Series in Mathematical Analysis and Applications

Edited by Ravi P. Agarwal and Donal O’Regan

VOLUME 10

TOPOLOGICAL
DEGREE THEORY
AND APPLICATIONS

Copyright 2006 by Taylor & Francis Group, LLC

 SERIES IN MATHEMATICAL ANALYSIS AND APPLICATIONS
Series in Mathematical Analysis and Applications (SIMAA) is edited by Ravi P. Agarwal,
Florida Institute of Technology, USA and Donal O’Regan, National University of Ireland,
Galway, Ireland.
The series is aimed at reporting on new developments in mathematical analysis and
applications of a high standard and or current interest. Each volume in the series is devoted
to a topic in analysis that has been applied, or is potentially applicable, to the solutions of
scientific, engineering and social problems.

Volume 1
Method of Variation of Parameters for Dynamic Systems
V. Lakshmikantham and S.G. Deo

Volume 2
Integral and Integrodifferential Equations: Theory, Methods and Applications
Edited by Ravi P. Agarwal and Donal O’Regan

Volume 3
Theorems of Leray-Schauder Type and Applications
Donal O’Regan and Radu Precup

Volume 4
Set Valued Mappings with Applications in Nonlinear Analysis
Edited by Ravi P. Agarwal and Donal O’Regan

Volume 5
Oscillation Theory for Second Order Dynamic Equations
Ravi P. Agarwal, Said R. Grace, and Donal O’Regan

Volume 6
Theory of Fuzzy Differential Equations and Inclusions
V. Lakshmikantham and Ram N. Mohapatra

Volume 7
Monotone Flows and Rapid Convergence for Nonlinear Partial Differential Equations
V. Lakshmikantham, S. Koksal, and Raymond Bonnett

Volume 8
Nonsmooth Critical Point Theory and Nonlinear Boundary Value Problems
Leszek Gasiński and Nikolaos S. Papageorgiou

Volume 9
Nonlinear Analysis
Leszek Gasiński and Nikolaos S. Papageorgiou

Volume 10
Topological Degree Theory and Applications
Donal O’Regan, Yeol Je Cho, and Yu-Qing Chen

Copyright 2006 by Taylor & Francis Group, LLC

 Series in Mathematical Analysis and Applications
Edited by Ravi P. Agarwal and Donal O’Regan

VOLUME 10

TOPOLOGICAL
DEGREE THEORY
AND APPLICATIONS

Donal O’Regan
Yeol Je Cho
Yu-Qing Chen

Boca Raton London New York

Copyright 2006 by Taylor & Francis Group, LLC

 C648X_Discl.fm Page 1 Wednesday, February 22, 2006 9:37 AM

Published in 2006 by
Chapman & Hall/CRC
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2006 by Taylor & Francis Group, LLC

Chapman & Hall/CRC is an imprint of Taylor & Francis Group
No claim to original U.S. Government works
Printed in the United States of America on acid-free paper
10 9 8 7 6 5 4 3 2 1
International Standard Book Number-10: 1-58488-648-X (Hardcover)
International Standard Book Number-13: 978-1-58488-648-8 (Hardcover)
This book contains information obtained from authentic and highly regarded sources. Reprinted material is
quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts
have been made to publish reliable data and information, but the author and the publisher cannot assume
responsibility for the validity of all materials or for the consequences of their use.
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, please access www.copyright.com
(http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive,
Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration
for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate
system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only
for identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Catalog record is available from the Library of Congress

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com
Taylor & Francis Group and the CRC Press Web site at
is the Academic Division of Informa plc. http://www.crcpress.com

Copyright 2006 by Taylor & Francis Group, LLC

 Contents

1 BROUWER DEGREE THEORY 1

1.1 Continuous and Differentiable Functions . . . . . . . . . . . 2
1.2 Construction of Brouwer Degree . . . . . . . . . . . . . . . . 4
1.3 Degree Theory for Functions in VMO . . . . . . . . . . . . . 15
1.4 Applications to ODEs . . . . . . . . . . . . . . . . . . . . . . 19
1.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2 LERAY SCHAUDER DEGREE THEORY 25

2.1 Compact Mappings . . . . . . . . . . . . . . . . . . . . . . . 25
2.2 Leray Schauder Degree . . . . . . . . . . . . . . . . . . . . . 30
2.3 Leray Schauder Degree for Multi-Valued Mappings . . . . . . 38
2.4 Applications to Bifurcations . . . . . . . . . . . . . . . . . . 43
2.5 Applications to ODEs and PDEs . . . . . . . . . . . . . . . . 46
2.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3 DEGREE THEORY FOR SET CONTRACTIVE MAPS 55

3.1 Measure of Noncompactness and Set Contractions . . . . . . 55
3.2 Degree Theory for Countably Condensing Mappings . . . . . 65
3.3 Applications to ODEs in Banach Spaces . . . . . . . . . . . . 68
3.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

4 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS

75
4.1 A-Proper Mappings . . . . . . . . . . . . . . . . . . . . . . . 75
4.2 Generalized Degree for A-Proper Mappings . . . . . . . . . . 80
4.3 Equations with Fredholm Mappings of Index Zero . . . . . . 82
4.4 Equations with Fredholm Mappings of Index Zero Type . . . 87
4.5 Applications of the Generalized Degree . . . . . . . . . . . . 95
4.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

5 COINCIDENCE DEGREE THEORY 105

5.1 Fredholm Mappings . . . . . . . . . . . . . . . . . . . . . . . 105
5.2 Coincidence Degree for L-Compact Mappings . . . . . . . . . 110
5.3 Existence Theorems for Operator Equations . . . . . . . . . 116
5.4 Applications to ODEs . . . . . . . . . . . . . . . . . . . . . . 119
5.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

Copyright 2006 by Taylor & Francis Group, LLC

 0

6 DEGREE THEORY FOR MONOTONE-TYPE MAPS 127

6.1 Monotone Type-Mappings in Reflexive Banach Spaces . . . . 128
6.2 Degree Theory for Mappings of Class (S+ ) . . . . . . . . . . 142
6.3 Degree for Perturbations of Monotone-Type Mappings . . . . 145
6.4 Degree Theory for Mappings of Class (S+ )L . . . . . . . . . 149
6.5 Coincidence Degree for Mappings of Class L-(S+ ) . . . . . . 152
6.6 Computation of Topological Degree . . . . . . . . . . . . . . 156
6.7 Applications to PDEs and Evolution Equations . . . . . . . 159
6.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

7 FIXED POINT INDEX THEORY 169

7.1 Cone in Normed Spaces . . . . . . . . . . . . . . . . . . . . . 169
7.2 Fixed Point Index Theory . . . . . . . . . . . . . . . . . . . . 176
7.3 Fixed Point Theorems in Cones . . . . . . . . . . . . . . . . 179
7.4 Perturbations of Condensing Mappings . . . . . . . . . . . . 186
7.5 Index Theory for Nonself Mappings . . . . . . . . . . . . . . 189
7.6 Applications to Integral and Differential Equations . . . . . . 191
7.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

8 REFERENCES 195

Copyright 2006 by Taylor & Francis Group, LLC

 Chapter 1
BROUWER DEGREE THEORY

Let R be the real numbers, Rn = {x = (x1 , x2 , · · · , xn ) : xi ∈ R for i =

Pn 1
1, 2, · · · , n} with |x| = ( i=1 x2i ) 2 and let Ω ⊂ Rn , and let f : Ω → Rn
be a continuous function. A basic mathematical problem is: Does f (x) = 0
have a solution in Ω? It is also of interest to know how many solutions are
distributed in Ω. In this chapter, we will present a number, the topological
degree of f with respect to Ω and 0, which is very useful in answering these
questions. To motivate the process, let us first recall the winding number of
plane curves, a basic topic in an elementary course in complex analysis. Let
C be the set of complex numbers, Γ ⊂ C an oriented closed C 1 curve and
a ∈ C \ Γ. Then the integer
Z
1 1
w(Γ, a) = dz (1)
2πi Γ z − a
is called the winding number of Γ with respect to a ∈ C \ Γ. Now, let G ⊂ C
be a simply connected region and f : G → C be analytic and Γ ⊂ G a closed
C 1 curve such that f (z) 6= 0 on Γ. Then we have
Z Z 0
1 1 1 f (z) X
w(f (Γ), 0) = dz = dz = w(Γ, zi )αi , (2)
2πi f (Γ) z 2πi Γ f (z) i

where zi are the zeros of f in the region enclosed by Γ and αi are the corre-
sponding multiplicites. If we assume in addition that Γ has positive orientation
and no intersection points, then we know from Jordan’s Theorem, which will
be proved later in this chapter, that w(Γ, zi ) = 1 for all zi . Thus (2) becomes
X
w(f (Γ), 0) = αi . (3)
i

So we may say that f has at least |w(f (Γ), 0)| zeros in G. The winding number
is a very old concept which goes back to Cauchy and Gauss. Kronecker,
Hadamard, Poincare, and others extended formula (1). In 1912, Brouwer [32]
introduced the so-called Brouwer degree in Rn (see Browder [35], Sieberg [277]
for historical developments). In this chapter, we introduce the Brouwer degree
theory and its generalization to functions in V M O. This chapter is organized
as follows:
In Section 1.1 we introduce the notion of a critical point for a differentiable
function f . We then prove Sard’s Lemma, which states that the set of critical

Copyright 2006 by Taylor & Francis Group, LLC

 2 Topological Degree Theory and Applications

points of a C 1 function is “small”. Our final result in this section shows how
a continuous function can be approximated by a C ∞ function.
In Section 1.2 we begin by defining the degree of a C 1 function using the
Jacobian. Also we present an integral representation which we use to define
the degree of a continuous function. Also in this section we present some
properties of our degree (see theorems 1.2.6, 1.2.12, and 1.2.13) and some
useful consequences. For example, we prove Brouwer’s and Borsuk’s fixed
point theorem, Jordan’s separation theorem and an open mapping theorem. In
addition we discuss the relation between the winding number and the degree.
In Section 1.3 we discuss some properties of the average value function and
then we introduce the degree for functions in V M O.
In Section 1.4 we use the degree theory in Section 1.2 to present some exis-
tence results for the periodic and anti-periodic first order ordinary differential
equations.

1.1 Continuous and Differentiable Functions

We begin with the following Bolzano’s intermediate value theorem:

Theorem 1.1.1. Let f : [a, b] → R be a continuous function, then, for m

between f (a) and f (b), there exists x0 ∈ [a, b] such that f (x0 ) = m.

Corollary 1.1.2. Let f : [a, b] → R be a continuous function such that

f (a)f (b) < 0. Then there exists x0 ∈ (a, b) such that f (x0 ) = 0.

Corollary 1.1.3. Let f : [a, b] → [a, b] be a continuous function. Then

there exists x0 ∈ [a, b] such that f (x0 ) = x0 .

Let Ω ⊂ Rn be an open subset. We recall that a function f : Ω → Rn

is differentiable at x0 ∈ Ω if there is a matrix f 0 (x0 ) such that f (x0 + h) =
f (x0 ) + f 0 (x0 )h + o(h), where x0 + h ∈ Ω and |o(h)|
|h| tends to zero as |h| → 0.
We use C k (Ω) to denote the space of k-times continuously differentiable
functions. If f is differentiable at x0 , we call Jf (x0 ) = detf 0 (x0 ) the Jacobian
of f at x0 . If Jf (x0 ) = 0, then x0 is said to be a critical point of f and we use
Sf (Ω) = {x ∈ Ω : Jf (x) = 0} to denote the set of critical points of f , in Ω. If
f −1 (y) ∩ Sf (Ω) = ∅, then y is said to be a regular value of f . Otherwise, y is
said to be a singular value of f .

Lemma 1.1.4. (Sard’s Lemma) Let Ω ⊂ Rn be open and f ∈ C 1 (Ω).

Then µn (f (Sf (Ω)) = 0, where µn is the n-dimensional Lebesgue measure.

Proof. Since Ω is open, Ω = ∪∞ i=1 Qi , where Qi is a cube for i = 1, 2, · · · .

We only need to show that µn (f (Sf (Q))) = 0 for a cube Q ⊂ Ω. In fact, let

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 3

l be the lateral length of Q. By the uniform continuity of f 0 on Q, for any

given  > 0, there exists an integer m > 0 such that

|f 0 (x) − f 0 (y)| ≤ 
√
nl
for all x, y ∈ Q with |x − y| ≤ m . Therefore, we have
Z 1
0
|f (x) − f (y) − f (y)(x − y)| ≤ |f 0 (y + t(x − y)) − f 0 (y)||x − y|dt
0
≤ |x − y|
√
for all x, y ∈ Q with |x − y| ≤ mnl . We decompose Q into r cubes, Qi , of
√
diameter mnl , i = 1, 2, · · · , r. Since ml is the lateral length of Qi , we have
r = mn . Now, suppose that Qi ∩Sf (Ω) 6= ∅. Choosing y ∈ Qi ∩Sf (Ω), we have √
f (y+x)−f (y) = f 0 (y)x+R(y, x) for all x ∈ Qi −y, where |R(y, x+y)| ≤  mnl .
Therefore, we have

f (Qi ) = f (y) + f 0 (y)(Qi − y) + R(y, Qi ).

But f 0 (y) = 0, so f 0 (y)(Qi −y) is contained in an (n−1)- dimensional subspace

of Rn . Thus, µn (f 0 (y)(Qi − y)) = 0, so we have
√
i n n nl n
µn (f (Q )) ≤ 2  ( ) .
m
Obviously, f (Sf (Q)) ⊂ ∪ri=1 f (Qi ), so we have
√
n n nl n √
µn (f (Sf (Q)) ≤ r2  ( ) = 2n n ( nl)n .
m
By letting  → 0+ , we obtain µn (f (Sf (Q))) = 0. Therefore, µn (f (Sf (Ω))) =
0. This completes the proof.
Proposition 1.1.5. Let K ⊂ Rn be a bounded closed subset, and f : K →
R continuous. Then there exists a continuous function f˜ : Rn → convf (K)
n

such that f˜(x) = f (x) for all x ∈ K, where convf (K) is the convex hull of
f (K).
Proof. Since K is bounded closed subset, there exists at most countable
{ki : i = 1, 2, · · · } ⊂ K such that {ki : i = 1, 2, · · · } = K. Put
|x − ki |
d(x, K) = inf |x − y|, αi (x) = max{2 − , 0}
y∈K d(x, A)
for any x ∈
/ K and

P
(
f (x), x ∈ K,
f˜(x) = P2
i≥1
−i

i≥1 2
αi (x)f (ki )
−i α (x)
i
, x∈
/ K.

Copyright 2006 by Taylor & Francis Group, LLC

 4 Topological Degree Theory and Applications

Then f˜ is the desired function.

Proposition 1.1.6. Let K ⊂ Rn be a bounded closed subset and f :
K → Rn continuous. Then there exists a function g ∈ C ∞ (Rn ) such that
|f (x) − g(x)| < .
Proof. By Proposition 1.1.5, there exists a continuous extension f˜ of f to
Rn . Define the following function
( 1
ce− 1−|x| , |x| < 1,
φ(x) = (1.1)
0, |x| ≥ 1,

where c satisfies Rn φ(x)dx = 1. Set φλ (x) = λ−n φ( λx ) for all x ∈ Rn and

R

Z
fλ (x) = f˜(y)φλ (y − x)dx for all x ∈ Rn , λ > 0.
Rn

It is obvious that suppfλ = {x ∈ Rn : fλ (x) 6= 0} = {x : |x| ≤ λ} for all

λ > 0. Consequently, we have fλ ∈ C ∞ and fλ (x) → f (x) uniformly on K
as λ → 0+ . Taking g as fλ for sufficiently small λ, g is the desired function.
This completes the proof.

1.2 Construction of Brouwer Degree

Now, we give the construction of Brouwer degree in this section as follows:
Definition 1.2.1. Let Ω ⊂ RN be open and bounded and f ∈ C 1 (Ω). If
p∈
/ f (∂Ω) and Jf (p) 6= 0, then we define
X
deg(f, Ω, p) = sgnJf (x),
x∈f −1 (p)

where deg(f, Ω, p) = 0 if f −1 (p) = ∅.

The next result gives another equivalent form of Definition 1.2.1.
Proposition 1.2.2. Let Ω, f and p be as in Definition 1.2.1 and let
1
(
−
c−n e 1−|−1 x|2 , |x| < 1,
φ (x) = (1.2)
0, otherwise,
R
where c is a constant such that Rn φ(x) = 1. Then there exists 0 = 0 (p, f )
such that
Z
deg(f, Ω, p) = φ (f (x) − p)Jf (x)dx for all  ∈ (0, 0 ).
Ω

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 5

Proof. The case f −1 (p) = ∅ is obvious. Assume that

f −1 (p) = {x1 , x2 , · · · , xn }.

We can find disjoint balls Br (xi ) and a neighborhood Vi of p such that f :

Br (xi ) → Vi is a homeomorphism and sgnJf (x) = sgnJf (xi ) in Bi (xi ). We
may take r0 > 0 such that Br0 (p) ⊂ ∩ni=1 Vi and set Ui = Br (xi )∩f −1 (Br0 (p)).
Then |f (x) − p| ≥ δ on Ω\ ∪ni=1 Ui for some δ > 0 and so, for any  < δ, we
have
Z n
X Z
φ (f (x) − p)Jf (x)dx = sgnJf (xi ) φ (f (x) − p)|Jf (x)|dx.
Ω i=1 Ui

But we have
Jf (x) = Jf −p (x),
Z Z
φ (f (x) − p)|Jf (x)|dx = φ (x)dx = 1,
Ui Br0

 < min{r0 , δ}.

This completes the proof.

Definition 1.2.3. Let Ω ⊂ RN be open and bounded and f ∈ C 2 (Ω). If
p∈
/ f (∂Ω). Then we define

deg(f, Ω, p) = deg(f, Ω, p0 ),

where p0 is any regular value of f that |p0 − p| < d(p, f (∂Ω)).

We need to check that, for any two regular values p1 and p2 of f ,

deg(f, Ω, p1 ) = deg(f, Ω, p2 ).

For any  < d(p, f (∂Ω)) − max{|p − pi | : i = 1, 2}, we have

Z
deg(f, Ω, pi ) = φ (f (x) − pi )Jf (x)dx for i = 1, 2.
Ω

Notice that
φ (x − p2 ) − φ (x − p1 ) = divw(x),
where Z 1
w(x) = (p1 − p2 ) φ (x − p1 + t(p1 − p2 ))dt.
0

We show that there exists a function v ∈ C 1 (RN ) such that supp(v) ⊂ Ω and

[φ (f (x) − p2 ) − φ (f (x) − p1 )]Jf (x) = divv(x) for all x ∈ Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 6 Topological Degree Theory and Applications

Lemma 1.2.4. Let Ω ⊂ RN be open, f ∈ C 2 (Ω) and let dij be the cofactor
∂f
of ∂xji in Jf (x) and
 PN
j=1 wj (f (x))dij (x) x ∈ Ω,
vi (x) =
0, otherwise.

Then (v1 (x), v1 (x), · · · , vN (x)) satisfies divv(x) = divw(f (x))Jf (x).

Proof. Since supp(w) ⊂ B(p, r) for r ≤ max{|p − pi | : i = 1, 2} +  <

d(p, ∂Ω), we have
supp(v) ⊂ Ω,
N
X N
X
∂i vi (x) = djk ∂k Wj (f (x))∂i fk (x) + Wj (f (x))∂i dij (x),
j,k=1 j=1

∂
where ∂k = ∂xk . Now, we claim that
X
N
∂i dij (x) = 0 for j = 1, 2, · · · , N.
i=1

For any given j, let fxk denote the column

(∂k f1 , · · · , ∂k fj−1 , ∂k fj+1 , · · · , ∂k fn ).

Then we have

dij (x) = (−1)i+j det(fx1 , · · · , fi−1 , fi+1 , · · · , fN ).

Therefore, it follows that

N
X
∂i dij (x) = (−1)i+j det(fx1 , · · · , fxi−1 , fxi+1 , · · · , ∂i fxk , · · · , fxN ).
k=1

Set

aki = det(∂i fxk , fx1 , · · · , fxi−1 , fxi+1 , · · · , fxk−1 , fxk+1 , · · · , fxN ),

then we have aki = aik and

N
X X
(−1)i+j ∂i dij (x) = (−1)k−1 aki + (−1)k−2 aki
i,k=1 k>i
N
X
= (−1)k−1 δki aki ,
k=1

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 7

where δki = 1 for k < i, δii = 0 and δki = −δik for i, k = 1, 2, · · · , N . Hence
we have
N
X N
X N
X
(−1)j ∂i dij (x) = (−1)k−1+i γki aki = (−1)i−1+k γik aik
i=1 i,k=1 k,i=1
N
X
=− (−1)k−1+i γki aki = 0.
i,k=1

Now, we have
N
X N
X
∂i vi (x) = di,j ∂k wj (f (x))∂i fk (x) + wj (f (x))∂i dij (x).
j,k=1 j=1

PN
On the other hand, i=1 dij ∂i fk (x) = δjk Jf (x) with Kronecker’s δjk .
Therefore, it follows that
N
X
divv(x) = ∂k wj (f (x))δjk Jf (x) = divw(f (x))Jf (x).
k,j=1

This completes the proof.

Finally, we are ready to introduce the following definition:
Definition 1.2.5. Let Ω ⊂ RN be open and bounded, f ∈ C(Ω) and
p∈
/ f (∂Ω). Then we define

deg(f, Ω, p) = deg(g, Ω, p),

where g ∈ C 2 (Ω) and |g − f | < d(p, f (∂Ω)).

Now, one may check the following properties by a reduction to the regular
case.
Theorem 1.2.6. Let Ω ⊂ RN be an open bounded subset and f : Ω →
N
R be a continuous mapping. If p ∈ / f (∂Ω), then there exists an integer
deg(f, Ω, p) satisfying the following properties:
(1) (N ormality) deg(I, Ω, p) = 1 if and only if p ∈ Ω, where I denotes the
identity mapping;
(2) (Solvability) If deg(f, Ω, p) 6= 0, then f (x) = p has a solution in Ω;
(3) (Homotopy) If ft (x) : [0, 1]×Ω → RN is continuous and p 6∈ ∪t∈[0,1] ft (∂Ω),
then deg(ft , Ω, p) does not depend on t ∈ [0, 1];
(4) (Additivity) Suppose that Ω1 , Ω2 are two disjoint open subsets of Ω
and p ∈
/ f (Ω−Ω1 ∪Ω2 ). Then deg(f, Ω, p) = deg(f, Ω1 , p)+deg(f, Ω2 , p);

Copyright 2006 by Taylor & Francis Group, LLC

 8 Topological Degree Theory and Applications

(5) deg(f, Ω, p) is a constant on any connected component of Rn \ f (∂Ω).

As consequences of Theorem 1.2.6, we have the following results:

Theorem 1.2.7. Let f : B(0, R) ⊂ Rn → B(0, R) be a continuous map-
ping. If |f (x)| ≤ R for all x ∈ ∂B(0, R), then f has a fixed point in B(0, R).
Proof. We may assume that x 6= f (x) for all x ∈ ∂B(0, R). Put H(t, x) =
x − tf (x) for all (t, x) ∈ [0, 1] × B(0, R). Then 0 6= H(t, x) for all [0, 1] ×
∂B(0, R). Therefore, we have

deg(I − f, B(0, R), 0) = deg(I, B(0, R), 0) = 1.

Hence f has a fixed point in B(0, R). This completes the proof.
From Theorem 1.2.7, we have the well-known Brouwer fixed point theorem:
Theorem 1.2.8. Let C ⊂ Rn be a nonempty bounded closed convex subset
and f : C → C be a continuous mapping. Then f has a fixed point in C.
Proof. Take B(0, R) such that C ⊂ B(0, R) and let r : B(0, R) → C be a
retraction. By Theorem 1.2.7, there exists x0 ∈ B(0, R) such that f rx0 = x0 .
Therefore, x0 ∈ C, and so we have rx0 = x0 . This completes the proof.
Theorem 1.2.9. Let f : Rn → Rn be a continuous mapping and 0 ∈ Ω ⊂
n
R with Ω an open bounded subset. If (f (x), x) > 0 for all x ∈ ∂Ω, then
deg(f, Ω, 0) = 1.
Proof. Put H(t, x) = tx + (1 − t)f (x) for all (t, x) ∈ [0, 1] × Ω. Then
0∈
/ H([0, 1] × ∂Ω), and so we have

deg(f, Ω, 0) = deg(I, Ω, 0) = 1.

This completes the proof.

Corollary 1.2.10. Let f : Rn → Rn be a continuous mapping. If

(f (x), x)
lim = +∞,
|x|→∞ |x|

then f (Rn ) = Rn .
Proof. For any p ∈ Rn , it is easy to see that there exists R > 0 such that
(f (x) − p, x) > 0 for all x ∈ ∂B(0, R), where B(0, R) is the open ball centered
at zero with radius R. By Theorem 1.2.9, we have

deg(f − p, B(0, R), 0) = 1

and so f (x) − p = 0 has a solution in B(0, R). This completes the proof.

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 9

Theorem 1.2.11. (Borsuk’s Theorem) Let Ω ⊂ Rn be open bounded

and symmetric with 0 ∈ Ω. If f ∈ C(Ω) is odd and 0 ∈
/ f (∂Ω), then d(f, Ω, 0)
is odd.
Proof. Without loss of generality, we may assume that f ∈ C 1 (Ω) with
Jf (0) 6= 0. Next, we define a mapping g ∈ C 1 (Ω) sufficiently close to f by
induction as follows:
Let φ ∈ C 1 (R) be an odd mapping with φ0 (0) = 0 and φ(t) = 0 if and
f (x)
only if t = 0. Put Ωk = {x ∈ Ω : xk 6= 0} and h(x) = φ(x 1)
for all x ∈ Ω1 .
Choose |y1 | sufficiently small such that y1 is a regular value for h on Ω1 . Put
g1 (x) = f (x) − φ(x1 )y1 , then 0 is a regular value for g1 on Ω1 .
Suppose that we have already an odd gk ∈ C 1 (Ω) close to f such that 0 is
a regular value for gk on Ωk . Then we define gk+1 (x) = gk (x) − φ(xk+1 )yk+1
with |yk+1 | small enough such that 0 is a regular value for gk+1 on Ωk+1 .
If x ∈ Ωk+1 and xk+1 = 0, then
0
x ∈ Ωk , gk+1 (x) = gk (x), gk+1 (x) = gk0 (x)

and hence Jgk+1 (x) 6= 0. By induction, we also have gn0 (0) = g10 (0) = f 0 (0)
and so 0 is a regular value for gn . By Definition 1.2.5 and Definition 1.2.1, we
know that
X
deg(f, Ω, 0) = deg(gn , Ω, 0) = sgnJgn (0) + sgnJgn (x)
x∈g −1 (0),x6=0

and thus deg(f, Ω, 0) is odd. This completes the proof.

The following theorem shows the relationship between Brouwer degrees in
different dimensional spaces:
Theorem 1.2.12. Let Ω ⊂ Rn be an open bounded subset, 1 ≤ m < n, let
f : Ω → Rm be a continuous function and let g = I − f . If y ∈
/ (I − f )(∂Ω),
then
deg(g, Ω, y) = deg(gm , Ω ∩ Rm , y),
where gm is the restriction of g on Ω ∩ Rm .
Proof. We may assume that f ∈ C 2 (Ω) and y is a regular value for g on
Ω. A direct computation yields that Jg (x) = Jgm (x) and so the conclusion
follows from Definition 1.2.1. This completes the proof.
Let Ω ⊂ Rn be open and bounded and let f ∈ C(Ω). By the homotopy
invariance of deg(f, Ω, y), we know that deg(f, Ω, y) is the same integer as y
ranges through the same connected component U of Rn \ f (∂Ω). Therefore, it
is reasonable to denote this integer by deg(f, Ω, U ). The unbounded connected
component is denoted by U∞ . Now, we have the product formula:
Theorem 1.2.13. Let Ω ⊂ Rn be an open bounded subset, f ∈ C(Ω),
g ∈ C(Rn ) and let Ui be the bounded connected components of Rn \ f (∂Ω).

Copyright 2006 by Taylor & Francis Group, LLC

 10 Topological Degree Theory and Applications

If p ∈
/ (gf )(∂Ω), then
X
deg(gf, Ω, p) = deg(f, Ω, Ui )deg(g, Ui , p), (1.2.1)
i

where only finitely many terms are not zero.

Proof. We first prove (1.2.1) only has finitely many non-zero terms. Take
r > 0 such that f (Ω) ⊂ Br (0). Then it follows that M = Br (0) ∩ g −1 (p)
is compact, M ⊂ Rn \ f (∂Ω) = ∪i≥1 Ui and there exists finitely many i, say
i = 1, 2, · · · , t, such that ∪t+1
i=1 Ui ⊇ M , where Ut+1 = U∞ ∩ Br+1 . We have

deg(f, Ω, Ut+1 ) = 0, deg(g, Ui , p) = 0

for i ≥ t + 2 since Uj ⊂ Br (0) and g −1 (y) ∩ Uj = ∅ for j ≥ t + 2. Therefore,

the right side of (1.2.1) has only finitely many terms different from zero.
We first suppose that f ∈ C 1 (Ω), g ∈ C 1 (Rn ) and p is a regular value of
gf , so we have
X X
deg(gf, Ω, p) = sgnJgf (x) = sgnJg (f (x))sgnJf (x)
x∈(gf )−1 (p) x∈(gf )−1 (p)

and note
X
sgnJg (z)sgnJf (x)
x∈f −1 (z),z∈g −1 (p)
X X
= sgnJg (z)[ sgnJf (x)]
z∈g −1 (p),z∈f (Ω) x∈f −1 (z)
X
= sgnJg (z)deg(f, Ω, z)
z∈f (Ω),g(z)=p
t X
X
= sgnJg (z)deg(f, Ω, z)
i=1 z∈Ui
X
= deg(f, Ω, Ui )deg(g, Ui , p).
i

For the general case f ∈ C(Ω) and g ∈ C(Rn ), Put

Vm = {z ∈ Br+1 (0) \ f (∂Ω) : deg(f, Ω, z) = m},

Nm = {i ∈ N : deg(f, Ω, Ui ) = m}.
Obviously, Vm = ∪i∈Nm Ui and thus we have
X X X X
deg(f, Ω, Ui )deg(g, Ui , p) = [ deg(g, Ui , p)] = deg(g, Vm , p).
i m i∈Nm m

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 11

Therefore, we need to show

X
deg(gf, Ω, p) = deg(g, Vm , p) (1.2.2)
m

Since ∂Vm ⊂ f (∂Ω), we take g0 ∈ C 1 (Rn ) such that

deg(g0 f, Ω, p) = deg(gf, Ω, p), deg(g0 , Vm , p) = deg(g, Vm , p).
We may assume that M0 = Br+1 (0) ∩ g0−1 (p) is not empty and then we have
ρ(M0 , f (∂Ω)) = inf{|x − z| : x ∈ M0 , z ∈ f (∂Ω)} > 0.
Now we take f0 ∈ C 1 (Ω) such that
max |f (x) − f0 (x)| < ρ(M0 , f (∂Ω)), f0 (Ω) ⊂ Br+1 (0)
x∈Ω

and put
Vm0 = {z ∈ Br+1 (0) \ f0 (∂Ω) : deg(f0 , Ω, z) = m}.
Then we have Vm ∩ M0 = Vm0 ∩ M0 and
deg(g0 , Vm , p) = deg(g0 , Vm ∩ Vm0 , p) = deg(g0 , Vm0 , p).
Therefore, we have
X X
deg(g0 f0 , Ω, p) = mdeg(g0 , Vm0 , p) = mdeg(g, Vm , p).
m m

By a simple homotopy argument, one gets deg(g0 f0 , Ω, p) = deg(g0 f, Ω, p).

Thus the conclusion of Theorem 1.2.13 is true. This completes the proof.
By using the product formula, we can prove the following version of Jordan’s
separation theorem:
Theorem 1.2.14. Let Ωi ⊂ Rn , i = 1, 2, be two compact sets which are
homeomorphic to each other. Then Rn \Ω1 and Rn \Ω2 have the same number
of connected components.
Proof. Let h : Ω1 → Ω2 be a homeomorphism onto Ω2 , h0 be a continuous
extension of h to Rn and h−1 be a continuous extension of h−1 : Ω2 → Ω1 to
Rn . Assume that Ui are bounded components of Rn \ Ω1 and Vj are bounded
components of Rn \ Ω2 . Notice that ∂Ui ⊂ Ω1 and Vj ⊂ Ω2 . For any fixed i,
let Wk denote the components of Rn \ h(∂Ui ). Since
∪j Vj = Rn \ Ω2 ⊂ Rn \ h(∂Ui )) = ∪k Wk
for each j, there exists a k such that Vj ⊂ Wk , so, in particular, V∞ ⊂ W∞ .
For any p ∈ Ui , notice that h−1 h0 x = x for x ∈ ∂Ui , so, by Theorem 1.2.13,
we have
X
1 = deg(h−1 h0 , Ui , p) = deg(h0 , Ui , Wk )deg(h−1 , Wk , p).
k

Copyright 2006 by Taylor & Francis Group, LLC

 12 Topological Degree Theory and Applications

Put Nk = {j : Vj ⊂ Wk }. Then we have

X
deg(h−1 , Wk , p) = deg(h−1 , Vj , p),
j∈Nk

deg(h0 , Ui , Wk ) = deg(h0 , Ui , Vj )
for all j ∈ Nk . Therefore, we have
X X
1= deg(h0 , Ui , Vj )deg(h−1 , Vj , p)
k j∈Nk
X (1.2.3)
= deg(h0 , Ui , Vj )deg(h−1 , Vj , Ui ),
j

since p ∈ Ui ⊂ Rn \ (h−1 (Ω)) ⊂ Rn \ h−1 (∂Vj ). For any fixed j, the same
argument implies that
X
1= deg(h0 , Ui , Vj )deg(h0 , Ui , Vj )deg(h−1 , Vj , Ui ). (1.2.4)
i

From (1.2.3) and (1.2.4), it follows that Rn \ Ω1 and Rn \ Ω2 have the same
number of connected components. This completes the proof.
Theorem 1.2.15. Let φ : Rn → R1 be continuously differentiable, gradφ(x) 6=
0 for |x| sufficiently large and lim|x|→∞ φ(x) = +∞. Then

lim deg(gradφ, Br (0), 0) = 1.

r→∞

Proof. We may assume that φ ∈ C ∞ (Rn ). Otherwise, we use the same

technique as in Proposition 1.1.5. Define
( 1
ce− 1−|x| , |x| < 1,
m(x) =
0, |x| ≥ 1,

where c satisfies Rn m(x)dx = 1. Set mλ (x) = λ−n m( λx ) for all x ∈ Rn and

R

Z
φλ (x) = φ(y)mλ (y − x)dx for all x ∈ Rn , λ > 0.
Rn

Then we consider φλ instead of φ for sufficiently small λ > 0. The initial

value problem (
u0 (t) = −gradφ(u(t)), t > 0,
(E 1.2.1)
u(0) = x ∈ Rn
has a unique local solution u(t, x) for all x ∈ Rn . Since ψ(t) = φ(u(t, x))
satisfies ψ 0 (t) = −|gradφ(u(t, x))|2 ≤ 0, we have φ(u(t, x)) ≤ φ(x) on the
interval, where u(t, x) exists. By assumption lim|x|→∞ φ(x) = +∞, u(t, x)

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 13

can be extended to a unique solution on [0, ∞). We may also assume that
φ(x) ≥ 0 by adding a constant. Take r0 > 0 such that gradφ(x) 6= 0 for
|x| ≥ r0 , put M = maxx∈Br (0) φ(x). Choose r1 > r0 such that φ(x) ≥ M + 1
0
for |x| ≥ r1 and set
M1 = max φ(x).
x∈∂Br1 (0)
Rt
Again, by φ(u(t, x)) = φ(x) − 0 |gradφ(u(s, x))|2 ds, we know that if x ∈
∂Br1 (0), then φ(u(t, x)) ≤ M1 . Let β = min{|gradφ(x)| : |x| ≥ r0 and
φ(x) ≤ M1 }. Then, if |x| = r1 , we have

φ(u(t, x)) ≤ M1 − β 2 t as long as |u(t, x))| ≥ r0 .

M1
Thus |u(t, x)| ≤ r0 for some t < β2 , which implies that

u(M1 β −2 , x) ∈ Br1 (0) for all x ∈ ∂Br1 (0).

This implies that the Poincaré operator P x = u(M1 β −2 , x) must satisfy

deg(I − P, Br1 (0), 0) = 1. Next, define
(
(x − u(M1 β −2 t, x)[t + (1 − t)(tM1 β −2 )−1 ], t 6= 0,
h(t, x) =
gradφ(x) t = 0.

It is easy to check that h is continuous and h(t, x) 6= 0 for all x ∈ ∂Br1 (0).
Thus we have deg(gradφ, Br1 (0), 0) = deg(I − P, Br1 (0), 0) = 1. This com-
pletes the proof.
Theorem 1.2.16. Let Ω ⊂ Rn be an open subset and f : Ω → Rn be
continuous and locally one to one. Then f is an open mapping.
Proof. For each x0 ∈ Ω, we prove that there exists r > 0 such that
f (Br (x0 )) contains a ball with center at f (x0 ). Without loss of generality,
we may assume that x0 = 0. Otherwise, put Ω1 = Ω − {x0 } and f1 (x) =
f (x + x0 ) − f (x0 ).
1
Choose r > 0 such that f is one to one on Br (0). Set h(t, x) = f ( 1+t x) −
t
f (− 1+t x) for all (t, x) ∈ [0, 1] × Br (0). Then h is continuous and h(t, x) 6= 0
for all (t, x) ∈ [0, 1] × ∂Br (0). Otherwise, h(t, x) = 0 for some (t, x) ∈ [0, 1] ×
1 t
∂Br (0) and so then 1+t x = − 1+t x since f is one to one and x = 0, which is
a contradiction. Therefore, we have

deg(f, Br (0), 0) = deg(h(1, ·), Br (0), 0) 6= 0

since h(1, ·) is odd. Choose t > 0 such that t < inf{|f (x)| : x ∈ ∂Br (0)}.
Then
deg(f, Br (0), y) = deg(f, Br (0), 0).
Now, we have Bt ⊂ f (Br (0)) and thus f is open. This completes the proof.

