Professional Documents
Culture Documents
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
Summable Spaces and Their Duals, Matrix Transforma!ons and Geometric Proper!es
Feyzi Basar, Hemen Du!a
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
Svetlin G. Georgiev
Khaled Zennir
First edition published 2020
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
c 2020 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.
Typeset in Nimbus
by Nova Techset Private Limited, Bengaluru & Chennai, India
Preface vii
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
v
vi Contents
Bibliography 293
Index 295
Preface
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
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).
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,
such that
Sj ⊆ E,
Tl ⊆ E,
diam(S j ) ≤ α(S) + ε
≤ η + ε,
diam(Tl ) ≤ α(T ) + ε
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.
where
S + T = {y + z : y ∈ S, z ∈ T }.
such that
Sj ⊆ E,
Tl ⊆ E,
diam(S j ) ≤ α(S) + ε,
and
diam(V jl ) ≤ diam(S j ) + diam(Tl )
≤ α(S) + ε + α(T ) + ε
= |a|α(S).
such that
Sj ⊆ E,
≤ |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),
6 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
whereupon
|a|α(S) ≤ α(aS).
Hence, and applying (1.3), we get
α(aS) = |a|α(S).
α(coS) = α(S).
Sj ⊆ E,
Then
diam(coS j ) = diam(S j )
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
Measures of Noncompactness 7
Take x, y ∈ X(λ ),
m
x = ∑ λ jx j,
j=1
m
y = ∑ λ jy 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
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,
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 .
Measures of Noncompactness 9
n
[
Because D is a compact set in Rm , there exists a partition D = D j such that
j=1
≤ 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).
Hence, and considering (1.4), we get
α(S) = α(coS).
This completes the proof.
Theorem 1.1.12 Let S and T be bounded subsets of E. Then
|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
Sj ⊆ E,
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
< 2η + diam(S j ).
Therefore
diam(T j ) ≤ 2η + diam(S j )
and
α(T ) ≤ 2dh (S, T ) + α(S) + 3ε.
As in the above, one can prove
Consequently
|α(S) − α(T )| ≤ 2dh (S, T ) + 3ε.
Because ε > 0 was arbitrarily chosen, we obtain
Bn = {x ∈ Rn : kxk < 1}
ft (x) = x + t( f (x) − x)
n
= x + tg(x), x∈B .
12 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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
= 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 ).
The Brouwer Fixed-Point Theorem. The Schauder Fixed-Point Theorem 13
Here, I is the identity map. Since g is C 1 , there is a point t0 > 0 such that
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→∞
lim xn = x0 .
n→∞
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 ).
14 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
n
Because f1 (x) = f (x) ∈ Sn−1 , x ∈ B , we obtain
n
= 1, 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.
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
The Brouwer Fixed-Point Theorem. The Schauder Fixed-Point Theorem 15
Then
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
hm (xm ) = xm , m ∈ N.
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 .
This completes the proof.
16 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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).
x = g(x)
= − f (x)
= −x
6= x.
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.
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.
The Brouwer Fixed-Point Theorem. The Schauder Fixed-Point Theorem 17
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.
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.
f ◦ r : Y → X ⊂ Y.
( f ◦ r)(y) = y and y ∈ X.
( f ◦ r)(y) = f (r(y))
= f (y)
= y.
Theorem 1.2.15 (Brouwer Fixed-Point Theorem) Every compact convex nonempty sub-
set A of Rn has the fixed-point property.
18 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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.
The Brouwer Fixed-Point Theorem. The Schauder Fixed-Point Theorem 19
< 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
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.
20 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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, . . .}.
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.
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.
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)),
22 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
[ [
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.
Hence, and applying Dabor’s Theorem, we get that A has a fixed point in D. This completes
the
proof.
24 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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
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 1.4.3 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
f1 (φ (s)) = cos s,
f2 (φ (s)) = sin s,
The Kronecker Index 27
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
= 1.
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
= −1.
Then
= cos(2s),
= 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.
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).
The Kronecker Index 29
Consequently
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
30 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
= m.
This completes the proof.
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
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
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.
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
32 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
!
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
1
Z ∞
= a(s)sn−1 ds
rn 0
= 0.
is such that Z ∞
suppb ⊂ [ε0 , µ0 ], b(r)rn−1 dr = 1,
0
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 ∗ σB
µn−1 ∂D
34 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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.
This completes the proof.
or
1
Z ∞
c(r)rn−1 dr = .
0 µn−1
= 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).
36 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
/ 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.
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 ).
dB ( f , D, z) = dB ( f (·) − z, D).
dB ( f , D, z) = dB ( f (·) − z, D)
Z
= c (k f (x) − zk) J f (x)dx,
D
= 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
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
38 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
Therefore
m
dB ( f , D, z) = ∑ sign(J f (x j ))
j=1
= N+ − N− .
Hence,
r s
mj
det(A) = ∏ λ j ∏ |µl |2nl .
j=1 l=1
The Brouwer Degree 39
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}.
= 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
40 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
!
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
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].
Proof 1.5.13 We have
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
> 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).
This completes the proof.
42 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
Proof 1.5.17 By Sard’s Theorem,2 it follows that there is a regular value u such that
Hence,
< µ.
dB ( f , D, z) = dB ( f , D, u).
≥ 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.
The Brouwer Degree 43
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 ( f , D, z) = dB (g, D, z),
µ
where g ∈ C 2 (D), max kg − f k < .
D 4
dB ( f , D, z) = dB ( f , D0 , z).
dB ( f , D, z) = dB (g, D, z),
dB ( f , D0 , z) = dB (g, D0 , z),
44 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
Hence,
dB ( f , D, z) = dB ( f , D0 , z).
This completes the proof.
dB ( f , D, z) = dB ( f , 0,
/ z)
= dB (g, 0,
/ z)
= 0.
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.
