x) + f (x)k
≤ k fm (x) − f (x)k + k f (x)k
1 n
≤ 1+ , m ∈ N, x∈B .
m
Let
m
hm = fm , m ∈ N.
m+1
Then
n
khm (x)k ≤ 1, m ∈ N, x∈B ,
i.e.,
n n
hm : B → B , m ∈ N,
and
hm → f as m → ∞ uniformly.
n n
Assume that hm , m ∈ N, has no fixed point in B . Let gm : B → Sn−1 , m ∈ N, be defined as
follows
gm (x) = point where the ray from hm (x) meets Sn−1 .
n
Because hm , m ∈ N, has no fixed point in B and hm is C 1 , we have that
gm (x) = x for any x ∈ Sn−1 , m ∈ N,

n
which is a contradiction of Lemma 1.2.1. Therefore hm , m ∈ N, has a fixed point in B . Let
n
xm ∈ B , m ∈ N, be such that
hm (xm ) = xm , m ∈ N.
Then there exists a subsequence, relabeled as {xm }m∈N , such that
xm → x0 as m → ∞,
n
for some x0 ∈ B . Then
hm (xm ) → f (x0 ) as m→∞
and
xm → x0 as m → ∞.
Consequently
f (x0 ) = x0 .
n
Corollary 1.2.5 There is no continuous map f : B → Sn−1 with f (x) = x for all x ∈ Sn−1 .
n
Proof 1.2.6 Assume that there is a continuous map f : B → Sn−1 such that f (x) = x for
n n
all x ∈ Sn−1 . Let g : B → B be defined by
n
g(x) = − f (x), x∈B .
n n n
We have g : B → Sn−1 and because g : B → B is a continuous map, by Theorem 1.2.3, it
n
follows that there is an x ∈ B such that
x = g(x).
Then x ∈ Sn−1 and
x = g(x)
= − f (x)
= −x
6= x.
This is a contradiction. This completes the proof.
Definition 1.2.7 A topological space X is said to possess the fixed-point property if any
continuous map of X into X has a fixed point in X.
Theorem 1.2.8 If a topological space X is homeomorphic of a topological space Y and X

has the fixed-point property, then Y has the fixed-point property.
Proof 1.2.9 Since X is homeomorphic of Y , there is a map f : X → Y such that

1. f is a bijection,
2. f is continuous,
3. there exists f −1 and f −1 is continuous.
Let g : Y → Y be a continuous map. Since f (X) = Y , we have that
f −1 ◦ g ◦ f : X → X
is a continuous map and since X has the fixed-point property, there exists a x ∈ X so that
f −1 (g( f (x))) = x.
Hence,
g( f (x)) = f (x),
i.e., f (x) is a fixed point of g. Because g : Y → Y was an arbitrarily chosen continuous map,
we conclude that Y has the fixed-point property. This completes the proof.
Definition 1.2.10 We say that X is a retract of Y if X ⊂ Y and there exists a continuous

map r : Y → X such that r = I on X. The map r is said to be a retraction.
Theorem 1.2.11 A closed convex nonempty subset A of Rn or of a Hilbert space is a retract

of any larger subset.
Proof 1.2.12 The required retraction map is obtained by mapping each point onto the near-
est point of X. This completes the proof.
Theorem 1.2.13 If Y has the fixed-point property and X is a retract of Y , then X has the
fixed-point property.
Proof 1.2.14 Since X is a retract of Y , there is a continuous map r : Y → X so that r = I

on X. Let f : X → X be any continuous map. Then
f ◦ r : Y → X ⊂ Y.
Hence, using that Y has the fixed-point property, there is y ∈ Y so that
( f ◦ r)(y) = y and y ∈ X.
From here, r(y) = y and
( f ◦ r)(y) = f (r(y))
= f (y)
= y.
Thus f has a fixed point in X. This completes the proof.
Theorem 1.2.15 (Brouwer Fixed-Point Theorem) Every compact convex nonempty sub-
set A of Rn has the fixed-point property.
n
Proof 1.2.16 Take k > 0 sufficiently large so that the ball kB of radius k contains A. By
n n n
Theorem 1.2.11, it follows that A is a retract of kB . Since kB and B are homeomorphic,
n
and by Theorem 1.2.3 we have that B has the fixed-point property, using Theorem 1.2.8,
n
we conclude that kB has the fixed-point property. Hence, and applying Theorem 1.2.13,
we get that A has the fixed-point property. This completes the proof.
Theorem 1.2.17 Let U be a metric space with a metric ρ. Suppose that T is a continuous
map of a closed subset M of U into a compact subset Z of U and that, for each ε > 0, there
exists x(ε) ∈ M such that
ρ(T x(ε), x(ε)) < ε. (1.5)
Then T has a fixed point.
Proof 1.2.18 We have that T x(ε) ∈ Z and Z is compact. Then we can assume that for some
sequence {εn }n∈N we have
T x(εn ) → y as n → ∞.
By (1.5), we get
ρ(T x(εn ), x(εn )) < εn .
Then
x(εn ) → y as n → ∞.
Hence, using that M is closed, we get that y ∈ M. Thus,

Ty = T lim x(εn )
n→∞
= lim T x(εn )
n→∞
= y.
Definition 1.2.19 The points x(ε) satisfying (1.5) will be called ε-fixed points.
Theorem 1.2.20 Let Y be a compact metric space with a metric ρ. For each ε > 0, let
Pε : Y → Y be a continuous map such that
ρ(Pε x, x) < ε, x ∈ Y.
Suppose that Pε Y has the fixed-point property. Then Y has the fixed-point property.
Proof 1.2.21 Let f : Y → Y be any continuous map. We have that

Pε ◦ f : Pε Y → Pε Y
is a continuous map. Because Pε Y has the fixed-point property, we get that there is an
xε ∈ Pε Y such that
Pε ( f (xε )) = xε .
Hence,
ρ(xε , f (xε )) = ρ(Pε ( f (xε )), f (xε ))
< ε.
Hence, and considering Theorem 1.2.17, it follows that f : Y → Y has a fixed point. This
completes the proof.
Definition 1.2.22 The Hilbert cube H0 is the subset of l 2 consisting of points a = {an }n∈N
such that
1
|an | ≤ , n ∈ N.
n
Theorem 1.2.23 Every compact subset K of a Banach space B is homeomorphic under a

linear mapping to a compact convex subset of H0 .
Proof 1.2.24 Without loss of generality, assume that K is a subset of the unit ball in B.
Since K and span(K) are separable, we can take a sequence {xn }n∈N dense in span(K).
For n ∈ N, define fn in the dual space B∗ such that
kxn k 1
fn (x) = , k fn k = , x ∈ K.
n n
Then
F : x → { fn (x)}n∈N
2
maps K into H0 and F : B → l is a bounded linear operator. For x, y ∈ K, x 6= y, we have
| fn (x) − fn (y)| ≥ | fn (x)| − | fn (x − y − xn )|
kxn k k(x − y) − xn k
≥ −
n n
> 0
if xn is sufficiently close to x − y. Thus, F : K → F(K) is a homeomorphism and F(K) is a
compact convex subset of H0 . This completes the proof.
Definition 1.2.25 For x = {xn }n∈N ∈ l 2 , define the projection Pn of l 2 into an n-dimensional
subspace by
Pn (x) = {x1 , x2 , . . . , xn , 0, 0, . . .}.
Theorem 1.2.26 The Hilbert cube H0 has the fixed-point property.
Proof 1.2.27 For any a = {an }n∈N ∈ H0 and for any n sufficiently large, we have
!1
∞ 2
1
kPn a − ak ≤ ∑ l2
l=n+1
< ε.
Since Pn H0 is compact, then H0 is compact. Because Pn H0 can be regarded as a compact

convex subset of Rn , using Theorem 1.2.15, we conclude that Pn H0 has the fixed-point prop-
erty. Hence, and considering Theorem 1.2.20, we get that H0 has the fixed-point property.
Theorem 1.2.28 Any nonempty compact convex subset X of the Hilbert cube H0 has the
fixed-point property.
Proof 1.2.29 By Theorem 1.2.11, it follows that X is a retract of H0 . From Theorem 1.2.26,
we have that H0 has the fixed-point property. Hence, and considering Theorem 1.2.13, it
follows that X has the fixed-point property. This completes the proof.
Theorem 1.2.30 (Schauder Fixed-Point Theorem) Any compact convex nonempty sub-
set Y of a normed space has the fixed-point property.
Proof 1.2.31 By Theorem 1.2.23, it follows that Y is homeomorphic to a compact convex

subset Z of H0 . From Theorem 1.2.28, we have that Z has the fixed-point property. Hence,
and applying Theorem 1.2.8, we obtain that Y has the fixed-point property. This completes
the proof.
1.3 Fixed Points of Strict Set Contractions

Suppose that E is a Banach space.
Definition 1.3.1 Let A : E → E be a mapping(operator).
Fixed Points of Strict Set Contractions 21
1. The mapping A is said to be compact if it maps every bounded subset of E

into a relatively compact subset of E.
2. The mapping A is said to be a completely continuous mapping if it is con-
tinuous and compact.
3. The mapping A is said to be a k-set contraction if it is continuous, bounded
and
α(A(S)) ≤ kα(S)
for any bounded set S ⊂ D. Especially, if k < 1, a k-set contraction A is said to
be a strict set contraction.
4. The mapping A is said to be a condensing operator if it is continuous,
bounded and
α(A(S)) < α(S)
for any non-relatively compact and bounded set S ⊂ D.
Theorem 1.3.2 Assume that Ω is a convex open set in E and x0 ∈ Ω. If a continuous

[
mapping A : Ω → Ω has the following property: C ⊂ Ω is countable, C ⊂ co({x0 } A(C))
implies C is relatively compact, then A has a fixed point in E.
Proof 1.3.3 Define the sets
D0 = {x0 },
[
Dn = co({x0 } A(Dn−1 )), n ∈ N.
We have
D0 ⊂ D1 ⊂ . . . ⊂ Dn ⊂ . . . ⊂ Ω.
Each set Dn , n ∈ N0 , is separable and there exists a countable set Cn , n ∈ N0 , such that
Cn = Dn , n ∈ N0 .
Let [ [
D= Dn , C= Cn .
n∈N0 n∈N0
We have
[ [
D = co({x0 } A(Dn−1 ))
n∈N
[
= co({x0 } A(D)),
[ [
Dn = Dn
n∈N0 n∈N0
= D,
[ [
Dn = Cn
n∈N0 n∈N0
[
= Cn
n∈N0
= C.
Thus
C ⊂ C
= D
[
= co({x0 } A(D)) (1.6)
[
= co({x0 } A(D))
[
= co({x0 } A(C)).
Since C is countable, using (1.6), we get that C is a compact set in E and so D is a compact
set in E. Because x0 ∈ D, we have that D is nonempty. Also, we have that
A(D) ⊂ D.
Hence, and applying Schauder’s fixed-point theorem, it follows that A has a fixed point in
D. This completes the proof.
Theorem 1.3.4 (Mönch’s Theorem) Assume that D is a closed convex set in E and x0 ∈
D. If a continuous
[ mapping A : D → D has the following property: C ⊂ D is countable,
C ⊂ co({x0 } A(C)) implies C is relatively compact, then the mapping A has a fixed point
in D.
Proof 1.3.5 By Dugundji’s extension theorem,1 it follows that there exists a continuous
mapping A1 : E → D such that
A1 x = Ax, x ∈ D.
1
Theorem 1.3.6 (Dugundji’s Extension Theorem) Let X be an arbitrary metric space, A a closed subset of X, L
a locally convex linear space, and f : A → L a continuous mapping. Then there exists an extension F : X → L of
f and
F(X) ⊂ co( f (A)).
Fixed Points of Strict Set Contractions 23
[
Assume that C ⊂ E is countable, C ⊂ co({x0 } A1 (C)). Since A1 (C) ⊂ D, it follows that
C ⊂ D and so A1 (C) = A(C). Hence, C is a relatively compact set. From here and from
Theorem 1.3.2, it follows that the mapping A1 has a fixed point x∗ ∈ E. We have that A1 x∗ ∈
D and then x∗ ∈ D. Therefore
x∗ = A1 x∗
= Ax∗ .
Theorem 1.3.7 (Dabor’s Theorem) Assume that D is a nonempty bounded and closed
convex set. If a continuous mapping A : D → D has the following property: C ⊂ D is count-
able and is not relatively compact implies α(A(C)) < α(C), then the mapping A has a fixed
point in D.
[
Proof 1.3.8 Choose x0 ∈ D and a countable set C ⊂ D with C ⊂ co({x0 } A(C)). Then,
using Theorem 1.1.2 and Theorem 1.1.4, we get
[
α(C) ≤ α(co({x0 } A(C)))
≤ α(A(C)).
Therefore C is a relatively compact set. Hence, and applying Mönch’s Theorem, it follows
that A has a fixed point in D. This completes the proof.
Theorem 1.3.9 (Sadovskii’s Theorem) Assume that D is a nonempty bounded and closed
convex set. If a mapping A : D → D is condensing, then A has a fixed point in D.
Proof 1.3.10 Since A : D → D is condensing, we have
α(A(C)) < α(C).
Hence, and applying Dabor’s Theorem, we get that A has a fixed point in D. This completes
the
proof.
Theorem 1.3.11 (Darbo’s Theorem) Assume that D is a nonempty bounded and closed
convex set. If a mapping A : D → D is a strict set contraction, then A has a fixed point
in D.
Proof 1.3.12 Since A : D → D is a strict set contraction, then it is condensing. Hence, and
applying Sadovskii’s Theorem, it follows that A has a fixed point in D. This completes the
proof.
Theorem 1.3.13 Assume that {Sn }n∈N is a sequence of nonempty bounded and closed sets
in E such that
S1 ⊃ S2 ⊃ . . . ⊃ Sn ⊃ . . . .
\
If α(Sn ) → 0, as n → ∞, then S = Sn is a nonempty compact set in E.
n∈N
Proof 1.3.14 Let {xn }n∈N be a sequence of elements of S. Then xn ∈ S j for any n, j ∈ N.
Let
jn
[
Sn = Snj , n ∈ N,
j=1
be such that
1
diam(Snj ) < α(Sn ) + , j ∈ {1, . . . , jn }.
n
Since {xn }n∈N ⊂ S1 , there is a subsequence {xn1 }n∈N of {xn }n∈N so that it is included in
some S1j and
diam({xn1 }n∈N ) < α(S1 ) + 1.
Since {xn1 }n∈N ⊂ S2 , there is a subsequence {xn2 }n∈N of {xn1 }n∈N such that
1
diam({xn2 }n∈N ) < α(S2 ) + .
2
And so on. We get a sequence {xnm }n∈N so that
1
diam({xnm }n∈N ) < α(Sm ) + .
m
In fact, for m > n, we get
1
kxnm − xnn k < α(Sn ) + .
n
Thus, the sequence {xnn }n∈N is a fundamental sequence in E. Hence, it is convergent to
m
x0 ∈ E. Because xm ∈ Sn , m ≥ n, m, n ∈ N, and Sn is closed, n ∈ N, we have that x0 ∈ Sn ,
n ∈ N. Then x0 ∈ S. Consequently, S is a nonempty compact set in E. This completes the
proof.
The Kronecker Index 25
1.4 The Kronecker Index

Let n ≥ 2, n ∈ N, and
w = dy1 ∧ . . . ∧ dyn
n
be the volume n-form in R , and let σ be the solid angle (n − 1)-form
n
σ= ∑ (−1) j−1 y j dy1 ∧ . . . ∧ dy
cj ∧ . . . ∧ dyn ,
j=1
where dy
cj means that dy j is missed. We have

1
w=d σ .
n
Let D ⊂ Rn be a bounded open set with oriented smooth boundary ∂ D and f ∈ C 1 (D),
f : D → Rn , be such that 0 ∈
/ f (∂ D). Then
µ = min k f k > 0.
∂D
Definition 1.4.1 The Kronecker index iK ( f , ∂ D) is defined by

1
Z
f ∗ k f k−n σ

iK ( f , ∂ D) =
µn−1 ∂D
! (1.7)
n
1 f
Z
j
=
µn−1 ∂D
∑ (−1) j−1 k f kn d f1 ∧ . . . ∧ dcf j ∧ . . . ∧ d fn ,
j=1
where d fk is the 1-differential form

n
d fk = ∑ ∂ j fk dxk , k ∈ {1, . . . , n},
j=1
and µn−1 is the (n − 1)-dimensional measure of the unit sphere Sn−1 .

Definition 1.4.2 If n = 1 and D = (a, b), the Kronecker index of a continuous function
f : [a, b] → R such that f (a) f (b) 6= 0 can be defined by

1 f (b) f (a)
iK ( f , ∂ (a, b)) = −
2 | f (b)| | f (a)|
1
= (sign f (b) − sign f (a)) .
2
Note the following

1. If f (a) f (b) > 0, then
iK ( f , ∂ (a, b)) = 0.
2. If f (a) < 0 < f (b), then

iK ( f , ∂ (a, b)) = 1.
3. If f (b) < 0 < f (a), then
iK ( f , ∂ (a, b)) = −1.
Definition 1.4.3 Let n = 2. Then the Kronecker index is called the winding number or
Poincaré index.
Now suppose that n = 2. Then the formula (1.7) takes the form
1 f1 d f2 − f2 d f1
Z
iK ( f , ∂ D) = .
2π ∂D f12 + f22
In this case, if ∂ D has the parametric representation φ : [0, 2π] → R2 with φ (0) = φ (2π),
then
d
d f1 = f1 (φ (s))ds,
ds
d
d f2 = f2 (φ (s))ds,
ds
and
d d
f2 (φ (s)) − f2 (φ (s)) ds
Z 2π
1 f1 (φ (s)) ds f1 (φ (s))
iK ( f , ∂ D) = ds.
2π 0 ( f1 (φ (s)))2 + ( f2 (φ (s)))2
Example 1.4.4 Let D = B2 . Then ∂ D = S1 and
φ (s) = (cos s, sin s), s ∈ [0, 2π].
Suppose that f = I. Then
f1 (φ (s)) = cos s,
f2 (φ (s)) = sin s,
d
f1 (φ (s)) = − sin s,
ds
d
f2 (φ (s)) = cos s, s ∈ [0, 2π].
ds
Hence,
Z 2π
1 cos s(cos s) − sin s(− sin s)
iK ( f , ∂ D) = ds
2π 0 (cos s)2 + (sin s)2
(cos s)2 + (sin s)2

Z 2π
1
= ds
2π 0 (cos s)2 + (sin s)2
= 1.
Example 1.4.5 Let D = B2 ,

f (x1 , x2 ) = (x1 , −x2 ).
Then
f1 (φ (s)) = cos s,
f2 (φ (s)) = − sin s,
d
f1 (φ (s)) = − sin s,
ds
d
f2 (φ (s)) = − cos s.
ds
Hence,
Z 2π
1 cos s(− cos s) + sin s(− sin s)
iK ( f , ∂ D) = ds
2π 0 (cos s)2 + (sin s)2
(cos s)2 + (sin s)2

Z 2π
1
= − ds
2π 0 (cos s)2 + (sin s)2
= −1.
Example 1.4.6 Let D = B2 and
f (x1 , x2 ) = (x12 − x22 , 2x1 x2 ).
Then
f1 (φ (s)) = (cos s)2 − (sin s)2

= cos(2s),
f2 (φ (s)) = 2 sin s cos s
= sin(2s),
d
f1 (φ (s)) = −2 sin(2s),
ds
d
f2 (φ (s)) = 2 cos(2s),
ds
and
Z 2π
1 cos(2s)(2 cos(2s)) − sin(2s)(−2 sin(2s))
iK ( f , ∂ D) = ds
2π 0 (cos(2s))2 + (sin(2s))2
(cos(2s))2 + (sin(2s))2
Z 2π
1
= ds
π 0 (cos(2s))2 + (sin(2s))2
= 2.
Now suppose that

f = f1 + i f2 .
Then
1 df 1 d f1 + id f2
Z Z
=
2πi ∂D f 2πi ∂D f1 + i f2
1 (d f1 + id f2 )( f1 − i f2 )
Z
=
2πi ∂ D ( f 1 + i f 2 )( f 1 − i f 2 )
1 f1 d f1 + i f1 d f2 − i f2 d f1 + f2 d f2
Z
=
2πi ∂D f12 + f22
1 f1 d f1 + f2 d f2 1 f1 d f2 − f2 d f1
Z Z
= 2 2
+
2πi ∂D f1 + f2 2π ∂D f12 + f22
1 d(k f k2 )
Z
= 2
+ iK ( f , ∂ D)
2πi ∂ D 2k f k
= iK ( f , ∂ D).
Consequently
1 ∂1 f (φ (s))φ10 (s) + ∂2 f (φ (s))φ20 (s)

Z
iK ( f , ∂ D) = ds
2πi ∂D f (φ (s))
Z 2π
1 ∂τ f (φ (s))
= ds,
2πi 0 f (φ (s))
where ∂τ denotes the tangential derivative of f along ∂ D.

Theorem 1.4.7 Let n = 2 and ∂ D have the parametric representation φ : [0, 2π] → R2 with
φ (0) = φ (2π). Then iK ( f , ∂ D) is an integer.
Proof 1.4.8 Let

Φ = φ1 + iφ2
and
Z t d
ds f (Φ(s))
h(t) = ds,
0 f (Φ(s))
g(t) = e−h(t) f (Φ(t)), t ∈ [0, 2π].
Then
d
f (Φ(t)) −h(t)
g0 (t) = − dt e f (Φ(t))
f (Φ(t))
d
+e−h(t) f (Φ(t))
dt
= 0, t ∈ [0, 2π].
So,
g(0) = g(2π).
Since
Φ(0) = Φ(2π) and h(0) = 0,
we get
f (Φ(0)) = e−h(2π) f (Φ(2π)),
whereupon
e−h(2π) = 1 and eh(2π) = 1.
Thus
h(2π) = 2mπi
for some m ∈ Z, and

Z 2π d
1 ds f (Φ(s))
iK ( f , ∂ D) = ds
2πi 0 f (Φ(s))
1
= 2mπi
2πi
= m.
Let n ≥ 2, n ∈ N, and D ⊆ Rn be a bounded open set with oriented smooth boundary ∂ D.
Suppose that g ∈ C 2 (D), g : D → Rn are such that
kg(x)k = 1 for all x ∈ ∂ D. (1.8)
With Jg we denote the Jacobian of g.
Theorem 1.4.9 Let (1.8) hold. Then
n
Z
iK (g, ∂ D) = Jg (x)dx. (1.9)
µn−1 D
Proof 1.4.10 By (1.7), we get

1
Z
iK (g, ∂ D) = g∗ (σ )
µn−1 ∂D
1
Z
= d(g∗ (σ ))
µn−1 D
1
Z
= g∗ (dσ )
µn−1 D
n
Z
= g∗ w
µn−1 D
n
Z
= dg1 ∧ . . . ∧ dgn
µn−1 D
n
Z
= Jg (x)dx.
µn−1 D

Theorem 1.4.11 Let n = 2 and (1.8) holds. Then
Z 2π
1
iK (g, ∂ D) = g(φ (s)) ∧ ∂τ g(φ (s))ds.
2π 0
Proof 1.4.12 We have

Z 2π
1 d d
iK (g, ∂ D) = g1 (φ (s)) g2 (φ (s)) − g2 (φ (s)) g1 (φ (s)) ds
2π 0 ds ds
1 2π
Z
= (g1 (φ (s))∂τ g2 (φ (s)) − g2 (φ (s))∂τ g1 (φ (s))) ds
2π 0
Z 2π
1
= g(φ (s)) ∧ ∂τ g(φ (s))ds.
2π 0
Let 0 < ε < µ0 , a ∈ C k ([0, ∞)), a : [0, ∞) → R, be such that suppa ⊂ [ε0 , µ0 ]. Consider
wa = a(kyk)dy1 ∧ . . . ∧ dyn
= a(kyk)w.
Theorem 1.4.13 The equality

wa = dσA ,
with
σA = A(kyk)σ ,
A ∈ C k+1 ([0, ∞)), A : [0, ∞) → R, suppA ⊂ [ε0 , ∞), holds if and only if
rA0 (r) + nA(r) = a(r),
i.e., if and only if A is given by

Z r
1
A(0) = 0, A(r) = a(s)sn−1 ds, r > 0.
rn 0
Proof 1.4.14 Let A ∈ C k+1 ([0, ∞)), A : [0, ∞) → R, suppA ⊂ [ε0 , ∞). Then σA ∈ C k−1 (Rn )
and σA = 0 on B(ε0 ). For y 6= 0, we have
dσA = d(A(kyk)σ )
= d(A(kyk)) ∧ σ + A(kyk)dσ
n
A0 (kyk)yk
= ∑ dyk ∧ σ + nA(kyk)w
k=1 kyk
!
0 1 n n
j−1
= A (kyk) ∑ yk dyk ∧
kyk k=1 ∑ (−1) y j dy1 ∧ . . . ∧ dy
cj ∧ . . . ∧ dyn
j=1
+nA(kyk)w
1 n 2
= A0 (kyk) ∑ yk w + nA(kyk)w
kyk k=1
= A0 (kyk)w + nA(kyk)w
A0 (kyk) + nA(kyk) w.

=
Consequently
wa = dσA
if and only if
a(kyk)w = A0 (kyk) + nA(kyk) w.

Hence,
wa = dσA
if and only if for any r ∈ [ε0 , ∞) we have
A0 (r) + nA(r) = a(r),
which is equivalent to, for r > 0,

d n
(r A(r)) = rn−1 a(r).
dr
Corollary 1.4.15 If we assume that

Z ∞
a(r)rn−1 dr = 0,
0
then suppA ⊂ [ε0 , µ0 ].
Proof 1.4.16 Let r > µ0 . Then

Z r
1
A(r) = a(s)sn−1 ds
rn 0
1 µ0
Z
= a(s)sn−1 ds
rn 0
1
Z ∞
= a(s)sn−1 ds
rn 0
= 0.
Suppose m > µ0 and define B ∈ C 1 ([0, ∞)), B : [0, ∞) → R, by



 0 if r ∈ [0, ε0 ]



B(r) = positive if r ∈ (ε0 , µ0 ) (1.10)



 −n

r if r ∈ [µ0 , ∞).
For r > µ0 , we have

rB0 (r) + nB(r) = 0.
Theorem 1.4.17 Let B be given by (1.10); b ∈ C ([0, ∞)) defined by
b(r) = rB0 (r) + nB(r)
is such that Z ∞
suppb ⊂ [ε0 , µ0 ], b(r)rn−1 dr = 1,
0
and if f ∈ C 2 (D), k f (x)k ≥ µ, then
1
Z
iK ( f , ∂ D) = f ∗ wb
µn−1 D
1
Z
= b(k f (x)k)J f (x)dx.
µn−1 D

1
Z
f ∗ kyk−n σ

iK ( f , ∂ D) =
µn−1 ∂D
1
Z
= f ∗ σB
µn−1 ∂D
1
Z
= d( f ∗ σB )
µn−1 D
1
Z
= f ∗ dσB
µn−1 D
1
Z
= f ∗ wb
µn−1 D
1
Z
= b(k f (x)k)J f (x)dx.
µn−1 D
[
Note that b(r) = 0 for r ∈ [0, ε0 ) (µ0 , ∞) and
Z ∞ Z ∞
b(r)rn−1 dr rB0 (r) + nB(r) rn−1 dr

=
0 0
d
Z ∞
= (rn B(r))
0 dr
= 1.
1.5 The Brouwer Degree

1.5.1 Smooth Mappings
Let n ∈ N, B be open and bounded set in Rn , f ∈ C 2 (D), f : D → Rn be such that 0 ∈
/ f (∂ D).
Then
µ = min k f k > 0.
∂D
Let also 0 < ε0 < µ0 < µ.
Definition 1.5.1 The Brouwer degree dB ( f , D) is defined by
Z
dB ( f , D) = f ∗ wc
D
Z (1.11)
= c(k f k)J f (x)dx,
D
where c ∈ C ([0, ∞)), c : [0, ∞) → R, suppc ⊂ [ε0 , µ0 ] and
Z
c(k f k)dx = 1
Rn
The Brouwer Degree 35
or
1
Z ∞
c(r)rn−1 dr = .
0 µn−1
Suppose that c̃ ∈ C ([0, ∞)), c̃ : [0, ∞) → R, suppc̃ ⊂ [ε0 , µ0 ] and

1
Z ∞
c̃(r)rn−1 dr = ,
0 µn−1
and Z
dB ( f , D) = f ∗ wc̃ .
D
Let a = c − c̃. Then
Z ∞ Z ∞ Z ∞
a(r)rn−1 dr = c(r)rn−1 dr − c̃(r)rn−1 dr
0 0 0
= 0.
Hence, and applying Corollary 1.4.15, it follows that suppA ⊂ [ε0 , µ0 ]. Then, using Theo-
rem 1.4.13, we obtain
Z Z Z Z
f ∗ wc − f ∗ wc̃ = c(k f k)J f (x)dx − c̃(k f k)J f (x)dx
D D D D
Z
= (c(k f k) − c̃(k f k)) J f (x)dx
D
Z
= a(k f k)J f (x)dx
D
Z
= f ∗ wa
D
Z
= d ( f ∗ σA )
D
= 0.
By Definition 1.5.1, it follows that
dB ( f , 0)
/ = 0.
Also, if ∂ D is smooth enough, by Theorem 1.4.17, it follows that
iK ( f , ∂ D) = dB ( f , D).
/ f (D\D0 ), then
Theorem 1.5.2 If the open set D0 ⊂ D is such that 0 ∈
dB ( f , D) = dB ( f , D0 ).
In particular, if 0 ∈
/ f (D), then dB ( f , D) = 0.
Proof 1.5.3 Let
µ0 = min k f k,
∂ D0
µ 00 = min k f k.
D\D0
Then
0 < µ 00 < µ 0 , µ 00 < µ,
and take 0 < ε0 < µ0 < µ 00 in the function c defining dB ( f , D). We have
c (k f (x)k) = 0, x ∈ D\D0 .
Consequently
Z
dB ( f , D) = c (k f (x)k) J f (x)dx
D
Z
= c (k f (x)k) J f (x)dx
D0
= dB ( f , D0 ).
Definition 1.5.4 Let D ⊂ Rn be open and bounded, f ∈ C 2 (D), f : D → Rn and z ∈

/ f (∂ D).
The Brouwer degree dB ( f , D, z) is defined as follows.
dB ( f , D, z) = dB ( f (·) − z, D).
By Definition 1.5.4, it follows that
dB ( f , D, z) = dB ( f (·) − z, D)
Z
= c (k f (x) − zk) J f (x)dx,
D
where c is the function defining dB ( f (·) − z, D).

Definition 1.5.5 z ∈ Rn is called a regular value for f if J f (x) 6= 0 when x ∈ f −1 (z).
Note that any z with f −1 (z) empty is a regular value.

Theorem 1.5.6 If z ∈
/ f (∂ D) is a regular value for f , then
dB ( f , D, z) = ∑ signJ f (x)
x∈ f −1 (z)
= N+ − N− ,
where N+ (resp. N− ) denotes the number of elements of f −1 (z) with positive (negative)
Jacobian.
Proof 1.5.7 If f −1 (z) = 0,
/ then the theorem is trivial. Let z ∈ / f (∂ D) be a regular value for
f and f −1 (z) 6= 0.
/ Then f −1 (z) ⊂ D is compact and discrete. Therefore f −1 (z) is finite and
let
f −1 (z) = {x1 , . . . , xm } .
By the inverse function theorem, for each j ∈ {1, . . . , m} there exists an
\ open neighbor-
hood U j ⊂ D of x j such that f is a diffeomorphism on U j and U j Uk = 0, / j 6= k,
m
\
j, k ∈ {1, . . . , m}. Then f (U j ) is an open neighborhood of z as well as N = f (U j ) and
j=1
N ⊂ Bz (r) for some r > 0. Let
m
[
K = D\ Uj.
j=1
Then f (x) 6= z, x ∈ K. Hence, and applying Theorem 1.5.2, it follows that

!
m
[
dB ( f , D, z) = dB f, Uj, z
j=1
Z
= Sm c (k f (x) − zk) J f (x)dx
j=1 U j
m Z
= ∑ c (k f (x) − zk) J f (x)dx
j=1 U j
m
= ∑ dB ( f ,U j , z),
j=1
where ( )
µ0 < min r, min k f (·) − zk
D\ mj=1 U j
S
in the definition of the function c. Next,

Z
dB ( f ,U j , z) = c (k f (x) − zk) J f (x)dx
Uj
Z
= sign(J f (x j )) c (k f (x) − zk) |J f (x)|dx
Uj
Z
= sign(J f (x j )) c(kyk)dy
f (U j )−z
Z
= sign(J f (x j )) c(kyk)dy
Rn
= sign(J f (x j )), j ∈ {1, . . . , m}.
Therefore
m
dB ( f , D, z) = ∑ sign(J f (x j ))
j=1
= N+ − N− .
Let A : Rn → Rn be a linear isomorphism. Then


 signdetA if z ∈ A(D)
dB (A, D, z) =
0 if z ∈/ A(D).

If λ1 , . . . , λr are the real eigenvalues of A with respective multiplicities m1 , . . . , mr , and if

µ1 , µ1 , . . . , µs , µs are the non-real eigenvalues of A with respective multiplicities n1 , . . . , ns ,
then
r + 2s = n
and
r s
det(A − µI) = ∏ (λ j − µ)m j ∏(µl − µ)nl (µl − µ)nl .
j=1 l=1
Hence,
r s
mj
det(A) = ∏ λ j ∏ |µl |2nl .
j=1 l=1
Consequently
dB (A, D, z) = signdetA
p
= (−1)∑ j=1 m j ,
where the sum is over the negative eigenvalues λ1 , . . ., λ p , p ∈ {1, . . . , r}.
1.5.2 Homotopy Invariance

Let D ⊂ Rn be open and bounded, w ∈ C 1 (Rn ), w : Rn → R, F ∈ C 2 (D × [0, 1]), F :
D × [0, 1] → Rn , and
n
dFj = ∑ ∂k Fj dxk , j ∈ {1, . . . , n}.
k=1
Theorem 1.5.8 For each λ ∈ [0, 1], we have

∂λ ((w ◦ F)dF1 ∧ . . . ∧ dFn )
!!
n
= d (w ◦ F) ∑ (−1) j−1 ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn .
j=1
Proof 1.5.9 We have

!
n
∂λ dFj = ∂λ ∑ ∂k Fj dxk
k=1
n
= ∑ ∂λ ∂k Fj dxk
k=1
n
= ∑ ∂k ∂λ Fj dxk
k=1
= d (∂λ Fj ) .
Now, we get
∂λ ((w ◦ F)dF1 ∧ . . . ∧ dFn ) = (∂λ (w ◦ F)) dF1 ∧ . . . ∧ dFn + (w ◦ F)∂λ (dF1 ∧ . . . ∧ dFn )
!
n n
= ∑ (∂λ w ◦ F)∂λ Fj dF1 ∧ . . . ∧ dFn + (w ◦ F) ∑ dF1 ∧ . . . ∧ ∂λ (dFj ) ∧ . . . ∧ dFn
j=1 j=1
n
= ∑ (−1) j−1 (∂ j w ◦ F)dFj ∧ ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn
j=1
!
n
j−1
+(w ◦ F) ∑ (−1) d(∂λ Fj )dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn
j=1
!
n n
= ∑ (−1) j−1 ∑ ∂k w ◦ FdFk ∧ ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn
j=1 k=1
!
n
j−1
+(w ◦ F) ∑ (−1) d ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn
j=1
!
n
j−1
= d(w ◦ F) ∧ ∑ (−1) ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn
j=1
!
n
j−1
+(w ◦ F)d ∑ (−1) ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn
j=1
!!
n
j−1
= d (w ◦ F) ∑ (−1) ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn .
j=1
Theorem 1.5.10 If \
F(∂ D × [0, 1]) suppw = 0,
/
then Iw (F(·, λ )) is independent of λ , where
Z
Iw (F(·, λ )) = w ◦ FdF1 ∧ . . . ∧ dFn .
D
Proof 1.5.11 By Theorem 1.5.8, we get

Z
d d
Iw (F(·, λ )) = w ◦ FdF1 ∧ . . . ∧ dFn
dλ dλ D
d
Z
= (w ◦ FdF1 ∧ . . . ∧ dFn )
D dλ
!!
Z n
j−1
= d (w ◦ F) ∑ (−1) ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn
D j=1
!
Z n
j−1
= (w ◦ F) ∑ (−1) ∂λ Fj dF1 ∧ . . . ∧ dF
dj ∧ . . . ∧ dFn
∂D j=1
= 0.

Theorem 1.5.12 If 0 ∈
/ F(∂ D × [0, 1]), then dB (F(·, λ ), D) is independent of λ ∈ [0, 1].
µ= inf kFk > 0.
∂ D×[0,1]
Take 0 < ε0 < µ0 < µ in the definition of the function c defining the Brouwer degree. Then
dB (F(·, λ ), D) = Iw (F(·, λ )).
Hence, and considering Theorem 1.5.10, it follows that dB (F(·, λ ), D) is independent of
λ ∈ [0, 1]. This completes the proof.
Theorem 1.5.14 Let f , g ∈ C 2 (D), f , g : D → Rn and
ν = max k f − gk < µ = min k f k.
∂D ∂D
Then
dB ( f , D) = dB (g, D).
Proof 1.5.15 Define F : D × [0, 1] → Rn as follows.
F(x, λ ) = (1 − λ ) f (x) + λ g(x).
Then
|F(x, λ )| = k(1 − λ ) f (x) + λ g(x)k
= k f (x) + λ (g(x) − f (x))k
≥ k f (x)k − λ kg(x) − f (x)k
≥ k f (x)k − kg(x) − f (x)k
> 0
for any (x, λ ) ∈ ∂ D × [0, 1]. Hence, and applying Theorem 1.5.12, we have that
dB (F(·, λ ), D) is independent of λ ∈ [0, 1]. Then
dB ( f , D) = dB (F(·, 0), D)
= dB (F(·, 1), D)
= dB (g, D).
Theorem 1.5.16 For each f ∈ C 2 (D), f : D → Rn , and z ∈

/ f (∂ D), dB ( f , D, z) is
an integer.
Proof 1.5.17 By Sard’s Theorem,2 it follows that there is a regular value u such that
ku − zk < µ = inf k f (·) − zk.

∂D
Hence,
max kz − uk = max k f (·) − z − ( f (·) − u)k

D D
< µ.
From here and from Theorem 1.5.14, we get
dB ( f , D, z) = dB ( f , D, u).
By Theorem 1.5.6, we have that dB ( f , D, u) is an integer. Therefore dB ( f , D, z) is an integer.

1.5.3 Continuous Mappings. Basic Properties

Let D ⊂ Rn be open and bounded, f ∈ C (D), f : D → Rn , z ∈
/ f (∂ D) so that
µ = min k f (·) − zk > 0.

∂D
Then there exist g, h ∈ C 2 (D) such that

µ µ
max kg − f k < , max kh − f k < .
D 4 D 4
Hence,
kg(·) − zk = kg(·) − f (·) + f (·) − zk
≥ k f (·) − zk − kg − f k
Theorem 1.5.18 (Sard’s Theorem) The set of critical values (that is, the image of the set of critical points)
of a smooth function f from one Euclidean space or manifold to another is a null set, i.e., it has Lebesgue
measure 0.
µ
≥ µ−
4
3µ
= ,
4
kg − hk = kg − f + f − hk
≤ kg − f k + k f − hk
µ µ
≤ +
4 4
µ
= .
2
Hence, and considering Theorem 1.5.14, it follows that
dB (g, D, z) = dB (h, D, z).
Definition 1.5.19 The Brouwer degree dB ( f , D, z) is defined by
dB ( f , D, z) = dB (g, D, z),
µ
where g ∈ C 2 (D), max kg − f k < .
D 4
Theorem 1.5.20 If the open subset D0 ⊂ D is such that z ∈

/ f (D\D0 ), then
dB ( f , D, z) = dB ( f , D0 , z).
Proof 1.5.21 Let g ∈ C 2 (D) be such that

µ
max kg − f k < and max kg − f k < min k f (·) − zk.
D 4 D D\D0
Then, by Definition 1.5.19, we get
dB ( f , D, z) = dB (g, D, z),
dB ( f , D0 , z) = dB (g, D0 , z),
and by Theorem 1.5.2, we obtain
dB (g, D0 , z) = dB (g, D, z).
Hence,
dB ( f , D, z) = dB ( f , D0 , z).
Theorem 1.5.22 (Existence Property) If z ∈ / f (D), then dB ( f , D, z) = 0. Equivalently, if

dB ( f , D, z) 6= 0, there exists at least one x ∈ D such that f (x) = z.
Proof 1.5.23 Take g ∈ C 2 (D) such that

µ
max kg − f k < and max kg − f k < min k f (·) − zk.
D 4 D D
Then, using Theorem 1.5.20, it follows that
dB ( f , D, z) = dB ( f , 0,
/ z)
= dB (g, 0,
/ z)
= 0.
Theorem 1.5.24 If F ∈ C (D × [0, 1]), F : D × [0, 1] → Rn and z ∈

/ F(∂ D × [0, 1]), then
dB (F(·, λ ), D, z) is independent of λ ∈ [0, 1].
Proof 1.5.25 Let

µ = min kF(·, ·) − zk > 0.
∂ D×[0,1]
Then there exists a G ∈ C 2 (D × [0, 1]) such that

µ
max kG − Fk < .
D×[0,1] 4
Consequently
dB (F(·, λ ), D, z) = dB (G(·, λ ), D, z)
for any λ ∈ [0, 1]. By Theorem 1.5.10, it follows that dB (G(·, λ ), D, z) is independent of
λ ∈ [0, 1]. Therefore dB (F(·, λ ), D, z) is independent of λ ∈ [0, 1]. This completes the proof.
Theorem 1.5.26 Let f , g ∈ C (D), f , g : D → Rn , z ∈

/ f (∂ D),
max k f − gk < µ = min k f (·) − zk.

∂D ∂D
Then
dB ( f , D, z) = dB (g, D, z).
Proof 1.5.27 Take
F(·, λ ) = (1 − λ ) f (·) + λ g(·), λ ∈ [0, 1].
We have
kF(·, λ ) − zk = k f (·) − z − λ ( f (·) − g(·))k
≥ k f (·) − zk − λ k f − gk
≥ k f (·) − zk − k f − gk
> 0.
Hence, and applying Theorem 1.5.24, we obtain
dB ( f , D, z) = dB (F(·, 0), D, z)
= dB (F(·, 1), D, z)
= dB (g, D, z).

Theorem 1.5.28 If f ∈ C (D), f : D → Rn , z ∈
/ f (∂ D), then for any y ∈ Bz min k f (·) − zk ,
∂D
we have
dB ( f , D, y) = dB ( f , D, z).
Proof 1.5.29 Note that
k f (·) − z − ( f (·) − y)k = kz − yk
< min k f (·) − zk.

∂D
Hence, and considering Theorem 1.5.26, it follows that

dB ( f (·) − y, D) = dB ( f (·) − z, D),
whereupon
dB ( f , D, y) = dB ( f , D, z).
Theorem 1.5.30 If f ∈ C (D), f : D → Rn , is such that 0 ∈
/ f (∂ D) and f (D) ⊂ Y , where Y
is a proper vector subspace of Rn , then
dB ( f , D, 0) = 0.
Proof 1.5.31 Let y ∈
/ Y be such that
kyk < min k f k.
∂D
Then, using Theorem 1.5.28, we get

dB ( f , D, 0) = dB ( f , D, y).
Since y ∈
/ Y , we have y ∈
/ f (D). Therefore, using Theorem 1.5.20, it follows that
dB ( f , D, y) = dB ( f , 0,
/ y)
= 0.
Hence,
dB ( f , D, 0) = 0.
Theorem 1.5.32 Let f ∈ C (D). Then dB ( f , D, z) is a constant on each connected compo-
nent of Rn \ f (∂ D).
Proof 1.5.33 Let Ω be a connected component of Rn \ f (∂ D). Then Ω is open and con-
nected and hence, it is path-connected. Let y, z ∈ Ω. Then there is a continuous path
θ : [0, 1] → Ω such that
θ (0) = y, θ (1) = z.
For any (x, λ ) ∈ ∂ D × [0, 1], one has f (x) 6= θ (λ ). Then, using Theorem 1.5.24, it follows
that
dB ( f , D, y) = dB ( f , D, θ (0))
= dB ( f (·) − θ (0), D, 0)
= dB ( f (·) − θ (1), D, 0)
= dB ( f , D, θ (1))
= dB ( f , D, z).
Theorem 1.5.34 Suppose!that {D j } j∈N is a sequence of open and mutually disjoint subsets
[
of D. If z ∈
/ f D\ Dk , then dB ( f , D j , z) = 0 for all but finitely many j ∈ N, and
k∈N
∞
dB ( f , D, z) = ∑ dB ( f , D j , z).
j=1
Proof 1.5.35 By these assumptions, we have
0 < µ0 = min
S k f (·) − zk
D\ j∈N D j
and
µ0 ≤ µ = min k f (·) − zk,

∂D
µ0 ≤ µ j = min k f (·) − zk, j ∈ N.

∂Dj
Let g ∈ C 2 (D), g : D → Rn , be such that

µ0
max k f − gk < ,
D 4
3µ0
≤ µ̃ = min kg(·) − zk,
4 ∂D
3µ0
≤ µ˜ j = min kg(·) − zk.
4 ∂ DJ
Then
dB ( f , D, z) = dB (g, D, z),
dB ( f , D j , z) = dB (g, D j , z), j ∈ N.
Take 0 < r < µ̃. By Sard’s Theorem, it follows that there exists a regular value u ∈ Bz (r) ⊂ D
of f . Hence,
dB (g, D, z) = dB (g, D, u),
dB (g, D j , z) = dB (g, D j , u), j ∈ N.

Since u is regular, g−1 (u) is finite and hence, it is contained in a finite number of D j ’s,
namely D j1 , . . . , D jm . Then
dB ( f , D, z) = dB (g, D, z)
= dB (g, D, u)
!
m
[
= dB g, D jl , u
l=1
m
= ∑ dB (g, D jl , u)
l=1
m
= ∑ dB (g, D jl , z)
l=1
m
= ∑ dB ( f , D jl , z).
l=1

Example 1.5.36 Let a < b and f : [a, b] → R be continuous. Then

1 f (b) f (a)
dB ( f , (a, b), 0) = − .
2 | f (b)| | f (a)|
Example 1.5.37 Let
n
p(x) = ∑ ak xk
k=1
be a real polynomial with an 6= 0. Let R > 0 be sufficiently large. Then

1 p(R) p(−R)
dB (p, (−R, R), 0) = −
2 |p(R)| |p(−R)|

 0 if n is even

= a
 n if n is odd.

|an |
1.5.4 The Case f : ∂ D ⊂ Rn → Sn−1

Let D ⊂ Rn be open and bounded, f : ∂ D → Sn−1 is continuous. Let also g, h : D → Rn be
two continuous extensions of f to D. Then, by Theorem 1.5.26, it follows that
dB (g, D) = dB (h, D).
Definition 1.5.38 The Brouwer degree dB ( f , ∂ D) is defined as follows.
dB ( f , ∂ D) = dB (g, D),
where g : D → Rn is any continuous extension of f to D.
Theorem 1.5.39 If f : ∂ D → Sn−1 is not onto, then
dB ( f , ∂ D) = 0.
Proof 1.5.40 Let y ∈ Sn−1 \ f (∂ D) and f˜ : Rn → B(1) be a continuous extension of f . De-

fine the homotopy F : B(1) × [0, 1] → Rn as follows.
F(x, λ ) = (1 − λ ) f˜(x) − λ y.
Suppose that there exists a (x, λ ) ∈ ∂ D × [0, 1] so that
F(x, λ ) = 0.
Then
0 = (1 − λ ) f˜(x) − λ y
= (1 − λ ) f (x) − λ y
or
(1 − λ ) f (x) = λ y.
Hence,
1−λ = (1 − λ )k f (x)k
= λ kyk
= λ.
1
Therefore λ = and f (x) = y, which is a contradiction. By Theorem 1.5.24, we get
2
dB ( f , ∂ D) = dB ( f˜, D)
= dB (F(·, 0), D)
= dB (F(·, 1), D)
= dB (−y, D)
= 0.
Corollary 1.5.41 If dB ( f , ∂ D) 6= 0, then f is onto.
Theorem 1.5.42 Let g ∈ C (Sn−1 ), g : Sn−1 → Sn−1 has no fixed points. Then
dB (g, Sn−1 ) = (−1)n .
Proof 1.5.43 Let g̃ : Rn → B(1) be a continuous extension of f . Consider the homotopy

F : B(1) × [0, 1] → Rn defined by
F(x, λ ) = (1 − λ )g̃(x) − λ x.
Assume that there is a (x, λ ) ∈ Sn−1 × [0, 1] such that
F(x, λ ) = 0.
Hence,
0 = (1 − λ )g(x) − λ x
or
(1 − λ )g(x) = λ x.
Then
1−λ = (1 − λ )kg(x)k
= λ kxk
= λ,
1
whereupon λ = and g(x) = x. This is a contradiction. Therefore
2
dB (g, Sn−1 ) = dB (g̃, B(1))
= dB (F(·, 0), B(1))
= dB (F(·, 1), B(1))
= dB (−I, B(1))
= (−1)n .
Corollary 1.5.44 Any mapping g ∈ C (Sn−1 ), g : Sn−1 → Sn−1 such that dB (g, Sn−1 ) 6=
(−1)n , has at least one fixed point.
Proof 1.5.45 Assume that g has no fixed points. By Theorem 1.5.42, it follows that
dB (g, Sn−1 ) = (−1)n . This is a contradiction. This completes the proof.
Definition 1.5.46 Let D ⊂ Rn be open and bounded, ∂ D is smooth and f ∈ C 1 (∂ D), f :

∂ D → Sn−1 . The Kronecker index iK ( f , ∂ D) is defined by
iK ( f , ∂ D) = iK (g, ∂ D),
where g is any C 1 extension of f to D.
Remark 1.5.47 Suppose that all conditions of Definition 1.5.38 hold. Then
dB ( f , ∂ D) = dB (g, D)
= iK (g, ∂ D)
= iK ( f , ∂ D).
1.6 The Leray-Schauder Degree

In this section we construct the Leray-Schauder degree. We need the following result on
the approximation of a compact mapping by finite dimensional mappings.
Theorem 1.6.1 Let E be a real Banach space, Ω ⊂ E be an open and bounded subset, and
T : Ω → E be a completely continuous mapping. Then for any ε > 0 there exists a finite
dimensional space F and a continuous mapping Tε : Ω → F such that
kTε x − T xk < ε, x ∈ Ω.
Proof 1.6.2 Since T : Ω → E is a completely continuous mapping, we have that T (Ω) is

relatively compact in E. Then for any ε > 0 there exist a finite number of elements of Ω,
{x1 , . . . , xm } ⊂ Ω, such that
m
[
T (Ω) ⊂ BT x j (ε).
j=1
Define
m
[
F = BT x j (ε),
j=1
φ j (x) = max{0, ε − kT x − T x j k}, j ∈ {1, . . . , m},
m
Γ(x) = ∑ φ j (x),
j=1
m
φ j (x)
Tε x = ∑ T x j.
j=1 Γ(x)
Let A ⊆ {1, . . . , m} so that for j ∈ A we have
kT x j − T xk < ε.
Then, for j ∈ {1, . . . , m}\A, we have φ j (x) = 0 and

m φ (x)
j
kTε x − T xk = ∑ T x j − T x

j=1 Γ(x)

m φ (x)
j
= ∑ (T x j − T x)

j=1 Γ(x)
The Leray-Schauder Degree 53
m
φ j (x)
≤ ∑ kT x j − T xk
j=1 Γ(x)
φ j (x)
= ∑ kT x j − T xk
j∈A Γ(x)
φ j (x)
< ε ∑
j∈A Γ(x)
m
φ j (x)
= ε ∑
j=1 Γ(x)
= ε, x ∈ Ω.
Theorem 1.6.3 Let E be a real Banach space, B ⊂ E be a closed bounded subset, and
T : B → E be a completely continuous mapping. Suppose that T x 6= x, x ∈ B. Then there
exists an ε0 > 0 such that
x 6= tTε1 x + (1 − t)Tε2 x, t ∈ [0, 1], x ∈ B,
where ε1 , ε2 ∈ (0, ε0 ), and Tεi : B → Fεi , i ∈ {1, 2}, are as in Theorem 1.6.1.
Proof 1.6.4 Suppose the contrary. There exist ε1j , ε2j → 0, t j → t0 , as j → ∞, t0 ∈ [0, 1],
such that
t j Tε j x j + (1 − t j )Tε j x j = x j , x j ∈ B, j ∈ N, (1.12)
1 2
and Tεj1 , Tεj2

are as in Theorem 1.6.1, j ∈ N. By the compactness of T , the sequence
{T x j } j∈N has a subsequence {T x jn }n∈N converging to y ∈ E and by Theorem 1.6.1, it
follows that
T jk x jk → y as k → ∞, i ∈ {1, 2}.
εi
Hence, and applying (1.12), we conclude that
x jk → y as k → ∞.
Therefore Ty = y, y ∈ B. This is a contradiction. This completes the proof.
Definition 1.6.5 Let E be a real Banach space, Ω ⊂ E be an open and bounded set, and
T : Ω → E be a completely continuous mapping. By Theorem 1.6.3, it follows that there
exists an ε0 > 0 such that
x 6= tTε1 x + (1 − t)Tε2 x, t ∈ [0, 1], x ∈ ∂ Ω,

where εi ∈ (0, ε0 ), Tεi : Ω → Fεi , i ∈ {1, 2}, as in Theorem 1.6.3. Then

\
dB (I − Tεi , Ω Fεi , 0), i ∈ {1, 2},
is well defined, and

\ \
dB (I − Tε1 , Ω Fε1 , 0) = dB (I − Tε2 , Ω Fε2 , 0).
Consequently, we define the Leray-Schauder degree as follows.

\
dLS (I − T, Ω, 0) = dB (I − Tε , Ω Fε , 0)
for any ε ∈ (0, ε0 ). For the general case, if p ∈

/ (I − T )(∂ Ω), we define
dLS (I − T, Ω, p) = dLS (I − T − p, Ω, 0).
Below we will list some of the properties of the Leray-Schauder degree.

Theorem 1.6.6 (Normality) We have
dLS (I, Ω, 0) = 1
if and only if 0 ∈ Ω.
Proof 1.6.7 Let ε0 > 0, ε ∈ (0, ε0 ), Oε , Fε be as in Theorem 1.6.1 and Theorem 1.6.3. Then
\
dLS (I, Ω, 0) = dB (I − Oε , Ω Fε , 0)
= 1
\
if and only if 0 ∈ Ω Fε . This completes the proof.
Theorem 1.6.8 (Solvability) If
dLS (I − T, Ω, 0) 6= 0,
then the equation T x = x has a solution in Ω.
Proof 1.6.9 Let ε0 > 0, ε ∈ (0, ε0 ), and Tε , Fε be as in Theorem 1.6.1 and Theorem 1.6.3.
Then \
dLS (I − T, Ω, 0) = dB (I − Tε , Ω Fε , 0).
The Leray-Schauder Degree 55
Since \
dB (I − Tε , Ω Fε , 0) 6= 0,
\
by Theorem 1.5.22, it follows that there exists an x ∈ Ω Fε such that
Tε x = x.
Hence, and applying
ε > kTε x − T xk
= k(Tε − I)x + (I − T )xk
= k(I − T )xk,
we conclude that T x = x. This completes the proof.
Theorem 1.6.10 (Homotopy) Let Tt : [0, 1] × Ω → E be a completely continuous mapping

and Tt x 6= x for any (t, x) ∈ [0, 1] × ∂ Ω. Then dLS (I − Tt , Ω, 0) doesn’t depend on t ∈ [0, 1].
Proof 1.6.11 Let ε0 > 0, ε ∈ (0, ε0 ), and Ttε , Ftε be as in Theorem 1.6.1 and Theorem 1.6.3.
Then [
dB (I − Ttε , Ω Ftε , 0)
is independent of t ∈ [0, 1]. Hence, and applying
\
dLS (I − Tt , Ω, 0) = dB (I − Ttε , Ω Ftε , 0),
we conclude that dLS (I − Tt , Ω, 0) is independent of t ∈ [0, 1]. This completes the proof.
\ 1.6.12 (Additivity) Let Ω1 , Ω2 be disjoint open subsets of Ω and 0 ∈

Theorem /
Ω\(Ω1 Ω2 ). Then
dLS (I − T, Ω, 0) = dLS (I − T, Ω1 , 0) + dLS (I − T, Ω2 , 0).
Proof 1.6.13 Let ε0 > 0, ε ∈ (0, ε0 ), Tε , Fε be as in Theorem 1.6.1 and Theorem 1.6.3.
Then
\
dLS (I − T, Ω, 0) = dB (I − Tε , Ω Fε , 0)
= dB (I − Tε , Ω1 , 0) + dB (I − Tε , Ω2 , 0)
= dLS (I − T, Ω1 , 0) + dLS (I − T, Ω2 , 0).