Copyright 2006 by Taylor & Francis Group, LLC

 14 Topological Degree Theory and Applications

Theorem 1.2.17. Let 0 ∈ Ω be open bounded and symmetric. If Ai ⊂ ∂Ω

is closed, Ai ∩ (−Ai ) = ∅ for i = 1, 2, · · · , k and ∪ki=1 Ai = ∂Ω, then k ≥ n + 1.
Proof. Assume the contrary, k ≤ n. Set fi (x) = 1 on Ai , fi (x) = −1
on −Ai for i = 1, 2 · · · , k − 1, fi (x) = 1 on Ω for i = k, · · · , n, and f =
(f1 , f2 , · · · , fn ). Extend f continuously to Ω. Then f (−x) 6= λf (x) on ∂Ω
for all λ ≥ 0. Otherwise, f (−x0 ) = λf (x0 ) for some λ ≥ 0 and x0 ∈ ∂Ω.
Now, λ > 0 since f (x) 6= 0 on ∂Ω. Also x0 ∈ / Ai ∪ (−Ai ) for i ≤ k − 1 since
fi (−x) = −fi (x). Thus x0 ∈ Ak . Also x0 ∈ / −Ak , so we have −x0 ∈ Ai
for some i ≤ k − 1 and thus x0 ∈ −Ai , which is a contradiction. Therefore,
f (−x) 6= λf (x) on ∂Ω for all λ ≥ 0. Thus we have deg(f, Ω, 0) = 0, i.e.,
f (x) = 0 for some x ∈ Ω, which is a contradiction to fn (x) = 1 on Ω. This
completes the proof.
Next, we prove that the winding number is a special case of the Brouwer
degree.
Theorem 1.2.18. Let B(0, 1) ⊂ C be the unit ball, Γ = ∂B(0, 1) and
f : B(0, 1) → C be a C 1 function. Assume that a ∈ / f (Γ). Then
Z
1 1
deg(f, B(0, 1), a) = dz. (1.2.5)
2πi f (Γ) z − a

Proof. It is sufficient to prove (1.2.5) in the case when a ∈

/ f (Sf ). Let
f −1 (a) = {z1 , z2 , · · · , zk }. Then we need to show
Z k
1 1 X
dz = sgnJf (zi ). (1.2.6)
2πi f (Γ) z−a i=1

Take  > 0 small enough such that the Vi ’s are disjoint, where Vi = B(zi , ),
sgnJf (z) = sgnJf (zi ) for z ∈ Vi , Vi ⊂ B(0, 1) and the restriction of f to Vi is
a homeomorphism for i = 1, 2, · · · , k. Put Si = ∂Vi . Then f (Si ) is a Jordan
curve such that a lies in its interior region, f (Si ) has the same orientation as
Si if Jf (zi ) > 0 and the opposite orientation if Jf (zi ) < 0.
Now, set U = B(0, 1) \ ∪ki=1 Vi . Then |f (z) − a| > α in U for some α > 0.
We can divide U into small rectangles Rj such that |f (z) − f (w)| < α on
each Rj . Since the image f (∂(Rj ∩ V )) does not wind around a, we have
w(f (∂(Rj ∩ V )), a) = 0, and summing over all Rj yields
Z k Z
1 X 1
dz = dz.
f (Γ) z−a i=1 f (Si ) z−a

Since the orientation of f (Si ) is determined by Jf (zi ), f (Si ) winds exactly

once around a. Thus, we have
Z
1
dz = sgnJf (zi )
f (Si ) z − a

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 15

and so (1.2.5) is true. This completes the proof.

Remark. (1.2.5) is also correct if f is continuous on B(0, 1).

P 1.2.19. Let B(0, 1) ⊂ C bePthe

Theorem
∞
unit ball, Γ = ∂B(0, 1) and
f (z) = a z
n=0 nP n : B(0, 1) → C with n=1 nan an < ∞. Suppose that
∞
f (Γ) ⊂ Γ. Then n=1 nan an is a non-negative integer.
Proof. Since 0 ∈
/ Γ, by Theorem 1.2.18, we have
Z
1 1
deg(f, B(0, 1), 0) = dz
2πi f (Γ) z
and
∞
f 0 (z)
Z Z Z
1 1 1 1 X
dz = dz = f 0 (z)f (z)dz = nan an .
2πi f (Γ) z 2πi Γ f (z) 2πi Γ n=1

Thus the conclusion is true. This completes the proof.

1.3 Degree Theory for Functions in VMO

Let Ω ⊂ Rn be an n
R open bounded subset and f : Ω → R a measurable
function such that Ω |f (x)|dx < ∞. For any ball Br (x) ⊂ Ω, we define
Ar f (x) to be the average value of f as follows:
Z
1
Ar f (x) = f (y)dy.
m(Br (x)) Br (x)

Lemma 1.3.1. Ar f (x) is continuous in r for each x and measurable in x

for each r.
Proof. Since m(Br (x)) = rn m(B(0, 1)) and m(∂Br (x)) = 0, we have
χBr (x) (y) → χBs (x) (y) almost everywhere as r → s, where χE (y) = 1 if
y ∈ E, while χE (y) = 0 if y ∈ / E. By Lebesgue’s dominated convergence
theorem, we know that Ar f (x) is continuous in r. Also, we have
Z
−n −1
Ar f (x) = r (m(B(0, 1))) χBr (x)(y) f (y)dy,
Ω

χBr (x)(y) is clearly measurable, so the measurability of Ar f (x) follows from

Fubini’s theorem. This completes the proof.
Lemma 1.3.2. Let φ be any collection of open balls in Rn and U = ∪B∈φ B.
If 0 < c < m(U ), then there exist disjoint B1 , B2 , · · · , Bk ∈ φ such that
Pk −n
i=1 m(Bi ) > 3 c.

Copyright 2006 by Taylor & Francis Group, LLC

 16 Topological Degree Theory and Applications

Proof. Since c < m(U ), there exists a compact subset K ⊂ U with m(K) >
c, and finitely many of the balls in φ, say, A1 , A2 , · · · , Am , cover K. Let B1 be
the largest of the Ai ’s (that is, choose B1 to have maximal radius), let B2 be
the largest of the Ai ’s which are disjoint from B1 and so on until the sequence
of Ai ’s is exhausted. If Aj is not one of the Bi ’s, there exists i such that
Aj ∩ Bi 6= ∅ and the radius of Aj is at most that of Bi . Therefore, Aj ⊂ Bi∗ ,
where Bi∗ is the ball concentric with Bi whose radius is three times that of
Bi . But then K ⊂ ∪i Bi∗ , so
X X
c < m(K) ≤ m(Bi∗ ) = 3n m(Bi ).
i i

This completes the proof.

Next, if f ∈ L1 (Ω), we define its Hardy Littlewood maximal function Hf
by Z
1
Hf (x) = sup Ar |f |(x) = sup |f (y)|dy.
r>0 r>0 m(Br (x)) Br (x)

Theorem 1.3.3. There is a constant β > 0 such that, for all f ∈ L1 (Ω)
and α > 0, Z
β
m({x : Hf (x) > α}) ≤ |f (x)|dx.
α Ω

Proof. Let Eα = {x : Hf (x) > α}. For each x ∈ Eα , we can choose

rx > 0 such that Arx |f |(x) > α. The balls Brx (x) cover Eα and so, by
Lemma 1.3.2, if c < m(Eα ), there exist x1 , x2 , · · · , xk ∈ Eα such that the
Pk
balls Bi = Brxi (xi ) are disjoint and i=1 m(Bi ) > 3−n c. But then

k k Z
3n X 3n
X Z
c < 3n m(Bi ) ≤ |f (y)|dy ≤ |f (y)|dy.
i=1
α i=1 Bi α Ω

By letting c → m(Eα ), we obtain the desired result. This completes the proof.
Theorem 1.3.4. If f ∈ L1 (Ω), then limr→0 Ar f (x) = f (x) for almost all
x ∈ Ω.
Proof. For any  > 0, there exists a continuous function g such that
Z
|f (x) − g(x)|dx < .
Ω

Continuity of g implies that for every x ∈ Ω and δ > 0, there exists r > 0
such that |g(y) − g(x)| < δ whenever |y − x| < r, and hence

|Ar g(x) − g(x)| < δ.

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 17

Therefore, Ar g(x) → g(x) as r → 0 for all x ∈ Ω, so we have

lim sup |Ar f (x) − f (x)|

r→0
= lim sup |Ar (f − g)(x) + (Ar g(x) − g(x)) + (g − f )(x)|
r→0
≤ H(f − g)(x) + |f (x) − g(x)|.

Hence, if

Eα = {x : lim sup |Ar (f − g)(x) > α}, Fα = {x : |f (x) − g(x)| > α},
r→0

then we have
α
Eα ⊂ F α2 ∪ {x : H(f − g)(x) > }.
2
R
However, αm(Fα ) ≤ Fα
|f (x) − g(x)|dx <  and so, by Theorem 1.3.3, we
have
2 2β
m(Eα ) ≤ + .
α α
By letting  → 0, we get m(Eα ) = 0 for all α > 0. This completes the proof.
In the following, let Ω be a smooth open bounded domain in Rn or a smooth
compact Riemannian manifold. (For the definition of Riemannian manifold,
we refer the reader to [171].
Next, we introduce the following concept:
1
Definition 1.3.5. Let
1
R f ∈ 1L (Ω)R and Br ⊂ Ω a (geodesic) ball with radius
r > 0. If supr>0 m(Br ) Br m(Br ) Br |f (x) − f (y)|dxdy < ∞, then f is called
a bounded mean oscillation function. The set of all bounded mean oscillation
functions is denotedR by BM O.
1 1
R
If limr→0 m(B r) Br m(Br ) Br
|f (x) − f (y)|dxdy = 0, then f is called a
vanishing mean oscillation function. The set of all vanishing mean oscillation
functions is denoted by V M O.
Example 1.3.6. If f ∈ L1 (Ω), then f ∈ BM O.
Example 1.3.7. f (x) = | log |x|| ∈ BM O.
In the following, 1 ≤ p < +∞, let W 1,p (Ω) = {u(·) : Ω → R such that u(·) ∈
L (Ω), u0 (·) ∈ Lp (Ω)}.
p

Proposition 1.3.8. If f (·) ∈ W 1,n , then f (·) ∈ V M O.

Proof. By Poincaré’s inequality, we have
Z Z Z
1 1
|f (y) − |f (x)|dx|dy ≤ cm(B) n |∇f (x)|dx.
Br m(Br ) Br Br

Copyright 2006 by Taylor & Francis Group, LLC

 18 Topological Degree Theory and Applications

Then we deduce, by using the Hölder inequality, that

Z Z Z
1 1 1
|f (y) − |f (x)| ≤ c( |∇f (x)|dx) n .
m(Br ) Br m(Br ) Br Ω

Thus the conclusion is true.

Recall that, for 0 < s < 1, 1 < p < +∞, the fractional Sobolev space
W s,p (Ω) is characterized by
|f (x) − f (y)|p
Z Z
W s,p (Ω) = {f (·) ∈ Lp (Ω), n+sp
< +∞}.
Ω Ω |x − y|

Proposition 1.3.9. If sp = n, then W s,p (Ω) ⊂ V M O.

Proof. Clearly, one has
Z Z
|f (x) − f (y)|dxdy
B B
|f (x) − f (y)|
Z Z
n
= (n
|x − y|( p )+s dxdy
|x − y| p )+s
B B
|f (x) − f (y)|
Z Z
1 s
+n
≤ c(m(B)) p
( n )+s
dxdy
B B |x − y| p

for some constant c > 0. By using Hölder’s inequality, we deduce that

|f (x) − f (y)| |f (x) − f (y)|p
Z Z Z Z
1 s 2
+n +2− p
n
( )+s
dxdy ≤ cm(B) p
n+sp
B B |x − y| p B B |x − y|

and thus, when sp = n,

|f (x) − f (y)|p p1
Z Z Z Z
1
2
|f (x) − f (y)|dxdy ≤ c[ n+sp
] .
m(B) B B B B |x − y|

This completes the proof.

Lemma 1.3.10. Let B(0, 1) be the unit ball of Rn+1 , S n = ∂B(0, 1) and
f ∈ L1 (S n , S n ) be such that f ∈ V M O. Then Ar f (x) = m(B1r (x)) Br (x) f (y)dy
R

is continuous in x for small r > 0 and limr→0 |Ar f (x)| = 1 uniformly on S n .

Proof. The continuity of Ar f (x) is guaranteed since f ∈ V M O. To prove
the uniform convergence of Ar f (x) on S n , we set
Z Z
1
δr (x) = |f (y) − f (z)|dydz.
m(Br (x))2 Br (x) Br (x)

Then δr (x) → 0 as r → 0 uniformly on S n and we have

1 − δr (x) ≤ |Ar f (x)| ≤ 1.

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 19

Thus |Ar f (x)| → 1 uniformly on S n as r → 0. This completes the proof.

Now, assume that f ∈ L1 (S n , S n ) and f ∈ V M O. By Lemma 1.3.10,
there exists r0 > 0 such that Ar f (x) 6= 0 for all x ∈ S n and 0 < r <
r0 . Let Ãr f (x) be a continuous extension of Ar f (x) to B(0, 1). Then the
Brouwer deg(Ãr f, B(0, 1), 0) is well defined for r ∈ (0, r0 ). It does not depend
on the extension Ãr f . Consider the homotopy {Atr1 +(1−t)r ˜ f }t∈[0,1] for all
2
r1 , r2 ∈ (0, r0 ). It follows from Lemma 1.3.10 that Atr1 +(1−t)r2 f (x) 6= 0 for all
(t, x) ∈ [0, 1]×S n and thus deg(A˜r f , B(0, 1), 0) does not depend on r ∈ (0, r0 ).
Now, we define the topological degree by

deg(f, S n ) = lim deg(A˜r f , B(0, 1), 0). (1.3.1)

r→0

Proposition 1.3.11. We have

Ãr f
lim deg(A˜r f , B(0, 1), 0) = lim deg( , B(0, 1), 0).
r→0 r→0 |Ãr f |

A˜r f (x)
Proof. Consider the homotopy Hr (t, x) = (1−t)+t|A˜r f (x)|
for all (t, x) ∈
[0, 1] × B(0, 1). By Lemma 1.3.10, we know that Hr (t, x) 6= 0 for all (t, x) ∈
[0, 1] × S n for r sufficiently small and so the conclusion follows from Theorem
1.2.6.
Remark. For more results regarding the degree defined by (1.3.1), we refer
the reader to Brezis and Nirenberg [29] (see also [25], [27], [28], [175] for more
results on the computation of the degree for Sobolev maps).

1.4 Applications to ODEs

In this section, we give some applications of results in section 1.2 to periodic
and anti-periodic problems of ordinary differential equations in Rn .
Theorem 1.4.1. Let f : R × Rn → Rn be a continuous function and
f (t + T, x) = f (t, x) for all (t, x) ∈ R × Rn . Suppose that the following
conditions are satisfied:

(1) There exists r > 0 such that (f (t, x), x) < 0 for all t ∈ [0, T ] and |x| = r.

(2) For each x ∈ Rn , there exist rx > 0, Lx > 0 such that

|f (t, y) − f (t, z)| ≤ Lx |y − z| for all t ∈ R, y, z ∈ B(x, rx ).

Copyright 2006 by Taylor & Francis Group, LLC

 20 Topological Degree Theory and Applications

Then the following equation:

(
x0 (t) = f (t, x(t), t ∈ [0, +∞),
(E 1.4.1)
x(0) = x(T )

has a solution.
Proof. For each x0 ∈ B(0, r), by Peano’s Theorem, the initial value prob-
lem (
x0 (t) = f (t, x(t)), t ∈ (0, t0 ),
(E 1.4.2)
x(0) = x0
has a solution for some t0 > 0. If (E 1.4.2) has two solutions x(·), y(·), then

d
|x(t) − y(t)|2 = 2(x0 (t) − y 0 (t), x(t) − y(t)) ≤ Lx0 |x(t) − y(t)| (1.4.1)
dt
for some t1 ∈ (0, t0 ) and t ∈ (0, t1 ). From (1.4.1), we get

|x(t) − y(t)| ≤ eLx0 t |x(0) − y(0)| for all t ∈ (0, t1 ),

so x(t) = y(t) for all t ∈ (0, t1 ). Therefore, x(t) = y(t) for t ∈ [0, t0 ], so the
solution of (E 1.4.2) is unique.
d
If x(t) = r, then dt |x(t)|2 = 2(x0 (t), x(t)) = (f (t, x(t)), x(t)) < 0. Thus x(t)
must stay in B(0, r) for all t ∈ [0, t0 ], so x(t) can be extended to [0, +∞) and
also x(t) ∈ B(0, r) for t ∈ [0, +∞).
Now, we define a mapping S : B(0, r) → B(0, r) as follows:

Sy = x(y, T ) for all y ∈ B(0, r),

where x(y, t) is the unique solution of (E 1.4.2) with initial value y.

Again, by using (1.4.1), one can easily prove that S is continuous. Thus, by
Brouwer’s fixed point theorem, S has a fixed point in B(0, r), i.e., (E 1.4.1)
has a solution. This completes the proof.
Theorem 1.4.2. Let G : Rn → R be an even continuous differentiable
function such that ∂G is Lipschitzian and f : R → Rn be a continuous
function such that f (t + T ) = −f (t) for all t ∈ R, then the following equation:
(
x0 (t) = ∂Gx(t) + f (t), t ∈ R,
(E1.4.3)
x(t + T ) = −x(t), t∈R

has a solution.
Proof. First, if x(t) is a solution of (E 1.4.3), then

|x0 (t)|2 = (∂Gx(t), x0 (t)) + (f (t), x0 (t)) for all t ∈ R.

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 21
RT
Integrate over [0, T ] and notice that 0
(∂Gx(t), x0 (t))dt = 0, we have
Z T Z T
|x0 (t)|2 dt = (f (t), x0 (t))dt
0 0
RT 1 RT 1
and thus ( 0 |x0 (t)|2 dt) 2 ≤ ( 0 |f (t)|2 dt) 2 .
Rt 0 Rt 0
In addition, 2x(t) = 0 x (s)ds − T x (s)ds and thus we have
√ Z T
T 1
max |x(t)| ≤ ( |f (t)|2 dt) 2 = M.
t∈[0,T ] 2 0

We take an even continuous differentiable function φ with φ(x) = 1 for |x| ≤

M and ∂φ(x) = 0 for |x| > 2M .
Now, we consider the following equation:
(
x0 (t) + x(t) = ∂(φ(x(t))[G(x(t)) + 21 |x(t)|2 ]) + f (t), t ∈ R,
(E 1.4.4)
x(0) = y ∈ Rn .

Since ∂φ is uniformly bounded, we have, for each solution x(t) of (E 1.4.4),

d
(|x(t)|2 ) + |x(t)|2 ≤ 2L|x(t)| + 2|f (t)||x(t)|
dt
for some constant L > 0. Thus, if there exists a N > 0 such that |x(0)| ≤ N ,
then |x(T )| ≤ N .
Now, we define a map S : Rn → Rn by Sy = −x(T ), where x(·) is the
unique solution of (E 1.4.4) with x(0) = y. It is obvious that S is continuous,
so, by Brouwer’s fixed point theorem, Theorem 1.2.7, there exists y ∈ Rn such
that Sy = y, i.e., x(T ) = −y. Thus |x(t)| ≤ M for all t ∈ R and, consequently,
it follows that
1
∂(φ(x(t))[G(x(t)) + |x(t)|2 ]) = ∂Gx(t) + x(t).
2
Therefore, x(·) is a solution of (E 1.4.3). This completes the proof.
Corollary 1.4.3. Let A be a n × n symmetric matrix and f : R → Rn
be a continuous function such that f (t + T ) = −f (t) for all t ∈ R. Then the
following equation:
(
x0 (t) = Ax(t) + f (t), t ∈ R,
(E 1.4.5)
x(t + T ) = −x(t), t∈R

has a solution.
Proof. Since A is symmetric, put Gu = 21 (Au, u) for u ∈ Rn and then
A = ∂G. Thus the conclusion follows from Theorem 1.4.2.

Copyright 2006 by Taylor & Francis Group, LLC

 22 Topological Degree Theory and Applications

Example 1.4.4. The following equation:


0
x1 (t) = 2x1 (t) − αx2 (t) + sint,
 t ∈ R,
0
x2 (t) = −αx1 (t) − 3x2 (t) + sin3 t, t ∈ R, (E 1.4.6)

x1 (t + π) = −x1 (t), x2 (t + π) = −x2 (t), t ∈ R,


has a solution, where α ∈ R is a constant. In fact, set

     
2 −α x1 sin t
A= , x= , f (t) = .
−α −3 x2 sin3 t
Then A is symmetric and the conclusion follows from Corollary 1.4.3.

1.5 Exercises
1. Let Ω ⊂ Rn be open bounded and 0 ∈ Ω and f : Ω → Rn be continuous
and (f (x), x) ≥ 0. Show 0 ∈ f (Ω).
2. Let Ω ⊂ R2 be open bounded and u(x, y), v(x, y) : Ω → R be contin-
uously differentiable functions with ux = vy and uy = −vx . Assume
that f (x, y) = (u(x, y), v(x, y)) : Ω → R2 has m many zero points in
Ω \ Sf (Ω). Show that deg(f, Ω, 0) = m.
3. Prove the fundamental theorem of algebra by using Brouwer degree.
4. Let B(0, 1) be the unit ball in Rn and f : Rn → Rn be a continuous func-
tion such that f (∂B(0, 1)) = ∂B(0, 1). Show that deg(f m , B(0, 1), 0) =
(deg(f, B(0, 1), 0)m for all positive integer m.
5. For any integer n, show that there exists an open bounded subset Ω ⊂ R
and a continuous function f : Ω → R such that deg(f, Ω, 0) = n.
6. Let Ω ⊂ Rn be open bounded, f, g ∈ C(Ω) and |g(x)| < |f (x)| for all
x ∈ ∂Ω. Show deg(f − g, Ω, 0) = deg(f, Ω, 0).
7. Let B(0, 1) be the unit ball of R2n+1 and f : ∂B(0, 1) → ∂B(0, 1) be
continuous. Show that there exists x0 ∈ ∂B(0, 1) such that f (x0 ) = x0
or f (x0 ) = −x0 .
8. Let Ω ⊂ Rn be open bounded symmetric, 0 ∈ Ω and f : ∂Ω → Rm be a
function with m < n. Show that f (x) = f (−x) for some x ∈ ∂Ω
9. Let Ω ⊂ Rn be open bounded and f ∈ C(Ω). Suppose that there exists
x0 ∈ Ω such that f satisfies the following condition:
f (x) − x0 = t(x − x0 ) for some x ∈ ∂Ω.
Then t ≤ 1. Show that f has a fixed point in Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 BROUWER DEGREE THEORY 23

10. Let φ : B(0, 1) ⊂ Rn → Rn be a continuous function such that φ(∂B(0, 1)) ⊂

φ(x)
Rn \ {0}, ψ(x) = |φ(x)| for all x ∈ ∂B(0, 1) and ∂B(0, 1) 6= ψ(∂B(0, 1)).
Show that
deg(φ, B(0, 1), 0) = deg(ψ, B(0, 1), p)
for all p ∈ ∂B(0, 1) \ ψ(∂B(0, 1)).
11. If f ∈ L1 (Ω), show that limr→0 m(B1r (x)) Br (x) |f (y) − f (x)|dy = 0 for
R

almost all x ∈ Ω.
P∞ P∞
12. If f (θ) = n=0 ai einθ ∈ L2 (S 1 , S 1 ) and n=0 n|an |2 < +∞, show that
∞
X
deg(f, S 1 ) = n|an |2 .
n=0

13. Show that the following equation:

(
x0 (t) = x3 (t) + sin5 t, t ∈ R,
x(t + π) = −x(t) t ∈ R,

has a solution.

Copyright 2006 by Taylor & Francis Group, LLC

 Chapter 2
LERAY SCHAUDER DEGREE
THEORY

Many problems in science lead to the equation T x = y in infinite dimensional

spaces rather than to the finite dimensional case in Chapter 1. In particu-
lar, ordinary and partial differential equations, and integral equations can be
formulated as abstract equations on infinite dimensional spaces of functions.
For the equation T x = y, we again are interested in the questions raised at
the beginning of Chapter 1.
In 1934, Leray and Schauder [185] generalized Brouwer degree theory to an
infinite Banach space and established the so-called the Leray Schauder degree.
It turns out that the Leray Schauder degree is a very powerful tool in proving
various existence results for nonlinear partial differential equations (see [135],
[185], [203], [228], etc.).
In this chapter, we will introduce the Leray Schauder degree. This chapter
consists of five sections.
Section 2.1 gathers together some well known results on compact maps.
In Section 2.2, we first show how a compact map can be approximated
by maps with finite dimensional ranges and from here we define the Leray
Schauder degree for compact maps. The main properties of this degree are
presented in theorems 2.2.4, 2.2.8 and 2.2.16. Also, various consequences, for
example, Schauder’s fixed point theorem, the Leray Schauder alternatives and
compression and expansion fixed point theorems, are presented in this section.
Section 2.3 presents a degree theory for multi-valued maps, and the theory
is based on the fact that a upper semicontinuous map admits an approximate
continuous selection (see Lemma 2.3.7). We use the degree theory in this
chapter to discuss bifurcation problems in Section 2.4 and ordinary (initial
and anti-periodic) and partial differential equations in Section 2.5.

2.1 Compact Mappings

In this section, we give some properties of compact operators in topological
spaces.

25

Copyright 2006 by Taylor & Francis Group, LLC

 26 Topological Degree Theory and Applications

Definition 2.1.1. Let X be a topological space. A subset M ⊂ X is

called compact if every open covering of M has an finite covering, i.e., if
M ⊂ ∪i∈I Vi , where Vi is an open subset of X for all i ∈ I, then there exist
ij ∈ I, j = 1, 2, · · · , k, such that M ⊂ ∪kj=1 Vij .
M is called relatively compact if M is compact.
Definition 2.1.2. Let X be a nonempty subset and d(·, ·) : X × X → R
be a function satisfying the following conditions:
(1) d(x, y) ≥ 0 for all x, y ∈ X and d(x, y) = 0 iff x = y;
(2) d(x, y) = d(y, x) for all x, y ∈ X;
(3) d(x, y) ≤ d(x, z) + d(z, y) for all x, y, z ∈ X.
Then we call d a metric on X and (X, d) a metric space.
Let (X, d) be a metric space, x ∈ X and r > 0. Let B(x, r) = {y ∈ X :
d(x, y) < r} be a open ball with center x and radius r.
Proposition 2.1.3. Let (E, d) be a metric space. Then, a subset M ⊂ X is
compact if and only if every infinite sequence (xn )∞
n=1 ⊂ M has a convergent
subsequence in M .
Proof. Assume that M is compact and (xn )∞ ∞
n=1 ⊂ M . If (xn )n=1 does
not have a convergent subsequence in M , then, for any y ∈ M , there exist
ry > 0 and an integer Ny > 0 such that B(y, ry ) ∩ {xn : n ≥ ny } = ∅.
Notice that ∪y∈M B(y, ry ) ⊃ M , so there exist finitely many y1 , y2 , · · · , yk
such that M ⊂ ∪ki=1 B(yi , ryi ). However, B(yi , ryi ) ∩ {xn : n ≥ m} = ∅ for
i = 1, 2, · · · , k, where m = max{nyi : i = 1, 2 · · · , k}. Thus

{xn : n ≥ m} = M ∩ {xn : n ≥ m}
⊂ ∪ki=1 B(yi , ryi ) ∩ {xn : n ≥ m}
= ∅,

which is a contradiction and so (xn )∞ n=1 has a convergent subsequence in M .

On the other hand, assume that every infinite sequence (xn )∞ n=1 ⊂ M has
a convergent subsequence in M . We prove that M is compact. Let (Ui )i∈I
be an open covering of M . For any x ∈ M , there exists Ui such that x ∈ Ui .
Since Ui is open, there exists r > 0 such that B(x, r) ⊂ Ui . Put rx = sup{r >
0 : B(x, r) ⊂ Bi for some i ∈ I} and set r0 = inf{rx : x ∈ M }. We next prove
that r0 > 0. There exists a sequence (xn )∞ n=1 ⊂ M such that ri = rxi → r0 .
By assumption, there exists a subsequence (xnk )∞ k=1 such that xnk → y0 ∈ M ,
so there exists an integer N > 0 such that xnk ∈ B(x0 , 4−1 rx0 ) for k > N .
Thus it follows that B(xnk , 4−1 rx0 ) ⊂ B(x0 , 2−1 rx0 ) ⊂ Bi for some i ∈ I.
Consequently, we have r0 ≥ 2−1 rx0 . For any x1 ∈ M , if B(x1 , 2−1 r0 ) 6⊃ M ,
there exists x2 ∈ M \ B(x1 , 2−1 r0 ), and if ∪2i=1 B(2−1 xi , r0 ) 6⊃ M , there exists
x3 ∈ M \ ∪2i=1 B(2−1 xi , r0 ), and we claim that this process will terminate at

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 27

some finite step. If not, there exist xn ∈ M \ ∪n−1

i=1 B(xi , 2
−1
r0 ) for n ≥ 4, so
we have
d(xn , xm ) ≥ 2−1 r0 f or n 6= m,

which is a contradiction to our assumption. Thus, there exist finitely many

x1 , x2 , · · · , xn such that M ⊂ ∪ni=1 B(xi , 2−1 r0 ) ⊂ ∪ni=1 Uji and M is compact.
This completes the proof.

Definition 2.1.4. Let E be a real vector space. A function k · k : E → R

satisfying the following conditions:

(1) kxk ≥ 0 for all x ∈ E, and kxk = 0 if and only if x = 0;

(2) kαxk = |α|kxk for all α ∈ R, x ∈ E,

(3) kx + yk ≤ kxk + kyk for all x, y ∈ E

is called a norm on E and (E, k · k) a real normed space or, simply, E is a

normed space. If E is also complete, then we say E a real Banach space.

Now, we give some well-known Banach spaces in functional analysis.

Example 2.1.5. Let 1 ≤ p < ∞ and

lp = {(xi ) : xi ∈ R for i = 1, 2, · · · , Σ∞ p
i=1 |xi | < +∞}.

Then lp is a Banach space.

Example 2.1.6. Let

c = {(xi ) : xi ∈ R for i = 1, 2, · · · , and lim xi exists}.

i→∞

Then c is a Banach space.

Example 2.1.7. Let c0 = {(xi ) : xi ∈ R for i = 1, 2, · · · , and limi→∞ xi =

0}. Then c0 is a Banach space.

Example 2.1.8. Let Ω ⊂ Rn be a bounded measurable subset, 1 ≤ p < ∞,

and Lp (Ω) = {f (·) : Ω → R such that Ω |f (x)|p dx < ∞}. Then Lp (Ω) is a
R

Banach space.

Lemma 2.1.9. (Riesz’s Theorem) Let E be a real normed space and

M ⊂ E be a proper closed subspace. Then, for any  ∈ (0, 1), there exists
x0 ∈ E such that kx0 k = 1 and d(x0 , M ) = inf y∈M kx0 − yk > .

Proof. Since M 6= E, there exists y0 ∈ E \ M such that d(y0 , M ) = δ0 > 0.

−y1
Take y1 ∈ M such that ky0 − y1 k < −1 δ0 and put x0 = kyy00 −y 1k
. Then

Copyright 2006 by Taylor & Francis Group, LLC

 28 Topological Degree Theory and Applications

kx0 k = 1 and
y0 − y1
d(x0 , M ) = inf k − yk
y∈M ky0 − y1 k
= inf ky0 − y1 k−1 ky0 − y1 − ky0 − y1 kyk
y∈M

> δ0−1 δ0
= ,

which is the desired result. This completes the proof.

As a consequence of Lemma 2.1.9, we get the following:
Proposition 2.1.10. Let E be a real normed space. Then, the unit closed
ball B(0, 1) = {x : kxk ≤ 1} is compact if and only if dim(E) < +∞.
Definiton 2.1.11. Let E be a real normed space. A mapping T : D(T ) ⊂
E → E is called compact if T maps every bounded subset of D(T ) to a
relatively compact subset in E. T is said to be completely continuous if T is
continuous and compact.
Definition 2.1.12. Let X, Y be two real Banach spaces and Ω ⊂ X be an
open subset. A mapping F : Ω → Y is said to be Fréchet differentiable at
x0 ∈ Ω if there is an F 0 (x0 ) ∈ L(X, Y ) such that

δ(x0 , h)
F (x0 + h) = F x0 + F 0 (x0 )h + δ(x0 , h), lim = 0,
h→0 khk

where L(X, Y ) is the set of all bounded linear operators from X → Y .

Proposition 2.1.13. Let X, Y be two real Banach spaces and Ω ⊂ X be
an open subset. If F : Ω → Y is a continuous compact mapping and F is
Fréchet differentiable at x0 ∈ Ω, then F 0 (x0 ) is compact.
Proof. Take any sequence (hn )∞ n=1 ⊂ X with khn k = 1 for n = 1, 2, · · · .
Let  > 0. There exists δ0 > 0 such that w(x0δ,δh) < 2 for all h ∈ X with
khk = 1 and δ ≤ δ0 . Since F is compact, (F (x0 + δ0 hn )) has a convergent
subsequence, say, (F (x0 + δ0 hnk ))∞
k=1 . We also have

F 0 (x0 )hnk − F 0 (x0 )xnl = δ0−1 [F (x0 + δ0 hnk ) − F (x0 + δ0 xnl )]

+δ0−1 [w(x0 , δ0 hnk ) − w(x0 , δ0 hnl )].

Then we deduce, by letting k, l → ∞, that

kF 0 (x0 )hnk − F 0 (x0 )xnl k ≤ .

Again, by letting  → 0, we get that (F 0 (x0 )hnk )∞ 0

k=1 is convergent; thus F (x0 )
is compact. This completes the proof.

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 29

Let E be a real Banach space and C([0, T ], E) be the space of continuous

functions from [0, T ] to E with the norm kx(·)k = maxt∈[0,T ] kx(t)k, then
C([0, T ], E) is a Banach space. A subset B ⊂ C([0, T ], E) is called equicontin-
uous if, for any  > 0, there exists δ() > 0 such that kx(t1 ) − x(t2 )k <  for
all x(·) ∈ B and t1 , t2 ∈ [0, T ] satisfying |t1 − t2 | < δ.
Theorem 2.1.14. (Ascoli Arzela’s Theorem) A subset M ⊂ C([0, T ], E)
is relatively compact if and only if
(1) M is equicontinuous and
(2) for each t ∈ [0, T ], M (t) = {x(t) : x(·) ∈ M } is relatively compact in E.
Proof. The proof follows easily from Theorem 3.1.16 in the next chapter.
Theorem 2.1.15. Let E be a real Banach space, T : E → E be a contin-
uous compact mapping with λ ∈ R, Tλ = T − λI and σ(T ) be the spectrum
of T . Then the following conclusions hold:

(1) If E is infinite dimensional, then 0 ∈ σ(T );

(2) If 0 6= λ ∈ σ(T ), then λ is an eigenvalue of T ;

(3) σ(T ) is a countable subset;

(4) If λ ∈
/ σ(T ), then Tλ is a homeomorphism onto X;

(5) E = N (Tλk ) ⊕ R(Tλk ) for all λ 6= 0, k > 0, where dim(N (Tλk )) < +∞,
and R(N (Tλk )) is closed in E.

Proof. (4) is obvious. For (1) suppose that 0 ∈ / σ(T ). Then T −1 : E → E is

continuous. Since E is infinite dimensional, there exists a sequence (xn )∞ n=1 ⊂
such that kxn k = 1 and kxn − xm k ≥ 21 for n 6= m. Put yn = T xn , then (yn )
has a convergence subsequence, but T −1 yn = xn will have a convergence
subsequence, which is a contradiction. Thus 0 ∈ σ(T ).
For (5) denote by Ni = N (Tλi ) and Ri = R(Tλi ) for i = 1, 2, · · · . The finite
dimensional property of Ni follows directly from the compactness of T . To
prove the closedness of Ri , first, note there exists a closed subspace M of E
such that E = Ni ⊕ M . Define an operator S : M → E by

Sx = Tλi x for all x ∈ M.

Observe that R(S) = R(Tλi ), so we only need to prove R(S) is closed. It is

obvious that S is one to one. Now, we prove that there exists γ > 0 such that

kSxk ≥ γkxk for all x ∈ M. (2.1.1)

If this is not true, there exist xn ∈ M with kxn k = 1 such that kSxn k ≤ n−1 .
Since S = (T − λI)i = (−λ)i I + Σij=1 Cij (−λ)i−j T j , one easily sees that
(xn )∞
n=1 has a convergence subsequence (xnk ) with xnk → x0 . Thus Sx0 = 0,

Copyright 2006 by Taylor & Francis Group, LLC

 30 Topological Degree Theory and Applications

which implies that x0 ∈ Ni and so x0 = 0, which contradicts kx0 k = 1. From

(2.1.1), we deduce that M = R(Tλi ) is closed.
For (2) notice that N1 ⊆ N2 ⊆ · · · and R1 ⊇ R2 ⊇ · · · . We cannot have
Ni 6= Ni+1 for all i. Otherwise, by Lemma 2.1.9, there exists a sequence of
xi ∈ Ni+1 \ Ni such that kxi k = 1 and kxi − xj k ≥ 21 for i 6= j and thus we
have
kT xi − T xj k ≥ 2−1 λ for all j < i,
which is impossible since T is compact. Thus Nj = Ni for some i and all
j > i and, consequently, Rj = Ri for all j > i. If λ 6= 0 is not an eigenvalue
of T , then Tλ is one to one. For any y ∈ Ri−1 , we have Tλ y ∈ Ri = Ri+1
and thus there exists x such that Tλ y = Tλi+1 x. Therefore, y = Tλi x ∈ Ri and
Ri−1 = Ri , so on. Thus we get R(Tλ ) = E. Therefore, λ ∈ / σ(T ).
To prove (3), we prove that, for any r > 0, {λ ∈ σ(T ) : |λ| > r} is a
finite subset. Suppose the contrary, i.e., there exist λn , |λn | > r, n = 1, 2, · · · ,
and xn ∈ E with |xn | = 1 such that T xn = λn xn and {x1 , x2 , · · · , xn } is
linearly independent. Put Mn = span{x1 , x2 , · · · , xn }. Then Mn ⊂ Mn+1 .
By Lemma 2.1.9, there exist yn+1 ∈ Mn+1 \ Mn such that kyn+1 k = 1 and
d(yn+1 , Mn ) ≥ 21 for n = 1, 2, · · · . Let yn+1 = Σn+1 n
i=1 αi xi . Then

λn+1 yn+1 − Ayn+1 = Σni=1 αin (λn+1 − λi )xi = zn ∈ Mn .