The Brouwer Degree 45
Then
dB ( f , D, z) = dB (g, D, z).
We have
≥ k f (·) − zk − λ k f − gk
≥ k f (·) − zk − k f − gk
> 0.
dB ( f , D, z) = dB (F(·, 0), D, z)
= dB (F(·, 1), D, z)
= dB (g, D, z).
= 0.
Hence,
dB ( f , D, 0) = 0.
This completes the proof.
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)
The Brouwer Degree 47
= 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
0 < µ0 = min
S k f (·) − zk
D\ j∈N D j
and
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.
48 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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, 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
dB ( f , ∂ D) = dB (g, D),
dB ( f , ∂ D) = 0.
F(x, λ ) = (1 − λ ) f˜(x) − λ y.
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.
50 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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.
Theorem 1.5.42 Let g ∈ C (Sn−1 ), g : Sn−1 → Sn−1 has no fixed points. Then
F(x, λ ) = (1 − λ )g̃(x) − λ x.
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,
The Brouwer Degree 51
1
whereupon λ = and g(x) = x. This is a contradiction. Therefore
2
= 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.
iK ( f , ∂ D) = iK (g, ∂ 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).
52 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
kTε x − T xk < ε, x ∈ Ω.
Define
m
[
F = BT x j (ε),
j=1
m
Γ(x) = ∑ φ j (x),
j=1
m
φ j (x)
Tε x = ∑ T x j.
j=1 Γ(x)
kT x j − T xk < ε.
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
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
x jk → y as k → ∞.
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
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.
dLS (I − T, Ω, 0) 6= 0,
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.
ε > kTε x − T xk
= k(I − T )xk,
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.
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)
= 1.
This completes the proof.
Theorem 1.7.4 (Solvability) If i(F, Ω, K) 6= 0, then Fx = x has a solution in Ω.
Proof 1.7.5 Let r : X → K be any retraction. Then
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
\
Dn = coA(Dn−1 U), n ∈ N, n ≥ 2.
58 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
\
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))
The Fixed-Point Index for Strict Set Contractions 59
\
≤ 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
A1 x0 = A2 x0 = Ax0 .
Therefore
x0 = t0 A1 x0 + (1 − t0 )A2 x0
= t0 Ax0 + (1 − t0 )Ax0
= Ax0 ,
60 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
which is a contradiction. By the homotopy invariance of the fixed-point index for completely
continuous operators, it follows that
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
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.
This completes the proof.
62 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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.
This completes the proof.
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.
This completes the proof.
The Fixed-Point Index for Strict Set Contractions 63
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)
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
i(A,U, X) = 0.
Definition 1.9.2 A cone P ⊂ E is said to be normal if there exists a constant N such that
A(tx) ≥ t α Ax
A(tx) ≤ t α Ax
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
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).
66 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
≥ 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
This completes the proof.
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∗ .
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)}.
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
T τ(z1 ) + z1 = τ(z1 ),
T τ(z2 ) + z2 = τ(z2 ),
we get
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∗
68 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
or
T x∗ + Sx∗ = x∗ .
This completes the proof.
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
Sx∗ + T x∗ = x∗ .
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.
(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∗ .
This completes the proof.
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)
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}.
ky2 − y1 k = kT −1 y1 − T −1 y0 k
≤ lky1 − y0 k,
ky3 − y2 k = kT −1 y2 − T −1 y1 k
70 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
≤ 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∗ .
This completes the proof.
For R > r > 0, denote
Pr = {x ∈ P : kxk < r},
∂ Pr = {x ∈ P : kxk = r},
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.
\
Fx 6= (I − T )(x − λ u0 )
72 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
i∗ (T + F, Pr Ω, P) = 0.
\
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
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 .
Multiple Fixed-Point Theorems 73
i∗ (T + F, Pr Ω, P) = i((I − T )−1 F, Pr , P)
\
= 1.
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
i∗ (T + F, Pr Ω, P) = i((I − T )−1 F, Pr , P)
\
= 0.
Fx0 = (I − T )(x0 − λ0 u0 ) ∈ Ω.
Then
x0 − λ0 u0 − Fx0 = T (x0 − λ0 u0 )
≤ 0,
Fx 6= (I − T )(x − λ u0 )
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.
i∗ (T + F, PR Ω, P) = 1,
\
i∗ (T + F, PL Ω, P) = 0.
\
= 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).
\
Multiple Fixed-Point Theorems 75
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.,
i∗ (T + F, Pr Ω, P) = 0,
\
i∗ (T + F, PR Ω, P) = 0,
\
i∗ (T + F, PL Ω, P) = 1.
\
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).
\
76 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
and
inf{kT x + Fxk : x ∈ P, kxk = ρ} > ρ. (1.21)
Fx ≥ Fi x, x ∈ P, i = 1, 2, (1.22)
(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
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
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 .
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
78 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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,
Fx2 ≤ x2 − λ0 u0 .
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,
Multiple Fixed-Point Theorems 79
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
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
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
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 },
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)
Hence,
mi = inf{kFi xk : x ∈ ∂ K , kxk = 1}
1
≥ Λi kFi (Λ0 e)kkek
N
> 0, i = 1, 2.
N3 N2
−1 = 1
−1
Λ1 kF1 (Λ0 e)kkek N Λ1 kF1 (Λ0 e)kkek
N2
≥ − 1,
m1
Multiple Fixed-Point Theorems 81
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
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
ξ1 (x,t) = t α−1 − 1,
Then
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
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.
82 Fixed-Point Index Theory. Multiple Fixed-Point Theorems
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
83
84 Applications to ODEs
Assume that the equation (2.1) has two ω-periodic solutions x1 and x2 . Set
v = x1 − x2 .
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
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
= −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]
86 Applications to ODEs
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]
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.
= (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
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
= −εR
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
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
−[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
Periodic Solutions for First Order ODEs 89
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.