1.7 The Fixed-Point Index for Completely Continuous Mappings

Let X be a Banach space, K is a closed convex subset of X, and Ω ⊂ K is open and bounded.
Definition 1.7.1 If F : Ω → K is a completely continuous mapping such that x 6= Fx, x ∈
∂ Ω, r1 , r2 : X → K are retractions, then
dLS (I − Fr1 , r1−1 (Ω), 0) and dLS (I − Fr2 , r2−1 (Ω), 0)
are well defined and
dLS (I − Fr1 , r1−1 (Ω), 0) = dLS (I − Fr2 , r2−1 (Ω), 0).
Define the fixed-point index i(F, Ω, K) as follows.
i(F, Ω, K) = dLS (I − Fr, r−1 (Ω), 0)
for any retraction r : X → K.
Below, suppose that F : Ω → K is a completely continuous operator such that x 6= Fx for

x ∈ ∂ Ω.
Theorem 1.7.2 (Normality) If F is a constant mapping, then
i(F, Ω, K) = 1.
Proof 1.7.3 Let r : X → K be any retraction. Then
dLS (I − Fr, r−1 (Ω), 0) = 1.
Hence,
i(F, Ω, K) = dLS (I − Fr, r−1 (Ω), 0)
= 1.
Theorem 1.7.4 (Solvability) If i(F, Ω, K) 6= 0, then Fx = x has a solution in Ω.
i(F, Ω, K) = dLS (I − Fr, r−1 (Ω), 0)
6= 0.
Hence, and applying Theorem 1.6.8, it follows that there exists an x ∈ r−1 (Ω) ⊆ Ω such
that Frx = x. Since r : X → K is a retraction and Ω ⊂ K, we get Fx = x. This completes the
proof.
The Fixed-Point Index for Strict Set Contractions 57
Theorem 1.7.6 (Homotopy) Let Tt : [0, 1] × Ω → E be a completely continuous mapping

and Tt x 6= x for any (t, x) ∈ [0, 1] × ∂ Ω. Then i(Tt , Ω, K) is independent of t ∈ [0, 1].
Proof 1.7.7 Let r : X → Y be any retraction. Then, by Theorem 1.6.10, it follows that
dLS (I − Tt r, r−1 (Ω), 0) is independent of t ∈ [0, 1]. Hence, and applying
i(Tt , Ω, K) = dLS (I − Tt r, r−1 (Ω), 0),
it follows that i(Tt , Ω, K) is independent of t ∈ [0, 1]. This completes the proof.
[ 1.7.8 (Additivity) Let Ω1 , Ω2 be two disjoint open subsets of Ω and 0 ∈

Theorem /
Ω\(Ω1 Ω2 ). Then
i(F, Ω, K) = i(F, Ω1 , K) + i(F, Ω2 , K).
i(F, Ω, K) = dLS (I − Fr, r−1 (Ω), 0),
i(F, Ω1 , K) = dLS (I − Fr, r−1 (Ω1 ), 0),
i(F, Ω2 , K) = dLS (I − Fr, r−1 (Ω2 ), 0).

Hence, and applying Theorem 1.6.12, it follows that
i(F, Ω, K) = dLS (I − Fr, r−1 (Ω), 0)
= dLS (I − Fr, r−1 (Ω1 ), 0) + dLS (I − Fr, r−1 (Ω2 ), 0)
= i(F, Ω1 , K) + i(F, Ω2 , K).

1.8 The Fixed-Point Index for Strict Set Contractions

Let E be a Banach space.
Definition 1.8.1 Suppose that X is a closed convex set in E, U is a bounded open set in X,
A : U → X is a k-set contraction, 0 ≤ k < 1, and Ax 6= x, x ∈ ∂U. Let
D1 = coA(U),
\
Dn = coA(Dn−1 U), n ∈ N, n ≥ 2.
\
If there is some n0 ∈ N, n0 ≥ 2, such that Dn0 U = 0, / then Dn = 0/ for any n ∈ N, n > n0 .
In this case, we define
i(A,U, X) = 0.
\
Assume that Dn U 6= 0/ for any n ∈ N. We have
\
D2 = coA(D1 U)
⊂ coA(U)
= D1 ,
and
\
α(D2 ) = α coA(D1 U)
\
= α(A(D1 U))
\
≤ kα(D1 U)
≤ kα(D1 ).
Assume that
D1 ⊃ D2 ⊃ . . . ⊃ Dn
and
α(Dn ) ≤ kn−1 α(D1 )
for some n ∈ N. Then
\
Dn+1 = coA(Dn U)
\
⊂ coA(Dn−1 U)
= Dn
and
\
α(Dn+1 ) = α coA(Dn U)
\
= α(A(Dn U))
\
≤ kα(Dn U)
≤ kα(Dn )
≤ kn α(D1 ).
Thus
Dn+1 ⊂ Dn
and
α(Dn+1 ) ≤ kn α(D1 )
for any n ∈ N, n ≥ 2. We have that
α(Dn ) → 0 as n → ∞.
Let \
D= Dn .
n∈N
By Theorem \ 1.3.13, it follows that D is a nonempty compact set in E. Also, D is convex.

Hence, A : D U → D is a completely continuous operator. Let A1 : U → D be a completely
\ of the operator A. Assume that there is a x0 ∈ ∂U such that A1 x0 = x0 .
continuous extension
Because x0 ∈ D U, we have that Ax0 = x0 . This is a contradiction. Assume that A2 : U →
D is another completely continuous extension of A. Define
H(t, x) = tA1 x + (1 − t)A2 x, t ∈ [0, 1], x ∈ U.
We have that H : [0, 1] ×U → X is completely continuous. Let (t0 , x0 ) ∈ [0, 1] × ∂U be such

that H(t0 , x0 ) = x0 . Then
x0 = t0 A1 x0 + (1 − t0 )A2 x0 .
\
Since D is convex and A1 , A2 : U → D, then x0 ∈ D. Thus x0 ∈ D U and
A1 x0 = A2 x0 = Ax0 .
Therefore
x0 = t0 A1 x0 + (1 − t0 )A2 x0
= t0 Ax0 + (1 − t0 )Ax0
= Ax0 ,
which is a contradiction. By the homotopy invariance of the fixed-point index for completely
continuous operators, it follows that
i(A1 ,U, X) = i(A2 ,U, X).
Then, we define
i(A,U, X) = i(A1 ,U, X).
By the properties of the fixed-point index for a completely continuous operator, we have
the following properties of the fixed-point index for strict set contractions.
1. Normality. If A : U → X is a constant mapping, then
i(A,U, X) = 1.
2. Additivity.
[ If U1 and U2 are disjoint open subsets of U such that A has no fixed points
in U\(U1 U2 ), then
i(A,U, X) = i(A,U1 , X) + i(A,U2 , X).
3. Homotopy Invariance. If we assume
1. H : [0, 1] ×U → X is continuous and H(t, x) is uniformly continuous in t,

2. H(t, ·) : U → X is a k-set contraction, 0 ≤ k < 1, where k is independent of
t ∈ [0, 1] and
3. H(t, x) 6= x for any t ∈ [0, 1], x ∈ ∂U,
then i(H(t, ·),U, X) is independent of t ∈ [0, 1].

4. Preservation. If Y is a closed convex subset of X and A(U) ⊂ Y , then
\
i(A,U, X) = i(A,U Y,Y ).
5. Excision Property. If V is open with respect to X, V ⊂ U, and A has no fixed point in

U\V , then
i(A,U, X) = i(A,V, X).
6. Solvability. If i(A,U, X) 6= 0, then A has at least one fixed point in U.

Theorem 1.8.2 Assume that X is a closed convex set in E, X1 is a bounded closed convex
subset of X, U is a nonempty open set of X with U ⊂ X1 . If A : X1 → X is a k-set contraction,
0 ≤ k < 1, A(X1 ) ⊂ X1 and A has no fixed point in X1 \U, then
i(A,U, X) = 1.
Proof 1.8.3 Since X1 is a closed subset of E, U ⊂ X1 , by the preservation property of the
fixed-point index, it follows that
i(A,U, X) = i(A,U, X1 ). (1.13)
Because A has no fixed points in X1 \U, we get
i(A,U, X1 ) = i(A, X1 , X1 ). (1.14)
Take x0 ∈ U ⊂ X1 and let
H(t, x) = tx0 + (1 − t)Ax, t ∈ [0, 1], x ∈ X1 .
We have that H : [0, 1] × X1 → X1 is continuous. Also, for t ∈ [0, 1] and S ⊂ X1 , we have
α(H(t, S)) ≤ (1 − t)α(A(S))
≤ (1 − t)kα(S).
So, H(t, ·) : X1 → X1 is a k-set contraction for any t ∈ [0, 1]. Hence, using the normality and
the homotopy invariance of the fixed-point index, we get
i(A, X1 , X1 ) = i(H(0, ·), X1 , X1 )
= i(H(1, ·), X1 , X1 )
= i(x0 , X1 , X1 )
= 1.
From here and from (1.13), (1.14), we arrive at
i(A,U, X) = i(A,U, X1 )
= i(A, X1 , X1 )
= 1.
Corollary 1.8.4 Assume that X is a closed convex set in E and U is a nonempty bounded
open convex subset of X. If A : U → X is a strict set contraction and A(U) ⊂ X, then
i(A,U, X) = 1.
Proof 1.8.5 We apply Theorem 1.8.2 for X1 = U. Then
i(A,U, X) = 1.
Theorem 1.8.6 Assume that X is a closed convex set in E and U is a bounded open set in
X with 0 ∈ U. If A : U → X is a k-set contraction, 0 ≤ k < 1, and
Ax 6= µx, x ∈ ∂U, µ ≥ 1, (1.15)
then
i(A,U, X) = 1.
Proof 1.8.7 Let
H(t, x) = tAx, t ∈ [0, 1], x ∈ U.
We have H : [0, 1] × U → X is continuous and H(t, ·) : U → X is a k-set contraction, 0 ≤
k < 1. Assume that there is a (t0 , x0 ) ∈ [0, 1] × ∂U such that
H(t0 , x0 ) = x0 .
Hence,
t0 Ax0 = x0 .
If t0 = 0, then x0 = 0. This is a contradiction because 0 ∈ U. If t0 6= 0, then
1
Ax0 = x0 ,
t0
which contradicts (1.15). Therefore H(t, x) 6= x for any (t, x) ∈ [0, 1] × ∂U. Thus, by the
homotopy invariance and the normality of the fixed-point index, it follows that
i(A,U, X) = i(H(1, x),U, X)
= i(H(0, x),U, X)
= i(0,U, X)
= 1.
Theorem 1.8.8 Assume that X is a closed convex set in E, U is a bounded open set in X,
A : U → X is a k-set contraction, and 0 ≤ k < 1. If there exists a u0 ∈ X, u0 6= 0, such that
λ u0 ∈ X for any λ ∈ [0, ∞) and
x − Ax 6= λ u0 , x ∈ ∂U, λ ∈ [0, ∞), (1.16)
then
i(A,U, X) = 0.
Proof 1.8.9 Assume that i(A,U, X) 6= 0. Choose λ0 > 0 such that
1
λ0 > sup (kxk + kAxk) . (1.17)
ku0 k x∈U
Let
H(t, x) = Ax + λ0tu0 , t ∈ [0, 1], x ∈ U.
Since X is a closed convex set, we have, for any n ∈ N,

1 1 1
1− Ax + λ0tu0 = 1 − Ax + nλ0tu0 ∈ X
n n n
for any t ∈ [0, 1] and for any x ∈ U. Letting n → ∞, we get
Ax + λ0tu0 ∈ X
for any x ∈ U and t ∈ [0, 1]. Thus H : [0, 1] × U → X. Also, H(t, ·) : U → X is a k-set
contraction, and 0 ≤ k < 1, for any t ∈ [0, 1]. Assume that there is an x0 ∈ ∂U and a
t0 ∈ [0, 1] such that
H(t0 , x0 ) = x0 .
Then
Ax0 + λ0t0 u0 = x0
or
λ0t0 u0 = x0 − Ax0 ,
which contradicts (1.16). Therefore H(t, x) 6= x for any t ∈ [0, 1] and x ∈ ∂U. By the homo-
topy invariance of the fixed-point index, we get
0 6= i(A,U, X)
= i(H(0, ·),U, X)
= i(H(1, ·),U, X)
= i(A + λ0 u0 ,U, X).

Hence, using the solvability of the fixed-point index, it follows that there is an x1 ∈ U so
that
Ax1 + λ0 u0 = x1 .
Thus
λ0 u0 = x1 − Ax1
and
λ0 ku0 k ≤ kx1 k + kAx1 k
≤ sup (kxk + kAxk) .

x∈U
The last inequality contradicts (1.17). Therefore
i(A,U, X) = 0.
1.9 Multiple Fixed-Point Theorems

Let E be a real Banach space.
Definition 1.9.1 A closed convex subset P of E is said to be a cone if
1. x ∈ P and λ ≥ 0 imply λ x ∈ P,
2. x ∈ P and −x ∈ P imply x = 0.
We will denote P ∗ = P\{0}. A partially ordered relation in E is given by x ≤ y, x, y ∈ E,
if y − x ∈ P.
Definition 1.9.2 A cone P ⊂ E is said to be normal if there exists a constant N such that
0≤x≤y imply kxk ≤ Nkyk.
The smallest constant N is called the normal constant of P.

Multiple Fixed-Point Theorems 65
Definition 1.9.3 Let A : P → E be a mapping and α ∈ R.

1. The mapping A is said to be α-concave if
A(tx) ≥ t α Ax
for any (x,t) ∈ P × [0, 1].

2. The mapping A : P → E is said to be α-convex if
A(tx) ≤ t α Ax
for any (x,t) ∈ P × [0, 1].
Definition 1.9.4 Let A : P → E be a mapping, e ∈ P ∗ ,
Ce = {x ∈ E : ∃α = α(x) > 0 and β = β (x) > 0 : αe ≤ x ≤ β e} .
1. The mapping A is called an e-concave mapping if, for any x ∈ P ∗ for which
Ax ∈ Ce and for any (x,t) ∈ Ce × (0, 1), there exists an η1 = η1 (x,t) > 0 such
that
A(tx) ≥ (1 + η1 )tAx.
2. The mapping A is called an e-convex mapping if, for any x ∈ P ∗ for which
Ax ∈ Ce and for any (x,t) ∈ Ce × (0, 1), there exists an η2 = η2 (x,t) > 0 such
that
A(tx) ≤ (1 − η2 )tAx.
Definition 1.9.5 Let X be a metric space, endowed with a metric d, and D be a subset of
X. The mapping T : D → X is said to be expansive if there exists a constant h > 1 such that
d(T x, Ty) ≥ hd(x, y), x, y ∈ D.
Theorem 1.9.6 Let X be a linear normed space, endowed with a norm k · k, and D ⊂ X.
Assume that T : D → X is expansive with a constant h > 1. Then the inverse of I − T : D →
(I − T )(D) exists and
1
k(I − T )−1 x − (I − T )−1 yk ≤ kx − yk
h−1
for any x, y ∈ (I − T )(D).
Proof 1.9.7 Let F = I − T . Since T : D → X is expansive with a constant h > 1, we have

that F −1 exists and
k(I − T )x − (I − T )yk = kx − y − (T x − Ty)k
≥ kT x − Tyk − kx − yk
≥ hkx − yk − kx − yk
= (h − 1)kx − yk, x, y ∈ D.
Hence,
kFx − Fyk ≥ (h − 1)kx − yk, x, y ∈ D. (1.18)
Therefore F is one-to-one. From here, the inverse of F : D → F(D) exists. Now, taking
x, y ∈ F(D), we have F −1 x, F −1 y ∈ D and using (1.18), we get
kx − yk ≥ (h − 1)kF −1 x − F −1 yk
or
1
kF −1 x − F −1 yk ≤ kx − yk.
h−1
Theorem 1.9.8 Let X be a complete metric space, provided with a metric d, D be a closed
subset of X. Assume that the mapping T : D → X is expansive and D ⊂ T (D). Then there
exists a unique point x∗ ∈ D such that T x∗ = x∗ .
Proof 1.9.9 Since T is expansive, it follows that T −1 : T (D) → D exists, it is a contraction

and hence, it is continuous. So, T (D) is a closed set. Because D ⊂ T (D), there exists a
unique x∗ ∈ T (D) such that
T −1 x∗ = x∗ .
Thus x∗ ∈ D and
T x∗ = x∗ .
Theorem 1.9.10 Let E be a Banach space and K ⊂ E be a closed convex subset. Suppose
that T and S map K into E such that
(i) S is continuous, S(K) resides in a compact subset of E,
(ii) T is an expansive mapping with a constant h > 1, and
(iii) z ∈ S(K) implies T (K) + z ⊃ K, where
T (K) + z = {y + z : y ∈ T (K)}.
Then there exists a point x∗ ∈ K such that
Sx∗ + T x∗ = x∗ .
Proof 1.9.11 From (ii) and (iii), for any z ∈ S(K), it follows that the mapping T +z : K → E
satisfies all assumptions of Theorem 1.9.8. Thus the equation
Tx+z = x
has a unique solution τ = τ(z) ∈ K. For z1 , z2 ∈ S(K), using that
T τ(z1 ) + z1 = τ(z1 ),
T τ(z2 ) + z2 = τ(z2 ),
we get
hkτ(z1 ) − τ(z2 )k ≤ kT τ(z1 ) − T τ(z2 )k
= kz2 − z1 + τ(z1 ) − τ(z2 )k
≤ kz1 − z2 k + kτ(z1 ) − τ(z2 )k,
whereupon
1
kτ(z1 ) − τ(z2 )k ≤ kz1 − z2 k.
h−1
Hence, τ : S(K) → K is continuous. Then τS : K → K is continuous and τS(K) resides in
a compact subset of E. From here and from the Schauder fixed-point theorem, we conclude
that there exists an x∗ ∈ K such that
τSx∗ = x∗ .
Therefore
T τSx∗ + Sx∗ = τSx∗
or
T x∗ + Sx∗ = x∗ .
Theorem 1.9.12 Let E be a Banach space and K ⊂ E be a nonempty closed convex subset.
Suppose that T and S map K into E such that
(i) S is continuous, S(K) resides in a compact subset of E and
(ii) T : K → E is expansive and onto.

Then there exists a point x∗ ∈ K such that
Sx∗ + T x∗ = x∗ .
Proof 1.9.13 Let z ∈ S(K) be fixed. Take y ∈ K arbitrarily. Then y − z ∈ E. Since T : K → E

is onto, we conclude that there is an x ∈ K such that
T x + z = y.
Therefore (iii) of Theorem 1.9.10 holds. Consequently there exists a point x∗ ∈ K such that
Sx∗ + T x∗ = x∗ .
Theorem 1.9.14 Let X be a nonempty closed convex subset of a Banach space E. Suppose
that T and S map X into E such that
(i) S is continuous and S(X) resides in a compact subset of E.
(ii) T : X → E is expansive.
(iii) S(X) ⊂ (I − T )(E) and [x = T x + Sy, y ∈ X] =⇒ x ∈ X(or S(X) ⊂ (I − T )(X)).
Then there exists a point x∗ ∈ X such that
Sx∗ + T x∗ = x∗ .
Proof 1.9.15 For each y ∈ X, by (iii), we have that there is an x ∈ E such that
x − T x = Sy.
By Theorem 1.9.6, we have

x = (I − T )−1 Sy ∈ E.
Because S(X) resides in a compact subset of E and
(I − T )−1 : X → (I − T )(X)
is continuous, it follows that (I − T )−1 S(X) resides in a compact subset of X. Hence, and
applying the Schauder fixed-point theorem, it follows that there is an x∗ ∈ X such that
x∗ = (I − T )−1 Sx∗ ,
whereupon
T x∗ + Sx∗ = x∗ .
Theorem 1.9.16 Let X be a nonempty closed convex subset of a Banach space E and Y is
a nonempty compact subset of E such that X ⊂ Y , Y 6= X. Suppose that T and S map X into
E such that
(i) S is continuous and S(X) resides in Y .
(ii) T : X → E is linear, continuous and expansive, and T : X → Y is onto and
{x − z : x ∈ X, z ∈ S(X)} ⊂ Y. (1.19)
Then there exists an x∗ ∈ X such that
T x∗ + Sx∗ = x∗ .
Proof 1.9.17 Since Y is compact and S(X) resides in Y , we have that the first condition of
Theorem 1.9.14 holds. Because T : X → E is expansive, we have that the second condition
of Theorem 1.9.14 holds. Note that T −1 : Y → E exists, it is linear and contractive with a
constant l ∈ (0, 1). Take y0 ∈ Y arbitrarily. Define the sequence {yn }n∈N as follows.
yn+1 = T −1 yn − z,
[
n∈N {0}.
Using (1.19) and T : X → Y is onto, we have y1 = T −1 y0 − z ∈ Y . Hence, again using (1.19)

and T : X → Y is onto, we obtain y2 ∈ Y and so on, yn ∈ Y for any n ∈ N. Then
ky2 − y1 k = kT −1 y1 − T −1 y0 k
≤ lky1 − y0 k,
ky3 − y2 k = kT −1 y2 − T −1 y1 k
≤ lky2 − y1 k
≤ l 2 ky1 − y0 k.
Using the principle of mathematical induction, we get
kyn+1 − yn k ≤ l n ky1 − y0 k, n ∈ N.
Now, for m > n, m, n ∈ N, we find
kym − yn k ≤ kym − ym−1 k + · · · + kyn+1 − yn k
l m−1 + · · · + l n ky1 − y0 k

≤
∞
≤ l n ∑ l j ky1 − y0 k
j=0
ln
= ky1 − y0 k.
1−l
Therefore {yn }n∈N is a Cauchy sequence of elements of Y ⊂ E. Since E is a Banach space, it
follows that the sequence {yn }n∈N is convergent to an element y∗ ∈ E. Because {yn }n∈N ⊂ Y
and Y ⊂ E is compact, we have that y∗ ∈ Y . Thus
y∗ = T −1 y∗ − z
or
z∗ = T z∗ + z, z∗ = T −1 y∗ ∈ X.
Because z ∈ S(X) was arbitrarily chosen, we conclude that S(X) ⊂ (I − T )(X), i.e., the
third condition of Theorem 1.9.14 holds. Hence, and applying Theorem 1.9.14, it follows
that there exists an x∗ ∈ X such that
T x∗ + Sx∗ = x∗ .
For R > r > 0, denote
Pr = {x ∈ P : kxk < r},
∂ Pr = {x ∈ P : kxk = r},
Pr,R = {x ∈ P : r < kxk < R}.

Let Ω be a subset of P, T : Ω → E is an expansive mapping with a constant h > 1, and

F : Pr → E is a k-set contraction with 0 ≤ k < h − 1. By Theorem 1.9.6, it follows that
1
(I − T )−1 exists, and it is -Lipschitzian on (I − T )(Ω). Suppose that F(Pr ) ⊂ (I −
h−1
−1
T )(Ω). Then (I − T ) F : Pr → P is continuous and bounded, and for any bounded set
B in Pr , we have
1
(I − T )−1 F (B) ≤

α α(F(B))
h−1
k
≤ α(B),
h−1
k
i.e., (I − T )−1 F : Pr → P is a strict -set contraction. Assume that x 6= T x + Fx,
h−1
x ∈ ∂ Pr . Then
x 6= (I − T )−1 Fx, x ∈ ∂ Pr .
−1
Hence, i (I − T ) F, Pr , P is well defined.

1. If Pr
\
Definition 1.9.18 Ω 6= 0,
/ then we define the generalized fixed-point
index of the sum T + F on Pr Ω with respect to the cone P as follows.
\
i∗ (T + F, Pr Ω, P) = i (I − T )−1 F, Pr , P .
\
2. If Pr
\
Ω = 0,
/ then we define
i∗ (T + F, Pr Ω, P) = 0.
\
Using the properties of the fixed-point index, one can deduce some of the properties of the
generalized fixed-point index such as: normality, additivity, homotopy invariance, preser-
vation, the excision property, and solvability.
Theorem 1.9.19 Assume that T : Ω ⊂ P → E is an expansive operator with a constant
h > 1, F : Pr → E is a k-set contraction with 0 ≤ k < h − 1, and F(Pr ) ⊂ (I − T )(Ω).
Then we have the following results.
(1) If 0 ∈ Ω, kFx + T 0k ≤ (h − 1)kxk, and Fx + T x 6= x for all x ∈ ∂ Pr
\
Ω, then
i∗ (T + F, Pr Ω, P) = 1.
\
(2) If there exists a u0 ∈ P ∗ such that
Fx 6= (I − T )(x − λ u0 )
for all λ ≥ 0 and x ∈ ∂ Pr

\
(Ω + λ u0 ), then
i∗ (T + F, Pr Ω, P) = 0.
\
(3) If there exists a u0 ∈ P ∗ such that T (x − λ u0 ) ∈ P and

Fx 6≤ x − λ u0
for all λ ≥ 0 and x ∈ ∂ Pr
\
(Ω + λ u0 ), then
i∗ (T + F, Pr Ω, P) = 0.
\
Proof 1.9.20 (1) Since F Pr ⊂ (I −T )(Ω), we have that for any x ∈ Pr there is a y ∈ P

such that
Fx = (I − T )y,
so
y = (I − T )−1 Fx ∈ Ω.
Hence, for each x ∈ Pr , we get
T (I − T )−1 Fx + Fx = T (I − T )−1 Fx + (I − T )(I − T )−1 Fx

= (T + I − T ) (I − T )−1 Fx

= (I − T )−1 Fx.
Thus
kT (I − T )−1 Fx − T 0k = k(I − T )−1 Fx − Fx − T 0k

≤ k(I − T )−1 Fxk + kFx + T 0k, x ∈ Pr .

Since T is expansive with a constant h > 1, we obtain
kT (I − T )−1 Fx − T 0k ≥ hk(I − T )−1 Fxk, x ∈ Pr .

Therefore
hk(I − T )−1 Fxk ≤ k(I − T )−1 Fxk + kFx + T 0k, x ∈ Pr ,
whereupon
1
k(I − T )−1 Fxk ≤ kFx + T 0k
h−1
≤ kxk, x ∈ Pr .
Hence, and applying Theorem 1.8.6, we get
i∗ (T + F, Pr Ω, P) = i((I − T )−1 F, Pr , P)
\
= 1.
(2) Choose u0 ∈ P ∗ . Assume that there exist x0 ∈ ∂ Pr and λ0 ≥ 0 such that
x0 − λ0 u0 = (I − T )−1 Fx0 .
Then
Fx0 = (I − T )(x0 − λ0 u0 ),
which is a contradiction. Therefore
x − (I − T )−1 Fx 6= λ u0
for any x ∈ ∂ Pr and λ ≥ 0. Hence, and applying Theorem 1.8.8, we get
i∗ (T + F, Pr Ω, P) = i((I − T )−1 F, Pr , P)
\
= 0.
(3) Choose u0 ∈ P ∗ . Assume that there exist λ0 ≥ 0 and x0 ∈ ∂ Pr

\
(Ω + λ0 u0 ) such that
Fx0 = (I − T )(x0 − λ0 u0 ) ∈ Ω.
Then
x0 − λ0 u0 − Fx0 = T (x0 − λ0 u0 )
≤ 0,
which is a contradiction because T (x0 − λ0 u0 ) ∈ P. Therefore
Fx 6= (I − T )(x − λ u0 )
for any λ ≥ 0 and x ∈ ∂ Pr

\
(Ω + λ u0 ). Hence, and considering and (2), we obtain
i∗ (T + F, Pr Ω, P) = 0.
\

Theorem 1.9.21 Let Ω be a subset of P, 0 ∈ Ω and 0 < r < L < R. Let also T : Ω → E
be an expansive operator with a constant h > 1, F : PR →\ E be a k-set contraction with
0 ≤ k < h − 1 and F(PR ) ⊂ (I − T )(Ω). Assume that Pr,L Ω 6= 0, / PL,R Ω 6= 0,
\
/ and
there exists a u0 ∈ P ∗ such that the following conditions hold.
(a) kFx + T 0k ≤ (h − 1)kxk for all x ∈ ∂ Pr
\
Ω.
(b) Fx 6= (I − T )(x − λ u0 ) for all λ ≥ 0 and x ∈ ∂ PL

\
(Ω + λ u0 ).
(c) kFx + T 0k ≤ (h − 1)kxk for all x ∈ ∂ PR

\
Ω.
Then T + F has at least three fixed points x1 , x2 , x3 ∈ P such that
0 ≤ kx1 k ≤ r ≤ kx2 k < L < kx3 k ≤ R.
Proof 1.9.22 Without loss of generality, assume that T x + Fx 6= x on ∂ Pr
\
Ω and T x +
Fx 6= x on ∂ PR
\
Ω. Otherwise the conclusion has been proved. By Theorem 1.9.19, it
follows that
i∗ (T + F, Pr Ω, P) = 1,
\
i∗ (T + F, PR Ω, P) = 1,
\
i∗ (T + F, PL Ω, P) = 0.
\
By the additivity of the generalized fixed-point index, we get

0 = i∗ (T + F, PL Ω, P)
\
= i∗ (T + F, Pr,L Ω, P) + i∗ (T + F, Pr Ω, P)
\ \
= i∗ (T + F, Pr,L Ω, P) + 1,
\
whereupon
i∗ (T + F, Pr,L Ω, P) = −1,
\
and
1 = i∗ (T + F, PR Ω, P)
\
= i∗ (T + F, PL,R Ω, P) + i∗ (T + F, PL Ω, P)
\ \
= i∗ (T + F, PL,R Ω, P).
\
By the solvability of the generalized \

fixed-point index, we conclude that T + F has at\
least
three positive fixed points x1 ∈ Pr Ω, x2 ∈ (PL \Pr ) Ω and x3 ∈ (PR \PL ) Ω.
\
Theorem 1.9.23 Let Ω be a subset of P, 0 ∈ Ω, and 0 < r < L < R are real constants.
Let also T : Ω → E be an expansive operator with a constant h > 1, F : PR → \ E be a k-
set contraction with 0 ≤ k < h − 1, and F(PR ) ⊂ (I − T )(Ω). Assume that Pr,L Ω 6= 0,
/
PL,R Ω 6= 0/ and there exists a u0 ∈ P ∗ such that T (x − λ u0 ) ∈ P for all λ ≥ 0 and
\
x ∈ ∂ Pr
\
(Ω + λ u0 ) and the following conditions hold.
(a) Fx 6= x − λ u0 , x ∈ ∂ Pr , λ ≥ 0.
(b) kFx + T 0k ≤ (h − 1)kxk and T x + Fx 6= x, x ∈ ∂ PL
\
Ω.
(c) Fx 6= x − λ u0 , x ∈ PR , λ ≥ 0.
Then T + F has at least two fixed points x1 ∈ Pr,L Ω, x2 ∈ PL,R
\ \
Ω, i.e.,
r < kx1 k < L < kx2 k < R.
Proof 1.9.24 By Theorem 1.9.19, it follows that
i∗ (T + F, Pr Ω, P) = 0,
\
i∗ (T + F, PR Ω, P) = 0,
\
i∗ (T + F, PL Ω, P) = 1.
\
By the additivity of the generalized fixed-point index, we get
0 = i∗ (T + F, PR Ω, P)
\
= i∗ (T + F, PL Ω, P) + i∗ (T + F, PL,R Ω, P)
\ \
= 1 + i∗ (T + F, PL,R Ω, P),
\
whereupon
i∗ (T + F, PL,R Ω, P) = −1,
\
and
1 = i∗ (T + F, PL Ω, P)
\
= i∗ (T + F, Pr Ω, P) + i∗ (T + F, Pr,L Ω, P)
\ \
= i∗ (T + F, Pr,L Ω, P).
\
By the solvability of the generalized \ it follows that T + F has at least two

\ fixed-point index,
positive fixed points x1 ∈ Pr,L Ω and x2 ∈ PL,R Ω. This completes the proof.
Theorem 1.9.25 Let P ⊂ E be a normal cone with a constant N, and Ω is a subset of P

such that 0 ∈ Ω. Suppose that the following conditions hold.
(H1) Let T : Ω → E be an expansive operator with a constant h > 1 and F : P → E be a
k-set contraction with 0 ≤ k < h − 1. Assume that for some real constant d > 0 sufficiently
∗
large, F(Pd ) ⊂ \(I − T )(Ω) and there is a u0 ∈ P such that T (x − λ u0 ) ∈ \ P for all
λ ≥ 0, x ∈ Pd (Ω + λ u0 ), there exists a constant 0 < ρ < d such that Pρ,d Ω 6= 0, /
sup{kFx + T 0k : x ∈ P, kxk = ρ} ≤ (h − 1)ρ, (1.20)
and
inf{kT x + Fxk : x ∈ P, kxk = ρ} > ρ. (1.21)
(H2) There exists two operators F1 , F2 : P → E such that
Fx ≥ Fi x, x ∈ P, i = 1, 2, (1.22)
and there exist real constants c1 , c2 > 0 such that
mi = inf{kFi xk : x ∈ P, kxk = ci } > 0, i = 1, 2. (1.23)
(H3) Suppose that for any x ∈ P ∗ and t ∈ (0, 1), there exist ξi = ξi (t, x) > 0, i = 1, 2, such
that
F1 (tx) ≥ t(1 + ξi )F1 x, (1.24)
where
c1 N 2
lim ξ1 (x,t) > − 1,
t→0+ m1
∗
uniformly for x ∈ P , and
F2 (tx) ≤ t(1 − ξ2 )F2 x, (1.25)
where
m2
limξ2 (x,t) > 1 − ,
t→0 c2 N 2
uniformly for x ∈ P ∗ .
Then T + F has at least two fixed points x1 , x2 ∈ P such that
0 < kx1 k < ρ < kx2 k < d.

Proof 1.9.26 By (1.25), for all t > 1, we have

1
F2 x = F2 tx
t

1 1
≤ 1 − ξ2 tx, F2 (tx), x ∈ P ∗.
t t
Thus −1
1
F2 (tx) ≥ t 1 − ξ2 tx, F2 x
t
for any t > 1 and x ∈ P ∗ . By the assumptions (H1) and (H3), it follows that we can choose
real numbers r < R ≤ d satisfying
0<r < min{ρ, c1 },
R > max{ρ, c2 },
and
Pr,ρ Pρ,R
\ \
Ω 6= 0,
/ Ω 6= 0,
/
such that
c1 N 2

r
ξ1 x, > − 1, x ∈ P ∗, (1.26)
c1 m1
and c
2 m2
ξ2 x, > 1− , x ∈ P ∗. (1.27)
R c2 N 2
Assume that there exist x1 ∈ ∂ Pr and λ0 ∈ [0, 1] such that
Fx1 ≤ x1 − λ0 u0 .
By (1.22) and (1.24), we obtain
x1 − λ0 u0 ≥ Fx1
≥ F1 x1

r c1 x1
= F1 ·
c1 r

r c1 x1 r c1 x1
≥ 1 + ξ1 , F1 .
c1 r c1 r
Now, using that P is a normal cone with a constant N and (1.26), we get
1
kx1 k ≥ kx1 − λ0 u0 k
N
−1
r c1 x1 r c1 x1
≥ 1 + ,

ξ1 1
c1 N 2
F
r c1 r
r c1 N 2
> m1
c1 N 2 m1
= r,
which contradicts x1 ∈ ∂ Pr . Thus
Fx 6≤ x − λ u0 , ∀x ∈ ∂ Pr , ∀λ ∈ [0, 1]. (1.28)
Assume that there exist x2 ∈ ∂ PR and λ0 ∈ [0, 1] such that
Fx2 ≤ x2 − λ0 u0 .
Now, using (1.22) and (1.27), we obtain
x2 − λ0 u0 ≥ Fx2
≥ F2 x2

R c2 x2
= F2
c2 R
R c −1 c x
2 2 2
≥ 1 − ξ2 x2 , F2 .
c2 R R
Hence,
1
kx2 k ≥ kx2 − λ0 u0 k
N
R c −1 c x
2 2 2
≥ 1 − ξ2 x 2 , kF2 k
c2 N 2 R R
R c2 N 2
≥ m2
c2 N 2 m2
= R,
which contradicts x2 ∈ ∂ PR . Therefore
Fx 6≤ x − λ u0 , ∀x ∈ ∂ PR , ∀λ ∈ [0, 1]. (1.29)
Suppose that there exists an x3 ∈ ∂ PR such that
kFx3 + T 0k > (h − 1)kx3 k
or
T x3 + Fx3 = x3 .
By (1.20) and (1.21), it follows that
kFx3 + T 0k ≤ (h − 1)ρ
and
kFx3 + T x3 k > ρ,
which is a contradiction. Consequently
kFx + T 0k ≤ (h − 1)kxk and Fx + T x 6= x, ∀x ∈ ∂ Pρ . (1.30)
By (1.28), (1.29) and (1.30), and Theorem 1.9.23, it follows that T + F has at least two
fixed points x1 , x2 ∈ P ∗ such that
r < kx1 k < ρ < kx2 k < R.
Theorem 1.9.27 Suppose that (H1) holds in Theorem 1.9.25. The second operator F in
the sum T + F can be written as F = F1 + F2 , where F1 : P → P is increasing e-concave,
F2 : P → P is increasing e-convex. If there exist Λi , i = 1, 2, such that
Fi (Λ0 e) ≥ Λi kFi (Λ0 e)ke, (1.31)
N3
lim η1 (x,t) ≥ − 1, uniformly for x ∈ Ce , (1.32)
t→0+ Λ1 kF1 (Λ0 e)kkek
Λ2
lim η2 (x,t) ≥ 1 − kF2 (Λ0 e)kkek, uniformly for x ∈ Ce , (1.33)
t→0+ N3
where η1 and η2 are given in Definition 1.9.4 and
Λ0 = min{Λ1 , Λ2 },
then T + F has at least two fixed points x1 , x2 ∈ P ∗ , such that
0 < kx1 k < ρ < kx2 k.

Proof 1.9.28 Define in E the cone K ⊂ P as follows
K = {x ∈ P : ∃Lx > 0 : Lx e ≥ x ≥ Λ0 kxke},

∗
K = K \{0}.
We have K ∗ ⊂ Ce and the cone K is a normal cone with the constant N. Since F1 is e-
concave, F2 is e-convex, we get F1 x, F2 x ∈ Ce , x ∈ K ∗ , and hence, Fx ∈ Ce , x ∈ K ∗ . Now,
using that Fi , i = 1, 2, are increasing, we have
Fi x ≥ Fi (Λ0 e)
≥ Λi kFi (Λ0 e)ke, i = 1, 2,
for any x ∈ ∂ K , kxk = 1. Therefore

1
kFi xk ≥ Λi kFi (Λ0 e)kkek, x ∈ ∂K , kxk = 1, i = 1, 2.
N
Since Fi (Λ0 e) ∈ Ce , i = 1, 2, we get
kFi (Λ0 e)k > 0, i = 1, 2.
Hence,
mi = inf{kFi xk : x ∈ ∂ K , kxk = 1}
1
≥ Λi kFi (Λ0 e)kkek
N
> 0, i = 1, 2.
Consequently (1.23) is satisfied for c1 = c2 = 1 and the cone K . Also,
F1 (tx) ≥ t(1 + η1 )F1 x,
F2 (tx) ≤ t(1 − η2 )F2 x, ∀(x,t) ∈ K ∗ × (0, 1).
By (1.32) and (1.33), we get
N3 N2
−1 = 1
−1
Λ1 kF1 (Λ0 e)kkek N Λ1 kF1 (Λ0 e)kkek
N2
≥ − 1,
m1
1
Λ2 N Λ2 kF2 (Λ0 e)kkek
1− kF2 (Λ0 e)kkek = 1 −
N3 N2
m2
≥ 1− ,
N2
respectively. Therefore (1.24) and (1.25) hold for the cone K . Hence, and applying Theo-
rem 1.9.25, it follows that T + F has at least two fixed points x1 , x2 ∈ K ∗ such that
0 < kx1 k < ρ < kx2 k.
Theorem 1.9.29 Assume (H1) in Theorem 1.9.25 holds. The second operator F in the
sum T + F can be written as F = F1 + F2 , where Fi : P → P, i = 1, 2, F1 is α-concave,
0 < α < 1, and F2 is β -convex, β > 1. If there exist positive constants c1 , c2 such that (1.23)
holds, then T + F has at least two fixed points x1 , x2 ∈ P ∗ , such that
0 < kx1 k < ρ < kx2 k.
Proof 1.9.30 Define
ξ1 (x,t) = t α−1 − 1,
ξ2 (x,t) = 1 − t β −1 , (x,t) ∈ P ∗ × (0, 1).
Then
lim ξ1 (x,t) = ∞, uniformly for x ∈ P ∗,

t→0+
lim ξ2 (x,t) = 1, uniformly for x ∈ P ∗.

t→0+
Thus (H3) in Theorem 1.9.25 holds. Hence, and applying Theorem 1.9.25, it follows that
T + F has at least two fixed points x1 , x2 ∈ P ∗ such that
0 < kx1 k < ρ < kx2 k.
Theorem 1.9.31 Assume that (H1) in Theorem 1.9.25 holds. The second operator F in the
sum T + F can be written as F = F1 + F2 , where Fi : P → P, i = 1, 2, F1 is increasing
α-concave, 0 ≤ α < 1, and F2 : P → P is increasing β -convex, β > 1. If there exists
e ∈ P ∗ such that Fi e ∈ P ∗ , i = 1, 2, then T + F has at least two fixed points x1 , x2 ∈ P ∗ ,
such that
0 < kx1 k < ρ < kx2 k, xi > kxi ke, i = 1, 2.
Proof 1.9.32 Define in E the cone Q ⊂ P by
Q = {x ∈ P : x ≥ kxke}.
We have that Q is a normal cone with the constant N. For any x ∈ ∂ Q, kxk = 1, we have
x ≥ e. Since Fi , i = 1, 2, are increasing and Q is normal, we get
Fi x ≥ Fi e,
1
kFi xk ≥ kFi ek
N
> 0, i = 1, 2, x ∈ ∂ Q, kxk = 1.
Therefore (H2) in Theorem 1.9.25 holds. Hence, and applying Theorem 1.9.29, it follows
that T + F has at least two fixed points x1 , x2 ∈ P ∗ such that
0 < kx1 k < ρ < kx2 k, xi > kxi ke, i = 1, 2.

2
Applications to ODEs
2.1 Periodic Solutions for First Order ODEs

In this section we will prove some criteria for multiple periodic solutions for a class of first
order ODEs. Let ω > 0 be arbitrarily chosen. For a continuous ω-periodic function a we
denote
1 ω
Z
[a] = a(τ)dτ.
ω 0
We will start with the following useful lemma.
Lemma 2.1.1 Let ε and φ be continuous ω-periodic functions and [ε] 6= 0. Then there
exists a unique ω-periodic solution to the equation
x0 = −ε(t)x + φ (t), t ∈ R, (2.1)
for which the following representation
e−[ε]ω ω R t+s
Z
x(t) = −[ε]ω
e t ε(τ)dτ φ (t + s)ds, t ∈ R, (2.2)
1−e 0
holds. Also, there exists a unique ω-periodic solution to the equation
x0 = ε(t)x − φ (t), t ∈ R, (2.3)
for which we have the following representation
1
Z ω Rt
x(t) = e t+s ε(τ)dτ φ (t + s)ds, t ∈ R. (2.4)
1 − e−[ε]ω 0
Proof 2.1.2 Observe that
Z t+ω+s Z t+s
ε(τ)dτ = ε(τ)dτ,
t+ω t
Z t+ω Z ω Z t+ω
ε(τ)dτ = ε(τ)dτ + ε(τ)dτ
0 0 ω
Z ω Z t
= ε(τ)dτ + ε(τ)dτ, t, s ∈ R.
0 0
83
84 Applications to ODEs
Assume that the equation (2.1) has two ω-periodic solutions x1 and x2 . Set
v = x1 − x2 .
Then v is an ω-periodic solution of the equation
v0 = −ε(t)v, t ∈ R.
Hence, Rt
v(t) = v(0)e− 0 ε(τ)dτ
and
Rt
v(0)e− 0 ε(τ)dτ = v(t)
= v(t + ω)
R t+ω
= v(0)e− 0 ε(τ)dτ
Rt Rω
= v(0)e− 0 ε(τ)dτ e− 0 ε(τ)dτ
, t ∈ R.
Therefore [ε] = 0, which is a contradiction. Now, we will prove that x, defined by (2.2), is
an ω-periodic solution of equation (2.1). We have
e−[ε]ω
Z ω R t+ω+s
ε(τ)dτ
x(t + ω) = e t+ω φ (t + ω + s)ds
1 − e−[ε]ω 0
e−[ε]ω
Z ω R t+s
ε(τ)dτ
= et φ (t + s)ds
1 − e−[ε]ω 0
= x(t), t ∈ R,
and
e−[ε]ω − R0t ε(τ)dτ

Z ω R t+s
ε(τ)dτ
x(t) = e e 0 φ (t + s)ds
1 − e−[ε]ω 0
e−[ε]ω − R0t ε(τ)dτ

Z t+ω Ry
ε(τ)dτ
= e e 0 φ (y)dy, t ∈ R.
1 − e−[ε]ω t
Periodic Solutions for First Order ODEs 85
From here,
e−[ε]ω
Z t+ω Ry
− 0t ε(τ)dτ
R
x0 (t) = (−ε(t))e e 0 ε(τ)dτ
φ (y)dy
1 − e−[ε]ω t
e−[ε]ω − R0t ε(τ)dτ R t+ω ε(τ)dτ Rt

0 ε(τ)dτ φ (t)

+ e e 0 φ (t + ω) − e
1 − e−[ε]ω
e−[ε]ω R0ω ε(τ)dτ

= −ε(t)x(t) + −[ε]ω
e − 1 φ (t)
1−e
= −ε(t)x(t) + φ (t), t ∈ R.
As in the above, one can prove that equation (2.3) has a unique ω-periodic solution given
by (2.4). This completes the proof.
Let ε, r1 , r2 , ρ, R, M, M0 , M1 , M2 , m and k be positive constants and a be a non-negative
continuous ω-periodic function such that
0 < r1 < ρ < r2 < R,
(2.5)
e−[a]ω Mω m k

e Mωri + M0 ω + M1 ωri + M2 ωri < ri , i = 1, 2,
1 − e−[a]ω
[a] ≤ M, a 6≡ 0, (2.6)
e−[a]ω Mω
e (MωR + M0 ω + M1 ωRm + M2 ωRk ) < ρ. (2.7)
1 − e−[a]ω
Consider the equation
x0 = f (t, x), t ∈ R, (2.8)
where
(H1) f : R × [0, ∞) → [0, ∞) is a continuous function that is ω-periodic in t and
0 ≤ f (t, x) ≤ a0 (t) + a1 (t)xm + a2 (t)xk , t ∈ R, x ∈ [0, ∞),
and ai , i = 0, 1, 2, are non-negative continuous ω-periodic functions such that
[ai ] ≤ Mi , i = 0, 1, 2.
Theorem 2.1.3 Assume that (H1), (2.5), (2.6) and (2.7) hold. Then equation (2.8) has at
least three solutions x1 , x2 , x3 such that
0 < max x1 (t) ≤ r1 ≤ max x2 (t) < ρ < max x3 (t) ≤ r3 .
t∈[0,ω] t∈[0,ω] t∈[0,ω]
Proof 2.1.4 Consider the Banach space E = C (ω) of all continuous ω-periodic functions,
endowed with the maximum norm
kxk = max |x(t)|,
t∈[0,ω]
the positive cone

P = {x ∈ E : x(t) ≥ 0, t ∈ R},
and the conical shells
PR = {x ∈ P : kxk < R},
Pρ = {x ∈ P : kxk < ρ},
Pri = {x ∈ P : kxk < ri }, i = 1, 2.

For x ∈ E, define the operators
T x(t) = (1 + ε)x(t) + εR,
e−[a]ω
Z ω R
t+s
a(τ)dτ
Fx(t) = −εR − ε e t a(t + s)x(t + s)
1 − e−[a]ω 0

+ f (t + s, x(t + s)) ds, t ∈ R.
Using Lemma 2.1.1, if x ∈ E is a fixed point of T + F, we have

x(t) = T x(t) + Fx(t)
= (1 + ε)x(t) − εR + εR
e−[a]ω
Z ω R t+s

a(τ)dτ
−ε et a(t + s)x(t + s)
1 − e−[a]ω 0

+ f (t + s, x(t + s)) ds,
whereupon
e−[a]ω
Z ω R t+s

a(τ)dτ
x(t) = e t a(t + s)x(t + s) + f (t + s, x(t + s)) ds, t ∈ R,
1 − e−[a]ω 0
and x satisfies (2.8). We will check all conditions of Theorem 1.9.21.

1. We have T : PR → E. Also, for any x, y ∈ P, we get

kT x − Tyk = (1 + ε)kx − yk,
that is, T : PR → E is expansive with a constant h = 1 + ε > 1.
2. We will show that F : PR → E is a 0-set contraction. Let x ∈ PR and

|Fx(t)| = − εR
e−[a]ω
Z ω R
t+s
a(τ)dτ

−ε e t a(t + s)x(t + s) + f (t + s, x(t + s)) ds
1 − e−[a]ω 0
≤ εR
e−[a]ω
Z ω R t+s
a(τ)dτ
+ε et (a(t + s)x(t + s) + f (t + s, x(t + s))) ds
1 − e−[a]ω 0
≤ εR
e−[a]ω Mω
Z ω
+ε e a(t + s)x(t + s) + a0 (t + s)
1 − e−[a]ω 0

+a1 (t + s)(x(t + s))m + a2 (t + s)(x(t + s))k ds
≤ εR
e−[a]ω Mω
+ε e (RMω + M0 ω + M1 ωRm + M2 ωRk ), t ∈ R.
1 − e−[a]ω
Hence,
kFxk ≤ εR
e−[a]ω Mω
+ε e (RMω + M0 ω + M1 ωRm + M2 ωRk ).
1 − e−[a]ω
This implies that F(PR ) is uniformly bounded. Also, we have
Fx(t) = −εR
e−[a]ω − R0t a(τ)dτ

Z ω R t+s

a(τ)dτ
−ε e e 0 a(t + s)x(t + s)
1 − e−[a]ω 0

+ f (t + s, x(t + s)) ds
= −εR
e−[a]ω − R0t a(τ)dτ

Z t+ω Ry
a(τ)dτ
−ε e e 0 (a(y)x(y) + f (y, x(y)))dy, t ∈ R,
1 − e−[a]ω t
and
d e−[a]ω t+ω R y
Rt Z
Fx(t) = ε −[a]ω
a(t)e− 0 a(τ)dτ e 0 a(τ)dτ (a(y)x(y) + f (y, x(y)))dy
dt 1−e t
e−[a]ω − R0t a(τ)dτ R t+ω a(τ)dτ

−ε e e0 (a(t)x(t) + f (t, x(t)))
1 − e−[a]ω
Rt

−e 0 a(τ)dτ (a(t)x(t) + f (t, x(t)))
e−[a]ω
Z ω R t+s
a(τ)dτ
=ε a(t) et (a(t + s)x(t + s) + f (t + s, x(t + s)))ds
1 − e−[a]ω 0
e−[a]ω [a]ω
−ε e − 1 (a(t)x(t) + f (t, x(t)))
1 − e−[a]ω
e−[a]ω
Z ω R t+s
a(τ)dτ
=ε a(t) et (a(t + s)x(t + s) + f (t + s, x(t + s)))ds
1 − e−[a]ω 0
−ε(a(t)x(t) + f (t, x(t))), t ∈ R,
−[a]ω

d
Fx(t) = ε e
Z ω R
t+s
e t a(τ)dτ (a(t + s)x(t + s) + f (t + s, x(t + s)))ds

dt 1 − e−[a]ω a(t)
0

−ε(a(t)x(t) + f (t, x(t)))
e−[a]ω
Z ω R t+s

a(τ)dτ
≤ε a(t) e t a(t + s)x(t + s)
1 − e−[a]ω 0

+a0 (t + s) + a1 (t + s)(x(t + s))m + a2 (t + s)(x(t + s))k ds

+ε a(t)x(t) + a0 (t) + a1 (t)(x(t))m + a2 (t)(x(t))k
e−[a]ω
≤ε NωeMω (MR + M0 + M1 Rm + M2 Rk )
1 − e−[a]ω
+ε(NR + N0 + N1 Rm + N2 Rk ), t ∈ R,
where
N = max a(t), Ni = max ai (t), i = {0, 1, 2}.
t∈[0,ω] t∈[0,ω]
Consequently
e−[a]ω
k(Fx)0 k ≤ ε NωeMω (MR + M0 + M1 Rm + M2 Rk )
1 − e−[a]ω
+ε(NR + N0 + N1 Rm + N2 Rk ).
Now, using the Arzela-Ascoli Theorem, we conclude that F : PR → E is a com-
pletely continuous mapping. Thus F : PR → E is a 0-set contraction.
3. For i = 1, 2 and x ∈ ∂ Pri , we have
e−[a]ω
Z ω R
t+s
a(τ)dτ

|Fx(t) − T 0(t)| = − ε

−[a]ω
e t a(t + s)x(t + s) + f (t + s, x(t + s)) ds
1−e 0
e−[a]ω
Z ω R t+s
a(τ)dτ
=ε et (a(t + s)x(t + s) + f (t + s, x(t + s)))ds
1 − e−[a]ω 0
e−[a]ω Mω

≤ε e Mωri
1 − e−[a]ω
Z ω
m k
+ (a0 (t + s) + a1 (t + s)(x(t + s)) + a2 (t + s)(x(t + s)) )ds
0
e−[a]ω Mω

≤ε e Mωri
1 − e−[a]ω

+M0 ω + M1 ωrim + M2 ωrik
≤ εri
= εkxk, t ∈ R.
Hence, conditions (a) and (c) of Theorem 1.9.21 are satisfied.

4. Now we will prove condition (b) of Theorem 1.9.21. Let u0 ∈ P ∗ and as-
sume that there exist x1 ∈ ∂ Pρ and λ1 ≥ 0 with x1 − λ1 u0 ∈ ∂ Pρ such that
Fx1 = (I − T )(x1 − λ1 u0 ).
Let t1 ∈ [0, ω] be such that
(x1 − λ1 u0 )(t1 ) = ρ.
Then
e−[a]ω Mω
εR + ε e (RMω + M0 ω + M1 ωRm + M2 ωRk ) ≥ −(Fx1 )(t1 )
1 − e−[a]ω
= −(I − T )(x1 − λ1 u0 )(t1 )
= ε(x1 − λ1 u0 )(t1 ) + εR
= ερ + εR,
i.e.,
e−[a]ω Mω
e (RMω + M0 ω + M1 ωRm + M2 ωRk ) ≥ ρ,
1 − e−[a]ω
which is a contradiction. Thus condition (b) of Theorem 1.9.21 is satisfied.
5. Now we will prove that
F(PR ) ⊂ (I − T )(PR ). (2.9)
Let z ∈ PR be fixed and consider the nonlinear equation on PR
x(t) = T x(t) + Fz(t)
= Ax(t), t ∈ R.
For any x, y ∈ PR , we get
kAx − Ayk = kT x − Tyk
= (1 + ε)kx − yk,
i.e., A : PR → E is an expansive mapping. We will apply Theorem 1.9.8. For
this aim we will prove that
PR ⊂ A(PR ).
For any given v ∈ PR , consider the following equation on PR

Ax(t) = v(t) or T x(t) = v(t) − Fz(t), t ∈ R. (2.10)
We have
v(t) − Fz(t) ≥ 0, t ∈ R,
and
e−[a]ω
v(t) − Fz(t) = v(t) + εR + ε
1 − e−[a]ω
Z ω R t+s
a(τ)dτ
× et (a(t + s)z(t + s) + f (t + s, z(t + s)))ds
0
e−[a]ω
≤ v(t) + εR + ε
1 − e−[a]ω
Z ω R
t+s
a(τ)dτ
× e t a(t + s)z(t + s) + a0 (t + s) + a1 (t + s)(z(t + s))m
0

+a2 (t + s)(z(t + s))k ds
e−[a]ω Mω

≤ R + εR + ε e MωR + M0 ω
1 − e−[a]ω

+M1 ωRm + M2 ωRk
= R1 , t ∈ R.
Thus, R1 > R and

P.
\
v − Fz ∈ B(R1 )
Note that T : PR → T (PR ) is a bijection, where
T (PR ) = {x ∈ P : kxk ≤ R1 }.
Therefore there is a x ∈ PR that satisfies (2.10). Hence, and applying Theorem
1.9.8, it follows that there is a w ∈ PR such that
Aw(t) = w(t), t ∈ R,
whereupon
w(t) = Tw(t) + T z(t), t ∈ R.
Since z ∈ PR was arbitrarily chosen, we get (2.9).
Now we apply Theorem 1.9.21 for r = r1 , R = r2 , L = ρ and Ω = PR . Therefore T + F has

at least three fixed points x1 , x2 , x3 such that
t∈[0,ω] t∈[0,ω] t∈[0,ω]

Let ε, M, M0 , M1 , M2 , L, r, R be positive constants such that
r < L < R, ε > 1, (2.11)
e−[a]ω Mω
e (MωL + M0 ω + M1 ωLm + M2 ωLk ) ≤ L. (2.12)
1 − e−[a]ω
Theorem 2.1.5 Suppose (2.6), (2.11), (2.12) and (H1). Then equation (2.8) has at least
two solutions x1 , x2 ∈ P such that
r < kx1 k < L < kx2 k < R.
Proof 2.1.6 Let E, P, PR be defined as in the proof of Theorem 2.1.3. Define on E the
operators
T x(t) = −(1 + ε)x(t) − εR,
e−[a]ω
Z ω R t+s
a(τ)dτ
Fx(t) = εR + ε et (a(t + s)x(t + s) + f (t + s, x(t + s)))ds, t ∈ R.
1 − e−[a]ω 0
Let also
Pr = {x ∈ P : kxk < r},
PL = {x ∈ P : kxk < L},
Pr,L = {x ∈ P : r < kxk < L},
PL,R = {x ∈ P : L < kxk < R},
Ω = PR .
We have
Pr,L
\
Ω 6= 0,
/
PL,R
\
Ω 6= 0.
/
We will check all conditions of Theorem 1.9.23.
1. As in the proof of Theorem 2.1.3, we have
F(PR ) ⊂ (I − T )(PR ).
Also, we have that F is a 0-set contraction.