From which we deduce, for m > n, that

kT ym − T yn k = k(λm ym − λn yn ) − (zm−1 − zn−1 )k

r
≥ |λm |d(ym , Mm−1 ) ≥ ,
2
which contradicts the compactness of T . Thus σ(T ) is countable. This com-
pletes the proof.

2.2 Leray Schauder Degree

In this section, we construct the Leray Schauder degree. First, we need
the following result on the approximation of a compact mapping by finite
dimensional mappings.
Lemma 2.2.1. Let E be a real Banach space, Ω ⊂ E be an open bounded
subset and T : Ω → E be a continuous compact mapping. Then, for any
 > 0, there exist a finite dimensional space F and a continuous mapping
T : Ω → F such that

kT x − T xk <  for all x ∈ Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 31

Proof. Since T Ω is relatively compact in E, for any  > 0, there exists a

finite subset {x1 , x2 , · · · , xn } ⊂ Ω such that
T (Ω) ⊂ ∪ni=1 B(T xi , ).
Now, we define a mapping T : Ω → F = span{T x1 , T x2 , · · · , T xn } as
follows:
φi (x)
T x = Σni=1 T xi for all x ∈ Ω,
Γ(x)
where φi (x) = max{0,  − kT x − T xi k} and Γ(x) = Σni=1 φi (x). Then it is easy
to check that T satisfies the conditions of Lemma 2.2.1, so the conclusion
follows. This completes the proof.
Lemma 2.2.2. Let E be a real Banach space, B ⊂ E be a closed bounded
subset and T : B → E be a continuous compact mapping. Suppose T x 6= x
for all x ∈ B. Then there exists 0 > 0 such that x 6= tT1 x + (1 − t)T2 x for
all t ∈ [0, 1] and x ∈ B, where i ∈ (0, 0 ) and Ti : B → Fi for i = 1, 2 as in
Lemma 2.2.1.
Proof. Suppose the conclusion is not true. There exist j1 → 0, j2 → 0,
tj → t0 , xj ∈ B such that tj Tj xj + (1 − tj )Tj xj = xj for j = 1, 2, · · · .
1 2
By compactness of T , (T xj )∞
j=1 has a subsequence, say (T xjk ), converging
to y ∈ E. By Lemma 2.2.1, Tjk xjk → y for i = 1, 2. Thus xjk → y ∈ B.
i
Therefore, T y = y, which is a contradiction.
Definition 2.2.3. Let E be a real Banach space, Ω ⊂ E be an open
bounded set and T : Ω → E be a continuous compact mapping. Now, suppose
that 0 6∈ (I − T )(∂Ω). Then, by Lemma 2.2.2, there exists 0 > 0 such that
x 6= tT1 x + (1 − t)T2 x for all t ∈ [0, 1], x ∈ ∂Ω,
where i ∈ (0, 0 ) and Ti : Ω → Fi for i = 1, 2 as in Lemma 2.2.1. Hence
Brouwer’s degree deg(I − T , Ω ∩ F , 0) is well defined, and so we define
deg(I − T, Ω, 0) = deg(I − T , Ω ∩ F , 0),
where  ∈ (0, 0 ).
By the homotopy property of Brouwer degree, we have
deg(I − T1 , Ω ∩ span{F1 ∪ F2 }, 0) = deg(I − T2 , Ω ∩ span{F1 ∪ F2 }, 0).
But Ti : Ω ∩ span{F1 ∪ F2 } :→ Fi for i = 1, 2, so by Theorem 1.2.12 we
have
deg(I − T1 , Ω ∩ span{F1 ∪ F2 }, 0) = deg(I − T1 , Ω ∩ F1 , 0)
and
deg(I − T2 , Ω ∩ span{F1 ∪ F2 }, 0) = deg(I − T2 , Ω ∩ F2 , 0).

Copyright 2006 by Taylor & Francis Group, LLC

 32 Topological Degree Theory and Applications

Thus we have

deg(I − T1 , Ω ∩ F1 , 0) = deg(I − T2 , Ω ∩ F2 , 0)

and the degree defined in Definition 2.2.3 is well defined. For the general case,
if p ∈
/ (I − T )(∂Ω), we define deg(I − T, Ω, p) = deg(I − T − p, Ω, 0).
We recall some properties of the Leray Schauder degree as follows:
Theorem 2.2.4. The Leray Schauder degree has the following properties:
(1) (N ormality) deg(I, Ω, 0) = 1 if and only if 0 ∈ Ω;
(2) (Solvability) If deg(I − T, Ω, 0) 6= 0, then T x = x has a solution in Ω,
(3) (Homotopy) Let Tt : [0, 1]×Ω → E be continuous compact and Tt x 6= x
for all (t, x) ∈ [0, 1] × ∂Ω. Then deg(I − Tt , Ω, 0) doesn’t depend on
t ∈ [0, 1];
(4) (Additivity) Let Ω1 , Ω2 be two disjoint open subsets of Ω and 0 ∈
/
(I − T )(Ω − Ω1 ∪ Ω2 ). Then

deg(I − T, Ω, 0) = deg(I − T, Ω1 , 0) + deg(I − T, Ω2 ).

Proof. The proof follows from the corresponding properties of the Brouwer
degree.
The following is the well-known Schauder fixed point theorem:
Theorem 2.2.5. Let C ⊂ E be a nonempty bounded closed convex subset
and T : C → C be a continuous compact mapping. Then T has a fixed point
in C.
Proof. The proof is the same as the proof of Brouwer’s fixed point theorem.
If we only require the continuous condition on the mapping T , then the
conclusion of Theorem 2.2.5 fails as the following example shows:
Example 2.2.6. Let T : l2 → l2 be a mapping defined by

T (x1 , x2 , · · · ) = (1 − kxk, x1 , x2 , · · · )

for all x = (x1 , x2 , · · · ) ∈ l2 . Then T : B(0, 1) → B(0, 1) is continuous without

a fixed point in B(0, R).
The Schauder fixed point theorem can be applied to yield the following
result on hyper-invariant subspaces of a linear bounded operator:
Theorem 2.2.7. Let E be a Banach space, T : E → E be a nonzero linear
continuous compact mapping and Γ(T ) = {S ∈ L(E) : T S = ST }. Then

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 33

there exists a nontrival hyper-invariant subspace F of T , i.e., SF ⊆ F for all

S ∈ Γ(T ).
Proof. Assume that the conclusion is not true. Then T does not have
eignvalues. Take x0 ∈ E such that kx0 k = 2. Then, for any x ∈ E, we have
T x 6= 0 and {ST x : S ∈ Γ(T )} is a space which is invariant under Γ(T ) and
so we have
{Sx : S ∈ Γ(T )} = E.
Now, for any y ∈ B(x0 , 1), there exists Sy ∈ Γ(T ) such that kSy T y − x0 k <
1. By continuity of S, there exists δ(y) > 0 such that kSy T x − x0 k < 1 for
all x ∈ B(x0 , 1). Since {B(y, δ(y)) : y ∈ B(x0 , 1)} is an open covering of
B(x0 , 1), there exists a locally finite open refinement {Vi }i∈I of {B(y, δ(y)) :
y ∈ B(x0 , 1)}.
Let {φi }i∈I be a partition of unity subordinated to {Vi }i∈I and define a
mapping K : B(x0 , 1) → B(x0 , 1) as follows:
Kx = Σi∈I φi (x)Si T x for all x ∈ B(x0 , 1).
Since T is compact, K is a continuous compact mapping and hence T has a
fixed point y0 ∈ B(x0 , 1), i.e.,
Σi∈I φi (y0 )Si T y0 = y0 .
Put Z = {y ∈ E : Σi∈I φi (y0 )Si T y = y}. Then Z is a finite dimensional
subspace of E, and T : Z → Z and hence T has an eigenvalue, which is a
contradiction. This completes the proof.
Theorem 2.2.8. Let E be a Banach space and Ω ⊂ E be an open bounded
subset. If T : Ω → E, S : E → E are continuous compact mappings and
p∈/ (I − S)(I − T )(∂Ω), then
deg((I − S)(I − T ), Ω, p) = Σi∈I deg(I − T, Ω, Ui )deg(I − S, Ui , p), (2.2.1)
where {Ui }i∈I are connected component of E\(I−T )(∂Ω) and deg(I−T, Ω, Ui )
is deg(I − T, Ω, z) for any z ∈ Ui .
Proof. We first prove that (2.2.1) only has finitely many nonzero terms.
Take r > 0 such that (I − T )(Ω) ⊂ Br (0), then M = Br (0) ∩ (I − S)−1 (p)
is compact, M ⊂ Rn \ f (∂Ω) = ∪i≥1 Ui and there exists finitely many i, say
i = 1, 2, · · · , t, such that ∪t+1
i=1 Ui ⊇ M , where Ut+1 = U∞ ∩ Br+1 .
We have deg(I − T, Ω, Ut+1 ) = 0 and deg(I − S, Ui , p) = 0 for i ≥ t + 2 since
Uj ⊂ Br (0) and g −1 (y) ∩ Uj = ∅ for j ≥ t + 2. Therefore, the right side of
(2.2.1) has only finitely many terms different from zero. Let
Vm = {z ∈ Br+1 (0) \ (I − T )(∂Ω) : deg(I − T, Ω, z) = m}.
The same proof as in Theorem 1.2.13 yields
Σi∈I deg(I − T, Ω, Ui )deg(I − S, Ui , p) = Σm mdeg(I − S, Vm , p).

Copyright 2006 by Taylor & Francis Group, LLC

 34 Topological Degree Theory and Applications

Now, we may choose  sufficiently small. Let F be a finite dimensional sub-

space, p ∈ F and T1 : Ω → F , S1 : Br+1 (0) → F be two continuous compact
mappings such that

kT x − T1 xk < , x ∈ Ω, kSy − S1 yk <  for all y ∈ Br+1 (0).

Then, by the homotopy argument, we have

deg((I − S)(I − T ), Ω, p) = deg((I − S1 )(I − T1 ), Ω, p)

and

Σm mdeg(I − S, Vm , p) = Σm mdeg(I − S1 , Vm , p)
= Σm mdeg(I − S1 , Vm ∩ F, p).

Therefore, by Theorem 1.2.13, we have

Σi∈I deg(I − T, Ω, Ui )deg(I − S, Ui , p) = deg((I − S1 )(I − T1 ), Ω ∩ F, p).

One easily gets

deg((I − S1 )(I − T1 ), Ω, p) = deg((I − S1 )(I − T1 ), Ω ∩ F, p).

Thus the conclusion of Theorem 2.2.8 is true. This completes the proof.
Theorem 2.2.9. Let E be a Banach space, E0 be a closed subspace of
E and Ω ⊂ E be an open bounded subset. If T : Ω → E0 is a continuous
compact mapping and p ∈ E0 , then deg(I − T, Ω, p) = deg(I − T, Ω ∩ E0 , p).
Proof. Since T (Ω) ⊂ E0 , we may choose a finite dimensional space F ⊂ E0
with p ∈ F and T1 : Ω → F such that kT x − T1 xk <  in Definition 2.2.3 for
small  > 0. Then we have

deg(I − T, Ω, p) = deg(I − T1 , Ω ∩ F, p) = deg(I − T, Ω ∩ E0 , p).

Theorem 2.2.10. Let E be a Banach space and 0 ∈ Ω ⊂ E with Ω be an

open bounded subset. If T : Ω → E is a continuous compact mapping, then
one of the following statements holds:
(1) T has a fixed point in Ω;
(2) There exist λ > 1 and x ∈ ∂Ω such that T x = λx.

Proof. If (2) holds, we are finished. Otherwise, put H(t, x) = x − tT x for

all (t, x) ∈ [0, 1] × Ω. If T x = x for some x ∈ ∂Ω, then (1) holds. Thus, we
may assume that T x 6= x for all x ∈ ∂Ω. Therefore, we have

x∈
/ tT x for all (t, x) ∈ [0, 1] × ∂Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 35

By Theorem 2.2.4, we have

deg(I − T, Ω, 0) = deg(I, Ω, 0) = 1

and so T has a fixed point in Ω. This completes the proof.

Lemma 2.2.11. Let E be an infinite dimensional Banach space and 0 ∈
/ ∂Ω
with Ω be an open bounded subset of E. Let T : Ω → E be a continuous
compact mapping. Suppose that T x 6= µx for all µ ∈ [0, 1], x ∈ ∂Ω and
0∈/ T ∂Ω. Then deg(I − T, Ω, 0) = 0.
Proof. First, we claim that there exists 0 > 0 such that

kT x − T xk < , µx 6= T x

for all µ ∈ [0, 1], x ∈ ∂Ω and  ∈ (0, ), where T is the same as in Lemma
2.2.1.
If this is not true, there exist j → 0, xj ∈ ∂Ω, µj → µ0 ∈ [0, 1] such that
µj xj = Tj xj and so we have T xj − µj xj → 0. Now, we have 0 ∈ / T ∂Ω and
so µ0 6= 0, (xj ) has a subsequence converging to x0 ∈ ∂Ω and T x0 = µ0 x0 ,
which is a contradiction.
From the definition of the Leray Schauder degree, we know that

deg(I − T, Ω, 0) = deg(I − T , Ω ∩ F, 0)

for sufficiently small  and any F ⊃ spanR(T ). The homotopy invariance of

Brouwer degree implies that

deg(I − T , Ω ∩ F, 0) = deg(−T , Ω ∩ F, 0).

Since E is an infinite dimensional Banach space, we may choose a finite di-

mensional subspace F of E such that spanR(T ) is a proper subspace of F ,
and deg(−T , Ω ∩ F, 0) = deg(−T , Ω ∩ F, p) for any p ∈ F with p sufficiently
close to 0, so we must have deg(−T , Ω ∩ F, 0) = 0. Thus it follows that

deg(I − T, Ω, 0) = 0.

This completes the proof.

Theorem 2.2.12. Let E be an infinite dimensional Banach space, 0 ∈ Ω0 ⊂
Ω with Ω0 and Ω be two open bounded subsets of E. Let T : Ω \ Ω0 → E
be a continuous compact mapping and suppose that the following conditions
hold:

(1) T x 6= λx for all λ > 1, x ∈ ∂Ω0 ,

(2) T x 6= µx for all µ ∈ [0, 1), x ∈ ∂Ω, and 0 ∈

/ T ∂Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 36 Topological Degree Theory and Applications

Then T has a fixed point in Ω \ Ω0 .

Proof. We may assume that T is defined on Ω. Also, we assume that
T x 6= x for all x ∈ ∂Ω0 ∪ ∂Ω. From (1) and Theorem 2.2.4, we have deg(I −
T, Ω0 , 0) = 1.
By (2) and Lemma 2.2.11, deg(I − T, Ω, 0) = 0. Therefore, we have

deg(I − T, Ω \ Ω0 , 0) = deg(I − T, Ω, 0) − deg(I − T, Ω0 , 0) = −1.

Consequently, T has a fixed point in Ω \ Ω0 . This completes the proof.

From Theorem 2.2.12, we easily get the following result of Guo [141]:
Corollary 2.2.13. Let E be a infinite Banach space and Ω0 ⊂ Ω with 0 ∈
Ω0 , Ω be two open bounded subsets of E. Let T : Ω \ Ω0 → E be a continuous
compact mapping and, further, suppose that the following conditions hold:
(1) kT xk ≤ kxk for all x ∈ ∂Ω0 ;
(2) kT xk ≥ kxk for all x ∈ ∂Ω.
Then T a fixed point in Ω \ Ω0 .
Remark. The conclusion of Theorem 2.2.12 fails if we drop the infinite
dimensional condition. The following example illustrates this:
Example 2.2.14. Let Ω = {(x, y) : x2 +y 2 < 4}, Ω1 = {(x, y) : x2 +y 2 < 1}
and T : Ω \ Ω1 → R2 be a rotation defined by
p  π p 2  π 
T (x, y) = x2 + y 2 cos θ + , x + y 2 sin θ +
4 4
p
for all (x, y) ∈ Ω \ Ω1 , where x + yi = x2 + y 2 eiθ . Then we have T (x, y) 6=
µ(x, y) for all µ ∈ [0, 1], x2 + y 2 = 1 and T (x, y) 6= λ(x, y) for all λ ≥ 1,
x2 + y 2 = 4. However, the mapping T does not have a fixed point in Ω \ Ω1 .
We have the following result in any dimensional Banach spaces:
Theorem 2.2.15. Let E be a Banach space and 0 ∈ Ω0 ⊂ Ω with Ω0 , Ω
two open bounded subsets of E. Let T : Ω \ Ω0 → E be a continuous compact
mapping and, further, suppose that the following conditions hold:
(1) T x 6= λx for all λ > 1 and x ∈ ∂Ω0 ;
(2) T x 6= µx for all µ ∈ [0, 1), x ∈ ∂Ω and 0 ∈
/ ConvT ∂Ω.
Then T has a fixed point in Ω \ Ω0 .
Proof. As in the proof of Theorem 2.2.12, we may assume that T x 6= x for
x ∈ ∂Ω0 ∪ ∂Ω. We only need to show that

deg(I − T, Ω, 0) = 0.

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 37

Assume that this is not true. Then there exists a compact mapping T1 : Ω →
ConvT ∂Ω such that T1 x = T x for x ∈ ∂Ω. For k > 1, it is easy to see that
x 6= tT x + (1 − t)kT1 x for all (t, x) ∈ [0, 1] × ∂Ω.
Thus we have
deg(I − kT1 , Ω, 0) = deg(I − T, Ω, 0) 6= 0
and so kT1 x = x has a solution in Ω for k > 1, which contradicts the fact
that Ω is unbounded. Hence we have deg(I − T, Ω, 0) = 1. This completes
the proof.
Theorem 2.2.16. Let E be a Banach space, Ω with 0 ∈ Ω be an open
bounded subset of E and L : Ω → E be a linear continuous compact mapping.
If λ 6= 0 and λ−1 is not an eigenvalue of L, then
deg(I − λL, Ω, 0) = (−1)m(λ) ,
where m(λ) is the sum of the algebraic multiplicities of the eigenvalues µ
satisfying µλ > 1, and, if L has no such eigenvalues µ, then m(λ) = 0.
Proof. Put S = I − λL, then S is a homeomorphism onto E. There are at
most finitely many eigenvalues of L such that µλ > 1, say, µi , i = 1, 2, · · · , k.
Set F = ⊕ki=1 N (µi ) and W = ∩ki=1 R(µi ), where N (µi ) = {x : Lx − µi x = 0}
and R(µi ) = (L − µi )(E) for i = 1, 2, · · · , k. We know from the spectral
theory of linear compact mappings that N (µi ) are finitely dimensional spaces,
i = 1, 2, · · · , k. It is easy to see that E = F ⊕ W . There are projections
P : E → F , and Q : E → W . Set L1 = SP + Q and L2 = P + SQ. Then we
have
I − L1 = −λLP1 , I − L2 = −λLQ, S = L1 L2 ,
(I − L1 )(F ) ⊂ F, (I − L2 )(W ) ⊂ W.
Moreover, Li is one to one for i = 1, 2. Thus, by product formula, we have
deg(S, Ω, 0) = deg(L1 L2 , Ω, 0) = deg(L1 , Ω, 0)deg(L2 , Ω, 0).
By Theorem 2.2.9, we have
deg(L1 , Ω, 0) = deg(L1 , Ω ∩ F, 0)
and
deg((L2 , Ω, 0) = deg(L2 , Ω ∩ W, 0).
But I − tλL has no solution on ∂Ω ∩ W and so we have
deg(L2 , Ω ∩ W, 0) = deg(I, Ω ∩ W, 0) = 1.
On the other hand, the eigenvalues of L1 are the eigenvalues of I − λL on
F , i.e., 1 − λµi for i = 1, 2, · · · , k. Thus we have
deg(L1 , Ω ∩ F, 0) = sgndet(L1 |F ) = sgnΠki=1 (1 − λµi )dim(N (µi )
k
= (−)Σi=1 dim(N (µi )) = (−1)m(λ) .

Copyright 2006 by Taylor & Francis Group, LLC

 38 Topological Degree Theory and Applications

2.3 Leray Schauder Degree for Multi-Valued Mappings

In this section, we describe the Leray Schauder degree for upper semicon-
tinuous compact mapping with closed convex values. First, we introduce
several multi-valued maps, which play very important roles in the study of
nonsmooth analysis, differential inclusions and nonlinear partial differential
equations. We begin with the following definitions:
Definition 2.3.1. Let X, Y be two topological spaces and T : X → 2Y be
a multi-valued mapping, i.e., T x is a subset of Y for all x ∈ X.
(1) T is said to be lower semicontinuous at x0 if, for any open set V of Y
with T x0 ∩V 6= ∅, the set U = {x : T x∩V 6= ∅} is open in X, and if T is lower
semicontinuous at every point of X, T is said to be lower semicontinuous on
X,
(2) T is said to be upper semicontinuous at x0 if, for any open neighborhood
V (T x0 ) of T x0 , there exists an open neighborhood of U (x0 ) such that T U ⊂
V , and T is said to be upper semicontinuous on X if T is upper semicontinuous
at every point of X,
(3) T is said to be continuous at x0 if T is both upper semicontinuous and
lower semicontinuous at x0 , and if T is continuous at all points of X, then T
is said to be continuous on X.
Definition 2.3.2. Let X be a topological space and Y be a normed space.
A mapping T : X → 2Y is called Hausdorff continuous at x0 if

lim H(T x, T x0 ) = 0
x→x0

and, if T is Hausdorff continuous at all points of X, then T is said to be

Hausdorff continuous on X.
Next, we give some examples of multi-valued mappings, which have ap-
peared in several fields of mathematics.
Example 2.3.3. Let f : Rn → Rn be a bounded function (not necessarily
n
continuous). We define a multi-valued mapping F : Rn → 2R as follows:

F x = {y : there exist xj ∈ Rn such that xj → x, f xj → y}.

Then F is upper semicontinuous. Such a mapping has been used to study

differential equations with discontinuous right-hand sides (see [12], [114]).
Example 2.3.4. Let X be a Banach space and f : X → R be a locally
Lipschitz function. The Clarke derivative of f at x in the direction v is defined
by
f (y + tv) − f (y)
f o (x; v) = lim sup
t↓0,y→x t

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 39

and the Clarke subdifferential of f at x is defined by

∂f (x) = {φ ∈ X ∗ : φ(v) ≤ f o (x; v)}.

This subdifferential is a multi-valued mapping and it is a powerful tool in

non-smooth analysis (see [72]).
Example 2.3.5. Let u : Rn → R be a continuous function. We define
n u(y) − u(x) − p(y − x) o
D+ u(x) = p ∈ RN : lim sup ≤0 ,
y→x,y∈Ω |y − x|
n u(y) − u(x) − p(y − x) o
D− u(x) = p ∈ RN : lim inf ≥0 .
y→x,y∈Ω |y − x|
D+ u and D− u are used to define the viscosity solutions for fully nonlinear
partial differential equations (see [75]).
Example 2.3.6. Let X be a infinite Banach space, r > 0 be a constant
and F : X → 2X be defined as follows:

F x = B(x, r) for all x ∈ X.

Then F is lower semicontinuous on X, but not upper semicontinuous.

Indeed, for any x0 ∈ X and y ∈ B(x0 , r), we have (y − x0 ) + x ∈ F (x) =
B(x, r) and so (y − x0 ) + x → y as x → x0 . Therefore, F is lower semicontin-
uous at x0 . To see that F is not upper semicontinuous, we know, since X is
infinite dimensional, that there exists a sequence {xn } ⊂ X such that

kxn k = 1, kxn − xm k ≥ ,

where  ∈ (0, 1) is a constant. Then we have (r + n1 )xn + x0 ∈ F (x0 + n1 xn ),

but V = X \ {(r + n1 )xn + x0 : n = 1, 2 · · · , } is open and F (x0 , r) ⊂ V .
Therefore, F (x0 + n1 xn ) 6⊂ V , i.e., F is not upper semicontinuous. Moreover,
it is easy to check that F is Hausdorff continuous on X.
Lemma 2.3.7. Let X be a metric space, Y be a normed space and T :
X → 2Y be an upper semicontinuous mapping with closed convex values.
Then, for any  > 0, there exists a continuous mapping f : X → conv(T X)
such that, for any x ∈ X, there exist y ∈ X and z ∈ T y such that

d(x, y) < , kf x − zk < .

Proof. For any x ∈ X and  > 0, there exists δx > 0 such that

T B(x, δx ) ⊂ B(T x, ).

We may require δx < . Let {Ui }i∈I be a locally finite open refinement of
{B(x, δ2x ) : x ∈ X} and {φi }i∈I be a partition of the unity subordinated
{Ui }i∈I .

Copyright 2006 by Taylor & Francis Group, LLC

 40 Topological Degree Theory and Applications

Now, we define a mapping f : X → Y as follows:

f x = Σi∈I φi (x)yi for all x ∈ X,

where Ui ⊂ B(xi , δ2x ) and yi ∈ T xi . It is obvious that f : X → conv(T X) is

continuous. For any given x ∈ X, let I0 = {i ∈ I : φi (x) 6= 0}. Then there
exists i0 ∈ I0 such that
δxi0 = max{δxi }.
i∈I0
δxi
We put y = xi0 . For i ∈ I0 , we have x ∈ Ui ⊂ B(xi , 2 ) and hence xi ∈
B(xi0 , δxi0 ). Therefore, we have

f x = Σi∈I0 φi (x)yi ∈ B(T y, ).

Take z ∈ T y such that kf x − zk < . This completes the proof.

Theorem 2.3.8. Let X be a metric space, Y be a Banach space and
T : X → 2Y be a lower semicontinuous mapping with closed convex values.
Then there exists a single-valued continuous mapping f : X → Y such that
f (x) ∈ T x for all x ∈ X.
Proof. First, we prove that, for any  > 0, there exists a continuous
mapping f : X → Y such that f (x) ∈ T x + B (0) for all x ∈ X. To see this,
if we put U (y) = {x ∈ X : x ∈ T −1 (y − B )}, then U (y) is open in X by
the lower semicontinuity of T . Now, since we have X = ∪y∈Y U (y), there is a
locally finite open refinement {Wi }i∈I of {U (y)}y∈Y and a partition of unity
{αi }i∈I subordinated to {Wi }i∈I . Pick yi such that Wi ⊂ U (yi ) and define a
mapping f : X → Y as follows:

f (x) = Σi∈I αi (x)yi for all x ∈ X.

Then f is continuous. Obviously, if αi (x) 6= 0, then x ∈ Wi ⊂ U (yi ) and so

we have yi ∈ F x + B (0). Thus it follows that

f (x) ∈ F (x) + B (0)

since the right hand side is convex. Now, for n = 2−n for n = 1, 2, · · · , by
the above conclusion, we have a sequence {fn } of continuous mappings such
that
fn ∈ fn−1 (x) + 2Bn (0), fn (x) ∈ T x + Bn (0)
for all x ∈ X. Indeed, assume that we have defined mappings f1 , f2 , · · · , fn
for some n > 1, respectively, and put Gx = T x ∩ (fn + B (0)) for all x ∈ X.
Then G is lower semicontinuous with convex values.
By the first step, there exists fn+1 : X → Y such that fn+1 (x) ∈ G(x) +
Bn+1 and so we have

fn+1 (x) ∈ T (x) + Bn+1 , fn+1 (x) ∈ fn (x) + 2Bn .

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 41

Evidently, since {fn } is a Cauchy sequence, let f (x) = limn→∞ fn (x). Then
f is continuous since the convergence is uniform and f (x) ∈ T x for all x ∈ X.
This completes the proof.
In the above result, we considered approximate selections and continuous
selections of multi-valued mappings. In some cases, we need to find a measur-
able selection for a given multi-valued mapping. Let (Ω, A) be a measurable
space, (X, d) be a separable metric space and f : Ω → 2X be a multi-valued
mapping. Then f is called a measurable function if f −1 (B) ∈ A for all open
subset B ⊂ X.
Theorem 2.3.9. Let (Ω, A) be a measurable space, (X, d) be a separable
complete metric space and F : Ω → 2X be a measurable multi-valued mapping
with closed values. Then there exists a single valued measurable mapping
f : Ω → X such that f (x) ∈ F x for all x ∈ Ω.
Proof. Since X is separable, there exists a countable subset {x1 , x2 , · · · }
of X such that
{x1 , x2 , · · · } = X.
We shall define a sequence {fn } of measurable functions satisfying the follow-
ing:
(1) d(fn (x), fn+1 (x)) < 2−n+1 for all x ∈ Ω;
(2) d(fn (x), F (x)) < 2−n for all x ∈ Ω
and define a mapping f1 : Ω → X by f1 (z) = xk if k is the smallest integer
such that
F z ∩ B(xk , 1) 6= ∅.
Since f −1 (xk ) = F −1 (B(xk , 1) \ ∪m<k F −1 (B(xm , 1)), it follows that f1 is
measurable.
Now, assume that we have defined f1 , f2 , · · · , fk satisfying (1) and (2)
for some k > 1. For each z ∈ Ω, we have z ∈ fk−1 (xi ) for some i. Now, if
z ∈ fk−1 (xi ), then we define

fk+1 (z) = xp ,

where p is the smallest integer such that

F z ∩ B(xi , 2−k ) ∩ B(xp , 2−k−1 ) 6= ∅.

Obviously, fk+1 is well defined and it is also measurable and, moreover, it

satisfies (1) and (2).
Finally, if we put f (x) = limn→∞ fn (x), then f is measurable and f (x) ∈
F x for all x ∈ Ω. This completes the proof.
Example 2.3.10. Let F : Rn → 2R \ ∅ be a upper semicontinuous function
with bounded closed values. Assume that |y| ≤ M |x| + f (x) for all y ∈ F x

Copyright 2006 by Taylor & Francis Group, LLC

 42 Topological Degree Theory and Applications

and x ∈ Rn , where M > 0 is a constant and f (·) ∈ Lp (Rn ) for p ∈ [1, ∞).
We define a mapping F : C([0, 1]; Rn ) → Lp (Rn ) by

Fu(t) = {g(·) ∈ Lp ([0, 1]) : g(t) ∈ F u(t), a.e.t ∈ [0, 1]}

for all u(·) ∈ C([0, 1]; Rn ). Since F u(t) : [0, 1] → 2R is upper semicontinuous
with closed values, it is measurable. By Theorem 2.3.9, we know that there
exists a measurable selection g(t) ∈ F u(t) and, by assumption, we have

|g(t)| ≤ M |u(t)| + f (u(t)) for all t ∈ [0, 1].

Therefore, it follows that g(·) ∈ Lp ([0, 1]) and hence F is well defined.
Next, we show that the Leray Schauder degree can be generalized to multi-
valued upper semicontinuous compact mappings with closed convex values:
Proposition 2.3.11. Let E be a real Banach space, Ω ⊂ E be an open
bounded set and T : Ω → 2E be an upper semicontinuous mapping with
closed convex values. If T Ω is relatively compact and x ∈
/ T x for all x ∈ ∂Ω,
then there exists 0 > 0 such that x 6= f x for all x ∈ ∂Ω and  ∈ (0, 0 ),
where f is defined as in Lemma 2.3.7.
Proof. Suppose that the conclusion is not true. Then there exist j → 0
and xj ∈ ∂Ω such that xj = fj xj . By Lemma 2.3.7, there exist yj ∈ Ω and
zj ∈ T yj such that

kxj − yj k ≤ j , kfj xj − zj k < j .

Since T is compact, we may assume that zj → z0 and hence we have

fj xj → z0 , xj → z0 ∈ ∂Ω

and thus yj → z0 . Using the upper semicontinuity of T , we get z0 ∈ T z0 ,

which is a contradiction. This completes the proof.
Definition 2.3.12. Let E be a real Banach space, Ω ⊂ E be an open
bounded set and T : Ω → 2E be an upper semicontinuous mapping with
closed convex values. Suppose that T Ω is relatively compact and x ∈
/ T x for
all x ∈ ∂Ω. Then we define

deg(I − T, Ω, 0) = lim deg(I − f , Ω, 0),

→0

where f is defined as in Lemma 2.3.7.

We show that Definition 2.3.12 is reasonable. By Proposition 2.3.11, there
exists 0 > 0 such that x 6= f x for all x ∈ ∂Ω and  ∈ (0, 0 ). It is obvious
that f (Ω) is compact. Thus deg(I − f , Ω, 0) is well defined for all  ∈ (0, 0 ).
We claim that there exists 1 < 0 such that tf x + (1 − t)fδ x 6= x for all

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 43

(t, x) ∈ [0, 1] × ∂Ω and , δ ∈ (0, 1 ). If not, there exist tj → t0 , j → 0,

δj → 0, and xj ∈ ∂Ω such that

tj fj xj + (1 − tj )fδj xj = xj .

By Lemma 2.3.7, there exist yj1 , yj2 and zj1 ∈ T yj1 , zj2 ∈ T yj2 satisfying

kyj1 − xj k < j , kyj2 − xj k < δj , kzj1 − fj xj k < j , kzj2 − fδj xj k < δj .

By compactness of T , we may assume that xj → x0 ∈ ∂Ω. Therefore, yj1 → x0

and yj2 → x0 . Consequently, (zj1 ) has a subsequence (zj1k ) with zj1k → z1 ∈
T x0 , (zj2 ) has a subsequence (zj2k ) with zj2k → z2 ∈ T x0 and tz1 + (1 − t)z2 =
x0 ∈ T x0 , which is a contradiction.
The following property follows from Theorem 2.2.4, and we leave the proofs
to the reader:
Theorem 2.3.13. The degree defined by Definition 2.3.12 has the following
properties:
(1) (N ormality) deg(I, Ω, 0) = 1 if and only if 0 ∈ Ω;
(2) (Solvability) If deg(I − T, Ω, 0) 6= 0, then x ∈ T x has a solution in Ω;
(3) (Homotopy) Let Tt : [0, 1]×Ω → E be a upper semicontinuous compact
mapping with closed convex values and x ∈/ Tt x for all (t, x) ∈ [0, 1]×∂Ω.
Then deg(I − Tt , Ω, 0) does not depend on t ∈ [0, 1];
(4) (Additivity) If Ω1 , Ω2 are two disjoint open subsets of Ω and 0 ∈
/ (I −
T )(Ω − Ω1 ∪ Ω2 ), then

deg(I − T, Ω, 0) = deg(I − T, Ω1 , 0) + deg(I − T, Ω2 ).

2.4 Applications to Bifurcations

In this section, we give some applications to bifurcation.
Definition 2.4.1. Let X, Y be two Banach spaces, α0 > 0, δ > 0, I =
(−α0 − δ, α0 + δ), x0 ∈ Ω ⊂ X be an open subset and F : I × Ω → Y be such
that F (α, x0 ) = 0 on I. If there exists αn → α0 , xn ∈ Ω \ {x0 }, xn → x0 such
that F (αn , xn ) = 0, then we call (α0 , x0 ) a bifurcation point for F (α, x) = 0.
Theorem 2.4.2. Let X be an infinite dimensional Banach space, Ω ⊂ X
be an open subset with 0 ∈ Ω and F : Ω → X be a continuous compact
operator with F 0 = 0. Suppose lim inf x→0 kF xk
kxk = +∞. Then (0, 0) is a
bifurcation point for x − αF x = 0.

Copyright 2006 by Taylor & Francis Group, LLC

 44 Topological Degree Theory and Applications

Proof. Take αn → 0. Consider the mapping H(t, x) = x − αn tF x for

(t, x) ∈ [0, 1] × Ω. Since lim inf x→0 kF xk 1
kxk = +∞, there exist rn < n such that

kF xk > αn−1 kxk, x ∈ ∂B(0, rn ).

By Lemma 2.2.11, we have deg(I − αn F, B(0, rn ), 0) = 0. However, it follows

that
deg(I, B(0, rn ), 0) = 1,
so there must exist tn ∈ (0, 1), xn ∈ ∂B(0, rn ) such that H(tn , xn ) = 0, i.e.,
xn − αn tn F xn = 0. Thus (0, 0) is a bifurcation point. This completes the
proof.
Lemma 2.4.3. Let (M, d) be a compact metric space, A1 ⊂ M be a
component and A2 ⊂ M closed such that A1 ∩ A2 = ∅. Then there exist
compact sets M1 , M2 such that Ai ⊂ Mi , i = 1, 2, M1 ∩ M2 = ∅ and M =
M1 ∪ M2 .
Proof. For  > 0, a, b ∈ M , if there exist finitely many xi ∈ M , i =
1, 2, · · · , n, such that x1 = a, xn = b and d(xi+1 , xi ) <  for i = 1, 2, · · · , n − 1,
then we call a, b  chainable. Put

A = {x ∈ M : there exists a ∈ A1 such that x and aare chainable}.

Clearly, A1 ⊂ A and A is both open and closed in M .