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
= 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 ).
Periodic Solutions for First Order ODEs 91
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.
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},
Ω = PR .
We have
Pr,L
\
Ω 6= 0,
/
PL,R
\
Ω 6= 0.
/
We will check all conditions of Theorem 1.9.23.
Periodic Solutions for First Order ODEs 93
F(PR ) ⊂ (I − T )(PR ).
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,
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.
94 Applications to ODEs
Hence,
kFx + T 0k ≤ εkxk.
Now, assume that x1 ∈ ∂ PL PR is such that
\
x1 = T x1 + Fx1 .
L = x1 (t1 )
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.
This completes the proof.
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],
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
Proof 2.1.8 Let E and P be as in the proof of Theorem 2.1.3. Let also
Ω = Pd .
96 Applications to ODEs
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
x = T x + Fz
= Ax.
We have
= (1 + ε)kx − yk, x, y ∈ Pd .
Pd ⊂ A(Pd ).
v = Au
= Tu + Fz
or
Tu = v − Fz. (2.20)
Periodic Solutions for First Order ODEs 97
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,
98 Applications to ODEs
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
Periodic Solutions for First Order ODEs 99
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.
µm = µ(1 + ξ1 (µ)),
100 Applications to ODEs
1
1 − ξ2 (µ) = (1 + µ k−1 )
2
≥ µ k−1 ,
Next,
F1 (µx) = µ m F1 x
= µ(1 + ξ1 (µ))F1 x,
F2 (µx) = µ k F2 x
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
Periodic Solutions for First Order ODEs 101
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)
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.
102 Applications to ODEs
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
Note that
P ≤ G(t, s) ≤ M, s,t ∈ [α, β ].
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
α0 + β0 e− α a(y)dy
Rs
γ + α0 e α a(y)dy Rβ
+β0 Rβ e− α a(y)dy
α0 + β0 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
104 Applications to ODEs
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
= −a(t)x(t)
= 1, 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
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.
106 Applications to ODEs
≤ 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
Hence,
≤ 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.
BVPs for First Order ODEs 107
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
< 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)
≤ ε(ρ − R − 1)
< 0, t ∈ [α, β ],
= Ax(t), t ∈ R.
= (1 + ε)kx − yk,
PR ⊂ A(PR ). (2.38)
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
≥ 0, t ∈ [α, β ],
and
= R1 .
BVPs for First Order ODEs 109
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},
Ω = PR .
110 Applications to ODEs
We have
Pr,L
\
Ω 6= 0,
/
PL,R
\
Ω 6= 0.
/
F(PR ) ⊂ (I − T )(PR ).
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
≤ εM(M0 + M1 Lm + M2 Lk + M̃L)(β − α)
≤ εL
= εkxk, t ∈ [α, β ].
Hence,
kFx + T 0k ≤ εkxk.
Now, assume that x1 ∈ ∂ PL
\
Ω is such that
x1 = T x1 + Fx1 .
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),
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)
α0 + β0 e α b(y)dy
where
(H4) f : [α, β ] → (−∞, 0],
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
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
Ω = Pd .
114 Applications to ODEs
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
kAx − Ayk = kT x − Tyk
= (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
BVPs for First Order ODEs 115
≥ ε(d + 1)
> 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
= d1 , t ∈ [α, β ].
We have d1 > d and
P,
\
v − Fz ∈ B(d1 )
and
T : Pd → T (Pd ) = B(d1 ) P
\
≤ ερ, t ∈ [α, β ],
116 Applications to ODEs
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.
Therefore (H2) of Theorem 1.9.25 holds.
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)),
BVPs for First Order ODEs 117
1
1 − ξ2 (µ) = (1 + µ k−1 )
2
≥ µ k−1 ,
Next,
F1 (µx) = µ m F1 x
= µ(1 + ξ1 (µ))F1 x,
F2 (µx) = µ k F2 x
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
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
120 Applications to ODEs
where
(H5) f : [α, β ] × [0, ∞) → [0, ∞) is a continuous function such that
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
and
P = {x ∈ E : x(t) ≥ 0, t ∈ [α, β ]}.
On P define the operators
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
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)
where
(H6) f : [α, β ] → (−∞, 0],
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.
x(0) = g,
x0 (1) = h,
where
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
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 tm ≤ t ≤ 1, and
s if 0≤s≤t ≤1
G(t, s) =
t if 0 ≤ t ≤ s ≤ 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
d
I(t, x(t)) = 0,
dt
126 Applications to ODEs
d2
I(t, x(t)) = 0,
dt 2
Z 1
x0 (t) = h + f (s, x(s))ds,
t
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 −)),
= vk (x(tk −)).
Also,
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 −)),
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
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
m i i
≤ ∑ Vj + ∑ Vj + ∑ U j
j=i+1 j=1 j=1
m
≤ ∑ (U j +V j ),
j=1
BVPs with Impulses 129
m
d
I(t, x(t)) = − ∑ v j (x(t j −))
dt
j=i+1
m
≤ ∑ Vj.
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
130 Applications to ODEs
Z 1
+ G(t, s) f (s, x1 (s))ds
0
= −(Fx1 )(t)
≤ ε(ρ − R − 1)
< 0,
= (1 + ε)kx − yk,
PR ⊂ A(PR ). (2.80)
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].
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
Proof 2.4.6 Let T and F be defined as in the proof of Theorem 2.4.3. Define
Ω = PR .
We have
Pr,L
\
Ω 6= 0,
/
PL,R
\
Ω 6= 0.
/
F(PR ) ⊂ (I − T )(PR ).
Fx1 = x1 − λ1 u0 .
134 Applications to ODEs
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
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
kFx + T 0k ≤ εkxk.
4. Assume that
x1 = T x1 + Fx1
for some x1 ∈ ∂ PL
\
Ω. Let t ∈ [0, 1] be such that
L = x1 (t).