2. Let u0 ∈ P ∗ and assume that there are a λ1 ≥ 0 and an x1 ∈ ∂ Pr such that
Fx1 ≤ x1 − λ1 u0 .
Then
r ≥ x1 (t) − λ1 u0 (t)
= Fx1 (t)
= εR
e−[a]ω
Z ω R t+s
a(τ)dτ
+ε et (a(t + s)x1 (t + s) + f (t + s, x1 (t + s)))ds
1 − e−[a]ω 0
≥ εR, t ∈ R,
which contradicts (2.11). Then condition (a) of Theorem 1.9.23 is satisfied. As

in the above, one can prove that condition (c) of Theorem 1.9.23 is satisfied.
3. Let x ∈ ∂ PL . Then
e−[a]ω
Z ω R t+s
a(τ)dτ
|Fx(t) + T 0(t)| = ε et (a(t + s)x(t + s) + f (t + s, x(t + s)))ds
1 − e−[a]ω 0
e−[a]ω Mω
Z ω
≤ε e a(s)x(s) + a0 (s)
1 − e−[a]ω 0

m k
+a1 (s)(x(s)) + a2 (s)(x(s)) ds
e−[a]ω Mω
≤ε e (MωL + M0 ω + M1 ωLm + M2 ωLk )
1 − e−[a]ω
≤ εL
= εkxk, t ∈ R.
Hence,
kFx + T 0k ≤ εkxk.
Now, assume that x1 ∈ ∂ PL PR is such that
\
x1 = T x1 + Fx1 .
Let t1 ∈ [0, ω] be such that

L = x1 (t1 ).
Then
L = x1 (t1 )
= −(1 + ε)x1 (t1 ) − εR + εR
e−[a]ω R t1 +s
Z ω
a(τ)dτ
+ε e t1 (a(t1 + s)x(t1 + s) + f (t1 + s, x(t1 + s)))ds
1 − e−[a]ω 0
e−[a]ω
Z ω R t1 +s

a(τ)dτ
≤ ε e t1
a(t1 + s)x(t1 + s)
1 − e−[a]ω 0

+a0 (t1 + s) + a1 (t1 + s)(x(t1 + s))m + a2 (t1 + s)(x(t1 + s))k ds − εL,
whereupon
e−[a]ω Mω
(1 + ε)L < ε e (MLω + M0 ω + M1 ωLm + M2 ωLk )
1 − e−[a]ω
or
ε e−[a]ω Mω
L< e (MLω + M0 ω + M1 ωLm + M2 ωLk ),
1 + ε 1 − e−[a]ω
which contradicts (2.12). Therefore, condition (b) of Theorem 1.9.23 holds.
By Theorem 1.9.23, it follows that the operator T + F has at least two fixed points x1 , x2 ∈
P such that
r < kx1 k < L < kx2 k < R.
Suppose that b, di , i = 0, 1, 2, are non-negative continuous ω-periodic functions, di 6≡ 0,

i = 0, 1, 2. Let also, ε, C0 , C1 , C2 , B, ρ, d, m and k be positive constants such that
0 < ρ < d, 0 < m < 1, k > 1, (2.13)

1
(C0 ω +C1 ωρ m +C2 ωρ k + Bωρ) < ρ, (2.14)
1 − e−[b]ω
1
(C0 ω +C1 ωd m +C2 ωd k + Bωd) < d + 1, (2.15)
1 − e−[b]ω

2m 1 − e−[b]ω
< 1, (2.16)
ε[di ]ωe−Bω
ε[d2 ]ωe−Bω
> 1, (2.17)
1 − e−[b]ω 2k
e−Bω ω
b(t) ≤ B, [d0 ] ≥ d + 1 t ∈ R. (2.18)
1 − e−[b]ω
Consider equation (2.8), where the function f satisfies the following conditions:
(H2) f : R × [0, ∞) → (−∞, 0],
| f (t, x)| ≤ c0 (t) + c1 (t)xm + c2 (t)xk ,
f (t, x) ≤ −d0 (t) − d1 (t)xm − d2 (t)xk , t ∈ R, x ∈ [0, ∞),
where ci , i = 0, 1, 2, are non-negative continuous ω-periodic functions and
ci (t) ≤ Ci , t ∈ R, i = 0, 1, 2.
Theorem 2.1.7 Suppose (H2), (2.13)-(2.18) hold. Then equation (2.8) has at least two
non-negative continuous ω-periodic solutions x1 , x2 such that
0 < kx1 k < ρ < kx2 k < d.
Proof 2.1.8 Let E and P be as in the proof of Theorem 2.1.3. Let also
Pρ = {x ∈ P : kxk < ρ},
Pd = {x ∈ P : kxk < d},
Pρ,d = {x ∈ P : ρ < kxk < d},
Ω = Pd .
On P define the operators
T x(t) = (1 + ε)x(t) + ε(d + 1),
Fx(t) = −ε(d + 1)
1
Z ω Rt
−ε e t+s b(τ)dτ (−b(t + s)x(t + s) + f (t + s, x(t + s))) ds, t ∈ R.
1 − e−[b]ω 0
Note that if x ∈ P is a fixed point of the operator T + F, using Lemma 2.1.1, we have that
x is a non-negative continuous ω-periodic solution of equation (2.8). We will use Theorem
1.9.25.
1. Firstly, we will prove that
F(Pd ) ⊂ (I − T )(Pd ). (2.19)
Let z ∈ Pd be arbitrarily chosen. For x ∈ Pd , consider the equation
x = T x + Fz
= Ax.
We have
= (1 + ε)kx − yk, x, y ∈ Pd .
Therefore A : Pd → E is an expansive operator. Now we will apply Theorem

1.9.8. For this aim we will prove that
Pd ⊂ A(Pd ).
For any v ∈ Pd , consider the following equation on Pd
v = Au
= Tu + Fz
or
Tu = v − Fz. (2.20)
We have, using (2.15),

v(t) − Fz(t) = v(t) + ε(d + 1)
1
Z ω Rt
+ε e t+s b(τ)dτ ( f (t + s, z(t + s)) − b(t + s)z(t + s))ds
1 − e−[b]ω 0
1
≥ ε(d + 1) − ε (C0 ω +C1 ωd m +C2 ωd k + Bωd)
1 − e−[b]ω
≥ 0, t ∈ R,
and
1
|v(t) − Fz(t)| ≤ v(t) + ε(d + 1) + ε
1 − e−[b]ω
Z ω Rt
× e t+s b(τ)dτ (| f (t + s, z(t + s))| + b(t + s)z(t + s)) ds
0
Z ω
1
≤ d + ε(d + 1) + ε c0 (t + s) + c1 (t + s)(z(t + s))m
1 − e−[b]ω 0

k
+c2 (t + s)(z(t + s)) + b(t + s)z(t + s) ds
≤ d + ε(d + 1)
1
m k

+ε C0 ω +C1 ωd +C2 ωd + Bωd
1 − e−[b]ω
= d1 , t ∈ R.
Thus
P.
\
v − Fz ∈ B(d1 )
Since
T : Pd → T (Pd )
= {x ∈ P : kxk ≤ d1 }
is a bijection, there is a u ∈ Pd such that (2.20) holds. Hence, and applying
Theorem 1.9.8, it follows that there is a w ∈ Pd so that
w(t) = Tw(t) + Fz(t), t ∈ R,
and (2.19) holds. As in the proof of Theorem 2.1.3, we have that F is a 0-set
contraction.
2. Let now x ∈ ∂ Pρ . Then, using (2.14), we get

1
Z ω R
t
t+s b(τ)dτ ( f (t + s, x(t + s)) − b(t + s)x(t + s))ds

|Fx(t) + T 0(t)| = ε −[b]ω
e
1−e 0

1
Z ω R
t
≤ε e t+s b(τ)dτ c (t + s) + c (t + s)(x(t + s))m
0 1
1 − e−[b]ω 0

k
+c2 (t + s)(x(t + s)) + b(t + s)x(t + s) ds
1
≤ε (C0 ω +C1 ωρ m +C2 ωρ k + Bωρ)
1 − e−[b]ω
≤ ερ, t ∈ R,
or
kFx + T 0k ≤ ερ,
and again using (2.14), we get
kFx + T xk ≥ (1 + ε)ρ
1
−ε (C0 ω +C1 ωρ m +C2 ωρ k + Bωρ)
1 − e−[b]ω
> (1 + ε)ρ − ερ
= ρ.
Thus, conditions (1.20) and (1.21) of Theorem 1.9.25 hold.
3. For x ∈ P, define the operators
1
Z ω Rt
F1 x(t) = ε e t+s b(τ)dτ d1 (t + s)(x(t + s))m ds,
1 − e−[b]ω 0
1
Z ω Rt
F2 x(t) = ε e t+s b(τ)dτ d2 (t + s)(x(t + s))k ds, t ∈ R.
1 − e−[b]ω 0
Then
Z ω
1
Fx(t) ≥ −ε(d + 1) + ε d0 (t + s) + d1 (t + s)(x(t + s))m
1 − e−[b]ω 0

+d2 (t + s)(x(t + s))k + b(t + s)x(t + s) ds
1
Z ω Rt
≥ −ε(d + 1) + ε e t+s b(τ)dτ d0 (t + s)ds
1 − e−[b]ω 0
e−Bω ω
≥ −ε(d + 1) + ε [d0 ]
1 − e−[b]ω
≥ 0, t ∈ R.
Hence,
Fx ≥ Fi x, i = 1, 2, x ∈ P.
Take c1 = c2 = 1. Let m1 and m2 be defined as in Theorem 1.9.25. Then
m Z ω
e−Bω 1
m1 ≥ ε d1 (s)ds
1 − e−[b]ω 2 0
m
e−Bω 1
= ε[d1 ]ω −[b]ω
1−e 2
> 0,
k Z ω
e−Bω 1
m2 ≥ ε d2 (s)ds
1 − e−[b]ω
2 0
k
e−Bω 1
= ε[d2 ]ω −[b]ω
1−e 2
> 0.
Therefore (H2) of Theorem 1.9.25 holds.

4. Let
1 − µ 1−m
ξ1 (µ) = ,
µ 1−m
1
ξ2 (µ) = 1 − µ k−1 , µ ∈ (0, 1).
2
Then
µm = µ(1 + ξ1 (µ)),
1
1 − ξ2 (µ) = (1 + µ k−1 )
2
≥ µ k−1 ,
µ(1 − ξ2 (µ)) ≥ µ k, µ ∈ (0, 1).
Note that P is a normal cone with the constant N = 1, and

1
− 1 < 0 and 1 − m2 < 0.
m1
We have
1
lim ξ1 (µ) > − 1,
µ→0+ m1
lim ξ2 (µ) > 1 − m2 .

µ→0+
Next,
F1 (µx) = µ m F1 x
= µ(1 + ξ1 (µ))F1 x,
F2 (µx) = µ k F2 x
≤ µ(1 − ξ2 (µ))F2 x, µ ∈ (0, 1).

Hence, and applying Theorem 1.9.25, it follows that equation (2.8) has at least two non-
negative continuous ω-periodic solutions x1 , x2 such that
0 < kx1 k < ρ < kx2 k < d.
Example 2.1.9 Consider the following model describing the population dynamic of
hematopoiesis (blow cell production)
xm
x0 = −a(t)x + b0 (t) + b1 (t) , (2.21)
1 + xn
where x denotes the density of mature cells in blood circulation, the cells are lost from the
circulation with a rate a(t) at time t, and
xm
b0 (t) + b1 (t)
1 + xn
is the flux of the cells into the circulation from the stem cell compartment. Suppose that a is
a positive continuous ω-periodic function, b0 is a positive continuous ω-periodic function,
b1 is a non-negative continuous ω-periodic function, m ≤ n, m, n > 0. Let M, B0 , B1 , B2 ,
ε, r1 , ρ, r2 , R be positive constants such that
0 < r1 < ρ < r2 < R, (2.22)
e−[a]ω Mω
e (Mωri + B1 ω + B2 ωrim ) < ri , i = 1, 2, (2.23)
1 − e−[a]ω
[a] ≤ M, (2.24)
e−[a]ω Mω
e (MωR + B1 ω + B2 ωRm ) < ρ. (2.25)
1 − e−[a]ω
Here
xm
f (t, x) = b0 (t) + b1 (t) , t ∈ R, x ∈ [0, ∞).
1 + xn
We have
b0 (t) ≤ f (t, x)
≤ b0 (t) + b1 (t)xm , t ∈ R, x ∈ [0, ∞).

By Theorem 2.1.3, it follows the following result.
Theorem 2.1.10 Suppose (2.22)-(2.25). Then equation (2.21) has at least three non-
negative continuous ω-periodic solutions x1 , x2 , x3 for which
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .
Now, suppose that ε, r, L, R, M, B0 , B1 , B2 are positive constants such that
0 < r < L < R, ε > 1, (2.26)
e−[a]ω Mω
e (MωL + B0 ω + B2 ωLm ) ≤ L. (2.27)
1 − e−[a]ω
By Theorem 2.1.5, we find the following result.
Theorem 2.1.11 Suppose (2.26) and (2.27). Then equation (2.21) has at least two non-
negative continuous ω-periodic solutions x1 , x2 such that
R < kx1 k < L < kx2 k < R.
2.2 BVPs for First Order ODEs

Let α, β ∈ R, α < β . Suppose that a : [α, β ] → [0, ∞) is a non-negative continuous function
and P, M, M̃ ∈ R are positive constants and α0 , β0 , γ ∈ R are constants such that
(γ, α0 ) 6= (0, 0), (α0 , β0 ) 6= (0, 0), (γ, β0 ) 6= (0, 0),
γ + α0 e
Rs
α a(y)dy Rt (2.28)
M≥ Rβ e− α a(y)dy
≥ P, s,t ∈ [α, β ],
α0 + β0 e− α a(y)dy
Rs Rβ
γ + α0 e α a(y)dy e− α a(y)dy Rt
M≥ Rβ e− α a(y)dy
≥ P, s,t ∈ [α, β ], (2.29)
α0 − β0 e− α a(y)dy
0 ≤ a(t) ≤ M̃, t ∈ [α, β ]. (2.30)

Let also ε, r1 , r2 , ρ, R, M0 , M1 , M2 , B0 , m and k be positive constants such that
0 < r1 < ρ < r2 < R, (2.31)
M(M0 + M1 rim + M2 rik + M̃ri )(β − α) < ri , i = 1, 2, (2.32)
P(β − α)B0 > R + 1. (2.33)

Consider the BVP
x0 = f (t, x), α < t < β,
Z β (2.34)
α0 x(α) + β0 x(β ) = γ ( f (s, x(s)) + a(s)x(s))ds,
α
where
(H3) f : [α, β ] × [0, ∞) → [0, ∞) is a continuous function such that
b0 (t) ≤ f (t, x) ≤ a0 (t) + a1 (t)xm + a2 (t)xk , t ∈ [α, β ], x ∈ [0, ∞),
b0 , ai : [α, β ] → [0, ∞), i = 0, 1, 2, are positive continuous functions such that
B0 ≤ b0 (t), ai (t) ≤ Mi , t ∈ [α, β ], i = 0, 1, 2.
Define the function

 Rs
 γ + α0 e α a(y)dy − Rαt a(y)dy


 Rβ e = g1 (t, s) if α ≤ s ≤ t ≤ β
 α0 + β0 e− α a(y)dy


G(t, s) =
Rs Rβ
α a(y)dy e− α a(y)dy

 γ − β 0 e Rt
e− α a(y)dy = g2 (t, s) if α ≤ t ≤ s ≤ β .



 Rβ

BVPs for First Order ODEs 103
Note that
P ≤ G(t, s) ≤ M, s,t ∈ [α, β ].
Lemma 2.2.1 Let a : [α, β ] → [0, ∞) be a non-negative continuous function and M, α0 , β0 ,

γ ∈ R satisfy (2.28) and (2.29). If x : [α, β ] → [0, ∞) is a continuous function that satisfies
the integral equation
Z β
x(t) = G(t, s)( f (s, x(s)) + a(s)x(s))ds, t ∈ [α, β ], (2.35)
α
then x is a solution of BVP (2.34).
Proof 2.2.2 We have

Rs Rβ
γ − β0 e α a(y)dy e− α a(y)dy
G(α, s) = Rβ ,
Rs
γ + α0 e α a(y)dy Rβ
G(β , s) = Rβ e− α a(y)dy
,
and
Rs Rβ
γ − β0 e α a(y)dy e− α a(y)dy
α0 G(α, s) + β0 G(β , s) = α0 Rβ
Rs
γ + α0 e α a(y)dy Rβ
+β0 Rβ e− α a(y)dy
= γ, s ∈ [α, β ].
Hence,
Z β
α0 x(α) + β0 x(β ) = α0 G(α, s)( f (s, x(s)) + a(s)x(s))ds
α
Z β
+β0 G(β , s)( f (s, x(s)) + a(s)x(s))ds
α
Z β
= (α0 G(α, s) + β0 G(β , s))( f (s, x(s)) + a(s)x(s))ds
α
Z β
= γ ( f (s, x(s)) + a(s)x(s))ds.
α
Since x ∈ C ([α, β ]) and x satisfies the integral equation (2.35), we have that x ∈ C 1 ([α, β ]).
Observe that
∂
g1 (t, s) = −a(t)g1 (t, s), α ≤ s ≤ t ≤ β,
∂t
∂
g2 (t, s) = −a(t)g2 (t, s), α ≤ t ≤ s ≤ β,
∂t
and
Z t
x(t) = G(t, s)( f (s, x(s)) + a(s)x(s))ds
α
Z β
+ G(t, s)( f (s, x(s)) + a(s)x(s))ds
t
Z t
= g1 (t, s)( f (s, x(s)) + a(s)x(s))ds
α
Z β
+ g2 (t, s)( f (s, x(s)) + a(s)x(s))ds, t ∈ [α, β ].
t
Then
x0 (t) = g1 (t,t)( f (t, x(t)) + a(t)x(t))
Z t
−a(t) g1 (t, s)( f (s, x(s)) + a(s)x(s))ds
α
−g2 (t,t)( f (t, x(t)) + a(t)x(t))

(2.36)
Z β
−a(t) g2 (t, s)( f (s, x(s)) + a(s)x(s))ds
t
= −a(t)x(t)
+(g1 (t,t) − g2 (t,t))( f (t, x(t)) + a(t)x(t)), t ∈ [α, β ].

Note that
Rt
γ + α0 e α a(y)dy Rt
g1 (t,t) = Rβ e− α a(y)dy
,
Rt Rβ
γ − β0 e α a(y)dy e− α a(y)dy Rt
g2 (t,t) = Rβ e− α a(y)dy
,
Rβ
g1 (t,t) − g2 (t,t) = Rβ
= 1, t ∈ [α, β ].
x0 (t) = −a(t)x(t) + f (t, x(t)) + a(t)x(t)
= f (t, x(t)), t ∈ [α, β ].
Let E = C ([α, β ]) be endowed with the norm
kxk = max |x(t)|

t∈[α,β ]
and
P = {x ∈ E : x(t) ≥ 0, t ∈ [α, β ]}.
Theorem 2.2.3 Suppose (H3), (2.28)-(2.33). Then, BVP (2.34) has at least three solutions
x1 , x2 , x3 ∈ P such that
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .
Proof 2.2.4 Define the conical shells
PR = {x ∈ P : kxk < R},
Pρ = {x ∈ P : kxk < ρ},
Pri = {x ∈ P : kxk < ri }, i = 1, 2.
T x(t) = (1 + ε)x(t) − ε(R + 1),
Z β
Fx(t) = ε(R + 1) − ε G(t, s)( f (s, x(s)) + a(s)x(s))ds, t ∈ [α, β ].
α
By Lemma 2.2.1, we have that any fixed point of the operator T + F is a solution of BVP
(2.34). To prove our result, we will apply Theorem 1.9.21.
1. We have T : PR → E. Next, for x, y ∈ P, we get
kT x − Tyk = (1 + ε)kx − yk,
i.e., T : PR → P is an expansive mapping with a constant 1 + ε.

2. We will prove that F : PR → E is a 0-set contraction. Let x ∈ PR . Then
Z β

|Fx(t)| = ε(R + 1) − ε G(t, s)( f (s, x(s)) + a(s)x(s))ds
α
Z β
≤ ε(R + 1) + ε G(t, s)( f (s, x(s)) + a(s)x(s))ds
α
≤ ε(R + 1)
Z β
+ε G(t, s)(a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k + a(s)x(s))ds
α
≤ ε(R + 1) + εM(M0 + M1 Rm + M2 Rk + M̃R)(β − α), t ∈ [α, β ].
Hence,
kFxk ≤ ε(R + 1) + εM(M0 + M1 Rm + M2 Rk + M̃R)(β − α)
and F : PR → E is uniformly bounded. Using the proof of Lemma 2.2.1, we get

d
Fx(t) = −ε f (t, x(t)),
dt

d
Fx(t) = ε f (t, x(t))
dt
≤ ε(a0 (t) + a1 (t)(x(t))m + a2 (t)(x(t))k )
≤ ε(M0 + M1 Rm + M2 Rk ), t ∈ [α, β ], x ∈ PR ,
whereupon
kF 0 xk ≤ ε(M0 + M1 Rm + M2 Rk ), x ∈ PR .
Using the Arzela-Ascoli Theorem, we conclude that F : PR → E is a completely
continuous mapping. Therefore F : PR → E is a 0-set contraction.
3. For i = 1, 2, x ∈ ∂ Pri , we have

Z β

|Fx(t) + T 0(t)| = ε(R + 1) − ε G(t, s)( f (s, x(s)) + a(s)x(s))ds
α

−ε(R + 1)
Z β
≤ ε G(t, s)( f (s, x(s)) + a(s)x(s))ds
α
Z β
≤ εM (a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k + a(s)x(s))ds
α
≤ εM(M0 + M1 rim + M2 rik + M̃ri )(β − α)
< εri , t ∈ R.
Therefore conditions (a) and (c) of Theorem 1.9.21 are satisfied.
4. Now we will prove condition (b) of Theorem 1.9.21. Let u0 ∈ P ∗ and as-
sume that there exist x1 ∈ P and λ1 ≥ 0 with x1 − λ1 u0 ∈ ∂ Pρ such that
Fx1 = (I − T )(x1 − λ1 u0 ).
Then
0 < −ε(R + 1) + εP(β − α)B0
Z β
≤ −ε(R + 1) + ε G(t, s)b0 (s)ds
α
Z β
≤ −ε(R + 1) + ε G(t, s)( f (s, x(s)) + a(s)x(s))ds
α
= −(Fx1 )(t)
= −(I − T )(x1 − λ1 u0 )(t)
= ε(x1 − λ1 u0 )(t) − ε(R + 1)
≤ εx1 (t) − ε(R + 1)

≤ ε(ρ − R − 1)
< 0, t ∈ [α, β ],
which is a contradiction. Therefore condition (b) of Theorem 1.9.21 is fulfilled.

5. Now we will prove
F(PR ) ⊂ (I − T )(PR ). (2.37)
Let z ∈ PR be fixed and consider the following nonlinear equation on PR
x(t) = T x(t) + Fz(t)
= Ax(t), t ∈ R.
= (1 + ε)kx − yk,
i.e., A : PR → P is an expansive mapping. Now we will prove that
PR ⊂ A(PR ). (2.38)
For any given v ∈ PR , consider the following equation on PR
Au(t) = v(t) or Tu(t) = v(t) − Fz(t), t ∈ [α, β ]. (2.39)
We have
Z β
v(t) − Fz(t) = v(t) − ε(R + 1) + ε G(t, s)( f (s, x(s)) + a(s)x(s))ds
α
≥ v(t) − ε(R + 1) + εPB0 (β − α)
≥ 0, t ∈ [α, β ],
and
v(t) − Fz(t) ≤ R + ε(R + 1) + εM(M0 + M1 Rm + M2 Rk + M̃R)(β − α)
= R1 .
Note that R1 > R and

P.
\
v − Fz ∈ B(R1 )
Next,
T : PR → T (PR ) = B(R1 ) P
\
is a bijection. Therefore there is a u ∈ PR that satisfies (2.39). Then (2.38)

holds. Hence, and applying Theorem 1.9.8, it follows that there is a w ∈ PR
such that
Aw(t) = w(t), t ∈ [α, β ],
whereupon
w(t) = Tw(t) + T z(t), t ∈ [α, β ].
Since z ∈ PR was arbitrarily chosen, we conclude that (2.37) holds.
By Theorem 1.9.21, we obtain that BVP (2.34) has at least three solutions x1 , x2 , x3 such
that
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .
Let ε, B0 , M, M̃, P, M0 , M1 , M2 , L, r and R be positive constants such that
ε(R + 1) − εM(M0 + M1 Rm + M2 Rk + M̃R)(β − α) > R, (2.40)
M(M0 + M1 Lm + M2 Lk + M̃L)(β − α) < L, r < L < R. (2.41)
Theorem 2.2.5 Suppose (H3), (2.28), (2.29), (2.30), (2.33), (2.40) and (2.41). Then BVP
(2.34) has at least two solutions x1 , x2 ∈ P such that
r < kx1 k < L < kx2 k < R.
Proof 2.2.6 Let PR , T and F be defined as in the proof of Theorem 2.2.3. Let also
Pr = {x ∈ P : kxk < r},
PL = {x ∈ P : kxk < L},
Pr,L = {x ∈ P : r < kxk < L},
PL,R = {x ∈ P : L < kxk < R},
Ω = PR .
We have
Pr,L
\
Ω 6= 0,
/
PL,R
\
Ω 6= 0.
/
To prove our result we will use Theorem 1.9.23.

1. As in the proof of Theorem 2.2.3, we get
F(PR ) ⊂ (I − T )(PR ).
2. Let u0 ∈ P ∗ and assume that there are a λ1 ≥ 0 and an x1 ∈ P with x1 −

λ1 u0 ∈ ∂ Pr such that
Fx1 = x1 − λ1 u0 .
Then
r ≥ x1 (t) − λ1 u0 (t)
= Fx1 (t)
Z β
= ε(R + 1) − ε G(t, s)( f (s, x(s)) + a(s)x(s))ds
α
Z β
≥ ε(R + 1) − εM (a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k + a(s)x(s))ds
α
≥ ε(R + 1) − εM(M0 + M1 rm + M2 rk + M̃r)(β − α), t ∈ [α, β ],
which is a contradiction. Therefore condition (a) of Theorem 1.9.23 is satisfied.

As in the above, one can prove that condition (c) of Theorem 1.9.23 is satisfied.
Z β
|Fx(t) + T 0(t)| = ε G(t, s)( f (s, x(s)) + a(s)x(s))ds
α
Z β
≤ εM (a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k + a(s)x(s))ds
α
≤ εM(M0 + M1 Lm + M2 Lk + M̃L)(β − α)
≤ εL
= εkxk, t ∈ [α, β ].
Hence,
Now, assume that x1 ∈ ∂ PL
\
Ω is such that
x1 = T x1 + Fx1 .
Let t1 ∈ [α, β ] be such that

L = x1 (t1 ).
Then
L = x1 (t1 )
Z β
= G(t1 , s)( f (s, x1 (s))) + a(s)x1 (s))ds
α
Z β
≤ M a0 (s) + a1 (s)(x1 (s))m + a2 (s)(x1 (s))k + a(s)x1 (s) ds
α

≤ M M0 + M1 Lm + M2 Lk + ML
e (β − α),
which is a contradiction. Therefore condition (b) of Theorem 1.9.23 is

fulfilled.
By Theorem 1.9.23, it follows that BVP (2.34) has at least two solutions x1 , x2 ∈ P such
that
r < kx1 k < L < kx2 k < R.
Suppose that b : [α, β ] → [0, ∞) is a positive continuous function and Q, N, Ñ ∈ R are

positive constants and α0 , β0 , γ ∈ R are constants such that
(γ, α0 ) 6= (0, 0), (γ, β0 ) 6= (0, 0), (α0 , β0 ) 6= (0, 0),
γ + α0 e−
Rs
α b(y)dy R t (2.42)
N≥ α b(y)dy
e ≥ Q, s,t ∈ [α, β ],
Rβ
α0 + β0 e α b(y)dy
Rs Rβ
γ − β0 e− α b(y)dy e α b(y)dy R t
α b(y)dy
N≥ Rβ e ≥ Q, s,t ∈ [α, β ], (2.43)
0 ≤ b(t) ≤ Ñ, t ∈ [α, β ]. (2.44)

Let also ε, C0 , C1 , C2 , B, ρ, d, m and k be positive constants such that
0 < ρ < d, 0 < m < 1, k > 1, (2.45)
d + 1 − N(C0 +C1 d m +C2 d k + Ñd)(β − α) > 0, (2.46)

ρ
N(C0 +C1 ρ m +C2 ρ k + Ñρ)(β − α) ≤ . (2.47)
2
Consider the BVP
x0 = f (t, x), α < t < β,
Z β (2.48)
α0 x(α) + β0 x(β ) = γ ( f (s, x(s)) − b(s)x(s))ds,
α
where
(H4) f : [α, β ] → (−∞, 0],
| f (t, x)| ≤ c0 (t) + c1 (t)xm + c2 (t)xk ,
f (t, x) ≤ −d0 (t) − d1 (t)xm − d2 (t)xk , t ∈ [α, β ], x ∈ [0, ∞),
where ci , di , i = 0, 1, 2, are non-negative continuous functions, ci , di 6≡ 0 on [α, β ],

i = 0, 1, 2, and
ci (t) ≤ Ci , t ∈ [α, β ], i = 0, 1, 2,
and m Z β
1
Q d1 (s)ds > 1, (2.49)
2 α
k Z β
1
Q d2 (s)ds > 1. (2.50)
2 α
Define the function

 Rs
 γ + α0 e− α b(y)dy Rαt b(y)dy


 Rβ e if α ≤ s ≤ t ≤ β
 α0 + β0 e α b(y)dy


G1 (t, s) =
Rs Rβ
γ − β0 e− α b(y)dy e α b(y)dy Rαt b(y)dy



e if α ≤ t ≤ s ≤ β .


 Rβ

Note that
Q ≤ G1 (t, s) ≤ N, s,t ∈ [α, β ].
As we have proved Lemma 2.2.1, one can prove the following lemma.
Lemma 2.2.7 Let b : [α, β ] → [0, ∞) be a positive continuous function and N, α0 , β0 , γ ∈ R
satisfy (2.42), (2.43). If x : [α, β ] → [0, ∞) is a continuous function that satisfies the integral
equation
Z β
x(t) = G1 (t, s)( f (s, x(s)) − b(s)x(s))ds, t ∈ [α, β ],
α
then x is a solution of BVP (2.48).
Theorem 2.2.8 Suppose (H4), (2.42)-(2.50). Then BVP (2.48) has at least two solutions
x1 , x2 ∈ C ([α, β ]) such that
0 < kx1 k < ρ < kx2 k < d.
Proof 2.2.9 Let
Pρ = {x ∈ P : kxk < ρ},
PR = {x ∈ P : kxk < d},
Pρ,d = {x ∈ P : ρ < kxk < d},
Ω = Pd .

T x(t) = (1 + ε)x(t) + ε(d + 1),
Z β
Fx(t) = −ε(d + 1) − ε G1 (t, s)( f (s, x(s)) − b(s)x(s))ds, t ∈ [α, β ].
α
Note that if x ∈ P is a fixed point of the operator T + F, using Lemma 2.2.7, then it is a
solution of BVP (2.48).
1. First, we will prove that
F(Pd ) ⊂ (I − T )(Pd ). (2.51)
Let z ∈ Pd be arbitrarily chosen and fixed. Consider on Pd the equation
x(t) = T x(t) + Fz(t)
= Ax(t), t ∈ [α, β ].
For any x, y ∈ Pd , we have
= (1 + ε)kx − yk,
i.e., A : Pd → P is an expansive mapping. Now we will prove that
Pd ⊂ A(Pd ). (2.52)
For any given v ∈ Pd , consider on Pd the equation
Au(t) = v(t) or Tu(t) = v(t) − Fz(t), t ∈ [α, β ]. (2.53)
We have
v(t) − Fz(t) = v(t) + ε(d + 1)
Z β
+ε G1 (t, s)( f (s, z(s)) − b(s)z(s))ds
α
≥ ε(d + 1)
Z β
−ε G1 (t, s)(c0 (s) + c1 (s)(z(s))m + c2 (s)(z(s))k + b(s)z(s))ds
α
≥ ε(d + 1)
−εN(C0 +C1 d m +C2 d k + Ñd)(β − α)
> 0, t ∈ [α, β ],
and
Z β
v(t) − Fz(t) ≤ v(t) + ε(d + 1) + ε G1 (t, s)(c0 (s) + c1 (s)(z(s))m
α
+c2 (s)(z(s))k + b(s)z(s))ds
≤ d + εN(C0 +C1 d m +C2 d k + Ñd) + ε(d + 1)
= d1 , t ∈ [α, β ].
We have d1 > d and
P,
\
v − Fz ∈ B(d1 )
and
T : Pd → T (Pd ) = B(d1 ) P
\
is a bijection. Therefore there is a u ∈ Pd that satisfies (2.53) and from here,

(2.52) holds. Hence, and applying Theorem 1.9.8, it follows that there is a w ∈
Pd such that
Aw(t) = w(t), t ∈ [α, β ],
whereupon
w(t) = Tw(t) + T z(t), t ∈ [α, β ].
Because z ∈ Pd was arbitrarily chosen, we conclude that (2.51) holds.
2. Let x ∈ ∂ Pρ . Then
Z β

|Fx(t) + T 0(t)| = − ε G1 (t, s)( f (s, x(s)) − b(s)x(s))ds
α
Z β
≤ε G1 (t, s)(c0 (s) + c1 (s)(x(s))m + c2 (s)(x(s))k + b(s)x(s))ds
α
≤ εN(C0 +C1 ρ m +C2 ρ k + Ñρ)(β − α)
≤ ερ, t ∈ [α, β ],
and
kFx + T xk ≥ (1 + ε)ρ − εN(C0 +C1 ρ m +C2 ρ k + Ñρ)(β − α)
ε
≥ (1 + ε)ρ − ρ
2
ε
= 1+ ρ.
2
Therefore conditions (1.20) and (1.21) of Theorem 1.9.25 hold. For x ∈ P,
define the operators
Z β
F1 x(t) = G1 (t, s)d1 (s)(x(s))m ds,
α
Z β
F2 x(t) = G1 (t, s)d2 (s)(x(s))k ds, t ∈ [α, β ].
α
Then
Fx ≥ Fi x, i = 1, 2, x ∈ P.
Take c1 = c2 = 1 in Theorem 1.9.25. Then
m Z β
1
m1 ≥ Q d1 (s)ds
2 α
> 1,
k Z β
1
m2 ≥ Q d2 (s)ds
2 α
> 1.
3. Let
1 − µ 1−m
ξ1 (µ) = ,
µ 1−m
1
ξ2 (µ) = 1 − µ k−1 , µ ∈ (0, 1).
2
Then
µm = µ(1 + ξ1 (µ)),
1
1 − ξ2 (µ) = (1 + µ k−1 )
2
≥ µ k−1 ,
µ(1 − ξ2 (µ)) ≥ µ k, µ ∈ (0, 1).
Note that P is a normal cone with constant N = 1, and

1
− 1 < 0, 1 − m2 < 0.
m1
We have
1
lim ξ1 (µ) > − 1,
µ→0+ m1
lim ξ2 (µ) > 1 − m2 .

µ→0+
Next,
F1 (µx) = µ m F1 x
= µ(1 + ξ1 (µ))F1 x,
F2 (µx) = µ k F2 x
≤ µ(1 − ξ2 (µ))F2 x, µ ∈ (0, 1).

Hence, and applying Theorem 1.9.25, it follows that BVP (2.48) has at least two solutions
x1 , x2 ∈ P such that
0 < kx1 k < ρ < kx2 k < d.
Example 2.2.10 Let γ = 0, α0 = 1 and β0 ∈ R, β0 < 0, in BVP (2.34). Suppose that a :

[α, β ] → [0, ∞) is a positive continuous function and P, M, M̃ ∈ R are positive constants
such that Rs
e α a(y)dy Rt
M≥ Rβ e− α a(y)dy ≥ P, s,t ∈ [α, β ], (2.54)
1 + β0 e− α a(y)dy
Rs Rβ
β0 e α a(y)dy e− α a(y)dy Rt
M≥− Rβ e− α a(y)dy
≥ P, s,t ∈ [α, β ], (2.55)
1 + β0 e− α a(y)dy
0 ≤ a(t) ≤ M̃, t ∈ [α, β ]. (2.56)

Let also ε, r1 , r2 , ρ, R, M0 , M1 , M2 , B0 , m and k be positive constants such that (2.31)-(2.33)
hold. Then, by Theorem 2.2.3, we get the following result.
Theorem 2.2.11 Suppose (H3), (2.31)-(2.33), (2.54)-(2.56). Then BVP (2.34) has at least
three solutions x1 , x2 , x3 ∈ P such that
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .
Now, suppose that ε, B0 , M, M̃, P, M0 , M1 , M2 , L, r, R are positive constants which satisfy

(2.40), (2.41). By Theorem 2.2.5, we get the following result.
Theorem 2.2.12 Suppose (H3), (2.40), (2.41), (2.54)-(2.56). Then BVP (2.34) has at least
r < kx1 k < L < kx2 k < R.
Suppose that b : [α, β ] → [0, ∞) is a continuous function and Q, N, Ñ, β0 ∈ R, β0 < 0,

satisfy the following conditions
Rs
b(y)dy e αt b(y)dy
R
e− α
N≥ Rβ ≥ Q, s,t ∈ [α, β ], (2.57)
1 + β0 e α b(y)dy
Rs
b(y)dy e α b(y)dy e αt b(y)dy
Rβ R
β0 e− α
N≥− Rβ ≥ Q, s,t ∈ [α, β ], (2.58)
1 + β0 e α b(y)dy
0 ≤ b(t) ≤ Ñ, t ∈ [α, β ]. (2.59)

Assume that ε, C0 , C1 , C2 , B, ρ, d, m, k are positive constants such that (2.45)-(2.50) hold.
By Theorem 2.2.8, we get the following result.
Theorem 2.2.13 Suppose (H4), (2.45)-(2.50), (2.57)-(2.59). Then BVP (2.48) has at least
0 < kx1 k < ρ < kx2 k < d.

BVPs for Second Order ODEs 119
2.3 BVPs for Second Order ODEs

Let α, β ∈ R, a ∈ R, a 6= 0, α < β , α0 , α1 α2 , β0 , β1 , β2 , A, B ∈ R, P, M, ε, r1 , r2 , ρ, R, M0 , M1 ,
M2 , B0 , m, k ∈ (0, ∞) be constants such that
P ≤ G(t, s) ≤ M, s,t ∈ [α, β ], (2.60)
0 < r1 < ρ < r2 < R, (2.61)
M(M0 + M1 rim + M2 rik + a2 ri )(β − α) < ri , i = 1, 2, (2.62)
P(β − α)B0 > R + 1, (2.63)

where 
 λ1 (s) sin(at) + λ2 (s) cos(at), α ≤ t ≤ s,
G(t, s) =
µ1 (s) sin(at) + µ2 (s) cos(at), s ≤ t ≤ β,

a1 = sin(aα) + aα0 cos(aα) + α1 sin(aβ ) + aα2 cos(aβ ),
a2 = cos(aα) − aα0 sin(aα) + α1 cos(aβ ) − aα2 sin(aβ ),
b1 = sin(aα) + aβ0 cos(aα) + β1 sin(aβ ) + aβ2 cos(aβ ),
b2 = cos(aα) − aβ0 sin(aα) + β1 cos(aβ ) − aβ2 sin(aβ ),
sin(a(α − s))
A1 = A− − α0 cos(a(α − s)),
a
sin(a(β − s))
B1 = B− − α0 cos(a(β − s)),
a
A1 b2 − a2 B1 cos(as)
λ1 (s) = + ,
a1 b2 − a2 b1 a
A1 b1 − a1 B1 sin(as)
λ2 (s) = − ,
b1 a2 − b2 a1 a
A1 b2 − a2 B1
µ1 (s) = ,
a1 b2 − a2 b1
A1 b1 − a1 B1
µ2 (s) = , s ∈ [α, β ].
b1 a2 − b2 a1
Consider the BVP

x00 = f (t, x), α < t < β,
Z β
x(α) + α0 x0 (α) + α1 x(β ) + α2 x0 (β ) = A ( f (s, x(s)) + a2 x(s))ds, (2.64)
α
Z β
0 0
x(α) + β0 x (α) + β1 x(β ) + β2 x (β ) = B ( f (s, x(s)) + a2 x(s))ds,
α
where
(H5) f : [α, β ] × [0, ∞) → [0, ∞) is a continuous function such that
b0 (t) ≤ f (t, x) ≤ a0 (t) + a1 (t)xm + a2 (t)xk , t ∈ [α, β ], x ∈ [0, ∞),
b0 , ai : [α, β ] → [0, ∞), i = 0, 1, 2, are positive continuous functions such that
B0 ≤ b0 (t), ai (t) ≤ Mi , t ∈ [α, β ], i = 0, 1, 2.
One can prove that any continuous solution of the following nonlinear integral equation
Z β
x(t) = G(t, s)( f (s, x(s)) + a2 x(s))ds, t ∈ [α, β ],
α
satisfies BVP (2.64). Let E = C ([α, β ]) be endowed with the maximum norm
kxk = max |x(t)|,

t∈[α,β ]
and
P = {x ∈ E : x(t) ≥ 0, t ∈ [α, β ]}.
T x(t) = (1 + ε)x(t) − ε(R + 1),
Z β
Fx(t) = ε(R + 1) − ε G(t, s)( f (s, x(s)) + a2 x(s))ds, t ∈ [α, β ].
α
Note that any fixed point in P of the operator T + F is a solution of BVP (2.64). As we
have proved Theorem 2.2.3, one can prove the following theorem.
Theorem 2.3.1 Suppose (H5), (2.60)-(2.63). Then BVP (2.64) has at least three solutions
x1 , x2 , x3 ∈ P such that
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .

BVPs for Second Order ODEs 121
Let α, β ∈ R, α < β , α0 , α1 α2 , β0 , β1 , β2 , A, B ∈ R, P, M, ε, r, L, R, M0 , M1 , M2 , B0 , m, k ∈
(0, ∞) be constants such that (2.60) and (2.63) hold and
r < L < R, (2.65)
ε(R + 1) − εM(M0 + M1 rm + M2 rk + a2 r)(β − α) > R, (2.66)
M(M0 + M1 Lm + M2 Lk + a2 L)(β − α) < L. (2.67)

As we have proved Theorem 2.2.5, one can prove the following theorem.
Theorem 2.3.2 Suppose (H5), (2.60), (2.63), (2.65)-(2.67). Then BVP (2.64) has at least
r < kx1 k < L < kx2 k < R.
Suppose that d1 , d2 : [α, β ] → (0, ∞) are continuous functions. Let α, β ∈ R, α < β ,
α0 , α1 α2 , β0 , β1 , β2 , A, B ∈ R, Q, N, ε, ρ, d,C0 ,C1 ,C2 , m, k ∈ (0, ∞) be constants such that
Q ≤ G1 (t, s) ≤ N, s,t ∈ [α, β ], (2.68)
0 < ρ < d, 0 < m < 1, k > 1, (2.69)
d + 1 − N(C0 +C1 d m +C2 d k + a2 d)(β − α) > 0, (2.70)

ρ
N(C0 +C1 ρ m +C2 ρ k + a2 ρ)(β − α) ≤ , (2.71)
2
m Z β
1
Q d1 (s)ds > 1, (2.72)
2 α
k Z β
1
Q d2 (s)ds > 1, (2.73)
2 α
where
 λ1˜(s)eat + λ2˜(s)e−at ,

α ≤ t ≤ s,
G1 (t, s) =
µ1˜(s)eat + µ2˜(s)e−at , s ≤ t ≤ β,

a˜1 = (1 + aα0 )eaα + (α1 + aα2 )eaβ ,
a˜2 = (1 − aα0 )e−aα + (α1 − aα2 )e−aβ ,
b˜1 = (1 + aβ0 )eaα + (β1 + aβ2 )eaβ ,
b˜2 = (1 − aβ0 )e−aα + (β1 − aβ0 )e−aβ ,

2(1 + aα0 )
A˜1 = A− sinh(a(α − s)),
a
2(1 + aβ0 )
B˜1 = B− sinh(a(α − s)),
a
A˜1 b˜2 − a˜2 B˜1 1

λ1˜(s) = + ,
a˜1 b˜2 − a˜2 b˜1 2aeas
A˜1 b˜1 − a˜1 B˜1 1

λ2˜(s) = + ,
a˜2 b1 − a˜1 b2 2ae−as
˜ ˜
A˜1 b˜2 − a˜2 B˜1

µ1˜(s) = ,
a˜1 b˜2 − a˜2 b˜1
A˜1 b˜1 − a˜1 B˜1

µ2˜(s) = .
a˜2 b˜1 − b˜2 a˜1
Consider the BVP
x00 = f (t, x), α < t < β,
Z β
x(α) + α0 x0 (α) + α1 x(β ) + α2 x0 (β ) = A ( f (s, x(s)) − a2 x(s))ds, (2.74)
α
Z β
x(α) + β0 x0 (α) + β1 x(β ) + β2 x0 (β ) = B ( f (s, x(s)) − a2 x(s))ds,
α
where
(H6) f : [α, β ] → (−∞, 0],
| f (t, x)| ≤ c0 (t) + c1 (t)xm + c2 (t)xk ,
f (t, x) ≤ −d0 (t) − d1 (t)xm − d2 (t)xk , t ∈ [α, β ], x ∈ [0, ∞),
where ci , di , i = 0, 1, 2, are non-negative continuous functions, ci , di 6≡ 0 on [α, β ], i =

0, 1, 2, and
ci (t) ≤ Ci , t ∈ [α, β ], i = 0, 1, 2.
Define on P the following operators
T x(t) = (1 + ε)x(t) + ε(d + 1),

BVPs with Impulses 123
Z β
Fx(t) = −ε(d + 1) − ε G1 (t, s)( f (s, x(s)) − a2 x(s))ds, t ∈ [α, β ].
α
Note that if x ∈ P is a fixed point of the operator T + F, then it is a solution of BVP (2.64).
As we have proved Theorem 2.2.8, one can prove the following theorem.
Theorem 2.3.3 Suppose (H6), (2.68)-(2.73). Then BVP (2.74) has at least two solutions
x1 , x2 ∈ P so that
0 < kx1 k < ρ < kx2 k < d.
2.4 BVPs with Impulses

Let ε, h, g, M0 , M1 ,U j ,V j , j ∈ {1, . . . , m}, r1 , r2 , ρ, R be positive constants such that
m
h + g + ∑ (U j +V j ) + M0 + M1 rim + M2 rik ≤ ri , i = 1, 2, (2.75)
j=1
0 < r1 < ρ < r2 < R, (2.76)
−(R + 1) + g > 0. (2.77)

Let
0 = t0 < t1 < . . . < tm < tm+1 = 1.
Assume
(H7) f : [0, 1] × [0, ∞) → [0, ∞), f ∈ C ([0, 1] × [0, ∞)),
f (t, x) ≤ a0 (t) + a1 (t)xm + a2 (t)xk , t, x ∈ [0, ∞),
ai : [0, ∞) → [0, ∞), b0 , ai ∈ C ([0, ∞)),
B0 ≤ b0 (t), ai (t) ≤ Mi , t ∈ [0, ∞), i = 0, 1, 2,
(H8) uk , −vk : [0, ∞) → [0, ∞), uk , −vk ∈ C ([0, ∞)),
uk (t) ≤ Uk , −vk (t) ≤ Vk , t ∈ [0, ∞), k ∈ {0, . . . , m}.

Consider the BVP

x00 + f (t, x) = 0, tk < tk+1 , k ∈ {0, . . . , m},
∆x(tk ) = uk (x(tk −)),
∆x0 (tk ) = vk (x(tk −)), k ∈ {1, . . . , m}, (2.78)
x(0) = g,
x0 (1) = h,
where
∆x(t) = lim x(s) − lim x(s),

s→t+ s→t−
x(tk −) = lim x(s), k ∈ {1, . . . , m}.

s→tk −
With PC([0, 1]) we will denote the set of all piecewise continuous functions on [0, 1]. For
x ∈ PC([0, 1]), for each tk ≤ t ≤ tk+1 , k ∈ {0, . . . , m}, define
tk ≤ t < tk+1 ,

lim x(s),
 s→t+

xk (t) =
 lim x(s),
 t = tk+1 .
s→tk+1
Define the Banach space
B = {x ∈ PC([0, 1]) : xk ∈ C ([tk ,tk+1 ]), k ∈ {0, . . . , m}},
endowed with the norm

kxk = max max |xk (t)|,
k∈{0,...,m} t∈[tk ,tk+1 ]
the cone
P = {x ∈ B, x(t) ≥ 0, t ∈ [0, 1]}.
Define
m i i
I(t, x) = − ∑ v j (x(t j −))t − ∑ v j (x(t j −))t j + ∑ u j (x(t j −))
j=i+1 j=1 j=1
if ti ≤ t < ti+1 , i ∈ {0, . . . , m − 1},

m
I(t, x) = ∑ (−v j (x(t j −))t j + u j (x(t j −)))
j=1
if tm ≤ t ≤ 1, and 
 s if 0≤s≤t ≤1
G(t, s) =
t if 0 ≤ t ≤ s ≤ 1.

Lemma 2.4.1 Suppose (H7), (H8). If x ∈ B is a solution of the integral equation

Z 1
x(t) = ht + g + I(t, x(t)) + G(t, s) f (s, x(s))ds, 0 ≤ t ≤ 1,
0
then it is a solution of BVP (2.78).

Proof 2.4.2 We have
Z t
x(t) = ht + g + I(t, x(t)) + s f (s, x(s))ds
0
Z 1
+t f (s, x(s))ds, t ∈ [0, 1].
t
Let t ∈ [ti ,ti+1 ], i ∈ {0, . . . , m − 1}. Then

m
d
I(t, x(t)) = − ∑ v j (x(t j −)),
dt j=i+1
d2
I(t, x(t)) = 0,
dt 2
Z 1
d
x0 (t) = h + I(t, x(t)) + t f (t, x(t)) + f (s, x(s))ds
dt t
−t f (t, x(t))
m Z 1
= h− ∑ v j (x(t j −)) + f (s, x(s))ds,
j=i+1 t
x00 (t) = − f (t, x(t)).

Let t ∈ [tm , 1]. Then
m
I(t, x(t)) = ∑ (−v j (x(t j −))t j + u j (x(t j ))) ,
j=1
d
I(t, x(t)) = 0,
dt
d2
I(t, x(t)) = 0,
dt 2
Z 1
x0 (t) = h + f (s, x(s))ds,
t
x00 (t) = − f (t, x(t)).
Next, for k ∈ {1, . . . , m}, we obtain
∆x(tk ) = lim x(y) − lim x(y)

y→tk + y→tk −
Z y
= lim hy + g + I(y, x(y)) + s f (s, x(s))ds
y→tk + 0
Z 1
+y f (s, x(s))ds
y
Z y
− lim hy + g + I(y, x(y)) + s f (s, x(s))ds
y→tk − 0
Z 1
+y f (s, x(s))ds
y
= lim I(y, x(y)) − lim I(y, x(y))

y→tk + y→tk −
m k k−1
= − ∑ v j (x(t j −))tk − ∑ v j (x(t j −))t j + ∑ u j (x(t j −)) + uk (x(tk −))
j=k+1 j=1 j=1
!
m k−1 k−1
− − ∑ v j (x(t j −))tk − ∑ v j (x(t j −))t j + ∑ u j (x(t j −))
j=k j=1 j=1
= uk (x(tk −)),
∆x0 (tk ) = lim x0 (y) − lim x0 (y)

y→tk + y→tk −
!
m Z 1
= lim h− ∑ v j (x(t j −)) + f (s, x(s))ds
y→tk + y
j=k+1
!
m Z 1
− lim h − ∑ v j (x(t j −)) + f (s, x(s))ds
y→tk − y
j=k
= vk (x(tk −)).
Also,
∆x(tm ) = lim x(y) − lim x(y)

y→tm + y→tm −
m
= lim hy + g + ∑ (−v j (x(t j −))t j + u j (x(t j −)))
y→tm +
j=1
Z y Z 1
+ s f (s, x(s))ds + y f (s, x(s))ds
0 y
m−1
− lim hy + g − vm (x(tm −))y − ∑ v j (x(t j −))t j
y→tm −
j=1
m−1 Z y Z 1
+ ∑ u j (x(t j −)) + 0
s f (s, x(s))ds + y
y
f (s, x(s))ds
j=1
= um (x(tm −)),
∆x0 (tm ) = lim x(y) − lim x0 (y)

y→tm + y→tm −
1
Z
= lim h + f (s, x(s))ds
y→tm + y
1
Z
− lim h − vm (x(tm −)) + f (s, x(s))ds
y→tm − y
= vm (x(tm −)),
and
x(0) = g,
x0 (1) = h.

Theorem 2.4.3 Suppose (H7), (H8), (2.75)-(2.77). Then BVP (2.78) has at least three
solutions x1 , x2 , x3 ∈ P such that
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .
PR = {x ∈ P : kxk < R},
Pρ = {x ∈ P : kxk < ρ},
Pri = {x ∈ P : kxk < ri }, i = 1, 2.
T x(t) = (1 + ε)x(t) − ε(R + 1),

Fx(t) = ε(R + 1) − ε ht + g + I(t, x(t))
Z 1
+ G(t, s) f (s, x(s))ds , t ∈ [0, 1].
0
By Lemma 2.4.1, it follows that any fixed point x ∈ E of the operator T + F is a solution of
BVP (2.78).
1. We have T : PR → E and
kT x − Tyk = (1 + ε)kx − yk, x, y ∈ PR .
Therefore T : PR → P is an expansive operator with a constant 1 + ε.

2. Now we will prove that F : PR → E is a 0-set contraction. Let x ∈ PR be
arbitrarily chosen. Note that for t ∈ [ti ,ti+1 ), i ∈ {0, . . . , m − 1}, we have
m i i
|I(t, x(t))| = − ∑ v j (x(t j −))t − ∑ v j (x(t j −))t j + ∑ u j (x(t j −))
j=i+1 j=1 j=1
m i i
≤ ∑ Vj + ∑ Vj + ∑ U j
j=i+1 j=1 j=1
m
≤ ∑ (U j +V j ),
j=1
m

d
I(t, x(t)) = − ∑ v j (x(t j −))
dt
j=i+1
m
≤ ∑ Vj.
j=1
For t ∈ [tm ,tm+1 ], we have

m
|I(t, x(t))| = ∑ (−v j (x(t j −))t j + u j (x(t j −)))
j=1
m
≤ ∑ (U j +V j ),
j=1

d
I(t, x(t)) = 0.
dt
Therefore, for any t ∈ [0, 1], we have

m
|I(t, x(t))| ≤ ∑ (U j +V j ),
j=1
m

d
I(t, x(t)) ≤ ∑ Vj.
dt
j=1
Hence,

|Fx(t)| = ε(R + 1) − ε ht + g + I(t, x(t))
Z 1

+ G(t, s) f (s, x(s))ds
0
m
≤ ε(R + 1) + ε h + g + ∑ (U j +V j )
j=1
Z 1
m k
+ a0 (s) + a1 (s)(x(s)) + a2 (s)(x(s)) ds
0
!
m
≤ ε(R + 1) + ε h + g + ∑ (U j +V j ) + M0 + M1 Rm + M2 Rk , t ∈ [0, 1].
j=1
Thus
!
m
m k
kFxk ≤ ε(R + 1) + ε h + g + ∑ (U j +V j ) + M0 + M1 R + M2 R ,
j=1
i.e., F : PR → E is uniformly bounded. Next,

Z 1
d
Fx(t) = −ε h + d I(t, x(t)) +

dt f (s, x(s))ds
dt t

d
≤ ε h + I(t, x(t))
dt
Z 1
m k
+ (a0 (s) + a1 (s)(x(s)) + a2 (s)(x(s)) )ds
0
!
m
m k
≤ ε h + ∑ V j + M0 + M1 R + M2 R , t ∈ [0, 1],
j=1
!
m
0 m k
kF xk ≤ ε h + ∑ V j + M0 + M1 R + M2 R .
j=1
By the Arzela-Ascoli Theorem, it follows that F : PR → E is a completely con-

tinuous mapping, from which we conclude that it is a 0-set contraction.
Z 1

|Fx(t) + T 0(t)| = −ε ht + g + I(t, x(t)) + G(t, s) f (s, x(s))ds
0
m
≤ ε h + g + ∑ (U j +V j )
j=1
Z 1
+ a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k ds
0
!
m
≤ ε h + g + ∑ (U j +V j ) + M0 + M1 rim + M2 rik
j=1
≤ εri , t ∈ [0, 1].