Now, we prove that there exists 0 > 0 such that A0 ∩ A2 = ∅. If this is not
true, for j → 0, there exist yj ∈ Aj ∩ A2 . So there exist aj ∈ A1 such that
yj , aj are j -chainable. However, A1 , A2 are compact, so we may assume that
aj → x0 , bj → y0 . Consequently, we have j chains joining x0 , y0 for every
j ≥ 1. We set

C = {x ∈ m : there exist jk → ∞, xjk ∈ Mjk such that xjk → x}.

Obviously, C is compact and x0 , y0 ∈ C. Notice that any two points in C

are j -chainable, so C must be connected. However, C ∩ A1 6= ∅, so we have
C ⊂ A1 , and thus y0 ∈ A1 ∩ A2 , which is a contradiction. Therefore, there
exist 0 > 0 such that A0 ∩ A2 = ∅. Set M1 = A0 and M2 = M \ M1 , then
we get the desired result.
Theorem 2.4.4. Let X be a real Banach space, Ω ⊂ R × X be a open
neighborhood of (α0 , 0) and T : Ω → X be a continuous compact mapping
with T (α, x) = o(kxk) as x → 0 uniformly in α. Let S : X → X be a
linear continuous compact mapping, α0−1 be an eigenvalue of odd algebraic
multiplicity and

M = {(α, x) ∈ Ω : x − αSx + T (α, x) = 0, x 6= 0}.

Then the component C of M containing (α0 , 0) has at least one of the following
properties:

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 45

(1) C ∩ ∂Ω 6= ∅.
(2) C contains an odd number of trivial zeros (αi , 0) 6= (α0 , 0), where αi−1
is an eigenvalue of S of odd algebraic multiplicity.
Proof. First, if C ∩ ∂Ω 6= ∅, then C is compact and contains another (α, 0)
with α 6= α0 . The compactness of C follows from the compactness of S and
T . Suppose C ∩ R × {0} = {(α0 , 0)}. For any δ > 0, set

Nδ = {(α, x) ∈ Ω : d((α, x), C) < δ}.

If Nδ ∩ M = C, we put Ω0 = Nδ . Otherwise, by Lemma 2.4.3, there exist

compact sets C1 , C2 such that C ⊂ C1 , M ∩ ∂Nδ ⊂ C2 , C1 ∩ C2 = ∅ and
C1 ∪ C2 = Nδ ∩ M . Set γ = d(C1 , C2 ), then γ > 0. We put Ω0 = Nδ ∩ {x :
d(x, C1 ) < γ2 }. It is obvious that

C ⊂ Ω0 ⊂ Ω0 ⊂ Ω, M ∩ ∂Ω0 = ∅.

Now, we may take δ > 0 small enough such that no other eigenvalue α−1 of
S satisfies |α − α0 | ≤ 2δ and the intersection of M and the real line is given
by I = [α0 − δ, α0 + δ]. Since M ∩ ∂Ω0 = ∅, deg(I − αS − T (α, ·), Ω(α), 0) is
constant on I, where Ω(α) = {x : (α, x) ∈ Ω0 }.
Now, we choose α0 − δ < α1 < α0 < α0 + δ. For r sufficiently small, we
have

deg(I − αi S − T (αi , ·), Ω(αi ), 0)

= deg(I − αi S − T (αi , ·), Ω(αi ) \ B(0, r), 0)
+deg(I − αi S − T (αi , ·), B(0, r), 0)

for i = 1, 2. However, deg(I − α1 S, T (α1 , ·), B(0, r), 0) is different to deg(I −

α2 S, T (α2 , ·), B(0, r)Ω, 0) by a factor −1 and deg(I − αi S − T (αi , ·), Ω(αi ) \
B(0, r), 0) = 0, which is a contradiction. Clearly, a bounded open neighbor-
hood Ω0 of C satisfying M ∩∂Ω0 = ∅ contains a finite number of points (αi , 0)
with αi−1 ∈ σ(S), say, α1 < · · · < αj−1 < α0 < αj+1 < · · · < αN . We may
take δ > 0 sufficiently small such that

Ω0 ∩ R × {0} = (∪N
i=1 [αi − δ, αi + δ]) × {0}.

Choose αj1 , αj2 such that

αj − δ < αj1 < αj < αj2 < αj + δ.

We have deg(I − αS − T (α, ·), Ω0 (α), 0) = m on [α1 − δ, αN + δ] for some

m ∈ Z and

m = deg(I − αji S − T (αji , ·), Ω0 (αji ), 0)

= deg(I − αji S − T (αji , ·), Ω0 (αji ) \ B(0, r), 0)
+deg(I − αji S − T (αji , ·), B(0, r), 0)

Copyright 2006 by Taylor & Francis Group, LLC

 46 Topological Degree Theory and Applications

for i = 1, 2, where r > 0 is sufficiently small.

Moreover, it follows that

deg(I − α11 S − T (α11 , ·), Ω0 (α11 ) \ B(0, r), 0)

= deg(I − αN 2 S − T (αN 2 , ·), Ω0 (αN 2 ) \ B(0, r), 0)
=0

and

deg(I − αj2 S − T (αj2 , ·), Ω0 (αj2 ) \ B(0, r), 0)

= deg(I − α(j+1)1 S − T (α(j+1)1 , ·), Ω0 (α(j+1)1 ) \ B(0, r), 0)
= 0.

Therefore, we have
−1
ΣN
j=1 [deg(I − αj2 S − T (αj2 , ·), B(0, r), 0)
−deg(I − αj1 S − T (αj1 , ·), B(0, r), 0) = 0.

Since the degree has a jump at α0 and the jumps occur only at eigenvalues
of odd algebraic multiplicity, it follows that the degree has an even number
of jumps. Consequently, C contains an odd number of trivial zeros (αi , 0) 6=
(α0 , 0).

2.5 Applications to ODEs and PDEs

In this section, we give some applications of the Leray Schauder theory to
existence of solutions for ordinary differential equations and partial differential
equations.
Theorem 2.5.1. (Peano’s Theorem) Let f : R × B(x0 , r) ⊂ Rn → Rn
be a continuous function. Then there exists t0 > 0 such that the following
equation: (
x0 (t) = f (t, x(t)), t ∈ (0, t0 ),
(E 2.5.1)
x(0) = x0
has a solution.
Proof. Since f is continuous, there exist t1 > 0, r1 < r, and M > 0 such
that
|f (t, x)| ≤ M, t ∈ [−t1 , t1 ], |x − x0 | ≤ r1 ,
where | · | is the norm in Rn . Take t0 ∈ (0, t1 ) such that M r0 < r1 . We set

E = C([0, t0 ], Rn ) = {x(t) : [0, t0 ] → Rn is continuous}

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 47

with the norm |x(·)| = maxt∈[0,t0 ] |x(t)|. Then C([0, t0 ], Rn ) is a Banach space.
Put
K = {x(·) ∈ E : x(0) = x0 , |x(t) − x0 | ≤ r1 , t ∈ [0, t0 ]}.
Then it is easy to see that K is a bounded closed convex subset.
Now, we define a mapping T : E → E by
Z t
T x(t) = x0 + f (s, x(s))ds for all x(·) ∈ E.
0

It is easy to check that T is continuous. Moreover, T : K → K is a mapping

and T K is compact. Thus, by Schauder’s
Rt fixed point theorem, T has a fixed
point x(·) in K, i.e., x(t) = x0 + 0 f (s, x(s))ds. Therefore,

x0 (t) = f (t, x(t)) for all t ∈ (0, t0 ).

Thus the problem (E 2.5.1) has a solution. This completes the proof.
In the following, let Ω ⊂ RN be an open bounded subset with smooth
boundary and ai , b : Ω×R×RN → R, i = 1, 2, · · · , N , be continuous functions
such that
∂ai
(1) ∂ηj ξi ξj ≥ |ξ|2 for all (x, z, ξ) ∈ Ω × R × Rn ;

(2) |ai (x, z, 0) ≤ g(z), i = 1, 2, · · · , N , where g(·) ∈ Lq (Ω), and q > N ;

(3) (1 + |ξ|2 )| ∂a ∂ai ∂ai 2

∂ξj | + (1 + |ξ|)(| ∂z | + |ai |) + | ∂xj | + |b| ≤ µ(|z|)(1 + |ξ| ) for
i

i, j = 1, 2 · · · , N , where µ : [0, +∞) → [0, +∞) is a increasing function;

(4) −b(x, z, ξ)signz ≤ L(|ξ| + f (x)) for all (x, z, ξ) ∈ Ω × R × RN , where

L > 0 is a constant.

Consider the following Dirichlet problem:

(
−Di ai (x, u(x), Du(x)) + b(x, u(x), Du(x)) = 0, x ∈ Ω,
(E 2.5.2)
u(x) = φ(x), x ∈ ∂Ω.

Theorem 2.5.2. Suppose that ∂Ω is smooth and (1)-(4) hold and ai ∈

C 1,α (Ω × R × RN ), b ∈ C 0,α (Ω × R × RN ), φ ∈ C 2,α (Ω), where α ∈ (0, 1) is
a constant. Then the problem (E 2.5.2) has a solution u ∈ C 2,α (Ω).
Proof. For each v ∈ C 1,α (Ω) and t ∈ [0, 1], consider the following linear
Dirichlet problem:
 ∂ai
−[t ∂ηj (x, v(x), Dv(x))Dij u(x) + (1 − t)∆u(x)]

∂ai
+t[ ∂xi
+ ∂a
∂zi ηi + b](x,z,η)=(x,v(x),Dv(x)) = 0,
i
x ∈ Ω, (E 2.5.3)

u(x) = tφ(x), x ∈ ∂Ω.


Copyright 2006 by Taylor & Francis Group, LLC

 48 Topological Degree Theory and Applications

Since the coefficients belong to C α (Ω), it is well known that (E 2.5.3) has a
unique solution u ∈ C 2,α (Ω) (see [132]). We define a mapping T : [0, 1] ×
C 1,α (Ω) → C 1,α (Ω) by

T (t, v) = u for all (t, v) ∈ [0, 1] × C 1,α (Ω).

It is easy to see that T is continuous compact. Also, by prior estimation,

there exist constants M > 0, 0 < γ < 1 such that |u|1,γ;Ω ≤ M (see [135]).
Therefore, u ∈ C αγ (Ω) and there exist a constant C which does not depend
on t and u such that
|u|2,αγ ≤ C.
Thus |u|1,α ≤ M . Therefore, T (·, 1) has a fixed point u, i.e., T (1, u) = u,
which is a solution of (E 2.5.2) and u ∈ C 2,α (Ω).
In the following, suppose that H is a real Hilbert space, A : D(A) ⊆ H → H
is a linear self-adjoint operator and F (t, u) : R × H → H is a nonlinear
mapping. Consider the anti-periodic problem:
(
u0 + Au(t) + ∂Gu(t) + F (t, u(t)) = 0, a.e. t ∈ R,
(E 2.5.4)
u(t) = −u(t + T ), t ∈ R.

Definition 2.5.3. A function u(·) is called a weak anti-periodic solution

RT
of (E 2.5.4) if u(t + T ) = −u(t) for t ∈ R, 0 |u0 (t)|2 dt < ∞ and

u0 + Au(t) + ∂Gu(t) + F (t, u(t)) = 0 for almost all t ∈ R.

Lemma 2.5.4. If u, u0 ∈ L2 (0, T ; H) and u(t + T ) = −u(t) for all t ∈ R,

then √ Z T
T  12
|u|∞ ≤ |u0 (s)|2 ds .
2 0
Rt 0 RT
Proof. Since u(t) = u(0) + 0 u (s)ds and u(t) = u(T ) − t u0 (s)ds, we
have
1 t 0
Z Z T
u(t) = [ u (s)ds − u0 (s)ds].
2 0 t
Thus the conclusion follows.
Lemma 2.5.5. Let H be a real separable Hilbert space and A : D(A) ⊆
H → H be a linear densely defined closed self-adjoint operator that only has a
point spectrum, i.e., eigenvalues. Suppose that f : R → H is a T -anti-periodic
function, i.e., f (T + t) = −f (t) for t ∈ R, and f (·) ∈ L2 (0, T ; H). Then the
following problem:
(
u0 + Au(t) + f (t) = 0, a.e.t ∈ R,
(E 2.5.5)
u(t + T ) = −u(t), t ∈ R,

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 49

has a unique weak solution.

Proof. Since σ(A) only has point spectrum and H is separable, A has
a countable family of eigenvalues {λi }∞ i=1 . Assume that {ei : i = 1, 2, · · · }
is the orthogonal family of eigenvectors associated with the eigenvalue λi
satisfying |ei | = 1 for each i, i = 1, 2 · · · . A is densely defined and we have
span{ei : i = 1, 2, · · · } = H, so f (t) = Σ∞ i=1 fi (t)ei , where fi : R → R satisfies
∞
RT 2
Σi=1 0 fi (t)dt < ∞. It is obvious that fi (t + T ) = −fi (t) for t ∈ R.
Now, we consider the one dimensional evolution equation:
(
u0i (t) + λi ui (t) + fi (t) = 0, t ∈ R,
(E 2.5.6)
ui (t + T ) = −ui (t), t ∈ R,

for each i = 1, 2, · · · . By Corollary 1.2 of [148] or by a direct computation,

the problem (E 2.5.6) has a unique solution ui (t).
Now, multiply (E 2.5.6) by u0i (t) and integrate over (0, T ) to get
Z T Z
|ui (t)| dt + fi (t)u0i (t)dt = 0.
0 2
0

Therefore, we have
Z T Z T
|u0i (t)|2 dt ≤ |fi (t)|2 dt. (2.5.1)
0 0
By Lemma 2.5.4, we get
T
T2
Z
|ui |2∞ ≤ |fi (t)|2 dt. (2.5.2)
4 0

Put u(t) = Σ∞ i=1 ui (t)ei . Then it follows from (2.5.2) that u is well defined. By
(2.5.1), we know that u0 (t) = Σ∞ 0 2
i=1 ui (t)ei belongs to L (0, T ; H). Therefore,
∞ 2
Σi=1 λi ui (t)ei belongs to L (0, T ; H). Since A is closed, u(t) ∈ D(A) for
almost all t ∈ R and

Au(t) = Σ∞
i=1 λi ui (t)ei for almost all t ∈ R.

In view of (E 2.5.6), we know that u is a weak solution of the problem (E

2.5.5).
If u and v are two weak anti-periodic solutions of (E 2.5.5), then
Z T
[|u0 (t) − v 0 (t)|2 + (Au(t) − Av(t), u0 (t) − v 0 (t))]dt = 0.
0
RT
However, 0 (Au(t) − Av(t), u0 (t) − v 0 (t))dt = 0, so the uniqueness is obvious.
This completes the proof.
Theorem 2.5.6. Let H be a real separable Hilbert space, A : D(A) ⊆
H → H be a linear densely defined closed self-adjoint operator that only has

Copyright 2006 by Taylor & Francis Group, LLC

 50 Topological Degree Theory and Applications

a point spectrum and G : H → R be a even continuous differentiable function

such that the gradient ∂G is continuous and bounded, i.e., maps bounded sets
of H to bounded sets in H. Suppose that F : R × H → H is a continuous
function and the following conditions are satisfied:

(1) D(A) is compactly embedded into H.

(2) F (t + T, −u) = −F (t, u) for all (t, u) ∈ R × H.

(3) |F (t, u)| ≤ f (t), a. e. t ∈ R, where f (·) ∈ L2 (0, T ; R) is non-negative.

Then the following equation:

(
u0 + Au(t) + ∂Gu(t) + F (t, u) = 0, a.e. t ∈ R,
(E 2.5.7)
u(t + T ) = −u(t), t ∈ R,

has a weak solution.

Proof. Let

Wa = {u : R → H is continuous, u(t + T ) = −u(t)},

Z T
Wa1,2 = {u ∈ Wa : |u0 (t)|2 dt < ∞)}.
0

For each v(·) ∈ Wa , we consider the following equation:

(
u0 + Au(t) + ∂Gv(t) + F (t, v(t)) = 0, a.e. t ∈ R,
(E 2.5.8)
u(t) = −u(t + T ), t ∈ R.

Since A only has a point spectrum, Lemma 2.5.5 implies that the problem
(E 2.5.8) has a unique solution u ∈ Wa1,2 . Now, we define a mapping K :
Wa → Wa as follows: For each v ∈ Wa , Kv is the unique solution of (E 2.5.8).
Next, we prove that K is continuous. Suppose that vn → v0 in Wa . Then,
by Lemma 2.5.4, |vn − v|∞ → 0 as n → ∞. Now, ∂G and F are continuous
functions, so

|∂Gvn (·) − ∂Gv0 (·)|∞ → 0, |F (·, vn (·)) − F (·, v0 (·))|∞ → 0

as n → ∞. Also, we have

(Kvn (t))0 − (Kv0 (t))0 + A(Kvn (t) − Kv0 (t))

+ (∂GKvn (t) − ∂GKv0 (t)) + F (t, vn (t)) − F (t, v0 (t)) (2.5.3)
= 0 for almost all t ∈ R.

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 51

Multiply both sides of (2.5.3) by (Kvn (t) − Kv0 (t))0 and integrate over (0, T ),
and we have
Z T
|(Kvn (t) − Kv0 (t))0 |2 dt
0
Z T
+ hA(Kvn (t)) − Kv0 (t)), (Kvn (t) − Kv0 (t))0 idt
0
Z T
+ h∂GKvn (t) − ∂GKv0 (t), (Kvn (t) − Kv0 (t))0 idt
0
Z T
+ hF (t, vn (t)) − F (t, v0 (t)), (Kvn (t) − Kv0 (t))0 idt = 0.
0

Thus we have
Z T  21
|(Kvn (t) − Kv0 (t))0 |2 dt
√ 0 √
≤ T |∂Gvn (·) − ∂Gv0 (·)|∞ + T |F (·, vn (·)) − F (·, v0 (·))|∞
→ 0 (n → ∞).

Therefore, it follows that Kvn (·) → Kv0 (·) in Wa . For each v ∈ Wa , again,
by (E 2.5.8), we get
Z T Z T Z T
0 2 0
((Kv(t)) ) dt + h∂Gv(t), (Kv(t)) idt + hF (t, v(t)), (Kv(t))0 idt = 0.
0 0 0

This and assumption (3) of Theorem 2.5.6 imply that

Z T  12 Z T  21 Z T  21
|(Kv(t))0 |2 dt ≤ |∂Gv(t)|2 dt + f 2 (t)dt . (2.5.4)
0 0 0

By (2.5.4) and the boundedness of ∂G, we know that K maps bounded sets of
Wa to bounded sets in Wa The compact embedding of D(A) into H implies
that K is a compact mapping. In view of Lemma 2.5.4, we may simply take
|u0 |L2 as the norm of u in Wa .
RT 1
Now, we prove that Kv 6= λv for v ∈ Wa with |v 0 |L2 > ( 0 f 2 (t)dt) 2 and
λ > 1. In fact, if this is not true, then there exist λ0 > 1 and v0 ∈ Wa with
RT 1
|v00 |L2 > ( 0 f 2 (t)dt) 2 such that Kv0 = λ0 v0 , i.e.,

λ0 v00 (t) + λ0 Av0 (t) + ∂Gv0 (t) + F (t, v0 (t)) = 0. (2.5.5)

Multiply both sides of (2.5.5) by v00 (t) and integrate over [0, T ], we obtain
Z T
hλ0 v00 (t) + λ0 Av0 (t) + ∂Gv0 (t) + F (t, v0 (t)), v00 (t)idt = 0.
0

Copyright 2006 by Taylor & Francis Group, LLC

 52 Topological Degree Theory and Applications
RT RT
However, 0
hAv0 (t), v00 (t)idt = 0 and 0
∂hGv0 (t), v00 (t)idt = 0 and so we
have Z T
hλ0 v00 (t) + F (t, v0 (t)), v00 (t)idt = 0.
0

Therefore, it follows that

Z T  12 Z T  12
λ0 |v00 (t)|2 dt ≤ f 2 (t)dt ,
0 0
RT 1
which is a contradiction. Now, if we take r0 > ( 0 f 2 (t)dt) 2 , then, by the
above arguments and the homotopy invariance property of the Leray Schauder
degree, we know that

deg(I − K, B(0, r0 ), 0) = deg(I, B(0, r0 ), 0) = 1,

where B(0, r0 ) is the open ball centered at 0 with radius r0 in Wa . Thus K

has a fixed point in B(0, r0 ), i.e., there exists v ∈ Wa such that Kv = v.
Hence the problem (E 2.5.7) has a solution. This completes the proof.
From Theorem 2.5.6, we have the following:
Corollary 2.5.7. Let H be a real separable Hilbert space, A : D(A) ⊆
H → H be a linear densely defined closed self-adjoint operator that only
has a point spectrum, 0 ∈ / σ(A) and G : H → R be a even continuous
differentiable function such that the gradient ∂G is Lipschitz. Suppose that
F : R×H → H is a continuous function and the conditions (1)-(3) in Theorem
2.5.6 are satisfied. Then the problem (E 2.5.7) has a solution.
Example 2.5.8. Consider the anti-periodic solution problem:
 0 2
u1 (t) = 2u1 (t) − αu2 (t) + 1+u22 (t) sin t,
 t ∈ R,
u02 (t) = −αu1 (t) − 5u2 (t) + 1+u14 (t) sin3 t, t ∈ R, (E 2.5.9)
 1

u1 (t + π) = −u1 (t), u2 (t + π) = −u2 (t), t ∈ R,



where α ∈ R is a constant. Set

2
!
sin t
   
2 −α u1 1+u22 (t)
A= , u= , F (t, u) = .
−α −5 u2 1
1+u41 (t)
sin3 t

Then A is a linear self-adjoint operator on R2 and F (t + π, −u) = −F (t, u)

for all (t, u) ∈ R × R2 . It is obvious that |F (t, u)| ≤ 3 for (t, u) ∈ R × R2 .
Now, the problem (E 2.5.9) is equivalent to the equation:
(
u0 (t) = Au(t) + F (t, u(t)), t ∈ R,
(E 2.5.10)
u(t + π) = −u(t), t ∈ R.

Copyright 2006 by Taylor & Francis Group, LLC

 LERAY SCHAUDER DEGREE THEORY 53

From Theorem 2.5.6, we know that the problem (E 2.5.10) has a solution, so
(E 2.5.9) has a solution.
Remark. The degree theory in this chapter can be established in locally
convex spaces (see [197]) or admissible topological vector spaces (see [170],
[235], [307]).

2.6 Exercises
1. Let f (x, y) : R × R → R be a continuous
Rs function and T : C[a, b] →
C[a, b] be defined by T x(·)(s) = a f (t, x(t))dt for all x(·) ∈ C[a, b] and
s ∈ [a, b]. Show that T is continuous and compact.

2. Let T : C[0, π] → C[0, π] be defined by

2 π
Z
T x(t) = [sin t sin s + c sin 2t sin 2s][2x(s) + x3 (s)]ds
π 0

for all x(·) ∈ C[0, π]. Compute T 0 (x) and the eigenvalues of T 0 (0).

3. Let T : c0 → c0 be defined by T (x1 , x2 , · · · ) = (x21 , x,2 · · · ). Show that

T 0 (x) is compact for all x ∈ c0 , but T is not compact.

4. Let E be a infinite dimensional Banach space. Show that the unit sphere
of E is not compact.

5. Let E be a real Banach space and Ω be an open subset of E. Let

T : Ω → E be a continuous compact mapping and I − T is locally one
to one. Show that I − T is an open mapping.

6. Let E be a real Banach space, T : B(0, R) → E be a continuous com-

pact mapping and L : X → X be a compact linear operator. Sup-
pose that kT x − Lxk < kx − Lxk for all x ∈ ∂B(0, R). Show that
deg(I − T, B(0, R), 0) is odd.

7. Let E be a real Banach space, Ω ⊂ E be an open bounded symmetric

subset with 0 ∈ Ω and T : Ω → E be a continuous compact mapping
such that 0 ∈
/ (I − T )(∂Ω) and (I − T )(−x) 6= t(I − T )(x) on ∂Ω for all
t ≥ 1. Show that deg(I − T, Ω, 0) is odd.

8. Let E be a real Banach space and T : E → E be a continuous compact

mapping such that limkxk→∞ kT xk
kxk = 0. Show (I − T )(E) = E

9. Let E be a real Banach space and A, B ⊂ E be two closed bounded

subsets. Suppose that there exists a continuous compact mapping T :

Copyright 2006 by Taylor & Francis Group, LLC

 54 Topological Degree Theory and Applications

A → E such that I − T : A → B is a homeomorphism. Show that E \ A

and E \ B have the same number of components.
10. Let E be a locally convex space and Ω ⊂ E be an open subset. Suppose
that T : Ω → E is continuous such that T (Ω) is compact in E and
0∈/ (I −T )(∂Ω). Show that there is a topological degree deg(I −T, Ω, 0).
11. Is a Hausdorff continuous mapping continuous?
12. Let E be a real Banach space, Ω ⊂ E be an open bounded symmetric
subset with 0 ∈ Ω and T : Ω → 2E be an upper semicontinuous compact
mapping with closed convex values such that 0 ∈ / (I − T )(∂Ω) and
T (−x) = −T x for all x ∈ ∂Ω. Show that deg(I − T, Ω, 0) is odd.
13. Let G : Rn → R be an even continuous differentiable function and
f : R → Rn be a continuous function such that f (t + T ) = −f (t) for all
t ∈ R. Show that the following equation
(
x0 (t) = ∂Gx(t) + f (t), t ∈ R,
x(t + T ) = −x(t), t ∈ R,

has a solution.
14. Let g ∈ C 1 ([0, a]), h ∈ C 1 ([0, b]) with g(0) = h(0) and f : [0, a] ×
[0, b] × R2 → R be continuous such that |f (x, y, z, 0)| ≤ M (1 + |z|)
and |f (x, y, z, u) − f (x, y, z, v)| ≤ L|u − v|, where a, b, L, M are positive
constants. Show that the following equation

∂u ∂u
 ∂x∂y = f (x, y, u(x, y), ∂x ),
 (x, y) ∈ [0, a] × [0, b],
u(x, 0) = g(x), x ∈ [0, a],

u(0, y) = h(y), y ∈ [0, b],


has a solution u ∈ C 1 ([0, a] × [0, b]).

Copyright 2006 by Taylor & Francis Group, LLC

 Chapter 3
DEGREE THEORY FOR SET
CONTRACTIVE MAPS

The Leray Schauder degree theory is very useful in solving an operator equa-
tion of the type (I − T )x = y, where T is compact. In many applications T
is not compact, so one may ask it is possible to give an analogue of the Leray
Schauder theory in the noncompact case. In 1936, Leray [184] constructed an
example to show that it is impossible to define a degree theory for mappings
with only a continuity condition. So a very natural question which arises is
the following:
For what kind of mappings in infinite dimensional spaces can we establish
a degree theory ?
Browder, Nussbaum, Sadovski, Vath, etc., showed that it is possible to
define a complete analogue of the Leray Schauder theory for condensing type
mappings T .
In this Chapter, we will introduce the degree theory for k-set contraction
mappings and condensing mappings. This chapter consists of three sections.
In Section 3.1, we define measures of non-compactness and present some
properties (see propositions 3.1.5, 3.1.7 and theorems 3.1.14, 3.1.15). Also,
countably condensing maps, etc., are defined here and, in particular, a fixed
point theorem for countably condensing self-mappings is presented in Corol-
lary 3.1.18.
Section 3.2 presents a degree theory for countably condensing mappings and
the theory is based on the use of retractions and the Leray Schauder degree.
Again, various properties and consequences are presented.
In Section 3.3, we use the degree of Section 3.2 to discuss the initial and
anti-periodic ordinary differential equations in Banach spaces.

3.1 Measure of Noncompactness and Set Contractions

Let (X, d) be a metric space and A ⊂ X be a subset. We call diam(A) =
supx,y∈A d(x, y) the diameter of A. If diam(A) < +∞, then we call A

55

Copyright 2006 by Taylor & Francis Group, LLC

 56 Topological Degree Theory and Applications

bounded. For two bounded sets A, B, the Hausdorff metric H is defined

by
H(A, B) = max{sup d(x, B), sup d(y, A)}.
x∈A y∈B

Let B(X) be the collection of all bounded subsets of X.

Proposition 3.1.1. If A ⊂ B, then diam(A) ≤ diam(B) and diam(A) =
diam(A).
Proposition 3.1.2. Let X be a Banach space and A, B ⊂ X. Then we
have the following:

(1) diam(λB) = |λ|diam(B);

(2) diam(x + B) = diam(B);

(3) diam(A + B) ≤ diam(A) + diam(B);

(4) diam(conv(A)) = diam(A).

Proof. (1)-(3) are obvious. To see (4), let x, y ∈ conv(A). There exist
si ∈ (0, 1), xi ∈ X, i = 1, 2, · · · , k, ti ∈ (0, 1), yi ∈ A, i = 1, 2, · · · , m, such
that x = Σki=1 si xi and y = Σmi=1 si yi . Now, we have

kx − yk = kΣki=1 si xi − Σmi=1 si yi k
= Σki=1 Σm k m
j=1 i j i − Σi=1 Σj=1 si tj yj k
s t x
≤ Σki=1 Σm
j=1 si tj kxi − yj k

≤ Σki=1 Σm
j=1 si tj diam(A).

Thus diam(conv(A)) ≤ diam(A), and the converse is obvious. Therefore, the

conclusion is true. This completes the proof.
Proposition 3.1.3. Let (X, d) be a metric space. Then (B(X), H) is a
metric space.
Proof. Obviously, H(A, B) ≥ 0 for any A, B ∈ B(X), H(A, B) = 0 if only
if A = B and H(A, B) = H(B, A).
For A, B, C ∈ B(X), we have

d(x, B) ≤ d(x, z) + d(z, B), d(y, A) ≤ d(y, z) + d(z, A)

for all z ∈ C, x ∈ A and y ∈ B. Thus

d(x, B) ≤ inf d(x, z) + sup d(z, B),

z∈C z∈C

d(y, A) ≤ inf d(y, z) + sup d(z, A)

z∈C z∈C

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 57

and so

H(A, B)
≤ max{sup d(x, C) + sup d(z, B), sup d(y, C) + supz∈C d(z, A)}
x∈A z∈C y∈B
≤ max{sup d(x, C), supz∈C d(z, A)} + max{sup d(z, B), sup d(y, C)}
x∈A z∈C y∈B
= H(A, C) + H(C, B).

Thus (B(X), H) is a metric space. This completes the proof.

Definition 3.1.4. Let (X, d) be a metric space, B be the collection of all
bounded subsets of X and A, B ∈ B. A function α : β → [0, +∞) defined by

α(A) = inf{δ > 0 : A is covered by finitely

many sets each having diameter less than δ}

is called the (Kuratowski) measure of noncompactness. If we replace α(A) by

β(A) = inf{δ > 0 : A is covered by finitely many balls with radius δ},

then we call β(A) as the ball (Hausdorff) measure of noncompactness.

The relation between α and β is given by the following inequality.
Proposition 3.1.5. β(A) ≤ α(A) ≤ 2β(A) for all A ∈ B.
Proof. For any δ > α(A), there exist finitely many sets A1 , A2 , · · · , Ak
such that A ⊂ ∪ki=1 Ai and diam(Ai ) < δ for i = 1, 2, · · · , k. Choose xi ∈ Ai .
Then Bδ (xi ) ⊃ Ai , so we have A ⊂ ∪ki=1 Bδ (xi ) and thus β(A) ≤ δ. By letting
δ → α(A), we get the first inequality.
For the second inequality, if δ > β(A), then there exist finitely many balls
Bδ (y1 ), Bδ (y2 ), · · · , Bδ (ym ) such that A ⊂ ∪m
i=1 Bδ (yi ). It is obvious that

diam(Bδ (yi )) = 2δ,

so we have α(A) ≤ 2δ and, by letting δ → β(A), we deduce the desired

inequality. This completes the proof.
Proposition 3.1.6. Let X be a metric space and B be the collection of
all bounded subsets of X. Let φ be the Kuratowski measure or Hausdorff
measure of noncompactness and A, B ∈ B. Then the following properties
hold:

(1) φ(A) = 0 if and only if A is relatively compact;

(2) φ(A) = φ(A);

(3) If A ⊂ B, then φ(A) ≤ φ(B);

Copyright 2006 by Taylor & Francis Group, LLC

 58 Topological Degree Theory and Applications

(4) φ(A ∪ B) = max{φ(A), φ(B)};

(5) φ(A ∩ B) ≤ min{φ(A), φ(B)};
(6) |φ(A) − φ(B)| ≤ 2H(A, B);
(7) If X is a Banach space, then

φ(λA) = |λ|φ(A), λ ∈ R, φ(A+B) ≤ φ(A)+φ(B), φ(conv(A)) = φ(A).

Proof. In (1)–(5), φ(λA) = |λ|φ(A) for any λ ∈ R and φ(A + B) ≤

φ(A) + φ(B) follow easily from the definition.
We only need to prove (6) and φ(conv(A)) = φ(A). Assume that φ = α.
The proof is the same for φ = β.
For any  > 0, there exists a finite cover {A1 , A2 , · · · , Ak } of A with
diam(Ai ) ≤ α(A) +  for i = 1, 2, · · · , k. Set

η = H(A, B) + , Bi = {y ∈ B : there exists x ∈ Ai , d(x, y) < η}

for i = 1, 2, · · · , k. Since H(A, B) < η, we have B ⊂ ∪ki=1 Bi . Obviously,

diam(Bi ) ≤ 2η + diam(Ai ) < 2H(A, B) + α(A) + 3

for i = 1, 2, · · · and thus α(B) ≤ 2H(A, B) + α(A).

Similarly, we have α(A) ≤ 2H(A, B) + α(B). Therefore, we have

|α(A) − α(B)| ≤ 2H(A, B).

For (7): Obviously, α(A) ≤ α(conv(A)). For any  > 0, there exists a finite
cover {A1 , A2 , · · · , Ak } of A with diam(Ai ) < α(A) +  for i = 1, 2, · · · , k.
We may also assume that Bi is convex since diam(conv(Bi )) = diam(Bi ) for
i = 1, 2, · · · , k. Put

Λ = {(λi , λ2 , · · · , λk ) : λi ≥ 0, i = 1, 2, · · · , k, Σki=1 λi = 1}

and B(λ) = Σki=1 λi Bi for each λ = (λ1 , λ2 , · · · , λk ) ∈ Λ. We have

α(B(λ)) ≤ α(A) +  for all λ ∈ Λ.

Now, we show that ∪λ∈Λ B(λ) is convex. For λ = (λ1 , λ2 , · · · , λk ), µ =

(µ1 , µ2 , · · · , µk ) ∈ Λ and x = Σki=1 λi xi ∈ B(λ), y = Σki=1 µi yi ∈ B(µ), where
xi , yi ∈ Bi for i = 1, 2, · · · , k, we have
tλi (1 − t)µi
tx + (1 − t)y = Σki=1 (tλi + (1 − t)µi )[ xi + yi ].
tλi + (1 − t)µi tλi + (1 − t)µi

Thus it follows that ∪λ∈Λ B(λ) is convex. Therefore, we have

conv(A) ⊂ conv(∪ki=1 Bi ) ⊂ ∪λ∈Λ B(λ).

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 59

Since Λ is compact, there exist finitely many λ1 , λ2 , · · · , λn ∈ Λ such that

∪λ∈Λ B(λ) ⊂ ∪ni=1 B(λi ) + B1 (0).

Therefore, we have

conv(A) ⊂ ∪ni=1 B(λi ) + B1 (0),

which implies that

α(conv(A)) ≤ α(A) + 3.
By letting  → 0+ , we get α(conv(A)) ≤ α(A). Thus α(conv(A)) = α(A)).
This completes the proof.
Proposition 3.1.7. Let X be a infinite dimensional Banach space and
B(0, 1) be the unit ball. Then α(B(0, 1)) = 2.
Proof. It is obvious that diam(B(0, 1)) = 2, so α(B(0, 1)) ≤ 2.
If α(B(0, 1)) < 2, then there exists A1 , A2 , · · · , Ak such that diam(Ai ) < 2
for i = 1, 2 · · · , k and B(0, 1) ⊂ ∪ki=1 Ai . Take a k-dimensional subspace Xk
of X and set Bk (0, 1) = B(0, 1) ∩ Xk , Bi = Ai ∩ Xk for i = 1, 2, · · · , k. Then
we have
Bk (0, 1) ⊂ ∪ki=1 Bi , diam(Bi ) < 2, i = 1, 2, · · · , k,
which is a contradiction to Theorem 1.2.17. Thus α(B(0, 1)) = 2. This
completes the proof.
Proposition 3.1.8. Let X be a separable Banach space and β be the
Hausdorff measure of noncompactness. Then there exists an increasing se-
quence of finite dimensional subspaces (Xn ) with X = ∪∞
n=1 Xn such that, for
any bounded countable subset {xn : 1 ≤ n < ∞},

β({xn }) = lim lim sup d(xm , Xn ).

n→∞ m→∞

Proof. Since X is separable, there exists a countable subset {y1 , y2 , · · · }

of X such that {y1 , y2 , · · · } = X. Put Xn = span{y1 , y2 , · · · , yn } for n =
1, 2, · · · . Then X1 ⊂ X2 ⊂ · · · and X = ∪∞
n=1 Xn . For any  > 0 and integer
n > 0, set rn = lim supm→∞ d(xm , Xn ). One can easily see that r1 ≥ r2 ≥ · · ·
since Xn is increasing. Choose an integer L > 0 such that d(xm , Xn ) < rn + 
for all m ≥ L.
Next, we define

Y = {y ∈ Xn : there exists m ≥ L, d(xm , Xn ) = kxm − yk}.

Then it follows that Y ∪ {x1 , x2 , · · · , xL } is compact. Thus there exist finitely

many {z1 , z2 , · · · , zk } such that

Y ∪ {x1 , x2 , · · · , xL } ⊂ ∪ki=1 B(zi , ).