BVPs with Impulses 135
Then
L = x1 (t)
+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.
This completes the proof.
3
Applications to FDEs
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)).
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 .
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
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
140 Applications to FDEs
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
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.
Global Existence for a Class of Fractional-Differential Equations 141
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.
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
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+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 )
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 )
t
≤
(a − γ − 1)(t + 1)10a
1
≤ , t ≥ 0,
a−γ −1
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.
144 Applications to FDEs
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
which completes the proof.
Z t
1 1
≤ dτ2
(t + 1)10a 0 (τ2 + 1)a−γ
t
≤
(t + 1)10a
≤ 1, t ≥ 0,
(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
× dzdsdτ2 dτ1 dwdy
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
× dzdsdτ2 dτ1 dwdy
(τ2 + 1)a+1 (τ2 − s)α ds (s − z)β
0 0 0 0
146 Applications to FDEs
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 )
Global Existence for a Class of Fractional-Differential Equations 147
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
× dzdsdτ2 dτ1 dwdy
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
148 Applications to FDEs
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
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
Global Existence for a Class of Fractional-Differential Equations 149
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 )
≤ 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.
Global Existence for a Class of Fractional-Differential Equations 151
Hence, using Lemma 3.1.7, Lemma 3.1.9, Lemma 3.1.11, Lemma 3.1.13 and
A1 ≥ 1, we obtain
(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
152 Applications to FDEs
((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.
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)
We will start with the following useful lemma.
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
Z t
1
= − (t − s)α−1 f (s, x(s))ds + c1t α−1 + c2t α−2 , 0 < t < 1,
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
BVP of Nonlinear Riemann-Liouville Fractional Differential Equations 155
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
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
solutions x1 , x2 , x3 ∈ P such that
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).
156 Applications to FDEs
1. We have T : PR → E and
kT x − Tyk = (1 + ε)kx − yk
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.
3. For i = 1, 2 and x ∈ ∂ Pri , we have
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
whereupon
kFx + T 0k ≤ εkxk.
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 )
= 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.
5. Now we will prove that
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)
= (1 + ε)kx − yk,
158 Applications to FDEs
PR ⊂ A(PR ). (3.14)
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].
tion, where
T (PR ) = {x ∈ P : kxk ≤ R1 }.
Therefore there exists a u ∈ PR such that
and then (3.14) holds. Hence, and applying Theorem 1.9.8, it follows that there
exists a w ∈ PR such that
Theorem 3.2.5 Suppose (H8), (3.8), (3.15), (3.16). Then BVP (3.11), (3.12) has at least
two solutions x1 , x2 ∈ P such that
r < max x(t) < L < max x(t) < R.
t∈[0,1] t∈[0,1]
Z 1
Fx(t) = εR + ε G(t, s) f (s, x(s))ds, t ∈ [0, 1].
0
1. As in the proof of Theorem 3.3.3, we get
F(PR ) ⊂ (I − T )(PR ).
= Fx1 (t)
160 Applications to FDEs
Z 1
= εR + ε G(t, s) f (s, x(s))ds
0
ε
≤ M0 + M1 Lm + M2 Lk
Γ(α)
≤ εL
Hence,
kFx + T 0k ≤ εkxk.
Now, assume that x1 ∈ ∂ PL such that
x1 = T x1 + Fx1 .
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)
BVP of Nonlinear Riemann-Liouville Fractional Differential Equations 161
≤ εL − (1 + ε)L
= −L,
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],
Theorem 3.2.7 Assume (H2), (3.17)-(3.23). Then BVP (3.11), (3.12) has at least two so-
lutions x1 , x2 ∈ P such that
0 < kx1 k < ρ < kx2 k < d.
Proof 3.2.8 Let
Pρ = {x ∈ P : kxk < ρ},
Z 1
Fx(t) = −εd − ε G(t, s) f (s, x(s))ds, t ∈ [0, 1].
0
= Ax.
We have
kAx − Ayk = kT x − Tyk
= (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
BVP of Nonlinear Riemann-Liouville Fractional Differential Equations 163
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 )
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
164 Applications to FDEs
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
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.
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.
BVP of Nonlinear Riemann-Liouville Fractional Differential Equations 165
> 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 ,
Next,
F1 (µx) = µ m F1 x
= µ(1 + ξ1 (µ))F1 x,
166 Applications to FDEs
F2 (µx) = µ k F2 x
c 1
Dα x(t) + f (t, x(t)) = 0, 0 < t < 1, (3.26)
2
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).
168 Applications to FDEs
Theorem 3.3.3 Suppose (H1), (3.7)-(3.10). Then BVP (3.26), (3.27) has at least three
solutions x1 , x2 , x3 ∈ P such that
Theorem 3.3.4 Suppose (H8), (3.8), (3.15), (3.16). Then BVP (3.26), (3.27) has at least
two solutions x1 , x2 ∈ P such that
Theorem 3.3.5 Assume (H2), (3.17)-(3.23). Then BVP (3.26), (3.27) has at least two so-
lutions x1 , x2 ∈ P such that
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
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
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
172 Applications to Parabolic Equations
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
≤ B.
Note that
Z xZ y
S11 (u)t (t, x) = −lut (t, x) + l ut (t, z)dzdy
0 0
≤ lB + lB + 3lB + lM11
= l(5B + M11 )
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
Z t
−l uxx (τ, x)dτ
0
Z t
−l f (τ, x, u(τ, x), ux (τ, x))dτ,
0
Differentiability of Classical Solutions of Initial IVP Conditions 175
Z t
−l uxx (τ, x)dτ
0
Z t
−l f (τ, x, u(τ, x), ux (τ, x))dτ
0
Z t
+l |uxx (τ, x)|dτ
0
Z t
+l | f (τ, x, u(τ, x), ux (τ, x))|dτ
0
≤ 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
From items a, b, and c, and from Theorem 1.9.12, it follows that there is u11 ∈ K11 such that
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
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
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.