Therefore conditions (a) and (c) of Theorem 1.9.21 hold.
4. Let u0 ∈ P ∗ and assume that there exists an x1 ∈ P and a λ1 > 0 with
x1 − λ1 u0 ∈ ∂ Pρ such that
Fx1 = (I − T )(x1 − λ1 u0 ).
Then
0 < −ε(R + 1) + εg

≤ −ε(R + 1) + ε ht + g + I(t, x1 (t))
Z 1
+ G(t, s) f (s, x1 (s))ds
0
= −(Fx1 )(t)
= −(I − T )(x1 − λ1 u0 )(t)
= ε(x1 − λ1 u0 )(t) − ε(R + 1)
≤ εx1 (t) − ε(R + 1)
≤ ε(ρ − R − 1)
< 0,
which is a contradiction. Therefore, condition (b) of Theorem 1.9.21 holds.

F(PR ) ⊂ (I − T )(PR ). (2.79)
Let z ∈ PR be fixed and consider on PR the nonlinear equation
x(t) = T x(t) + Fz(t)
= Ax(t), t ∈ [0, 1].
= (1 + ε)kx − yk,
i.e., A : PR → E is an expansive mapping. Now we will prove that
PR ⊂ A(PR ). (2.80)
For any v ∈ PR , consider on PR the equation
Au(t) = v(t) or Tu(t) = v(t) − Fz(t), t ∈ [0, 1]. (2.81)

We have

v(t) − Fz(t) = v(t) − ε(R + 1) + ε ht + g + I(t, x(t))
Z 1
+ G(t, s) f (s, x(s))ds
0
≥ v(t) − ε(R + 1) + εg
≥ 0, t ∈ [0, 1],
and
m
v(t) − Fz(t) ≤ R + ε(R + 1) + ε h + g + ∑ (U j +V j )
j=1
Z 1
m k
+ a0 (s) + a1 (s)(x(s)) + a2 (s)(x(s)) ds
0
!
m
m k
≤ R + ε(R + 1) + ε h + g + ∑ (U j +V j ) + M0 + M1 R + M2 R
j=1
= R1 , t ∈ [0, 1].
Note that R1 > R,

P,
\
v − Fz ∈ B(R1 )
and
T : PR → T (PR ) = B(R1 ) P
\
is a bijection. Therefore, equation (2.81) has a solution u ∈ PR . Hence, and

applying Theorem 1.9.8, it follows that there is a w ∈ PR which is a fixed point
of the operator A and (2.79) holds.
By Theorem 1.9.21, it follows that BVP (2.78) has three solutions x1 , x2 , x3 ∈ P such that
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .
Let ε, r, R, L, h, g, M0 , M1 , M2 ,U j ,V j , j ∈ {1, . . . , m}, be positive constants such that
0 < r < L < R, ε > 1, (2.82)

!
m
m k
ε(R + 1) − ε h + g + ∑ (U j +V j ) + M0 + M1 R + M2 R > L, (2.83)
j=1
m
h + g + ∑ (U j +V j ) + M0 + M1 Lm + M2 Lk ≤ L, (2.84)
j=1
g > L. (2.85)
Theorem 2.4.5 Suppose (H7), (H8), (2.82)-(2.85). Then BVP (2.78) has at least two so-
lutions x1 , x2 ∈ P such that
r < kx1 k < L < kx2 k < R.
Proof 2.4.6 Let T and F be defined as in the proof of Theorem 2.4.3. Define
Pr = {x ∈ P : kxk < r},
PL = {x ∈ P : kxk < L},
Pr,L = {x ∈ P : r < kxk < L},
PL,R = {x ∈ P : L < kxk < R},
Ω = PR .
We have
Pr,L
\
Ω 6= 0,
/
PL,R
\
Ω 6= 0.
/
1. As in the proof of Theorem 2.4.3, it follows that
F(PR ) ⊂ (I − T )(PR ).
2. Let u0 ∈ P ∗ and assume that there is a λ1 ≥ 0 and an x1 ∈ ∂ Pr such that
Fx1 = x1 − λ1 u0 .
Then
r ≥ x1 (t)
≥ x1 (t) − λ1 u0 (t)
= Fx1 (t)
Z 1
= ε(R + 1) − ε ht + g + I(t, x(t)) + G(t, s) f (s, x(s))ds
0
m
≥ ε(R + 1) − ε h + g + ∑ (U j +V j ) + M0 + M1 rm + M2 rk , t ∈ [0, 1],
j=1
which is a contradiction. Thus condition (a) of Theorem 1.9.23 holds. As in the

above, one can prove that condition (c) of Theorem 1.9.23 holds.

|Fx(t) + T 0(t)| = − ε(R + 1) + ε(R + 1) − ε ht + g + I(t, x(t))

Z 1

+ G(t, s) f (s, x(s))ds
0
Z 1
= ε ht + g + I(t, x(t)) + G(t, s) f (s, x(s))ds
0
m
m k
≤ ε h + g + ∑ (U j +V j ) + M0 + M1 L + M2 L
j=1
≤ εL
= εkxk.
Therefore
4. Assume that
x1 = T x1 + Fx1
for some x1 ∈ ∂ PL
\
Ω. Let t ∈ [0, 1] be such that
L = x1 (t).
Then
L = x1 (t)
= T x1 (t) + Fx1 (t)

Z 1
= (1 + ε)x1 (t) − ε(R + 1) − ε ht + g + I(t, x(t)) + G(t, s) f (s, x(s))ds
0
+ε(R + 1)
≤ (1 + ε)L − εg,
whereupon
g ≤ L.
This is a contradiction and hence, condition (b) of Theorem 1.9.23 holds.
By Theorem 1.9.23, it follows that BVP (2.78) has at least two solutions x1 , x2 ∈ P such
that
r < kx1 k < L < kx2 k < R.
3
Applications to FDEs
3.1 Global Existence for a Class of Fractional-Differential Equations

Consider the following problem
Dα0+ (t k D0+ u) = t γ ∆u + |u| p+1 , t > 0, x ∈ RN ,

β
(3.1)
lim u(x,t) = u(x, 0), x ∈ RN , (3.2)

t→0
β
where 0 < α, β < 1, p > 0 , γ ∈ R, k ∈ R . We denote with Dα0+ and D0+ the left Riemann-
n
∂2
Liouville fractional derivatives in t, ∆ = ∑ 2 is the Laplacian in the space variables. For
i=1 ∂ xi
convenience, we restrict our investigations for the case N = 1. In this case, problem (3.1)
and (3.2) can be written as follows.
Dα0+ (t k D0+ u) = t γ uxx + |u| p+1 ,

β
t > 0, x ∈ R, (3.3)
lim u(x,t) = u(x, 0), x ∈ R. (3.4)

t→0
Theorem 3.1.1 Let α, β ∈ (0, 1), γ ≥ 0, k ≥ β and p > 0 be fixed. Then problem (3.3),
(3.4) has a solution u ∈ C1 ((0, +∞),C2 (R)).
Proof 3.1.2 Let b > 3 and Q > 0 be fixed. We set
n Zx Zy 1 Zx 1
A1 = max 1, sup dwdy sup dw

2
(w + 1)b , 2
(w + 1)b ,
x∈R x∈R
0 0 0
Zx Zy |1 − (1 + 2b)w2 | Zx |1 − (1 + 2b)w2 | o
sup dwdy, sup dw .

(w2 + 1)b+2 (w2 + 1)b+2
x∈R x∈R
0 0 0
137
138 Applications to FDEs
Because b > 3, we have that A1 < ∞. We choose a constant a > 3 such that
a − γ > 5, a > k+4−α −β, 11a > k + 4 − α − β . (3.5)
Let ε, q ∈ (0, 1) be chosen so that
(a(6 − 3β ) + 2 − β )(1 − q)


 εQ + A1 Q 2



 (a − 1)(1 − α)(1 − β )Γ(1 − α)Γ(1 − β )



 + (1 − q)Q (1 + 6bA1 ) + (1 − q)A1 Q p+1 ≤ Q,



a−γ −1









 2(1 − q)A1 Q (30 − 12β )a(1 − q)A1 Q
εQ + +


(1 − α)Γ(1 − α)Γ(1 − β ) (a − 1)(1 − α)(1 − β )Γ(1 − α)Γ(1 − β )



(3.6)
(40 − 20β )a2 (1 − q)A1 Q


+ 2 + (1 − q)Q






 (a − 1)(1 − α)(1 − β )Γ(1 − α)Γ(1 − β )



10a(1 − q)Q 40ab(1 − q)A1 Q


+ + 4b(1 − q)A1 Q +


a−γ −1 a−γ −1








 20abA1 Q
+ 2(1 − q)A1 Q p+1 ≤ Q.

 +2b(1 − q)A1 Q +

a−γ −1
Let C 1 ([0, ∞), C 2 (R1 )) be endowed with the norm
n
||u|| = max sup |u(x,t)|, sup |ux (x,t)|,
t∈[0,∞),x∈R t∈[0,∞),x∈R
o
sup |ut (x,t)|, sup |uxx (x,t)| .
t∈[0,∞),x∈R t∈[0,∞),x∈R
With Ñ we will denote the set of all equi-continuous families in C 1 ([0, ∞), C 2 (R)), i.e.,
if F ∈ Ñ is a family of elements of C 1 ([0, ∞), C 2 (R)), then for every ε > 0 there exists
δ = δ (ε) > 0 such that
|v(x1 ,t1 ) − v(x2 ,t2 )| < ε, |vt (x1 ,t1 ) − vt (x2 ,t2 )| < ε,
|vx (x1 ,t1 ) − vx (x2 ,t2 )| < ε, |vxx (x1 ,t1 ) − vxx (x2 ,t2 )| < ε
for every v ∈ F , whenever |t1 −t2 | < δ , |x1 − x2 | < δ , t1 ,t2 ∈ [0, ∞), x1 , x2 ∈ R. Let N 1 = Ñ,
where Ñ is the closure of Ñ, and
N = u ∈ N 1 : ||u|| ≤ Q , N1 = u ∈ N 1 : ||u|| ≤ (1 + ε)Q .

Note that the set N is a compact subset of N1 .

Global Existence for a Class of Fractional-Differential Equations 139
For u ∈ N1 , we define the operators.
T (u)(x,t) = (1 + ε)u(x,t),
S(u)(x,t) = −εu(x,t)
(1 − q) 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a
sk
Z xZ y Z t Z s
1 1 d u(w, z)
Z τ
× dzdsdτdwdy
0 0 (w2 + 1)b 0 (τ + 1)a 0 (τ − s)α ds 0 (s − z)β
a(1 − q) 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a
sk
Z xZ y Z t Z τ1 Z s
1 1 d u(w, z)
Z τ2
× dzdsdτ2 dτ1 dwdy
0 0 (w2 + 1)b 0 0 (τ2 + 1)a+1 0 (τ2 − s)α ds 0 (s − z)β
Z t Z τ1 γ
(1 − q) 1 τ2
− b (t + 1)10a
u(x, τ2 )dτ2 dτ1
2
(x + 1) 0 0 (τ2 + 1)a
Z x Z t Z τ1 γ
1 y τ2
−4b(1 − q) u(y, τ2 )dτ2 dτ1 dy
(t + 1)10a 0 (y2 + 1)b+1 0 0 (τ2 + 1)a
γ
1 − (1 + 2b)w2 t τ1
Z xZ y
1
Z Z
τ2
+2b(1 − q) u(w, τ2 )dτ2 dτ1 dwdy
(t + 1)10a 0 0 (w2 + 1)b+2 0 0 (τ2 + 1)a
Z xZ y Z t Z τ1
1 1 1
−(1 − q) |u(w, τ2 )| p+1 dτ2 dτ1 dwdy,
(t + 1)10a 0 0
2
(w + 1)b 0 0 (τ2 + 1)a
R(u)(x,t) = T (u)(x,t) + S(u)(x,t), x ∈ R, t ∈ [0, ∞).
Our first observation is as follows.

Lemma 3.1.3 The fixed points of the operator R are solutions of problem (3.3) and (3.4).
Proof 3.1.4 Let u ∈ N1 be such that R(u)(x,t) = u(x,t), x ∈ R, t ≥ 0. Then
sk
Z xZ y Z t Z s
1 1 1 d u(w, z)
Z τ
dzdsdτdwdy
Γ(1 − α)Γ(1 − β ) 0 0 (w2 + 1)b 0 (τ + 1)a 0 (τ − s)α ds 0 (s − z)β
a
+
Γ(1 − α)Γ(1 − β )
sk
Z xZ y Z t Z τ1 Z s
1 1 d u(w, z)
Z τ2
× 2
dzdsdτ2 dτ1 dwdy
0 0 (w + 1)b 0 0 (τ2 + 1)a+1 0 (τ2 − s) ds
α
0 (s − z)β
Z t Z τ1 γ
1 τ2
− u(x, τ2 )dτ2 dτ1
(x2 + 1)b 0 0 (τ2 + 1)a
Z x Z t Z τ1 γ
y τ2
−4b u(y, τ2 )dτ2 dτ1 dy
0 (y2 + 1)b+1 0 0 (τ2 + 1)a
γ
1 − (1 + 2b)w2 t τ1
Z xZ y Z Z
τ2
+2b u(w, τ2 )dτ2 dτ1 dwdy
0 0 (w2 + 1)b+2 0 0 (τ2 + 1)a
|u(w, τ2 )| p+1
Z xZ y Z t Z τ1
1
− 2
dτ2 dτ1 dwdy = 0, x ∈ R, t ≥ 0.
0 0 (w + 1)b 0 0 (τ2 + 1)a
The last equation we differentiate twice in x and we get
sk
Z t Z s
1 1 d u(x, z)
Z τ
dzdsdτ
Γ(1 − α)Γ(1 − β ) 0 (τ + 1)a 0 (τ − s)α ds 0 (s − z)β
sk
Z t Z τ1 Z s
a 1 d u(x, z)
Z τ2
+ dzdsdτ2 dτ1
Γ(1 − α)Γ(1 − β ) 0 0 (τ2 + 1)a+1 0 (τ2 − s) ds
α
0 (s − z)β
γ
|u(x, τ2 )| p+1
Z t Z τ1 Z t Z τ1
τ2
− uxx (x, τ2 )dτ2 dτ1 − dτ2 dτ1 = 0, x ∈ R, t ≥ 0,
0 0 (τ2 + 1)a 0 0 (τ2 + 1)a
which we differentiate twice in t and we find
sk
Z t Z s
1 d d u(x, z)
dzds − t γ uxx (x,t) − |u(x,t)| p+1 = 0,
Γ(1 − α)Γ(1 − β ) dt 0 (t − s)α ds 0 (s − z)β
x ∈ R,t ≥ 0.
Therefore, u satisfies equation (3.3). Since u ∈ N1 , we have u ∈ C1 [0, +∞),C2 (R) and

hence
lim u(x,t) = u(x, 0).
t→0
Lemma 3.1.3 motivates us to search fixed points of the operator R. To prove that R has fixed
points, we will use Theorem 1.9.12.
Lemma 3.1.5 The operator T : N → N1 is an expansive operator and onto.
Proof 3.1.6 Let u ∈ N. Then T (u) ∈ C1 ([0, ∞),C2 (R)) and ||T (u)|| ≤ (1 + ε)Q. Therefore
T : N → N1 . Since ||T (u) − T (v)|| = (1 + ε)||u − v|| for u, v ∈ N, we conclude that T : N →
v
N1 is an expansive operator with constant h = 1+ε. Also, for any v ∈ N1 we set u =
(1 + ε)
and we get T (u) = v. Thus T : N → N1 is onto.
Before proving that R : N → N, we need the following estimates.

Lemma 3.1.7 For u ∈ N we have
d Zs u(x, z)
!
−β s1−β
dz ≤ Q s + for s ≥ 0, x ∈ R.

ds (s − z)β 1−β

0
Proof 3.1.8 Using (3.4), we have

Z s Z s
d u(x, z)
dz = u(x, 0)s−β + uz (x, z)(s − z)−β dz, s ≥ 0, x ∈ R.
ds 0 (s − z)β 0
Hence,
d Z s u(x, z) Z s
u(x, 0)s−β + uz (x, z)(s − z)−β dz

dz =

ds 0 (s − z)
β

0

Z s
≤ |u(x, 0)|s−β + |uz (x, z)|(s − z)−β dz
0
Z s
≤ Qs−β + Q (s − z)−β dz
0
Q z=s
= Qs−β − (s − z)1−β

1−β z=0
Q 1−β
= Qs−β + s
1−β
!
−β s1−β
= Q s + , s ≥ 0, x ∈ R,
1−β
which completes the proof.
Lemma 3.1.9 For k ≥ β we have

!
sk s1−β
Z t
1 1 2−β
Z τ
−β
s + dsdτ ≤ , t ≥ 0.
(t + 1)a 0 (τ + 1)a 0 (τ − s)α 1−β (a − 1)(1 − α)(1 − β )

!
sk s1−β
Z t
1 1
Z τ
−β
s + dsdτ
(t + 1)a 0 (τ + 1)a 0 (τ − s)α 1−β
sk−β
Z t
1 1
Z τ
= dsdτ
(t + 1)a 0 (τ + 1)a 0 (τ − s)α
Z t Z τ k+1−β
1 1 s
+ dsdτ
(t + 1)a (1 − β ) 0 (τ + 1)a 0 (τ − s)α
t k−β t k+1−β
Z t Z t
1 ds 1 ds
Z τ Z τ
≤ dτ + dτ
(t + 1)a 0 (τ + 1)a 0 (τ − s)α (1 − β )(t + 1)a 0 (τ + 1)a 0 (τ − s)α
t k−β τ 1−α t k+1−β τ 1−α

Z t Z t
= dτ + dτ
(1 − α)(t + 1)a 0 (τ + 1) a (1 − α)(1 − β )(t + 1)a 0 (τ + 1)a
t k+1−α−β t k+2−α−β
Z t Z t
dτ dτ
≤ +
(1 − α)(t + 1)a 0 (τ + 1)a (1 − α)(1 − β )(t + 1)a 0 (τ + 1)a
t k+1−α−β t k+2−α−β
≤ +
(a − 1)(1 − α)(t + 1)a−1 (a − 1)(1 − α)(1 − β )(t + 1)a−1
1 1
≤ +
(a − 1)(1 − α) (a − 1)(1 − α)(1 − β )
2−β
= , t ≥ 0,
(a − 1)(1 − α)(1 − β )
Lemma 3.1.11 Let k ≥ β . Then

!
sk s1−β
Z t Z τ1
1 1 4 − 2β
Z τ2
−β
s + dsdτ2 dτ1 ≤ .
(t + 1)10a 0 0 (τ2 + 1)a+1 0 (τ2 − s)α 1−β (a2 − 1)(1 − α)(1 − β )
Proof 3.1.12 Using Lemma 3.1.9, we get

!
sk s1−β
Z t Z τ1
1 1
Z τ2
−β
s + dsdτ2 dτ1
(t + 1)10a 0 0 (τ2 + 1)a+1 0 (τ2 − s)α 1−β
!
(τ1 + 1)a τ1 sk s1−β
Z t
1 1
Z Z τ2
−β
= s + dsdτ2 dτ1
(t + 1)10a 0 (τ1 + 1)a 0 (τ2 + 1)a+1 0 (τ2 − s)α 1−β
Z t
1 2−β
≤ (τ1 + 1)a dτ1
(t + 1)10a (a − 1)(1 − α)(1 − β ) 0
1 2−β τ =t
a+1 1
= (τ 1 + 1)
(t + 1)10a (a − 1)(a + 1)(1 − α)(1 − β )

τ1 =0
(4 − 2β ) (t + 1)a+1
≤
(a2 − 1)(1 − α)(1 − β ) (t + 1)10a
(4 − 2β )
≤ , t ≥ 0,
(a2 − 1)(1 − α)(1 − β )
Lemma 3.1.13 We have

Z t Z τ1 γ
1 τ2 1
dτ2 dτ1 ≤ , t ≥ 0.
(t + 1)10a 0 0 (τ2 + 1)a a−γ −1

Z t Z τ1 γ Z t Z τ1 γ
1 τ2 1 τ2
dτ2 dτ1 = dτ2 dτ1
(t + 1)10a 0 0 (τ2 + 1)a (t + 1)10a 0 0 (τ2 + 1)a−γ (τ2 + 1)γ
Z t Z τ1
1 dτ2
≤ dτ1
(t + 1)10a 0 0 (τ2 + 1)a−γ
Z t
1
≤ dτ1
(a − γ − 1)(t + 1)10a 0
t
≤
(a − γ − 1)(t + 1)10a
1
≤ , t ≥ 0,
a−γ −1
Lemma 3.1.15 Let k ≥ β . Then for every u ∈ N we have
sk d s u(w, z)
Z t Z
1 2Q
11a
dz ds ≤
(t + 1) 0 (t − s) ds 0 (s − z) 1−α
α β
for t ≥ 0, w ∈ R.
Proof 3.1.16 By Lemma 3.1.7 and our choice of a, we get

!
sk sk s1−β
Z t Zs Z t
1 d u(w, z) Q −β
dz ds ≤ s + ds
(t + 1)11a 0 (t − s)α ds 0 (s − z)β (t + 1)11a 0 (t − s)α 1−β
sk−β
Z t Z t k+1−β
Q Q s
= ds + ds
(t + 1)11a 0 (t − s)α 11a
(t + 1) (1 − β ) 0 (t − s)α
Qt k−β
Z t Z t
Q 1 k+1−β
≤ (t − s)−α ds + t (t − s)−α ds
(t + 1)11a 0
11a
(t + 1) 1 − β 0
!
Q t k+1−α−β t k+2−α−β
= +
1−α (t + 1)11a (t + 1)11a
2Q
≤ , t ≥ 0, w ∈ R,
1−α
Lemma 3.1.17 We have

Z t γ
1 τ2
dτ2 ≤ 1, t ≥ 0.
(t + 1)10a 0 (τ2 + 1)a
Z t γ Z t γ
1 τ2 1 τ2 1
dτ2 = dτ2
(t + 1)10a 0 (τ2 + 1)a (t + 1)10a 0 (τ2 + 1) (τ2 + 1)a−γ
γ
Z t
1 1
≤ dτ2
(t + 1)10a 0 (τ2 + 1)a−γ
t
≤
(t + 1)10a
≤ 1, t ≥ 0,
Lemma 3.1.19 The operator S : N → N is a continuous operator and S(N) resides in a

compact subset of N1 .
Proof 3.1.20 1. For u ∈ N we have that S(u) ∈ C 1 ([0, ∞), C 2 (R)).

Below, without loss of generality, we suppose that x ≥ 0.
2. Let u ∈ N. Then, applying Lemma 3.1.7, Lemma 3.1.9, Lemma 3.1.11,

Lemma 3.1.13, Lemma 3.1.15 and Lemma 3.1.17 we get

|S(u)(x,t)| = −εu(x,t)

(1 − q) 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a
sk
Z xZ y Z t Z s
1 1 d u(w, z)
Z τ
× dzdsdτdwdy
0 0 (w2 + 1)b 0 (τ + 1)a 0 (τ − s)α ds 0 (s − z)β
Z xZ y
a(1 − q) 1 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a 0 0 (w2 + 1)b
sk
Z t Z τ1 Z s
1 d u(w, z)
Z τ2
0 0 (τ2 + 1)a+1 0 (τ2 − s)α ds 0 (s − z)β
Z t Z τ1 γ
(1 − q) 1 τ2
− b (t + 1)10a
u(x, τ2 )dτ2 dτ1
2
(x + 1) 0 0 (τ2 + 1)a
Z x Z t Z τ1 γ
1 y τ2
−4b(1 − q) u(y, τ2 )dτ2 dτ1 dy
(t + 1)10a 0 (y2 + 1)b+1 0 0 (τ2 + 1)a
γ
1 − (1 + 2b)w2 t τ1
Z xZ y
1
Z Z
τ2
+2b(1 − q) u(w, τ2 )dτ2 dτ1 dwdy
(t + 1)10a 0 0 (w2 + 1)b+2 0 0 (τ2 + 1)
a
Z xZ y Z t Z τ1
1 1 1
−(1 − q) |u(w, τ2 )| p+1 dτ2 dτ1 dwdy

(t + 1)10a 0 0 (w2 + 1)b 0 0 (τ2 + 1)a
≤ ε|u(x,t)|
(1 − q) 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a
sk d
Z xZ y Z t Z s
1 1 u(w, z)
Z τ
× dz

(w2 + 1)b (τ + 1)a (τ − s)α ds (s − z)
β
dsdτdwdy
0 0 0 0 0
Z xZ y
a(1 − q) 1 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a 0 0 (w2 + 1)b
sk
Z t Z τ1 Z s
1 d u(w, z)
Z τ2

(τ2 + 1)a+1 (τ2 − s)α ds (s − z)β

0 0 0 0
Z t Z τ1 γ
(1 − q) 1 τ2
+ b (t + 1)10a
|u(x, τ2 )|dτ2 dτ1
2
(x + 1) 0 0 (τ2 + 1)a
Z x Z t Z τ1 γ
1 y τ2
+4b(1 − q) |u(y, τ2 )|dτ2 dτ1 dy
(t + 1)10a 0 (y2 + 1)b+1 0 0 (τ2 + 1)a
γ
|1 − (1 + 2b)w2 | t τ1
Z xZ y
1
Z Z
τ2
+2b(1 − q) |u(w, τ2 )|dτ2 dτ1 dwdy
(t + 1)10a 0 0 (w2 + 1)b+2 0 0 (τ2 + 1)a
Z xZ y Z t Z τ1
1 1 1
+(1 − q) |u(w, τ2 )| p+1 dτ2 dτ1 dwdy
(t + 1)10a 0 0 (w2 + 1)b 0 0 (τ2 + 1)a
1−q 1
≤ εQ + A1 Q
Γ(1 − α)Γ(1 − β ) (t + 1)10a
!
sk s1−β
Z t
1
Z τ
−β
× s + dsdτ
0 (τ + 1)a 0 (τ − s)α 1−β
a(1 − q) 1
+ A1 Q
Γ(1 − α)Γ(1 − β ) (t + 1)10a
!
sk s1−β
Z t Z τ1
1
Z τ2
−β
× s + dsdτ2 dτ1
0 0 (τ2 + 1)a+1 0 (τ2 − s)α 1−β
Z t Z τ1 γ
1 τ2
+(1 − q)Q dτ2 dτ1
(t + 1)10a 0 0 (τ2 + 1)a
Z t Z τ1 γ
1 τ2
+4b(1 − q)QA1 dτ2 dτ1
(t + 1)10a 0 0 (τ2 + 1)a
Z t Z τ1 γ
1 τ2
+2b(1 − q)A1 Q dτ2 dτ1
(t + 1)10a 0 0 (τ2 + 1)a
Z t Z τ1
1 1
+(1 − q)A1 Q p+1 dτ1 dτ1
(t + 1)10a 0 0 (τ2 + 1)a
1−q 2−β
≤ εQ + A1 Q
Γ(1 − α)Γ(1 − β ) (a − 1)(1 − α)(1 − β )
a(1 − q) 4 − 2β
+A1 Q
Γ(1 − α)Γ(1 − β ) (a2 − 1)(1 − α)(1 − β )
1 1
+(1 − q)Q + 4b(1 − q)A1 Q
a−γ −1 a−γ −1
1
+2b(1 − q)A1 Q + (1 − q)A1 Q p+1
a−γ −1
(a(6 − 3β ) + 2 − β )(1 − q)
= εQ + A1 Q
(a2 − 1)(1 − α)(1 − β )Γ(1 − α)Γ(1 − β )
(1 − q)Q
+ (1 + 6bA1 ) + (1 − q)A1 Q p+1
a−γ −1
≤ Q, t ≥ 0, x ≥ 0.
3. Let u ∈ N. Then,
S(u)t (x,t) = −εut (x,t)
1−q 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)11a
sk
Z xZ y Z t Z s
1 d u(w, z)
× dzdsdwdy
0 0 (w2 + 1)b 0 (t − s)α ds 0 (s − z)β
Z xZ y
10a(1 − q) 1
−
Γ(1 − α)Γ(1 − β )(t + 1)10a+1 0 0 (w2 + 1)b
sk
Z t
1 d s u(w, z)
Z τ Z
× dzdsdτdwdy
0 (τ + 1)a 0 (τ − s)α ds 0 (s − z)β
Z xZ y
a(1 − q) 1
+
Γ(1 − α)Γ(1 − β )(t + 1)10a 0 0 (w2 + 1)b
sk
Z t
1 d s u(w, z)
Z τ2 Z
× dzdsdτ2 dwdy
0 (τ2 + 1)a+1 0 (τ2 − s)α ds 0 (s − z)β
10a2 (1 − q)
Z xZ y
1
−
Γ(1 − α)Γ(1 − β )(t + 1)10a+1 0 0 (w2 + 1)b
sk
Z t Z τ1 Z s
1 d u(w, z)
Z τ2
0 0 (τ2 + 1)a+1 0 (τ2 − s)α ds 0 (s − z)β
Z t γ
1−q 1 τ2
− u(x, τ2 )dτ2
(x2 + 1)b (t + 1)10a 0 (τ2 + 1)a
Z t Z τ1 γ
10a(1 − q) τ2
+ u(x, τ2 )dτ2 dτ1
(x + 1)b (t + 1)10a+1
2
0 0 (τ2 + 1)a
γ
4b(1 − q) x
Z t
y
Z
τ2
− u(y, τ2 )dτ2 dy
(t + 1)10a 0 (y2 + 1)b+1 0 (τ2 + 1)a
Z x Z t Z τ1 γ
40ab(1 − q) y τ2
+ u(y, τ2 )dτ2 dτ1 dy
(t + 1)10a+1 0 (y2 + 1)b+1 0 0 (τ2 + 1)a
γ
1 − (1 + 2b)w2 t
Z xZ y
2b(1 − q)
Z
τ2
+ u(w, τ2 )dτ2 dwdy
(t + 1)10a 0 0 (w2 + 1)b+2 0 (τ2 + 1)a
γ
20ab(1 − q) x y 1 − (1 + 2b)w2 t τ1
Z Z Z Z
τ2
− u(w, τ2 )dτ2 dτ1 dwdy
(t + 1)10a+1 0 0 (w2 + 1)b+2 0 0 (τ2 + 1)a
Z xZ y Z t
1−q 1 1
− |u(w, τ2 )| p+1 dτ2 dwdy
(t + 1)10a 0 0 (w2 + 1)b 0 (τ2 + 1)a
10a(1 − q) x y
Z t Z τ1
1 1
Z Z
+ |u(w, τ2 )| p+1 dτ2 dτ1 dwdy,
(t + 1)10a+1 0 0
2
(w + 1)b 0 0 (τ2 + 1)a
x ≥ 0, t ≥ 0.
Hence, using Lemma 3.1.7, Lemma 3.1.9, Lemma 3.1.11, Lemma 3.1.13, Lemma
3.1.15 and Lemma 3.1.17, we obtain
|S(u)t (x,t)| ≤ ε|ut (x,t)|
1−q 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)11a
sk
Z xZ y Z t Zs
1 d u(w, z)
× 2 ds 0 (s − z)β dz dsdwdy

0 0 (w + 1)b 0 (t − s)α
10a(1 − q) x y 1
Z Z
+ 10a+1 2 b
Γ(1 − α)Γ(1 − β )(t + 1) 0 0 (w + 1)
sk
Z t Zs
1 d u(w, z)
Z τ
× a
dz dsdτdwdy
0 (τ + 1) 0 (τ − s) 0 (s − z)β
α ds
Z xZ y
a(1 − q) 1
+
Γ(1 − α)Γ(1 − β )(t + 1)10a 0 0 (w2 + 1)b
sk
Z t Zs
1 d u(w, z)
Z τ2
× a+1
dz dsdτ2 dwdy
0 (τ2 + 1) 0 (τ2 − s) 0 (s − z)β
α ds
10a2 (1 − q)
Z xZ y
1
+
Γ(1 − α)Γ(1 − β )(t + 1)10a+1 0 0 (w2 + 1)b
sk
Z t Z τ1 Zs
1 d u(w, z)
Z τ2
× ds 0 (s − z)β dz dsdτ2 dτ1 dwdy

0 0 (τ2 + 1)a+1 0 (τ2 − s)α
Z t γ
1−q 1 τ2
+ |u(x, τ2 )|dτ2
(x2 + 1)b (t + 1)10a 0 (τ2 + 1)a
Z t Z τ1 γ
10a(1 − q) τ2
+ |u(x, τ2 )|dτ2 dτ1
(x2 + 1)b (t + 1)10a+1 0 0 (τ2 + 1)a
Z x Z t γ
4b(1 − q) y τ2
+ |u(y, τ2 )|dτ2 dy
(t + 1)10a 0
2
(y + 1)b+1 0 (τ2 + 1)a
γ
40ab(1 − q) x
Z t Z τ1
y
Z
τ2
+ |u(y, τ2 )|dτ2 dτ1 dy
(t + 1)10a+1 0 (y2 + 1)b+1 0 0 (τ2 + 1)a
γ
2b(1 − q) x y |1 − (1 + 2b)w2 | t
Z Z Z
τ2
+ |u(w, τ2 )|dτ2 dwdy
(t + 1)10a 0 0 (w2 + 1)b+2 0 (τ2 + 1)a
γ
|1 − (1 + 2b)w2 | t τ1
Z xZ y
20ab(1 − q)
Z Z
τ2
+ |u(w, τ2 )|dτ2 dτ1 dwdy
(t + 1)10a+1 0 0 (w2 + 1)b+2 0 0 (τ2 + 1)a
Z xZ y Z t
1−q 1 1
+ |u(w, τ2 )| p+1 dτ2 dwdy
(t + 1)10a 0 0 (w2 + 1)b 0 (τ2 + 1)a
10a(1 − q) x y
Z t Z τ1
1 1
Z Z
+ |u(w, τ2 )| p+1 dτ2 dτ1 dwdy
(t + 1)10a+1 0 0 (w2 + 1)b 0 0 (τ2 + 1)a
2(1 − q)A1 Q
≤ εQ +
(1 − α)Γ(1 − α)Γ(1 − β )
(20 − 10β )a(1 − q)A1 Q
+
(a − 1)(1 − α)(1 − β )Γ(1 − α)Γ(1 − β )
a(2 − β )(1 − q)A1 Q
+
(a − 1)(1 − α)(1 − β )Γ(1 − α)Γ(1 − β )
(40 − 20β )a2 (1 − q)A1 Q 10a(1 − q)Q

+ + (1 − q)Q +
(a2 − 1)(1 − α)(1 − β )Γ(1 − α)Γ(1 − β ) a−γ −1
40ab(1 − q)A1 Q
+4b(1 − q)A1 Q +
a−γ −1
20abA1 Q
+2b(1 − q)A1 Q +
a−γ −1
+2(1 − q)A1 Q p+1
≤ Q, t ≥ 0, x ≥ 0.
4. Let u ∈ N. Then
S(u)x (x,t) = −εux (x,t)
Z x
(1 − q) 1 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a 0 (w2 + 1)b
sk
Z t Z s
1 d u(w, z)
Z τ
× dzdsdτdw
0 (τ + 1)a 0 (τ − s)α ds 0 (s − z)β
Z x
a(1 − q) 1 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a 0 (w2 + 1)b
sk
Z t Z τ1 Z s
1 d u(w, z)
Z τ2
× dzdsdτ2 dτ1 dw
0 0 (τ2 + 1)a+1 0 (τ2 − s)α ds 0 (s − z)β
Z t Z τ1 γ
2(1 − q)bx 1 τ2
+ u(x, τ2 )dτ2 dτ1
2
(x + 1)b+1 (t + 1)10a 0 0 (τ2 + 1)a
Z t Z τ1 γ
(1 − q) 1 τ2
− b (t + 1)10a
ux (x, τ2 )dτ2 dτ1
2
(x + 1) 0 0 (τ2 + 1)a
Z t Z τ1 γ
1 x τ2
−4b(1 − q) u(x, τ2 )dτ2 dτ1
(t + 1)10a (x2 + 1)b+1 0 0 (τ2 + 1)a
γ
1 − (1 + 2b)w2 t τ1
Z x
1
Z Z
τ2
+2b(1 − q) u(w, τ2 )dτ2 dτ1 dw
(t + 1)10a 0 (w2 + 1)b+2 0 0 (τ2 + 1)a
Z x Z t Z τ1
1 1 1
−(1 − q) |u(w, τ2 )| p+1 dτ2 dτ1 dw,
(t + 1)10a 0 (w2 + 1)b 0 0 (τ2 + 1)a
t ≥ 0, x ≥ 0.
Hence, using Lemma 3.1.7, Lemma 3.1.9, Lemma 3.1.11, Lemma 3.1.13 and
A1 ≥ 1, we obtain
|S(u)x (x,t)| ≤ ε|ux (x,t)|
(1 − q) 1 x 1
Z
+ 10a 2 b
Γ(1 − α)Γ(1 − β ) (t + 1) 0 (w + 1)
sk
Z t Zs
1 d u(w, z)
Z τ
× a
dz dsdτdw
0 (τ + 1) 0 (τ − s) 0 (s − z)β
α ds
a(1 − q) 1 x 1
Z t Z Z τ1
1
+ 10a 2
Γ(1 − α)Γ(1 − β ) (t + 1) 0 (w + 1)
b
0 0 (τ2 + 1)a+1
sk

Z τ2 d Z s u(w, z)
× dz dsdτ2 dτ1 dw
0 (τ2 − s) 0 (s − z)β
α ds
Z t Z τ1 γ
2(1 − q)bx 1 τ2
+ |u(x, τ2 )|dτ2 dτ1
(x2 + 1)b+1 (t + 1)10a 0 0 (τ2 + 1)a
Z t Z τ1 γ
(1 − q) 1 τ2
+ b (t + 1)10a
|ux (x, τ2 )|dτ2 dτ1
2
(x + 1) 0 0 (τ2 + 1)a
Z t Z τ1 γ
1 x τ2
+4b(1 − q) |u(y, τ2 )|dτ2 dτ1
(t + 1)10a (x2 + 1)b+1 0 0 (τ2 + 1)a
γ
|1 − (1 + 2b)w2 | t τ1
Z x
1
Z Z
τ2
+2b(1 − q) |u(w, τ2 )|dτ2 dτ1 dw
(t + 1)10a 0 (w2 + 1)b+2 0 0 (τ2 + 1)a
Z x Z t Z τ1
1 1 1
+(1 − q) |u(w, τ2 )| p+1 dτ2 dτ1 dw
(t + 1)10a 0 (w2 + 1)b 0 0 (τ2 + 1)a
1−q 2−β
≤ εQ + A1 Q
Γ(1 − α)Γ(1 − β ) (a − 1)(1 − α)(1 − β )
a(1 − q) 4 − 2β 2(1 − q)bQ
+A1 Q 2
+
Γ(1 − α)Γ(1 − β ) (a − 1)(1 − α)(1 − β ) a−γ −1
1 1
+(1 − q)Q + 4b(1 − q)A1 Q
a−γ −1 a−γ −1
1
+2b(1 − q)A1 Q + (1 − q)A1 Q p+1
a−γ −1
((6 − 3β )a + 2 − β )(1 − q)
= εQ + A1 Q
(a2 − 1)(1 − α)(1 − β )Γ(1 − α)Γ(1 − β )
(1 − q)Q
+ (1 + 8bA1 ) + (1 − q)A1 Q p+1
a−γ −1
≤ Q, t ≥ 0, x ≥ 0.
5. For u ∈ N we have
S(u)xx (x,t) = −εuxx (x,t)
Z t
1 1−q 1 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a (x2 + 1)b 0 (τ + 1)a
sk
Z s
d u(w, z)
Z τ
× dzdsdτ
0 (τ − s)α ds 0 (s − z)β
a 1−q 1
+
Γ(1 − α)Γ(1 − β ) (t + 1)10a (x2 + 1)b
sk
Z t Z τ1 Z s
1 d u(w, z)
Z τ2
× dzdsdτ2 dτ1
0 0 (τ2 + 1)a+1 0 (τ2 − s) ds
α
0 (s − z)β
Z t Z τ1 γ
1−q 1 τ2
− uxx (x, τ2 )dτ2 dτ1
(t + 1)10a (x2 + 1)b 0 0 (τ2 + 1)a
|u(x, τ2 )| p+1
Z t Z τ1
1−q
− dτ2 dτ1 , x ≥ 0, t ≥ 0.
(t + 1)10a 0 0 (τ2 + 1)a
Hence, using Lemma 3.1.7, Lemma 3.1.9, Lemma 3.1.11, Lemma 3.1.13, Lemma
3.1.15 and Lemma 3.1.17 we obtain
|S(u)xx (x,t)| ≤ ε|uxx (x,t)|
1 1 1−q
+
Γ(1 − α)Γ(1 − β ) (x + 1) (t + 1)10a
2 b
sk
Z t Zs
1 d u(w, z)
Z τ
× a
dz dsdτ
0 (τ + 1) 0 (τ − s) 0 (s − z)β
α ds
a 1−q 1
+ 10a 2
Γ(1 − α)Γ(1 − β ) (t + 1) (x + 1)b
sk
Z t Z τ1 Zs
1 d u(w, z)
Z τ2
× a+1
dz dsdτ2 dτ1
0 0 (τ2 + 1) 0 (τ2 − s) 0 (s − z)β
α ds
BVP of Nonlinear Riemann-Liouville Fractional Differential Equations 153
Z t Z τ1 γ
1−q 1 τ2
+ 10a 2
|uxx (x, τ2 )|dτ2 dτ1
(t + 1) (x + 1)b 0 0 (τ2 + 1)a
|u(x, τ2 )| p+1
Z t Z τ1
1−q
+ dτ2 dτ1
(t + 1)10a 0 0 (τ2 + 1)a
1−q 2−β
≤ εQ + Q
Γ(1 − α)Γ(1 − β ) (a − 1)(1 − α)(1 − β )
a(1 − q) 4 − 2β
+ Q
Γ(1 − α)Γ(1 − β ) (a2 − 1)(1 − α)(1 − β )
1−q
+ Q + (1 − q)Q p+1
a−γ −1
≤ Q, t ≥ 0, x ≥ 0.
From items 1-5 it follows that S : N → N and it is continuous. Since N is a compact subset
of N1 , we have that S(N) resides in a compact subset of N1 . This completes the proof.
From Lemma 3.1.5, Lemma 3.1.19 and Theorem 1.9.12 it follows that the operator R has a
fixed point u ∈ N. Hence, using Lemma 3.1.3, we conclude that u is a C 1 ([0, ∞), C 2 (R))-
solution to the problem (3.3), (3.4). This completes the proof.
3.2 Multiple Solutions for a BVP of Nonlinear Riemann-Liouville

Fractional Differential Equations
Let ε, r1 , r2 , ρ, R, s, M0 , M1 , M2 be positive constants which satisfy
1
0 < r1 < ρ < r2 < R, 0<s< , (3.7)
2
1
m k

R+ M0 + M1 R + M2 R < s, (3.8)
Γ(α)
M0 + M1 rim + M2 rik < ri , i = 1, 2, (3.9)
1
M0 + M1 Rm + M2 Rk < ρ. (3.10)
Γ(α)
Consider the BVP
Dα0+ x(t) + f (t, x(t)) = 0, 0 < t < 1, (3.11)
x(0) = x(1) = 0, (3.12)

where 1 < α ≤ 2, Dα0+ is the standard Riemann-Liouville fractional differentiation,
(H1) f ∈ C ([0, 1] × [0, ∞)), f : [0, 1] × [0, ∞) → [0, ∞),
0 ≤ f (t, x) ≤ a0 (t) + a1 (t)xm + a2 (t)xk , t ∈ [0, 1], x ∈ [0, ∞),
ai ∈ C ([0, 1]), ai : [0, 1] → [0, ∞),
ai (t) ≤ Mi , t ∈ [0, 1], i = 0, 1, 2.
Define 
(t(1 − s))α−1 − (t − s)α−1
, 0 ≤ s ≤ t ≤ 1,





 Γ(α)
G(t, s) =

(t(1 − s))α−1


, 0 ≤ t ≤ s ≤ 1.



Γ(α)
Lemma 3.2.1 Suppose (H1), 1 < α ≤ 2. Then the unique solution of BVP (3.11), (3.12) is
given by
Z 1
x(t) = G(t, s) f (s, x(s))ds, 0 ≤ t ≤ 1.
0
Proof 3.2.2 Equation (3.11) is equivalent to the following integral equation

α
x(t) = −I0+ f (t, x(t)) + c1t α−1 + c2t α−2
Z t
1
= − (t − s)α−1 f (s, x(s))ds + c1t α−1 + c2t α−2 , 0 < t < 1,
Γ(α) 0
where c1 , c2 ∈ R. By (3.12), we get c2 = 0 and

Z 1
1
x(1) = − (1 − s)α−1 f (s, x(s))ds + c1
Γ(α) 0
= 0.
Then Z 1
1
c1 = (1 − s)α−1 f (s, x(s))ds.
Γ(α) 0
Hence,
Z t Z 1
1 1
x(t) = − (t − s)α−1 f (s, x(s))ds + (1 − s)α−1t α−1 f (s, x(s))ds
Γ(α) 0 Γ(α) 0
Z t Z t
1 1
= − (t − s)α−1 f (s, x(s))ds + (1 − s)α−1t α−1 f (s, x(s))ds
Γ(α) 0 Γ(α) 0
Z 1
1
+ (1 − s)α−1t α−1 f (s, x(s))ds
Γ(α) t
Z 1
= G(t, s) f (s, x(s))ds, 0 ≤ t ≤ 1.
0
Note that
1
0 ≤ G(t, s) ≤ , s,t ∈ [0, 1].
Γ(α)
Let E = C ([0, 1]) be endowed with the maximum norm
kxk = max |x(t)|.

t∈[0,1]
Define
P
f = {x ∈ E : x(t) ≥ 0, t ∈ [0, 1]}.
Let P be the set of all equi-continuous families in P.
f
Theorem 3.2.3 Suppose (H1), (3.7)-(3.10). Then BVP (3.11), (3.12) has at least three

t∈[0,1] t∈[0,1] t∈[0,1]
Proof 3.2.4 Define
PR = {x ∈ P : kxk < R},
Pρ = {x ∈ P : kxk < ρ},
Pri = {x ∈ P : kxk < ri }, i = 1, 2.
T x(t) = (1 + ε)x(t) + εR,
Z 1
Fx(t) = −εR − ε G(t, s) f (s, x(s))ds, t ∈ [0, 1].
0
Using Lemma 3.2.1, we have that if x ∈ E is a fixed point of the operator T + F, then it is
a solution of BVP (3.11), (3.12).
1. We have T : PR → E and
kT x − Tyk = (1 + ε)kx − yk
for any x, y ∈ PR , i.e., T : PR → E is an expansive operator with a constant

h = 1 + ε.
2. Let x ∈ PR . Then
Z 1

|Fx(t)| = −εR − ε
G(t, s) f (s, x(s))ds
0
Z 1
≤ εR + ε G(t, s) f (s, x(s))ds
0
Z 1
≤ εR + ε G(t, s) a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k ds
0
ε
≤ εR + M0 + M1 Rm + M2 Rk
Γ(α)
< εs
ε
< , t ∈ [0, 1].
2
Then F : PR → E is uniformly bounded. Since F : PR → E is continuous,
we conclude that F(PR ) is equi-continuous. Therefore F : PR → E is 0-set
contraction.
Z1

|Fx(t) + T 0(t)| = −ε G(t, s) f (s, x(s))ds
0
Z 1
= ε G(t, s) f (s, x(s))ds
0
Z 1
≤ ε G(t, s) a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k ds
0
1
≤ ε M0 + M1 rim + M2 rik
Γ(α)
≤ εri
= εkxk, t ∈ [0, 1],

whereupon
Hence, conditions (a) and (c) of Theorem 1.9.21 hold.
4. Let u0 ∈ P ∗ and assume that there exist x1 ∈ P and λ1 ≥ 0 with x1 −λ1 u0 ∈
∂ Pρ such that
Fx1 = (I − T )(x1 − λ1 u0 ).
Let t1 ∈ [0, 1] be such that
ρ = x1 (t1 ) − λ1 u0 (t1 ).
Then
Z 1
ε
εR + M0 + M1 Rm + M2 Rk ≥ εR + ε G(t1 , s) f (s, x1 (s))ds
Γ(α) 0
= −(Fx1 )(t1 )
= −(I − T )(x1 − λ1 u0 )(t1 )
= ε(x1 − λ1 u0 )(t1 ) + εR
= ερ + εR,
whereupon
1
M0 + M1 Rm + M2 Rk ≥ ρ,
Γ(α)
which is a contradiction. Therefore condition (b) of Theorem 1.9.21 holds.
F(PR ) ⊂ (I − T )(PR ). (3.13)
Let x ∈ PR be fixed and consider on PR the following nonlinear integral equa-
tion
x(t) = T x(t) + Fz(t)
= Ax(t), t ∈ [0, 1].

= (1 + ε)kx − yk,
PR ⊂ A(PR ). (3.14)
For any given v ∈ PR consider on PR the following equation
Ax(t) = v(t) or T x(t) = v(t) − Fz(t), t ∈ [0, 1].
We have
v(t) − Fz(t) ≥ 0, t ∈ R,
and
Z 1
v(t) − Fz(t) = v(t) + εR + ε G(t, s) f (s, x(s))ds
0
Z 1
ε
≤ (1 + ε)R + a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k ds
Γ(α) 0
ε
≤ (1 + ε)R + M0 + M1 Rm + M2 Rk
Γ(α)
= R1 , t ∈ [0, 1].
Note that R1 > R and v − Fz ∈ B(R1 ) P. Next, T : PR → T (PR ) is a bijec-

\
tion, where
T (PR ) = {x ∈ P : kxk ≤ R1 }.
Therefore there exists a u ∈ PR such that
Tu(t) = v(t) − Fz(t), t ∈ [0, 1],
and then (3.14) holds. Hence, and applying Theorem 1.9.8, it follows that there
exists a w ∈ PR such that
w(t) = Aw(t), t ∈ [0, 1],
and from here, (3.13) holds.

From Theorem 1.9.21, we get that BVP (3.11), (3.12) has at least three solutions x1 , x2 , x3 ∈
P such that

t∈[0,1] t∈[0,1] t∈[0,1]

Let ε, r, R, L, M0 , M1 , M2 , s be positive constants such that

1
r < L < R, s< , ε > 1, (3.15)
2
1
M0 + M1 Lm + M2 Lk ≤ L. (3.16)
Γ(α)
Theorem 3.2.5 Suppose (H8), (3.8), (3.15), (3.16). Then BVP (3.11), (3.12) has at least
r < max x(t) < L < max x(t) < R.
t∈[0,1] t∈[0,1]
Proof 3.2.6 Let PR be as in the proof of Theorem 3.3.3. Let also

Pr = {x ∈ P : kxk < r},
PL = {x ∈ P : kxk < L},
Pr,L = {x ∈ P : r < kxk < L},
PL,R = {x ∈ P : L < kxk < R}.

T x(t) = −(1 + ε)x(t) − εR,
Z 1
Fx(t) = εR + ε G(t, s) f (s, x(s))ds, t ∈ [0, 1].
0
F(PR ) ⊂ (I − T )(PR ).
2. Let u0 ∈ P ∗ and assume that there are a λ1 ≥ 0 and an x1 ∈ P with x1 −

λ1 u0 ∈ ∂ Pr and
Fx1 = x1 − λ1 u0 .
Then
r ≥ x1 (t) − λ1 u0 (t)
= Fx1 (t)
Z 1
= εR + ε G(t, s) f (s, x(s))ds
0
≥ εR, t ∈ [0, 1],
which is a contradiction. Therefore condition (a) of Theorem 1.9.23 holds. As

in the above, one can prove condition (c) of Theorem 1.9.23 holds.
Z 1
|Fx(t) + T 0(t)| = ε G(t, s) f (s, x(s))ds
0
Z 1
≤ ε G(t, s) a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k ds
0
ε
≤ M0 + M1 Lm + M2 Lk
Γ(α)
≤ εL
= εkxk, t ∈ [0, 1].
Hence,
Now, assume that x1 ∈ ∂ PL such that
x1 = T x1 + Fx1 .
Let t1 ∈ [0, 1] be such that

L = x1 (t1 ).
Then
L = x1 (t1 )
Z 1
= −(1 + ε)x1 (t1 ) − εR + εR + ε G(t1 , s) f (s, x(s))ds
0
Z 1
≤ ε G(t1 , s) a0 (s) + a1 (s)(x(s))m + a2 (s)(x(s))k ds − (1 + ε)L
0
ε
≤ M0 + M1 Lm + M2 Lk − (1 + ε)L
Γ(α)
≤ εL − (1 + ε)L
= −L,
which is a contradiction. Therefore condition (b) of Theorem 1.9.23 holds.

By Theorem 1.9.23, it follows that BVP (3.11), (3.12) has at least two solutions x1 , x2 ∈ P
such that
r < max x1 (t) < L < max x(t) < R.
t∈[0,1] t∈[0,1]
Let ε, ρ, d, s,C0 ,C1 ,C2 , m, k be positive constants and d0 , d1 , d2 ∈ C ([0, 1]) be non-negative
functions such that
1
0 < ρ < d, s > , (3.17)
2
1
C0 +C1 d m +C2 d k ≤ d, (3.18)
Γ(α)
1
d+ C0 +C1 d m +C2 d k < s, (3.19)
Γ(α)
1
C0 +C1 ρ m +C2 ρ k < ρ, (3.20)
Γ(α)
Z 1
1
d0 (s)ds ≥ d, (3.21)
Γ(α) 0
Z 1
ε
m
d1 (s)ds > 1, (3.22)
2 Γ(α) 0
Z 1
ε
k
d2 (s)ds > 1. (3.23)
2 Γ(α) 0
Suppose that
(H2) f : [0, 1] × [0, ∞) → (−∞, 0],
| f (t, x)| ≤ c0 (t) + c1 (t)xm + c2 (t)xk ,
f (t, x) ≤ −d0 (t) − d1 (t)xm − d2 (t)xk , t ∈ [0, 1], x ∈ [0, ∞),
where ci ∈ C ([0, 1]), ci : [0, 1] → [0, ∞),
ci (t) ≤ Ci , t ∈ [0, 1], i = 0, 1, 2.

Theorem 3.2.7 Assume (H2), (3.17)-(3.23). Then BVP (3.11), (3.12) has at least two so-
0 < kx1 k < ρ < kx2 k < d.
Proof 3.2.8 Let
Pρ = {x ∈ P : kxk < ρ},
Pd = {x ∈ P : kxk < d},
Pρ,d = {x ∈ P : ρ < kxk < d}.

T x(t) = (1 + ε)x(t) + εd,
Z 1
Fx(t) = −εd − ε G(t, s) f (s, x(s))ds, t ∈ [0, 1].
0
1. We will prove that

F(Pd ) ⊂ (I − T )(Pd ). (3.24)
Let z ∈ Pd be arbitrarily chosen. For x ∈ Pd consider the equation
x = T x + Fz
= Ax.
We have
= (1 + ε)kx − yk, x, y ∈ Pd .
Therefore A : Pd → E is an expansive operator. Now we will prove that
Pd ⊂ A(Pd ).
For any v ∈ Pd consider on Pd the equation
v = Au
= Tu + Fz
or
Tu = v − Fz. (3.25)
We have
Z 1
v(t) − Fz(t) = v(t) + εd + ε G(t, s) f (s, x(s))ds
0
ε
≥ εd − C0 +C1 d m +C2 d k
Γ(α)
≥ 0, t ∈ [0, 1],
and
Z 1
|v(t) − Fz(t)| ≤ v(t) + εd + ε G(t, s) f (s, x(s))ds
0
Z 1
≤ d + εd + ε G(t, s) c0 (s) + c1 (s)(x(s))m + c2 (s)(x(s))k ds
0

≤ (1 + ε)d + ε C0 +C1 d m +C2 d k
= d1 , t ∈ [0, 1].
Thus
P.
\
v − Fz ∈ B(d1 )
Since F : Pd → T (Pd ) = B(d1 ) P is a bijection, there is a u ∈ Pd such

\
that (3.25) holds. Hence, and applying Theorem 1.9.8, it follows that there is a
w ∈ Pd so that
w(t) = Tw(t) + Fz(t), t ∈ [0, 1],
and (3.24) holds.
2. As in the proof of Theorem 3.2.3, we have that F is a 0-set contraction.
3. Let x ∈ ∂ Pρ . Then
Z1

|Fx(t) + T 0(t)| = −ε
G(t, s) f (s, x(s))ds
0
Z 1
≤ ε G(t, s)| f (s, x(s))|ds
0
Z 1
≤ ε G(t, s) c0 (s) + c1 (s)(x(s))m + c2 (s)(x(s))k ds
0
ε
≤ C0 +C1 ρ m +C2 ρ k
Γ(α)
≤ ερ
= εkxk, t ∈ [0, 1],
whereupon
kFx + T 0k ≤ εkxk,
Z 1
kFx + T xk ≥ (1 + ε)ρ − ε G(t, s)| f (s, x(s))|ds
0
ε
≥ (1 + ε)ρ − C0 +C1 ρ m +C2 ρ k
Γ(α)
> (1 + ε)ρ − ερ
= ρ.
For x ∈ P, define the operators

Z 1
ε
F1 x(t) = d1 (s)(x(s))m ds,
Γ(α) 0
Z 1
ε
F2 x(t) = d2 (s)(x(s))k ds, t ∈ [0, 1].
Γ(α) 0
Then
Z 1
ε
Fx(t) ≥ −εd + d0 (s) + d1 (s)(x(s))m + d2 (s)(x(s))k ds
Γ(α) 0
Z 1
ε
≥ −εd + d0 (s)ds
Γ(α) 0
≥ 0, t ∈ [0, 1],
and
Fx(t) ≥ Fi x(t), t ∈ [0, 1], i = 1, 2.
Take c1 = c2 = 1. Let m1 and m2 be defined as in Theorem 1.9.25. Then

m Z 1
1 1
m1 ≥ ε d1 (s)ds
2 Γ(α) 0
> 1,
k Z 1
1 1
m2 ≥ ε d2 (s)ds
2 Γ(α) 0
> 1.
4. Let
1 − µ 1−m
ξ1 (µ) = ,
µ 1−m
1
ξ2 (µ) = 1 − µ k−1 , µ ∈ (0, 1).
2
Then
µm = µ(1 + ξ1 (µ)),
1
1 − ξ2 (µ) = (1 + µ k−1 )
2
≥ µ k−1 ,
µ(1 − ξ2 (µ)) ≥ µ k, µ ∈ (0, 1).