Copyright 2006 by Taylor & Francis Group, LLC

 60 Topological Degree Theory and Applications

Therefore, we have
{x1 , x2 , · · · } ⊂ {x1 , x2 , · · · , xL } ∪ B(Y, rn + ) ⊂ ∪ki=1 B(zi , rn + 2).
Thus β({xn }) ≤ rn + 2, i.e.,
β({xn }) ≤ inf{rn , n ≥ 1} = lim lim sup d(xm , Xn ).
n→∞ m→∞

On the other hand, for  > 0, put r = β({xi : i ≥ 1}) and there exists
finitely many wi , 1 ≤ i ≤ s, such that {x1 , x2 , · · · } ⊂ ∪si=1 B(wi , r + ). By
the construction of Xn , we know that there exists an integer K > 0 such that
d(wi , Xn ) <  for i = 1, 2, · · · , s and n > K. Therefore, we have
d(xm , Xn ) ≤ inf{kxm − wi k : 1 ≤ i ≤ s} + sup{d(wi , Xn ) : 1 ≤ i ≤ s}
≤ r + 2
for m ≥ 1 and n > K. From this, we get
lim lim sup d(xm , Xn ) ≤ r.
n→∞ m→∞

Thus we have limn→∞ lim supm→∞ d(xm , Xn ) = β({xi : i ≥ 1}). This com-
pletes the proof.
Definition 3.1.9. Let X be a real normed space, T : D → X be a mapping
and α be the measure of noncompactness.
(1) T is called a k-set contraction if α(T B) ≤ kα(B) for all bounded subsets
B ⊂ D, where k > 0 is a constant;
(2) T is said to be condensing if α(T B) < α(B) for all bounded subsets
B ⊂ D with α(B) > 0.

Example 3.1.10. Let X be a real normed space and T : X → X be a

linear bounded operator. Then L is a kLk-set contraction.
Example 3.1.11. Let X be a real normed space and T : D ⊂ X → X be a
Lipschitz mapping with Lipschitz constant l. Then T is an l-set contraction.
Proposition 3.1.12. Let X be a real normed space, B(0, 1) be the unit
ball of X and T : X → B(0, 1) be defined by
(
x
, kxk ≥ 1,
T x = kxk
x, x ∈ B(0, 1).
Then T is an 1-set contraction.
Proof. Let A ⊂ X be a bounded set. It is obvious that T (A) ⊂ conv({0} ∪
A), so we have
α(T (A)) ≤ α(conv({0} ∪ A)) = α({0} ∪ A) = α(A).

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 61

Proposition 3.1.13. Let X be a infinite dimensional Banach space, φ :

[0, 1] → [0, 1] be a strictly decreasing and continuous function, B(0, 1) be the
unit ball of X and T : B(0, 1) → B(0, 1) be defined by
T x = φ(|x|)x for all x ∈ B(0, 1).
Then α(T (B)) < α(B) for B ⊂ B(0, 1) with α(B) > 0.
Proof. Let B ⊂ B(0, 1) be such that α(B) = c > 0. Take r ∈ (0, 2c )
and define the sets B1 = B ∩ B(0, r), B2 = B \ B(0, r). It is obvious that
T (B) = T (B1 ) ∪ T (B2 ), so we have
α(T (B)) = α(T (B1 ) ∪ T (B2 )) ≤ max{α(T (B1 )), α(T (B2 ))}.
Moreover, it follows that
α(T (B1 )) ≤ α(conv({0} ∪ B1 )) = α(B1 ) ≤ diam(B1 ) ≤ 2r < c
and
T (B2 ) ⊂ conv({0} ∪ φ(r)B),
so α(T (B2 )) ≤ α(φ(r)B) < α(B). Thus we have α(T (B)) < α(B). This
completes the proof.
Definition 3.1.14. Let X be a real Banach space, T : D → X be a
mapping and α be the measure of noncompactness.
(1) T is called a countably k-set contraction if α(T B) ≤ kα(B) for all
countably bounded subsets B ⊂ D, where k > 0 is a constant;
(2) T is said to be countably condensing if α(T B) < α(B) for all countably
bounded subsets B ⊂ D with α(B) > 0;
(3) H(t, x) : [0, 1]×D → X is said to be a homotopy of countably condensing
mappings if α(H([0, 1] × B)) < α(B) for all countably bounded subsets
B ⊂ D with α(B) > 0.

One can easily see that a condensing mapping is a countably condensing

mapping.
Theorem 3.1.15. Let E be a Banach space and B ⊂ C([a, b], E) be a
bounded equicontinuous subset. Then α(B(t)) is continuous on [a, b], where
B(t) = {x(t) : x(·) ∈ B}, and
Z b Z b
α({ x(t)dt : x(·) ∈ B}) ≤ α(B(t))dt.
a a

Proof. First, we prove that α({x(t) : x(·) ∈ B}) is a continuous function

on [a, b]. For any  > 0, since B is equicontinuous, there exists γ > 0 such
that kx(t) − x(t0 )k <  for all t, t0 ∈ [a, b] satisfying |t − t0 | < γ. Thus we have
H({x(t) : x(·) ∈ B}, {x(t0 ) : x(·) ∈ B}) ≤ 

Copyright 2006 by Taylor & Francis Group, LLC

 62 Topological Degree Theory and Applications

for t, t0 ∈ [a, b] satisfying |t − t0 | < γ. From this and (6) of Proposition 3.1.6,
we infer that

|α({x(t) : x(·) ∈ B}) − α({x(t0 ) : x(·) ∈ B})| ≤ 2

for t, t0 ∈ [a, b] satisfying |t−t0 | < γ. Thus α({x(t) : x(·) ∈ B}) is a continuous
function on [a, b].
For any division of [a, b]: a = t0 < t1 < · · · < tn = b, where ti = a + i b−an ,
i = 0, 1, · · · , n. For any  > 0, from the equicontinuity of B, there exists
N > 0 such that, if n > N , then

kx(ti ) − x(t)k <  for all x(·) ∈ B, t ∈ [ti−1 , ti ]

for i = 1, 2, · · · , n. Thus we have

b ti
−a
Z Z
b
kΣni=1 x(ti ) − x(t)dtk = kΣni=1 (x(ti ) − x(t))dtk < (b − a)
n a ti−1

for all n > N . Therefore, we have

Z b
b−a
|α({Σni=1 x(ti ) : x(·) ∈ B}) − α({ x(t)dt : x(·) ∈ B})| ≤ 2(b − a),
n a

i.e.,
b
b−a
Z
lim α({Σni=1 x(ti ) : x(·) ∈ B}) = α({ x(t)dt : x(·) ∈ B}).
n→∞ n a

On the other hand, by (6) of Proposition 3.1.6, we have

b−a b−a
α({Σni=1 x(ti ) : x(·) ∈ B}) ≤ Σni=1 α({x(ti ) : x(·) ∈ B}) .
n n
Therefore, it follows that
Z b Z b
α({ x(t)dt : x(·) ∈ B}) ≤ α(B(t))dt.
a a

This completes the proof.

Theorem 3.1.16. Let E be a Banach space and B ⊂ C([a, b], E) be a
bounded equicontinuous subset, where a, b ∈ R. Then

α(B) = max α({x(t) : x(·) ∈ B}).

t∈[a,b]

Proof. First, by Theorem 3.1.15, we know that α({x(t) : x(·) ∈ B}) is a

continuous function on [a, b].

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 63

Next, for α(B) < δ, there exist B1 , B2 , · · · , Bm ⊂ C([a, b), E) such that
diam(Bi ) ≤ δ and B ⊂ ∪m
i=1 Bi . Hence, for each t ∈ [a, b], we have

{x(t) : x(·) ∈ B} ⊂ ∪m
i=1 {x(t) : x(·) ∈ Bi }.

Also, we have

diam({x(t) : x(·) ∈ Bi } = sup kx(t) − y(t)k ≤ diam(Bi ) ≤ δ.

x(·),y(·)∈Bi

Therefore, α({x(t) : x(·) ∈ B}) ≤ δ for all t ∈ [a, b] and thus

max α({x(t) : x(·) ∈ B}) ≤ α(B).

On the other hand, since B is equicontinuous, there exist finitely many

t1 , t2 , · · · , tn ∈ [a, b] such that

{x(t) : x(·) ∈ B} ⊂ ∪ni=1 {x(ti ) : x(·) ∈ B} + B (0) for all t ∈ [a, b].

If δ > maxt∈[a,b] α({x(t) : x(·) ∈ B}), we can find finitely many subsets
A1 , A2 , · · · , As ⊂ E such that

diam(Ai ) ≤ δ, ∪ni=1 {x(ti ) : x(·) ∈ B} ⊂ ∪si=1 Ai .

Obviously, B is the union of finitely many sets {x(·) ∈ B : x(ti ) ∈ Aji , i =

1, 2, · · · , n} and each of these sets has diameter less than δ + 2. Thus we have
α(B) ≤ δ + 2 and we get the desired result. This completes the proof.
Proposition 3.1.17. Let E be a Banach space, Ω ⊂ E be an bounded
subset and T : Ω → E be a countably condensing mapping. Put F = {x ∈
Ω : T x = x}. Then there exists a convex compact subset C such that
(1) F ⊆ C;
(2) if x0 ∈ conv(C ∪ {T x0 }), then x0 ∈ C;

(3) C = conv(T (C ∩ Ω)).

Proof. Put

F = {K : F ⊂ K ⊂ E closed convex, T (K ∩ Ω) ⊆ K and (2) holds for K}.

Then F is nonempty since conv(T Ω) ∈ F. We set C = ∩K∈F K. Obviously,

C satisfies (1),(2), (3), and C is closed convex.
Now, we prove that C is compact. Assume that this is not true. Then
there exists C1 = {x1 , x2 , · · · } ⊂ C without a Cauchy subsequence. Note that
C = conv(T (C ∩ Ω)). Thus there exists a countable subset A1 ∈ C ∩ Ω such
that C1 ⊆ conv(T A1 ). One can easily prove that H1 = conv(T (C1 ∩ Ω)) is

Copyright 2006 by Taylor & Francis Group, LLC

 64 Topological Degree Theory and Applications

separable and H1 ∩ Ω is separable, so there exist countable subsets B1 ⊂ H1

and D1 ⊂ H1 ∩Ω such that B1 = H1 , D1 = H1 ∩Ω. Put C2 = C1 ∪A1 ∪B1 ∪D1 .
One can check that
C1 ⊂ C2 , (3.1.1)

conv(T (C1 ∩ Ω)) ⊂ C2 , (3.1.2)

conv(T (C1 ∩ Ω) ∩ Ω ⊂ C2 ∩ Ω. (3.1.3)

In general, we define inductively a sequence (Cn ) of countable subsets Cn of
C satisfying
Cn ⊂ Cn+1 , (3.1.4)

conv(T (Cn ∩ Ω)) ⊂ Cn+1 , (3.1.5)

conv(T (Cn ∩ Ω) ∩ Ω ⊂ Cn+1 ∩ Ω. (3.1.6)

Finally, we put L = ∪∞
n=1 Cn . Then

L ⊆ convT (L ∩ Ω.

Thus we have
α(L) ≤ α(convT (L ∩ Ω) < α(L ∩ Ω),
which is a contradiction. Therefore, C is compact. This completes the proof.
Proposition 3.1.18. [291] Let E be a Banach space, Ω ⊂ E be a bounded
subset and T : Ω → E be a countably condensing mapping. Set C1 =
conv(T Ω), Cn+1 = conv(T (Cn ∩ Ω)) for n ≥ 1 and C = ∩∞ n=1 Cn . Then
C is convex and compact.
Proof. This is a special case of Lemma 7.2.1.
Corollary 3.1.19. Let C ⊂ E be a nonempty bounded closed convex
subset and let T : C → C be a continuous countably condensing mapping.
Then T has a fixed point in C.
Proof. We first assume that T is k-set countably condensing for some
k ∈ [0, 1). Let C1 = convT C and Ci+1 = convT Ci for i = 1, 2, · · · . By
Proposition 3.1.18, K = ∩∞ i=1 Ci is convex and compact and also T : K → K
is a mapping.
Now, we prove that K is non-empty. Take x0 ∈ C, then T i x0 ∈ Ci for
i ≥ 1. We have α({T i x0 , i ≥ n}) ≤ k n α({T0i , i ≥ 0}) for n ≥ 1. Obviously,

α({T i x0 , i ≥ n}) = α({T i x0 , i ≥ 0}),

so α({T i x0 , i ≥ 0}) = 0. Therefore, (T i x0 )∞

i=1 has a subsequence which
converges to a point in K and hence K is nonempty. Therefore, T has a fixed
point.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 65

Now, assume that T is countably condensing. Take x0 ∈ C and, for any

k ∈ (0, 1), put Sx = kT x + (1 − k)x0 for all x ∈ C. Then S is k-set countably
condensing, so S has a fixed point in C. Let kn → 1. Then there are xn ∈ C
such that
kn T xn + (1 − kn )x0 = xn ,
which implies that α({T xn : n ≥ 1}) = α({xn : n ≥ 1}). Since T is countably
condensing, (xn )∞
n=1 has a subsequence xnj → y ∈ C. By the continuity of T ,
we have T y = y. This completes the proof.

3.2 Degree Theory for Countably Condensing Mappings

In this section, we introduce a degree theory for condensing mappings.

Construction of the degree for a countably condensing mapping.
Let E be a Banach space and Ω ⊂ E be an open bounded subset. Let
T : Ω → E be a continuous and countably condensing mapping and 0 ∈ /
(I − T )(∂Ω). If 0 ∈ / (I − T )(Ω), we define deg(I − T, Ω, 0) = 0. Otherwise,
put F = {x ∈ Ω : T x = x} and let C be the convex compact subset satisfying
Proposition 3.1.17. Now, C is nonempty since F ⊂ C. Obviously, T : C ∩Ω →
C is a mapping. If r : E → C is a retraction, then T r is compact and
r−1 (Ω) is open in E. By assumption, 0 ∈ / (I − T )(∂Ω) and we know that 0 ∈
/
(I −T r)(∂(r−1 (Ω)∩Ω), so the Leray Schauder degree deg(I −T r, r−1 (Ω)∩Ω, 0)
is well defined for each retraction r. Now, we define
deg(I − T, Ω, 0) = deg(I − T r, r−1 (Ω) ∩ Ω, 0), (3.2.1)
where deg(I − T r, r−1 (Ω) ∩ Ω, 0) is the Leray Schauder degree.
To see this definition is reasonable, we show that, if r1 , r2 : X → C are two
retractions, then
deg(I − T r1 , r1−1 (Ω) ∩ Ω, 0) = deg(I − T r2 , r2−1 (Ω) ∩ Ω, 0).
Put r(t, x) = tr1 x + (1 − t)r2 x for all (t, x) ∈ [0, 1] × E. Then r(t, ·) :
E → C is a retraction for each t ∈ [0, 1]. Obviously, x 6= T r(t, x) for
(t, x) ∈ [0, 1]∂(r1−1 (Ω) ∩ r2−1 (Ω) ∩ Ω). Thus, by the homotopy property of
Leray Schauder degree, we have
deg(I − T r1 , r1−1 (Ω) ∩ r2−1 (Ω) ∩ Ω, 0)
= deg(I − T r2 , r1−1 (Ω) ∩ r2−1 (Ω) ∩ Ω, 0).
It is simple to check that

/ r1−1 (Ω) ∩ Ω \ r1−1 (Ω) ∩ r2−1 (Ω) ∩ Ω

0∈

Copyright 2006 by Taylor & Francis Group, LLC

 66 Topological Degree Theory and Applications

and
/ r2−1 (Ω) ∩ Ω \ r1−1 (Ω) ∩ r2−1 (Ω) ∩ Ω.
0∈
Thus the excision property of the Leray Schauder degree implies that

deg(I − T r1 , r1−1 (Ω) ∩ Ω, 0) = deg(I − T r1 , r1−1 (Ω) ∩ r2−1 (Ω) ∩ Ω, 0)

and

deg(I − T r2 , r2−1 (Ω) ∩ Ω, 0) = deg(I − T r2 , r1−1 (Ω) ∩ r2−1 (Ω) ∩ Ω, 0).

Therefore, we have

deg(I − T r1 , r1−1 (Ω) ∩ Ω, 0) = deg(I − T r2 , r1−1 (Ω) ∩ r2−1 (Ω) ∩ Ω, 0).

One may also define a degree by taking the set C as in Proposition 3.1.18.
This degree will coincide with the above one by the excision property of the
Leray Schauder degree.
Theorem 3.2.1. The degree defined by 3.2.1 has the following properties:
(1) (N ormality) deg(I, Ω, 0) = 1 if and only if 0 ∈ Ω;
(2) (Solvability) If deg(I − T, Ω, 0) 6= 0, then T x = x has a solution in Ω;
(3) (Homotopy) Let H(t, x) : [0, 1] × Ω → E be a continuous and countably
condensing mappings, i.e., α([0, 1] × B) < α(B) for all countable subset
B of Ω with α(B) > 0 and H(t, x) 6= x for all (t, x) ∈ [0, 1] × ∂Ω. Then
deg(I − H(t, ·), Ω, 0) doesn’t depend on t ∈ [0, 1];
(4) (Additivity) Let Ω1 , Ω2 be two disjoint open subsets of Ω and 0 ∈
/
(I − T )(Ω − Ω1 ∪ Ω2 ). Then

deg(I − T, Ω, 0) = deg(I − T, Ω1 , 0) + deg(I − T, Ω2 ).

Proof. (1), (2), and (4) follow directly from the definition and properties
of the Leray Schauder degree.
To prove (3), we set

C0 = conv(H([0, 1] × Ω)), Cn = conv(H([0, 1] × (Cn−1 ∩ Ω)))

for n ≥ 1. Then C = ∩∞ n=0 Cn is compact by virtue of Proposition 3.1.18 and

H([0, 1] × C) → C. Let r : E → C be a retraction. Then x 6= H(t, rx) for all
x ∈ ∂(r−1 (Ω∩C)∩Ω). Thus deg(I −H(t, ·)r, r−1 (C ∩Ω)∩Ω, 0) doesn’t depend
on t. Therefore, the conclusion follows from the definition of the degree and
the excision property of the Leray Schauder degree. This completes the proof.
Theorem 3.2.2. Let E be a Banach space, Ω ⊂ E be an open bounded
subset with 0 ∈ Ω and T : Ω → E be a continuous and countably condensing

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 67

mapping. Suppose x 6= λT x for all λ ∈ [0, 1), x ∈ ∂Ω. Then T has a fixed
point in Ω.
Proof. We may assume that T x 6= x for all x ∈ ∂Ω. Put H(t, x) = tT x
for all (t, x) ∈ [0, 1] × Ω. It is easy to see that {H(t, ·)}t∈[0,1] is a homotopy
of countably condensing mappings. By assumption, we have H(t, x) 6= x for
all x ∈ ∂Ω. As a result, deg(I − T, Ω, 0) = deg(I, Ω, 0) = 1. Thus T x = x has
a solution in Ω, which is the desired result. This completes the proof.
Corollary 3.2.3. Let E be a Banach space, Ω ⊂ E be an open bounded
subset with 0 ∈ Ω and T : Ω → E be a continuous and countably condensing
mapping. Suppose kT xk ≤ kxk for all x ∈ ∂Ω. Then T has a fixed point in
Ω.
Proof. We may assume that T x 6= x for all x ∈ ∂Ω. Otherwise, the
conclusion is true. Thus we have T x 6= λx for all x ∈ ∂Ω and λ ≤ 1. By
Theorem 3.2.2, T has a fixed point in Ω.
Theorem 3.2.4. Let E be a Banach space and T : E → E be a continuous
and countably condensing mapping. Then one of the following conclusions
holds:
(1) T has a fixed point in E;
(2) {x : T x = λx for some λ > 1} is unbounded.

Proof. Assume that {x : T x = λx for some λ > 1} is bounded. Take

r > 0 such that {x : T x = λx for some λ > 1} ⊂ B(0, r). If T x = x
for some x ∈ ∂B(0, r), then (1) holds. So we may assume that T x 6= x
for all x ∈ ∂B(0, r). Thus x 6= tT x for all t ∈ [0, 1] and x ∈ ∂B(0, r), so
deg(I − T, B(0, r), 0) = 1 and T has a fixed point in B(0, r). This completes
the proof.
Theorem 3.2.5. Let E be a infinite dimensional Banach space, Ω ⊂ E
be an open bounded subset with 0 ∈ Ω, T : Ω → E be a continuous and
countably condensing mapping and S : ∂Ω → E be a continuous compact
mapping. Suppose tT x + (1 − t)Sx 6= x and kSxk ≥ kxk for all x ∈ ∂Ω and
t ∈ [0, 1]. Then deg(I − T, Ω, 0) = 0.
Proof. First, there exists a continuous compact mapping L : Ω → E such
that Lx = Sx for all x ∈ ∂Ω. Obviously, we have tT x + (1 − t)Lx 6= x for all
(t, x) ∈ [0, 1] × ∂Ω. Thus

deg(I − T, Ω, 0) = deg(I − L, Ω, 0).

By Lemma 2.2.11, we have deg(I − L, Ω, 0) = 0 and so deg(I − T, Ω, 0) = 0.

Corollary. 3.2.6. Let E be a infinite dimensional Banach space, Ω ⊂ E an
open bounded subset with 0 ∈ Ω, x0 ∈ E such that kx0 k > supx∈∂Ω kxk, and

Copyright 2006 by Taylor & Francis Group, LLC

 68 Topological Degree Theory and Applications

let T : Ω → E be a continuous and countably condensing mapping. Suppose

kT xk > kxk and T x 6= λx + (λ − 1)x0 for all x ∈ ∂Ω, and λ > 1. Then
deg(I − T, Ω, 0) = 0. This completes the proof.
Proof. Define a mapping S : ∂Ω → E by Sx = x0 . Then S is continuous,
compact and tT x + (1 − t)Sx 6= x for all x ∈ ∂Ω and t ∈ [0, 1]. Obviously,
kSxk > kxk for all x ∈ ∂Ω. Thus the conclusion follows from Theorem 3.2.5.
Theorem 3.2.7. Let E be a infinite dimensional Banach space, Ωi ⊂ E,
i = 1, 2, be two open bounded subsets such that 0 ∈ Ω1 ⊂ Ω2 , x0 ∈ E be such
that kx0 k > supx∈∂Ω2 kxk and T : Ω2 → E be a continuous and countably
condensing mapping. Suppose that the following conditions are satisfied:
(1) kT xk ≤ kxk for all x ∈ ∂Ω1 ;
(2) kT xk ≥ kx|| and T x 6= λx + (λ − 1)x0 for all x ∈ ∂Ω and λ > 1.
Then T has a fixed point in Ω2 \ Ω1 .
Proof. We may assume that T x 6= x for x ∈ ∂Ω1 ∪∂Ω2 . By assumption (1),
we know x − tT x 6= 0 for t ∈ [0, 1] and x ∈ ∂Ω1 . Thus deg(I − T, Ω1 , 0) = 1.
On the other hand, by assumption (2) and Corollary 3.2.6, we have deg(I −
T, Ω2 , 0) = 0. Thus

deg(I − T, Ω2 \ Ω1 , 0) = deg(I − T, Ω2 , 0) − deg(I − T, Ω1 , 0) = −1

and so T has a fixed point in Ω2 \ Ω1 . This completes the proof.

Remark. One can define a degree theory for the so-called fundamentally
restrictive mapping (see [291]).

3.3 Applications to ODEs in Banach Spaces

In this section, we give some applications to the ordinary differential equa-
tions in Banach spaces by using the degree theory for set contractive mappings.
Theorem 3.3.1. Let E be a Banach space and f (t, x) : [0, 1] × B(x0 , r0 ) →
E be a continuous mapping satisfying

α(f ([0, 1] × B(x0 , r)) ≤ kα(B(x0 , r)) for all r ∈ (0, r0 ),

where k ∈ (0, 1) and r0 > 0 are constants. Then there exists t0 ∈ (0, 1] such
that the initial value problem
(
x0 (t) = f (t, x(t)), t ∈ (0, t0 ),
(E 3.3.1)
x(0) = x0

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 69

has a solution.
Proof. Set
r0
M = sup{kf (t, x)k : (t, x) ∈ [0, 1] × B(x0 , r0 }, t0 = min{1, }.
M
Obviously, (E 3.3.1) is equivalent to the following integral equation:
Z t
x(t) = x0 + f (s, x(s))ds. (E 3.3.2)
0

Put X = C([0, t0 ], E) with the norm kx(·)k = max{kx(t)k : t ∈ [0, t0 ]}. Then
X is a Banach space. We also set K = {x(·) ∈ X : x(t0 ) = x0 , kx(t)−x(t0 )k ≤
r0 }. Then K is a bounded closed convex subset of X.
Now, we define a mapping T : K → K by
Z t
T x(t) = x0 + f (s, x(s))ds for all x(·) ∈ K.
0

It is easy to see that T is continuous. Now, we prove that T is condensing.

In fact, for any subset B of K with α(B) > 0, we have
α(T B) = sup α({T x(t) : x(·) ∈ B})
t∈[0,t0 ]
Z t
= sup α({x0 + f (s, x(s)) : x(·) ∈ B})
t∈[0,t0 ] 0

≤ sup α({x0 + tconv(f (s, x(s)) : s ∈ [0, t0 ])}) : x(·) ∈ B})

t∈[0,t0 ]

≤ sup α({x0 + tconv(f ([0, t0 ] × B))})

t∈[0,t0 ]

≤ t0 sup α(f ([0, t0 ] × B))

t∈[0,t0 ]

≤ t0 kα(B).
Thus T is a condensing mapping, so T has a fixed point in K, i.e., (E 3.3.2)
has a solution. Consequently, (E 3.3.1) has a solution. This completes the
proof.
Theorem 3.3.2. Let H be a real Hilbert space, T > 0 be a constant,
f (t, x) : R × H → H be a continuous mapping satisfying
α(f ([0, T ] × B) ≤ kα(B)
for all bounded subsets B of H, where k ∈ (0, 1) is a constant, and kT < 1.
If f (t + T, −x) = −f (t, x) and kf (t, x)k ≤ M kxk + g(t) for all (t, x) ∈ R × E,
where 0 ≤ M T < 2 is a constant, and g(·) ∈ L2 (0, T ), then the following
problem (
x0 (t) = f (t, x(t)), t ∈ R,
(E 3.3.3)
x(t + T ) = −x(t)

Copyright 2006 by Taylor & Francis Group, LLC

 70 Topological Degree Theory and Applications

has a solution.
Proof. Let Ca = {x(·) : R → H is continuous , x(t + T ) = −x(t), t ∈ R}.
Define kx(·)ka = maxt∈[0,T ] for x(·) ∈ Ca , and it is easy to check that Ca
is a Banach space under this norm. It is simple to check that (E 3.3.3) is
equivalent to the following equation:

1 T 1 t
Z Z
x(t) = − f (s, x(s))dt + f (s, x(s))ds. (E 3.3.4)
2 t 2 0
We define a mapping S : Ca → Ca by

1 T 1 t
Z Z
Sx(t) = − f (s, x(s))dt + f (s, x(s))ds for all x(·) ∈ Ca .
2 t 2 0
For any bounded subset U of Ca , we have, by Theorem 3.1.16, that

α(SU ) = max {α({Sx(t) : x(·) ∈ U })}.

t∈[0,T ]

By the same reasoning as in Theorem 3.3.1, we get

α(SU ) ≤ kT α(U ).

Thus S is a kT -set contraction.

Now, we prove that Sx(·) 6= λx(·) for all λ > 1 and x(·) ∈ Ca with
√ Z T
M T −1 T 1
kx(·)ka > (1 − ) ( g 2 (t)dt) 2 .
2 2 0

In fact, suppose that there exists x(·) ∈ Ca with

√ Z T
M T −1 T 1
kx(·)k > (1 − ) ( g 2 (t)dt) 2
2 2 0

such that Sx(·) = λx(·). Then we have

λx0 (t) = f (t, x(t)). (3.3.1)

Multiply both sides of (3.3.1) by x0 (t) and integrate over [0, T ], we have
Z T Z T
0
λ 2
kx (t)k dt = f (t, x(t))x0 (t)dt
0 0
Z T Z T
0
≤M kx(t)kkx (t)kdt + g(t)kx0 (t)kdt.
0 0

In view of Lemma 2.5.4, we have

Z T
MT T 0
Z Z T Z T
1 1
λ kx0 (t)k2 dt ≤ kx (t)k2 dt + ( g 2 (t)dt) 2 ( kx0 (t)k2 dt) 2 ,
0 2 0 0 0

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 71

i.e.,
T
M T −1 T 2
Z Z
1 1
0 2
( kx (t)k dt) ≤ (1 −
2 ) ( g (t)dt) 2 .
0 2 0

Again, by Lemma 2.5.4, we get

√ Z T
M T −1 T 1
kx(·)ka ≤ (1 − ) ( g 2 (t)dt) 2 ,
2 2 0

which is a contradiction. Thus by Theorem 3.2.4, S has a fixed point in Ca ,

i.e., the problem (E 3.3.3) has a solution. This completes the proof.

3.4 Exercises
1. Let Ai ⊂ C([0, 1]), i = 1, 2, 3, be defined by
A1 = {x(·) : x(0) = 0, x(1) = 1 and 0 ≤ x(t) ≤ 1 for t ∈ [0, 1]};
A2 = {x(·) : 0 ≤ x(t) ≤ 12 , t ∈ [0, 21 ], 1
2 ≤ x(t) ≤ 1, t ∈ [ 12 , 1]} ∩ A1 ;
A3 = x(·) : 0 ≤ x(t) ≤ 32 , t ∈ [0, 21 ], 1
3 ≤ x(t) ≤ 1, t ∈ [ 21 , 1]} ∩ A1 ;
Show β(A1 ) = 12 , i = 1, 2, 3, and α(A1 ) = 1, α(A2 ) = 12 , α(A3 ) = 32 .

2. Let E be a Banach space, C 1 ([a, b], E) be the space of continuously

differentiable functions with the norm

kx(·)k1 = max kx(t)k + max kx0 (t)k

t∈[a,b] t∈[a,b]

and α1 is the Kuratowski measure of noncompactness in C 1 ([a, b], E).

Let B ⊂ C 1 ([a, b], E) be a bounded subset such that B 0 = {x0 (·) : x(·) ∈
B} is equicontinuous. Show that

α1 (B) = max{ max α({x(t) : x(·) ∈ B}), max α({x0 (t) : x(·) ∈ B})}.
t∈[a,b] t∈[a,b]

3. Let X be a complete metric space and Ai ⊂ X be a closed subset for

i = 1, 2, · · · . Suppose that A1 ⊃ A2 ⊃ A3 ⊃ · · · , and limn→∞ φ(An ) =
0, where φ is the Kuratowski measure or the Hausdorff measure of non-
compactness. Show that ∩∞ n=1 An 6= ∅.

4. Let X be a Banach space and T : X → X be a continuous and compact

mapping. Suppose that there exists a linear bounded mapping L : X →
X such that limkxk→∞ kT x−Lxk
kxk = 0, i.e., T is asymptotically linear.
Show that L is compact.

Copyright 2006 by Taylor & Francis Group, LLC

 72 Topological Degree Theory and Applications

5. Let B be the closed unit ball of l2 and T : B → B is defined by

p
T (x1 , x2 , · · · ) = ( 1 − kxk2 , x1 , x2 , · · · )

for all x = (x1 , x2 , · · · ) ∈ B. Show that T is an 1-set contraction.

6. Let X be a real Banach space, B(0, R) ⊂ X be the open ball of origin

with radius R and T : B(0, R) → X be a condensing mapping. Suppose
that one of the following condition is satisfied:
(1) T (∂B(0, R)) ⊂ B(0, R);
(2) kT x − xk2 ≥ kT xk2 − kxk2 for all x ∈ ∂B(0, R).
Show that T has a fixed point in B(0, R).

7. Let X be a real Banach space, Ω ⊂ X be an open bounded subset with

0 ∈ Ω and T : Ω → X be an 1-set contraction such that (I − T )(Ω) is
closed and, if T x = λx for all x ∈ ∂Ω, then λ ≤ 1. Show that T has a
fixed point in Ω.

8. Let X be a Banach space with a Schauder basis {ei : i ∈ N } for each

x = Σni=1 ai (x)ei and Rn x = Σ∞
i=n+1 ai (x)ei for n = 1, 2, · · · . For each
bounded subset B ⊂ X, define µ(B) = lim supn→∞ sup{kRn xk : x ∈
B}. Show that µ has the following properties:
(1) µ(B) = 0 if and only if B is relatively compact;
(2) µ(A ∪ B) = max{µ(A), µ(B)};
(3) µ(conv(A)) = µ(A).

9. Let X be a real Banach space, C ⊂ X be a bounded closed convex subset

and T : C → C be an 1-set contraction. Show that inf x∈C {kx − T xk} =
0.

10. Let X be a real Banach space, Ω ⊂ X an open bounded subset and

T : Ω → X be a k-set contraction with k ∈ (0, 1). Suppose that I −T is a
homeomorphism from Ω → (I −T )(Ω). Show that α([I −(I −T )−1 ]B) ≤
k(1 − k)−1 α(B) for all bounded subset B of Ω.

11. Let X be a real Banach space, Ω ⊂ X be an open bounded subset

with 0 ∈ Ω, T : Ω → X be a countably condensing mapping and
A : D(A) ⊆ E → 2E be a m−accretive mapping such that 0 ∈ A0.
Suppose that kT xk ≤ kxk for all x ∈ ∂Ω. Show that −A + T has a fixed
point in Ω ∩ D(A).

12. Let X be a complete metric space, and Ai ⊂ X a clsoed subset for i =

1, 2, · · · . Suppose that A1 ⊃ A2 ⊃ A3 ⊃ · · · , and limn→∞ α(An ) = 0,
A∞ = ∩∞ n=1 An . Show limn→∞ H(An , A∞ ) = 0.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR SET CONTRACTIVE MAPS 73

13. Show that the equation

 ∂u(s,t)
1 1 1
 ∂s = 2 u(s, 2t) + 2 u(s, 0) + sin πs, t ∈ [0, 2 ],
 s∈R
∂u(s,t)
∂s = 12 u(s, 2t − 1) + 21 u(s, 1) + sin πs, t ∈ [ 12 , 1], s ∈ R

u(s + 1, t) = −u(s, t), t ∈ [0, 1], s ∈ R,


has a solution u(s, t) such that u(s, ·) ∈ C([0, 1]) for all s ∈ R.

Copyright 2006 by Taylor & Francis Group, LLC

Copyright 2006 by Taylor & Francis Group, LLC
 Chapter 4
GENERALIZED DEGREE
THEORY FOR A-PROPER MAPS

To solve an infinite dimensional equation T x = y, a very natural method is

to approximate the original equation by finite dimensional equations, as we
have seen in the Leray Schauder theory. The well-known Galerkin method has
proved to be a very efficient tool in finite dimensional approximation. In the
1960s, Browder and Petryshyn systematically studied the finite dimensional
method for a large class of mappings, which they called A-proper mappings,
and they developed a similar theory to the Brouwer degree.
Our goal of this chapter is to introduce Petryshyn’s generalized degree the-
ory for A-proper mappings. This chapter has five main sections.
In Section 4.1, we define projection schemes and A-Proper mappings. Var-
ious examples (see lemmas 4.1.8 and 4.1.10) are also discussed.
Section 4.2 presents a degree theory for A-proper mappings and various
properties are presented (see Theorem 4.2.3 and Proposition 4.2.5).
A variety of existence results are presented in Section 4.3 for the semilinear
situation S − L where L is a Fredholm map of index zero and S − λN is
A-Proper for each λ ∈ (0, 1], where N is a mapping satisfying some specific
conditions (see theorems 4.3.3 and 4.3.4).
In Section 4.4, we introduce the notion of a Fredholm mapping of index
zero type. We present a degree (coincidence) theory for maps L − N where L
is a Fredholm mapping of index zero type and N is such that either (i) N is
L-A-Proper or (ii) L + λJP − N is A-proper for all λ ∈ (0, λ0 ) with λ0 > 0
or (iii) I − (L + λJP )−1 (N + λJP ) is A-proper for some λ > 0 hold.
The results of Section 4.3 and 4.4 are used in Section 4.5 to present the
existence results for the periodic semilinear ordinary and partial differential
equations.

4.1 A-Proper Mappings

Definition 4.1.1. Let X and Y be real separable Banach spaces.
(1) If there is a sequence of finite dimensional subspaces Xn ⊂ X and a

75

Copyright 2006 by Taylor & Francis Group, LLC

 76 Topological Degree Theory and Applications

sequence {Pn } of linear projections Pn : X → Xn such that Pn x → x

for all x ∈ X, then we say that X has a projection scheme {Xn , Pn }.
(2) If X and Y have projection schemes {Xn , Pn } and {Yn , Qn }, respec-
tively, and dimXn = dimYn for all positive integers n, then we call
Π = {Xn , Pn ; Yn , Qn } an operator projection scheme.

Example 4.1.2. Let X = C([0, 1]). For n ∈ N , partition [0, 1] into n

equal parts and set t0 = 0 < t1 = n1 < t2 = n2 < · · · < tn = 1. Let Xn be
the subspace of all x ∈ X which are linear in every subinterval [ti , ti+1 ] and
P : X → Xn be the projection satisfying Pn x(ti ) = x(ti ) for i = 1, 2, · · · , n.
Then {Xn , Pn } is a projection scheme for X.
Example 4.1.3. Let X be a Banach space with a Schauder basis {ei : i ∈
N }. Then X has a projection scheme {Xn , Pn } defined by

Xn = span{e1 , e2 , · · · , en }, Pn x = Σni=1 αi (x)ei

for x = Σ∞ i=1 αi (x)ei . In the case of a separable Hilbert space, we may choose
an orthonormal basis {ei : i ∈ N }, then the projection Pn x = Σni=1 (x, ei )ei
satisfies Pn∗ = Pn and kPn k = 1.
Example 4.1.4. Let X be a reflexive Banach space with a projection
scheme such that Pn Pm = Pmin{m,n} . Then {Pn∗ X ∗ , Pn∗ } is a projection
scheme for X ∗ .
Proof. Notice that Pn∗ Pn∗ f (x) = f (Pn2 x) = Pn∗ f (x) on X and thus Pn∗ is a
projection. We also have

dimPn∗ X = dimN (I ∗ − P ∗ ) = dimN (I − Pn ) = dimXn .