Differentiability of Classical Solutions of Initial IVP Conditions 177
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.
ut − uxx = f (t, x, u(t, x), ux (t, x)) in (0, m] × [1, 2], (4.10)
n
||u|| = max max |u(t, x)|, max |ut (t, x)|,
(t,x)∈[0,m]×[1,2] (t,x)∈[0,m]×[1,2]
o
max |ux (t, x)|, max |uxx (t, x)| .
(t,x)∈[0,m]×[1,2] (t,x)∈[0,m]×[1,2]
Let also
0
L12 = {u ∈ K12 : ||u|| ≤ (1 + l1 )B}.
We note that K12 is a closed convex subset of L12 .
For u ∈ L12 , we define the operators
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
− f (τ, z, u12 (τ, z), u12x (τ, z))dzdydτ
0 1 1
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))
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
180 Applications to Parabolic Equations
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
− f (τ, z, u13 (τ, z), u13x (τ, z))dzdydτ
0 2 2
lem
ut − uxx = f (t, x, u(t, x), ux (t, x)) in (0, m] × R,
u(0, x) = φ (x) in R.
This completes the proof.
∂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
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
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 },
182 Applications to Parabolic Equations
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).
This completes the proof.
n
+ · · · + ani (t, x)|uxn (t, x)| pi ,
Local Existence of Classical Solutions for an IBVP 183
(H2)
u0 ∈ C 2 (Rn ),
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
184 Applications to Parabolic Equations
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
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
Local Existence of Classical Solutions for an IBVP 185
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
B1 = {(x1 , x2 , . . . , xn ) ∈ Rn : 0 ≤ xi ≤ Ai , i = 1, . . . , n} .
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,
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 ),
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
Local Existence of Classical Solutions for an IBVP 187
u(0, x) = u0 (x).
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 ).
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
−∑ 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
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).
where
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
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
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
Local Existence of Classical Solutions for an IBVP 189
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,
Therefore
max |M11 (u)(t, x)| ≤ (1 + ε)P,
t∈[0,m],x∈B1
Consequently
M11 : D11 → D̃11 .
For u, v ∈ D11 we have that
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.
Let u ∈ D11 .
190 Applications to Parabolic Equations
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
Local Existence of Classical Solutions for an IBVP 191
≤ 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.
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
n
+ ∑ A1 (A2 . . . Ai−1 Ai+1 . . . An )2 Pm
i=2
192 Applications to Parabolic Equations
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 )|
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
Local Existence of Classical Solutions for an IBVP 193
n
+ ∑ (A1 . . . Ai−1 Ai+1 . . . A j−1 A j+1 . . . An )2 A j mP
i=2,i6= j
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 .
In the last inequality we used inequality (4.23) and the conditions in (H1) for
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
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
194 Applications to Parabolic Equations
m2
0 n
+m(A2 . . . An )2 G ∑ P pi + · · · + P pi
i=1
≤ P, (t, x) ∈ [0, m] × B1 .
In the last inequality we used inequality (4.24) and the conditions in (H1) for
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 )|
Z t Z xj Z sj
+ |ux j x j (τ, σ̃ j )dσ j ds j dτ
0 0 0
Local Existence of Classical Solutions for an IBVP 195
Z t Z xj Z sj
+ | f (τ, σ̃ j , u, Du)|dσ j ds j dτ
0 0 0
n
+ ∑ (A1 . . . Ai−1 Ai+1 . . . A j−1 A j+1 . . . An )2 mP
i=2,i6= j
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 .
In the last inequality we used inequality (4.25) and the conditions in (H1) for
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} ,
196 Applications to Parabolic Equations
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
−∑ u(τ, σ̃i )dσ i dsi dτ
i=2 0 A A
Z tZ xZ s
− f (τ, σ , u, Du)dσ dsdτ,
0 A A
n Z t Z x i Z si
−∑ u(τ, σ̃i )dσ i dsi dτ
i=2 0 A A
Z tZ xZ s
− f (τ, σ , u, Du)dσ dsdτ,
0 A A
A = (A1 , 0, . . . , 0),
the sets
n
K12 = u ∈ C 1 ([0, m], C 2 (B2 )) : max |u(t, x)| ≤ P,
t∈[0,m],x∈B2
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
u(0, x) = u0 (x), x ∈ B2 ,
0 ≤ xi ≤ Ai , i = 2, . . . , n,
198 Applications to Parabolic Equations
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 differentiate the last equality once in t, twice in x2 , and etc., twice in xn , we obtain
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
Local Existence of Classical Solutions for an IBVP 199
u12 (τ, A1 , σ2 , . . . , σi−1 , xi , σi+1 , . . . , σn )
−u11 (τ, A1 , σ2 , . . . , σi−1 , xi , σi+1 , . . . , σn )
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 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
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 ,
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+ ,
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
(H1)
b0 (t, x) ≤ f (t, x, u, v1 , . . . , vn )
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 ,
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).
Periodic Solutions 203
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 .
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
+ b(t, z)a(t + s)u(t + s, z)dzdy
−∞ −∞
204 Applications to Parabolic Equations
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 , (t, x) ∈ [0, ω] × Rn .
−∞ −∞
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
+ 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 , (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
+ 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, 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 −∞ −∞
Periodic Solutions 205
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, (t, x) ∈ [0, ω] × Rn .