Note that P is a normal cone with the constant N = 1, and
1
− 1 < 0 and 1 − m2 < 0.
m1
We have
1
lim ξ1 (µ) > − 1,
µ→0+ m1
lim ξ2 (µ) > 1 − m2 .

µ→0+
Next,
F1 (µx) = µ m F1 x
= µ(1 + ξ1 (µ))F1 x,
F2 (µx) = µ k F2 x
≤ µ(1 − ξ2 (µ))F2 x, µ ∈ (0, 1).

Hence, and applying Theorem 1.9.25, it follows that BVP (3.11), (3.12) has at least two
non-negative continuous solutions x1 , x2 such that
0 < kx1 k < ρ < kx2 k < d.
3.3 Multiple Solutions for a BVP of Nonlinear Caputo Fractional

Differential Equations
Consider the following BVP
c 1
Dα x(t) + f (t, x(t)) = 0, 0 < t < 1, (3.26)
2
x(0) = x(1) = 0, x00 (0) = 0, (3.27)

where f ∈ C ([0, 1] × [0, ∞)), 2 < α ≤ 3, c Dα is the standard Caputo fractional differentia-
tion. Define

t(1 − s)α−1 − (t − s)α−1
, 0 ≤ s ≤ t ≤ 1,




 2Γ(α)
G(t, s) =
(1 − t)(1 − s)α−1


, 0 ≤ t ≤ s ≤ 1.



2Γ(α)

Lemma 3.3.1 BVP (3.26), (3.27) has a unique solution given by
Z 1
x(t) = G(t, s) f (s, x(s))ds, t ∈ [0, 1].
0
BVP of Nonlinear Caputo Fractional Differential Equations 167
Proof 3.3.2 Equation (3.26) is equivalent to the following integral equation

Z t
1
x(t) = − (t − s)α−1 f (s, x(s))ds + c1 + c2t, t ∈ [0, 1],
2Γ(α) 0
where c1 , c2 ∈ R. We have
x(0) = c1 = 0.
Then Z t
1
x(t) = − (t − s)α−1 f (s, x(s))ds + c2t, t ∈ [0, 1].
2Γ(α) 0
Hence,
0 = x(1)
Z 1
1
= − (1 − s)α−1 f (s, x(s))ds + c2 ,
2Γ(α) 0
whereupon
Z 1
1
c2 = (1 − s)α−1 f (s, x(s))ds.
2Γ(α) 0
Therefore
Z t Z 1
1 t
x(t) = − (t − s)α−1 f (s, x(s))ds + (1 − s)α−1 f (s, x(s))ds
2Γ(α) 0 2Γ(α) 0
Z t Z t
1 t
= − (t − s)α−1 f (s, x(s))ds + (1 − s)α−1 f (s, x(s))ds
2Γ(α) 0 2Γ(α) 0
Z 1
t
+ (1 − s)α−1 f (s, x(s))ds
2Γ(α) t
Z t Z 1
1 t
= t(1 − s)α−1 − (t − s)α−1 f (s, x(s))ds + (1 − s)α−1 f (s, x(s))ds
2Γ(α) 0 2Γ(α) t
Z 1
= G(t, s) f (s, x(s))ds, t ∈ [0, 1].
0
We have
1
0 ≤ G(t, s) ≤ , t, s ∈ [0, 1].
Γ(α)
As we have proved the main results in the previous section, one can deduct the following
results for BVP (3.26), (3.27).
Theorem 3.3.3 Suppose (H1), (3.7)-(3.10). Then BVP (3.26), (3.27) has at least three

t∈[0,1] t∈[0,1] t∈[0,1]
Theorem 3.3.4 Suppose (H8), (3.8), (3.15), (3.16). Then BVP (3.26), (3.27) has at least
r < max x(t) < L < max x(t) < R.

t∈[0,1] t∈[0,1]
Theorem 3.3.5 Assume (H2), (3.17)-(3.23). Then BVP (3.26), (3.27) has at least two so-
0 < kx1 k < ρ < kx2 k < d.

4
Applications to Parabolic Equations

Conditions of an IVP
Here, we consider the Cauchy problem
ut − uxx = f (t, x, u, ux ) in (0, ∞) × R, (4.1)
u(0, x) = φ (x) in R, (4.2)

where φ ∈ C 2 (R), f : [0, ∞) × R × R × R → C is a given continuous function, u : [0, ∞) ×
R → C is the main unknown. We will start our investigations in this section with the fol-
lowing useful lemma.
Lemma 4.1.1 Let f ∈ C ([a, b] × [c, d] × R × R), g ∈ C 2 ([c, d]). Then the function u ∈
C 1 ([a, b], C 2 ([c, d])) is a solution to the problem
ut − uxx = f (t, x, u, ux ) in (a, b] × [c, d], (4.3)
u(a, x) = g(x) in [c, d], (4.4)

if and only if it is a solution to the integral equation
Z xZ y Z t
(u(t, z) − g(z)) dzdy − (u(τ, x) − u(τ, c) − (x − c)ux (τ, c)) dτ
c c a
Z tZ xZ y
(4.5)
= f (τ, z, u(τ, z), ux (τ, z))dzdydτ, x ∈ [c, d], t ∈ [a, b].
a c c
Proof 4.1.2 1. Let u ∈ C 1 ([a, b], C 2 ([c, d])) be a solution to problem (4.3),
(4.4).
We integrate equation (4.3) with respect to x and we get
Z x Z x
ut (t, z)dz − uxx (t, z)dz
c c
Z x
= f (t, z, u(t, z), ux (t, z))dz, x ∈ [c, d], t ∈ [a, b],
c
169
170 Applications to Parabolic Equations
or Z x
ut (t, z)dz − ux (t, x) + ux (t, c)
c
Z x
= f (t, z, u(t, z), ux (t, z))dz, x ∈ [c, d], t ∈ [a, b].
c
Now we integrate the last equation with respect to x and we find
Z xZ y Z x
ut (t, z)dzdy − (ux (t, z) − ux (t, c)) dz
c c c
Z xZ y
= f (t, z, u(t, z), ux (t, z))dzdy, x ∈ [c, d], t ∈ [a, b],
c c
or
Z xZ y
ut (t, z)dzdy − u(t, x) + u(t, c) + (x − c)ux (t, c)
c c
Z xZ y
= f (t, z, u(t, z), ux (t, z))dzdy, x ∈ [c, d], t ∈ [a, b].
c c
We integrate the last equality with respect to t and we obtain

Z tZ xZ y Z t
ut (s, z)dzdyds − (u(s, x) − u(s, c) − (x − c)ux (s, c)) ds
a c c a
Z tZ xZ y
= f (s, z, u(s, z), ux (s, z))dzdyds, x ∈ [c, d], t ∈ [a, b],
a c c
or
Z xZ y Z t
(u(t, z) − g(z)) dzdy − (u(s, x) − u(s, c) − (x − c)ux (s, c)) ds
c c a
Z tZ xZ y
= f (s, z, u(s, z), ux (s, z))dzdyds, x ∈ [c, d], t ∈ [a, b],
a c c
i.e., u satisfies equation (4.5).

2. Let u ∈ C 1 ([a, b], C 2 ([c, d])) be a solution to the integral equation (4.5).
We differentiate equation (4.5) with respect to x and we get
Z x Z t
(u(t, z) − g(z)) dz − (ux (s, x) − ux (s, c)) ds
c a
Z tZ x
= f (s, z, u(s, z), ux (s, z))dzds, x ∈ [c, d], t ∈ [a, b].
a c
Differentiability of Classical Solutions of Initial IVP Conditions 171
Again we differentiate with respect to x and we find

Z t
u(t, x) − g(x) − uxx (s, x)ds
a
Z t
(4.6)
= f (s, x, u(s, x), ux (s, x))ds, x ∈ [c, d], t ∈ [a, b].
a
Now we put t = a in the last equation and we find

u(a, x) = g(x), x ∈ [c, d],
i.e., the function u satisfies (4.4).
Now we differentiate equation (4.6) with respect to t and we find
ut (t, x) − uxx (t, x) = f (t, x, u(t, x), ux (t, x)), x ∈ [c, d],t ∈ [a, b].
Theorem 4.1.3 Let f ∈ C ([0, ∞) × R × R × R), φ ∈ C 2 (R). Then there exists m ∈ (0, 1)
such that problem (4.1), (4.2) has a solution u ∈ C 1 ([0, m], C 2 ([0, 1])).
Proof 4.1.4 Let B > kφ kC 2 ([0,1]) be arbitrarily chosen. Since φ ∈ C ([0, 1]), f ∈ C ([0, 1] ×
[0, 1] × [−B, B] × [−B, B]) we have that there exists a constant M11 > 0 such that
|φ | ≤ M11 in [0, 1],
| f | ≤ M11 in [0, 1] × [0, 1] × [−B, B] × [−B, B].

We take l, m ∈ (0, 1) so that
lB + l(B + M11 ) + 3lBm + lM11 m ≤ B
(4.7)
l(5B + 2M11 ) ≤ B.
Let E11 = C 1 ([0, m], C 2 ([0, 1])) be endowed with the norm
n
||u|| = max max |u(t, x)|, max |ut (t, x)|,
(t,x)∈[0,m]×[0,1] (t,x)∈[0,m]×[0,1]
o
max |ux (t, x)|, max |uxx (t, x)| .
(t,x)∈[0,m]×[0,1] (t,x)∈[0,m]×[0,1]
By Ke11 we denote the set of all equi-continuous families in E11 , i.e., for every ε > 0 there
exists δ = δ (ε) > 0 such that
|u(t1 , x1 ) − u(t2 , x2 )| < ε, |ut (t1 , x1 ) − ut (t2 , x2 )| < ε,
|ux (t1 , x1 ) − ux (t2 , x2 )| < ε, |uxx (t1 , x1 ) − uxx (t2 , x2 )| < ε
whenever |t1 − t2 | < δ , |x1 − x2 | < δ . Let also
0 0
K11 =K
e11 , K11 = {u ∈ K11 : ||u|| ≤ B}
and
0
L11 = {u ∈ K11 : ||u|| ≤ (1 + l)B}.
We note that K11 is a closed convex subset of L11 .
For u ∈ L11 , we define the operators
T11 (u)(t, x) = (1 + l)u(t, x),

Z xZ y
S11 (u)(t, x) = −lu(t, x) + l (u(t, z) − φ (z))dzdy
0 0
Z t
−l (u(τ, x) − u(τ, 0) − xux (τ, 0))dτ
0
Z tZ xZ y
−l f (τ, z, u(τ, z), ux (τ, z))dzdydτ, (t, x) ∈ [0, m] × [0, 1].
0 0 0
We will prove that the problem
ut − uxx = f (t, x, ux ) in [0, m] × [0, 1], (4.8)
u(0, x) = φ (x) in [0, 1], (4.9)

has a solution u ∈ C 1 ([0, m], C 2 ([0, 1])).
a) S11 : K11 → K11 . Let u ∈ K11 . Then S11 (u) ∈ C 1 ([0, m], C 2 ([0, 1])) and for (t, x) ∈
[0, m] × [0, 1], using the first inequality of (4.7), we get
Z xZ y
|S11 (u)(t, x)| = −lu(t, x) + l (u(t, z) − φ (z))dzdy

0 0
Z t
−l (u(τ, x) − u(τ, 0) − xux (τ, 0))dτ
0
Z tZ xZ y
−l f (τ, z, u(τ, z), ux (τ, z))dzdydτ

0 0 0
Z xZ y
≤ l|u(t, x)| + l (|u(t, z)| + |φ (z)|) dzdy
0 0
Z t
+l (|u(τ, x)| + |u(τ, 0)| + x|ux (τ, 0)|) dτ
0
Z tZ xZ y
+l | f (τ, z, u(τ, z), ux (τ, z))|dzdydτ
0 0 0
≤ lB + l(B + M11 ) + 3lBm + lM11 m
≤ B.
Note that
Z xZ y
S11 (u)t (t, x) = −lut (t, x) + l ut (t, z)dzdy
0 0
−l(u(t, x) − u(t, 0) − xux (t, 0))

Z xZ y
−l f (t, z, u(t, z), ux (t, z))dzdy,
0 0
(t, x) ∈ [0, m] × [0, 1].
Then, using the second inequality of (4.7), we obtain

Z xZ y
|S11 (u)t (t, x)| = −lut (t, x) + l ut (t, z)dzdy

0 0
−l(u(t, x) − u(t, 0) − xux (t, 0))

Z xZ y
−l f (t, z, u(t, z), ux (t, z))dzdy

0 0
Z xZ y
≤ l|ut (t, x)| + l |ut (t, z)|dzdy
0 0
+l (|u(t, x)| + |u(t, 0)| + x|ux (t, 0)|)

Z xZ y
+l | f (t, z, u(t, z), ux (t, z))|dzdy
0 0
≤ lB + lB + 3lB + lM11
= l(5B + M11 )
≤ B, (t, x) ∈ [0, m] × [0, 1].

Also,
Z x
S11 (u)x (t, x) = −lux (t, x) + l (u(t, z) − φ (z))dz
0
Z t
−l (ux (τ, x) − ux (τ, 0))dτ
0
Z tZ x
−l f (τ, z, u(τ, z), ux (τ, z))dzdτ,
0 0
(t, x) ∈ [0, m] × [0, 1].
Hence, using the first inequality of (4.7),

Z x
|S11 (u)x (t, x)| = −lux (t, x) + l (u(t, z) − φ (z))dz

0
Z t
−l (ux (τ, x) − ux (τ, 0))dτ
0
Z tZ x
−l f (τ, z, u(τ, z), ux (τ, z))dzdτ

0 0
Z x
≤ l|ux (t, x)| + l (|u(t, z)| + |φ (z)|) dz
0
Z t
+l (|ux (τ, x)| + |ux (τ, 0)|) dτ
0
Z tZ x
+l | f (τ, z, u(τ, z), ux (τ, z))|dzdτ
0 0
≤ lB + l(B + M11 ) + 2lBm + lM11 m
≤ B, (t, x) ∈ [0, m] × [0, 1].
For (t, x) ∈ [0, m] × [0, 1] we have
S11 (u)xx (t, x) = −luxx (t, x) + l(u(t, x) − φ (x))
Z t
−l uxx (τ, x)dτ
0
Z t
−l f (τ, x, u(τ, x), ux (τ, x))dτ,
0
from where, using the first inequality of (4.7),

|S11 (u)xx (t, x)| = −luxx (t, x) + l(u(t, x) − φ (x))

Z t
−l uxx (τ, x)dτ
0
Z t
−l f (τ, x, u(τ, x), ux (τ, x))dτ

0
≤ l|uxx (t, x)| + l (|u(t, x)| + |φ (x)|)
Z t
+l |uxx (τ, x)|dτ
0
Z t
+l | f (τ, x, u(τ, x), ux (τ, x))|dτ
0
≤ lB + l(B + M11 ) + lBm + lM11 m
≤ B.
We note that {S11 (u) : u ∈ K11 } is an equi-continuous family in E11 . Consequently S11 :
K11 → K11 . Also, S11 (K11 ) ⊂ K11 ⊂ L11 , i.e., S11 (K11 ) resides in a compact subset of L11 .
b) S11 : K11 → K11 is a continuous operator. We note that if {un }∞
n=1 is a sequence of el-
ements of K11 such that un −→ u in K11 as n −→ ∞, then S11 (un ) −→ S11 (u) in K11 as
n −→ ∞. Therefore S11 : K11 → K11 is a continuous operator.
c) T11 : K11 → L11 is an expansive operator and onto. For u, v ∈ K11 we have that
||T11 (u) − T11 (v)|| = (1 + l)||u − v||,
i.e., T11 : K11 → L11 is an expansive operator with constant 1 + l.

v
Let v ∈ L11 . Then ∈ K11 and
1+l

v
T11 = v,
1+l
i.e., T11 : K11 → L11 is onto.

From items a, b, and c, and from Theorem 1.9.12, it follows that there is u11 ∈ K11 such that
T11 u11 + S11 u11 = u11
or
Z xZ y
(1 + l)u11 (t, x) − lu11 (t, x) + l (u11 (t, z) − φ (z))dzdy
0 0
Z t
−l (u11 (τ, x) − u11 (τ, 0) − xu11x (τ, 0))dτ
0
Z tZ xZ y
−l f (τ, z, u11 (τ, z), u11x (τ, z))dzdydτ
0 0 0
= u11 (t, x),
or
Z xZ y Z t
(u11 (t, z) − φ (z))dzdy − (u11 (τ, x) − u11 (τ, 0) − xu11x (τ, 0))dτ
0 0 0
Z tZ xZ y
− f (τ, z, u11 (τ, z), u11x (τ, z))dzdydτ
0 0 0
= 0, (t, x) ∈ [0, m] × [0, 1],
whereupon, using Lemma 4.1.1, we conclude that u11 ∈ C 1 ([0, 1], C 2 ([0, 1])) is a solution
to problem (4.8), (4.9). This completes the proof.
Theorem 4.1.5 Let f ∈ C ([0, ∞) × R × R × R), φ ∈ C 2 (R). Then there exists m ∈ (0, 1)
such that problem (4.1), (4.2) has a solution u ∈ C 1 ([0, m], C 2 (R)).
1
Example 4.1.6 Let p > 1 and a ∈ C be chosen so that a p−1 = − . Consider the
p−1
Cauchy problem
ut − uxx = u p in (0, ∞) × R
u(0, x) = a in R.
1
Then u(t, x) = a(t + 1)− p−1 is its solution. Actually,
a p
ut (t, x) = − (t + 1)− p−1 ,
p−1
and
uxx (t, x) = 0,
and
a p
(u(t, x)) p = − (t + 1)− p−1 .
p−1
Therefore
ut (t, x) − uxx (t, x) = (u(t, x)) p in (0, ∞) × R
and
u(0, x) = a in R.
Proof 4.1.7 Consider
ut − uxx = f (t, x, u(t, x), ux (t, x)) in (0, m] × [1, 2], (4.10)
u(0, x) = φ (x) in [1, 2]. (4.11)

Let E12 = C ([0, m], C ([1, 2])) be endowed with the norm
1 2
n
(t,x)∈[0,m]×[1,2] (t,x)∈[0,m]×[1,2]
o
(t,x)∈[0,m]×[1,2] (t,x)∈[0,m]×[1,2]
By Ke12 we denote the set of all equi-continuous families in E12 .

0
Let K12 =K e12 ,
0
K12 = {u ∈ K12 : ||u|| ≤ B}.
Since φ ∈ C ([1, 2]), f ∈ C ([0, m] × [1, 2] × [−B, B] × [−B, B]) we have that there exists a
constant M12 > 0 such that
|φ (x)| ≤ M12 in [1, 2],
| f (t, x, y, z)| ≤ M12 in [0, m] × [1, 2] × [−B, B] × [−B, B].
Let l1 > 0 be chosen so that

l1 (5B + 2M12 ) ≤ B,
l1 B + l1 (B + M12 ) + 3l1 Bm + l1 M12 m ≤ B.

Let also
0
L12 = {u ∈ K12 : ||u|| ≤ (1 + l1 )B}.
T12 (u)(t, x) = (1 + l1 )u(t, x),

Z xZ y
S12 (u)(t, x) = −l1 u(t, x) + l1 (u(t, z) − φ (z))dzdy
1 1
Z t
−l1 (u(τ, x) − u11 (τ, 1) − (x − 1)u11x (τ, 1))dτ
0
Z tZ xZ y
−l1 f (τ, z, u(τ, z), ux (τ, z))dzdydτ, (t, x) ∈ [0, m] × [1, 2].
0 1 1
As in the previous theorem, one can prove that there is u12 ∈ C 1 ([0, 1], C 2 ([1, 2])) which is
a solution to problem (4.10), (4.11). This solution u12 satisfies the integral equation
Z xZ y
(u12 (t, z) − φ (z))dzdy
1 1
Z t
− (u12 (τ, x) − u11 (τ, 1) − (x − 1)u11x (τ, 1))dτ
0
(4.12)
Z tZ xZ y
0 1 1
= 0, (t, x) ∈ [0, m] × [1, 2].
Now we put x = 1 in (4.12) and we find

Z t
(u12 (τ, 1) − u11 (τ, 1))dτ = 0,
0
which we differentiate with respect to t and we get
u12 (t, 1) = u11 (t, 1) in [0, m]. (4.13)
Now we differentiate (4.12) with respect to x and we find

Z x Z t
(u12 (t, z) − φ (z))dz − (u12x (τ, x) − u11x (τ, 1))dτ
1 0
Z tZ x
− f (τ, z, u12 (τ, z), u12x (τ, z))dzdτ = 0, (t, x) ∈ [0, m] × [1, 2].
0 1
In the last equation we put x = 1 and we obtain

Z t
(u12x (τ, x) − u11x (τ, 1))dτ = 0, (t, x) ∈ [0, m] × [1, 2],
0
which we differentiate with respect to t and we find
u12x (t, 1) = u11x (t, 1) in [0, m]. (4.14)
Now we differentiate (4.13) with respect to t and we get
u12t (t, 1) = u11t (t, 1) in [0, m].
Hence, (4.13), (4.14) and
f (t, 1, u11 (t, 1), u11x (t, 1)) = f (t, 1, u12 (t, 1), u12x (t, 1)),
we find
u12xx (t, 1) = u12t (t, 1) − f (t, 1, u12 (t, 1), u12x (t, 1))
= u11t (t, 1) − f (t, 1, u11 (t, 1), u11x (t, 1))
= u11xx (t, 1) in [0, m].
Consequently, the function


 u11 (t, x) in [0, m] × [0, 1]
u(t, x) =
u12 (t, x) in [0, m] × [1, 2],

is a C 1 ([0, m], C 2 ([0, 2]))-solution to the problem
ut − uxx = f (t, x, u(t, x), ux (t, x)) in (0, m] × [0, 2],
u(0, x) = φ (x) in [0, 2].
Then we consider the problem
ut − uxx = f (t, x, u(t, x), ux (t, x)) in (0, m] × [2, 3]

(4.15)
u(0, x) = φ (x) in [2, 3].
As in the above there is u13 ∈ C 1 ([0, m], C 2 ([2, 3])) which is a solution to problem (4.15)
and satisfies the integral equation
Z xZ y
(u13 (t, z) − φ (z))dzdy
2 2
Z t
− (u13 (τ, x) − u12 (τ, 2) − (x − 2)u12x (τ, 2))dτ
0
Z tZ xZ y
0 2 2
= 0, t ∈ [0, m], x ∈ [2, 3].

The function 

 u11 (t, x) in [0, m] × [0, 1]



u(t, x) = u12 (t, x) in [0, m] × [1, 2]




u13 (t, x) in [0, m] × [2, 3]

is a C ([0, m], C ([0, 3]))-solution to the problem

1 2
ut − uxx = f (t, x, u(t, x), ux (t, x)) in [0, m] × [0, 3],

u(0, x) = φ (x) in [0, 3],
and so on. We construct a solution u1 ∈ C ([0, m], C (R)) which is a solution to the prob-
1 2
lem
ut − uxx = f (t, x, u(t, x), ux (t, x)) in (0, m] × R,
u(0, x) = φ (x) in R.
∂f ∂f
Theorem 4.1.8 Let f ∈ C ([0, ∞) × R × R × R), , exist and are continuous in
∂ u ∂ ux
[0, ∞) × R × R × R, φ ∈ C 2 (R). Let also u(t, x, φ ) ∈ C 1 ([0, m], C 2 ([c, d])) be a solution
to problem (4.1), (4.2) for some m ∈ (0, 1) and for some [c, d] ⊂ R. Then u(t, x, φ ) is dif-
∂u
ferentiable with respect to φ and v(t, x) = (t, x, φ ) satisfies the following initial value
∂φ
problem
∂f
vt − vxx = (t, x, u(t, x, φ ), ux (t, x, φ ))v
∂u
(4.16)
∂f
+ (t, x, u(t, x, φ ), ux (t, x, φ ))vx in [0, m] × [c, d],
∂ ux
v(0, x) = 1 in [c, d]. (4.17)

Proof 4.1.9 We have that the solution u(t, x, φ ) satisfies the following integral equation:
Z xZ y
Q(φ ) = (u(t, z, φ (z)) − φ (z))dzdy
c c
Z t
− (u(τ, x, φ (x)) − u(τ, c, φ (c)) − (x − c)ux (τ, c, φ (c)))dτ
0
Z tZ xZ y
− f (τ, z, u(τ, z, φ (z)), ux (τ, z, φ (z)))dz
0 c c
= 0, t ∈ [0, m], x ∈ [c, d].
Then
Z xZ y
Q(φ ) − Q(φ1 ) = (u(t, z, φ (z)) − u(t, z, φ1 (z)) − (φ (z) − φ1 (z)))dzdy
c c
Z t
− (u(τ, x, φ (x)) − u(τ, x, φ1 (x)))dτ
0
Z t
+ (u(τ, c, φ (c)) − u(τ, c, φ1 (c)))dτ
0
Z t
+ (x − c)(ux (τ, c, φ (c)) − ux (τ, c, φ1 (c)))dτ
0
Z t Z x Z y
− f (τ, z, u(τ, z, φ (z)), ux (τ, z, φ (z)))
0 c c

− f (τ, z, u(τ, z, φ1 (z)), ux (τ, z, φ1 (z))) dzdydτ
Z xZ y
∂u
= (t, z, φ (z)) − 1 dzdy
c c ∂φ
Z t Z t Z t
∂u ∂u ∂u
− (τ, x, φ (x))dτ + (τ, c, φ (c))dτ + (x − c) (τ, c, φ (c))dτ
0 ∂φ 0 ∂φ 0 ∂φ x
Z tZ xZ y
∂f ∂u
− (τ, z, u(τ, z, φ (z)), ux (τ, z, φ (z)))
(τ, z, φ (z))dzdydτ
0 c c ∂u ∂φ
Z tZ xZ y
∂f ∂u
− (τ, z, u(τ, z, φ (z)), ux (τ, z, φ (z))) (τ, z, φ (z))dzdydτ
0 c c ∂ ux ∂φ x
+δ {φ , φ1 },
where δ {φ , φ1 } −→ 0 as φ (x) −→ φ1 (x) for every x ∈ [c, d]. Hence, when φ (x) −→ φ1 (x)
for every x ∈ [c, d], we get
Z xZ y Z t
0 = (v(t, z) − 1)dzdy − v(τ, x)dτ
c c 0
Z t Z t
+ v(τ, c)dτ + (x − c)vx (τ, c)dτ
0 0
Z tZ xZ y (4.18)
∂f
− (τ, z, u(τ, z, φ (z)), ux (τ, z, φ (z)))v(τ, z)dzdydτ
0 c c ∂u
Z tZ xZ y
∂f
− (τ, z, u(τ, z, φ (z)), ux (τ, z, φ (z)))vx (τ, z)dzdydτ,
0 c c ∂ ux
which we differentiate twice in x and once in t and we get that v satisfies (4.16). Now we
put t = 0 in (4.18) and then we differentiate twice in x, and we find that v satisfies (4.17).
4.2 Local Existence of Classical Solutions for an IBVP

In this section we investigate the following IBVP
ut − 4u = f (t, x, u, Du), t ≥ 0, x1 ≥ 0, (x2 , . . . , xn ) ∈ Rn−1 ,

u(0, x) = u0 (x), x1 ≥ 0, (x2 , . . . , xn ) ∈ Rn−1 , (4.19)
n−1
u(t, 0, x2 , . . . , xn ) = v(t, x2 , . . . , xn ), t ≥ 0, (x2 , . . . , xn ) ∈ R ,
n
where n ≥ 2, 4u = ∑ uxi xi , Du = (ux1 , ux2 , . . . , uxn ), u : Rn+1 → R is an unknown function,
i=1
f : R × Rn × R × Rn → R, u0 : Rn → R and v : Rn → R are given functions which satisfy
the following conditions
(H1)
f ∈ C (R, Rn , R, Rn ),
m2
0 1
| f (t, x, u, Du)| ≤ ∑ a0i (t, x)|u(t, x)| pi + a1i (t, x)|ux1 (t, x)| pi
i=1
n

+ · · · + ani (t, x)|uxn (t, x)| pi ,
Local Existence of Classical Solutions for an IBVP 183
where aij : Rn+1 → R, i = 1, . . . , m2 , j = 0, 1, . . . , n, are positive continuous functions on

Rn+1 for which sup aij (t, x) < ∞, pij ∈ N ∪ {0}, m2 ∈ N.
R×Rn
(H2)
u0 ∈ C 2 (Rn ),
||u0 ||C 2 (Rn ) ≤ P for ∀ x ∈ Rn ,
v ∈ C 1 (R1 , C 2 (Rn−1 )),
||v||C1 (R1 ,C 2 (Rn−1 )) ≤ P for ∀ (t, x2 , . . . , xn ) ∈ Rn ,
v(0, x2 , . . . , xn ) = u0 (0, x2 , . . . , xn ) for ∀ (x2 , . . . , xn ) ∈ Rn−1 .

Here P is a fixed positive constant.
For O1 ⊂ [0, ∞) and O2 ⊂ Rn with C 1 (O1 , C 2 (O2 )) we denote the space of all continuous
functions u : O1 × O2 → R such that ut , uxi , uxi xi , i ∈ {1, . . . , n}, exist and are continuous
on O1 × O2 .
As an example for the function f that satisfies (H1) we consider the function f =
λ |u| p−1 u, p ∈ N, λ ∈ R. Let ε ∈ (0, 1) be fixed and G = max sup aij (t, x).
i∈{1,...,m2 }, j∈{0,1,...,n} Rn+1
We choose the positive constants m, Ai , i = 1, 2, . . . , n, in the following way
A21

2 2
εP + 2 (A1 A2 . . . An ) P + 2P (A2 . . . An ) 1 + A1 + m
2
n
A2i

2
+2 ∑ (A1 . . . Ai−1 Ai+1 . . . An ) 1 + Ai + Pm (4.20)
i=2 2
m2
0 1 n
+ (A1 . . . An )2 Gm ∑ P pi + P pi + · · · + P pi ≤ P,
i=1
A2

εP + (A1 A2 . . . An )2 P + 2P(A2 . . . An )2 1 + A1 + 1
2
n
A2

+2 ∑ (A1 . . . Ai−1 Ai+1 . . . An )2 1 + Ai + i P (4.21)
i=2 2
m2
0 1 n
+ (A1 . . . An )2 G ∑ P pi + P pi + · · · + P pi ≤ P,
i=1
A21

2 2
εP + 2A1 (A2 . . . An ) P + 2 (A2 . . . An ) 1 + A1 + mP
2
n
A2i

2
+2 ∑ A1 (A2 . . . Ai−1 Ai+1 . . . An ) 1 + Ai + mP (4.22)
i=2 2
m2
0 1 n
+A1 (A2 . . . An )2 Gm ∑ P pi + P pi + · · · + P pi ≤ P,
i=1
εP + 2Ai (A1 . . . Ai−1 Ai+1 . . . An )2 P
A2

+2 (A2 . . . Ai−1 Ai+1 . . . An )2 1 + A1 + 1 mPAi
2
!
n 2 A2j
+2 ∑ Ai A1 . . . Ai−1 Ai+1 . . . A j−1 A j+1 . . . An 1+Aj + mP
j=2, j6=i 2
(4.23)
A2i

2
+2(A1 . . . Ai−1 Ai+1 . . . An ) 1 + Ai + mP
2
m2
0 1 n
+Ai (A1 . . . Ai−1 Ai+1 . . . An ) mG ∑ P pi + P pi + · · · + P pi ≤ P,
i=1
i = 2, . . . , n,
A21

2 2
εP + 2(A2 . . . An ) P + (A2 . . . An ) 1 + A1 + mP
2
n
A2i

2
+2m ∑ (A2 . . . Ai−1 Ai+1 . . . An ) 1 + Ai + P (4.24)
i=2 2
m2
0 n
+m(A2 . . . An )2 G ∑ P pi + · · · + P pi ≤ P,
i=1
εP + 2(A1 . . . A j−1 A j+1 . . . An )2 P
A21

2
+2(A2 . . . A j−1 A j+1 . . . An ) 1 + A1 + mP
2
n
A2i

2
+2 ∑ (A1 . . . Ai−1 Ai+1 . . . A j−1 A j+1 . . . An ) 1 + Ai + mP
i=2,i6= j 2 (4.25)
!
A2j
+2(A1 . . . A j−1 A j+1 . . . An )2 1 + A j + mP
2
m2
0 n
+(A1 . . . A j−1 A j+1 . . . An )2 mG ∑ P pi + · · · + P pi ≤ P.
i=1
√
n
Example 4.2.1 The constants m = 1 − ε,
√
2n
1−ε
Ai = 0 ,
n(1 + P)(1 + G) 1 + ∑m pi pni 1 + ∑ni=1 Ai + 12 ∑ni=1 A2i
i=1 P + · · · + P
2
i ∈ {1, . . . , n}, satisfy conditions (4.20)-(4.25).

For fixed positive constants m, A1 , A2 , . . . , An that satisfy (4.20)-(4.25), we denote the set
B1 = {(x1 , x2 , . . . , xn ) ∈ Rn : 0 ≤ xi ≤ Ai , i = 1, . . . , n} .
We will prove that the IBVP
ut − 4u = f (t, x, u, Du), t ∈ [0, m], x ∈ B1 ,
u(0, x) = u0 (x), x ∈ B1 ,
(4.26)
u(t, 0, x2 , . . . , xn ) = v(t, x2 , . . . , xn ), t ∈ [0, m], 0 ≤ xi ≤ Ai ,
i = 2, . . . , n,
has a solution u ∈ C 1 ([0, m], C 2 (B1 )).

The main key of our proof is the following lemma.
Lemma 4.2.2 Let u ∈ C 1 ([0, m], C 2 (B1 )) satisfies the integral equation
Z xZ s Z xZ s
0 = u(t, σ )dσ ds − u0 (σ )dσ ds
0 0 0 0
Z t Z x 1 Z s1
− (u(τ, σ̃1 ) − v(τ, σ 1 )) dσ 1 ds1 dτ
0 0 0
(4.27)
n Z t Z x i Z si
−∑ u(τ, σ̃i )dσ i dsi dτ
i=2 0 0 0
Z tZ xZ s
− f (τ, σ , u, Du)dσ dsdτ.
0 0 0
Then u satisfies IBVP (4.26).
Here
Z x Z x1 Z xn
= ··· ,
c c1 cn
Z xi Z x1 Z xi−1 Z xi+1 Z xn
= ··· ··· ,
c c1 ci−1 ci+1 cn
c = (c1 , . . . , cn ),
si = (s1 , . . . , si−1 , si+1 , . . . , sn ),
σ̃i = (σ1 , . . . , σi−1 , xi , σi+1 , . . . , σn ),
dσ = dσn dσn−1 . . . dσ1 ,
dσ i = dσn . . . dσi+1 dσi−1 . . . dσ1 ,
i = 1, . . . , n.
Proof 4.2.3 For t ∈ [0, m], x ∈ B1 , after we differentiate once in t, then twice in x1 , twice in
x2 , and etc., twice in xn , equation (4.27), and we obtain
ut − 4u = f (t, x, u, Du).
Now we put t = 0 in equation (4.27) and we obtain
Z xZ s
(u(0, σ ) − u0 (σ )) dσ ds = 0,
0 0
which we differentiate twice in x1 , twice in x2 , and etc., twice in xn , and we get
u(0, x) = u0 (x).
We put x1 = 0 in equation (4.27) and we obtain

Z t Z x2 Z xn
··· (u(τ, 0, s2 , . . . , sn ) − v(τ, s2 , . . . , sn )) dsn . . . ds2 dτ = 0.
0 0 0
The last equality we differentiate once in t, twice in x2 , and etc., twice in xn , and we find
u(t, 0, x2 , . . . , xn ) = v(t, x2 , . . . , xn ).
The above lemma motivates us to define the operator
Z xZ s Z xZ s
L11 (u)(t, x) = u(t, x) + u(t, σ )dσ ds − u0 (σ )dσ ds
0 0 0 0
Z t Z x1 Z s1
− (u(τ, σ̃1 ) − v(τ, σ 1 )) dσ 1 ds1 dτ
0 0 0
n Z t Z x i Z si
i=2 0 0 0
Z tZ xZ s
− f (τ, σ , u, Du)dσ dsdτ,
0 0 0
(t, x) ∈ [0, m] × B1 , u ∈ C 1 ([0, m], C 2 (B1 )).
Note that every fixed point of the operator L11 satisfies equation (4.27) and from here it
follows that every fixed point of operator L11 is a solution to IBVP (4.26).
Theorem 4.2.4 Let n ≥ 2 be fixed, f satisfy (H1), u0 and v satisfy (H2). Then there exist
positive constants m, A1 , . . ., An so that there exists a solution u ∈ C 1 ([0, m], C 2 ([0, A1 ] ×
· · · × [0, An ])) to problem (4.19).
Proof 4.2.5 We represent the operator L11 in the following way
L11 (u)(t, x) = M11 (u)(t, x) + N11 (u)(t, x),
where
M11 (u)(t, x) = (1 + ε)u(t, x),

Z xZ s Z xZ s
N11 (u)(t, x) = −εu(t, x) + u(t, σ )dσ ds − u0 (σ )dσ ds
0 0 0 0
Z t Z x 1 Z s1
− (u(τ, σ̃1 ) − v(τ, σ 1 )) dσ 1 ds1 dτ
0 0 0
n Z t Z x i Z si
i=2 0 0 0
Z tZ xZ s
0 0 0
(t, x) ∈ [0, m] × B1 , u ∈ C 1 ([0, m], C 2 (B1 )).
We define the sets

n
K11 = u ∈ C 1 ([0, m], C 2 (B1 )) : max |u(t, x)| ≤ P,
t∈[0,m],x∈B1
max |uxi (t, x)| ≤ P, i = 0, 1, . . . , n,

t∈[0,m],x∈B1
o
max |uxi xi (t, x)| ≤ P, i = 1, . . . , n ,
t∈[0,m],x∈B1
n
K
e11 = u ∈ C 1 ([0, m], C 2 (B1 )) : max |u(t, x)| ≤ (1 + ε)P,
t∈[0,m],x∈B1
max |uxi (t, x)| ≤ (1 + ε)P, i = 0, 1, . . . , n,

t∈[0,m],x∈B1
o
max |uxi xi (t, x)| ≤ (1 + ε)P, i = 1, . . . , n ,
t∈[0,m],x∈B1
where ux0 = ut .
In these sets we define a norm as follows
n
kuk = sup max |u(t, x)|, max |ut (t, x)|,
t∈[0,m],x∈B1 t∈[0,m],x∈B1
max |uxi (t, x)|, max |uxi xi (t, x)|,

t∈[0,m],x∈B1 t∈[0,m],x∈B1
o
i ∈ {1, . . . , n} .
α
Here Dα u = ∂t 0 u, ∂xα11 u, . . . , ∂xαnn u . The sets K11 and K
e11 are completely normed spaces
1 1
with respect to this norm. Let K11 and K e11 denote the sets of all equi-continuous fami-
lies in K11 and K11 , respectively. Let also D11 = K11 1 and D̃ = K e1 , where K 1 and Ke1
e 11 11 11 11
1 1
are the closures of K11 and K
e11 , respectively. Note that D11 and D̃11 are compact sets in
C ([0, m], C (B1 )).
1 2
Lemma 4.2.6 The operator M11 : D11 → D̃11 is an expansive operator and onto.
Proof 4.2.7 Let u ∈ D11 . Then u ∈ C 1 ([0, m], C 2 (B1 )). From here, (1 + ε)u ∈
C 1 ([0, m], C 2 (B1 )) and M11 (u) ∈ C 1 ([0, m], C 2 (B1 )). Also,
max |u(t, x)| ≤ P,

t∈[0,m],x∈B1
max |uxi (t, x)| ≤ P, i = 0, 1, . . . , n,

t∈[0,m],x∈B1
max |uxi xi (t, x)| ≤ P, i = 1, . . . , n.

t∈[0,m],x∈B1
Therefore
max |M11 (u)(t, x)| ≤ (1 + ε)P,
t∈[0,m],x∈B1
max |M11 (u)xi (t, x)| ≤ (1 + ε)P, i = 0, 1, . . . , n,

t∈[0,m],x∈B1
max |M11 (u)xi xi (t, x)| ≤ (1 + ε)P, i = 1, . . . , n.

t∈[0,m],x∈B1
Consequently
M11 : D11 → D̃11 .
For u, v ∈ D11 we have that
kM11 (u) − M11 (v)k = (1 + ε)ku − vk,
from which it follows that the operator M11 is an expansive operator with constant 1 + ε.
v
Also, for v ∈ D̃11 , we take u = . Then |u(t, x)| ≤ P,|uxi (t, x)| ≤ P, |ux j x j (t, x)| ≤ P,
1+ε
i = 0, 1, . . . , n, j = 1, . . . , n, for every t ∈ [0, m] and x ∈ B1 , i.e., u ∈ D11 and M11 (u) = v.
Consequently M11 : D11 → D̃11 is onto.
Lemma 4.2.8 The operator

N11 : D11 → D11
is continuous.
Proof 4.2.9 First, we will prove that
N11 : D11 → D11
Let u ∈ D11 .
1.
Z xZ s Z xZ s
|N11 (u)(t, x)| ≤ ε|u(t, x)| + |u(t, σ )|dσ ds + |u0 (σ )|dσ ds
0 0 0 0
Z t Z x 1 Z s1
+ (|u(τ, σ̃1 )| + |v(τ, σ 1 )|) dσ 1 ds1 dτ
0 0 0
n Z t Z xi Z si
+∑ |u(τ, σ̃i )|dσ i dsi dτ
i=2 0 0 0
Z t Z x Z s m2
0 1
+ ∑ a0i (τ, σ )|u(τ, σ )| pi + a1i (τ, σ )|ux1 (τ, σ )| pi
0 0 0 i=1
n

+ · · · + ani (τ, σ )|uxn (τ, σ )| pi dσ dsdτ
≤ εP + 2(A1 . . . An )2 P + 2P(A2 . . . An )2 m
n
+ ∑ (A1 . . . Ai−1 Ai+1 . . . An )2 Pm
i=2
m2
0 1 n
+(A1 . . . An )2 Gm ∑ P pi + P pi + · · · + P pi
i=1
≤ P, (t, x) ∈ [0, m] × B1 .
In the last inequality we used the conditions in (H1) for the function f and
inequality (4.20).
2.
Z xZ s
|(N11 (u))t (t, x)| ≤ ε|ut (t, x)| + |ut (t, σ )|dσ ds
0 0
Z x 1 Z s1
+ (|u(t, σ̃1 )| + |v(t, σ 1 )|) dσ 1 ds1
0 0
n Z x i Z si
+∑ |u(t, σ̃i )|dσ i dsi
i=2 0 0
Z x Z s m2
0 1
+ ∑ a0i (t, σ )|u(t, σ | pi + a1i (t, σ )|ux1 (t, σ )| pi
0 0 i=1
n

+ · · · + ani (t, σ )|uxn (t, σ )| pi dσ ds
≤ εP + (A1 . . . An )2 P + 2(A2 . . . An )2 P
n
+ ∑ (A1 . . . Ai−1 Ai+1 . . . An )2 P
i=2
m2
0 1 n
+(A1 . . . An )2 G ∑ P pi + P pi + · · · + P pi
i=1
≤ P, (t, x) ∈ [0, m] × B1 .
In the last inequality we used inequality (4.21) and the conditions in (H1) for
the function f .
3.
|N11 (u)x1 (t, x)| ≤ ε|ux1 (t, x)|

Z x2 Z xn Z x1 Z s2 Z sn
+ ··· ··· |u(t, σ )|dσ ds1
0 0 0 0 0
Z x2 Z xn Z x1 Z s2 Z sn
+ ··· ··· |u0 (σ )|dσ ds1
0 0 0 0 0
Z t Z x 1 Z s1
+ |ux1 (τ, σ̃1 )|dσ 1 ds1 dτ
0 0 0
n Z t Z x2 Z xi−1 Z xi+1 Z xn Z x1 Z s2
+∑ ··· ···
i=2 0 0 0 0 0 0 0
Z si−1 Z si+1 Z sn
··· ··· |u(τ, σ̃i )|dσ i dsn · · · dsi+1 dsi−1 · · · ds2 dτ
0 0 0
Z t Z x2 Z xn Z x1 Z s2 Z sn m2
0
+ ··· ··· ∑ a0i (τ, σ )|u(τ, σ )| pi
0 0 0 0 0 0 i=1
1 n

+a1i (τ, σ )|ux1 (τ, σ )| pi + · · · + ani (τ, σ )|uxn (τ, σ )| pi dσ ds1 dτ
≤ εP + 2A1 (A2 . . . An )2 P + (A2 . . . An )2 Pm
n
+ ∑ A1 (A2 . . . Ai−1 Ai+1 . . . An )2 Pm
i=2
m2
0 1 n
+A1 (A2 . . . An )2 Gm ∑ P pi + P pi + · · · + P pi
i=1
≤ P, (t, x) ∈ [0, m] × B1 .
In the last inequality we used the conditions in (H1) for the function f and
inequality (4.22).
4. For j = 2, . . . , n, we have
Z x j Z s1 Z s j−1 Z x j Z s j+1 Z sn
|N11 (u)x j | ≤ ε|ux j (t, x)| + ... ... |u(t, σ )|dσ ds j
0 0 0 0 0 0
Z x j Z s1 Z s j−1 Z x j Z s j+1 Z sn
+ ... ... |u0 (σ )|dσ ds j
0 0 0 0 0 0
Z t Z x2 Z x j−1 Z x j+1 Z xn
+ ... ...
0 0 0 0 0
Z s2 Z s j−1 Z x j Z s j+1 Z sn
... ...
0 0 0 0 0

|u(τ, x1 , σ2 , . . . , σ j−1 , σ j , σ j+1 , . . . , σn )|

+|v(τ, σ2 , . . . , σ j−1 , σ j , σ j+1 , . . . , σn )|
dσ 1 dsn . . . ds j+1 ds j−1 . . . ds2 dτ
n Z t Z x1 Z xi−1 Z xi+1 Z x j−1 Z x j+1 Z xn

+ ∑ ... ... ...
i=2,i6= j 0 0 0 0 0 0 0
Z s1 Z si−1 Z si+1 Z s j−1 Z x j Z s j+1 Z sn
... ... ...
0 0 0 0 0 0 0
|u(τ, σ̃i )|dσ i dsn . . . ds j+1 ds j−1 . . . dsi+1 dsi−1 . . . ds1 dτ
Z t Z xj Z sj
+ |ux j (τ, σ̃ j )|dσ j ds j dτ
0 0 0
Z t Z x j Z s1 Z s j−1 Z x j Z s j+1 Z sn
+ ... ... | f (τ, σ , u, Du)|dσ ds j dτ
0 0 0 0 0 0 0
≤ εP + 2A j (A1 . . . A j−1 A j+1 . . . An )2 P
+2(A2 . . . A j−1 A j+1 . . . An )2 mP
n
+ ∑ (A1 . . . Ai−1 Ai+1 . . . A j−1 A j+1 . . . An )2 A j mP
i=2,i6= j
+(A1 . . . A j−1 A j+1 . . . An )2 mP
m2
0 n
+(A1 . . . A j−1 A j+1 . . . An )2 A j mG ∑ P pi + · · · + P pi
i=1
≤ P, (t, x) ∈ [0, m] × B1 .
the function f .
5.
Z x 1 Z s1
|N11 (u)x1 x1 (t, x)| ≤ ε|ux1 x1 (t, x)| + |u(t, σ̃1 )|dσ 1 ds1
0 0
Z x1 Z s1 Z t Z x 1 Z s1
+ |u0 (σ̃1 )|dσ 1 ds1 + |ux1 x1 (τ, σ̃1 )dσ 1 ds1 dτ
0 0 0 0 0
n Z t Z x2 Z xi−1 Z xi+1 Z xn
+∑ ... ...
i=2 0 0 0 0 0
Z s2 Z si−1 Z si+1 Z sn
... ...
0 0 0 0
|u(τ, x1 , σ2 , . . . , σi−1 , xi , σi+1 , . . . , σn )|dσn . . . dσi+1 dσi−1 . . . dσ2
dsn . . . dsi+1 dsi−1 . . . ds2 dτ
Z t Z x 1 Z s1
+ | f (τ, σ̃1 , u, Du)|dσ 1 ds1 dτ
0 0 0
≤ εP + 2(A2 . . . An )2 P + (A2 . . . An )2 mP
n
+m ∑ (A2 . . . Ai−1 Ai+1 . . . An )2 P
i=2
m2
0 n
+m(A2 . . . An )2 G ∑ P pi + · · · + P pi
i=1
≤ P, (t, x) ∈ [0, m] × B1 .
the function f .
6. For j = 2, . . . , n, we have
Z xj Z sj
|N11 (u)x j x j | ≤ ε|ux j x j (t, x)| + |u(t, σ̃ j )|dσ j ds j
0 0
Z xj Z sj
+ |u0 (σ̃ j )|dσ j ds j
0 0
Z t Z x2 Z x j−1 Z x j+1 Z xn
+ ... ...
0 0 0 0 0
Z s2 Z s j−1 Z s j+1 Z sn
... ...
0 0 0 0

|u(τ, x1 , σ2 , . . . , σ j1 , x j , σ j+1 , . . . , σn )|

+|v(τ, σ2 , . . . , σ j−1 , x j , σ j+1 , . . . , σn )|
dσn . . . dσ j+1 dσ j−1 . . . dσ2 dsn . . . ds j+1 ds j−1 . . . , ds2 dτ
n Z t Z x1 Z xi−1 Z xi+1 Z x j−1 Z x j+1 Z xn

+ ∑ ... ... ...
i=2,i6= j 0 0 0 0 0 0 0
Z s1 Z si−1 Z si+1 Z s j−1 Z s j+1 Z sn
... ... ...
0 0 0 0 0 0
|u(τ, σ1 , . . . , σi−1 , xi , σi+1 , . . . , σ j−1 , x j , σ j+1 , . . . , σn )|
dσn . . . dσ j+1 dσ j−1 . . . dσi+1 dσi−1 . . . dσ1
dsn . . . ds j+1 ds j−1 . . . dsi+1 dsi−1 . . . ds1 dτ
Z t Z xj Z sj
+ |ux j x j (τ, σ̃ j )dσ j ds j dτ
0 0 0
Z t Z xj Z sj
+ | f (τ, σ̃ j , u, Du)|dσ j ds j dτ
0 0 0
≤ εP + 2(A1 . . . A j−1 A j+1 . . . An )2 P
+2(A2 . . . A j−1 A j+1 . . . An )2 mP
n
+ ∑ (A1 . . . Ai−1 Ai+1 . . . A j−1 A j+1 . . . An )2 mP
i=2,i6= j
+(A1 . . . A j−1 A j+1 . . . An )2 mP
m2
0 n
+(A1 . . . A j−1 A j+1 . . . An )2 mG ∑ P pi + · · · + P pi
i=1
≤ P, (t, x) ∈ [0, m] × B1 .
the function f .
Consequently
N11 : D11 → D11 .
From the above estimates it follows that if un → u, un ∈ D11 , u ∈ D11 , as n → ∞, in the
sense of the topology in D11 , since f is a continuous function of its arguments, we have
N11 (un ) → N11 (u), as n → ∞, in the sense of the topology of the space D11 . Therefore
N11 : D11 → D11 is a continuous operator.
From Lemma 4.2.2, Theorem 1.9.12, Lemma 4.2.6 and Lemma 4.2.8 it follows that
IBVP (4.26) has a solution u ∈ C 1 ([0, m], C 2 (B1 )). This completes the proof.
Theorem 4.2.10 Let n ≥ 2 be fixed, f satisfy (H1), u0 and v satisfy (H2). Then there exists
a positive constant m such that there exists a solution u ∈ C 1 ([0, m], C 2 (Rn1+ )) to problem
(4.19).
Proof 4.2.11 Let u11 be the solution which is obtained in the previous theorem.
Now we define the set
B2 =!‘ {x ∈ Rn : A1 ≤ x1 ≤ 2A1 , 0 ≤ xi ≤ Ai , i = 2, . . . , n} ,
the operators
Z xZ s Z xZ s
L12 (u)(t, x) = u(t, x) + u(t, σ )dσ ds − u0 (σ )dσ ds
A A A A
Z t Z x 1 Z s1
u(τ, σ̃1 ) − ũ11 (τ, σ̃1 ) dσ 1 ds1 dτ

−
0 A A
n Z t Z x i Z si
i=2 0 A A
Z tZ xZ s
0 A A
M12 (u)(t, x) = (1 + ε)u(t, x),

Z xZ s Z xZ s
N12 (u)(t, x) = −εu(t, x) + u(t, σ )dσ ds − u0 (σ )dσ ds
A A A A
Z t Z x 1 Z s1
u(τ, σ̃1 ) − ũ11 (τ, σ̃1 ) dσ 1 ds1 dτ

−
0 A A
n Z t Z x i Z si
i=2 0 A A
Z tZ xZ s
0 A A
L12 (u)(t, x) = M12 (u)(t, x) + N12 (u)(t, x),
(t, x) ∈ [0, m] × B2 , u ∈ C 1 ([0, m], C 2 (B2 )),
A = (A1 , 0, . . . , 0),
ũ11 (t, x) = u11 (t, A1 , x2 , . . . , xn ) + (x1 − A1 )u11

x1 (t, A1 , x2 , . . . , xn )
the sets
n
K12 = u ∈ C 1 ([0, m], C 2 (B2 )) : max |u(t, x)| ≤ P,
t∈[0,m],x∈B2
max |uxi (t, x)| ≤ P, i = 0, 1, . . . , n,

t∈[0,m],x∈B2
o
max |uxi xi (t, x)| ≤ P, i = 1, . . . , n ,
t∈[0,m],x∈B2
n
K
e12 = u ∈ C 1 ([0, m], C 2 (B2 )) : max |u(t, x)| ≤ (1 + ε)P,
t∈[0,m],x∈B2
max |uxi (t, x)| ≤ (1 + ε)P, i = 0, 1, . . . , n,

t∈[0,m],x∈B2
o
max |uxi xi (t, x)| ≤ (1 + ε)P, i = 1, . . . , n ,
t∈[0,m],x∈B2
which are completely normed spaces with the norm

n
kuk = sup max |u(t, x)|, max |ut (t, x)|,
t∈[0,m],x∈B2 t∈[0,m],x∈B2
max |uxi (t, x)|, max |uxi xi (t, x)|,

t∈[0,m],x∈B2 t∈[0,m],x∈B2
o
i ∈ {1, . . . , n} .
1 1
Let K12 and Ke12 denote the sets of all equi-continuous families in K12 and Ke12 , respec-
1 and D̃ = K e1 , where K 1 and Ke1 are the closures of K121
tively. Let also D12 = K12 12 12 12 12 and
Ke12 , respectively. Note that D12 and D̃12 are compact sets in C ([0, m], C (B2 )). As in the
1 1 2
previous section we prove that IBVP
ut − 4u = f (t, x, u, Du), t ∈ [0, m], x ∈ B2 ,
u(0, x) = u0 (x), x ∈ B2 ,
u(t, A1 , x2 , . . . , xn ) = u11 (t, A1 , x2 , . . . , xn ), t ∈ [0, m],
0 ≤ xi ≤ Ai , i = 2, . . . , n,
has a solution u12 ∈ C 1 ([0, m], C 2 (B2 )). For it we have

Z xZ s Z xZ s
u12 (t, σ )dσ ds − u0 (σ )dσ ds
A A A A
Z t Z x1 Z s1
u12 (τ, σ̃1 ) − ũ11 (τ, σ̃1 ) dσ 1 ds1 dτ

−
0 A A
(4.28)
n Z t Z x i Z si
12
−∑ u (τ, σ̃i dσ i dsi dτ
i=2 0 A A
Z tZ xZ s
− f (τ, σ , u12 , Du12 )dσ dsdτ = 0.
0 A A
We put x1 = A1 in the last equality and we obtain

Z t Z x1 Z s1
u12 (τ, A1 , σ2 , . . . , σn ) − u11 (τ, A1 , σ2 , . . . , σn ) dσ 1 ds1 dτ = 0.