We claim that X ∗ = ∪∞ ∗ ∗
i=1 Pi X . If not, there is x0 ∈ X \ {0} such that
f (x) = 0 for all f ∈ ∪∞ P ∗ X ∗ since X ∗∗ = J(X), where J(x)(f ) = f (x) for
i=1 i
all f ∈ X ∗ and x ∈ X. Thus, we have f (Pn x) = 0 for all n and f ∈ X ∗ , so
f (x) = 0 for all f ∈ X ∗ and so x = 0, which is a contradiction. Therefore,
X ∗ = ∪∞ ∗ ∗ ∗ ∗ ∗ ∗
i=1 Pi X . We also have Pn X ⊂ Pm X for n < m. Thus, for any
f ∈ X ∗ and  > 0, we may choose g ∈ Pn∗ X ∗ such that kf − gk <  and we
then have

kPm f − f k ≤ kPm (f − g)k + kg − f k ≤ (sup kPn k + 1),

n≥1

∗
which gives Pm f → f as m → ∞.
Example 4.1.5. If both X and Y have Schauder basis, then there exists
an operator projection scheme.
Proof. Let {en } be a Schauder basis of X and {e0n } be a Schauder basis
of Y . Put Xn = span{e1 , e2 , · · · , en } and Yn = span{e01 , e02 , · · · , e0n }. For

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 77

x = Σ∞ ∞ 0 n n 0
i=1 αi ei and y = Σi=1 βi ei , set Pn x = Σi=1 αi ei and Qn y = Σi=1 βi ei .
Then Π = {Xn , Pn ; Yn , Qn } is an operator projection scheme.
Definition 4.1.6. Let X, Y be real Banach spaces and Π = {Xn , Pn ; Yn ,Qn }
be an operator projection scheme. Then a mapping T : D ⊂ X → Y is
called A- proper (respectively, pseudo A-proper) with respect to Π if, for any
bounded xm ∈ D ∩ Xm and Qm T xm → y, there exists a subsequence {xmk }
such that xmk → x ∈ D and T x = y, (respectively, there exists x ∈ D(T ),
such that T x = y). We denote by AΠ (D, Y ) the class of all A-proper mappings
F :D →Y.
Recall that, in a normed space X, the semi-inner products (·, ·)− and (·, ·)+
are defined by

(x, y)− = lim h−1 (kxk − kx − hyk) for all x, y ∈ X,

h→0+

(x, y)+ = lim h−1 (kx + hyk − kxk) for all x, y ∈ X.

h→0+

An operator T : D(T ) ⊆ X → X is called accretive if

(x − y, T x − T y)+ ≥ 0 for all x, y ∈ D(T )

and strongly accretive if

(x − y, T x − T y)+ ≥ ckx − yk2 for some constant c > 0.

For some properties of accretive operators, we refer the reader to [17] and
[60]. One can show that a continuous strongly accretive operator T : X → X
is A-proper, and the proof is left to the reader as an exercise.
Definition 4.1.7. Let X be a separable Banach space with a projection
scheme Π = {Xn , Pn }. Then T : D ⊆ X → X is called a P1 compact mapping
if λI − T is A-proper with respect to Π for all λ ≥ 1.
In the following, let X, Y be separable Banach spaces, S : X → Y be a linear
Fredholm mapping of index zero with N (S) 6= {0} and C : D ⊂ X → Y be
a nonlinear mapping. Consider the semilinear problem Sx − Cx = y for all
x ∈ D(L) ∩ D and y ∈ Y . Since S is Fredholm of index zero (see Chapter V),
there exist closed subspaces X 0 of X and Y 0 of Y with dimY 0 = dimN (S)
such that
X = N (S) ⊕ X 0 , Y = Y 0 ⊕ R(S).
Let P : X → N (S) be a projection, Q : Y → Y 0 be a projection and
M : N (S) → Y 0 be a isomorphism. Put T = M P . Then T is a compact
linear operator. It is known that S + T is also a Fredholm mapping with
ind(S + T ) = ind(S) = 0 and S + T is bijective with (S + T )−1 : Y → X
bounded. Set S1 = S|X 0 ∩D(S) . Then S1 is injective and closed and so S1−1 is
continuous on R(S).

Copyright 2006 by Taylor & Francis Group, LLC

 78 Topological Degree Theory and Applications

Suppose Y has a sequence of finite dimensional subspaces Yn with a se-

quence of projections Qn : Y → Yn such that Qn y → y for all y ∈ Y
as n → ∞. If S : X → Y is a Fredholm mapping of index zero, we set
Xn = (S + T )−1 Yn and then ΓS = {Xn , Yn , Qn } is an admissible scheme for
(X, Y ).
Lemma 4.1.8. Let S : X → Y be a Fredholm mapping of index zero.
Then S is A-proper with respect to ΓS .
Proof. Obviously, Qn S : Xn → Yn is continuous. Let (xnj ) ∈ Xnj be a
bounded sequence such that Qnj Sxnj → y ∈ Y . Since Qn (S +T )x = (S +T )x
for all x ∈ Xn , we have

Qnj (S + T )xnj = (S + T )xnj .

Since T is compact, we may assume that T xn → z by taking a subsequence.

Therefore, we have

xnj = (S + T )−1 [Qnj Sxnj + Qnj T xnj ] → x = (S + T )−1 (y + z)

and T x = z. Thus Sx = y, i.e., S is A-proper with respect to ΠS .

Definition 4.1.9. Let X, Y be real Banach spaces, Π = {Xn , Pn ; Yn ,Qn }
be an operator projection scheme and D ⊂ X. A family of mappings H(t, x) :
[0, 1] × D → Y is called A-proper homotopy with respect to Π if

(1) for any bounded sequence (xm ) in D∩Xm , tm → t0 and Qm H(tm , xm ) →

y, there exists a subsequence (xmk ) of (xm ) such that xmk → x ∈ D
and H(t0 , x) = y;

(2) Qn H(t, x) : [0, 1] × D ∩ Xn → Yn is continuous for n = 1, 2, · · · .

If S : D(S) ⊂ X → Y is a Fredholm mapping, then it is known from [283]

that

l(S) = sup{r ∈ R+ : rγ(B) ≤ γ(S(B)), B ⊂ D(S) is bounded} > 0.

Lemma 4.1.10. Suppose that S is Fredholm of index zero, Ω ⊂ X is

an open bounded subset, ΓS is as above, N : Ω ∩ D(S) → Y is a bounded
continuous mapping and Tλ = S − λN for λ ∈ (0, 1]. Assume that one of the
following conditions holds:

(1) N or S1−1 : R(S) → X is compact;

(2) N is k-ball contractive with k ∈ [0, l(S)) and kQn k = 1;

(3) N (S + T )−1 : (S + T )(Ω ∩ D(S)) → Y is ball condensing and kQn k = 1.

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 79

Then Tλ is A-proper with respect to ΓS for each λ ∈ (0, 1].

Proof. (1) If N is compact, then it is trivial that Sλ N is A-proper with

respect to ΠS . Suppose now that S1−1 is compact. Let (xnj ) ∈ Ω ∩ Xnj be
any sequence such that

gnj = Qnj Sxnj − λQnj N xnj → g ∈ Y.

Since Qn (S + T )x = (S + T )x for all x ∈ Xn , we have

gnj = (S + T )xnj − λQnj N xnj − Qnj T xnj → g.

But T is compact, so we have Sxnj − λQnj N xnj → g. Therefore, it follows

that
(I − Q)Sxnj − λ(I − Q)Qnj N xnj → (I − Q)g ∈ R(S).

Thus we have

(I − P )xnj − λS1−1 (I − Q)Qnj (N xnj ) → S1−1 (I − Q)g.

Since P and S1−1 (I − Q) are compact, it is easy to see that (xnj )∞ j=1 has a
subsequence (x0nj )∞
j=1 converging to x0 . From the continuity of N , it follows
that Sx0nj → g + λN x0 . The closedness of S implies that

Sx0 − λN x0 = g

and thus S − λN is A-proper.

(2) Let xnj ∈ Ω ∩ Xnj be any bounded sequence such that Qnj Sxnj −
λQnj N xnj → g. As in (1), we have Sxnj − λQnj N xnj → g.
On the other hand, γ({Sxnj }) ≤ λkγ({xnj }) ≤ kγ({xnj }). This and the
assumption that k < l(S) imply that γ({xnj }) = 0, so we may assume that
xnj → x0 and, consequently, Sxnj → g + λN x0 . Hence Sx0 − λN x0 = g and
so S − λN is A-proper.
(3) Let (xnj ) be any bounded sequence in Ω ∩ Xnj such that Qnj Sxnj −
λQnj N xnj → g. Then we have

(S + T )xnj − λQnj N xnj − Qnj T xnj → g.

Set ynj = (S + T )xnj . We have

ynj − λQnj N (S + T )−1 (ynj ) − Qnj T (S + T )−1 (ynj ) → g.

By the compactness of T (S + T )−1 and the assumption (3), it follows that

(ynj )∞ 0 0 0
j=1 has a subsequence (ynj ) with ynj → y0 . Thus xnj = (S + T )
−1
y nj →
x0 , and it is simple to check that Sx0 − λN x0 = g. This completes the proof.

Copyright 2006 by Taylor & Francis Group, LLC

 80 Topological Degree Theory and Applications

4.2 Generalized Degree for A-Proper Mappings

Let X, Y be real separable Banach spaces and Π = {Xn , Pn ; Yn , Qn } be an

operator projection scheme. Let Ω ⊂ X be an open bounded subset and L be
a dense subspace of X with ∪∞n=1 Xn ⊂ L. We let ΩL = Ω ∩ L.

Lemma 4.2.1. Let T ∈ AΠ (Ω ∩ L, Y ). Suppose that p ∈

/ T (∂Ω ∩ L). Then
there exists an integer n0 > 0 such that

Qn p ∈
/ Qn T (∂(Ω ∩ Xn )) for all n > n0 .

Proof. Suppose that the assertion of Lemma 4.2.1 is not true. Then
there exists nk → ∞ and xnk ∈ ∂Ω ∩ Xnk such that Qnk p = Qnk T xnk .
Obviously, xnk ∈ ∂Ω ∩ L. Thus we have Qnk T xnk → p as k → ∞ and the A-
properness of T guarantees the existence of a subsequence (xnkl )∞
l=1 such that
xnkl → x0 ∈ ∂Ω ∩ L, and T x0 = p, which is a contradiction. This completes
the proof.
Definition 4.2.2. Let T ∈ AΠ (Ω ∩ L, Y ). Suppose that p ∈ / T (∂Ω ∩ L)
and Qn T is continuous. We define a generalized degree D(T, Ω, p) by

Deg(T, Ω ∩ L, p) = {k ∈ Z ∪ {±∞} : deg(Qnj T, Ω ∩ Xnj , Qnj p) → k

for some nj → ∞},

where Z is the set of all integers.

By Lemma 4.2.1, we know that there exists an integer n0 > 0 such that p ∈/
Qn T (∂Ω ∩ Xn ) and Qn T is continuous, so the Brouwer degree deg(Qn T, Ω ∩
Xn , Qn p is well defined for n > n0 . Thus Deg(T, Ω ∩ L, p) is nonempty and
the definition is well defined.
Theorem 4.2.3. Let T ∈ AΠ (Ω ∩ L, Y ) and p ∈ / T (∂Ω ∩ L). Then the
generalized degree has the following properties:

(1) If Deg(T, Ω ∩ L, p) 6= {0}, then T x = p has a solution in Ω ∩ L;

(2) If Ωi ⊂ Ω for i = 1, 2, Ω1 ∩ Ω2 = ∅ and p 6∈ (Ω \ Ω1 ∪ Ω2 ) ∩ L, then

Deg(T, Ω ∩ L, p) ⊆ Deg(T, Ω1 ∩ L, p) + Deg(T, Ω2 ∩ L, p),

here we use the convention that +∞ + (−∞) = Z ∪ {±∞};

(3) If H(t, x) : [0, 1] × Ω ∩ L → Y is a A-proper homotopy and p ∈ / H(t, x)

for all (t, x) ∈ [0, 1]×∂Ω∩L, then Deg(H(t, ·), Ω∩L, p) does not depend
on t ∈ [0, 1];

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 81

(4) If 0 ∈ Ω, Ω is symmetric about 0, T : Ω ∩ L → Y is an odd A-proper

mapping and 0 ∈ / T (∂Ω ∩ L), then Deg(T, Ω ∩ L, 0) contains no even
numbers.

Proof. (1) If Deg(T, Ω ∩ L, p) 6= {0}, then there exists nk → ∞ such that

deg(Qnk T, Ω ∩ Xnk , Qnk p) 6= 0.

Thus there exists xnk ∈ Ω ∩ L such that Qnk T xnk = Qnk p. By the A-
properness of T , there is a subsequence (xnkj ) with xnkj → x0 ∈ Ω ∩ L and
T x0 = p.
(2) Since p 6∈ (Ω \ Ω1 ∪ Ω2 ) ∩ L, there exists n0 > 0 such that

/ (Ω \ Ω1 ∪ Ω2 ) ∩ Xn
Qn p ∈ for all n > n0 .

Therefore, we have

deg(Qn T, Ω, p) = deg(Qn T, Ω1 , p) + deg(Qn T, Ω2 , p) for all n > n0 .

If k = limj→∞ deg(Qnj T, Ω ∩ Xnj , Qnj p), then we have

k = lim [deg(Qnj T, Ω1 ∩ Xnj , Qnj p) + deg(Qnj T, Ω2 ∩ Xnj , Qnj p).

j→∞

If limj→∞ deg(Qnj T, Ω1 ∩ Xnj , Qnj p) and limj→∞ [deg(Qnj T, Ω2 ∩ Xnj , Qnj p)

are both equal to +∞ or −∞, then k = +∞ or k = −∞ and so the conclusion
holds. If one of them equals to +∞ and the other one is −∞, then, by the
convention, we have k ∈ Z ∪ {±∞}. For the case,

lim sup |deg(Qnj T, Ω1 ∩ Xnj , Qnj p)| < +∞

j→∞

and
lim sup |deg(Qnj T, Ω2 ∩ Xnj , Qnj p)| < +∞
j→∞

and so the conclusion is obvious.

(3) We claim that there exists an integer n0 > 0 such that

Qn p ∈
/ ∪t∈[0,1] H(t, ∂Ω ∩ Xn ) for all n > n0 .

Assume the assertion is false. Then there exist nj → ∞, tj → t0 , xnj ∈

∂Ω ∩ Xnj such that Qnj p = Qnj H(tj , xnj ). Therefore, (xnj ) has a subse-
quence (x0nj ) which converges to x0 ∈ ∂Ω ∩ L and H(t0 , x0 ) = p, which is a
contradiction. Thus the Brouwer degree deg(Qn H(t, ·), Ω ∩ Xn , Qn p) does not
depend on t ∈ [0, 1] for n > n0 , so Deg(H(t, ·), Ω ∩ L, p) does not depend on
t ∈ [0, 1].
(4) Since Ω ∩ Xn is symmetric about 0, by Borsuk’s theorem, we have
deg(Qn T, Ω ∩ Xn , 0) is odd for n sufficiently large. Thus Deg(T, Ω ∩ L, 0)
contains no even numbers. This completes the proof.

Copyright 2006 by Taylor & Francis Group, LLC

 82 Topological Degree Theory and Applications

Corollary 4.2.4. Let X, Y be separable Banach spaces, S : D(S) ⊆ X →

Y be a Fredholm mapping of index zero and (Yn , Qn ) be a projection scheme
for Y . If Ω is an open bounded subset of X with Ω ∩ D(L) 6= ∅ and N :
Ω ∩ D(S) → Y is a nonlinear mapping such that S − N is A-proper with
respect to ΓS , then we have the following:

(1) If p ∈
/ (S − N )(∂Ω ∩ D(S)) and Deg(S − N, Ω ∩ L, p) 6= {0}, then
Sx − N x = p has a solution in Ω ∩ D(S);

(2) If Ω is symmetric about 0, N is odd and 0 ∈/ (S − N )(∂Ω ∩ L), then

Deg(S − N, Ω ∩ D(S), 0) contains no even numbers.

Proof. The proof follows from Theorem 4.2.3.

Proposition 4.2.5. Let X, Y be separable Banach spaces, S : D(S) ⊆
X → Y be a Fredholm mapping of index zero and (Yn , Qn ) be a projection
scheme for Y . If Ω is an open bounded subset of X with Ω ∩ D(L) 6= ∅ and
A : X → Y is a linear continuous compact mapping, then S − A is A-proper
with respect to ΓS . If N (S − A) = {0}, then
(
{0}, 0∈/ Ω,
Deg(S − A, Ω ∩ D(S), 0) =
{1} or {−1}, 0 ∈ Ω.

Proof. It is easy to see that S − A is A-proper with respect to ΠS . Since

N (S − A) = {0}, S − A is injective, it follows that Sx − Ax 6= 0 for all
x ∈ ∂Ω ∩ D(S). Thus, if 0 ∈ / Ω, then Deg(S − A, Ω ∩ D(S), 0) 6= 0 would imply
that Sx−Ax = 0 has a solution in Ω∩D(S), which contradicts N (S−A) = {0}.
On the other hand, if 0 ∈ Ω, then 0 ∈ Ω∩Xn for all n, Qn (S −A) : Xn → Yn
is injective for sufficiently large n and we have

deg(Qn (S − A), Ω ∩ Xn , 0) = 1 or − 1.

This completes the proof.

4.3 Equations with Fredholm Mappings of Index Zero

In this section, all the notations are the same as in previous sections.
Proposition 4.3.1. Let X, Y be separable Banach spaces, S : D(S) ⊆
X → Y be a Fredholm mapping of index zero and (Yn , Qn ) be a projection
scheme for Y . Suppose that Ω ⊂ X is symmetric about 0. Let N : Ω∩D(S) →
Y be a mapping such that

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 83

1 t
(1) H(t, x) = Sx − [ 1+t N (x) − 1+t N (−x)] is A-proper with respect to ΓS
for all t ∈ [0, 1];

(2) kN (x) + N (−x)k ≤ d for all x ∈ ∂Ω ∩ D(S) and some d > 0;

(3) S(x) − N (x) 6= λ[S(−x) − T (−x)] for all x ∈ ∂Ω ∩ D(S) and λ ∈ [0, 1].

Then Sx − N x = 0 has a solution in D(S).

Proof. By (1) and (2), it is simple to check that (H(t, ·))t∈[0,1] is an
A-proper homotopy. Moreover, in view of (3), H(t, x) 6= 0 for all (t, x) ∈
[0, 1]×∂Ω∩D(S). Thus Deg(H(t, ·), Ω∩D(S), 0) does not depend on t ∈ [0, 1].
But H(1, x) = Sx − 12 [N (x) − N (−x)] is an odd mapping, so Deg(H(1, ·), Ω ∩
D(S), 0) contains no even numbers. Thus Deg(L − N, Ω ∩ D(S), 0) contains
no even numbers, so Lx − N x has a solution in Ω ∩ D(S). This completes the
proof.
Proposition 4.3.2. Assume that the following conditions hold:

(a) S −λN : Ω∩D(S) → Y is A-proper with respect to ΓS for each λ ∈ (0, 1]

with N (Ω ∩ D(S)) bounded;

(b) Sx 6= λN x + λp for λ ∈ (0, 1) and x ∈ ∂Ω ∩ D(S);

(c) QN x + Qp 6= 0 for x ∈ S −1 (0) ∩ ∂Ω ∩ D(S), where Q is the projection

of Y onto Y 0 ;

(d) Deg(S − [QN + Qp], Ω ∩ D(S), 0) 6= {0}.

Then there exists x ∈ Ω ∩ D(S) such that Sx − N x = p.

Proof. Since QN is compact and S − tN is A-proper with respect to
ΓS for all t ∈ [0, 1], it follows from N (Ω ∩ D(S)) bounded that H(t, x) =
Sx − (1 − t)[QN x + Qp] − tN x − tp is an A-proper homotopy with respect to
ΓS .
We may assume that Sx − N x ∈ / p for all x ∈ ∂Ω ∩ D(S) (otherwise,
Sx − N x = p has a solution, and we are done). We claim that H(t, x) 6= 0
for all (t, x) ∈ [0, 1] × ∂Ω ∩ D(S). Indeed, if H(t0 , x0 ) = 0 for some (t0 , x0 ) ∈
[0, 1] × ∂Ω ∩ D(S), then t0 ∈ [0, 1).
Case (1) If t0 = 0, then Sx0 = QN x0 + Qp, but Y 0 ∩ R(S) = {0}, so
Sx0 = 0, which contradicts (c).
Case (2) If t0 ∈ (0, 1), then Sx0 = (1 − t0 )[QN x0 + Qp] + t0 N x + t0 p. By
(b), QN x0 + Qp 6= 0, so we have

0 = QSx0 = (1 − t0 )Q[QN x0 + Qp] + t0 QN x0 + t0 Qp = QN x0 + Qp,

which is a contradiction. Thus the claim is true and, as a consequence,

Deg(S − QN − Qp, Ω ∩ D(S), 0) = Deg(S − N − p, Ω ∩ D(S), 0).

Copyright 2006 by Taylor & Francis Group, LLC

 84 Topological Degree Theory and Applications

By the assumption (d), Sx − N x = p has a solution in Ω ∩ D(S). This

completes the proof.
In the following, we assume that there exists a continuous bilinear form [·, ·]
on Y × X such that

y ∈ R(S) if and only if [y, x] = 0 for all x ∈ N (S). (4.3.1)

If {e1 , e2 , · · · , em } is a basis in N (S), then (4.3.1) implies that J : R(Q) =

Y 0 → N (S) given by
Jy = Σmi=1 [y, ei ]ei
−1
is an isomorphism and, if y = Σmi=1 ci ei , then [J ei , ej ] = δij and [J −1 y, ei ] =
ci for 1 ≤ i, j ≤ m.
For subsequent use, let P : X → N (S), Q : Y → Y 0 be the projections and
set A = J −1 P .
Theorem 4.3.3 Let X, Y be separable Banach spaces, S : D(S) ⊆ X → Y
be a Fredholm mapping of index zero and (Yn , Qn ) be a projection scheme for
Y . If Ω is an open bounded subset of X with 0 ∈ Ω and [·, ·] is a continuous
bilinear form on Y ×X such that (4.3.1) holds. Also, assume that the following
conditions hold:
(1) S −λN : Ω∩D(S) → Y is A-proper with respect to ΓS for each λ ∈ (0, 1]
with N (Ω) bounded;
(2) Sx 6= λN x + λp for all x ∈ N (S) ∩ ∂Ω ∩ D(S) and λ ∈ (0, 1);
(3) QN x + Qp 6= 0 for all x ∈ N (S) ∩ ∂Ω;
(4) One of the following conditions holds:
(4a) [QN x + Qp, x] ≥ 0 for all x ∈ N (S) ∩ ∂Ω;
(4b) [QN x + Qp, x] ≤ 0 for all x ∈ N (S) ∩ ∂Ω.
Then Sx − N x = p has a solution.
Proof. Consider the homotopy H : [0, 1] × Ω ∩ D(S) → Y given by

H(t, x) = Sx − (1 − t)(QN x + Qp) − tN x − tp

for all (t, x) ∈ [0, 1] × Ω ∩ D(S). Since QN is compact, it follows from (1) that
H is an A-proper homotopy.
Now, we claim that H(t, x) 6= 0 for all (t, x) ∈ [0, 1] × ∂Ω ∩ D(S). If this is
not true, then there exists (t0 , x0 ) ∈ [0, 1] × ∂Ω ∩ D(S) such that

Sx0 = (1 − t0 )(QN x0 + Qp) + t0 N x0 + t0 p.

If t0 = 0, then Sx0 = QN x0 + Qp, so we have 0 = Sx0 = QN x0 + Qp, which

contradicts (3). If t0 = 1, then Sx0 = N x0 + p, which contradicts (2).

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 85

If t0 ∈ (0, 1), then Sx0 −t0 N x0 −t0 p = (1−t0 )(QN x0 +Qp), so QN x0 +Qp 6=
0. Therefore, we have

0 = QSx0 = QN x0 + Qp,

which contradicts (3). Therefore, the claim is true. Consequently, we have

Deg(S − QN − Qp, Ω ∩ D(S), 0) = Deg(S − N − p, Ω ∩ D(S), 0).

Now, we prove Deg(S − QN − Qp, Ω ∩ D(S), 0) 6= {0}. To achieve this goal,

we consider the homotopy H1 (t, x) : [0, 1] × Ω ∩ D(S) → Y given by

H1 (t, x) = Sx − (1 − t)Ax − t(QN x + Qp)

for all (t, x) ∈ [0, 1]×Ω∩D(S), where A = J −1 P if (4a) holds and A = −J −1 P

if (4b) holds.
Also, we prove that H1 (t, x) 6= 0 for all (t, x) ∈ [0, 1] × ∂Ω ∩ D(S). If this
is not true, then there exists (t0 , x0 ) ∈ [0, 1] × ∂Ω ∩ D(S) such that

Sx0 − (1 − t)Ax0 − t0 (QN x0 + Qp) = 0.

Since N (S − A) = {0}, it follows that t0 6= 0. If t0 = 1, then Sx0 = QN x0 +

Qp = 0, which contradicts (3). Therefore, t0 ∈ (0, 1), so we have Sx0 = 0 =
(1 − t0 ) + Ax0 + t0 (QN x0 + Qp).
Assume that (4a) holds. Since P x = x, we have (1 − t0 )J −1 x0 + t0 (QN x0 +
Qp) = 0 and x0 6= 0. Therefore, we have

(1 − t0 )[J −1 x0 , x0 ] + t0 [QN x0 + Qp, x0 ] = 0.

Thus x0 = 0, which is impossible.

If (4b) holds, then a similar argument shows that x0 = 0, which is a con-
tradiction. Thus we have

Deg(S − QN − Qp, Ω ∩ D(S), 0) = Deg(S − A, Ω ∩ D(S), 0).

But S − A is a linear injection, so Deg(S − A, Ω ∩ D(S), 0) 6= {0}. Therefore,

it follows that Deg(S − N − p, Ω ∩ D(S), 0) 6= 0. Consequently, Sx − N x = p
has a solution. This completes the proof.
Theorem 4.3.4. Let X, Y be separable Banach spaces, S : D(S) ⊆ X → Y
be a Fredholm mapping of index zero, (Yn , Qn ) be a projection scheme for Y ,
Ω be open bounded and symmetric about 0 ∈ Ω and [·, ·] be a continuous
bilinear form on Y × X. If N : Ω ∩ D(S) → Y is a bounded continuous
mapping such that

(1) S − λN : Ω ∩ D(S) → Y is A-proper with respect to ΓS for all λ ∈ (0, 1];

(2) Sx 6= λN x + λp for all x ∈ N (S) ∩ ∂Ω ∩ D(S) and λ ∈ (0, 1);

Copyright 2006 by Taylor & Francis Group, LLC

 86 Topological Degree Theory and Applications

(3) [Q(N x + p), x][Q(N (−x) + p), x] < 0 for all x ∈ N (S) ∩ ∂Ω.
Then Sx − N x = p has a solution in Ω ∩ D(S).
Proof.If Sx − N x = p has a solution on ∂Ω ∩ D(S), we are done, so we
may assume that Sx − N x 6= p for x ∈ ∂Ω ∩ D(S).
Consider the homotopy H : [0, 1] × Ω ∩ D(S) → Y given by

H(t, x) = Sx − (1 − t)[Q(N (x) + p)] − tN x − tp

for all (t, x) ∈ [0, 1] × Ω ∩ D(S). Since QN is compact, it follows from (1)
that H(t, ·) is A-proper for all t ∈ [0, 1] and N is bounded. Therefore, H is
an A-proper homotopy.
Now, we prove that H(t, x) 6= 0 for (t, x) ∈ [0, 1]×∂Ω∩D(S). In fact, assume
the contrary, there exists (t0 , x0 ) ∈ [0, 1] × ∂Ω ∩ D(S) such that H(t0 , x0 ) = 0,
i.e.,
Sx0 − (1 − t)[Q(N (x0 ) + p)] − t0 N (x0 ) − t0 p = 0.
By (2), we know that t0 6= 1. If t0 = 0, then Sx0 = Q(N (x0 ) + p). But
R(L) ∩ Y 0 = {0}, so we have Sx0 = 0 and Q(N (x0 ) + p) = 0. Therefore,
[Q(N (x0 ) + p), x0 ] = 0, which contradicts (3). Thus we must have t0 ∈ (0, 1).
Therefore,

0 = QSx0 = (1 − t0 )Q(N (x0 ) + p) + t0 Q(N (x0 ) + p),

i.e., Q(N (x0 )+p) = 0 which contradicts (3) again. By the homotopy property
of the generalized degree, we get

Deg(S − N − p, Ω ∩ D(S), 0) = Deg(S − Q(N + p), Ω ∩ D(S), 0).

Now, we prove that {0} =6 Deg(S − Q(N + p), Ω ∩ D(S), 0). To reach this
goal, we consider the homotopy H1 (t, x) : [0, 1] × Ω ∩ D(S) → Y given by
1
H1 (t, x) = Sx − [Q(N (x) + p) − t(Q(N (−x) + p)]
1+t

for all (t, x) ∈ [0, 1] × Ω ∩ D(S). Obviously, H1 is an A-proper homotopy. We

claim that H1 (t, x) 6= 0 for all (t, x) ∈ [0, 1] × ∂Ω ∩ D(S). If not, there exists
(t1 , x1 ) ∈ [0, 1] × ∂Ω ∩ D(S) such that H1 (t1 , x1 ) = 0, i.e.,
1
Sx − [Q(N (x1 ) + p) − t1 (Q(N (−x1 ) + p)] = 0.
1 + t1
If t1 = 0, then Sx1 = Q(N (x1 ) + p) = 0, which contradicts (3). So t1 6= 1,
thus Lx0 = 0 and Q(N (x1 ) + p) − t1 (Q(N (−x1 ) + p) = 0, which lead again
to a contradiction to (3). Therefore, we have
1
Deg(S − Q(N + p), Ω ∩ D(S), 0) = Deg(S − [Q(N (·) + p) − Q(N (−·) + p)].
2

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 87

But S − 12 [Q(N (·) + p) − Q(N (−·) + p)] is odd, so

1
/ Deg(S − [Q(N (·) + p) − Q(N (−·) + p)].
0∈
2
Thus we have Deg(S −N −p, Ω∩D(S)) 6= {0} and, consequently, Sx−N x = p
has a solution. This completes the proof.

4.4 Equations with Fredholm Mappings of Index Zero

Type
In this section, we show the existence of solutions of the equations with
Fredholm mappings of index zero type in Banach spaces.
Definition 4.4.1. Let X, Y be two real Banach spaces and L : D(L) ⊆
X → Y be a linear mapping. We say that L is a Fredholm mapping of index
zero type if
(1) Ker(L) = {x ∈ X : Lx = 0} and Im(L) = {Lx : x ∈ D(L)} are closed
in H;
(2) X = Ker(L) ⊕ X1 for some subspace X1 of X and Y = Y1 ⊕ Im(L) for
some subspace Y1 of Y ;
(3) Ker(L) is linearly homeomorphic to Coker(L) = Y /Im(L).

Remark. Obviously, if X is linearly homeomorphic to Y , L = 0 is a

Fredholm mapping of index zero type but not a Fredholm mapping of index
zero. If L is a Fredholm mapping of index zero, then

dim(Ker(L)) = dim(Coker(L)) < +∞,

so Ker(L) is linearly homeomorphic to Coker(L) and thus L is a Fredholm

mapping of index zero type.
Now, assume that L : D(L) ⊂ X → Y is a Fredholm mapping of index zero
type. Then there exist linear projections P : X → X and Q : Y → Y such
that
Im(P ) = Ker(L), Im(Q) = Y1 .
Obviously, the restriction of LP of L to D(L) ∩ Ker(P ) is one to one and
onto Im(L), so its inverse KP : Im(L) → D(L) ∩ Ker(P ) is defined. Let
J : Ker(L) → Y1 be a linear homeomorphism and set KP Q = KP (I − Q)
Proposition 4.4.2. L + λJP : X → Y is a bijective mapping for each
λ 6= 0.

Copyright 2006 by Taylor & Francis Group, LLC

 88 Topological Degree Theory and Applications

Proof. For each λ 6= 0, if Lx + λJP x = 0, then JP x = 0 and Lx = 0,

so x ∈ Ker(L). Thus x = 0. On the other hand, for y = y1 + y2 ∈ Y ,
y1 ∈ Y1 , y2 ∈ Im(L), put x = λ−1 J −1 y1 + KP y2 , then Lx + λJP x = y.
Therefore, L + λJP is bijective.
Proposition 4.4.3. Let X, Y be real separable Banach spaces, (Yn , Qn )
be a projection scheme for Y and L : D(L) ⊂ X → Y be a Fredholm mapping
of zero index type. Then, for each λ 6= 0, there exists a projection scheme
Γλ,L for (X, Y ).
Proof. For each λ 6= 0, put Kλ = L + λJP . By Proposition 4.4.2, Kλ is
bijective. Set Xn = Kλ−1 Yn for n = 1, 2, · · · . Obviously, we have

dim(Xn ) = dim(Yn ), X = ∪∞
n=1 Xn .

Thus Γλ,L = {Xn , Yn , Qn } is a projection scheme for (X, Y ).

Proposition 4.4.4. Let L : D(L) ⊂ X → Y be a Fredholm mapping of
zero index type. Assume that X is reflexive. If G ⊂ X is bounded closed and
convex, then L : G ∩ D(L) → Y is pseudo A-proper with respect to Γλ,L for
each λ 6= 0.
Proof. For any sequence (xnk ) in G ∩ D(L) ∩ Xnk with Qnk Lxnk → y, we
may assume that xnk * x0 ∈ G by taking subsequences. Notice that

Qnk (Lxnk + λJP xnk ) = Lxnk + λJP xnk , JP xnk * JP x0 ,

so we have

xnk = (L + JP )−1 (Qnk (Lxnk + JP xnk ) * (L + JP )−1 (y + JP x0 ) = x0 .

Thus x0 ∈ D(L) and Lx0 = y. Therefore, L is pseudo A-proper with respect

to Γλ,L .
Definition 4.4.5. Let X be a real separable Banach space, Γ0 = (Xn , Pn )
be a projection scheme for X, Y be a real Banach space, L : D(L) ⊂ X → Y
be a Fredholm mapping of zero index type and N : D ⊂ X → Y be a mapping.
(1) If I − P − (J −1 Q + KP Q )N is A-proper with respect to Γ0 , then we say
that N is L-A-proper with respect to Γ0 ;
(2) If I − P − (J −1 Q + KP Q )N is pseudo A-proper with respect to Γ0 , then
we say that N is pseudo L-A-proper with respect to Γ0 ;
(3) A family of mappings H(t, x) : [0, 1] × D → Y is called a homotopy
of L-A-proper mappings with respect to Γ0 if H(t, ·) is a L-A-proper
mapping with respect to Γ0 for all t ∈ [0, 1].

Proposition 4.4.6. Let L : D(L) ⊆ X → Y be a linear mapping with

Ker(L) = {0} and Im(L) = Y , then the following conclusions hold:

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 89

(1) If Γ0 = (Xn , Pn ) is a projection scheme for X, then 0 is L-A-proper

with respect to Γ0 ;
(2) If (Yn , Qn ) is a projection scheme for Y and L−1 is continuous, then
L is A-proper with respect to Γ1,L , where Γ1,L is constructed as in
Proposition 4.4.3.

Proof. (1) We have P = 0, Q = 0 and the identity mapping I : X → X

is obviously A-proper with respect to Γ0 . Thus 0 is L−A-proper with respect
to Γ0 .
(2) Since Ker(L) = {0}, the mapping K in the proof of Proposition 4.4.3 is
just the mapping L. Thus Xn = L−1 Yn . If xnk ∈ Xnk such that Qnk Lxnk →
y, then Lxnk = Qnk Lxnk → y. Therefore, we have xnk → L−1 y. The
conclusion is true.
Proposition 4.4.7. Let L : D(L) ⊂ X → Y be a Fredholm mapping of
zero index type and Γ0 = (Xn , Pn ) be a projection scheme for X. If G ⊂ X is
a bounded closed convex subset, T : G → Y is a weakly continuous mapping
and X is reflexive, then T is pseudo L-A-proper with respect to Γ0 .
Proof. For any subsequence (xnk ) in Xnk such that Pnk (I − P − J −1 QT −
KP Q T )xnk → y, we may assume that xnk * x0 ∈ G by taking a subsequence.
Thus we have

(I − P )xnk * x0 , J −1 QT xnk * J −1 QT x0

and
KP Q T xnk * KP Q T x0
and, consequently, (I − P − J −1 QT − KP Q T )x0 = y. So T is pseudo L−A-
proper with respect to Γ0 .
Proposition 4.4.8. Let X, Y be real separable Banach spaces, (Yn , Qn ) be
a projection scheme for Y . Let L : D(L) ⊂ X → Y be a Fredholm mapping
of zero index type, G ⊂ X be a bounded closed subset and N : G → Y be a
continuous compact mapping. Then L + λJP − N is A-proper with respect
Γλ,L for each λ > 0.
Proof. For any sequence (xnk ) in G ∩ D(L) ∩ Xnk with Qnk (L + λJP −
N )xnk → y, in view of the compactness of N , we may assume that N xnk →
y0 ∈ Y by taking a subsequence. Notice that

Qnk (Lxnk + λJP xnk ) = Lxnk + λJP xnk ,

so we have

xnk = (L + JP )−1 [Qnk (L + λJP − N )xnk + Qnk N xnk ]

→ (L + λJP )−1 (y + y0 ) = x0 .