−∞ −∞
−Gu(t, x) ≥ R
n
−M nωAQ + ωAMQ + M A + AQm + AQk + ∑ A(Qn j + Ql j ) ,
j=1
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)
−∞ −∞ −∞ −∞
206 Applications to Parabolic Equations
× 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
+ b(t, z)a(t + s)u(t + s, z)dzdy + 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 dzdy ds
j=1 j=1
Z x Z y
+ b(t, z)u(t, z)dzdy
−∞ −∞
≤ R + (1 + A)Q
n
m k nj lj
+M nωAQ + ωAMQ + M A + AQ + AQ + ∑ A(Q + Q ) ,
j=1
∂
Gu(t, x) = −uxi (t, x)
∂ xi
e−[a]ω xˆi
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 xˆi Z y˜x Z xˆi Z y˜x
i i
+ b(t, z)a(t + s)u(t + s, z)dzdyi + b(t, z)
−∞ −∞ −∞ −∞
i
× f (t + s, z, u(t + s, z), ux1 (t + s, z), . . . , uxn (t + s, z))dzdy ds
Z xˆi Z y˜x
i
− b(t, z)u(t, z)dzdyi ,
−∞ −∞
Periodic Solutions 207
∂
∂ xi Gu(t, x) ≤ |uxi (t, x)|
e−[a]ω xˆi
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 xˆi Z y˜x Z xˆi 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 xˆi Z y˜x
i
+ b(t, z)u(t, z)dzdyi
−∞ −∞
≤ (1 + A)Q
n
m k nj lj
+M nωAQ + ωAMQ + M A + AQ + AQ + ∑ A(Q + Q ) ,
j=1
∂2
Gu(t, x) = −uxi xi (t, x)
∂ xi2
e−[a]ω xˆi
Z ω R t+s
Z Z yˆi
a(τ)dτ
+ et b(t, z˜xi )∆u(t + s, z˜xi )dzi dyi
1 − e−[a]ω 0 −∞ −∞
Z xˆi Z yˆi Z xˆi 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 xˆi Z yˆi
− b(t, z˜xi )u(t, z˜xi )dzi dyi ,
−∞ −∞
2
∂
∂ x2 Gu(t, x) ≤ |uxi xi (t, x)|
i
e−[a]ω xˆi
Z ω R t+s
Z Z yˆi
a(τ)dτ
+ et b(t, z˜xi )|∆u(t + s, z˜xi )|dzi dyi
1 − e−[a]ω 0 −∞ −∞
Z xˆi Z yˆi Z xˆi Z yˆi
i i
+ b(t, z˜xi )a(t + s)u(t + s, z˜xi )dz dy + b(t, z˜xi )
−∞ −∞ −∞ −∞
208 Applications to Parabolic Equations
× 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 xˆi Z yˆi
+ b(t, z˜xi )u(t, z˜xi )dzi dyi
−∞ −∞
≤ (1 + A)Q
n
m k nj lj
+M nωAQ + ωAMQ + M A + AQ + AQ + ∑ A(Q + Q ) ,
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
+ 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
−∞ −∞
≥ 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
−∞ −∞
Periodic Solutions 209
n
≥ R − M nωAQ + ωAMQ + M A + AQm + AQk + ∑ A(Qn j + Ql j ) ,
j=1
Theorem 4.3.5 Suppose (4.33)-(4.36), (H1)-(H3). Then equation (4.37) has at least three
solutions x1 , x2 , x3 ∈ P such that
< sε
1
< ε.
2
As in the proof of Theorem 3.2.3. we have that F : PR → E is 0-set contraction.
210 Applications to Parabolic Equations
Z x Z y
− b(t, z)u(t, z)dzdy
−∞ −∞
−[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
+ 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
−∞ −∞
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
+ 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
−∞ −∞
Periodic Solutions 211
n
m k nj lj
≤ εM nωAri + ωAMri + M A + Ari + Ari + ∑ A ri + ri
j=1
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)
= ε(u1 − λ1 u0 )(t, x) + εR
whereupon
ε
ρ< R.
p−ε
This is a contradiction. Therefore condition (b) of Theorem 1.9.21 holds.
212 Applications to Parabolic Equations
= (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
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
Define
P
f = {u ∈ E : u(t, x) ≥ 0, (t, x) ∈ [0, T ] × Ω}.
Let P be the set of all equi-continuous families in P.
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).
Multiple Solutions for an IBVP with Robin Boundary Conditions 215
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 .
Proof 4.4.3 Define the conical shells
PR = {u ∈ P : kuk < R},
≤ 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
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
≤ −εR + εφ (x)
≤ −Fu1 (t, x)
≤ εu1 (t, x) − εR
≤ ε(ρ − R)
For u, v ∈ PR , we obtain
= (1 + ε)ku − vk,
Multiple Solutions for an IBVP with Robin Boundary Conditions 217
PR ⊂ A(PR ). (4.48)
≥ v(t, x) − εR + εφ (x)
≥ v(t, x) − εR + εB0
≥ 0, (t, x) ∈ [0, T ] × Ω,
and
= R1 , (t, x) ∈ [0, T ] × Ω.
P, and
\
Note that R1 > R and v − Fz ∈ B(R1 )
T : PR → B(R1 ) P
\
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,R PR
\
6= 0.
/
To prove our result we will use Theorem 1.9.23.
1. As in the proof of Theorem 4.4.2, we get
F(PR ) ⊂ (I − T )(PR ).
≥ u1 (t, x) − λ1 u0 (t, x)
Multiple Solutions for an IBVP with Robin Boundary Conditions 219
= Fu1 (t, x)
≤ εL
= (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)
whereupon
ε(L − B0 ) ≥ 0.
This is a contradiction. Therefore condition (b) of Theorem 1.9.23 holds.
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.
This completes the proof.
5
Applications to Hyperbolic Equations
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
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
utt (t, x) − uxx (t, x) = f (t, x, u(t, x), ut (t, x), ux (t, x)), t ∈ [a, b], x ∈ [c, d],
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
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
max |ux (t, x)|, max |uxx (t, x)| .