0 A A
We differentiate the last equality once in t, twice in x2 , and etc., twice in xn , we obtain
u12 (t, A1 , x2 , . . . , xn ) = u11 (t, A1 , x2 , . . . , xn ), t ∈ [0, m], 0 ≤ xi ≤ Ai ,
i ∈ {2, . . . , n}, from here,
ut12 (t, A1 , x2 , . . . , xn ) = ut11 (t, A1 , x2 , . . . , xn ), t ∈ [0, m], 0 ≤ xi ≤ Ai , (4.29)
i ∈ {2, . . . , n}.
Now we differentiate equality (4.28) in x1 , then we put x1 = A1 , we get
Z t Z x1 Z s1
u12 11

x1 (τ, σ2 , . . . , σn ) − ux1 (τ, A1 , σ2 , . . . , σn ) dσ 1 ds1 dτ = 0.
0 A A
The last equality we differentiate once in t, twice in x2 , and etc., twice in xn , we get
u12 11
x1 (t, A1 , x2 , . . . , xn ) = ux1 (t, A1 , x2 , . . . , xn ), t ∈ [0, m], 0 ≤ xi ≤ Ai , (4.30)
i ∈ {2, . . . , n}.
Now we differentiate (4.28) twice in xi , i ∈ {2, . . . , n}, then we put x1 = A1 , we get
Z t Z x2 Z xi−1 Z xi+1 Z x n Z s2 Z si−1 Z si+1 Z sn
... ... ... ...
0 0 0 0 0 0 0 0 0

u12 (τ, A1 , σ2 , . . . , σi−1 , xi , σi+1 , . . . , σn )

−u11 (τ, A1 , σ2 , . . . , σi−1 , xi , σi+1 , . . . , σn )
dσn . . . dσi+1 dσi−1 . . . dσ2
dsn . . . dsi+1 dsi−1 . . . ds2 = 0,
which we differentiate once in t, twice in x2 , and etc., twice in xi−1 , twice in xi+1 , and etc.,
twice in xn , we get
u12 (t, A1 , x2 , . . . , xn ) = u11 (t, A1 , x2 , . . . , xn ), t ∈ [0, m],
0 ≤ xi ≤ Ai , i ∈ {2, . . . , n}. Hence,
u12 11
xi (t, A1 , x2 , . . . , xn ) = uxi (t, A1 , x2 , . . . , xn ), t ∈ [0, m], (4.31)
0 ≤ xi ≤ Ai , i ∈ {2, . . . , n}, and
u12 11
xi xi (t, A1 , x2 , . . . , xn ) = uxi xi (t, A1 , x2 , . . . , xn ), t ∈ [0, m], (4.32)
0 ≤ xi ≤ Ai , i ∈ {2, . . . , n}.
From (4.30) and (4.31) we get
f t, A1 , x2 , . . . , xn , u11 (t, A1 , x2 , . . . , xn ), Du11 (t, A1 , x2 , . . . , xn )

f t, A1 , x2 , . . . , xn , u12 (t, A1 , x2 , . . . , xn ), Du12 (t, A1 , x2 , . . . , xn ) ,

=
t ∈ [0, m], 0 ≤ xi ≤ Ai , i ∈ {2, . . . , n}. Hence, (4.29) and (4.32) we obtain
u12 11
x1 x1 (t, A1 , x2 , . . . , xn ) = ux1 x1 (t, A1 , x2 , . . . , xn ), t ∈ [0, m],
0 ≤ xi ≤ Ai , i ∈ {2, . . . , n}.
In this way we obtain that the function
(
u11 t ∈ [0, m], x ∈ B1 ,
u=
u12 t ∈ [0, m], x ∈ B2 ,
is a solution to the IBVP

ut − 4u = f (t, x, u, Du), t ∈ [0, m], x ∈ B1 ∪ B2
u(0, x) = u0 (x), x ∈ B1 ∪ B2 ,
u(t, 0, x2 , . . . , xn ) = v(t, x2 , . . . , xn ), t ∈ [0, m], 0 ≤ xi ≤ Ai , i = 2, . . . , n.
Repeat the above steps in x1 , x2 , and etc., and in xn we obtain that the IBVP
ut − 4u = f (t, x, u, Du), t ∈ [0, m], x ∈ Rn1+ ,
u(0, x) = u0 (x), x ∈ Rn1+ ,
u(t, 0, x2 , . . . , xn ) = v(t, x2 , . . . , xn ), t ∈ [0, m], (x2 , . . . , xn ) ∈ Rn−1 ,
has a solution u1 ∈ C 1 ([0, m], C 2 (Rn1+ )).
4.3 Periodic Solutions

Let ω, p, ε, s, A, B, m, k, M, N, r1 , ρ, r2 , R, n j , l j , j ∈ {1, . . . , n} be positive constants such that
1
r1 < ρ < r2 < R, s< , (4.33)
2
n
m k nj lj
(2 + A)R + M nωAR + ωAMR + M A + A + AR + ∑ A R + R < s, (4.34)
j=1
n
m k nj lj
M nωAri + ωAMri + M A + Ari + Ari + ∑ A ri + ri < ri , i = 1, 2, (4.35)
j=1
n
m k nj lj
ε R − M nωAR + ωAMR + M A + A + AR + ∑ A R + R > pρ,
j=1
p p 1 ε
ε< < < p< , ρ> R,
2 p+1 2 p−ε
NB > (1 + A)R + MAnωR.

(4.36)
For x = (x1 , . . . , xn ), y = (y1 , . . . , yn ), z = (z1 , . . . , zn ) ∈ Rn , introduce the following notations
Z xˆj Z x1 Z x j−1 Z x j+1 Z xn
= ... ... ,
−∞ −∞ −∞ −∞ −∞
Z yx˜ Z y1 Z y j−1 Z x j Z y j+1 Z yn
j
= ... ... ,
−∞ −∞ −∞ −∞ −∞ −∞
Z x Z x1 Z xn
= ... ,
−∞ −∞ −∞
Periodic Solutions 201
dx j = dxn . . . dx j+1 dx j−1 . . . dx1 ,
z˜x j = (z1 , . . . , z j−1 , x j , z j+1 , . . . , zn ),
dx = dxn . . . dx1 , j ∈ {1, . . . , n}.
Consider the equation
ut − ∆u = f (t, x, u, ux ), (t, x) ∈ R × Rn , (4.37)

n
where ∆u = ∑ ux j x j , ux = (ux1 , . . . , uxn ),
j=1
(H1)
b0 (t, x) ≤ f (t, x, u, v1 , . . . , vn )
≤ a0 (t, x) + a1 (t, x)|u|m + a2 (t, x)|u|k
n n
+ ∑ b j (t, x)|v j |n j + ∑ c j (t, x)|v j |l j , (t, x) ∈ [0, ω] × Rn ,
j=1 j=1
u ∈ R, (v1 , . . . , vn ) ∈ Rn ,
ai : R × Rn → [0, ∞), i ∈ {0, 1, 2},

b0 , b j , c j : R × Rn → [0, ∞), j ∈ {1, . . . , n},
b0 , ai , b j , c j ∈ C (R × R ), n
i ∈ {0, 1, 2}, j ∈ {1, . . . , n},
Z ωZ x Z y
ai (t, z)dzdydt ≤ A,
0 −∞ −∞
Z ω Z xˆl Z y˜x
l
ai (t, z)dzdyl dt ≤ A,
0 −∞ −∞
Z ω Z xˆl Z yˆl
ai (t, z˜xl )dzl dyl dt ≤ A, i ∈ {0, 1, 2}, l ∈ {1, . . . , n}, (t, x) ∈ R × Rn ,
0 −∞ −∞
Z ωZ x Z y
b j (t, z)dzdydt ≤ A,
0 −∞ −∞
Z ω Z xˆl Z y˜x
l
b j (t, z)dzdyl dt ≤ A,
0 −∞ −∞
Z ω Z xˆl Z yˆl
b0 (t, z˜xl )dzl dyl dt ≤ A, l ∈ {1, . . . , n}, (t, x) ∈ R × Rn ,
0 −∞ −∞
Z ωZ x Z y
c j (t, z)dzdydt ≤ A,
0 −∞ −∞
Z ω Z xˆl Z y˜xl
c j (t, z)dzdyl dt ≤ A,
0 −∞ −∞
Z ω Z xˆl Z yˆl
c j (t, z˜xl )dzl dyl dt ≤ A, j, l ∈ {1, . . . , n}, (t, x) ∈ [0, ω] × Rn .
0 −∞ −∞
Suppose that
(H2) a : R → [0, ∞) is a continuous ω-periodic function such that
0 ≤ a(t) ≤ M,
e−[a]ω Rtt+s a(τ)dτ
N≤ e ≤ M, t, s ∈ [0, ω].
1 − e−[a]ω
(H3) b : R × Rn → (0, ∞), b ∈ C (R × Rn ), b is ω-periodic with respect to t, b(t, x) 6= 0,
t ∈ R, x ∈ Rn , Z Z x y
b(t, z)dzdy ≤ A,
−∞ −∞
Z xˆl Z y˜xl
b(t, z)dzdyl dt ≤ A,
−∞ −∞
Z ωZ x Z y
b(t, z)b0 (t + s, z)dzdydt ≥ B,
0 −∞ −∞
Z xˆl Z yˆl
b(t, z˜xl )dzl dyl dt ≤ A, t ∈ [0, ω], x ∈ Rn , l ∈ {1, . . . , n}.
−∞ −∞
With C 1 ([0, ω], C 2 (Rn )) we will denote the space of all elements of the space
C 1 (R, C 2 (Rn )) which are ω-periodic with respect to the first variable.
Lemma 4.3.1 Let u ∈ C 1 ([0, ω], C 2 (Rn )) be a solution of the equation
e−[a]ω
Z ω R Z x Z y
t+s
a(τ)dτ
e t b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y
+ b(t, z)a(t + s)u(t + s, z)dzdy
−∞ −∞
Z x Z y
+ b(t, z) f (t + s, z, u(t + s, z), ux1 (t + s, z), . . . , uxn (t + s, z))dzdy ds
−∞ −∞
Z x Z y
− b(t, z)u(t, z)dzdy = 0, t ∈ [0, ω], x ∈ Rn ,
−∞ −∞
(4.38)
then it is an ω-periodic solution with respect to t of equation (4.37).
Proof 4.3.2 Differentiate twice in x1 , . . . , xn equation (4.38) and we get

e−[a]ω
Z ω R
t+s
a(τ)dτ
b(t, x)u(t, x) = e t b(t, x)∆u(t + s, x) + b(t, x)a(t + s)u(t + s, x)
1 − e−[a]ω 0

+b(t, x) f (t + s, x, u(t + s, x), ux (t + s, x)) ds, t ∈ [0, ω], x ∈ Rn ,
whereupon
e−[a]ω
Z ω R t+s

a(τ)dτ
u(t, x) = et ∆u(t + s, x) + a(t + s)u(t + s, x)
1 − e−[a]ω 0

+ f (t + s, x, u(t + s, x), ux (t + s, x)) ds, t ∈ [0, ω], x ∈ Rn .
Hence, and applying Lemma 2.1.1, we get

ut (t, x) = −a(t)u(t, x) + a(t)u(t, x) + ∆u(t, x) + f (t, x, u(t, x), ux (t, x))
= ∆u(t, x) + f (t, x, u(t, x), ux (t, x)), t ∈ [0, ω], x ∈ Rn .

Let E = C 1 ([0, ω], C 2 (Rn )) be endowed with the norm

kuk = max sup |u(t, x)|, sup |uxi (t, x)|,
(t,x)∈[0,ω]×Rn (t,x)∈[0,ω]×Rn

sup |uxi xi (t, x)|, i ∈ {1, . . . , n} .
(t,x)∈[0,ω]×Rn
Define the cone

P
f = u ∈ E : u(t, x) ≥ 0, (t, x) ∈ [0, ω] × R . n

f
For u ∈ E, define the operators
Tu(t, x) = (1 + ε)u(t, x) + εR,
Fu(t, x) = −εu(t, x) − εR
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
+ε et b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y
−∞ −∞
Z x Z y
−∞ −∞
Z x Z y
− b(t, z)u(t, z)dzdy , (t, x) ∈ [0, ω] × Rn .
−∞ −∞
Let u ∈ E be a fixed point of the operator T + F. Then

u(t, x) = (1 + ε)u(t, x) − εu(t, x)
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
+ε e t b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y
−∞ −∞
Z x Z y
−∞ −∞
Z x Z y
− b(t, z)u(t, z)dzdy , (t, x) ∈ [0, ω] × Rn ,
−∞ −∞
whereupon
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
et b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y
−∞ −∞
Z x Z y
−∞ −∞
Z x Z y
− b(t, z)u(t, z)dzdy = 0, (t, x) ∈ [0, ω] × Rn .
−∞ −∞
Therefore, u is an ω-periodic solution to the integral equation (4.38). Hence, and applying
Lemma 4.3.1, it follows that u is an ω-periodic solution of equation (4.37). Let Q > 0 be
an arbitrarily chosen constant. Define the conical shell
PQ = {u ∈ P : kuk < Q}.
On PQ define the operator
Gu(t, x) = −R − u(t, x)
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
+ et b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y
−∞ −∞
Z x Z y
−∞ −∞
Z x Z y
− b(t, z)u(t, z)dzdy, (t, x) ∈ [0, ω] × Rn .
−∞ −∞
Lemma 4.3.3 For u ∈ PQ we have
|Gu(t, x)| ≤ R + (1 + A)Q

n
+M nωAQ + ωAMQ + M A + AQm + AQk + ∑ A(Qn j + Ql j ) ,
j=1

∂
Gu(t, x) ≤ R + (1 + A)Q
∂ xi
n
+M nωAQ + ωAMQ + M A + AQm + AQk + ∑ A(Qn j + Ql j ) ,
j=1
2
∂
Gu(t, x) ≤ R + (1 + A)Q
∂ x2
i
n
m k nj lj
+M nωAQ + ωAMQ + M A + AQ + AQ + ∑ A(Q + Q ) ,
j=1
−Gu(t, x) ≥ R
n
−M nωAQ + ωAMQ + M A + AQm + AQk + ∑ A(Qn j + Ql j ) ,
j=1
(t, x) ∈ [0, ω] × Rn , i ∈ {1, . . . , n}.
Proof 4.3.4 We have
|Gu(t, x)| ≤ R + u(t, x)
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
+ e t b(t, z)|∆u(t + s, z)|dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y Z x Z y
+ b(t, z)a(t + s)u(t + s, z)dzdy + b(t, z)
−∞ −∞ −∞ −∞

× f (t + s, z, u(t + s, z), ux1 (t + s, z), . . . , uxn (t + s, z))dzdy ds
Z x Z y
+ b(t, z)u(t, z)dzdy
−∞ −∞
≤ u(t, x) + R
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
+ e t b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y Z x Z y
−∞ −∞ −∞ −∞

× a0 (t + s, z) + a1 (t + s, z)(u(t + s, z))m + a2 (t + s, z)(u(t + s, z))k
n n
+ ∑ b j (t + s, z)|ux j (t + s, z)|n j + ∑ c j (t + s, z)|ux j (t + s, z)|l j dzdy ds
j=1 j=1
Z x Z y
−∞ −∞
≤ R + (1 + A)Q
n
m k nj lj
j=1
∂
Gu(t, x) = −uxi (t, x)
∂ xi
e−[a]ω xî
Z ω R t+s
Z Z y˜x
a(τ)dτ i
+ et b(t, z)∆u(t + s, z)dzdyi
1 − e−[a]ω 0 −∞ −∞
Z xî Z y˜x Z xî Z y˜x
i i
+ b(t, z)a(t + s)u(t + s, z)dzdyi + b(t, z)
−∞ −∞ −∞ −∞

i
Z xî Z y˜x
i
− b(t, z)u(t, z)dzdyi ,
−∞ −∞

∂
∂ xi Gu(t, x) ≤ |uxi (t, x)|

e−[a]ω xî
Z ω R t+s
Z Z y˜x
a(τ)dτ i
+ e t b(t, z)|∆u(t + s, z)|dzdyi
1 − e−[a]ω 0 −∞ −∞
Z xî Z y˜x Z xî Z y˜x
i i
+ b(t, z)a(t + s)u(t + s, z)dzdyi + b(t, z)
−∞ −∞ −∞ −∞

× a0 (t + s, z) + a1 (t + s, z)(u(t + s, z))m + a2 (t + s, z)(u(t + s, z))k
n n
+ ∑ b j (t + s, z)|ux j (t + s, z)|n j + ∑ c j (t + s, z)|ux j (t + s, z)|l j dzdyi ds
j=1 j=1
Z xî Z y˜x
i
+ b(t, z)u(t, z)dzdyi
−∞ −∞
≤ (1 + A)Q
n
m k nj lj
j=1
∂2
Gu(t, x) = −uxi xi (t, x)
∂ xi2
e−[a]ω xî
Z ω R t+s
Z Z yî
a(τ)dτ
+ et b(t, z˜xi )∆u(t + s, z˜xi )dzi dyi
1 − e−[a]ω 0 −∞ −∞
Z xî Z yî Z xî Z yˆx
i
+ b(t, z˜xi )a(t + s)u(t + s, z˜xi )dzi dyi + b(t, z˜xi )
−∞ −∞ −∞ −∞

× f (t + s, z˜xi , u(t + s, z˜xi ), ux1 (t + s, z˜xi ), . . . , uxn (t + s, z˜xi ))dzi dyi ds
Z xî Z yî
− b(t, z˜xi )u(t, z˜xi )dzi dyi ,
−∞ −∞
2
∂
∂ x2 Gu(t, x) ≤ |uxi xi (t, x)|

i
e−[a]ω xî
Z ω R t+s
Z Z yî
a(τ)dτ
+ et b(t, z˜xi )|∆u(t + s, z˜xi )|dzi dyi
1 − e−[a]ω 0 −∞ −∞
Z xî Z yî Z xî Z yî
i i
+ b(t, z˜xi )a(t + s)u(t + s, z˜xi )dz dy + b(t, z˜xi )
−∞ −∞ −∞ −∞

× a0 (t + s, z) + a1 (t + s, z)|u(t + s, z˜xi )|m + a2 (t + s, z˜xi )|u(t + s, z˜xi )|k
n
+ ∑ b j (t + s, z˜xi )|ux j (t + s, z˜xi )|n j
j=1
n
i ilj
+ ∑ c j (t + s, z˜xi )|ux j (t + s, z˜xi )| dz dy ds
j=1
Z xî Z yî
+ b(t, z˜xi )u(t, z˜xi )dzi dyi
−∞ −∞
≤ (1 + A)Q
n
m k nj lj
j=1
−Gu(t, x) = R + u(t, x)
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
− et b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y Z x Z y
−∞ −∞ −∞ −∞

Z x Z y
−∞ −∞
≥ R + u(t, x)
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
− e t b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y Z x Z y
−∞ −∞ −∞ −∞

Z x Z y
− b(t, z)u(t, z)dzdy
−∞ −∞
n
≥ R − M nωAQ + ωAMQ + M A + AQm + AQk + ∑ A(Qn j + Ql j ) ,
j=1
(t, x) ∈ [0, ω] × Rn , i ∈ {1, . . . , n}.
Theorem 4.3.5 Suppose (4.33)-(4.36), (H1)-(H3). Then equation (4.37) has at least three
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .
PR = {x ∈ P : kxk < R},
Pρ = {x ∈ P : kxk < ρ},
Pri = {x ∈ P : kxk < ri }, i = 1, 2.
1. We have T : PR → E. Next, for u, v ∈ P, we get
kTu − T vk = (1 + ε)ku − vk,
i.e., T : PR → E is an expansive mapping with a constant 1 + ε.

2. By Lemma 4.3.3, we get, for u ∈ PR ,

kFuk ≤ ε (2 + A)R + M nωAR + ωAMR + M A + ARm + ARk
n
nj lj
+ ∑ A R +R
j=1
< sε
1
< ε.
2
As in the proof of Theorem 3.2.3. we have that F : PR → E is 0-set contraction.
3. For i = 1, 2, u ∈ ∂ Pri , using Lemma 4.3.3, we have

−[a]ω Z ω R
e t+s
e t a(τ)dτ

|Fu(t, x) + T 0(t, x)| = − εu(t, x) + ε

−[a]ω
1−e 0
Z x Z y
× b(t, z)∆u(t + s, z)dzdy
−∞ −∞
Z x Z y Z x Z y
−∞ −∞ −∞ −∞

Z x Z y

−∞ −∞
−[a]ω Z ω R
e t+s
= −εu(t, x) + ε −[a]ω
e t a(τ)dτ
1−e 0
Z x Z y
× b(t, z)∆u(t + s, z)dzdy
−∞ −∞
Z x Z y Z x Z y
−∞ −∞ −∞ −∞

Z x Z y
−∞ −∞
e−[a]ω x
Z ω R t+s
Z Z y
a(τ)dτ
≤ε et b(t, z)∆u(t + s, z)dzdy
1 − e−[a]ω 0 −∞ −∞
Z x Z y Z x Z y
−∞ −∞ −∞ −∞

Z x Z y
−∞ −∞
n
m k nj lj
≤ εM nωAri + ωAMri + M A + Ari + Ari + ∑ A ri + ri
j=1
< εri , i = 1, 2, t ∈ [0, ω], x ∈ Rn .

4. We will prove condition (b) of Theorem 1.9.21. Let u0 ∈ P ∗ and assume
that there exists a u1 ∈ ∂ Pρ and a λ1 > 0 with u1 − λ1 u0 ∈ Pρ such that
Fu1 = (I − T )(u1 − λ1 u0 ).
Hence,
0 < pρ
n
≤ ε R − M nωAR + ωAMR + M A + ARm + ARk + ∑ A Rn j + Rl j
j=1

≤ ε R + u1 (t, x) − M nωAρ + ωAMρ
n
m k nj lj
+M A + Aρ + Aρ + ∑ A ρ + ρ
j=1
≤ −εGu1 (t, x)
= −Fu1 (t, x)
= −(I − T )(u1 − λ1 u0 )(t, x)
= ε(u1 − λ1 u0 )(t, x) + εR
≤ ερ + εR, (t, x) ∈ [0, ω] × Rn ,
whereupon
ε
ρ< R.
p−ε
This is a contradiction. Therefore condition (b) of Theorem 1.9.21 holds.

F(PR ) ⊂ (I − T )(PR ). (4.39)
Let z ∈ PR be fixed and consider on PR the following equation
u(t, x) = Tu(t, x) + Fz(t, x)
= Su(t, x), t ∈ [0, ω], x ∈ Rn .

For any u, v ∈ PR , we have
kSu − Svk = kTu − T vk
= (1 + ε)ku − vk,
i.e., S : PR → P is an expansive mapping. Now we will prove that
PR ⊂ S(PR ). (4.40)
For any given v ∈ PR , we consider the equation on PR
Su(t, x) = v(t, x) or Tu(t, x) = v(t, x) − Fz(t, x), t ∈ [0, ω], x ∈ Rn .
(4.41)
Using Lemma 4.3.3, we have
v(t, x) − Fz(t, x) = v(t, x) − εGz(t, x)
≥ 0, (t, x) ∈ [0, ω] × Rn ,
and
kv − Fzk ≤ kvk + kFzk

≤ R + ε (2 + A)R + M nωAR + ωAMR
n
m k nj lj
+M A + AR + AR + ∑ A R + R
j=1
= R1 , t ∈ [0, ω], x ∈ Rn .
P. Note that
\
We have R1 > R and v − Fz ∈ B(R1 )
T : PR → T (PR ) = B(R1 ) P
\
Multiple Solutions for an IBVP with Robin Boundary Conditions 213
is a bijection. Therefore there is a u ∈ PR that satisfies (4.41) and then (4.40)

holds. Hence, and applying Theorem 1.9.8, it follows that there is a w ∈ PR
such that
Aw(t, x) = w(t, x), t ∈ [0, ω], x ∈ Rn .
Therefore (4.39) holds.
By Theorem 1.9.21, it follows that equation (4.37) has at least three solutions x1 , x2 , x3 ∈ P
such that
0 < kx1 k ≤ r1 ≤ kx2 k < ρ < kx3 k ≤ r2 .
4.4 Multiple Solutions for an IBVP with Robin Boundary Conditions

Let the positive constants ε, m, k, s, A, T, B0 , R, r1 , r2 , ρ and a ≥ 0 satisfy the following con-
ditions
1
0 < r1 < ρ < r2 < R, 0 < s < , (4.42)
2
R + 3A + aT R + ARm + ARk < s, (4.43)
3A + aTri + Arim + Arik < ri , B0 > R. (4.44)

Consider the IBVP
ut = ∆u + f (t, x, u), (t, x) ∈ [0, T ] × Ω,
u(0, x) = φ (x), x ∈ Ω, (4.45)

∂u
= au + g on [0, T ] × ∂ Ω,
∂n
where Ω ⊂ Rn is a bounded open set, ∂ Ω is smooth,
Z
dσ ≤ A,
∂Ω
dσ is a vector element of ∂ Ω, and

(H1) f : [0, T ] × Ω × R → [0, ∞), f ∈ C ([0, T ] × Ω × R),
f (t, x, u) ≤ a0 (t, x) + a1 (t, x)um + a2 (t, x)uk , (t, x) ∈ [0, T ] × Ω, u ∈ [0, ∞),
a j ∈ C ([0, T ] × Ω), j = 0, 1, 2,
Z TZ
a j (τ, x)dxdτ ≤ A, j = 0, 1, 2,
0 Ω
(H2) g : [0, T ] × ∂ Ω → [0, ∞), g ∈ C ([0, T ] × ∂ Ω),

Z TZ
g(τ, x)dσ dτ ≤ A,
0 ∂Ω
(H3) φ ∈ C (Ω), B0 ≤ φ (x) ≤ A, x ∈ Ω.
Definition 4.4.1 We say that u ∈ C ([0, T ] × Ω) is a solution of IBVP (4.45), if it satisfies

Z tZ Z tZ
u(t, x) = φ (x) + a u(τ, x)dσ dτ + g(τ, x)dσ dτ
0 ∂Ω 0 ∂Ω
Z tZ
+ f (τ, y, u(τ, y))dydτ, (t, x) ∈ [0, T ] × Ω.
0 Ω
Let E = C ([0, T ] × Ω) be endowed with the norm
kuk = max |u(t, x)|.

(t,x)∈[0,T ]×Ω
Define
P
f = {u ∈ E : u(t, x) ≥ 0, (t, x) ∈ [0, T ] × Ω}.
f
For u ∈ E define the operators
Tu(t, x) = (1 + ε)u(t, x) − R,
Z tZ Z tZ
Fu(t, x) = εR − ε φ (x) + a u(τ, x)dσ dτ + g(τ, x)dσ dτ
0 ∂Ω 0 ∂Ω
Z tZ
+ f (τ, y, u(τ, y))dydτ , (t, x) ∈ [0, T ] × Ω.
0 Ω
Note that any fixed point u ∈ E of the operator T + F is a solution of IBVP (4.45).
Theorem 4.4.2 Suppose (4.42)-(4.44), (H1)-(H3). Then IBVP (4.45) has at least three
solutions u1 , u2 , u3 ∈ P such that
0 < ku1 k ≤ r1 ≤ ku2 k < ρ < ku3 k ≤ r2 .
PR = {u ∈ P : kuk < R},
Pρ = {u ∈ P : kuk < ρ},
Pri = {u ∈ P : kuk < ri }, i = 1, 2.

1. Let u, v ∈ P. Then
kTu − T vk = (1 + ε)ku − vk.
Therefore T : PR → E is an expansive operator.
2. For u ∈ PR , we have
Z TZ
|Fu(t, x)| ≤ εR + εφ (x) + εa u(τ, x)dσ dτ
0 ∂Ω
Z TZ Z TZ
+ε g(τ, x)dσ dτ + ε f (τ, y, u(τ, y))dydτ
0 ∂Ω 0 Ω
≤ εR + εA + εaT RA + εA
Z TZ Z TZ
+ε a0 (τ, y)dydτ + ε a1 (τ, y)(u(τ, y))m dydτ
0 Ω 0 Ω
Z TZ
+ε a2 (τ, y)(u(τ, y))k dydτ
0 Ω
≤ εR + εA + εaT RA + εA + εA + εARm + εARk
= ε(R + 3A + aT RA + ARm + ARk )
< sε, (t, x) ∈ [0, T ] × Ω.

(4.46)
As in the proof of Theorem 3.2.3, we have that F : PR → E is a 0-set contrac-
tion.
3. For i = 1, 2, x ∈ ∂ Pri , we get, using (4.46),
|Fu(t, x) + T 0(t, x)| ≤ ε(3A + aTri A + Arim + Arik )
< εri , (t, x) ∈ [0, T ] × Ω.
Thus conditions (a) and (c) of Theorem 1.9.21 hold.
4. Now we will check condition (b) of Theorem 1.9.21. Let u0 ∈ P ∗ and as-
sume that there exists a u1 ∈ P and a λ1 ≥ 0 with u1 − λ1 u0 ∈ ∂ Pρ such that
Fu1 = (I − T )(u1 − λ1 u0 ).
Then
0 < −εR + εB0
≤ −εR + εφ (x)
≤ −Fu1 (t, x)
= −(I − T )(u1 − λ1 u0 )(t, x)
= ε(u1 (t, x) − λ1 u0 (t, x)) − εR
≤ εu1 (t, x) − εR
≤ ε(ρ − R)
< 0, (t, x) ∈ [0, T ] × Ω.

F(PR ) ⊂ (I − T )(PR ). (4.47)
Let z ∈ PR be fixed and consider on PR the equation
u(t, x) = Tu(t, x) + Fz(t, x)
= Au(t, x), (t, x) ∈ [0, T ] × Ω.
For u, v ∈ PR , we obtain
kAu − Avk = kTu − T vk
= (1 + ε)ku − vk,
PR ⊂ A(PR ). (4.48)
For any given v ∈ PR , consider on PR , the equation
Au(t, x) = v(t, x) or Tu(t, x) = v(t, x) − Fz(t, x), (t, x) ∈ [0, T ] × Ω.

(4.49)
We have
Z tZ
v(t, x) − Fz(t, x) = v(t, x) − εR + ε φ (x) + a u(τ, x)dσ dτ
0 ∂Ω
Z tZ Z tZ
+ g(τ, x)dσ dτ + f (τ, y, u(τ, y))dydτ
0 ∂Ω 0 Ω
≥ v(t, x) − εR + εφ (x)
≥ v(t, x) − εR + εB0
≥ 0, (t, x) ∈ [0, T ] × Ω,
and
v(t, x) − Fz(t, x) ≤ R + ε(R + 3A + aT RA + ARm + ARk )
= R1 , (t, x) ∈ [0, T ] × Ω.
P, and
\
Note that R1 > R and v − Fz ∈ B(R1 )
T : PR → B(R1 ) P
\
is a bijection. Thus, there is a u ∈ PR which satisfies (4.49) and then (4.48)

holds. Hence, and applying Theorem 1.9.8, it follows that there exists a w ∈ PR
so that
w(t, x) = Aw(t, x), (t, x) ∈ [0, T ] × Ω,
and (4.47) holds.
By Theorem 1.9.21, it follows that IBVP (4.45) has at least three solutions u1 , u2 , u3 ∈ P
so that
0 < ku1 k ≤ r1 ≤ ku2 k < ρ < ku3 k ≤ r2 .
Let the positive constants ε, A, B0 , r, L, R, m, k, T and a ≥ 0 satisfy the following conditions

B0 > R, r < L < R, (4.50)
R < εR − ε(3A + aT RA + ARm + ARk ), (4.51)
3A + aT LA + ALm + ALk ≤ L. (4.52)
Theorem 4.4.4 Suppose (4.50)-(4.52), (H1)-(H3). Then IBVP (4.45) has at least two so-
lutions u1 , u2 ∈ P such that
r < ku1 k < L < ku3 k < R.
Proof 4.4.5 Define
Pr = {x ∈ P : kxk < r},
PL = {x ∈ P : kxk < L},
Pr,L = {x ∈ P : r < kxk < L},
PL,R = {x ∈ P : L < kxk < R}.

We have
Pr,L PR
\
6= 0,
/
PL,R PR
\
6= 0.
/
F(PR ) ⊂ (I − T )(PR ).
2. Let u0 ∈ P ∗ and assume that there is a λ1 ≥ 0 and a u1 ∈ P with u1 −

Fu1 = u1 − λ1 u0 .
Then, using (4.46), we have
r ≥ u1 (t, x)
≥ u1 (t, x) − λ1 u0 (t, x)
= Fu1 (t, x)
≥ εR − ε(3A + aTrA + Arm + Ark ), (t, x) ∈ [0, T ] × Ω.

This is a contradiction. Therefore condition (a) of Theorem 1.9.23 holds. As in
the above, one can prove that condition (c) of Theorem 1.9.23 holds.
3. Let u ∈ ∂ PL . Using (4.46), we get
|Fu(t, x) + T 0(t, x)| ≤ ε(3A + aT LA + ALm + ALk )
≤ εL
= εkuk, (t, x) ∈ [0, T ] × Ω.

Hence,
kFu + T 0k ≤ εkuk.
Now, assume that there is a u1 ∈ ∂ PL
\
Ω such that
u1 = Tu1 + Fu1 .
Let (t1 , x1 ) ∈ [0, T ] × Ω be such that
L = u1 (t1 , x1 ).
Then
L = u1 (t1 , x1 )
= Tu1 (t1 , x1 ) + Fu1 (t1 , x1 )
= (1 + ε)u1 (t1 , x1 ) − εR
Z tZ
+εR − ε φ (x) + a u(τ, x)dσ dτ
0 ∂Ω
Z tZ Z tZ
+ g(τ, x)dσ dτ + f (τ, y, u(τ, y))dydτ
0 ∂Ω 0 Ω
≤ (1 + ε)L − εφ (x)
≤ (1 + ε)L − εB0 , (t, x) ∈ [0, T ] × Ω,

whereupon
ε(L − B0 ) ≥ 0.
Hence, and applying Theorem 1.9.23, it follows that IBVP (4.45) has at least two solutions
u1 , u2 ∈ P such that
r < ku1 k < L < ku2 k < R.
5
Applications to Hyperbolic Equations

Conditions for an IVP for a Class of Hyperbolic Equations
In this section we investigate the Cauchy problem
utt − uxx = f (t, x, u, ut , ux ) in (0, ∞) × R, (5.1)
u(0, x) = φ (x), ut (0, x) = ψ(x) in R, (5.2)

where φ , ψ ∈ C (R), f : [0, ∞) × R × R × R × R → R is a given continuous function,
2
u : [0, ∞) × R → R is unknown. We will start with the following useful lemma.

Lemma 5.1.1 Let f ∈ C ([a, b] × [c, d] × R × R × R), g ∈ C 2 ([c, d]), h ∈ C 1 ([c, d]). Then
the function u ∈ C 1 ([a, b], C 2 ([c, d])) is a solution to the problem
utt − uxx = f (t, x, u, ut , ux ) in (a, b] × [c, d], (5.3)
u(a, x) = g(x), ut (a, x) = h(x) in [c, d], (5.4)

if and only if it is a solution to the integral equation
Z xZ y
(u(t, z) − g(z) − (t − a)h(z)) dzdy
c c
Z t Z t1
− (u(τ, x) − u(τ, c) − (x − c)ux (τ, c)) dτdt1 (5.5)
a a
Z t Z t1 Z x Z y
= f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))dzdydτdt1 , x ∈ [c, d],t ∈ [a, b].
a a c c
Proof 5.1.2 1. Let u ∈ C 1 ([a, b], C 2 ([c, d])) be a solution to problem (5.3), (5.4).
We integrate equation (5.3) with respect to x and we get
Z x Z x
utt (t, z)dz − uxx (t, z)dz
c c
Z x
= f (t, z, u(t, z), ut (t, z), ux (t, z))dz or
c
221
222 Applications to Hyperbolic Equations
Z x
utt (t, z)dz − ux (t, x) + ux (t, c)
c
Z x
= f (t, z, u(t, z), ut (t, z), ux (t, z))dz, x ∈ [c, d],t ∈ [a, b].
c
Now we integrate the last equation with respect to x and we find
Z xZ y Z x
utt (t, z)dzdy − (ux (t, z) − ux (t, c)) dz
c c c
Z xZ y
= f (t, z, u(t, z), ut (t, z), ux (t, z))dzdy
c c
or Z xZ y
utt (t, z)dzdy − u(t, x) + u(t, c) + (x − c)ux (t, c)
c c
Z xZ y
= f (t, z, u(t, z), ut (t, z), ux (t, z))dzdy, x ∈ [c, d],t ∈ [a, b].
c c
We integrate the last equality with respect to t and we obtain
Z tZ xZ y Z t
utt (s, z)dzdyds − (u(s, x) − u(s, c) − (x − c)ux (s, c)) ds
a c c a
Z tZ xZ y
= f (s, z, u(s, z), ut (s, z), ux (s, z))dzdyds
a c c
or
Z xZ y Z t
(ut (t, z) − h(z)) dzdy − (u(s, x) − u(s, c) − (x − c)ux (s, c)) ds
c c a
Z tZ xZ y
= f (s, z, u(s, z), ut (s, z), ux (s, z))dzdyds, x ∈ [c, d],t ∈ [a, b].
a c c
Now we integrate the last equation with respect to t and we obtain

Z tZ xZ y
(ut (s, z) − h(z)) dzdyds
a c c
Z t Z t1
− (u(s, x) − u(s, c) − (x − c)ux (s, c))dsdt1
a a
Z t Z t1 Z x Z y
= f (s, z, u(s, z), ut (s, z), ux (s, z))dzdydsdt1 , x ∈ [c, d],t ∈ [a, b],
a a c c
Differentiability of Classical Solutions of an IVP for Hyperbolic Equations 223
or
Z xZ y
(u(t, z) − g(z) − (t − a)h(z)) dzdy
c c
Z t Z t1
− (u(s, x) − u(s, c) − (x − c)ux (s, c))dsdt1
a a
Z t Z t1 Z x Z y
= f (s, z, u(s, z), ut (s, z), ux (s, z))dzdydsdt1 , x ∈ [c, d],t ∈ [a, b],
a a c c
i.e., u satisfies equation (5.5).

2. Let u ∈ C 2 ([a, b], C 2 ([c, d])) be a solution to integral equation (5.5).
We differentiate equation (5.5) with respect to x and we get
Z x Z t Z t1
(u(t, z) − g(z) − (t − a)h(z)) dz − (ux (s, x) − ux (s, c)) dsdt1
c a a
Z t Z t1 Z x
= f (s, z, u(s, z), ut (s, z), ux (s, z))dzdsdt1 , x ∈ [c, d],t ∈ [a, b].
a a c
Again we differentiate with respect to x and we find

Z t Z t1
u(t, x) − g(x) − (t − a)h(x) − uxx (s, x)dsdt1
a a
Z t Z t1
(5.6)
= f (s, x, u(s, x), ut (s, x), ux (s, x))dsdt1 , x ∈ [c, d],t ∈ [a, b].
a a
Now we put t = a in the last equation and we find
u(a, x) = g(x), x ∈ [c, d]. (5.7)
Now we differentiate equation (5.6) with respect to t and we find

Z t
ut (t, x) − h(x) − uxx (τ, x)dτ
a
Z t
(5.8)
= f (τ, x, u(τ, x), ut (τ, x), ux (τ, x))dτ, x ∈ [c, d],t ∈ [a, b].
a
After we put t = a in (5.8), we get
ut (a, x) = h(x). (5.9)
From (5.7) and (5.9), we conclude that u(t, x) satisfies (5.4).

After we differentiate (5.8) with respect to t we get
utt (t, x) − uxx (t, x) = f (t, x, u(t, x), ut (t, x), ux (t, x)), t ∈ [a, b], x ∈ [c, d],
i.e., u(t, x) satisfies (5.3).
Theorem 5.1.3 Let f ∈ C ([0, ∞) × R × R × R × R), φ , ψ ∈ C 2 (R). Then problem (5.1),

(5.2) has a solution u ∈ C 2 ([0, ∞), C 2 (R)).
1
2(p + 1) p−1
Example 5.1.4 Let p > 1 and a = . Consider the Cauchy problem
(p − 1)2
utt − uxx = up in (0, ∞) × R,
2a
u(0, x) = a, ut (0, x) = − in R.
p−1
a
Then u(t, x) = 2 is its solution. Really,
(t + 1) p−1
2a 1 2a(p + 1) 1
ut (t, x) = − , utt (t, x) = 2p ,
p − 1 (t + 1) p+1
p−1 (p − 1)2 (t + 1) p−1
2a(p + 1) 1 aa p−1 p
utt − uxx = 2p = 2p = u ,
(p − 1)2 (t + 1) p−1 (t + 1) p−1
2a
u(0, x) = a, ut (0, x) = −
.
p−1
√
2
In particular, when p = 3, we have that u = is a solution to the problem
t +1
utt − uxx = u3 in (0, ∞) × R,

√ √
u(0, x) = 2, ut (0, x) = − 2 in R.
Proof 5.1.5 Step 1. Firstly, we will prove that the problem
utt − uxx = f (t, x, u, ut , ux ) in (0, 1] × [0, 1], (5.10)
u(0, x) = φ (x), ut (0, x) = ψ(x) in [0, 1] (5.11)

has a solution u ∈ C ([0, 1], C ([0, 1])).
2 2
Let E11 = C 2 ([0, 1], C 2 ([0, 1])) be endowed with the norm
n
||u|| = max max |u(t, x)|, max |ut (t, x)|, max |utt (t, x)|,
t,x∈[0,1] t,x∈[0,1] t,x∈[0,1]
o
t,x∈[0,1] t,x∈[0,1]
With Ke11 we denote the set of all equi-continuous families in E11 , i.e., for every ε > 0
there exists δ = δ (ε) > 0 such that
|u(t1 , x1 ) − u(t2 , x2 )| < ε, |ut (t1 , x1 ) − ut (t2 , x2 )| < ε,
|utt (t1 , x1 ) − utt (t2 , x2 )| < ε, |ux (t1 , x1 ) − ux (t2 , x2 )| < ε,
|uxx (t1 , x1 ) − uxx (t2 , x2 )| < ε
whenever |t1 − t2 | < δ , |x1 − x2 | < δ , for all u ∈ K

e11 .
0
Let B > max{||φ ||C 2 ([0,1]) , ||ψ||C 1 ([0,1]) } be arbitrarily chosen, K11 =K
e11 ,
0
K11 = {u ∈ K11 : ||u|| ≤ B}.
Since φ , ψ ∈ C ([0, 1]), f ∈ C ([0, 1] × [0, 1] × [−B, B] × [−B, B] × [−B, B]) we have that
there exists a constant M11 > 0 such that
|φ (x)|, |ψ(x)| ≤ M11 in [0, 1],
| f (t, x, y, z, z1 )| ≤ M11 in [0, 1] × [0, 1] × [−B, B] × [−B, B] × [−B, B].
Let l > 0 be chosen so that
l(5B + 3M11 ) ≤ B. (5.12)
Let also
0
L11 = {u ∈ K11 : ||u|| ≤ (1 + l)B}.
T11 (u)(t, x) = (1 + l)u(t, x),
Z xZ y
S11 (u)(t, x) = −lu(t, x) + l (u(t, z) − φ (z) − tψ(z))dzdy
0 0
Z t Z t1
−l (u(τ, x) − u(τ, 0) − xux (τ, 0))dτdt1
0 0
Z t Z t1 Z x Z y
−l f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))dzdydτdt1 .
0 0 0 0
a) S11 : K11 → K11 . Let u ∈ K11 . Then S11 (u) ∈ C 2 ([0, 1], C 2 ([0, 1])) and for (t, x) ∈
[0, 1] × [0, 1], using the choice (5.12) of the constant l, we have
Z xZ y
|S11 (u)(t, x)| = −lu(t, x) + l (u(t, z) − φ (z) − tψ(z))dzdy

0 0
Z t Z t1
−l (u(τ, x) − u(τ, 0) − xux (τ, 0))dτdt1
0 0
Z t Z t1 Z x Z y
−l f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))dzdydτdt1

0 0 0 0
Z xZ y
≤ l|u(t, x)| + l (|u(t, z)| + |φ (z)| + t|ψ(z)|) dzdy
0 0
Z t Z t1
+l (|u(τ, x)| + |u(τ, 0)| + x|ux (τ, 0)|) dτdt1
0 0
Z t Z t1 Z x Z y
+l | f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))|dzdydτdt1
0 0 0 0
Z xZ y
≤ lB + l (B + 2M11 )dzdy
0 0
Z t Z t1 Z t Z t1 Z x Z y
+l (2B + xB)dτdt1 + lM11 dzdydτdt1
0 0 0 0 0 0
≤ lB + l(B + 2M11 ) + 3lB + lM11
= l(5B + 3M11 )
≤ B,
Z xZ y
S11 (u)t (t, x) = −lut (t, x) + l (ut (t, z) − ψ(z))dzdy
0 0
Z t
−l (u(τ, x) − u(τ, 0) − xux (τ, 0))dτ
0
Z tZ xZ y
−l f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))dzdydτ,
0 0 0
Z xZ y
|S11 (u)t (t, x)| = −lut (t, x) + l (ut (t, z) − ψ(z))dzdy

0 0
Z t
−l (u(τ, x) − u(τ, 0) − xux (τ, 0))dτ
0
Z tZ xZ y
−l f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))dzdydτ

0 0 0
Z xZ y
≤ l|ut (t, x)| + l (|ut (t, z)| + |ψ(z)|)dzdy
0 0
Z t
+l (|u(τ, x)| + |u(τ, 0)| + x|ux (τ, 0)|) dτ
0
Z tZ xZ y
+l | f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))|dzdydτ
0 0 0
Z xZ y
≤ lB + l (B + M11 )dzdy
0 0
Z t Z tZ xZ y
+l (2B + xB)dτ + lM11 dzdydτ
0 0 0 0
≤ lB + l(B + M11 ) + 3lB + lM11
= l(5B + 2M11 )
≤ B,
Z xZ y
S11 (u)tt (t, x) = −lutt (t, x) + l utt (t, z)dzdy
0 0
−l(u(t, x) − u(t, 0) − xux (t, 0))

Z xZ y
−l f (t, z, u(t, z), ut (t, z), ux (t, z))dzdy,
0 0
Z xZ y
|S11 (u)tt (t, x)| = −lutt (t, x) + l utt (t, z)dzdy

0 0
−l(u(t, x) − u(t, 0) − xux (t, 0))

Z xZ y
−l f (t, z, u(t, z), ut (t, z), ux (t, z))dzdy

0 0
Z xZ y
≤ l|utt (t, x)| + l |utt (t, z)|dzdy
0 0
+l(|u(t, x)| + |u(t, 0)| + x|ux (t, 0)|)

Z xZ y
+l | f (t, z, u(t, z), ut (t, z), ux (t, z))|dzdy
0 0
≤ lB + lB + 3lB + lM11
= l(5B + M11 )
≤ B,
Z x
S11 (u)x (t, x) = −lux (t, x) + l (u(t, z) − φ (z) − tψ(z))dz
0
Z t Z t1
−l (ux (τ, x) − ux (τ, 0))dτdt1
0 0
Z t Z t1 Z x
−l f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))dzdτdt1 ,
0 0 0
Z x
|S11 (u)x (t, x)| = −lux (t, x) + l (u(t, z) − φ (z) − tψ(z))dz

0
Z t Z t1
−l (ux (τ, x) − ux (τ, 0))dτdt1
0 0
Z t Z t1 Z x
−l f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))dzdτdt1

0 0 0
Z x
≤ l|ux (t, x)| + l (|u(t, z)| + |φ (z)| + t|ψ(z)|) dz
0
Z t Z t1
+l (|ux (τ, x)| + |ux (τ, 0)|) dτdt1
0 0
Z t Z t1 Z x
+l | f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))|dzdτdt1
0 0 0
≤ lB + l(B + 2M11 ) + 2lB + lM11

= l(4B + 3M11 )
≤ B,
S11 (u)xx (t, x) = −luxx (t, x) + l(u(t, x) − φ (x) − tψ(x))
Z t Z t1
−l uxx (τ, x)dτdt1
0 0
Z t Z t1
−l f (τ, x, u(τ, x), ut (τ, x), ux (τ, x))dτdt1 ,
0 0

|S11 (u)xx (t, x)| = −luxx (t, x) + l(u(t, x) − φ (x) − tψ(x))

Z t Z t1
−l uxx (τ, x)dτdt1
0 0
Z t Z t1
−l f (τ, x, u(τ, x), ut (τ, x), ux (τ, x))dτdt1

0 0
≤ l|uxx (t, x)| + l (|u(t, x)| + |φ (x)| + t|ψ(x)|)
Z t Z t1
+l |uxx (τ, x)|dτdt1
0 0
Z t Z t1
+l | f (τ, x, u(τ, x), ut (τ, x), ux (τ, x))|dτdt1
0 0
≤ lB + l(B + 2M11 ) + lB + lM11
= l(3B + 3M11 )
≤ B.
We note that {S11 (u) : u ∈ K11 } is an equi-continuous family in E11 . Consequently

S11 : K11 → K11 . Also, S11 (K11 ) ⊂ K11 ⊂ L11 , i.e., S11 (K11 ) resides in a compact
subset of L11 .
b) S11 : K11 → K11 is a continuous operator. We note that if {un }∞
n=1 is a sequence of
elements of K11 such that un −→ u in K11 as n −→ ∞, then S11 (un ) −→ S11 (u) in
K11 as n −→ ∞. Therefore S11 : K11 7−→ K11 is a continuous operator.
c) T11 : K11 → L11 is an expansive operator and onto. For u, v ∈ K11 we have that
||T11 (u) − T11 (v)|| = (1 + l)||u − v||,
i.e., T11 : K11 → L11 is an expansive operator with constant 1 + l.
v
Let v ∈ L11 . Then ∈ K11 and
1+l

v
T11 = v,
1+l
i.e., T11 : K11 → L11 is onto.
From items a, b, and c and from Theorem 1.9.12, it follows that there is u11 ∈ K11 such
that
T11 u11 + S11 u11 = u11
or
u11 (t, x) = (1 + l)u11 (t, x) − lu11 (t, x)
Z xZ y
+l (u11 (t, z) − φ (z) − tψ(z))dzdy
0 0
Z t Z t1
−l (u11 (τ, x) − u11 (τ, 0) − xu11x (τ, 0))dτdt1
0 0
Z t Z t1 Z x Z y
−l f (τ, z, u11 (τ, z), u11t (τ, z)u11x (τ, z))dzdydτdt1 ,
0 0 0 0
or
Z xZ y
0 = (u11 (t, z) − φ (z) − tψ(z))dzdy
0 0
Z t Z t1
− (u11 (τ, x) − u11 (τ, 0) − xu11x (τ, 0))dτdt1
0 0
Z t Z t1 Z x Z y
− f (τ, z, u11 (τ, z), u11t (τ, z), u11x (τ, z))dzdydτdt1 ,
0 0 0 0
whereupon, using Lemma 5.1.1, we conclude that u11 ∈ C 2 ([0, 1], C 2 ([0, 1])) is a solution
to problem (5.10), (5.11).
Step 2. Now we consider the problem
utt − uxx = f (t, x, u(t, x), ut (t, x), ux (t, x)) in (0, 1] × [1, 2], (5.13)
u(0, x) = φ (x), ut (0, x) = ψ(x) in [1, 2]. (5.14)

Let E12 = C 2 ([0, 1], C 2 ([1, 2])) be endowed with the norm
n
(t,x)∈[0,1]×[1,2] (t,x)∈[0,1]×[1,2]
max |utt (t, x)|, max |ux (t, x)|,

(t,x)∈[0,1]×[1,2] (t,x)∈[0,1]×[1,2]
o
max |uxx (t, x)| .
(t,x)∈[0,1]×[1,2]
With K
e12 we denote the set of all equi-continuous families in E12 .
0
Let K12 =K
e12 ,
0
K12 = {u ∈ K12 : ||u|| ≤ B}.
Since φ , ψ ∈ C ([1, 2]), f ∈ C ([0, 1] × [1, 2] × [−B, B] × [−B, B] × [−B, B]), we have that
there exists a constant M12 > 0 such that
|φ (x)|, |ψ(x)| ≤ M12 in [1, 2],
| f (t, x, y, z, z1 )| ≤ M12 in [0, 1] × [1, 2] × [−B, B] × [−B, B] × [−B, B].

Let l1 > 0 be chosen so that
l1 (5B + 3M12 ) ≤ B. (5.15)
Let also
0
L12 = {u ∈ K12 : ||u|| ≤ (1 + l1 )B}.
T12 (u)(t, x) = (1 + l1 )u(t, x),
Z xZ y
S12 (u)(t, x) = −l1 u(t, x) + l1 (u(t, z) − φ (z) − tψ(z))dzdy
1 1
Z t Z t1
−l1 (u(τ, x) − u11 (τ, 1) − (x − 1)u11x (τ, 1))dτdt1
0 0
Z t Z t1 Z x Z y
−l1 f (τ, z, u(τ, z), ut (τ, z), ux (τ, z))dzdydτdt1 .
0 0 1 1
As in Step 1, one can prove that there is u12 ∈ C 2 ([0, 1], C 2 ([1, 2])) which is a solution to
problem (5.13), (5.14). This solution u12 satisfies the integral equation
Z xZ y
0 = (u12 (t, z) − φ (z) − tψ(z))dzdy
1 1
Z t Z t1
− (u12 (τ, x) − u11 (τ, 1) − (x − 1)u11x (τ, 1))dτdt1 (5.16)
0 0
Z t Z t1 Z x Z y
− f (τ, z, u12 (τ, z), u12t (τ, z), u12x (τ, z))dzdydτdt1 .
0 0 1 1
Now we put x = 1 in (5.16) and we find

Z t Z t1
(u12 (τ, 1) − u11 (τ, 1))dτdt1 = 0,
0 0
which we differentiate twice with respect to t and we get
u12 (t, 1) = u11 (t, 1) in [0, 1]. (5.17)
Now we differentiate (5.16) with respect to x and we find

Z x
0 = (u12 (t, z) − φ (z) − tψ(z))dz
1
Z t Z t1
− (u12x (τ, x) − u11x (τ, 1))dτdt1
0 0
Z t Z t1 Z x
− f (τ, z, u12 (τ, z), u12t (τ, z), u12x (τ, z))dzdτdt1 .
0 0 1
In the last equation we put x = 1 and we obtain

Z t Z t1
(u12x (τ, 1) − u11x (τ, 1))dτdt1 = 0,
0 0
which we differentiate twice with respect to t and we find
u12x (t, 1) = u11x (t, 1) in [0, 1]. (5.18)
Now we differentiate (5.17) with respect to t and we get
u12t (t, 1) = u11t (t, 1) in [0, 1],
which we differentiate with respect to t and we find
u12tt (t, 1) = u11tt (t, 1).

Hence, based on (5.17), (5.18) and

f (t, 1, u11 (t, 1), u11t (t, 1), u11x (t, 1)) = f (t, 1, u12 (t, 1), u12t (t, 1), u12x (t, 1)),
we find
u12xx (t, 1) = u12tt (t, 1) − f (t, 1, u12 (t, 1), u12t (t, 1), u12x (t, 1))
= u11tt (t, 1) − f (t, 1, u11 (t, 1), u11t (t, 1), u11x (t, 1))
= u11xx (t, 1) in [0, 1].


 u11 (t, x) in [0, 1] × [0, 1]
u(t, x) =
u12 (t, x) in [0, 1] × [1, 2],

is a C ([0, 1], C ([0, 2]))-solution to the problem

2 2
utt − uxx = f (t, x, u(t, x), ut (t, x), ux (t, x)) in (0, 1] × [0, 2],
u(0, x) = φ (x), ut (0, x) = ψ(x) in [0, 2].

(5.19)
u(0, x) = φ (x), ut (0, x) = ψ(x) in [2, 3].
As in the above, there is u13 ∈ C 2 ([0, 1], C 2 ([2, 3])) which is a solution to problem (5.19)
and satisfies the integral equation
Z xZ y
0 = (u13 (t, z) − φ (z) − tψ(z))dzdy
2 2
Z t Z t1
− (u13 (τ, x) − u12 (τ, 2) − (x − 2)u12x (τ, 2))dτdt1
0 0
Z t Z t1 Z x Z y
0 0 2 2
The function 

 u11 (t, x) in [0, 1] × [0, 1]



u(t, x) = u12 (t, x) in [0, 1] × [1, 2]




u13 (t, x) in [0, 1] × [2, 3]

is a C 2 ([0, 1], C 2 ([0, 3]))-solution to the problem

utt − uxx = f (t, x, u(t, x), ut (t, x), ux (t, x)) in [0, 1] × [0, 3],
u(0, x) = φ (x), ut (0, x) = ψ(x) in [0, 3],
and so on. We construct a solution u1 ∈ C 2 ([0, 1], C 2 (R)) to the problem

utt − uxx = f (t, x, u(t, x), ut (t, x), ux (t, x)) in (0, 1] × R,
u(0, x) = φ (x), ut (0, x) = ψ(x) in R.

u(1, x) = u1 (1, x), ut (1, x) = u1t (1, x) in [0, 1].