Copyright 2006 by Taylor & Francis Group, LLC

 90 Topological Degree Theory and Applications

Thus x0 ∈ D(L), N x0 = y0 and (L+λJP −N )x0 = y. Therefore, L+λJP −N

is A-proper with respect to Γλ,L .
In the following, suppose that X, Y are real separable Banach spaces, L :
D(L) ⊆ X → Y is a Fredholm mapping of index zero type with D(L) dense
in X and N : Ω ⊂ X → Y is a nonlinear mapping. We consider the semilinear
operator equation Lx − N x = 0 and we apply the generalized degree theory
in Section 4.2 to study such an equation.
Lemma 4.4.9. Let L : D(L) ⊆ X → Y be a Fredholm mapping of index
zero type, Ω ⊂ X an open bounded subset and N : Ω → Y be a mapping. If
0∈ / [I − P − (J −1 Q + KP Q )N ](∂Ω).
/ (L − N )(∂Ω ∩ D(L)), then 0 ∈
Proof. Suppose the contrary. Then there exists x0 ∈ ∂Ω such that 0 ∈
x0 − P x0 − (J −1 Q + KP Q )N x0 . Since J −1 QT x0 ∈ Ker(L) = Im(P ), x0 −
P x0 ∈ Ker(P ) and KpQ T x0 ∈ D(L) ∩ Ker(P ), we must have
J −1 QN x0 = 0, x0 − P x0 − KP Q N x0 = 0.
Therefore, we have
QN x0 = 0, x0 − P x0 − KP N x0 = 0, i.e., Lx0 − N x0 = 0,
which contradicts 0 ∈
/ (L − N )(∂Ω ∩ D(L)). This completes the proof.
Now, let L : D(L) ⊆ X → Y be a Fredholm mapping of index zero type,
Γ0 = (Xn , Pn ) be a projection scheme for X, Ω ⊂ X be an open bounded
subset and N : Ω → Y be a L-A-proper mapping respect to Γ0 . Suppose that
0∈/ (L − T )(∂Ω ∩ D(L)). By Lemma 4.4.9, we have
/ [I − P − (J −1 Q + KP Q )N ](∂Ω).
0∈
Since I − P − (J −1 Q + KP Q )N is an A-proper mapping with respect to Γ0 ,
the generalized degree deg(I − P − (J −1 Q + KP Q )N, Ω, 0) is well defined and
we define
degΓ0 ,J (L − N, Ω, 0) = deg(I − P − (J −1 Q + KP Q )N, Ω, 0), (4.4.1)
which is called the generalized coincidence degree of L and N on Ω.
Theorem 4.4.10. The generalized coincidence degree of L and N defined
by (4.4.1) on Ω has the following properties:
(1) If Ω1 and Ω2 are disjoint open subsets of Ω such that 0 ∈
/ (L−N )(D(L)∩
Ω \ (Ω1 ∪ Ω2 ), then
degΓ0 ,J (L − N, Ω, 0) ⊆ degΓ0 ,J (L − N, Ω1 ) + degΓ0 ,J (L − N, Ω2 , 0);

(2) If H(t, x) : [0, 1] × Ω → Y is a homotopy of L-A-proper mappings with

respect to Γ0 and 0 6= Lx − H(t, x) for all (t, x) ∈ [0, 1] × ∂Ω ∩ D(L),
then degΓ0 ,J (L − H(t, ·), Ω, 0) doesn’t depend on t ∈ [0, 1];

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 91

(3) If degΓ0 (L − N, Ω, 0) 6= {0}, then 0 ∈ (L − N )(D(L) ∩ Ω);

(4) If L : D(L) ⊆ X → Y is a linear mapping such that L−1 : Y → D(L) is

continuous, then degΓ0 ,J (L, Ω, 0) = {1} if 0 ∈ Ω;

(5) If Ω is a symmetric neighborhood of 0 and N : Ω → Y is an odd L-

A-proper mapping respect to Γ0 with 0 ∈ / (L − N )(∂Ω ∩ D(L)), then
degΓ0 ,J (L − N, Ω, 0) does not contain even numbers.

Proof. (1)-(3) follows directly from the definition and the properties of
generalized degree.
(4) Since Ker(L) = {0}, P = 0 and Q = 0, the zero mapping is L−A-proper
with respect to Γ0 . Thus degΓ0 ,J (L, Ω, 0) = deg(I, Ω, 0) = {1}.
(5) Since N is odd, the mapping I − P − (J −1 Q + KP Q )N is odd and
thus deg(I − P − (J −1 Q + KP Q )N, Ω, 0) doesn’t contain even numbers. The
conclusion follows by definition.
Corollary 4.4.11. Let L : D(L) ⊆ X → Y be a linear mapping such that
L−1 : Y → D(L) is continuous, Ω ⊂ X be an open bounded subset with 0 ∈ Ω
and N : Ω → Y be a mapping such that {L − tN }t∈[0,1] is a homotopy of L-A-
proper mappings respect to Γ0 . If Lx ∈
/ tN x for all (t, x) ∈ [0, 1] × ∂Ω ∩ D(L),
then deg(L − N, Ω, 0) = 1.
In the following, let L : D(L) ⊂ X → Y be a densely defined Fredholm
mapping of zero index type. We assume that Γ0 = (Yn , Qn ) is a projection
scheme for Y , Γλ,L is the same as in Proposition 4.4.3 and L + λJP − N is
A-proper with respect to Γλ,L for λ ∈ (0, λ0 ), where λ0 > 0 is a constant.
Suppose that 0 ∈/ (L − N )(D(L) ∩ ∂Ω). Then there exists λ1 < λ0 such that

0∈
/ (L + λJP − N )(D(L) ∩ ∂Ω) for all λ ∈ (0, λ1 ).

We define a generalized degree by

deg(L − N, Ω, 0) = ∩0<λ<λ1 ∪0<≤λ deg(L + JP − N, Ω, 0), (4.4.2)

where deg(L+JP −N, Ω, 0) is the generalized degree for A-proper mappings.

Notice that, if 0 ∈
/ (L + λJP − N )(D(L) ∩ ∂Ω) for all λ ∈ (0, λ2 ), then it is
easy to check that

∩0<λ<λ1 ∪0<≤λ deg(L + JP − N, Ω, 0)

= ∩0<λ<λ2 ∪0<≤λ deg(L + JP − N, Ω, 0).

Thus (4.4.2) is well defined.

Theorem 4.4.12. The generalized degree defined by (4.4.2) has the fol-
lowing properties:

Copyright 2006 by Taylor & Francis Group, LLC

 92 Topological Degree Theory and Applications

(1) If Ω1 and Ω2 are two open subsets of Ω such that Ω1 ∩ Ω2 = ∅ and

0∈ / (L − N )(D(L) ∩ Ω \ (Ω1 ∪ Ω2 )), then

deg(L − N, Ω, 0) ⊆ deg(L − N, Ω1 ) + deg(L − N, Ω2 , 0);

(2) If H(t, x) : [0, 1] × Ω → Y satisfies 0 ∈

/ ∪t∈[0,1] (L − H(t, ·)(D(L) ∩ ∂Ω)
and {L + λJP − H(t, ·)}t∈[0,1] is a homotopy of A-proper mappings with
respect to Γλ,L for each λ ∈ (0, λ0 ), where λ0 > 0 is a constant, then
deg(L − H(t, ·), Ω, 0) does not depend on t ∈ [0, 1];
(3) If degΓ0 (L − N, Ω, 0) 6= {0}, then 0 ∈ (L − N )(D(L) ∩ Ω);
(4) If Ω is a symmetric neighborhood of 0 and N : Ω → Y is an odd mapping
such that L + λJP − N is A-proper with respect to Γλ,L for all λ > 0,
and 0 ∈ / (L − N )(∂Ω ∩ D(L)), then deg(L − N, Ω, 0) does not contain
even numbers;
(5) deg(L, Ω, 0) ⊆ {±1} if 0 ∈ Ω.

Proof. (1) By assumption, there exists λ0 > 0 such that

0∈
/ (L + λJP − N )(D(L) ∩ Ω \ (Ω1 ∪ Ω2 ))

for all λ ∈ (0, λ0 ). If m ∈ deg(L − N, Ω, 0), then there exist λj → 0+ , λj < λ0 ,

j = 1, 2, · · · , such that m ∈ deg(L + λj JP − N, Ω, 0). By Theorem 4.2.3, we
have

deg(L + λj JP − N, Ω, 0)
⊆ deg(L + λj JP − N, Ω1 , 0) + deg(L + λj JP − N, Ω2 , 0)

for j = 1, 2, · · · . By (4.4.2), (1) is true.

(2) Since 0 ∈ / ∪t∈[0,1] (L − H(t, ·)(D(L) ∩ ∂Ω), there exists λ1 > 0 such that

0∈
/ ∪t∈[0,1] (L + λJP − H(t, ·))(∂Ω ∩ D(L)) for all λ ∈ (0, λ1 ).

By Theorem 4.2.3, deg(L + λJP − H(t, ·), Ω, 0) does not depend on t ∈ [0, 1]
for λ ∈ (0, min{λ0 , λ1 }). So the conclusion of (2) follows from (4.4.2).
(3) If degΓ0 (L−N, Ω, 0) 6= {0}, then there exists 0 6= m ∈ degΓ0 (L−N, Ω, 0),
so there exists λj → 0+ such that m ∈ deg(L + λj JP − N, Ω, 0). Therefore,
(L + λj JP − N )x has a solution in Ω ∩ D(L) for j = 1, 2, · · · . By letting
j → ∞, we get 0 ∈ (L − N )(D(L) ∩ Ω).
(4) The proof is left to the reader.
(5) Now, L+λJP is A-proper with respect to Γλ,L and 0 ∈ / (L+λJP )(∂Ω∩
D(L)) for all λ > 0. Since L + λJP ) is bijective, deg(L + λJP, Ω, 0) ⊆ {±1}
for all λ > 0, so we have

deg(L − N, Ω, 0) ⊆ {±1}.

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 93

This completes the proof.

Theorem 4.4.13. Let X, Y be real separable Banach spaces and (Yn , Qn )
be a projection scheme for Y . Let L : D(L) ⊂ X → Y be a Fredholm
mapping of zero index type, 0 ∈ Ω ⊂ X be a bounded subset and N : Ω → Y
be a continuous compact mapping. Suppose that the following conditions are
satisfied:
(1) 0 ∈
/ (L − N )(∂Ω ∩ D(L));
/ QN (∂Ω ∩ D(L)).
(2) 0 ∈
Then deg(L − N, Ω, 0) = deg(L − QN, Ω, 0).
Proof. For each λ ∈ (0, λ0 ), a similar proof to Proposition 4.4.3 shows that
{L + λJP − tN − (1 − t)QN }t∈[0,1] is a homotopy of A-proper mappings with
respect to Γλ,L .
Now, we claim that
0∈
/ ∪t∈[0,1] (L − tN − (1 − t)QN )(D(L) ∩ ∂Ω).
If this is not true, then there exist tj ∈ [0, 1] with tj → t0 and xj ∈ ∂Ω ∩ D(L)
such that Lxj − tj N xj − (1 − tj )QN xj → 0.
Case (1) If t0 = 1, then Lxj − N xj → 0, which contradicts the assumption
(1).
Case (2) If t0 6= 1, then QLxj − QN xj → 0 and thus we have QN xj → 0
and xj ∈ D(L), which contradicts the assumption (2).
By (2) of Theorem 4.4.12, we get deg(L − N, Ω, 0) = deg(L − QN, Ω, 0).
Finally, let L : D(L) ⊆ X → Y be a Fredholm mapping of index zero type,
Γ0 = (Xn , Pn ) be a projection scheme for X and Ω ⊂ X an open bounded
subset. Let N : Ω → Y be a mapping such that I − (L + λJP )−1 (N + λJP )
is A-proper with respect Γ0 for some λ > 0. One can easily see that
0 ∈ Lx − N x if and only if 0 ∈ (I − (L + λJP )−1 (N + λJP ))x.
Assume that 0 ∈
/ (L − N )(∂Ω ∩ D(L)). Then we have
/ (I − (L + λJP )−1 (N + λJP ))(∂Ω)
0∈ for all λ > 0
and we define a generalized degree by
degΓ0 (L − N, Ω, 0) = ∪0<λ deg(I − (L + λJP )−1 (N + λJP ), Ω, 0), (4.4.3)
−1
where deg(I − (L + λJP ) (N + λJP ), Ω, 0) is the generalized degree for A-
proper mappings if I − (L + λJP )−1 (N + λJP ) is A-proper. Otherwise, we
have
deg(I − (L + λJP )−1 (N + λJP ), Ω, 0) = ∅.

Theorem 4.4.14. The generalized degree defined by (4.4.3) has the fol-
lowing properties:

Copyright 2006 by Taylor & Francis Group, LLC

 94 Topological Degree Theory and Applications

(1) If Ω1 and Ω2 are disjoint open subsets of Ω such that 0 ∈

/ (L−N )(D(L)∩
Ω \ (Ω1 ∪ Ω2 ), then

deg(L − N, Ω, 0) ⊆ deg(L − N, Ω1 ) + deg(L − N, Ω2 , 0);

(2) If H(t, x) : [0, 1] × Ω → Y satisfies 0 ∈

/ ∪t∈[0,1] (L − H(t, ·)(D(L) ∩ ∂Ω)
and {I − (L + λJP )−1 (H(t, ·) + λJP )}t∈[0,1] is a homotopy of A-proper
mappings with respect to Γλ,L for all λ > 0, then deg(L − H(t, ·), Ω, 0)
does not depend on t ∈ [0, 1];

(3) If deg(L − N, Ω, 0) 6= {0}, then 0 ∈ (L − N )(D(L) ∩ Ω);

(4) If Ω is a symmetric neighborhood of 0 and N : Ω → Y is an odd mapping

such that I − (L + λJP )−1 (N + λJP ) is A-proper with respect to Γ0
for some λ > 0 and 0 ∈ / (L − N )(∂Ω ∩ D(L)), then deg(L − N, Ω, 0) does
not contain even numbers.

Proof. The proof is standard. We prove (2) and skip the others. Since
0∈
/ ∪t∈[0,1] (L − H(t, ·)(D(L) ∩ ∂Ω), it follows that

/ ∪t∈[0,1] (I − (L + λJP )−1 (H(t, ·) + λJP )(∂Ω)

0∈ for all λ > 0.

By Theorem 4.2.3, we know that

deg(I − (L + λJP )−1 (H(t, ·) + λJP ), Ω, 0)

does not depend on t ∈ [0, 1] for each λ > 0. Thus the conclusion of (2) follows
from (4.4.3). This completes the proof.
Theorem 4.4.15. Suppose that (L + λJP )−1 : Y → X is a continuous
compact mapping for each λ > 0, Ω ⊂ X is an open bounded subset with
0 ∈ Ω and N : Ω → Y is a continuous bounded mapping such that Lx 6= N x
and QN x 6= ηJP x for all x ∈ ∂Ω∩D(L) and η > 0, where P, Q are projections
as in the beginning of this section. Then deg(L − N, Ω, 0) = {1}.
Proof. Since (L + λJP )−1 : Y → X is continuous and compact for each
λ > 0, it follows that {I − (L + λJP )−1 t(N + λJP )}t∈[0,1] is a homotopy of
A-proper mappings.
Now, we claim that

x 6= (L + λJP )−1 t(N + λJP )x

for all (t, x) ∈ [0, 1]×(∂Ω∩D(L)) and λ > 0. If this is not true, then there exist
λ0 > 0 and (t0 , x0 ) ∈ [0, 1)×∂Ω such that x0 = (L+λ0 JP )−1 t0 (N x0 +λJP x0 ).
Thus we have x0 ∈ D(L) and

Lx0 + λ0 JP x0 = t0 (N x0 + λ0 JP x0 ).

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 95

Obviously, t0 6= 1 and so (1 − t0 )λ0 JP x0 = t0 QN x0 , which contradicts our

assumption. Consequently, the A-proper degree

deg(I − (L + λJP )−1 (N + λJP ), Ω, 0) = deg(I, Ω, 0) = {1}.

Therefore, from (4.4.3), we get

deg(L − T, Ω, 0) = {1}.

This completes the proof.

Corollary 4.4.16. Suppose that H is a separable Hilbert space, (L +
λJP )−1 : H → X is a continuous compact mapping for each λ > 0 and
Ω ⊂ X is an open bounded subset 0 ∈ Ω, N : Ω → H is a continuous
bounded mapping such that Lx 6= N x for all x ∈ ∂Ω ∩ D(L), QN x 6= 0
for x ∈ ∂Ω ∩ D(L) ∩ Ker(P ) and (QN x, JP x) < 0 for all x ∈ ∂Ω ∩ D(L) ∩
(Ker(P ))c , where P, Q are projections as in the begining of this section. Then
deg(L − N, Ω, 0) = {1}.
Proof. From our assumption, we have QN x 6= ηJP x for all x ∈ ∂Ω ∩ D(L)
and η > 0. Thus the conclusion follows from Theorem 4.4.15.

4.5 Applications of the Generalized Degree

In this section, we apply the results in Sections 4.3 and 4.4 to the periodic
semilinear ordinary and partial differential equations.
First, consider the periodic ordinary differential equations:
(
x00 (t) = f (t, x(t), x0 (t), x00 (t)) − g(t), t ∈ [0, T ],
(E 4.5.1)
x(0) = x(T ), x0 (0) = x0 (T )

where f : [0, T ] × R3 → R is a continuous function.

Lemma 4.5.1. Let f : [0, T ] × R3 → R be a continuous function satisfying
the following:
(a) there exist constants M > 0 and c, d ∈ R with c ≤ gm ≤ gM ≤ d, and
x ≤ M implies that d < f (t, x, 0, r) for all t, ∈ [0, T ] and r ∈ R3 , while
x ≤ −M implies that f (t, x, 0, r) < c for all t ∈ [0, T ] and r ∈ R3 , where
gm = min{g(t) : t ∈ [0, T ]} and gM = max{g(t) : t ∈ [0, T ]}.
If x(t) is a C 2 solution of (E 4.5.1) and |x(t)| does not achieve its maximum
at t = 0 or t = T , then

|x(t)| ≤ M for all t ∈ [0, T ].

Copyright 2006 by Taylor & Francis Group, LLC

 96 Topological Degree Theory and Applications

Proof. Assume that |x(t)| achieves its maximum at t0 ∈ (0, T ). We claim

that |x(t0 )| ≤ M . If not, then |x(t0 )| > M .
Case (1) If x(t0 ) > M , then x0 (t0 ) = 0 and x00 (t0 ) ≤ 0. Therefore, we have

f (t0 , x(t0 ), 0, x00 (t0 )) − g(t0 ) ≤ 0.

By (a), we have d − g(t0 ) < 0, so g(t0 ) < d, which contradicts gM ≤ d.

Case (2) If x(t0 ) < −M , then x0 (t0 ) = 0 and x00 (t0 ) ≥ 0. Therefore, we
have
f (t0 , x(t0 ), 0, x00 (t0 )) − g(t0 ) ≥ 0.
By (a), we have b − g(t0 ) > 0, so c > g(t0 ), which contradicts gm > c.
Therefore, from the above arguments, we know that |x(t0 )| ≤ M and
Lemma 4.5.1 is proved. This completes the proof.
Lemma 4.5.2. Suppose that the condition (a) of Lemma 4.5.1 holds. Then
any solution x(·) of the problem (E 4.5.1) satisfies

|x(t)| ≤ M for all t ∈ [0, T ].

Proof. Let x(·) be a solution of (E 4.5.1). Assume that |x(t)| achieves its
maximum at t = 0. Then x0 (0) = 0. Otherwise, |x(t)| can not achieve its
maximum at t = 0. Therefore, |x(0)| = |x(T )| ≤ M .
Lemma 4.5.3. Suppose the following conditions hold:

(1) There exists a constant M > 0 such that, for each solution x(·) of (E
4.4.1), |x(t)| ≤ M for t ∈ [0, T ];

(2) There exist constants A, C > 0 and B ∈ [0, 1] such that

|f (t, x, r, q)| ≤ Ar2 + B|q| + C

for all (t, x) ∈ [0, T ] × [−M, M ] and r, q ∈ R.

Then there are constants M1 , M2 > 0 depending only on M, A, B, C and gM

such that
|x0 (t)| ≤ M1 , |x00 (t)| ≤ M2 for all t ∈ [0, T ]
for each solution x(·) of (E 4.5.1).
Proof. Since x0 (t) vanishes at least once in [0, T ], each point t ∈ [0, T ] for
which x0 (t) 6= 0 belongs to an interval [µ, γ] such that x0 (t) maintains a fixed
sign on [µ, γ] and x(µ) or x(γ) is 0. Without loss of generality, we may assume
that x(µ) = 0 and x0 (t) ≥ 0 for t ∈ [µ, γ]. Therefore, it follows from (2) that

|x00 (t)| ≤ Ax0 (t)2 + B|x00 (t)| + C + D, (4.5.1)

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 97

where D = max{|gm |, |gM |}. Multiply (4.5.1) by x0 (t) and rearrange the
terms, we obtain

(1 − B)|x00 (t)x0 (t)| ≤ (Ax0 (t)2 + C + D)x0 (t).

A C+D
Set α = 1−B and β = 1−B , we get

2αx00 (t)x0 (t)

≤ 2αx0 (t).
αx0 (t)2 + β

Integrate the last inequality over [µ, t] and use |x(t)| ≤ M and x0 (µ) = 0, we
obtain
αx0 (t)2 + β
ln( ) ≤ 2αM.
β
Therefore, we have

β 1
|x0 (t)| ≤ [ (e2αM − 1)] 2 = M1 .
α
This and the condition (2) implies that |x00 (t)| ≤ M2 for some M2 depending
on M, A, B, and C. This completes the proof.
Combine Lemma 4.5.1 and Lemma 4.5.2 with Lemma 4.5.3, we get the
following:
Proposition 4.5.4. Assume that the condition of Lemma 4.4.1 holds.
If there are continuous functions A(t, x), C(t, x) > 0 which are bounded on
compact subsets of [0, T ] × R and a constant B ∈ [0, 1] such that

|f (t, x, r, q)| ≤ A(t, x)r2 + B|q| + C(t, x)

for all (t, x) ∈ [0, T ] × [−M, M ] and r, q ∈ R, then there are constants M1 and
M2 such that, for any solution of (E 4.5.1),

|x(t)| ≤ M, |x0 (t)| ≤ M1 , |x00 (t)| ≤ M2 for all t ∈ [0, T ].

Now, we consider a family of the periodic problems:

(
x00 (t) = λf (t, x(t), x0 (t), x00 (t)) − λg(t), t ∈ [0, T ],
(E 4.5.2)
x(0) = x(T ), x0 (0) = x0 (T ),

where λ ∈ [0, 1].

Lemma 4.5.5. Suppose that the following conditions hold:

(1) Let M > 0 and c, d be the same as in Lemma 4.5.1 and the condition of
Lemma 4.5.1 holds;

Copyright 2006 by Taylor & Francis Group, LLC

 98 Topological Degree Theory and Applications

(2) Let A, C > 0 and B ∈ [0, 1] be such that

|f (t, x, r, q)| ≤ Ar2 + B|q| + C

for all (t, x) ∈ [0, T ] × [−M, M ] and r, q ∈ R.

Then there are constants M1 , M2 > 0 such that, for λ ∈ [0, 1] and any solution
xλ (·) of (E 4.5.2),

|xλ (t)| ≤ M, |x0λ (t)| ≤ M1 , |x00λ (t)| ≤ M2 for all t ∈ [0, T ].

Proof. Replace f by λf and g by λg in Lemma 4.5.2 and Lemma 4.5.3,

respectively, one easily gets the conclusions.
To formulate an existence result for the problem (E 4.5.1), let Y = C[0, T ]
be the Banach space of continuous functions on [0, T ] with the supremum
norm and C k ([0, T ]) be the Banach space of k-times continuously differentiable
functions with the norm kx(·)kk = max{kx(i) (·)k : 0 ≤ i ≤ k}, where k · k is
the norm in C([0, T ]). Set

X = {x(·) ∈ C 2 ([0, T ]) : x(0) = x(T ), x0 (0) = x0 (T )},

and let S : X → Y be a mapping defined by

Sx(t) = x00 (t) for all x(·) ∈ X, t ∈ [0, T ].

It is well known that S is a Fredholm mapping of index zero, N (S) = {x(·) ∈

RT
X : x(t) is constant}, R(S) = {y(·) ∈ Y : 0 y(t)dt = 0}, X = N (S) ⊕ X 0
and Y = N (S) ⊕ R(S). It is easy to see that K = S − I : X → Y is
a linear isomorphism. Let (Yn , Qn ) be a projection scheme for Y and set
Xn = K −1 Yn , then d(x, Xn ) → 0 as n → ∞, so ΠS = (Xn , Yn , Qn ) is
admissible for mappings from X to Y . Now, S is A-proper with respect to
ΠS .
Finally, we set N : X → Y by

N x(t) = f (t, x(t), x0 (t), x00 (t)) for all t ∈ [0, T ], x(·) ∈ X.

Then N is continuous and maps bounded subsets in X to bounded subsets in

Y.
Now, we have the following:
Theorem 4.5.6. Let g(·) ∈ Y , f (t, x, r, q) : [0, T ]×R3 → R be a continuous
function and S, ΠS , N be the same as the above. Suppose that the following
conditions hold:

(1) S − λN is A-proper with respect to ΠS for each λ ∈ (0, 1];

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 99

(2) There exist constants M > 0 and c, d ∈ R with c ≤ gm ≤ gM ≤ d, and

x ≤ M implies that d < f (t, x, 0, r) for all t ∈ [0, T ] and r ∈ R3 , while
x ≤ −M implies that f (t, x, 0, r) < c for all t ∈ [0, T ] and r ∈ R3 , where
gm = min{g(t) : t ∈ [0, T ]} and gM = max{g(t) : t ∈ [0, T ]};

(3) There are continuous functions A(t, x), C(t, x) > 0 which are bounded
on compact subsets of [0, T ] × R and a constant B ∈ [0, 1] such that

|f (t, x, r, q)| ≤ A(t, x)r2 + B|q| + C(t, x)

for all (t, x) ∈ [0, T ] × [−M, M ] and r, q ∈ R.

Then the problem (E 4.5.1) has a solution.

Proof. Take r > {M, M1 , M2 }, where M, M1 , M2 are the same as in
Lemma 4.5.5. Put Ω = B(0, r) = {x(·) : kx(·)k2 < r}. Then Sx 6= λN x − λg
RT
for all λ ∈ (0, 1] and x ∈ ∂Ω. Set Qu = T1 0 u(t)dt for all u ∈ Y , then
Q : Y → N (S) is a projection.
We define a bilinear form on Y × X by
Z T
[u, x] = u(t)x(t)dt for all (u, x) ∈ Y × X.
0

If x(·) ∈ ∂Ω∩N (S), then kxk2 = r > M , so x ≡ r or −r. Thus the assumption
(2) implies that
Z T
QN (c) − Qg = [f (t, c, 0, 0) − g(t)]dt 6= 0
0

and Z T
[QN (c) − Qg, c] = [f (t, c, 0, 0) − g(t)]cdt = 0,
0

where c = r or −r. From Theorem 4.3.3, we know that the problem (E 4.5.1)
has a solution. This completes the proof.
A special case of (E 4.5.1) is the following:
(
x00 (t) = f (t, x(t), x0 (t)) − g(t), t ∈ [0, T ],
(E 4.5.3)
x(0) = x(T ), x0 (0) = x0 (T ).

In this case, the mapping given by N (x)(t) = f (t, x(t), x0 (t)) for all t ∈ [0, T ]
is compact, so S − λN is A-proper with respect to ΠS . From Theorem 4.5.6,
the following immediately holds:
Corollary 4.5.7. Let g(·) ∈ Y , f (t, x, r) : [0, T ] × R2 → R be a continuous
function and S, ΠS , N be the same as above. Suppose that the following
conditions hold:

Copyright 2006 by Taylor & Francis Group, LLC

 100 Topological Degree Theory and Applications

(1) There exist constants M > 0 and c, d ∈ R with c ≤ gm ≤ gM ≤ d, and

x ≤ M implies that d < f (t, x, 0) for all t ∈ [0, T ], while x ≤ −M implies
that f (t, x, 0) < c for all t ∈ [0, T ], where gm = min{g(t) : t ∈ [0, T ]}
and gM = max{g(t) : t ∈ [0, T ]};
(2) There are continuous functions A(t, x), C(t, x) > 0 which are bounded
on compact subsets of [0, T ] × R such that

|f (t, x, r)| ≤ A(t, x)r2 + C(t, x) for all (t, x) ∈ [0, T ] × [−M, M ].

Then the problem (E 4.5.3) has a solution.

Next, we consider the wave equation:


 utt (t, x) − uxx (t, x) − h(u(t, x)) = f (t, x),

t ∈ (0, 2π), x ∈ (0, π),
(E 4.5.4)
u(t, 0) = u(t, π) = 0, t ∈ (0, 2π),


u(0, x) = u(2π, x), x ∈ (0, π),


where h : R → R is a continuous function satisfying

|h(u)| ≤ δ|u| + γ (4.5.4)

and f (·) ∈ L2 ((0, 2π) × (0, π)), where δ > 0 and γ > 0 are constants.
We say that u ∈ L2 ((0, 2π) × (0, π)) is a weak solution of the problem (E
4.5.4) if
(u, vtt − vxx ) − (h(u(t, x)), v) = (f (t, x), v)
for all v ∈ C 2 ([0, 2π] × [0, π]) with v(t, 0) = v(t, π) = 0 for all t ∈ [0, 2π] and
v(2π, x) = v(0, x) for all x ∈ [0, π].
Let L : D(L) ⊂ L2 ((0, 2π)×(0, π)) → L2 ((0, 2π)×(0, π)) be the wave opera-
tor Lu = utt −uxx . Then it is well known that L is self-adjoint, densely defined,
and closed, and Ker(L) is infinite dimensional with Ker(L)⊥ = Im(L). Thus
L is a Fredholm mapping of zero index type. Let P : L2 ((0, 2π) × (0, π)) →
Ker(L) be the projection. Then (L + λP )−1 : L2 ((0, 2π) × (0, π)) → D(L) is
compact for all λ > 0.
For each η > 0, consider the following equation:


 utt (t, x) − uxx (t, x) + ηu(t, x) − h(u(t, x)) = f (t, x),

t ∈ (0, 2π), x ∈ (0, π),
(E 4.5.5)
u(t, 0) = u(t, π) = 0, t ∈ (0, 2π),


u(0, x) = u(2π, x), x ∈ (0, π),


where h, f is the same as in (E 4.5.4). Let uη be the weak solution of (E 4.5.5)

if it exists, and we set S = {uη : η > 0}.

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 101

Theorem 4.5.8. One of the following conclusion holds:

(1) The problem (E 4.5.4) has a weak solution;
(2) S is unbounded in L2 ((0, 2π) × (0, π).
Proof. Let N : L2 ((0, 2π) × (0, π)) → L2 ((0, 2π) × (0, π)) be defined by

N u(t, x) = h(u(t, x)) + f (t, x) for all u(t, x) ∈ L2 ((0, 2π) × (0, π)).

By (4.5.4), N is a bounded and continuous mapping. We may assume that

(2) does not hold, i.e., S is bounded, so there exists r0 > 0 such that

kuη kL2 < r0 for all uη ∈ S. (4.5.5)

Let Ω = {u(t, x) ∈ L2 ((0, 2π)×(0, π)) : kukL2 < r0 }. By (4.5.5), we know that
P N u 6= ηP u for all u ∈ C 2 ([0, 2π] × [0, π]) ∩ ∂Ω and η > 0. We may assume
that Lu 6= N u for all u ∈ C 2 ([0, 2π] × [0, π]) ∩ ∂Ω. By Theorem 4.4.16, we
have deg(L − N, Ω, 0) = {1}. Thus the problem (E 4.5.4) has a weak solution.
This completes the proof.
Remark. The results of Sections 4.1-4.3 in this chapter can be found in
[239].

4.6 Exercises
1. Let X be a separable Banach space with a projection scheme Π =
{Xn , Pn }, Ω ⊂ X be an open bounded subset with x0 ∈ Ω and T :
Ω → X be a P1 compact mapping satisfying

T x − x0 6= λ(x − x0 ) for all x ∈ ∂Ω, λ > 1.

Show that T has a fixed point in Ω.

2. Let X be a separable Banach space with a projection scheme Π =

{Xn , Pn }, Ω ⊂ X be an open bounded subset with 0 ∈ Ω and tΩ ⊂ Ω
for λ ∈ (0, 1) and T : Ω → X be a P1 compact mapping satisfying
T (∂Ω) ⊆ Ω. Show that T has a fixed point in Ω.

3. Let X be a separable Banach space with a projection scheme Π =

{Xn , Pn }, Ω ⊂ X be an open bounded subset with 0 ∈ Ω and T : Ω → X
be a P1 compact mapping satisfying

kx − N xk2 ≥ kN xk2 − kxk2 for all x ∈ ∂Ω.

Show that T has a fixed point in Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 102 Topological Degree Theory and Applications

4. Let H be a real separable Hilbert space and T : H → H be a continuous

mapping satisfying (T x − T y, x − y) ≥ ckx − yk2 for all x, y ∈ H and
some c > 0. Show that T is A-proper.
5. Let X be a separable Banach space with a projection scheme Π =
{Xn , Pn }, Ω ⊂ X be an open bounded subset and T : Ω → X be a
continuous A-proper mapping. Show that T −1 y for all y ∈ T (Ω) is
compact.
6. Let X be a separable Banach space with a projection scheme Π =
{Xn , Pn }, c = supn kPn k = c, D ⊂ X be closed and T : D → X
be a mapping satisfying β(T (B)) ≤ kβ(B) for all bounded subset B of
D. Show that λI − T is A-proper for each λ > kc.
7. Let X be a separable Banach space with an operator scheme Π =
{Xn , Pn } and T : X → X be a continuous strongly accretive mapping,
i.e.,
(x − y, T x − T y)+ ≥ ckx − yk2 for all x, y ∈ X, some c > 0.
Show that T is A-proper.
8. Let X be a separable Banach space with an operator scheme Π =
{Xn , Pn } and T : X → X be a linear bounded operator. Show that T is
A-proper and one to one if and only if R(T ) = X and kPn T xk ≥ ckxk
on Xn for some c > 0 and all n ≥ n0 .
9. Let L : D(L) ⊂ X → Y be a Fredholm mapping of zero index type
and {Yn , Qn } be a projection scheme for Y . Assume that X is reflexive.
Show that, if G ⊂ X is bounded closed, convex and N : G → Y is
continuous compact, then L + kN : G ∩ D(L) → Y is pseudo A-proper
with respect to Γλ,L for each λ 6= 0.
10. Let L : D(L) ⊂ X → Y be a Fredholm mapping of zero index, {Yn , Qn }
be a projection scheme for Y and Ω ⊂ be an open bounded subset. Let
N : Ω → Y be a bounded mapping such that the following conditions
are satisfied:
(1) L − λN is A-proper with respect to ΓL for each λ ∈ (0, 1];
(2) There are sets E1 ⊂ E2 ⊂ R(L) such that 0 ∈ E1 and λE2 ⊂ E1
for all λ ∈ (0, 1) and, for all x ∈ ∂(Ω∩D(L)) such that N x ∈ R(L),
one has Lx ∈ / E1 and N x ∈ E2 .
If Deg(L−QN, D(L)∩Ω, 0) 6= 0, show that Lx−N x = 0 has a solution,
where Q is as in Section 4.1.
11. Let L : D(L) ⊂ X → Y be a Fredholm mapping of zero index, {Yn , Qn }
be a projection scheme for Y , Ω ⊂ be an open bounded subset. Let
N : Ω → Y be a bounded mapping such that the following conditions
are satisfied:

Copyright 2006 by Taylor & Francis Group, LLC

 GENERALIZED DEGREE THEORY FOR A-PROPER MAPS 103

(1) L − λN is A-proper with respect to ΓL for each λ ∈ (0, 1];

/ R(L) for all x ∈ L−1 ∩ ∂Ω;
(2) N x ∈
(3) kLx − N xk2 ≥ kN xk2 − kLxk2 for all x ∈ D(L) \ (N (L) ∩ ∂Ω);
(4) Deg(L − QN, D(L) ∩ Ω, 0) 6= 0.
Show Lx = N x has a solution in Ω ∩ D(L).

Copyright 2006 by Taylor & Francis Group, LLC

 Chapter 5
COINCIDENCE DEGREE
THEORY

In the 1970s, Mawhin systematically studied a class of mappings of the form

L + T , where L is a Fredholm mapping of index zero and T is a nonlinear
mapping, which he called a L- compact mapping. Based on the Lyapunov-
Schmidt method, he was able to construct a degree theory for such mapping.
The goal of this chapter is to introduce Mawhin’s degree theory for L-
compact mappings. This chapter has four main sections.
We present some introductory material on Fredholm mappings and their
relations with A-proper mappings in Section 5.1.
In Section 5.2, we define L-compact mappings (here L is a Fredholm map-
ping) and we introduce the coincidence degree. Various properties of this
degree (see Theorem 5.2.2 and Lemma 5.2.6) are also discussed in this sec-
tion.
In Section 5.3, various consequences of the degree theory in Section 5.2 are
presented (see, in particular, Theorem 5.3.5).
An application to the periodic ordinary differential equations is presented
in Section 5.4.

5.1 Fredholm Mappings

Definition 5.1.1. Let X and Y be normed spaces. A linear mapping
L : D(L) ⊂ X → Y is called a Fredholm mapping if

(1) Ker(L) has finite dimension;

(2) Im(L) is closed and has finite codimension.

Proposition 5.1.2. Let X be a Banach space and T : X → X be a linear

bounded mapping. Then dim(Ker(T )) < ∞ and Im(T ) is closed if and
only if, for xn ∈ B(0, 1) such that T xn → y, thus (xn )∞
n=1 has a convergent
subsequence.

105

Copyright 2006 by Taylor & Francis Group, LLC

 106 Topological Degree Theory and Applications

Proof. For the ”if” part, we know from our assumption that {x : T x =
0, kxk ≤ 1} is compact and thus Ker(T ) is finite dimensional. We have
X = Ker(T ) ⊕ M for some closed subspace M of X. Obviously, we have
T (M ) = Im(T ). Since T : M → Im(T ) is one to one, it follows that

kT xk ≥ ckxk for all x ∈ M, some c > 0

and, from this, we deduce that T (M ) is closed. Therefore, Im(T ) is closed.