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,
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
= 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
Differentiability of Classical Solutions of an IVP for Hyperbolic Equations 227
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
= l(5B + 2M11 )
≤ B,
Z xZ y
S11 (u)tt (t, x) = −lutt (t, x) + l utt (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
= l(4B + 3M11 )
≤ B,
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
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(3B + 3M11 )
≤ B.
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)
Let E12 = C 2 ([0, 1], C 2 ([1, 2])) be endowed with the norm
n
||u|| = max max |u(t, x)|, max |ut (t, x)|,
(t,x)∈[0,1]×[1,2] (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
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
232 Applications to Hyperbolic Equations
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
= u11tt (t, 1) − f (t, 1, u11 (t, 1), u11t (t, 1), u11x (t, 1))
utt − uxx = f (t, x, u(t, x), ut (t, x), ux (t, x)) in (0, 1] × [0, 2],
u21xx (1, x) = u1xx (1, x) = u11xx (1, x), u21t (1, x) = u1t (1, x) = u11t (1, x),
with the convention um0 (τ, 0) = um1 (τ, 0), m ∈ N, u0 (0, z) = u1 (0, z), u0t (0, z) = u1t (0, z).
This completes the proof.
∂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
∂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.
236 Applications to Hyperbolic Equations
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
− (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 ∂ ut
∂u
× (τ, z, φ (z), ψ(z))dzdydτdt1
∂φ t
Differentiability of Classical Solutions of an IVP for Hyperbolic Equations 237
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 ∂ ux
∂u
× (τ, z, φ (z), ψ(z))dzdydτdt1
∂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
− (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
×v(τ, z)dzdydτdt1
Z t Z t1 Z x Z y (5.22)
∂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 ∂ ut
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
238 Applications to Hyperbolic Equations
∂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.
u(0, x) = φ (x), x ∈ Ω,
(5.26)
ut (0, x) = ψ(x), x ∈ Ω,
∂u
= au + g on [0, T ] × ∂ Ω,
∂n
where Ω ⊂ Rn is a bounded open set, ∂ Ω is smooth,
Z
dσ ≤ A,
∂Ω
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
Define
P
f = {u ∈ E : u(t, x) ≥ 0, (t, x) ∈ [0, T ] × Ω}.
Let P be the set of all equi-continuous families in P.
f
For u ∈ E define the operators
Note that any fixed point u ∈ E of the operator T + F is a solution of IBVP (5.26).
240 Applications to Hyperbolic Equations
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
1. Let u, v ∈ P. Then
≤ 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 + εφ (x)
≤ −Fu1 (t, x)
≤ ε(ρ − R)
= (1 + ε)ku − vk,
242 Applications to Hyperbolic Equations
PR ⊂ A(PR ). (5.29)
≥ v(t, x) − εR + εφ (x)
≥ v(t, x) − εR + εB0
and
= R1 , (t, x) ∈ [0, T ] × Ω.
P, and
\
Note that R1 > R and v − Fz ∈ B(R1 )
T : PR → B(R1 ) P
\
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
We have
Pr,L PR
\
6= 0,
/
PL,R PR
\
6= 0.
/
F(PR ) ⊂ (I − T )(PR ).
r ≥ u1 (t, x) − λ1 u0 (t, x)
= Fu1 (t, x)
≤ εL
Hence,
kFu + T 0k ≤ εkuk.
Now, assume that there is a u1 ∈ ∂ PL
\
Ω such that
u1 = Tu1 + Fu1 .
L = u1 (t1 , x1 ).
Then
L = u1 (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
Periodic Solutions 245
≤ (1 + ε)L − εφ (x)
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
246 Applications to Hyperbolic Equations
∂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 } .
E n+2 = E × . . . × E,
| {z }
n+2
P n+2 = P ×...×P .
| {z }
n+2
(H2) Suppose that a, a j : R → [0, ∞), j ∈ {0, . . . , n} are continuous ω-periodic functions
such that
N ≤ a(t) ≤ M,
N ≤ a j (t) ≤ M,
(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
= ... ,
−∞ −∞ −∞
248 Applications to Hyperbolic Equations
Z xˆl Z x1 Z xl−1 Z xl+1 Z xn
= ... ... ,
−∞ −∞ −∞ −∞ −∞
dz = dzn . . . dz1 ,
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
Periodic Solutions 249
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
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},
−∞
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
+ 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 ,
−∞
250 Applications to Hyperbolic Equations
T̃ = (T, T0 , T1 , . . . , Tn ),
−Fu(t, x) ≥ εN − εR,
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 + εR + εAQ(ωM(1 + M) + 1) + εM,
252 Applications to Hyperbolic Equations
−[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 + εAQ(ωM(1 + M) + 1),
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
+ 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
−∞
≤ 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
Periodic Solutions 253
∂ ∂ ∂
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
−∞
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
PRn+2 = PR × . . . × PR ,
| {z }
n+2
Pρn+2 = Pρ × . . . × Pρ ,
| {z }
n+2
Prn+2
i
= Pri × . . . × Pri .
| {z }
n+2
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,
Hence,
kT̃ u − T̃ vk = (1 + ε)ku − vk,
Periodic Solutions 255
2εR + εM + εAR(nωM + 1)
!
n
2 n mj
+εM 1 + AR + A ∑ R kuk,
j=0
≤ εskuk
< εkuk.
such that
F̃u1 = (I − T )(u1 − λ1 u0 ).
256 Applications to Hyperbolic Equations
Hence,
0 < εN − εAQ − εR
≤ −F̃u1 (t, x)
≤ ερ − εR + εM
< 0.
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 .
This completes the proof.
6
Applications to Elliptic Equations
257
258 Applications to Elliptic Equations
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
Define
P
f = {u ∈ E : u(x) ≥ 0, x ∈ Ω}.
Let P be the set of all equi-continuous families in P.
f
For u ∈ E define the operators
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
1. Let u, v ∈ P. Then
2. For u ∈ PR , we have
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
< sε, x ∈ Ω.