As in the above, this problem has a solution u21 ∈ C 2 ([1, 2], C 2 ([0, 1])) which satisfies
Z xZ y
0 = (u21 (t, z) − u1 (1, z) − (t − 1)u1t (1, z))dzdy
0 0
Z t Z t1
− (u21 (τ, x) − u21 (τ, 0) − xu21x (τ, 0))dτdt1
1 1
Z t Z t1 Z x Z y
1 1 0 0
We have that, x ∈ [0, 1],

u21 (1, x) = u1 (1, x) = u11 (1, x), u21x (1, x) = u1x (1, x) = u11x (1, x),
u21xx (1, x) = u1xx (1, x) = u11xx (1, x), u21t (1, x) = u1t (1, x) = u11t (1, x),
u21tt (1, x) = u1tt (1, x) = u11tt (1, x).

Similarly, we construct a solution u ∈ C 2 ([0, 2], C 2 (R)) of the problem
utt − uxx = f (t, x, u(t, x), ut (t, x), ux (t, x)) in (0, 2] × R,
u(0, x) = φ (x), ut (0, x) = ψ(x) in R.

Similarly, we construct a solution u ∈ C 2 ([0, ∞), C 2 (R)) to problem (5.1), (5.2).

We note that umn , m, n ∈ N, satisfies the integral equation
Z x Z y
0 = (umn (t, z) − um−1 (m − 1, z) − (t − (m − 1))um−1t (m − 1, z))dzdy
n−1 n−1
Z t Z t
− (umn (τ, x) − umn−1 (τ, n − 1) − (x − (n − 1))umn−1x (τ, n − 1))dτdt1
m−1 m−1
Z t Z t1 Z x Z y
− f (τ, z, umn (τ, z), umnt (τ, z), umnx (τ, z))dzdydτdt1
m−1 m−1 n−1 n−1
with the convention um0 (τ, 0) = um1 (τ, 0), m ∈ N, u0 (0, z) = u1 (0, z), u0t (0, z) = u1t (0, z).
∂f ∂f ∂f
Theorem 5.1.6 Let f ∈ C ([0, ∞) × R × R × R × R), and , , exist and are con-
∂ u ∂ ut ∂ ux
tinuous in [0, ∞)×R ×R ×R ×R, φ , ψ ∈ C (R). Let also u(t, x, φ , ψ) ∈ C 2 ([0, ∞), C 2 (R))
2
be a solution to problem (5.1), (5.2). Then u(t, x, φ , ψ) is differentiable with respect to φ

∂u
and v(t, x) = (t, x, φ , ψ) satisfies the following initial value problem
∂φ
∂f
vtt − vxx = (t, x, u(t, x, φ , ψ), ut (t, x, φ , ψ), ux (t, x, φ , ψ))v
∂u
∂f
+ (t, x, u(t, x, φ , ψ), ut (t, x, φ , ψ), ux (t, x, φ , ψ))vt (5.20)
∂ ut
∂f
+ (t, x, u(t, x, φ , ψ), ut (t, x, φ , ψ), ux (t, x, φ , ψ))vx
∂ ux
in [0, ∞) × R,
v(0, x) = 1, vt (0, x) = 0, in R. (5.21)
Proof 5.1.7 We have that the solution u(t, x, φ , ψ) satisfies the following integral equation
Z xZ y
Q(φ ) = (u(t, z, φ (z), ψ(z)) − φ (z) − tψ(z))dzdy
0 0
Z t Z t1
− (u(τ, x, φ (x), ψ(x)) − u(τ, 0, φ (0), ψ(0)) − xux (τ, 0, φ (0), ψ(0)))dτdt1
0 0
Z t Z t1 Z x Z y
− f (τ, z, u(τ, z, φ (z), ψ(z)), ut (τ, z, φ (z), ψ(z)), ux (τ, z, φ (z), ψ(z)))
0 0 0 0
dzdydτdt1 = 0.
Then
Z xZ y
Q(φ ) − Q(φ1 ) = (u(t, z, φ (z), ψ(z)) − u(t, z, φ1 (z), ψ(z)) − (φ (z) − φ1 (z)))dzdy
0 0
Z t Z t1
− (u(τ, x, φ (x), ψ(x)) − u(τ, x, φ1 (x), ψ(x)))dτdt1
0 0
Z t Z t1
+ (u(τ, 0, φ (0), ψ(0)) − u(τ, 0, φ1 (0), ψ(0)))dτdt1
0 0
Z t Z t1
+ x(ux (τ, 0, φ (0), ψ(0)) − ux (τ, 0, φ1 (0), ψ(0)))dτdt1
0 0
Z t Z t1 Z x Z y
− f (τ, z, u(τ, z, φ (z), ψ(z)), ut (τ, z, φ (z), ψ(z)), ux (τ, z, φ1 (z), ψ(z)))
0 0 0 0

− f (τ, z, u(τ, z, φ1 (z), ψ(z)), ut (τ, z, φ1 (z), ψ(z)), ux (τ, z, φ1 (z), ψ(z))) dzdydτdt1
Z xZ y
∂u
= (t, z, φ (z), ψ(z)) − 1 dzdy
0 0 ∂φ
Z t Z t1 Z t Z t1
∂u ∂u
− (τ, x, φ (x), ψ(x))dτdt1 + (τ, 0, φ (0), ψ(0))dτdt1
0 ∂φ 0 0 0 ∂φ
Z t Z t1
∂u
+ x (τ, 0, φ (0), ψ(0))dτdt1
0 0 ∂φ x
Z t Z t1 Z x Z y
∂f
− (τ, z, u(τ, z, φ (z), ψ(z)), ut (τ, z, φ (z), ψ(z)), ux (τ, z, φ (z), ψ(z)))
0 0 0 0 ∂u
∂u
× (τ, z, φ (z), ψ(z))dzdydτdt1
∂φ
Z t Z t1 Z x Z y
∂f
0 0 0 0 ∂ ut

∂u
∂φ t
Z t Z t1 Z x Z y
∂f
0 0 0 0 ∂ ux

∂u
∂φ x
+δ {φ , φ1 },
where δ {φ , φ1 } −→ 0 as φ (x) −→ φ1 (x) for every x ∈ R. Hence, when φ (x) −→ φ1 (x) for
every x ∈ R, we get
Z xZ y Z t Z t1
0= (v(t, z) − 1)dzdy − v(τ, x)dτdt1
0 0 0 0
Z t Z t1 Z t Z t1
+ v(τ, 0)dτdt1 + xvx (τ, 0)dτdt1
0 0 0 0
Z t Z t1 Z x Z y
∂f
0 0 0 0 ∂u
×v(τ, z)dzdydτdt1
Z t Z t1 Z x Z y (5.22)
∂f
0 0 0 0 ∂ ut
×vt (τ, z)dzdydτdt1

Z t Z t1 Z x Z y
∂f
0 0 0 0 ∂ ux
×vx (τ, z)dzdydτdt1 ,

which we differentiate twice in x and t and we get that v satisfies (5.20). Now we put t = 0
in (5.22) and then we differentiate twice in x, after which we differentiate (5.22) in t and
then we put t = 0, and we find that v satisfies (5.21). This completes the proof.
Remark 5.1.8 As in the above one can prove the following theorem.
∂f ∂f ∂f
Theorem 5.1.9 Let f ∈ C ([0, ∞) × R × R × R × R), and , , exist and are con-
∂ u ∂ ut ∂ ux
tinuous in [0, ∞)×R ×R ×R ×R, φ , ψ ∈ C 2 (R). Let also u(t, x, φ , ψ) ∈ C 2 ([0, ∞), C 2 (R))
be a solution to problem (5.1), (5.2). Then u(t, x, φ , ψ) is differentiable with respect to ψ
∂u
and v(t, x) = (t, x, φ , ψ) satisfies the following initial value problem
∂ψ
∂f
vtt − vxx = (t, x, u(t, x, φ , ψ), ut (t, x, φ , ψ), ux (t, x, φ , ψ))v
∂u
∂f
+ (t, x, u(t, x, φ , ψ), ut (t, x, φ , ψ), ux (t, x, φ , ψ))vt
∂ ut
∂f
+ (t, x, u(t, x, φ , ψ), ut (t, x, φ , ψ), ux (t, x, φ , ψ))vx in [0, ∞) × R,
∂ ux
v(0, x) = 0, vt (0, x) = 1 in R.
5.2 Multiple Solutions for an IBVP with Robin Boundary Conditions

Let the positive constants ε, m, k, s, A, T, B0 , R, r1 , r2 , ρ and a ≥ 0 satisfy the following con-
ditions
1
0 < r1 < ρ < r2 < R, 0 < s < , (5.23)
2
R + 3A + AT + aT 2 R + ARm + ARk < s, (5.24)
3A + AT + aT 2 ri + Arim + Arik < ri , B0 > R. (5.25)

Consider the IBVP
utt = ∆u + f (t, x, u), (t, x) ∈ [0, T ] × Ω,
u(0, x) = φ (x), x ∈ Ω,
(5.26)
ut (0, x) = ψ(x), x ∈ Ω,
∂u
= au + g on [0, T ] × ∂ Ω,
∂n
Z
dσ ≤ A,
∂Ω
and dσ is a vector element of ∂ Ω,

(H1) f : [0, T ] × Ω × R → [0, ∞), f ∈ C ([0, T ] × Ω × R),
f (t, x, u) ≤ a0 (t, x) + a1 (t, x)um + a2 (t, x)uk , (t, x) ∈ [0, T ] × Ω, u ∈ [0, ∞),
a j ∈ C ([0, T ] × Ω), j = 0, 1, 2,
Z T Z
(T − τ) a j (τ, x)dxdτ ≤ A, j = 0, 1, 2,
0 Ω
(H2) g : [0, T ] × ∂ Ω → [0, ∞), g ∈ C ([0, T ] × ∂ Ω),

Z T Z
(T − τ) g(τ, x)dσ dτ ≤ A,
0 ∂Ω
(H3) φ , ψ ∈ C (Ω), B0 ≤ φ (x) ≤ A, 0 ≤ ψ(x) ≤ A, x ∈ Ω.
Definition 5.2.1 We say that u ∈ C ([0, T ] × Ω) is a solution of IBVP (5.26), if it satisfies

the following integral equation
Z t Z Z t Z
u(t, x) = φ (x) + tψ(x) + a (t − τ) u(τ, x)dσ dτ + (t − τ) g(τ, x)dσ dτ
0 ∂Ω 0 ∂Ω
Z t Z
+ (t − τ) f (τ, y, u(τ, y))dydτ, (t, x) ∈ [0, T ] × Ω.
0 Ω
Let E = C ([0, T ] × Ω) be endowed with the norm
kuk = max |u(t, x)|.

(t,x)∈[0,T ]×Ω
Define
P
f = {u ∈ E : u(t, x) ≥ 0, (t, x) ∈ [0, T ] × Ω}.
f
Tu(t, x) = (1 + ε)u(t, x) − εR,

Z t Z
Fu(t, x) = εR − ε φ (x) + tψ(x) + a (t − τ) u(τ, x)dσ dτ
0 ∂Ω
Z t Z
+ (t − τ) g(τ, x)dσ dτ
0 ∂Ω
Z t Z
+ (t − τ) f (τ, y, u(τ, y))dydτ , (t, x) ∈ [0, T ] × Ω.
0 Ω
Note that any fixed point u ∈ E of the operator T + F is a solution of IBVP (5.26).
Theorem 5.2.2 Suppose (5.23)-(5.25), (H1)-(H3). Then IBVP (5.26) has at least three
solutions u1 , u2 , u3 ∈ P such that
0 < ku1 k ≤ r1 ≤ ku2 k < ρ < ku3 k ≤ r2 .
PR = {u ∈ P : kuk < R},
Pρ = {u ∈ P : kuk < ρ},
Pri = {u ∈ P : kuk < ri }, i = 1, 2.
kTu − T vk = (1 + ε)ku − vk.

Z T Z
|Fu(t, x)| ≤ εR + εφ (x) + εtψ(x) + εa (T − τ) u(τ, x)dσ dτ
0 ∂Ω
Z T Z Z T Z
+ε (T − τ) g(τ, x)dσ dτ + ε (T − τ) f (τ, y, u(τ, y))dydτ
0 ∂Ω 0 Ω
≤ εR + εA + εTA + εaT 2 RA + εA
Z T Z Z T Z
+ε (T − τ) a0 (τ, y)dydτ + ε (T − τ) a1 (τ, y)(u(τ, y))m dydτ
0 Ω 0 Ω
Z T Z
+ε (T − τ) a2 (τ, y)(u(τ, y))k dydτ
0 Ω
≤ εR + εA + εAT + εaT 2 RA + εA + εA + εARm + εARk
= ε(R + 3A + AT + aT 2 RA + ARm + ARk )
< sε, (t, x) ∈ [0, T ] × Ω.

(5.27)
tion.

|Fu(t, x) + T 0(t, x)| ≤ ε(3A + AT + aT 2 ri A + Arim + Arik )
< εri , (t, x) ∈ [0, T ] × Ω.

Fu1 = (I − T )(u1 − λ1 u0 ).
Then
0 < −εR + εB0
≤ −εR + εφ (x)
≤ −Fu1 (t, x)
= −(I − T )(u1 − λ1 u0 )(t, x)
= ε(u1 (t, x) − λ1 u0 (t, x)) − εR
≤ ε(ρ − R)
< 0, (t, x) ∈ [0, T ] × Ω.

F(PR ) ⊂ (I − T )(PR ). (5.28)
u(t, x) = Tu(t, x) + Fz(t, x)
= Au(t, x), (t, x) ∈ [0, T ] × Ω.

= (1 + ε)ku − vk,
PR ⊂ A(PR ). (5.29)
Au(t, x) = v(t, x) or Tu(t, x) = v(t, x) − Fz(t, x), (t, x) ∈ [0, T ] × Ω.

(5.30)
We have
Z t Z
v(t, x) − Fz(t, x) = v(t, x) − εR + ε φ (x) + tψ(x) + a (t − τ) u(τ, x)dσ dτ
0 ∂Ω
Z t Z Z t Z
+ (t − τ) g(τ, x)dσ dτ + (t − τ) f (τ, y, u(τ, y))dydτ
0 ∂Ω 0 Ω
≥ v(t, x) − εR + εφ (x)
≥ v(t, x) − εR + εB0
> 0, (t, x) ∈ [0, T ] × Ω,
and
v(t, x) − Fz(t, x) ≤ R + ε(R + 3A + AT + aT 2 RA + ARm + ARk )
= R1 , (t, x) ∈ [0, T ] × Ω.
P, and
\
T : PR → B(R1 ) P
\
is a bijection. Thus, there is a u ∈ PR which satisfies (5.30) and then (5.29)

holds. Hence, and applying Theorem 1.9.8, it follows that there exists a w ∈ PR
so that
w(t, x) = Aw(t, x), (t, x) ∈ [0, T ] × Ω,
and (5.28) holds.
By Theorem 1.9.21, it follows that IBVP (5.26) has at least three solutions u1 , u2 , u3 ∈ P
so that
0 < ku1 k ≤ r1 ≤ ku2 k < ρ < ku3 k ≤ r2 .
Let the positive constants ε, A, B0 , r, L, R, m, k, T and a ≥ 0 satisfy the following conditions
B0 > R, r < L < R, (5.31)
R < εR − ε(3A + AT + aT 2 RA + ARm + ARk ), (5.32)
3A + AT + aT 2 LA + ALm + ALk ≤ L. (5.33)
Theorem 5.2.4 Suppose (5.31)-(5.33), (H1)-(H3). Then IBVP (5.26) has at least two so-
lutions u1 , u2 ∈ P such that
r < ku1 k < L < ku2 k < R.
Proof 5.2.5 Define
Pr = {x ∈ P : kxk < r},
PL = {x ∈ P : kxk < L},
Pr,L = {x ∈ P : R < kxk < L},
PL,R = {x ∈ P : L < kxk < R}.
We have
Pr,L PR
\
6= 0,
/
PL,R PR
\
6= 0.
/

F(PR ) ⊂ (I − T )(PR ).
2. Let u0 ∈ P ∗ and assume that there are a λ1 ≥ 0 and a u1 ∈ P with u1 −

Fu1 = u1 − λ1 u0 .
r ≥ u1 (t, x) − λ1 u0 (t, x)
= Fu1 (t, x)
≥ εR − ε(3A + AT + aT 2 rA + Arm + Ark ), (t, x) ∈ [0, T ] × Ω.

|Fu(t, x) + T 0(t, x)| ≤ ε(3A + AT + aT 2 LA + ALm + ALk )
≤ εL
= εkuk, (t, x) ∈ [0, T ] × Ω.
Hence,
Now, assume that there is a u1 ∈ ∂ PL
\
Ω such that
u1 = Tu1 + Fu1 .
Let (t1 , x1 ) ∈ [0, T ] × Ω be such that
L = u1 (t1 , x1 ).
Then
L = u1 (t1 , x1 )
= Tu1 (t1 , x1 ) + Fu1 (t1 , x1 )
= (1 + ε)u1 (t1 , x1 ) − εR
Z t Z
+εR − ε φ (x) + tψ(x) + a (t − τ) u(τ, x)dσ dτ
0 ∂Ω
Z t Z Z t Z
+ (t − τ) g(τ, x)dσ dτ + (t − τ) f (τ, y, u(τ, y))dydτ
0 ∂Ω 0 Ω
≤ (1 + ε)L − εφ (x)
≤ (1 + ε)L − εB0 , (t, x) ∈ [0, T ] × Ω,

whereupon
0 < ε(L − B0 ).
Hence, and applying Theorem 1.9.23, it follows that IBVP (5.26) has at least two solutions
r < ku1 k < L < ku2 k < R.
5.3 Periodic Solutions

Let the positive constants ε, M, ω, N, A, m, m j , j ∈ {0, . . . , n}, r1 , ρ, r2 , R, s be chosen so that
1
0 < r1 < ρ < r2 < R, 0<s< , (5.34)
2
R + ωRM(1 + M) + M < s,
2R + M + AR(ωM(1 + M) + 1) < s,
(5.35)
!
n
2R + M + AR(nωM + 1) + M 2 A + ARm + A ∑ Rm j < s,
j=0
N − AR − R > 0, M ≥ N, ρ + M < R. (5.36)
Consider the wave equation
utt − ∆u = f (t, x, u, ut , ux ), t ∈ R, x ∈ Rn ,
n
where ux = (ux1 , . . . , uxn ), ∆u = ∑ ux j x j . This equation is equivalent to the system
j=1
∂u
= u0
∂t
∂uj ∂ u0
= , j ∈ {1, . . . , n}, (5.37)
∂t ∂xj
n
∂ u0 ∂uj
= ∑ ∂ x j + f (t, x, u, u0 , u1 , . . . , un ), t ∈ R, x ∈ Rn ,
∂t j=1
∂u
where u j = , j ∈ {1, . . . , n}. Let E = C ([0, ω], C 1 (Rn )) be the space of all functions
∂xj
u ∈ C (R, C 1 (Rn )) which are ω-periodic with respect to the first variable. In E define a
norm
( )
∂
kuk = max sup |u(t, x)|, sup u(t, x) , j ∈ {1, . . . , n} .

(t,x)∈[0,ω]×Rn ∂ x j

(t,x)∈[0,ω]×Rn
Define
P
f = {v ∈ E : v(t, x) ≥ 0, (t, x) ∈ [0, ω] × Rn } .
Let P be the set of all equi-continuous families in P

f and
E n+2 = E × . . . × E,
| {z }
n+2
P n+2 = P ×...×P .
| {z }
n+2
For u = (u, u0 , u1 , . . . , un ) ∈ E n+2 define a norm

kuk = max{kuk, ku0 k, ku1 k, . . . , kun k}.
For u = (u, u0 , u1 , . . . , un ), v = (v, v0 , v1 , . . . , vn ) ∈ E n+2 we write u ≥ v if u ≥ v, u j ≥ v j ,
j ∈ {0, . . . , n}.
(H1) Suppose f ∈ C (R × R2n+2 ), f : R × R2n+2 → [0, ∞) is ω-periodic with respect to the
first variable and
n
m
f (t, x, u, u0 , . . . , un ) ≤ p(t, x)+c(t, x)um + ∑ c j (t, x)u j j , (t, x, u, u0 , . . . , un ) ∈ R×R2n+2 ,
j=0
where p, c, c j ∈ C (R × Rn ), j ∈ {0, . . . , n}, p, c, c j : R × Rn → [0, ∞) are ω-periodic with

respect to the first variable, j ∈ {0, . . . , n},
Z ωZ x
p(t, z)dzdt ≤ A,
0 −∞
Z ω Z xˆl
p(t, z˜xl )dzl dt ≤ A,
0 −∞
Z ωZ x
c(t, z)dzdt ≤ A,
0 −∞
Z ω Z xˆl
c(t, z˜xl )dzl dt ≤ A,
0 −∞
Z ωZ x
c j (t, z)dzdt ≤ A,
0 −∞
Z ω Z xˆl
c j (t, z˜xl )dzl dt ≤ A, x ∈ Rn , j ∈ {0, . . . , n}.
0 −∞
(H2) Suppose that a, a j : R → [0, ∞), j ∈ {0, . . . , n} are continuous ω-periodic functions
such that
N ≤ a(t) ≤ M,
e−[a]ω Rtt+s a(τ)dτ

N ≤ e ≤ M,
1 − e−[a]ω
N ≤ a j (t) ≤ M,
e−[a j ]ω Rtt+s a j (τ)dτ

N ≤ e ≤ M, t, s ∈ [0, ω], j ∈ {0, . . . , n}.
1 − e−[a j ]ω
(H3) Suppose that b, b j : R × Rn → [0, ∞), j ∈ {0, . . . , n} are continuous functions which
are ω-periodic with respect to the first variable and
N ≤ b(t, x) ≤ M,
Z x
b j (t, z)dz ≤ A,
−∞
Z xˆl
b(t, z˜xl )dzl ≤ A,
−∞
Z xˆl
b j (t, z˜xl )dzl ≤ A, (t, x) ∈ [0, ω] × Rn , j ∈ {0, . . . , n}, l ∈ {1, . . . , n}.
−∞
(H4) Suppose that h ∈ C (R, C 1 (Rn )), h : R × Rn → [0, ∞) is ω-periodic with respect to
the first variable and
N ≤ h(t, x) ≤ M,

∂
(t, x) ∈ [0, ω] × Rn .

h(t, x) ≤ M, l ∈ {1, . . . , n},
∂x
l
Here
Z x Z x1 Z xn
= ... ,
−∞ −∞ −∞
Z xˆl Z x1 Z xl−1 Z xl+1 Z xn
= ... ... ,
−∞ −∞ −∞ −∞ −∞
dz = dzn . . . dz1 ,
dzl = dzn . . . dzl+1 dzl−1 . . . dz1 ,
z˜xl = (z1 , . . . , zl−1 , xl , zl+1 , . . . , zn ), l ∈ {1, . . . , n}.
Lemma 5.3.1 Let (u, u0 , u1 , . . . , un ) ∈ E n+2 be a solution to the system
e−[a]ω
Z ω R t+s
a(τ)dτ
et (u0 (t + s, x) + a(t + s, x)u(t + s, x))ds − u(t, x) = 0,
1 − e−[a]ω 0
e−[a j ]ω
Z x
∂ u0
Z ω R t+s
a j (τ)dτ
e t b(t, z) (t + s, z) + a(t + s)u j (t + s, z) dzds
1 − e−[a j ]ω 0 −∞ ∂xj
Z x
− b(t, z)u j (t, z)dz = 0, j ∈ {1, . . . , n},
−∞
e−[a0 ]ω
Z x n
∂uj
Z ω R t+s
a0 (τ)dτ
1 − e−[a0 ]ω 0
e t
−∞
∑ ∂ x j (t + s, z)
j=1

+ f (t + s, z, u(t + s, z), u0 (t + s, z), u1 (t + s, z), . . . , un (t + s, z)) dzds
Z x
− b(t, z)u0 (t, z)dz = 0, (t, x) ∈ [0, ω] × Rn ,
−∞
(5.38)
then it is an ω-periodic solution with respect to the first variable of system (5.37).
Proof 5.3.2 From Lemma 2.1.1 and the first equation of system (5.38), we get the first
equation of system (5.37). Now we differentiate with respect to x1 , x2 , . . ., xn the second,
third, and so on to the last equation of system (5.38) and we get
e−[a j ]ω

∂ u0
Z ω R t+s
a j (τ)dτ
e t (t + s, x) + a(t + s)u j (t + s, x) ds = u j (t, x), j ∈ {1, . . . , n},
1 − e−[a j ]ω 0 ∂xj
e−[a0 ]ω n

∂uj
Z ω R t+s
a0 (τ)dτ
1 − e−[a0 ]ω 0
e t
∑ ∂ x j (t + s, x)
j=1

+ f (t + s, x, u(t + s, x), u0 (t + s, x), u1 (t + s, x), . . . , un (t + s, x)) ds = u0 (t, x), (t, x) ∈ [0, ω] × Rn .
Hence, and applying Lemma 2.1.1, we get the second, third, and so on to the last equation
of system (5.37). This completes the proof.
For u = (u, u0 , u1 , . . . , un ) ∈ E n+2 , define the operators
T u(t, x) = (1 + ε)u(t, x) − εR + εh(t, x),
Fu(t, x) = εR − εh(t, x)
e−[a]ω
Z ω R t+s

a(τ)dτ
−ε e t u0 (t + s, x) + a(t + s, x)u(t + s, x) ds,
1 − e−[a]ω 0
T j u(t, x) = (1 + ε)u j (t, x) − εR + εh(t, x),
Fj u(t, x) = −εu j (t, x) + εR − εh(t, x)
e−[a j ]ω
Z x
∂ u0
Z ω R t+s
a j (τ)dτ
−ε et b(t, z) (t + s, z)
1 − e−[a j ]ω 0 −∞ ∂xj
x
Z
+a(t + s)u j (t + s, z) dzds − b(t, z)u j (t, z)dz , j ∈ {1, . . . , n},
−∞
T0 u(t, x) = (1 + ε)u0 (t, x) − εR + εh(t, x),
F0 u(t, x) = −εu0 (t, x) + εR − εh(t, x)
e−[a0 ]ω
n Z x
∂uj
Z ω R t+s
a0 (τ)dτ
−ε et b(t, z) ∑ (t + s, z)
1 − e−[a0 ]ω0 −∞ j=1 ∂ x j

Z x
− b(t, z)u0 (t, z)dz ,
−∞
T̃ = (T, T0 , T1 , . . . , Tn ),
F̃ = (F, F0 , F1 , . . . , Fn ), (t, x) ∈ [0, ω] × Rn .
For Q > 0, define

PQ = {u ∈ E : kuk < Q}.
By Lemma 5.3.1 it follows that any fixed point u = (u, u0 , u1 , . . . , un ) ∈ E n+2 of T̃ + F̃ is an
ω-periodic solution of system (5.37).
Lemma 5.3.3 For u = (u, u0 , u1 , . . . , un ) ∈ PQn+2 , we have the following estimates
|Fu(t, x)| ≤ εR + εωQM(1 + M) + εM,

∂
∂ x Fu(t, x) ≤ εM + εωQM(1 + M),

l
|Fj u(t, x)| ≤ εR + εQ + εM + εAQ(ωM(1 + M) + 1), j ∈ {1, . . . , n},

∂
∂ x Fj u(t, x) ≤ εQ + εM + εAQ(ωM(1 + M) + 1),

l
|F0 u(t, x)| ≤ εR + εQ + εM + εAQ(nωM + 1)

!
n
2 m mj
+εM A + AQ + A ∑ Q ,
j=0

∂
F u(t, x) ≤ εQ + εM + εAQ(nωM + 1)
∂xj 0

!
n
2 m mj
+εM A + AQ + A ∑ Q ,
j=0
−Fu(t, x) ≥ εN − εR,
−Fj u(t, x) ≥ εN − εAQ − εR,
−F0 u(t, x) ≥ εN − εAQ − εR, (t, x) ∈ [0, ω] × Rn , l, j ∈ {1, . . . , n}.

Proof 5.3.4 We have
|Fu(t, x)| ≤ εR + εh(t, x)
e−[a]ω
Z ω R t+s
a(τ)dτ
+ε et (u0 (t + s, x) + a(t + s)u(t + s, x))ds
1 − e−[a]ω 0
≤ εR + εM + εMQω(1 + M),
e−[a]ω ω R t+sZ
∂ ∂
Fu(t, x) = −ε h(t, x) − ε e t a(τ)dτ
∂ xl ∂ xl 1 − e−[a]ω 0

∂ ∂
× u0 (t + s, x) + a(t + s) u(t + s, x) ds,
∂ xl ∂ xl
−[a]ω

≤ ε ∂ h(t, x) + ε e
∂ Z ω R
t+s
e t a(τ)dτ

Fu(t, x) −[a]ω
∂x
l
∂x
l
1−e 0

∂ ∂
× u0 (t + s, x) + a(t + s) u(t + s, x) ds
∂ xl ∂ xl
≤ εM + εMω(Q + MQ)
= εM + εωQM(1 + M),
e−[a j ]ω
Z ω R t+s
a j (τ)dτ
|Fj u(t, x)| ≤ εu(t, x) + εR + εh(t, x) + ε et
1 − e−[a j ]ω 0
Z x
∂ u0
× b(t, z) (t + s, z) + a(t + s)u j (t + s, z) dzds
−∞ ∂xj
Z x
+ b(t, z)u j (t, z)dz
−∞
≤ εQ + εM + εR + ε(MωA(Q + MQ) + AQ)
= εQ + εR + εAQ(ωM(1 + M) + 1) + εM,
−[a ]ω Z ω R
∂ ∂ ∂ e j t+s
Fj u(t, x) = −ε u j (t, x) − ε h(t, x) − ε −[a ]ω
e t a j (τ)dτ
∂ xl ∂ xl ∂ xl 1−e j 0
Z xˆl
∂ u0
× b(t, z˜xl ) (t + s, z˜xl ) + a(t + s)u j (t + s, z˜xl ) dzl ds
−∞ ∂xj
Z xˆl
l
− b(t, z˜xl )u j (t, z˜xl )dz ,
−∞
−[a ]ω Z ω R
∂ ∂ ∂ e j t+s
a j (τ)dτ
∂ x Fj u(t, x) ≤ ε ∂ x u j (t, x) + ε ∂ x h(t, x) + ε 1 − e−[a j ]ω 0 e
t
l l l
Z xˆl
∂ u0
(t + s, z˜xl ) + a(t + s)u j (t + s, z˜xl ) dzl ds

× b(t, z˜xl )
−∞ ∂xj
Z xˆl
l
+ b(t, z˜xl )u j (t, z˜xl )dz
−∞
≤ εQ + εM + εMωA(Q + MQ) + εAQ
= εQ + εM + εAQ(ωM(1 + M) + 1),
|F0 u(t, x)| ≤ εu0 (t, x) + εR + εh(t, x)
e−[a0 ]ω
n
Z x
Z ω R t+s
a0 (τ)dτ
∂uj
+ε e t b(t, z) ∑ (t + s, z)
1 − e−[a0 ]ω 0 −∞ j=1 ∂ x j

Z x
+ b(t, z)u0 (t, z)dz
−∞
≤ εQ + εR + εM
! !
n
2 m mj
+ε MωAnQ + M A + AQ + A ∑ Q + AQ
j=1
= εQ + εR + εM + εAQ(nωM + 1)
!
n
2 m mj
+εM A + AQ + A ∑ Q ,
j=0
∂ ∂ ∂
F0 u(t, x) = −ε u0 (t, x) − ε h(t, x)
∂ xl ∂ xl ∂ xl
−[a0 ]ω Z ω R Z xˆl n
e t+s
a0 (τ)dτ ∂uj
−ε e t b(t, z˜xl ) ∑ ∂ x j (t + s, z˜xl )
1 − e−[a0 ]ω 0 −∞ j=1

+ f (t + s, z˜xl , u(t + s, z˜xl ), u0 (t + s, z˜xl ), u1 (t + s, z˜xl ), . . . , un (t + s, z˜xl )) dzl ds
Z xˆl
− b(t, z˜xl )u0 (t, z˜xl )dzl ,
−∞

∂ ∂ ∂
∂ x F0 u(t, x)
≤ ε
u0 (t, x) + ε
h(t, x)
l ∂x l ∂x l
e−[a0 ]ω
Z xˆl n
Z ω R t+s
a0 (τ)dτ
∂uj
+ε e t b(t, z˜xl ) ∑ (t + s, z˜xl )
1 − e−[a0 ]ω 0 −∞ j=1 ∂ x j

+ f (t + s, z˜xl , u(t + s, z˜xl ), u0 (t + s, z˜xl ), u1 (t + s, z˜xl ), . . . , un (t + s, z˜xl )) dzl ds
Z xˆl
+ b(t, z˜xl )u0 (t, z˜xl )dzl
−∞
≤ εQ + εM
n
+ε M nωAQ + MA + MAQm + M ∑ AQmj
+ AQ
j=0
!
n
= εQ + εM + εAQ(nωM + 1) + εM 2 A + AQm + A ∑ Qm j ,
j=0
−Fu(t, x) ≥ εh(t, x) − εR
≥ εN − εR,
Z x
−Fj u(t, x) ≥ εh(t, x) − ε b(t, z)u j (t, z)dz − εR
−∞
≥ εN − εAQ − εR,
Z x
−F0 u(t, x) ≥ εh(t, x) − ε b(t, z)u0 (t, z)dz − εR
−∞
≥ εN − εAQ − εR, (t, x) ∈ [0, ω] × Rn .

Theorem 5.3.5 Suppose (5.34)-(5.36), (H1)-(H4). Then system (5.37) has at least three
solutions u1 , u2 , u3 ∈ P n+2 so that
0 < ku1 k ≤ r1 ≤ ku2 k < ρ < ku3 k ≤ r2 .
Proof 5.3.6 Define
PR = {x ∈ P : kxk < R},
PRn+2 = PR × . . . × PR ,
| {z }
n+2
Pρ = {x ∈ P : kxk < ρ},
Pρn+2 = Pρ × . . . × Pρ ,
| {z }
n+2
Pri = {x ∈ P : kxk < ri },
Prn+2
i
= Pri × . . . × Pri .
| {z }
n+2
1. T̃ : PRn+2 → E n+2 . Next, let
u = (u, u0 , u1 , . . . , un ) ∈ E n+2 ,
v = (v, v0 , v1 , . . . , vn ) ∈ E n+2 .
Then
kT u − T vk = (1 + ε)ku − vk,
kT j u − T j vk = (1 + ε)ku j − v j k, j ∈ {0, . . . , n}.
Hence,
kT̃ u − T̃ vk = (1 + ε)ku − vk,
i.e., T̃ : PRn+2 → E n+2 is 1 + ε-expansive operator. By Lemma 5.3.3, it follows

that

kF̃uk ≤ max εR + εωRM(1 + M) + εM,
2εR + εM + εAR(ωM(1 + M) + 1),
2εR + εM + εAR(nωM + 1)
!
n
2 n mj
+εM 1 + AR + A ∑ R kuk,
j=0
u = (u, u0 , u1 , . . . , un ) ∈ PRn+2 . As in the proof of Theorem 3.2.3, we get that

F : PRn+2 → E n+2 is a 0-set contraction.
2. For i = 1, 2, u ∈ ∂ Prn+2
i
, using Lemma 5.3.3, we have

kF̃u + T 0k ≤ max εR + εωri M(1 + M) + εM,
2εR + εM + εAri (ωM(1 + M) + 1),
2εR + εM + εAri (nωM + 1)

!
n
2 m
+εM 1 + Arin + A ∑ ri j kuk
j=0
≤ εskuk
< εkuk.

3. Now we will check condition (b) of Theorem 1.9.21. Let u0 ∈ P
n+2
and
n+2
assume that there exists a u ∈ P
1 n+2
and a λ1 ≥ 0 with u −λ1 u ∈ ∂ PR
1 0
such that
F̃u1 = (I − T )(u1 − λ1 u0 ).
Hence,
0 < εN − εAQ − εR
≤ −F̃u1 (t, x)
= −(I − T̃ )(u1 − λ1 u0 )(t, x)
= ε(u1 − λ1 u0 )(t, x) − εR + εh(t, x)
≤ ερ − εR + εM
< 0.


F̃ PRn+2 ⊂ (I − T̃ ) PRn+2 .
By Theorem 1.9.21, it follows that system (5.37) has at least three solutions u1 , u2 , u3 ∈
P n+2 so that
0 < ku1 k ≤ r1 ≤ ku2 k < ρ < ku3 k ≤ r2 .
6
Applications to Elliptic Equations
6.1 Multiple Solutions for a BVP with Robin Boundary Conditions

Let the positive constants ε, m, k, s, A, B0 , R, r1 , r2 , ρ and a ≥ 0 satisfy the following condi-
tions
1
0 < r1 < ρ < r2 < R, 0 < s < , (6.1)
2
R + 3A + aR + ARm + ARk < s, ρ + A < 2B0 , (6.2)
3A + ari + Arim + Arik < ri , B0 > R. (6.3)

Consider the BVP
0 = ∆u + f (x, u), x ∈ Ω,
(6.4)
∂u
= au + g on ∂ Ω,
∂n
Z
dσ ≤ A,
∂Ω
and dσ is a vector element of ∂ Ω,

(H1) f : Ω × R → [0, ∞), f ∈ C (Ω × R),
f (x, u) ≤ a0 (x) + a1 (x)um + a2 (x)uk , x ∈ Ω, u ∈ [0, ∞),
a j ∈ C (Ω), j = 0, 1, 2,
Z
a j (x)dx ≤ A, j = 0, 1, 2,
Ω
(H2) g : ∂ Ω → [0, ∞), g ∈ C (∂ Ω),

Z
B0 ≤ g(x)dσ ≤ A,
∂Ω
257
258 Applications to Elliptic Equations
(H3) φ ∈ C (Ω), B0 ≤ φ (x) ≤ A, x ∈ Ω.
Definition 6.1.1 We say that u ∈ C (Ω) is a solution of BVP (6.4), if it satisfies the integral
equation
Z Z Z
0 = a u(τ)dσ + g(x)dσ + f (y, u(y))dy, x ∈ Ω.
∂Ω ∂Ω Ω
Let E = C (Ω) be endowed with the norm
kuk = max |u(x)|.

x∈Ω
Define
P
f = {u ∈ E : u(x) ≥ 0, x ∈ Ω}.
f
Tu(x) = (1 + ε)u(x) − εR + εφ (x),

Z Z
Fu(x) = εR − ε φ (x) + a u(x)dσ + g(x)dσ
∂Ω ∂Ω
Z
+ f (y, u(y))dy , x ∈ Ω.
Ω
Note that any fixed point u ∈ E of the operator T + F is a solution of BVP (6.4).
Theorem 6.1.2 Suppose (6.1)-(6.3), (H1)-(H3). Then BVP (6.4) has at least three solu-
tions u1 , u2 , u3 ∈ P such that
0 < ku1 k ≤ r1 ≤ ku2 k < ρ < ku3 k ≤ r2 .
PR = {u ∈ P : kuk < R},
Pρ = {u ∈ P : kuk < ρ},
Pri = {u ∈ P : kuk < ri }, i = 1, 2.
kTu − T vk = (1 + ε)ku − vk.

Multiple Solutions for a BVP with Robin Boundary Conditions 259
Z
|Fu(x)| ≤ εR + εφ (x) + εa u(x)dσ
∂Ω
Z Z
+ε g(x)dσ + ε f (y, u(y))dy
∂Ω Ω
≤ εR + εA + εaRA + εA
Z Z
+ε a0 (y)dy + ε a1 (y)(u(y))m dy
Ω Ω (6.5)
Z
+ε a2 (y)(u(y))k dy
Ω
≤ εR + εA + εaRA + εA + εA + εARm + εARk
= ε(R + 3A + aRA + ARm + ARk )
< sε, x ∈ Ω.

tion.
|Fu(x) + T 0(x)| ≤ ε(3A + ari A + Arim + Arik )
< εri , x ∈ Ω.

Fu1 = (I − T )(u1 − λ1 u0 ).
Then
Z
−εR + 2εB0 ≤ −εR + εφ (x) + ε g(x)dσ
∂Ω
≤ −Fu1 (x)
= −(I − T )(u1 − λ1 u0 )(x)

= ε(u1 (x) − λ1 u0 (x)) − εR + εφ (x)
≤ ε(ρ + A − R), x ∈ Ω,
whereupon
2B0 ≤ ρ + A.
F(PR ) ⊂ (I − T )(PR ). (6.6)
u(x) = Tu(x) + Fz(x)
= Au(x), x ∈ Ω.
= (1 + ε)ku − vk,
PR ⊂ A(PR ). (6.7)
Au(x) = v(x) or Tu(x) = v(x) − Fz(x), x ∈ Ω. (6.8)
We have
Z
v(x) − Fz(x) = v(x) − εR + ε φ (x) + a u(x)dσ
∂Ω
Z Z
+ g(x)dσ + f (y, u(y))dy
∂Ω Ω
≥ v(x) − εR + εφ (x)
≥ v(x) − εR + εB0
≥ 0, x ∈ Ω,
and
v(x) − Fz(x) ≤ R + ε(R + 3A + aRA + ARm + ARk )
= R1 , x ∈ Ω.
P, and
\
T : PR → B(R1 ) P
\
is a bijection. Thus, there is a u ∈ PR which satisfies (6.8) and then (6.7) holds.
Hence, and applying Theorem 1.9.8, it follows that there exists a w ∈ PR so
that
w(x) = Aw(x), x ∈ Ω,
and (6.6) holds.
By Theorem 1.9.21, it follows that BVP (6.4) has at least three solutions u1 , u2 , u3 ∈ P so
that
0 < ku1 k ≤ r1 ≤ ku2 k < ρ < ku3 k ≤ r2 .
Let the positive constants ε, A, B0 , r, L, R, m, k and a ≥ 0 satisfy the following conditions
B0 > R, r < L < R, (6.9)
R < εR − ε(3A + aRA + ARm + ARk ), (6.10)
3A + aLA + ALm + ALk ≤ L. (6.11)
Theorem 6.1.4 Suppose (6.9)-(6.11), (H1)-(H3). Then BVP (6.4) has at least two solu-
tions u1 , u2 ∈ P such that
r < ku1 k < L < ku2 k < R.
Proof 6.1.5 Define
Pr = {x ∈ P : kxk < r},
PL = {x ∈ P : kxk < L},

Pr,L = {x ∈ P : r < kxk < L},
PL,R = {x ∈ P : L < kxk < R}.
We have
Pr,L PR
\
6= 0,
/
PL,R PR
\
6= 0.
/

F(PR ) ⊂ (I − T )(PR ).
2. Let u0 ∈ P ∗ and assume that there are a λ1 ≥ 0 and a u1 ∈ ∂ Pr such that
Fu1 = u1 − λ1 u0 .
r ≥ u1 (x)
≥ u1 (x) − λ1 u0 (x)
= Fu1 (x)
≥ εR − ε(3A + arA + Arm + Ark ), x ∈ Ω.

|Fu(x) + T 0(x)| ≤ ε(3A + aLA + ALm + ALk )
≤ εL
= εkuk, x ∈ Ω.
Hence,
Now, assume that there is a u1 ∈ ∂ PL such that
u1 = Tu1 + Fu1 .
Let x1 ∈ Ω be such that

L = u1 (x1 ).
Then
L = u1 (x1 )
= Tu1 (x1 ) + Fu1 (x1 )
= (1 + ε)u1 (x1 ) − εR + εφ (x)

Z
+εR − ε φ (x) + a u(x)dσ
∂Ω
Z Z
+ g(x)dσ + f (y, u(y))dy
∂Ω Ω
Z
≤ εR − ε g(x)dσ
∂Ω
≤ εR − εB0
< 0, x ∈ Ω.

Hence, and applying Theorem 1.9.23, it follows that BVP (6.4) has at least two solutions
r < ku1 k < L < ku2 k < R.
6.2 Existence and Smoothness of Navier-Stokes Equations

In this section we investigate the following Navier-Stokes equations
n

∂ ui ∂ ui ∂p
+ uj = ν∆ui − + fi (t, x), i ∈ {1, . . . , n},



 ∂t ∑ ∂ x j ∂ xi
j=1


(6.12)
n
∂ ui


x ∈ Rn , t ≥ 0,



 ∑ ∂ xi = 0,
i=1
subject to the initial condition
u(0, x) = u0 (x), x ∈ Rn , (6.13)
where
 ui0 ∈ C0 (R ),
n
 ∞
suppui0 ⊂ B,
(H1)
|∂xα ui0 (x)| ≤ Cα,K (1 + |x|)−K on Rn , i ∈ {1, . . . , n},

for any multi-index α and for any positive constant K,
 fi ∈ C ([0, ∞), C0 (R )) , suppx fi ⊂ B,

n
 ∞ ∞
(H2)
|∂t ∂x fi (t, x)| ≤ CαmK (1 + |x| + t)−K on [0, ∞) × Rn ,
 m α
i ∈ {1, . . . , n},
for any α, m, K, u = (u1 , . . . , un ) and p are unknown, ν > 0 is a constant, n ≥ 3. Here B is a

compact subset of Rn and ∂xα = ∂xα11 . . . ∂xαnn , α = (α1 , . . . , αn ), αi ∈ N0 , i ∈ {1, . . . , n}, and
n
∂2
∆ = ∑ 2 is the Laplacian in the space variables.
i=1 ∂ xi
By the last equation of system (6.12), we get
n
∂ ui
uj ∑ =0 for any j ∈ {1, . . . , n}.
i=1 ∂ xi
Then
n n n
∂ ui ∂ ∂uj
∑ uj = ∑ ∂xj (u j ui ) − ui ∑ ∂xj
j=1 ∂xj j=1 j=1
n
∂
= ∑ ∂ x j (u j ui ) .
j=1
Existence and Smoothness of Navier-Stokes Equations 265
From here, it follows that we can rewrite system (6.12) in the following form
n n

∂ ui ∂ ∂2 ∂p
+∑ (ui u j ) − ν ∑ 2 ui + = fi (t, x), i ∈ {1, . . . , n},





 ∂t j=1 ∂ x j j=1 ∂ x j ∂ xi
(6.14)
n
∂ ui


n
(t, x) ∈ (0, ∞) × R .



 ∑ ∂ xi = 0,
i=1
Lemma 6.2.1 Let x0 = (x10 , . . . , xn0 ) ∈ ∂ B and fi satisfies (H2), i ∈ {1, . . . , n}. If p and
ui ∈ C 1 ([a, b], C02 (Rn )), suppx ui ⊂ B, suppx p ⊂ B, i ∈ {1, . . . , n}, satisfy the system
 Z xZ s
 0 0 (ui (t, σ ) − gi (σ )) dσ ds




 x x


n Z t Z x Z sj



+ ui τ, σ̃s j u j τ, σ̃s j dσ j dsdτ


 ∑ 0 0
j=1 a x x j









 n Z t Z xj Z sj
−ν ∑ ui τ, σ̃x j dσ j ds j dτ



 0 0
j=1 a x j xj


(6.15)

 Z t Z x Z si
+ p (τ, σ̃si ) dσ i dsdτ



a x0 x0i







 Z tZ xZ s

fi (τ, σ )dσ dsdτ, i ∈ {1, . . . , n},


 =
a x0 x0






n Z t Z x Z sj


n

 ∑ a x0 x0 u j τ, σ̃s j dσ j dsdτ = 0, t ∈ [a, b], x ∈ R ,



j=1 j
then p and ui , i ∈ {1, . . . , n}, satisfy the problem

n n

∂ ui ∂ ∂2 ∂p
+∑ (ui u j ) − ν ∑ 2 ui + = fi (t, x), i ∈ {1, . . . , n},



 ∂t

 j=1 ∂ x j j=1 ∂ x j ∂ xi
(6.16)
n
∂ ui


(t, x) ∈ [a, b] × Rn ,



 ∑ ∂ xi = 0,
i=1
u(a, x) = g(x), g(x) = (g1 (x), . . . , gn (x)), x ∈ Rn (6.17)
where gi ∈ C02 (Rn ), suppgi ⊂ B, i ∈ {1, . . . , n}.
Here
Z x Z x1 Z xn
= ... ,
x0 x10 xn0
dσ = dσn . . . dσ1 ,
ds = dsn . . . ds1 ,
Z sj Z s1 Z s j−1 Z s j+1 Z sn
= ... ... ,
x0j x10 x0j−1 x0j+1 xn0
Z xj Z x1 Z x j−1 Z x j+1 Z xn
= ... ... ,
x0j x10 x0j−1 x0j+1 xn0

σ̃s j = σ1 , . . . , σ j−1 , s j , σ j+1 , . . . , σn ,
dσ j = dσn . . . dσ j+1 dσ j−1 . . . dσ1 ,

σ̃x j = σ1 , . . . , σ j−1 , x j , σ j+1 , . . . , σn ,
ds j = dsn . . . ds j+1 ds j−1 . . . ds1 ,
σ̃si = (σ1 , . . . , σi−1 , si , σi+1 , . . . , σn ) , i, j ∈ {1, . . . , n}.
Proof 6.2.2 We differentiate once in t and twice x1 , . . ., xn , all equations of system (6.15)
and we see that the function u satisfies the system (6.16). Now we put t = a in the first n
equations of system (6.15) and we get
Z xZ s
(ui (a, σ ) − gi (σ )) dσ ds = 0, i ∈ {1, . . . , n}.
x0 x0
We differentiate the last system twice in x1 , . . ., xn , and we obtain
ui (a, x) = gi (x), x ∈ Rn , i ∈ {1, . . . , n},
i.e., the function ui , i ∈ {1, . . . , n}, satisfies the initial condition (6.17). This completes the
proof.
Below we will suppose that x0 = x10 , . . . , xn0 ∈ ∂ B is arbitrarily chosen and fixed. Also, we

will use the following notations

Z s j,m Z s1 Z s j−1 Z s j+1 Z sm−1 Z sm+1 Z sn
= ... ... ... ,
x0j,m x10 x0j−1 x0j+1 0
xm−1 0
xm+1 xn0
Z s j,m,r Z s1 Z s j−1 Z s j+1 Z sm−1 Z sm+1 Z sr−1 Z sr+1 Z sn
= ... ... ... ... ,
x0j,m,r x10 x0j−1 x0j+1 0
xm−1 0
xm+1 0
xr−1 0
xr+1 xn0

σ̃s j ,sm = σ1 , . . . , σ j−1 , s j , σ j+1 , . . . , σm−1 , sm , σm+1 , . . . , σn ,
σ̃si ,sm ,sr = (σ1 , . . . , σi−1 , si , σi+1 , . . . , σm−1 , sm , σm+1 , . . . , σr−1 , sr , σr+1 , . . . , σn ),
i, j, m, r ∈ {1, . . . , n}.
With µ(A) we will denote the measure of a set A ⊂ Rk , k ≥ 1. Let

n
B∗ = max 1, µ(B), µ B ∩ Rx1 × . . . × Rxk −1 × Rxk +1 × . . . × Rxn ,
1 1

µ B ∩ Rx1 × . . . × Rxk −1
× Rxk +1
× . . . × Rxk −1
× Rxk +1
× . . . × Rxn ,
1 1 2 2

µ B∩ Rx1 × . . . × Rxk × Rxk × . . . × Rxk × Rxk
1 −1 1 +1 2 −1 2 +1

× . . . × Rxk −1
× Rxk +1
× . . . × Rxn ,
3 3
o
k1 , k2 , k3 ∈ {1, . . . , n} , Rn = Rx1 × · · · × Rxn ,
as we set
µ B∩ Rx1 × . . . × Rxk × Rxk × . . . × Rxk × Rxk
1 −1 1 +1 2 −1 2 +1

× . . . × Rxk −1
× Rxk +1
× . . . × Rxn := 1,
3 3
k1 , k2 , k3 ∈ {1, 2, 3} for n = 3.
Theorem 6.2.3 Suppose that ui0 satisfies (H1) and fi satisfies (H2), i ∈ {1, . . . , n}. Then
the Cauchy problem (6.12), (6.13) has a solution p, ui ∈ C ∞ ([0, ∞) × Rn ), i ∈ {1, . . . , n},
such that Z
|u(t, x)|2 dx ≤ C for all t ≥ 0,
Rn
q
where |u(t, x)| = (u1 (t, x))2 + · · · + (un (t, x))2 .
Proof 6.2.4 Firstly we will prove that the Cauchy problem

n

∂ ui ∂ ui ∂p
+ uj = ν∆ui − + fi (t, x), i ∈ {1, . . . , n},



 ∂t
 ∑ ∂xj ∂ xi
j=1

(6.18)
n
∂ ui


∑ ∂ xi (t, x) = 0 in [0, 1] × Rn ,




i=1
u(0, x) = u0 (x) in Rn , (6.19)

has a C 1
([0, 1], C02 (Rn ))-solution 1 1
(u , p ) such that
suppx u1i , suppx p1 ⊂ B, i ∈ {1, . . . , n}.
Let Q > 0 be chosen such that

|u0 (x)| ≤ Q, u0ix ≤ Q,
l
u0ix x ≤ Q,
l m
x ∈ B,
i, l, m ∈ {1, . . . , n}, and

M1 = max | f (t, x)|.
t∈[0,1],x∈B
We choose the constant l > 0 such that
ln Q + 3n(1 + ν)(B∗ )2 Q + n2 Q2 (B∗ )2 + (B∗ )2 M 1 ≤ Q.

(6.20)
Let n
E1 = v = (v1 , . . . , vn+1 ) : vi ∈ C 1 ([0, 1], C02 (Rn )), suppx vi ⊂ B,
o
i ∈ {1, . . . , n + 1}
be endowed with the norm
n
||v|| = max max |vq (t, x)|, max |vqt (t, x)|,
q∈{1,...,n+1} t∈[0,1],x∈B t∈[0,1],x∈B
max |vqxi (t, x)|, max |vqxi x j (t, x)|,

t∈[0,1],x∈B t∈[0,1],x∈B
o
i, j ∈ {1, . . . , n} .
With Ke1 we denote the set of all equi-continuous families in E 1 , i.e., if F ⊂ K

e1 is a family
1
of elements of E , then for every ε > 0 there exists δ = δ (ε) > 0 such that
|vq (t1 , x1 ) − vq (t2 , x2 )| < ε, |vqt (t1 , x1 ) − vqt (t2 , x2 )| < ε,
|vqxi (t1 , x1 ) − vqxi (t2 , x2 )| < ε, |vqxi x j (t1 , x1 ) − vqxi x j (t2 , x2 )| < ε
for any i, j ∈ {1, . . . , n}, for any q ∈ {1, . . . , n + 1}, and for any v ∈ F , whenever |t1 −t2 | <
δ , |x1 − x2 | < δ . Let also
1 1
e1 ,
K =K K 1 = {v ∈ K : ||v|| ≤ Q},
1
Q1 = {v ∈ K : ||v|| ≤ (1 + l)Q}.
Here Ke1 is the closure of Ke1 . Here, by Arzela’s Theorem, we have that K 1 is a compact
subset of Q . For (u, p) ∈ Q1 , we define the operators.
1
Z xZ s
Li1 (u, p)(t, x) = −lui (t, x) + l (ui (t, σ ) − ui0 (σ )) dσ ds
x0 x0
n Z t Z x Z sj
+l ∑ ui τ, σ̃s j u j τ, σ̃s j dσ j dsdτ
j=1 0 x0 x0j
n Z t Z xj Z sj
−νl ∑ ui τ, σ̃x j dσ j ds j dτ
j=1 0 x0j x0j
Z t Z x Z si
+l p (τ, σ̃si ) dσ i dsdτ
0 x0 x0i
Z tZ xZ s
−l f (τ, σ )dσ dsdτ,
0 x0 x0
n Z t Z x Z sj
1

Ln+1 (u, p)(t, x) = −l p(t, x) + l ∑ u j τ, σ̃s j dσ j dsdτ,
j=1 0 x0 x0j
Mi1 (u, p)(t, x) = (1 + l)ui (t, x), i ∈ {1, . . . , n},
1
Mn+1 (u, p)(t, x) = (1 + l)p(t, x),
L1 (u, p)(t, x) = L11 (u, p)(t, x), . . . , Ln+1

1

(u, p)(t, x) ,
1
M11 (u, p)(t, x), . . . , Mn+1
1

M (u, p)(t, x) = (u, p)(t, x) ,
(t, x) ∈ [0, 1] × Rn .
Proposition 6.2.5 L1 : K 1 → K 1 is continuous and L1 (K 1 ) resides in a compact subset of

Q1 .
Proof 6.2.6 1. For (u, p) ∈ K 1 we have that Li1 (u, p) ∈ C 1 ([0, 1], C 2 (Rn )),
i ∈ {1, . . . , n + 1} and for x ∈ Rn \B, using the definition of L1 , we have that
L1 (u, p)(t, x) = 0 for all t ∈ [0, 1]. Hence, Li1 (u, p) ∈ C 1 ([0, 1], C02 (Rn )) and
suppx Li1 (u, p) ⊂ B, i ∈ {1, . . . , n + 1}.
2. Let (u, p) ∈ K 1 . Then
∂ 1 ∂
L (u, p)(t, x) = −l ui (t, x)
∂t i ∂t
Z xZ s
∂
+l ui (t, σ )dσ ds
x0 x0 ∂t
n Z x Z sj
+l ∑ ui t, σ̃s j u j t, σ̃s j dσ j ds
j=1 x0 x0j
n Z xj Z sj
−lν ∑ ui t, σ̃x j dσ j ds j
j=1 x0j x0j
Z x Z si
+l p (t, σ̃si ) dσ i ds
x0 x0i
Z xZ s
−l f (t, σ )dσ ds, i ∈ {1, . . . , n}, (t, x) ∈ [0, 1] × Rn .
x0 x0
Hence, using (6.20),

∂ 1
Li (u, p)(t, x) ≤ l ∂ ui (t, x)

∂t ∂t
Z x Z s
∂ ui (t, σ ) dσ ds

+l
x0 x0 ∂t

n Z x Z sj
+l ∑ ui t, σ̃s u j t, σ̃s dσ j ds
j j
j=1 x0 x0j
n Z xj Z sj
+lν ∑ ui t, σ̃x dσ j ds j
j
j=1 x0j x0j
Z x Z si
+l |p (t, σ̃si )| dσ i ds
x0 x0i
Z xZ s
+l | f (t, σ )| dσ ds
x0 x0
≤ lQ + l(B∗ )2 Q + nl(B∗ )2 Q2 + νln(B∗ )2 Q + l(B∗ )2 Q + l(B∗ )2 M 1
= l Q + n(B∗ )2 Q2 + (νn + 2)(B∗ )2 Q + (B∗ )2 M 1

≤ Q, i ∈ {1, . . . , n}, (t, x) ∈ [0, 1] × Rn .
3. Let (u, p) ∈ K 1 . Then

∂ 1 ∂
L (u, p)(t, x) = −l p(t, x)
∂t n+1 ∂t
n Z x Z sj
(t, x) ∈ [0, 1] × Rn .

+l ∑ u j t, σ̃s j dσ j ds,
j=1 x
0 x0j

∂ 1
Ln+1 (u, p)(t, x) ≤ l ∂ |p(t, x)|

∂t ∂t
n Z x Z sj
+l ∑ u j t, σ̃s dσ j ds
j
j=1 x0 x0j
≤ lQ + ln(B∗ )2 Q
= l Q + n(B∗ )2 Q

≤
Q, (t, x) ∈ [0, 1] × Rn .
4. Let (u, p) ∈ K 1 and m 6= i, m, i ∈ {1, . . . , n}. Then

∂ 1 ∂
Li (u, p)(t, x) = −l ui (t, x)
∂ xm ∂ xm
Z x Z sm
+l (ui (t, σ̃sm ) − ui0 (σ̃sm )) dσ m ds
x0 x0m
n Z t Z x Z sm, j
+l ∑ ui τ, σ̃sm ,s j u j τ, σ̃sm ,s j dσ m, j dsdτ
j=1, j6=m 0 x0 x0m, j
Z t Z x m Z sm
+l ui (τ, σ̃xm ) um (τ, σ̃xm ) dσ m dsm dτ
0 x0m x0m
n Z t Z x j Z sm, j
−νl ∑ ui τ, σ̃sm ,x j dσ j,m ds j dτ
j=1, j6=m 0 x0j x0m, j
Z t Z x m Z sm
∂
−νl ui (τ, σ̃xm ) dσ m dsm dτ
0 x0m x0m ∂ xm
Z t Z x Z sm,i
+l p (τ, σ̃sm ,si ) dσ i,m dsdτ
0 x0 x0m,i
Z t Z x Z sm
−l f (τ, σ̃sm ) dσ m dsdτ, (t, x) ∈ [0, 1] × Rn .
0 x0 x0m

∂ 1 ∂
∂ xm Li (u, p)(t, x) ≤ l ∂ xm |ui (t, x)|

Z x Z sm
+l |ui (t, σ̃sm ) − ui0 (σ̃sm )| dσ m ds
x0 x0m
n Z t Z x Z sm, j
+l ∑ ui τ, σ̃s ,s u j τ, σ̃s ,s dσ m, j dsdτ
m j m j
j=1, j6=m 0 x0 x0m, j
Z t Z x m Z sm
+l |ui (τ, σ̃xm ) um (τ, σ̃xm )| dσ m dsm dτ
0 x0m x0m
n Z t Z x j Z sm, j
+νl ∑ ui τ, σ̃s ,x dσ j,m ds j dτ
m j
j=1, j6=m 0 x0j x0m, j
Z t Z xm Z sm
∂

+νl ∂ xm ui (τ, σ̃xm ) dσ m dsm dτ
0 x0m x0m
Z t Z x Z sm,i
+l |p (τ, σ̃sm ,si )| dσ i,m dsdτ
0 x0 x0m,i
Z t Z x Z sm
+l | f (τ, σ̃sm )| dσ m dsdτ
0 x0 x0m
≤ lQ + 2l(B∗ )2 Q + (n − 1)l(B∗ )2 Q2 + l(B∗ )2 Q2
+νl(n − 1)(B∗ )2 Q + νl(B∗ )2 Q + l(B∗ )2 Q + l(B∗ )2 M 1

= l Q + (3 + νn)(B∗ )2 Q + n(B∗ )2 Q2 + (B∗ )2 M 1

≤ Q, (t, x) ∈ [0, 1] × Rn .
5. Let (u, p) ∈ K 1 . Then
∂ 1 ∂
L (u, p)(t, x) = −l ui (t, x)
∂ xi i ∂ xi
Z x Z si
+l (ui (t, σ̃si ) − ui0 (σ̃si )) dσ i ds
x0 x0i
n Z t Z x Z si, j
+l ∑ ui τ, σ̃si ,s j u j τ, σ̃si ,s j dσ i, j dsdτ
j=1, j6=i 0 x0 x0i, j
Z t Z x i Z si
+l (ui (τ, σ̃xi ))2 dσ i dsi dτ
0 x0i x0i
n Z t Z x j Z si, j
−νl ∑ ui τ, σ̃si ,x j dσ i, j ds j dτ
j=1, j6=i 0 x0j x0i, j
Z t Z x i Z si
∂
−νl ui (τ, σ̃xi ) dσ i dsi dτ
0 x0i x0i ∂ xi
Z t Z x i Z si
+l p (τ, σ̃xi ) dσ i dsi dτ
0 x0i x0i
Z t Z x Z si
−l f (τ, σ̃si ) dσ i dsdτ, (t, x) ∈ [0, 1] × Rn .
0 x0 x0i

∂ 1 ∂
∂ xi Li (u, p)(t, x) ≤ l ∂ xi ui (t, x)

Z x Z si
+l |ui (t, σ̃si ) − ui0 (σ̃si )| dσ i ds
x0 x0i
n Z t Z x Z si, j
+l ∑ ui τ, σ̃s ,s u j τ, σ̃s ,s dσ i, j dsdτ
i j i j
j=1, j6=i 0 x0 x0i, j
Z t Z x i Z si
+l |ui (τ, σ̃xi )|2 dσ i dsi dτ
0 x0i x0i
n Z t Z x j Z si, j
+νl ∑ ui τ, σ̃s ,x dσ i, j ds j dτ
i j
j=1, j6=i 0 x0j x0i, j
Z t Z xi Z si
∂

+νl ∂ xi ui (τ, σ̃xi ) dσ i dsi dτ
0 x0i x0i
Z t Z x i Z si
+l |p (τ, σ̃xi )| dσ i dsi dτ
0 x0i x0i
Z t Z x Z si
+l | f (τ, σ̃si )| dσ i dsdτ
0 x0 x0i
≤ lQ + 2l(B∗ )2 Q + (n − 1)l(B∗ )2 Q2 + l(B∗ )2 Q2
= l Q + (3 + νn)(B∗ )2 Q + n(B∗ )2 Q2 + (B∗ )2 M 1

≤ Q, (t, x) ∈ [0, 1] × Rn .
6. Let (u, p) ∈ K 1 . Then

∂ 1 ∂
L (u, p)(t, x) = −l p(t, x)
∂ xm n+1 ∂ xm
n Z t Z x Z sm, j
+l ∑ u j τ, σ̃sm ,s j dσ j,m dsdτ
j=1, j6=m 0 x0 x0m, j
Z t Z x m Z sm
+l um (τ, σ̃xm ) dσ m dsm dτ, (t, x) ∈ [0, 1] × Rn .
0 x0m x0m

∂ 1 ∂
∂ xm Ln+1 (u, p)(t, x) ≤ l ∂ xm p(t, x)

n Z t Z x Z sm, j
+l ∑ u j τ, σ̃s ,s dσ j,m dsdτ
m j
j=1, j6=m 0 x0 x0m, j
Z t Z x m Z sm
+l |um (τ, σ̃xm )| dσ m dsm dτ
0 x0m x0m
≤ lQ + l(n − 1)(B∗ )2 Q + l(B∗ )2 Q

≤ l Q + n(B∗ )2 Q

≤ Q, (t, x) ∈ [0, 1] × Rn .
∂2 1 ∂2
2
Li (u, p)(t, x) = −l 2 ui (t, x)
∂ xm ∂ xm
Z x m Z sm
+l (ui (t, σ̃xm ) − ui0 (σ̃xm )) dσ m dsm
x0m x0m
n Z t Z xm Z sm, j
+l ∑ ui τ, σ̃xm ,s j u j τ, σ̃xm ,s j dσ j,m dsm dτ
j=1, j6=m 0 x0m x0m, j
Z t Z x m Z sm
∂
+l (ui (τ, σ̃xm ) u j (τ, σ̃xm )) dσ m dsm dτ
0 x0m x0m ∂ xm
n Z t Z xm, j Z sm, j
−νl ∑ ui τ, σ̃xm ,x j dσ j,m ds j,m dτ
j=1, j6=m 0 x0m, j x0m, j
Z t Z x m Z sm 2
∂
−νl 2
ui (τ, σ̃xm ) dσ m dsm dτ
0 x0m x0m ∂ xm
Z t Z xm Z sm,i
+l p (τ, σ̃xm ,si ) dσ m,i dsm dτ
0 x0m x0m,i
Z t Z x m Z sm
−l f (τ, σ̃xm ) dσ m dsm dτ, (t, x) ∈ [0, 1] × Rn .
0 x0m x0m

2 2
∂ 1
∂
∂ x2 Li (u, p)(t, x) ≤ l ∂ x2 ui (t, x)

m m
Z x m Z sm
+l |ui (t, σ̃xm ) − ui0 (σ̃xm )| dσ m dsm
x0m x0m
n Z t Z xm Z sm, j
+l ∑ ui τ, σ̃x ,s u j τ, σ̃x ,s dσ j,m dsm dτ
m j m j
j=1, j6=m 0 x0m x0m, j
Z t Z xm Z sm
∂

+l ∂ xm (ui (τ, σ̃xm ) u j (τ, σ̃xm )) dσ m dsm dτ
0 x0m x0m
n Z t Z xm, j Z sm, j
+νl ∑ ui τ, σ̃x ,x dσ j,m ds j,m dτ
m j
j=1, j6=m 0 x0m, j x0m, j
Z t Z xm Z sm 2
∂

+νl ∂ x2 ui (τ, σ̃xm ) dσ m dsm dτ
0 x0m x0m m
Z t Z xm Z sm,i
+l |p (τ, σ̃xm ,si )| dσ m,i dsm dτ
0 x0m x0m,i
Z t Z xm Z sm
+l | f (τ, σ̃xm )| dσ m dsm dτ
0 x0m x0m
≤ lQ + 2l(B∗ )2 Q + (n − 1)l(B∗ )2 Q2 + 2l(B∗ )2 Q2
= l Q + (3 + νn)(B∗ )2 Q + (n + 1)(B∗ )2 Q2 + (B∗ )2 M 1

≤ Q, (t, x) ∈ [0, 1] × Rn .
8. Let (u, p) ∈ K 1 . Then
∂2 1 ∂2
L (u, p)(t, x) = −l ui (t, x)
∂ xi2 i ∂ xi2
Z x i Z si
+l (ui (t, σ̃xi ) − ui0 (σ̃xi )) dσ i dsi
x0i x0i
n Z t Z xi Z si, j
+ ∑ l ui τ, σ̃xi ,s j u j τ, σ̃xi ,s j dσ i, j dsi dτ
j=1, j6=i 0 x0i x0i, j
Z t Z x i Z si
∂
+2l ui (τ, σ̃xi ) ui (τ, σ̃xi ) dσ i dsi dτ
0 x0i x0i ∂ xi
n Z t Z xi, j Z si, j
−νl ∑ ui τ, σ̃xi ,x j dσ i, j dsi, j dτ
j=1, j6=i 0 x0i, j x0i, j
Z t Z x i Z si 2
∂
−νl ui (τ, σ̃xi ) dσ i dsi dτ
0 x0i x0i ∂ xi2
Z t Z x i Z si
∂
+l p (τ, σ̃xi ) dσ i dsi dτ
0 x0i x0i ∂ xi
Z t Z x i Z si
−l f (τ, .σ̃xi ) dσ i dsi dτ, (t, x) ∈ [0, 1] × Rn .
0 x0i x0i

2 2
∂ 1 ∂
∂ x2 Li (u, p)(t, x) ≤ l ∂ x2 ui (t, x)

i i
Z x i Z si
+l |ui (t, σ̃xi ) − ui0 (σ̃xi )| dσ i dsi
x0i x0i
n Z t Z xi Z si, j
+ ∑ l ui τ, σ̃x ,s u j τ, σ̃x ,s dσ i, j dsi dτ
i j i j
j=1, j6=i 0 x0i x0i, j
Z t Z xi Z si
ui (τ, σ̃x ) ∂ ui (τ, σ̃x ) dσ i dsi dτ

+2l i i
0 x0i x0i
∂ xi
n Z t Z xi, j Z si, j
+νl ∑ ui τ, σ̃x ,x dσ i, j dsi, j dτ
i j
j=1, j6=i 0 x0i, j x0i, j
Z t Z xi Z si 2
∂

+νl ∂x 2
ui (τ, σ̃xi ) dσ i dsi dτ
0 x0i x0i

i
Z t Z xi Z si
∂

+l ∂ xi p (τ, σ̃xi ) dσ i dsi dτ
0 x0i x0i
Z t Z x i Z si
+l | f (τ, .σ̃xi )| dσ i dsi dτ
0 x0i x0i
≤ lQ + 2l(B∗ )2 Q + (n − 1)l(B∗ )2 Q2 + 2l(B∗ )2 Q2
= l Q + (3 + νn)(B∗ )2 Q + (n + 1)(B∗ )2 Q2 + (B∗ )2 M 1

≤ Q, , (t, x) ∈ [0, 1] × Rn ,
9. Let (u, p) ∈ K 1 and r 6= m, m 6= i, r 6= i, r, m, i ∈ {1, . . . , n}. Then

∂2 ∂2
Li1 (u, p)(t, x) = −l ui (t, x)
∂ xr ∂ xm ∂ xm ∂ xr
Z x Z sr,m
+l (ui (t, σ̃sr ,sm ) − ui0 (σ̃sr ,sm )) dσ r,m ds
x0 x0r,m
n Z t Z x Z sr,m, j
+l ∑ ui τ, σ̃sr ,sm ,s j u j τ, σ̃sr ,sm ,s j dσ r,m, j dsdτ
j=1, j6=m, j6=r 0 x0 x0r,m, j
Z t Z xr Z sr,m
+l ui (τ, σ̃xr ,sm ) ur (τ, σ̃xr ,sm ) dσ r,m dsr dτ
0 x0r x0r,m
Z t Z xm Z sr,m
+l ui (τ, σ̃sr ,xm ) um (τ, σ̃sr ,xm ) dσ r,m dsm dτ
0 x0m x0r,m
n Z t Z x j Z sr,m, j
−νl ∑ ui τ, σ̃sr ,sm ,x j dσ j,m,r ds j dτ
j=1, j6=r, j6=m 0 x0j xr,m, j
Z t Z xr Z sr,m
∂
−νl ui (τ, σ̃sm ,xr ) dσ r,m dsr dτ
0 x0r x0r,m ∂ xr
∂2
Z t Z xm Z sr,m
−νl ui (τ, σ̃sr ,xm ) dσ m,r dsm dτ
0 x0m x0r,m ∂ xr ∂ xm
Z t Z x Z sr,m,i
+l p (τ, σ̃sr ,sm ,si ) dσ r,m,i dsdτ
0 x0 xr,m,i
Z t Z x Z xr,m
−l f (τ, σ̃sr ,sm ) dσ m,r dsdτ, (t, x) ∈ [0, 1] × Rn .
0 x0 x0r,m

∂2 ∂2

1

L
∂ xr ∂ xm i (u, p)(t, x)
≤ l u
∂ xm ∂ xr i
(t, x)

Z x Z sr,m
+l |ui (t, σ̃sr ,sm ) − ui0 (σ̃sr ,sm )| dσ r,m ds
x0 x0r,m
n Z t Z x Z sr,m, j
+l ∑ ui τ, σ̃s ,s ,s u j τ, σ̃s ,s ,s dσ r,m, j dsdτ
r m j r m j
j=1, j6=m, j6=r 0 x0 x0r,m, j
Z t Z xr Z sr,m
+l |ui (τ, σ̃xr ,sm ) ur (τ, σ̃xr ,sm )| dσ r,m dsr dτ
0 x0r x0r,m
Z t Z xm Z sr,m
+l |ui (τ, σ̃sr ,xm ) um (τ, σ̃sr ,xm )| dσ r,m dsm dτ
0 x0m x0r,m
n Z t Z x j Z sr,m, j
+νl ∑ ui τ, σ̃s ,s ,x dσ j,m,r ds j dτ
r m j
j=1, j6=r, j6=m 0 x0j xr,m, j
Z t Z xr Z sr,m
∂
+νl ui (τ, σ̃sm ,xr ) dσ r,m dsr dτ
0 x0r x0r,m ∂ xr

∂2
Z t Z xm Z sr,m

+νl ui (τ, σ̃sr ,xm ) dσ m,r dsm dτ
0 xm xr,m ∂ xr ∂ xm
0 0
Z t Z x Z sr,m,i
+l |p (τ, σ̃sr ,sm ,si )| dσ r,m,i dsdτ
0 x0 xr,m,i
Z t Z x Z xr,m
+l | f (τ, σ̃sr ,sm )| dσ m,r dsdτ
0 x0 x0r,m
≤ lQ + 2l(B∗ )2 Q + (n − 2)l(B∗ )2 Q2 + l(B∗ )2 Q2
+l(B∗ )2 Q2 + νl(n − 2)(B∗ )2 Q + νl(B∗ )2 Q
+νl(B∗ )2 Q + l(B∗ )2 Q + l(B∗ )2 M 1
= l Q + (3 + νn)(B∗ )2 Q + n(B∗ )2 Q2 + (B∗ )2 M 1

≤ Q, (t, x) ∈ [0, 1] × Rn ,
∂ 2 Li1 ∂2
(u, p)(t, x) = −l ui (t, x)
∂ xi ∂ xm ∂ xi ∂ xm
Z x Z si,m
+l (ui (t, σ̃si ,sm ) − ui0 (σ̃si ,sm )) dσ m,i ds
x0 x0i,m
n Z t Z x Z si,m, j
+l ∑ ui τ, σ̃si ,sm ,s j u j τ, σ̃si ,sm ,s j dσ j,m,i dsdτ
j=1, j6=m, j6=i 0 x0 x0i,m, j
Z t Z xi Z sm,i
+l (ui (τ, σ̃sm ,xi ))2 dσ i,m dsi dτ
0 x0i x0m,i
Z t Z xm Z si,m
+l ui (τ, σ̃si ,xm ) u j (τ, σ̃si ,xm ) dσ m,i dsm dτ
0 x0m x0i,m
n Z t Z x j Z si,m, j
−νl ∑ ui τ, σ̃si ,sm ,x j dσ j,m,i ds j dτ
j=1, j6=m, j6=i 0 x0j x0i,m, j
Z t Z xi Z sm,i
∂ ui
−νl (τ, σ̃sm ,xi ) dσ i,m dsi dτ
0 x0i x0m,i ∂ xi
Z t Z xm Z si,m
∂
−νl ui (τ, σ̃si ,xm ) dσ m,i dsm dτ
0 x0m x0i,m ∂ xm
Z t Z xi Z sm,i
+l p (τ, σ̃sm ,xi ) dσ i,m dsi dτ
0 x0i x0m,i
Z t Z x Z si,m
−l f (τ, σ̃si ,sm ) dσ i,m dsdτ, (t, x) ∈ [0, 1] × Rn .
0 x0 x0i,m

2 1
∂2

∂ Li
∂ xi ∂ xm (u, p)(t, x) ≤ l ∂ xi ∂ xm ui (t, x)

Z x Z si,m
+l |ui (t, σ̃si ,sm ) − ui0 (σ̃si ,sm )| dσ m,i ds
x0 x0i,m
n Z t Z x Z si,m, j
+l ∑ ui τ, σ̃s ,s ,s u j τ, σ̃s ,s ,s dσ j,m,i dsdτ
i m j i m j
j=1, j6=m, j6=i 0 x0 x0i,m, j
Z t Z xi Z sm,i
+l (ui (τ, σ̃sm ,xi ))2 dσ i,m dsi dτ
0 x0i x0m,i
Z t Z xm Z si,m
+l ui (τ, σ̃s ,x ) u j (τ, σ̃s ,x ) dσ m,i dsm dτ
i m i m
0 x0m x0i,m
n Z t Z x j Z si,m, j
+νl ∑ ui τ, σ̃s ,s ,x dσ j,m,i ds j dτ
i m j
j=1, j6=m, j6=i 0 x0j x0i,m, j
Z t Z xi Z sm,i
∂ ui
+νl
∂ xi (τ, σ̃ )
sm ,xi dσ i,m dsi dτ

0 x0i x0m,i
Z t Z xm Z si,m
∂

+νl ∂ xm ui (τ, σ̃si ,xm ) dσ m,i dsm dτ
0 x0m x0i,m
Z t Z xi Z sm,i
+l |p (τ, σ̃sm ,xi )| dσ i,m dsi dτ
0 x0i x0m,i
Z t Z x Z si,m
+l | f (τ, σ̃si ,sm )| dσ i,m dsdτ
0 x0 x0i,m
≤ lQ + 2l(B∗ )2 Q + (n − 2)l(B∗ )2 Q2 + l(B∗ )2 Q2
+l(B∗ )2 Q2 + νl(B∗ )2 Q + νl(n − 2)(B∗ )2 Q + νl(B∗ )2 Q
+l(B∗ )2 Q + l(B∗ )2 M 1
= l Q + (3 + νn)(B∗ )2 Q + n(B∗ )2 Q2 + (B∗ )2 M 1

≤ Q, (t, x) ∈ [0, 1] × Rn .
11. Let (u, p) ∈ K 1 and r 6= i, m = i, m, i ∈ {1, . . . , n}. Then
∂2 ∂2
Li1 (u, p)(t, x) = −l ui (t, x)
∂ xr ∂ xi ∂ xi ∂ xr
Z x Z sr,i
+l (ui (t, σ̃sr ,si ) − ui0 (σ̃sr ,si )) dσ i,r ds
x0 x0r,i
n Z t Z x Z sr,i, j
+l ∑ ui τ, σ̃sr ,si ,s j u j τ, σ̃sr ,si ,s j dσ j,i,r dsdτ
j=1, j6=i, j6=r 0 x0 x0r,i, j
Z t Z xr Z si,r
+l ui (τ, σ̃si ,xr ) u j (τ, σ̃si ,xr ) dσ r,i dsr dτ
0 x0r x0i,r
Z t Z xi Z sr,i
+l (ui (τ, σ̃sr ,xi ))2 dσ i,r dsi dτ
0 x0i x0r,i
n Z t Z x j Z sr,i, j
−νl ∑ ui τ, σ̃sr ,si ,x j dσ r,i, j ds j dτ
j=1, j6=i, j6=r 0 x0j x0r,i, j
Z t Z xr Z si,r
∂
−νl ui (τ, σ̃si ,xr ) dσ r,i dsr dτ
0 x0r x0i,r ∂ xr
Z t Z xi Z si,r
∂
−νl ui (τ, σ̃xi ,sr ) dσ r,i dsi dτ
0 x0i x0i,r ∂ xi
Z t Z xi Z si,r
+l p (τ, σ̃xi ,sr ) dσ i,r dsi dτ
0 x0i x0i,r
Z t Z x Z si,r
−l f (τ, σ̃si ,sr ) dσ i,r dsdτ, (t, x) ∈ [0, 1] × Rn .
0 x0 x0i,r

∂2 ∂2

1

L
∂ xr ∂ xi i (u, p)(t, x)
≤ l u
∂ xi ∂ xr i
(t, x)

Z x Z sr,i
+l |ui (t, σ̃sr ,si ) − ui0 (σ̃sr ,si )| dσ i,r ds
x0 x0r,i
n Z t Z x Z sr,i, j
+l ∑ ui τ, σ̃s ,s ,s u j τ, σ̃s ,s ,s dσ j,i,r dsdτ
r i j r i j
j=1, j6=i, j6=r 0 x0 x0r,i, j
Z t Z xr Z si,r
+l ui (τ, σ̃s ,x ) u j (τ, σ̃s ,x ) dσ r,i dsr dτ
i r i r
0 x0r x0i,r
Z t Z xi Z sr,i
+l (ui (τ, σ̃sr ,xi ))2 dσ i,r dsi dτ
0 x0i x0r,i
n Z t Z x j Z sr,i, j
+νl ∑ ui τ, σ̃s ,s ,x dσ r,i, j ds j dτ
r i j
j=1, j6=i, j6=r 0 x0j x0r,i, j
Z t Z xr Z si,r
∂

+νl ∂ xr ui (τ, σ̃si ,xr ) dσ r,i dsr dτ
0 x0r x0i,r
Z t Z xi Z si,r
∂

+νl ∂ xi ui (τ, σ̃xi ,sr ) dσ r,i dsi dτ
0 x0i x0i,r
Z t Z xi Z si,r
+l |p (τ, σ̃xi ,sr )| dσ i,r dsi dτ
0 x0i x0i,r
Z t Z x Z si,r
+l | f (τ, σ̃si ,sr )| dσ i,r dsdτ
0 x0 x0i,r
≤ lQ + 2l(B∗ )2 Q + (n − 2)l(B∗ )2 Q2
+l(B∗ )2 Q2 + l(B∗ )2 Q2 + νl(n − 2)(B∗ )2 Q
+νl(B∗ )2 Q + νl(B∗ )2 Q + l(B∗ )2 Q + l(B∗ )2 M 1
= l Q + (3 + νn)(B∗ )2 Q + n(B∗ )2 Q2 + (B∗ )2 M 1

≤ Q, (t, x) ∈ [0, 1] × Rn .
12. Let (u, p) ∈ K 1 and m ∈ {1, . . . , n}. Then
∂2 1 ∂2
2
L n+1 (u, p)(t, x) = −l 2
p(t, x)
∂ xm ∂ xm
n Z t Z xm Z sm, j
+l ∑ u j τ, σ̃xm ,s j dσ j,m dsm dτ
j=1, j6=m 0 x0m x0m, j
Z t Z x m Z sm
∂
+l um (τ, σ̃xm ) dσ m dsm dτ, (t, x) ∈ [0, 1] × Rn .
0 x0m x0m ∂ xm

2 2
∂ 1
∂
∂ x2 Ln+1 (u, p)(t, x) ≤ l ∂ x2 p(t, x)

m m
n Z t Z xm Z sm, j
+l ∑ u j τ, σ̃x ,s dσ j,m dsm dτ
m j
j=1, j6=m 0 x0m x0m, j
Z t Z xm Z sm
∂

+l ∂ xm um (τ, σ̃xm ) dσ m dsm dτ
0 x0m x0m
≤ lQ + l(n − 1)(B∗ )2 Q + l(B∗ )2 Q
= l Q + n(B∗ )2 Q

≤ Q, (t, x) ∈ [0, 1] × Rn .
13. Let (u, p) ∈ K 1 and m 6= r, m, r ∈ {1, . . . , n}. Then
∂2 1 ∂2
Ln+1 (u, p)(t, x) = −l p(t, x)
∂ xm ∂ xr ∂ xm ∂ xr
n Z t Z x Z sm, j,r
+l ∑ u j τ, σ̃sm ,s j ,sr dσ j,m,r dsdτ
j=1, j6=m, j6=r 0 x0 x0m, j,r
Z t Z xr Z sm,r
+l u j (τ, σ̃sm ,xr ) dσ m,r dsr dτ
0 x0r x0m,r
Z t Z xm Z sm,r
+l um (τ, σ̃xm ,sr ) dσ m,r dsm dτ. (t, x) ∈ [0, 1] × Rn .
0 x0m x0m,r

∂2 ∂2

1

∂ xm ∂ xr Ln+1 (u, p)(t, x) ≤ l ∂ xm ∂ xr p(t, x)

n Z t Z x Z sm, j,r
+l ∑ u j τ, σ̃s ,s ,s dσ j,m,r dsdτ
m j r
j=1, j6=m, j6=r 0 x0 x0m, j,r
Z t Z xr Z sm,r
+l u j (τ, σ̃s ,x ) dσ m,r dsr dτ
m r
0 x0r x0m,r
Z t Z xm Z sm,r
+l |um (τ, σ̃xm ,sr )| dσ m,r dsm dτ
0 x0m x0m,r
≤ lQ + l(n − 2)(B∗ )2 Q + l(B∗ )2 Q + l(B∗ )2 Q
= l Q + n(B∗ )2 Q

≤ Q, (t, x) ∈ [0, 1] × Rn .
From items 1-13, it follows that L1 : K 1 → K 1 and it is continuous. Since K 1 is a compact

subset of Q1 , we have that L1 (K 1 ) resides in a compact subset of Q1 . This completes the
proof.
1
Proposition 6.2.7 M : K 1 → Q1 is an expansive operator and onto.
Proof 6.2.8 For U = (u, p), Ũ = (ũ, p̃) ∈ K 1 we have

1 1
||M (U) − M (Ũ)|| = (1 + l)||U − Ũ||,
1 1 1 1
M : K → Q is an expansive operator with h = 1 + l. Let (ũ, p̃) ∈ Q . Then
i.e.,
ũ p̃
, ∈ K 1 and
1+l 1+l
1 ũ p̃
M , = (ũ, p̃),
1+l 1+l
1
i.e., the operator M : K 1 → Q1 is onto. This completes the proof.
By Proposition 6.2.5, Proposition 6.2.7 and Theorem 1.9.12 it follows that the operator
1
L1 + M has a fixed point (u1 , p1 ) ∈ K 1 . For it we have
Z xZ s
u1i (t, σ ) − ui0 (σ ) dσ ds

x0 x0
n Z t Z x Z sj
u1i τ, σ̃s j u1j τ, σ̃s j dσ j dsdτ

+∑
j=1 0 x0 x0j
n Z t Z xj Z sj
u1i τ, σ̃x j dσ j ds j dτ

−ν ∑
j=1 0 x0j x0j
Z t Z x Z si
+ p1 (τ, σ̃si ) dσ i dsdτ,
0 x0 x0i
Z tZ xZ s
= fi (τ, σ )dσ dsdτ, i ∈ {1, . . . , n},
0 x0 x0
n Z t Z x Z sj
u1j τ, σ̃s j dσ j dsdτ = 0, x ∈ Rn .

∑ t ∈ [0, 1],
j=1 0 x0 x0j
Hence by Lemma 6.2.1 it follows that (u1 , p1 ) is a C 1 ([0, 1], C02 (Rn )) solution of problem
(6.18), (6.19) for which suppx p1 , suppx u1i ⊂ B, i ∈ {1, . . . , n}. Consider the Cauchy problem
n

∂ ui ∂ ui ∂p
+ ∑ uj = ν∆ui − + fi (t, x), i ∈ {1, . . . , n},





 ∂t j=1 ∂ x j ∂ xi
(6.21)
 n ∂ ui


n
 ∑ ∂ x (t, x) = 0 in [1, 2] × R ,


i=1 i
u(1, x) = u1 (1, x) in Rn . (6.22)

Let
M2 = max | f (t, x)|.
t∈[1,2],x∈B
We choose the constant l1 > 0 such that

l1 n Q + 3n(1 + ν)(B∗ )2 Q + n2 Q2 (B∗ )2 + (B∗ )2 M 2 ≤ Q.

Let n
o
i ∈ {1, . . . , n + 1}

n
q∈{1,...,n+1} t∈[1,2],x∈B t∈[1,2],x∈B

t∈[1,2],x∈B t∈[1,2],x∈B
o
i, j ∈ {1, . . . , n} .
e2 we denote the set of all equi-continuous families in E 2 . Let also

With K
2 2
e2 ,
K =K K 2 = {v ∈ K : ||v|| ≤ Q},
2
Q2 = {v ∈ K : ||v|| ≤ (1 + l1 )Q}.
e2 . By Arzela’s Theorem, we have that K 2 is a compact subset of

e2 is the closure of K
Here K
2 2
Q . For (u, p) ∈ Q , we define the operators.
Z xZ s
Li2 (u, p)(t, x) = −l1 ui (t, x) + l1 ui (t, σ ) − u1i (1, σ ) dσ ds

x0 x0
n Z t Z x Z sj
+l1 ∑ ui τ, σ̃s j u j τ, σ̃s j dσ j dsdτ
j=1 1 x0 x0j
n Z t Z xj Z sj
−νl1 ∑ ui τ, σ̃x j dσ j ds j dτ
j=1 1 x0j x0j
Z t Z x Z si
+l1 p (τ, σ̃si ) dσ i dsdτ
1 x0 x0i
Z tZ xZ s
−l1 f (τ, σ )dσ dsdτ,
1 x0 x0
n Z t Z y Z x Z sj
2

Ln+1 (u, p)(t, x) = −l1 p(t, x) + l1 ∑ u j τ, σ̃s j dσ j dsdτdy
j=1 1 1 x0 x0j
Z xZ s
pt (1, σ ) − pt1 (1, σ ) dσ ds

+l1 (t − 1)
x0 x0
Z xZ s
p(1, σ ) − p1 (1, σ ) dσ ds,

+l1
x0 x0
Mi2 (u, p)(t, x) = (1 + l1 )ui (t, x), i ∈ {1, . . . , n},

2
Mn+1 (u, p)(t, x) = (1 + l1 )p(t, x),
L2 (u, p)(t, x) = L12 (u, p)(t, x), . . . , Ln+1

2

(u, p)(t, x) ,
2
M12 (u, p)(t, x), . . . , Mn+1
2

M (u, p)(t, x) = (u, p)(t, x) ,
(t, x) ∈ [1, 2] × Rn .
2
As in the above, we obtain that the operator L2 + M has a fixed point (u2 , p2 ) ∈ K 2 . For it
we have
Z xZ s
u2i (t, σ ) − u1i (1, σ ) dσ ds

0 =
x0 x0
n Z t Z x Z sj
u2i τ, σ̃s j u2j τ, σ̃s j dσ j dsdτ

+∑
j=1 1 x0 x0j
n Z t Z xj Z sj
u2i τ, σ̃x j dσ j ds j dτ

−ν ∑
j=1 1 x0j x0j
Z t Z x Z si
+ p2 (τ, σ̃si ) dσ i dsdτ
1 x0 x0i
(6.23)
Z tZ xZ s
− f (τ, σ )dσ dsdτ, i ∈ {1, . . . , n},
1 x0 x0
n Z t Z y Z x Z sj
u2j τ, σ̃s j dσ j dsdτdy

0 = ∑
j=1 1 1 x0 x0j
Z xZ s
pt2 (1, σ ) − pt1 (1, σ ) dσ ds

+(t − 1)
x0 x0
Z xZ s
p2 (1, σ ) − p1 (1, σ ) dσ ds, (t, x) ∈ [1, 2] × Rn .

+
x0 x0
We differentiate the first n equations of system (6.23) once in t, twice in x1 , .., xn , and we get
n n
∂ 2 ∂ ∂2 ∂
u2i u2j − ν ∑ ∂ x2 u2i + ∂ xi p2 = fi (t, x),

ui + ∑ (6.24)
∂t j=1 ∂ x j j=1 j
i ∈ {1, . . . , n}, (t, x) ∈ [1, 2] × Rn .

Now we differentiate the last equation of (6.23) twice in t, twice in x1 , .., xn , and we get
n
∂
∑ ∂ x j u2j (t, x) = 0, (t, x) ∈ [1, 2] × Rn .
j=1
For i ∈ {1, . . . , n} we put t = 1 in (6.23) and we find

Z xZ s
u2i (1, σ ) − u1i (1, σ ) dσ ds = 0,

x0 x0
which we differentiate twice in x1 , and so on, twice in xn , and we find
u2i (1, x) = u1i (1, x), x ∈ Rn , (6.25)
whereupon
∂ 2 ∂ 1
u (1, x) = u (1, x), x ∈ Rn , (6.26)
∂xj i ∂xj i
and
∂2 ∂2
u2i (1, x) = u1 (1, x), i, m, r ∈ {1, . . . , n}, x ∈ Rn . (6.27)
∂ xm ∂ xr ∂ xm ∂ xr i
In the (n + 1)-th equation of system (6.23) we put t = 1 and we find
Z xZ s
p2 (1, σ ) − p1 (1, σ ) dσ ds = 0,

x0 x0
which we differentiate twice in x1 , and so on, twice in xn , and we find
p2 (1, x) = p1 (1, x), x ∈ Rn .
Hence,
∂ p2 ∂ p1
(1, x) = (1, x), x ∈ Rn , i ∈ {1, . . . , n}, (6.28)
∂ xi ∂ xi
and
∂ 2 p2 ∂ 2 p1
(1, x) = (1, x), x ∈ Rn , r, m ∈ {1, . . . , n}. (6.29)
∂ xm ∂ xr ∂ xm ∂ xr
Now we differentiate in t the n + 1-th equation of system (6.23) and we get
n Z t Z x Z sj Z xZ s
u2j τ, σ̃s j dσ j dsdτ + pt2 (1, σ ) − pt1 (1, σ ) dσ ds = 0,

∑ x0 x0j x0 x0
j=1 1
in which we put t = 1 and we get

Z xZ s
pt2 (1, σ ) − pt1 (1, σ ) dσ ds = 0.

x0 x0
The last equation we differentiate twice in x1 , and so on, twice in xn , and we get
pt2 (1, x) = pt1 (1, x), x ∈ Rn . (6.30)
By (6.25)-(6.30), using (6.24), we obtain
n
∂ 2 ∂
u2i (1, x)u2j (1, x)

ui (1, x) = − ∑
∂t j=1 ∂ x j
n
∂2 ∂ p2
+ν ∑ ∂ x2 u2i (1, x) − ∂ xi (1, x) + fi (1, x)
j=1 j
n
∂
u1i (1, x)u1j (1, x)

= −∑
j=1 ∂ x j
n
∂2 ∂ p1
+ν ∑ ∂ x2 u1i (1, x) − ∂ xi (1, x) + fi (1, x)
j=1 j
∂ 1
= u (1, x), x ∈ Rn , i ∈ {1, . . . , n}.
∂t i
Consequently the function
 1 1
 (u , p ) for t ∈ [0, 1], x ∈ Rn ,
(u, p) =
u2 , p2 x ∈ Rn ,

for t ∈ [1, 2],

is a solution to the problem

n

∂ ui ∂ ui ∂p
+ ∑ uj = ν∆ui − + fi (t, x), i ∈ {1, . . . , n},





 ∂t j=1 ∂ x j ∂ xi
n
∂ ui


in [0, 2] × Rn ,



 ∑ ∂ xi (t, x) = 0
i=1
u(0, x) = u0 (x) in Rn ,
such that ui , p ∈ C 1 ([0, 2], C02 (Rn )), suppx ui , suppx p ⊂ B, i ∈ {1, . . . , n}. Note that
|ui |, |p| ≤ Q on [0, 2] × Rn , i ∈ {1, . . . , n}.
Now we consider the Cauchy problem
n

∂ ui ∂ ui ∂p
+ uj = ν∆ui − + fi (t, x), i ∈ {1, . . . , n},



 ∂t
 ∑ ∂xj ∂ xi
j=1

n
∂ ui


in [2, 3] × Rn ,



 ∑ ∂ xi (t, x) = 0
i=1
u(2, x) = u2 (2, x) in Rn ,
Let
M3 = max | f (t, x)|.
t∈[2,3],x∈B
We choose the constant l2 > 0 such that
l2 n Q + 3n(1 + ν)(B∗ )2 Q + n2 Q2 (B∗ )2 + (B∗ )2 M 3 ≤ Q.

Let n
o
i ∈ {1, . . . , n + 1}
n
q∈{1,...,n+1} t∈[2,3],x∈B t∈[2,3],x∈B

t∈[2,3],x∈B t∈[2,3],x∈B
o
i, j ∈ {1, . . . , n} .
e3 we denote the set of all equi-continuous families in E 3 . Let also

With K
3 3
e3 ,
K =K K 3 = {v ∈ K : ||v|| ≤ Q},
3
Q3 = {v ∈ K : ||v|| ≤ (1 + l2 )Q}.
e3 . By Arzela’s Theorem, we have that K 3 is a compact subset of

e3 is the closure of K
Here K
3 3
Q . For (u, p) ∈ Q , we define the operators.
Z xZ s
Li3 (u, p)(t, x) = −l2 ui (t, x) + l2 ui (t, σ ) − u2i (2, σ ) dσ ds

x0 x0
n Z t Z x Z sj
+l2 ∑ ui τ, σ̃s j u j τ, σ̃s j dσ j dsdτ
j=1 2 x0 x0j
n Z t Z xj Z sj
−νl2 ∑ ui τ, σ̃x j dσ j ds j dτ
j=1 2 x0j x0j
Z t Z x Z si
+l2 p (τ, σ̃si ) dσ i dsdτ
2 x0 x0i
Z tZ xZ s
−l2 f (τ, σ )dσ dsdτ,
2 x0 x0
n Z t Z y Z x Z sj
3

Ln+1 (u, p)(t, x) = −l2 p(t, x) + l2 ∑ u j τ, σ̃s j dσ j dsdτdy
j=1 2 2 x0 x0j
Z xZ s
pt (2, σ ) − pt2 (2, σ ) dσ ds

+l2 (t − 2)
x0 x0
Z xZ s
p(2, σ ) − p2 (2, σ ) dσ ds,

+l2
x0 x0
Mi3 (u, p)(t, x) = (1 + l2 )ui (t, s), i ∈ {1, . . . , n},
3
Mn+1 (u, p)(t, x) = (1 + l2 )p(t, x),
L3 (u, p)(t, x) = L13 (u, p)(t, x), . . . , Ln+1

3

(u, p)(t, x) ,
3
M13 (u, p)(t, x), . . . , Mn+1
3

M (u, p)(t, x) = (u, p)(t, x) ,
(t, x) ∈ [2, 3] × Rn ,
and so on.
 1 1

 (u , p ) for t ∈ [0, 1], x ∈ Rn ,




 u2 , p2 for t ∈ [1, 2], x ∈ Rn ,


(ũ, p̃) =
u3 , p3 x ∈ Rn ,

for t ∈ [2, 3],








···

is a solution to problem (6.12), (6.13) such that ũi , p̃ ∈ C 1 ([0, ∞), C02 (Rn )), suppx ũi ,
suppx p̃ ⊂ B, i ∈ {1, . . . , n}. Since fi ∈ C 2 ([0, ∞), C0∞ (Rn )), using (6.12), we conclude that
p̃, ũi ∈ C ∞ ([0, ∞) × Rn ), i ∈ {1, . . . , n}. Note that
|ũi |, | p̃| ≤ Q on [0, ∞) × Rn , i ∈ {1, . . . , n}.
Therefore Z Z
|ũ(t, x)|2 dx = |ũ(t, x)|2 dx ≤ nQ2 µ(B) < ∞.
Rn B
Bibliography
[1] A. Babakhani and V.D. Gejji. Existence of positive solutions of nonlinear fractional
differential equations, J. Math. Anal. Appl. 278 (2003) 434–442.
[2] L. Benzenati and K. Mebarki. Multiple positive fixed points for the sum of expansive
mappings and k-set contractions, MMAS, 2019, Vol. 42, No. 13 (1919), 4412–4426.
[3] D. Delbosco. Fractional calculus and function spaces, J. Fract. Calc. 6 (1996) 45–53.
[4] D. Delbosco and L. Rodino. Existence and uniqueness for a nonlinear fractional dif-
ferential equation, J. Math. Anal. Appl. 204 (1996) 609–625.
[5] J. Dugundji. An extension of Tietze’s theorem. Pacific Journal of Mathematics, Vol. 1,
No. 3, 1951.
[6] G. Dinca and J. Mawhin. Brouwer degree and applications, 2009, a private copy.
[7] K. Deimling. Nonlinear Functional Analysis, Springer, 1985.
[8] S. Georgiev. Fundamental existence theorems for first order nonlinear ordinary differ-
ential equations. AMSA (Japan), 2010.
[9] S. Georgiev. Global existence for nonlinear wave equation. Dynam. PDE, Vol. 7,
No. 3, pp. 207–216, 2010.
[10] S. Georgiev. A note on the linear systems of quaternions. Progress in Analysis and Its
Applications, Vol. 73, No. 2, pp. 29–42, 2010.
[11] S. Georgiev. Periodic solutions to the nonlinear parabolic equations. Int. J. Evol. Eq.,
Vol. V, No. 1, 2010, pp. 53–60.
[12] S. Georgiev. Blow up of the solutions of the nonlinear parabolic equation, Dyn. of
PDE, Vol. 6, No. 1, pp. 1–14, 2009.
[13] S. Georgiev and P. LeFloch. Generalised time-periodic solutions to the Euler equa-
tions of compressible fluids, DEA, Vol. 1, No. 3, 2009, pp. 413–426.
[14] S. Georgiev. Positive periodic solutions to the nonlinear wave equation, DEA, Vol. 2,
No. 2, 2010, pp. 283–298.
293
294 Bibliography
[15] S. Georgiev. Positive periodic solutions to the Kolmogorov’s system, Nonlinear Stud-
ies, Vol. 16, No. 4, 2009, pp. 381–402.
[16] S. Georgiev. Existence of classical solutions for a class nonlinear wave equations.
GJSFR(F), Vol. 12, No. 11, Version 1, pp. 44–57.
[17] S. Georgiev. Existence of solutions for first order nonlinear ordinary differential equa-
tions, Nonlinear Studies, Vol. 19, No. 3, 2012, pp. 457–460
[18] S. Georgiev. Periodic solutions for a class nonlinear parabolic equations, Applicable
Analysis, Vol. 92, No. 8, 2013, pp. 1643–1654.
[19] S. Georgiev. Global existence of classical solutions for a class nonlinear parabolic
equations, GJSFR(F), Vol. 12, No. 8, Version 1, pp. 1–12, 2012.
[20] D. Guo, Y. Cho and J. Zhu. Partial Ordering Methods in Nonlinear Problems, Nova
Science Publishers, 2004.
[21] I. Gyori and G. Ladas. Oscillation Theory of Delay Differential Equations with
Applications, Clarendon, Oxford, 1991.
[22] A.A. Kilbas, O.I. Marichev and S.G. Samko. Fractional Integral and Derivatives
(Theory and Applications), Gordon and Breach, Switzerland, 1993.
[23] M.A. Krasnosel’skii. Positive Solutions of Operator Equations, Noordhoff, Gronin-
gen, 1964.
[24] R.W. Leggett and L.R. Williams. Multiple positive fixed points of nonlinear operators
on ordered Banach spaces, Indiana Univ. Math. J. 28 (1979) 673–688.
[25] M.C. Mackey and L. Glass. Oscillations and chaos in physiological control systems,
Sciences 197 (1977) 287–289.
[26] K.S. Miller. Fractional differential equations, J. Fract. Calc. 3 (1993) 49–57.
[27] K.S. Miller and B. Ross. An Introduction to the Fractional Calculus and Fractional
Differential Equations, Wiley, New York, 1993.
[28] I. Podlubny. Fractional Differential Equations, Mathematics in Science and Engineer-
ing, Academic Press, New York, 1999.
[29] I. Podlubny. The Laplace transform method for linear differential equations of the
fractional order, Inst. Expe. Phys., Slov. Acad. Sci., UEF-02-94, Kosice, 1994.
[30] T. Xiang and R. Yuan. A Class of Expansive-Type Krasnosel’skii Fixed Point Theo-
rems, Nonlinear Analysis, 71(2009), pp. 3229–3239.
[31] T. Xiang and S. Georgiev. Noncompact-type Krasnoselskii fixed point theorems and
their applications. MMAS, Vol. 39, No. 4, 2016, pp. 833-863. (wileyonlinelibrary.com)
DOI: 10.1002/mma.3525.
Index
Brouwer Degree, 34 FDEs. See Fractional-differential equations

continuous mappings, 42–51 First order ODEs
homotopy invariance, 39–42 BVPs for, 102–118
smooth mappings, 34–39 periodic solutions for, 83–101
Brouwer Fixed-Point Theorem, 11–20 Fixed-point index theory
BVP Brouwer Degree, 34–51
for first order ODEs, 102–118 Brouwer Fixed-Point Theorem, 11–20
with impulses, 123–135 fixed-point index for completely
multiple solutions with Robin continuous mappings, 56–57
boundary conditions, 257–263 fixed-point index for strict set
of nonlinear Caputo FDEs, 166–168 contractions, 57–64
of nonlinear Riemann-Liouville FDEs, fixed points of strict set contractions,
153–166 20–24
for second order ODEs, 119–123 Kronecker Index, 25–34
Leray-Schauder Degree, 52–55
Cauchy problem, 169 measures of noncompactness, 1–11
Completely continuous mappings, multiple fixed-point theorems, 64–82
fixed-point index for, 56–57 Schauder Fixed-Point Theorem, 11–20
Condensing operator, 21 Fixed point property, 16, 17, 18
Cone, 64 Fractional-differential equations (FDEs),
137
Dabor’s Theorem, 23–24 BVP of nonlinear Caputo FDEs,
Dugundji’s Extension Theorem, 22 166–168
BVP of nonlinear Riemann-Liouville
e-concave mapping, 65 FDEs, 153–166
e-convex mapping, 65 global existence for class of, 137–153
ε-fixed points, 18
Elliptic equations Generalized fixed-point index, 71, 74, 75
Navier-Stokes equations, 264–291
solutions for BVP with Robin Hilbert Cube, 19, 20
boundary conditions, 257–263 Homotopy invariance, 39–42
Existence Property, 44 Hyperbolic equations
Expansive operator, 71, 74, 75, 141, 162, IVP for, 221–238
175, 189, 230 periodic solutions, 245–256
295
296 Index
solutions for IBVP with Robin periodic solutions for first order
boundary conditions, 238–245 ODEs, 83–101
IBVP Parabolic equations

classical solutions for, 182–200 classical solutions for IBVP, 182–200
with Robin boundary conditions, classical solutions for IVP, 169–182
213–220, 238–245 periodic solutions, 200–213
Impulses, BVPs with, 123–135 solutions for IBVP with Robin
IVP boundary conditions, 213–220
classical solutions for, 169–182 Periodic solutions, 200–213
for hyperbolic equations, 221–238 for first order ODEs, 83–101
hyperbolic equations, 245–256
Kronecker Index, 25–34 Poincaré index, 26
Kuratowski measure, 1 Population dynamic of hematopoiesis, 100
Projection, 20
Leray-Schauder Degree, 52–55
Regular value, 37
Mönch’s Theorem, 22–23 Retract, 17
Multiple fixed-point theorems, 64–82 Retraction, 17
Robin boundary conditions
Navier-Stokes equations, 264–291 elliptic equations for BVP, 257–263
Noncompactness measure, 1 hyperbolic equations for IBVP,
Noncompactness, measures of, 1–11 238–245
Nonlinear Caputo FDEs, 166–168 IBVP with, 213–220
Nonlinear Riemann-Liouville FDEs,
153–166 Sadovskii’s Theorem, 23
Normal Cone, 76, 78, 80, 82, 100, 165 Sard’s Theorem, 42
Normal Constant, 64 Schauder Fixed-Point Theorem, 11–20
Second order ODEs, BVPs for, 119–123
ODEs, applications to Solid angle [n–1]-form, 25
BVPs for first order ODEs, 102–118 Strict set contractions, fixed-point index
BVPs for second order ODEs, for, 57–64
119–123
BVPs with impulses, 123–135 Winding number, 26

(Chapman &amp; Hall_CRC Monographs and Research Notes in Mathematics) Svetlin G. Georgiev, Khaled Zennir - Multiple Fixed-Point Theorems and Applications in the Theory of ODEs, FDEs and PDEs-CRC Press (20

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Chapman &amp; Hall_CRC Monographs and Research Notes in Mathematics) Svetlin G. Georgiev, Khaled Zennir - Multiple Fixed-Point Theorems and Applications in the Theory of ODEs, FDEs and PDEs-CRC Press (20

Uploaded by

Copyright:

Available Formats

Multiple Fixed-Point

Theorems and Applications

About the Series

La"ce Point Iden!!es and Shannon-Type Sampling

Spectral Geometry of Par!al Diﬀeren!al Operators (Open Access)

Linear Groups: The Accent on Inﬁnite Dimensionality

For more informa!on about this series please visit: h$ps://www.crcpress.com/Chapman--HallCRC-Monographs-and-Research-Notes-in-Mathema!cs/book-

and by CRC Press

CRC Press is an imprint of Taylor & Francis Group, LLC

ISBN: 978-0-367-46432-5 (hbk)

Visit the Taylor & Francis Web site at

and the CRC Press Web site at

1 Fixed-Point Index Theory. Multiple Fixed-Point Theorems 1

3 Applications to FDEs 137

4 Applications to Parabolic Equations 169

4.2 Local Existence of Classical Solutions for an IBVP . . . . . . . . . . . . 182

5 Applications to Hyperbolic Equations 221

6 Applications to Elliptic Equations 257

Many problems in science lead to nonlinear equations T x + Fx = x posed in some closed

1.1 Measures of Noncompactness

Because S is bounded, we have

Definition 1.1.1 α(S) is said to be the Kuratowski measure of noncompactness and it is

Proof 1.1.3 Since S ⊆ S, we get

diam(S j ) < α(S) + ε, j ∈ {1, . . . , m}.

< α(S) + ε, j ∈ {1, . . . , m},

≤ η + ε, j ∈ {1, . . . , m}, l ∈ {1, . . . , k}.

Hence, and considering (1.2), we arrive at

This completes the proof.

Theorem 1.1.6 Let S and T be bounded subsets of E. Then

α(S + T ) ≤ α(S) + α(T ),

Proof 1.1.7 Take ε > 0 arbitrarily. Then there are partitions

diam(Tl ) ≤ α(T ) + ε, j ∈ {1, . . . , m}, l ∈ {1, . . . , k}.

= α(S) + α(T ) + 2ε, j ∈ {1, . . . , m}, l ∈ {1, . . . , k}.

Proof 1.1.9 1. Let a = 0. Then

2. Let a 6= 0. Take ε > 0 arbitrarily. Then there is a partition

diam(S j ) ≤ α(S) + ε, j ∈ {1, . . . , m}.

This completes the proof.

Theorem 1.1.10 Let S be a bounded set of E. Then

Proof 1.1.11 Note that S ⊆ coS. Therefore

α(S) ⊆ α(coS). (1.4)

diam(S j ) < a1 , j ∈ {1, . . . , m}.

Take a2 ∈ (0, a1 ) such that

diam(S j ) < a2 , j ∈ {1, . . . , m}.

< a2 , j ∈ {1, . . . , m}.

where x j , y j ∈ coS j , j ∈ {1, . . . , m}. Then

and λ = (λ1 , . . . , λm ), µ = (µ1 , . . . , µm ) ∈ D. Letting

z = hx + (1 − h)y, h ∈ (0, 1),

zj ∈ coS j , j ∈ {1, . . . , m}.

Thus z ∈ X(D) and X(D) is convex. Therefore

diam(D j ) < δ , j ∈ {1, . . . , n}.

where d(·, ·) denotes the distance from an element of E to a subset of E.

diam(S j ) < α(S) + ε, j ∈ {1, . . . , m}.

≤ ky1 − x1 k + kx1 − x2 k + kx2 − y2 k

< 2dh (S, T ) + 2ε + α(S) + ε

= 2dh (S, T ) + α(S) + 3ε

α(S) ≤ 2dh (S, T ) + α(T ) + 3ε.

|α(S) − α(T )| ≤ 2dh (S, T ).

This completes the proof.

1.2 The Brouwer Fixed-Point Theorem. The Schauder Fixed-Point

be the open unit ball,

For t ∈ [0, 1], define

kg(x2 ) − g(x1 )k ≤ Ckx2 − x1 k

We have that ft0 : Rn → Rn and

ft0 (x) = I + tg0 (x), x ∈ Rn .

(Chapman & Hall_CRC Monographs and Research Notes in Mathematics) Svetlin G. Georgiev, Khaled Zennir - Multiple Fixed-Point Theorems and Applications in the Theory of ODEs, FDEs and PDEs-CRC Press (20

(Chapman & Hall_CRC Monographs and Research Notes in Mathematics) Svetlin G. Georgiev, Khaled Zennir - Multiple Fixed-Point Theorems and Applications in the Theory of ODEs, FDEs and PDEs-CRC Press (20