< εri , x ∈ Ω.
Fu1 = (I − T )(u1 − λ1 u0 ).
Then
Z
−εR + 2εB0 ≤ −εR + εφ (x) + ε g(x)dσ
∂Ω
≤ −Fu1 (x)
260 Applications to Elliptic Equations
≤ ε(ρ + A − R), x ∈ Ω,
whereupon
2B0 ≤ ρ + A.
This is a contradiction. Therefore condition (b) of Theorem 1.9.21 holds.
5. Now we will prove that
= Au(x), x ∈ Ω.
For u, v ∈ PR , we obtain
= (1 + ε)ku − vk,
PR ⊂ A(PR ). (6.7)
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)
Multiple Solutions for a BVP with Robin Boundary Conditions 261
≥ v(x) − εR + εB0
≥ 0, x ∈ Ω,
and
v(x) − Fz(x) ≤ R + ε(R + 3A + aRA + ARm + ARk )
= R1 , x ∈ 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 (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 .
This completes the proof.
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)
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},
We have
Pr,L PR
\
6= 0,
/
PL,R PR
\
6= 0.
/
F(PR ) ⊂ (I − T )(PR ).
Fu1 = u1 − λ1 u0 .
r ≥ u1 (x)
≥ u1 (x) − λ1 u0 (x)
= Fu1 (x)
≤ εL
= εkuk, x ∈ Ω.
Multiple Solutions for a BVP with Robin Boundary Conditions 263
Hence,
kFu + T 0k ≤ εkuk.
Now, assume that there is a u1 ∈ ∂ PL such that
u1 = Tu1 + Fu1 .
L = u1 (x1 )
≤ εR − εB0
< 0, x ∈ Ω.
where
ui0 ∈ C0 (R ),
n
∞
suppui0 ⊂ B,
(H1)
|∂xα ui0 (x)| ≤ Cα,K (1 + |x|)−K on Rn , i ∈ {1, . . . , n},
(H2)
|∂t ∂x fi (t, x)| ≤ CαmK (1 + |x| + t)−K on [0, ∞) × Rn ,
m α
i ∈ {1, . . . , n},
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
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 ,
σ̃x j = σ1 , . . . , σ j−1 , x j , σ j+1 , . . . , σn ,
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
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
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}.
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 .
268 Applications to Elliptic Equations
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
|vqxi (t1 , x1 ) − vqxi (t2 , x2 )| < ε, |vqxi x j (t1 , x1 ) − vqxi x j (t2 , x2 )| < ε
Existence and Smoothness of Navier-Stokes Equations 269
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
1
Mn+1 (u, p)(t, x) = (1 + l)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 .
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
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
Existence and Smoothness of Navier-Stokes Equations 271
≤ lQ + ln(B∗ )2 Q
= l Q + n(B∗ )2 Q
≤
Q, (t, x) ∈ [0, 1] × Rn .
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
272 Applications to Elliptic Equations
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
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
≤ Q, (t, x) ∈ [0, 1] × Rn .
∂ 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
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
274 Applications to Elliptic Equations
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
≤ Q, (t, x) ∈ [0, 1] × Rn .
Z t Z x m Z sm
+l um (τ, σ̃xm ) dσ m dsm dτ, (t, x) ∈ [0, 1] × Rn .
0 x0m x0m
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
≤ 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
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
276 Applications to Elliptic Equations
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
≤ Q, (t, x) ∈ [0, 1] × Rn .
∂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
Existence and Smoothness of Navier-Stokes Equations 277
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
≤ Q, , (t, x) ∈ [0, 1] × Rn ,
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
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
Existence and Smoothness of Navier-Stokes Equations 279
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
≤ 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
280 Applications to Elliptic Equations
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
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
Existence and Smoothness of Navier-Stokes Equations 281
+l(B∗ )2 Q + l(B∗ )2 M 1
≤ Q, (t, x) ∈ [0, 1] × Rn .
∂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
282 Applications to Elliptic Equations
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
≤ Q, (t, x) ∈ [0, 1] × Rn .
Existence and Smoothness of Navier-Stokes Equations 283
∂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
Z t Z xm Z sm
∂
+l ∂ xm um (τ, σ̃xm ) dσ m dsm dτ
0 x0m x0m
= l Q + n(B∗ )2 Q
≤ Q, (t, x) ∈ [0, 1] × Rn .
∂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
284 Applications to Elliptic Equations
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
= l Q + n(B∗ )2 Q
≤ Q, (t, x) ∈ [0, 1] × Rn .
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
Let n
E2 = v = (v1 , . . . , vn+1 ) : vi ∈ C 1 ([1, 2], C02 (Rn )), suppx vi ⊂ B,
o
i ∈ {1, . . . , n + 1}
286 Applications to Elliptic Equations
2
Q2 = {v ∈ K : ||v|| ≤ (1 + l1 )Q}.
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
Existence and Smoothness of Navier-Stokes Equations 287
(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
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
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
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
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],
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 ,
290 Applications to Elliptic Equations
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
E3 = v = (v1 , . . . , vn+1 ) : vi ∈ C 1 ([2, 3], 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∈[2,3],x∈B t∈[2,3],x∈B
3
Q3 = {v ∈ K : ||v|| ≤ (1 + l2 )Q}.
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
Existence and Smoothness of Navier-Stokes Equations 291
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
3
Mn+1 (u, p)(t, x) = (1 + l2 )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.
Consequently, the function
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
Therefore Z Z
|ũ(t, x)|2 dx = |ũ(t, x)|2 dx ≤ nQ2 µ(B) < ∞.
Rn B
This completes the proof.
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
295
296 Index
solutions for IBVP with Robin periodic solutions for first order
boundary conditions, 238–245 ODEs, 83–101