For the ”only if” part, assume that xn ∈ B(0, 1) such that T xn → y. As
before, X = Ker(T ) ⊕ M , so xn = zn + mn for some xn ∈ Ker(T ) and
mn ∈ M . Thus T mn → y. However, the restriction of T to M is continuous,
one to one and onto Im(T ), so mn → m ∈ M . Recall that dim(N (T )) is
finite, so (xn )∞
n=1 has a convergent subsequence. This completes the proof.

Proposition 5.1.3. Let X be a Banach space, T : X → X be a linear

bounded Fredholm operator and K : X → X be a linear continuous compact
mapping. Then T + K is a Fredholm mapping.
Proof. The proof is left to the reader as an exercise.
Recall that the codimension of Im(L) is the dimension of Coker(L) =
Y /Im(L). If L is a Fredholm mapping, then its index is defined by

Ind(L) = dim(Ker(L)) − dim(Coker(L)).

Now, assume that L is a Fredholm mapping. Then there exist two linear
continuous projections P : X → X and Q : Y → Y such that

Im(P ) = Ker(L), Ker(Q) = Im(L).

Also, we have

X = Ker(L) ⊕ Ker(P ), Y = Im(L) ⊕ Im(Q)

as the topological direct sums.

Obviously, the restriction of LP of L to D(L) ∩ Ker(P ) is one to one and
onto Im(L) and so its inverse KP : Im(L) → D(L) ∩ Ker(P ) is defined. We
denote by KP Q : Y → D(L) ∩ Ker(P ) the generalized inverse of L defined by
KP Q = KP (I − Q).
Proposition 5.1.4. Let X, Y be separable Banach spaces and L : D(L) ⊂
X → Y be a densely defined Fredholm mapping with Ind(L) = m ≥ 0.
Then there exist a sequence of monotonically increasing finite dimensional
subspaces (Xn )∞ ∞ ∞
n=1 ⊂ D(L) such that ∪n=1 Xn is dense in X, and let (Pn )n=1
be a sequence of linear continuous projections on X with Im(Pn ) = Xn for
each n ≥ 1 and Pn x → x for all x ∈ X as n → ∞ and (Qn )∞ n=1 be a sequence
of linear continuous projections on Y with Im(Qn ) = Yn for each n ≥ 1 such
that dim(Xn ) − dim(Yn ) = m for each n ≥ 1.

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 107

Proof. Since L is Fredholm, there exist two linear continuous projections

P : X → X and Q : Y → Y such that

Im(P ) = Ker(L), Ker(Q) = Im(L),

X = Ker(L) ⊕ Ker(P ), Y = Im(L) ⊕ Im(Q).

By the assumption D(L) is dense in X, we may choose a a sequence (Xn )∞ n=1
of monotonically increasing finite dimensional subspaces of D(L) such that
∪∞ ∞
n=1 Xn is dense in X and Ker(L) ⊂ Xn , a sequence (Pn )n=1 of linear con-
tinuous projections on X with Im(Pn ) = Xn for each n ≥ 1 and Pn x → x
for each x ∈ X as n → ∞. Obviously, P Pn = P , Pn (Ker(P )) ⊂ Ker(P ) and
(I − Pn )(X) ⊂ Ker(P ). Set Qn = Q + LPn KP Q , then Qn is continuous. It is
easy to check that Q2n = Qn for each n ≥ 1. Thus, Qn is a linear continuous
projection for each n ≥ 1. Finally, set Yn = Qn (Y ); then Im(Q) ⊂ Qn ,
Qn Lx = LPn x for all x ∈ D(L) and dim(Xn ) − dim(Yn ) = m for each n ≥ 1.
This completes the proof.
In the sequel, we denote the approximation scheme constructed in Propo-
sition 5.1.4 by Γm = {Xn , Pn ; Yn , Qn }.
Definition 5.1.5. Let G ⊂ X be a non-empty set, Gn = G ∩ Xn for
n = 1, 2, · · · . A mapping T : G → Y is said to be A-proper (respectively,
pseudo A-proper) with respect to Γm if Tn = Qn T : Gn → Yn is continuous
and, if xnk ∈ Gnk such that (xnk )∞k=1 is bounded and Qnk (T xnk − g) → 0
as k → ∞ for some g ∈ Y , then there exists a subsequence (xnkl ) with
xnkl → x0 ∈ G (resp., x0 ∈ G exists) such that T x = g.
Proposition 5.1.6. Let X, Y be separable Banach spaces and L : D(L) ⊂
X → Y be a densely defined Fredholm mapping of index m > 0. Then L is
A-proper with respect to Γm .
Proof. Let xnk ∈ Gnk be such that (xnk )∞
k=1 is bounded and Qnk (T xnk −
g) → 0 as k → ∞ for some g ∈ Y . Notice that Qnk = Q + LPnk KP Q , so we
get
gnk = Qnk (T xnk − g) = Lxnk − Qg − LPnk KP Q g → 0.
So L(xnk − Pnk KP Q g) = gnk + Qg → Qg. However, Im(L) is closed, so we
have Qg = 0 and L(xnk − Pnk KP Q g) → 0. From which we deduce that

KP L(xnk − Pnk KP Q g) = (I − P )(xnk − Pnk )KP Q g) → 0.

Now, the compactness of P implies that (xnk ) has a subsequence (xnkl ) such
that xnkl → x0 and x0 − KP Q g = P x0 − P KP Q g. Since L is closed, we have
L(x0 − KP Q g) = 0. Thus Lx0 = g. This completes the proof.
More precisely, we have the following result between the A-proper mapping
and the Fredholm mapping:

Copyright 2006 by Taylor & Francis Group, LLC

 108 Topological Degree Theory and Applications

Proposition 5.1.7. Let X, Y be separable Banach spaces and L : X → Y

be a linear bounded mapping. Then L is a Fredholm mapping of index m ≥ 0
if and only if L is A-proper with respect to some projectional scheme.
Proposition 5.1.8. Let X, Y be Banach spaces, L : D(L) ⊂ X → Y be a
Fredholm mapping of index m > 0, J 0 : Im(Q → Ker(L) be a monomorphism
and N : D(N ) ⊂ X → Y be a mapping. Then Lx − N x = y if and only if

x − P x − J 0 QN x − KP Q N x = KP Q y + J 0 Qy.

Proof. If Lx − N x = y, then we have

L(I − P )x − QN x − (I − Q)N x = Qy + (I − Q)y,

so −QN x = Qy and L(I − P )x − (I − Q)N x = (I − Q)y, i.e., −J 0 QN x = JQy

and x − P x − KP Q N x = KP Q y. Thus we have

x − P x − J 0 QN x − KP Q N x = KP Q y + J 0 Qy.

On the other hand, if x − P x − J 0 QN x − KP Q N x = KP Q y + J 0 Qy, then,

since J 0 N x ∈ Ker(L) = Im(P ) and KP Q N x ∈ D(L) ∩ Ker(P ), we have
x − P x − KP Q N x = KP Q y and −J 0 QN x = J 0 Qy and so Lx − (I − Q)N x =
(I − Q)y and −QN x = Qy. Thus Lx − N x = y. This completes the proof.
Proposition 5.1.9. Let X, Y be separable Banach spaces, L : D(L) ⊂
X → Y be a densely defined Fredholm mapping of index m > 0 and Γm
the same as Proposition 5.1.4. Let Ω ⊂ X be an open bounded subset and
N : Ω → Y be a bounded mapping. If I − P − J 0 QN − KpQ N is A-proper
with respect to Γ = {Xn , Pn }, then L − N is A-proper with respect to Γm .
Proof. For any xnk ∈ Ω ∩ D(L) ∩ Xnk such that Qnk (Lxnk − N xnk − y → 0
as k → ∞ for some y ∈ Y , we recall that Qnk = Q + LPnk KpQ , so we have

Qnk (Lxnk − N xnk − y)

= L(I − Pnk )xnk − Q(N xnk + y) − LPnk (I − Q)(N xnk + y)
→0

as k → ∞. Therefore, we have

yk = L(I − Pnk )xnk − LPnk (I − Q)(N xnk + y) → 0,

zk = QN xnk + Qy → 0
and thus

hk = KP yk = (I − P )xnk − Pk KP Q (N xnk + y) → 0,

wk = J 0 zk = J 0 QN xnk + J 0 Qy → 0.

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 109

From which we deduce that

xnk − P xnk − Pnk Kp N xnk → KP Q y,

J 0 QN xnk → −J 0 Qy,
which immediately implies that

Pnk (I − P − J 0 QN − KpQ N )xnk → KpQ y + J 0 Qy.

So (xnk )∞
k=1 has a convergence subsequence (xnkl ) with xnkl → x0 and x0 −
P x0 − J 0 QN x0 − KpQ N x0 = KpQ y + J 0 Qy. By Proposition 5.1.8, we have
Lx0 − N x0 = y. Thus L − N is A-proper with respect to Γm . This completes
the proof.
Assume now that L is a Fredholm mapping of index zero. Then, for any
isomorphism J : Im(Q) → Ker(L), the mapping JQ+KP Q is an isomorphism
from Y onto D(L) and

(JQ + KP Q )−1 x = (L + J −1 P )x for all x ∈ D(L).

In fact, if y ∈ Y , we have

(JQ + KP Q )y = x ⇔ JQy = P x,

KP Q y = (I − P )x ⇔ Qy = J −1 P x,
LP KP Q y = L(I − P )x ⇔ Qy = J −1 P x,
(I − Q)y = Lx ⇔ y = (J −1 P + L)x.
Example 5.1.10. Let X be a real Banach space and T : X → X be a
linear continuous compact mapping. Then, by (5) of Theorem 2.1.15, we know
that dim(ker(I + T )) = dim(codim(I + T )) < +∞, so I + T is a Fredholm
mapping of index zero.
Example 5.1.11. Let f : [0, T ] → R be in L1 and consider the following
problem: (
x0 (t) = f (t), t ∈ (0, T ),
(E 5.1.1)
x(0) = x(T ).
We set X = C([0, T ], R), the space of all continuous function from [0, T ] to
R, Y = L1 ([0, T ], R) × R, y0 = (f (·), 0) and define a mapping L : X → Y by

Lx(·) = (x0 (·), x(0) − x(T )) for all x(·) ∈ dom(L),

where dom(L) = {x(·) ∈ C([0, T ], R) : x0 (·) ∈ L1 ([0, T ], R)}.

One can easily see that (E 5.1.1) is equivalent to the following equation:

Lx(·) = y0 .

Copyright 2006 by Taylor & Francis Group, LLC

 110 Topological Degree Theory and Applications

It is easy to see that

ker(L) = {x(·) ∈ C([0, T ], R) : x(t) = c, t ∈ [0, T ], c ∈ R},

Z T
Im(L) = (f (·), − f (s)ds) for all f (·) ∈ L1 ([0, T ], R).
0

Obviously, Im(L) is closed and dim(Ker(L)) = dim(Coker(L)) = 1. Thus L

is a Fredholm mapping of index zero.

5.2 Coincidence Degree for L-Compact Mappings

In this section, we define coincidence degree for L-compact mappings and
give some properties of coincidence degree.
Definition 5.2.1. Let L : D(L) ⊂ X → Y be a Fredholm mapping, E be
a metric space and T : E → Y be a mapping. We say that T is L-compact
on E if QT : E → Y , KP Q T : E → X are continuous and QT (E), KP Q T (E)
are compact, where all notations are the same as in Section 5.1.
Now, we are ready to introduce the conincidence degree:
Coincidence degree. Let X, Y be real normed spaces, L : D(L) ⊂ X → Y
be a Fredholm mapping of index zero and Ω be an open bounded subset of X.
Suppose that F = L + T : D(L) ∩ Ω → Y is a mapping and T : Ω → Y is L-
compact on Ω. Suppose also that 0 ∈
/ F (D(L)∩∂Ω). Let J : Im(Q) → Ker(L)
be an isomorphism. Put HPJ Q = JQ + KP Q . It is easy to check that

HPJ Q F = KP Q L + HPJ Q T = I − P + (JQ + KP Q )T.

Consequently, 0 ∈/ HPJ Q F (D(L) ∩ ∂Ω) (if 0 ∈ HPJ Q F (D(L) ∩ ∂Ω, then 0 =

KP Q (Lx + T x) + JQT x for some x ∈ D(L) ∩ ∂Ω, so QT x = 0 and (I −
Q)(Lx + T x) = 0. Thus Lx + T x = 0, which is a contradiction). By the L-
compactness of T , the Leray Schauder degree deg(I − P + (JQ + KP Q )T, Ω, 0)
is well defined.
Now, we define a degree by

DJ (L + T, Ω, 0) = deg(I − P + (JQ + KP Q )T, Ω, 0),

which is called the coincidence degree of L and −T on Ω∩D(L). One can easily
prove that this definition does not depend on the choice of P, Q. It is known
that DJ (L + T, Ω, 0) is a constant for some J depending on orientations on
Ker(L) and Coker(L) (see [203]), so the coincidence degree in [203] is defined
only for those J 0 s. The definition given here depends on the J.

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 111

Remark. (1) If dim(X) = dim(Y ) < +∞ and we take L = 0, then any

continuous mapping T on Ω is L-compact. If we take P = I and Q = I, then
it follows that KP Q = 0, so HPJ Q F = JT and thus we have

DJ (T, Ω, 0) = deg(JT, Ω, 0) = sign(detJ)deg(T, Ω, 0).

Therefore, if we only take those J such that detJ > 0, then we have DJ (T, Ω, 0) =
deg(T, Ω, 0), which is the Brouwer degree.
(2) If X = Y and we take L = I, then any continuous compact mapping T
on Ω is L-compact. If we take P = Q = 0, then KP Q = I, J = 0 : {0} → {0}
and HPJ Q F = I + T . Thus DJ (I + T, Ω, 0) = deg(I + T, Ω, 0), which is the
Leray Schauder degree.
Theorem 5.2.2. The coincidence degree of L and −T on Ω has the fol-
lowing properties:
(1) If Ω1 and Ω2 are disjoint open subsets of Ω such that 0 ∈
/ F (D(L) ∩
Ω \ (Ω1 ∪ Ω2 ), then

DJ (L + T, Ω, 0) = DJ (L + T, Ω1 ) + DJ (L + T, Ω2 , 0);

(2) If H(t, x) : [0, 1]×Ω → Y is L-compact on [0, 1]×Ω and 0 6= Lx+H(t, x)

for all (t, x) ∈ [0, 1] × ∂Ω, then DJ (L + H(t, ·), Ω, 0) does not depend on
t ∈ [0, 1];
(3) If DJ (L + T, Ω, 0) 6= 0, then 0 ∈ (L + T )(D(L) ∩ Ω).

Corollary 5.2.3. If T1 , T2 are L-compact mappings on Ω and T1 x = T2 x

for all x ∈ D(L) ∩ ∂Ω, then

DJ (L + T1 , Ω, 0) = DJ (L + T2 , Ω, 0).

Proof. We define H(t, x) : [0, 1] × Ω → Y by

H(t, x) = tT1 x + (1 − t)T2 x for all (t, x) ∈ [0, 1] × Ω.

Then H is L-compact. Therefore, it follows from (3) of Theorem 5.2.2 that

DJ (L + T1 , Ω, 0) = DJ (L + T2 , Ω, 0).
Proposition 5.2.4. Let X, Y be real normed spaces, L : D(L) ⊂ X → Y
be a Fredholm mapping of index zero, Y0 be a finite dimensional subspace of
Y satisfying Y = Im(L) ⊕ Y0 algebraically and Ω be an open bounded subset
of X. If T is L-compact on Ω ∩ D(L), and T (Ω) ⊂ Y0 , then

DJ (L + T, Ω, 0) = signdet(J)deg(T, Ω ∩ Ker(L), 0),

where deg(·, ·, ·) is the Brouwer degree.

Copyright 2006 by Taylor & Francis Group, LLC

 112 Topological Degree Theory and Applications

Proof. Since L is a Fredholm mapping of index zero, we get Y = Im(L)⊕Y0

topologically. Take Q : Y → Y with Im(Q) = Y0 , then QT x = T x for x ∈ Ω
and
HPJ Q F = (JQ + KP Q )(L + T ) = I − P + JT.
Note that (P − JT )(Ω) ⊂ Ker(L) and I = P on Ker(L); thus, by Theorem
2.2.9, we have
DJ (L + T, Ω, 0) = deg(I − P + JT, Ω, 0) = deg(I − P + JT, Ω ∩ Ker(L), 0).
But I = P on Ker(L) and thus we get
DJ (L + T, Ω, 0) = deg(JT, Ω ∩ Ker(L), 0)
= signdet(J)deg(T, Ω ∩ Ker(L), 0).
This completes the proof.
Lemma 5.2.5. Let A : Im(Q) → D(L) be a linear mapping such that
P A : Im(Q) → Ker(L) is an isomorphism. Then HPAQ = AQ + KP Q is an
algebraic isomorphism from Y onto D(L) and
(HPAQ )−1 = L − LA(P A)−1 P + (P A)−1 P.

Moreover, if T is L-compact on Ω, then

HPAQ (L + T ) = I − P + HpQ
A
T

with HPAQ T : Ω → X continuous and compact.

Proof. For any x ∈ D(L), HPAQ z = x if and only if P AQz = P x and

(I − P )AQz + KP Q z = (I − P )x, i.e.,
Qz = (P A)−1 Ox, LAQz + (I − Q)z = Lx,
z = Lx − LA(P A)−1 P x + (P A)−1 P x.
Finally, we have
HPAQ (L + T ) = (AQ + KP Q )(L + T )
= KP Q L + (AQ + KP Q )T
= (I − P ) + HPAQ T.

By the assumption, QT and KP Q T are continuous and compact on Ω, so

HPAQ T is continuous and compact on Ω. This completes the proof.
Lemma 5.2.6. Let A, B : Im(Q) → D(L) be linear mappings such that
P A, P Q : Im(Q) → Ker(L) are isomorphisms and T is L-compact. Then
deg(I − P + HPBQ T, Ω, 0)
= deg(I − (A − B)(P A)−1 P, B(0, r), 0)deg(I − P + KPAQ T, Ω, 0)

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 113

for any r > 0.

Proof. First, we have

(I − (A − B)(P A)−1 P )(I − P + KPAQ T )

= I − P + AQ + KpQ )T − (A − B)(P A)−1 P AQT
= I − P + HPBQ T.

Therefore, it follows that

deg(I − P + HPBQ T, Ω, 0) = deg(I − (A − B)(P A)−1 P )(I − P + KPAQ T ), Ω, 0).

Moreover, if x − (A − B)(P A)−1 P x = 0, then P B(P A)−1 P x = 0, so P x = 0.

Thus x = 0 and so I − (A − B)(P A)−1 P is a homeomorphism of X onto itself.
By the product formula of Theorem 2.2.8, we know that

deg(I − (A − B)(P A)−1 P )(I − P + KPAQ T ), Ω, 0)

= Σi deg(I − (A − B)(P A)−1 P, (Ui , 0)deg(I − P + KPAQ T, Ω, Ui ),
A
where Ui are connected components of X \ (I − P + KpQ T )(∂Ω).
There are now two cases:
/ Uk for k 6= i. Thus deg(I−(A−B)(P A)−1 P, Uk , 0) =
Case (1) If 0 ∈ Ui , then 0 ∈
0, so we have

deg(I − (A − B)(P A)−1 P )(I − P + KPAQ T ), Ω, 0)

= deg(I − (A − B)(P A)−1 P, Ui , 0)deg(I − P + KPAQ T, Ω, 0).

By the excision property of Leray Schauder degree, the conclusion is true.

/ Ui for all i ≥ 1, then deg(I − (A − B)(P A)−1 P, Ui , 0) = 0
Case (2) If 0 ∈
for all i ≥ 1 and thus the conclusion holds. This completes the proof.
From Lemma 5.2.6, the following holds immediately:
Corollary 5.2.7. Let T be L-compact on Ω ∩ D(L) and A be the same as
in Lemma 5.2.6. Then

DJ (L + T, Ω, 0)
= deg(I − (A − J)(P A)−1 P, B(0, r))deg(I − P + HPAQ T, Ω, 0)

for all r > 0.

Proposition 5.2.8. If T : X → Y is linear and L-completely continuous,
Ker(L + T ) = {0} and Ω ⊂ X is a nonempty open bounded subset such that
0 6∈ ∂Ω, then (
0, 0∈/ Ω,
|DJ (L + T, Ω, 0)| =
1, 0 ∈ Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 114 Topological Degree Theory and Applications

Proof. By definition, we have

DJ (L + T, Ω, 0) = deg(I − P + KP Q T, Ω, 0).

Thus the conclusion follows from the assumption and Theorem 2.2.4.
Theorem 5.2.9. If Ω ⊂ X is open bounded with 0 ∈ Ω, Ω is symmetric
with respect to 0 and T is L-compact on Ω ∩ D(L) such that T (−x) = −T x
for all x ∈ ∂Ω ∩ D(L), then |DJ (L + T, Ω, 0)| is an odd number.
Proof. Since DJ (L + T, Ω, 0) = deg(I − P + KP Q T, Ω, 0), by the definition
of the Leray Schauder degree and Borsuk’s Theorem (Theorem 1.2.11), we
know that the conclusion is true.
In the following, let L : D(L) ⊂ X → Y be a Fredholm mapping of index
zero, and L = L1 + L2 , where L1 , L2 satisfying the following conditions:

(1) L1 : D(L) → Y is a Fredholm mapping of index zero;

(2) L2 : X → Y is linear and L1 -completely continuous on X.

Now, assume that Ω ⊂ X is a nonempty open bounded subset of X, T :

Ω → Y is L-compact on Ω and T is also L1 -compact on Ω. Set T1 = L2 + T .
Then T1 is L1 -compact on D(L) ∩ Ω and L + T = L1 + T1 .
Let P, Q, J be the linear mapping associated with L as in Section 5.1 and
P1 , Q1 , J1 be the corresponding ones for L1 . Put H = JQ + KP Q and H1 =
J1 Q1 + KP1 Q1 . Then H, H1 : Y → D(L) are algebraic isomorphisms, H −1 =
L+J −1 P and H1−1 = L1 +J1−1 P1 . We set K1 = L2 +J −1 P . It is easy to check
that J −1 P is L1 -completely continuous and hence K1 is also L1 -completely
continuous.
We have seen that

H(L + T ) = I − P + HT, H1 (L + T ) = H1 (L1 + T1 ) = I − P1 + H1 T1 ,

(I − P + HT )(D(L) ∩ Ω) = H(L + T )(D(L) ∩ Ω) ⊂ D(L),

so we have
I − P1 + H1 T1 = H1 H −1 (I − P + HT ). (5.2.1)
On the other hand, H1 H −1 = H1 (L1 + K1 ) = (I − P1 + H1 K1 ). Therefore,
we have come to the following conclusion:
Proposition 5.2.10. I − P1 + H1 T1 = (I − P1 + H1 K1 )(I − P + HT ) and
I − P1 + H1 K1 is a linear homeomorphism on X.
Proof. We have seen that

I − P1 + H1 T1 = (I − P1 + H1 K1 )(I − P + HT ).

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 115

Note that Ker(I − P1 + H1 T1 = Ker(H1 H −1 ) = {0} and P1 , H1 T1 are

compact linear mappings, so I − P1 + H1 K1 is a linear homeomorphism on
X.
Corollary 5.2.11. Under the above assumptions, we have

|DJ1 (L1 + T1 , Ω, 0)| = |DJ (L + T, Ω, 0)|.

Proof. By the definition of coincidence degree and Proposition 5.2.10, we

have

DJ1 (L1 + T1 , Ω, 0)
= deg((I − P1 + H1 K1 )(I − P + HT ), Ω, 0)
= Σi deg(I − P1 + H1 K1 , Ui , 0)deg(I − P + HT, Ω, Ui ),

where Ui are connected components of X \ (I − P + HT )(∂Ω).

We have the following two cases:
Case (1) If 0 ∈ Ui , then 0 ∈
/ Uk for k 6= i, so we have

DJ1 (L1 + T1 , Ω, 0) = deg(I − P1 + H1 K1 , Ui , 0)deg(I − P + HT, Ω, 0).

Since I − P1 + H1 K1 is a homeomorphism on X, by Theorem 2.2.4, we know

that |DJ1 (L1 + T1 , Ω, 0)| = |DJ (L + T, Ω, 0)|.
Case (2) If 0 ∈
/ Ui for all i ≥ 1, then 0 ∈
/ (I − P + HT )(Ω), so we have

|DJ1 (L1 + T1 , Ω, 0)| = |DJ (L + T, Ω, 0)| = 0.

This completes the proof.

Let X, Y be real normed spaces, L : D(L) ⊂ X → Y be a Fredholm
mapping of index zero, Ω ⊂ X be an open bounded subset and T : Ω → Y be
L-compact. Assume that a is an isolated zero of L − T , then we define

iJ (L − T, a) = lim DJ (L − T, B(a, r) ∩ D(L), 0),

r→0

which is called the coincidence index of L and T at a. One may easily see
that this definition is well defined by using the excision property of coincidence
degree.
The following result follows immediately from the definition:
Proposition 5.2.12. Let Ω ⊂ X be an open bounded subset and T : Ω →
Y be L-compact. If (L − T )−1 (0) = {a1 , a2 , · · · , ak } ⊂ Ω, then

DJ (L − T, Ω ∩ D(L), 0) = Σki=1 iJ (L − T, ai ).

Proposition 5.2.13. Let Ω ⊂ X be an open bounded subset with 0 ∈ Ω

and T : Ω → Y be L-compact. Suppose that A : X → Y is a linear mapping

Copyright 2006 by Taylor & Francis Group, LLC

 116 Topological Degree Theory and Applications

such that A is L-compact on any bounded subset of X. Assume that

kQT xk + kKP Q T xk
lim = 0, (5.2.2)
kxk→0 kxk

and Ker(L − A) = {0}. Then 0 is an isolated zero of L − A − T and

iJ (L − A − T, 0) = iJ (L − A, 0).

Proof. By assumption, A is L-compact on any bounded suubset, so I −P −

JQA − KP Q A is a linear completely continuous perturbation of the identity.
Also, Ker(I − P − JQA − KP Q A) = {0} and thus there exists c > 0 such that

k(I − P − JQA − KP Q A)xk ≥ ckxk, x ∈ X.

From the assumption (5.2.2), we know that there exists r > 0 such that

B(0, r) ⊂ Ω, kQT xk + kKP Q T xk ≤ 2−1 ckxk for all x ∈ B(0, r),

which implies that

k(I − P − JQA − KP Q A − tJQT − tKP Q T )xk ≥ 2−1 ckxk

for all x ∈ B(0, r) and t ∈ [0, 1]. Thus we have

DJ (L − A − T, B(0, r) ∩ D(L), 0) = DJ (L − A, B(0, r) ∩ D(L), 0),

i.e., iJ (L − A − T, 0) = iJ (L − A, 0). This completes the proof.

5.3 Existence Theorems for Operator Equations

Let X, Y be real normed spaces, L : D(L) ⊆ X → Y be a linear Fredholm

mapping of index zero and Ω ⊂ X be an open bounded subset with D(L)∩Ω 6=
∅.
Theorem 5.3.1. Let 0 ∈ Ω, and Ω symmetric with respect to 0 and
T : Ω → Y be L-compact. If Lx − T x 6= t(−Lx − T (−x)) for all (t, x) ∈
(0, 1] × D(L) ∩ ∂Ω, then Lx − T x = 0 has a solution in D(L) ∩ Ω.
Proof. Let a mapping H(t, x) : [0, 1] × Ω → Y be defined by

1+t 1−t
H(t, x) = Tx − T (−x) for all (t, x) ∈ [0, 1] × Ω.
2 2

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 117

Then H(t, ·) is a homotopy of L-compact mappings. If Lx − H(t, x) = 0 for

some (t, x) ∈ [0, 1) × D(L) ∩ ∂Ω, then

1−t
Lx − T x = (−Lx − T (−x)),
1+t

which is a contradiction. We may also assume that Lx − T x 6= 0 for x ∈

∂Ω. Otherwise, the conclusion is true. By Theorem 5.2.9, we have DJ (L −
T, Ω, 0) 6= 0, thus Lx − N x = 0 has a solution in D(L) ∩ Ω. This completes
the proof.
Theorem 5.3.2. Let T1 , T2 : Ω → Y be L-compact. If the following
conditions are satisfied:

(1) Lx − tT1 x + (1 − t)T2 x 6= 0 for all (t, x) ∈ (0, 1) × D(L) ∩ ∂Ω;

(2) DJ (L + T2 , Ω, 0) 6= 0;

then Lx − T1 x has a solution in D(L) ∩ Ω.

Proof. We may also assume that Lx − T1 x 6= 0 for all x ∈ ∂Ω. Otherwise,
the conclusion is true. Let H(t, x) : [0, 1] × Ω → Y be defined by

H(t, x) = tT1 x − (1 − t)T2 x for all (t, x) ∈ (t, x) ∈ [0, 1] × Ω.

By assumption, we have Lx − H(t, x) 6= 0 for all (t, x) ∈ [0, 1] × D(L) ∩ ∂Ω.

Thus we have
DJ (L − T1 , Ω, 0) = DJ (L − T2 , Ω, 0) 6= 0

and so Lx − T1 x has a solution in D(L) ∩ Ω. This completes the proof.

Theorem 5.3.3. Let T1 , T2 : Ω → Y be L-compact. If L + T2 is one to
one on Ω and

Lx − t1 T x + (1 − t)(T2 x − p) for all (t, x) ∈ (0, 1) × D(L) ∩ ∂Ω,

where p ∈ (L + T2 )(D(L) ∩ Ω), then Lx − T1 x has a solution in D(L) ∩ Ω.

Proof. Since L + T2 is one to one and p ∈ (L + T2 )(D(L) ∩ Ω), we have
|DJ (L + T2 − p, Ω, 0)| = 1. Thus, the conclusion follows from Theorem 5.3.2.
Theorem 5.3.4. Let A : X → Y be a linear continuous L-compact map-
ping with Ker(L − A) = {0} and T : Ω → Y be L-compact. Suppose that
the following conditions hold:

(1) 0 ∈ Ω and λ∂Ω ⊂ Ω for all λ ∈ (0, 1);

(2) (T − A)(D(L) ∩ ∂Ω) ⊂ (L − A)(D(L) ∩ Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 118 Topological Degree Theory and Applications

Then Lx = T x has a solution in D(L) ∩ Ω.

Proof. Put H(t, x) = (1 − t)Ax + tT x for all (t, x) ∈ [0, 1] × Ω. We claim
that
Lx 6= H(t, x) for all (t, x) ∈ (0, 1) × D(L) ∩ ∂Ω.
If this is not true, then there exist (t, x) ∈ (0, 1) × D(L) ∩ ∂Ω such that
Lx 6= H(t, x). Then we have

(L − A)x = λ(T − A)x

∈ λ(T − A)(D(L) ∩ ∂Ω
⊂ λ(T − A)(D(L) ∩ Ω
= (L − A)(D(L) ∩ λΩ
⊂ (L − A)(D(L) ∩ Ω),

which is impossible because L − A is one to one.

By assumption and Proposition 5.2.8, |DJ (L−A, Ω, 0)| = 1. If Lx−T x = 0
for some x ∈ D(L) ∩ ∂Ω, then the conclusion is true. Otherwise, we have

DJ (I − T, Ω, 0) = DJ (L − A, Ω, 0) 6= 0,

thus Lx − T x = 0 has a solution in D(L) ∩ Ω. This completes the proof.

Theorem 5.3.5. Let T1 , T2 : Ω → Y be L-compact. Let Z ⊂ Y be a
subspace with Y = Im(L) ⊕ Z algebraically and T2 (Ω) ⊂ Z. Suppose that
the following conditions hold:

(1) Lx − (1 − t)T2 − tT1 6= 0 for all (t, x) ∈ (0, 1) × D(L) ∩ ∂Ω;

(2) T2 x 6= 0 for all x ∈ Ker(L) ∩ ∂Ω;

(3) deg(TKer(L) , Ω ∩ Ker(L), 0) 6= 0, where TKer(L) is the restriction of T2

to Ker(L) ∩ Ω.

Then Lx = T1 x has a solution in D(L) ∩ Ω.

Proof. Put H = L − T2 x and let Q : Y → Y be the projection such that
Im(Q) = Z and Ker(Q) = Im(L). Then QT2 = T2 and Hx = 0 if and only
if QHx = 0, (I − Q)Hx = 0, i.e., T2 x = 0 and Lx = 0. Therefore, by the
assumption (2) and Proposition 5.2.4, we have

|DJ (L − T2 , Ω, 0)| = |deg(TKer(L) , Ω ∩ Ker(L), 0)| =

6 0.

Thus, it follows from Theorem 5.3.2 that Lx = T1 x has a solution in D(L)∩Ω.

This completes the proof.
Corollary 5.3.6. Let T : Ω → Y be L-compact. Suppose that the follow-
ing conditions hold:

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 119

(1) Lx − tT 6= 0 for all (t, x) ∈ (0, 1) × (D(L) \ Ker(L)) ∩ ∂Ω;

(2) T x ∈
/ Im(L) = 0 for all x ∈ Ker(L) ∩ ∂Ω;

(3) deg(QTKer(L) , Ω ∩ Ker(L), 0) 6= 0, where Q : Y → Y is the projection

such that Ker(Q) = Im(L).

Then Lx = T x has a solution in D(L) ∩ Ω.

Proof. Put Z = Im(Q) and T2 = QT in Theorem 5.3.5. By the assumption
(2), we know that

QT x 6= 0 for all x ∈ Ker(L) ∩ ∂Ω.

Now, if Lx − (1 − t)QT x − tT x = 0 for some (t, x) ∈ (0, 1) ∩ D(L) ∩ ∂Ω, then

we have
QN x = 0, Lx − tT x = 0.
It easily follows that x ∈ D(L) \ Ker(L)) ∩ ∂Ω, which contradicts the assump-
tion (1). Thus the conditions of Theorem 5.3.5 are satisfied and, consequently,
Lx = T x has a solution in D(L) ∩ Ω.

5.4 Applications to ODEs

In this section, we give some applications of the results to differential equa-
tions.
Let f (t, x, y) : [0, π]×Rn ×Rn → Rn be a function satisfying the Caradéodory
condition, i.e.,

(1) For almost all t ∈ [0, π], f (t, x, y) is continuous in (x, y);

(2) For all (x, y) ∈ Rn × Rn , f (t, x, y) is measurable in t;

(3) For all r > 0, there exists gr (·) ∈ L1 ([0, π], [0, +∞)) such that, For
almost all t ∈ [0, π],

|f (t, x, y)| ≤ gr (t) for all x, y ∈ Rn , |x| ≤ r, |x| ≤ r.

Consider the Picard boundary value problem:

(
−x00 (t) = f (t, x(t), x0 (t)), t ∈ [0, π],
(E 5.4.1)
x(0) = x(π) = 0.

Copyright 2006 by Taylor & Francis Group, LLC

 120 Topological Degree Theory and Applications

Put X = C01 ([0, π], Rn ), Y = L1 ([0, π], Rn ) and let L : D(L) ⊂ X → Y be

defined by Lx(·) = x00 (·), where

D(L) = {x ∈ X : x0 (·) is absolutely continuous in [0, π], x00 (·) ∈ Y }.

Then Ker(L) = {0} and Im(L) = Y . Let a mapping N : C01 ([0, π], Rn ) → Y
be defined by

(N x)(t) = f (t, x(t), x0 (t)) for all t ∈ [0, π].

Then, by Lebesgue’s Theorem, N is a continuous bounded mapping. Now,

the problem (E 5.4.1) is equivalent to the following equation:

Lx = N x, x ∈ D(L). (E 5.4.2)

Since L−1 : Y → X0 is given by

t π s
Z Z Z tZ s
−1
(L y)(t) = y(l)dl − y(l)dl,
π 0 0 0 0

i.e.,
1 t
Z Z π
−1
(L y)(t) = [ s(π − t)y(s)ds + t(π − s)y(s)ds,
π 0 x

L−1 is continuous and compact, and consequently, it follows that N is L-

compact on any bounded subset of X.
Theorem 5.4.1. Suppose that the following conditions hold:
(1) There exist a, b > 0 such that a + b < 1 and g(·) ∈ L1 ([0, π], R+ ) such
that, for almost all t ∈ [0, π],

(x, f (t, x, y)) ≤ a|x|2 + b|x||y| + g(t)|x| for all (x, y) ∈ Rn × Rn ;

(2) There exist c ≥ 0 and h(·) ∈ L1 ([0, π], R+ ) such that, for all x ∈ Rn
with |x| ≤ π(1 − a − b)−1 kgkL1 ,

|f (t, x, y)| ≤ c|y|2 + h(t) for almost all t ∈ [0, π], y ∈ Rn .

Then the problem (E 5.4.1) has a solution.

Proof. Consider the family of problem:
(
−x00 (t) = λf (t, x(t), x0 (t)), t ∈ [0, π],
(E 5.4.3)
x(0) = x(π) = 0.

Let x be a possible solution of (E 5.4.3) for some λ ∈ (0, 1). Then we have

−(x00 (t), x(t)) = λ(f (t, x(t), x0 (t)), x0 (t))

≤ a|x(t)|2 + b|x(t)||x0 (t)| + g(t)|x(t)|.

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 121

So, by integrating over [0, π], it follows that

kx0 k2L2 ≤ akxk2L2 + bkxkL2 kx0 kL2 + kgkL1 kxk0 , (5.4.1)

where
√ kxk0 = maxt∈[0,π] kx(t)k. Notice that kxkL2 ≤ kx0 kL2 and kxk0 ≤
0
πkx kL2 , so (5.4.1) implies that

kxk20 ≤ π(1 − a − b)−1 kgkL1 kxk0 ,

and thus we have

kxk0 ≤ π(1 − a − b)−1 kgkL1 = r1 . (5.4.2)

By (5.4.2) and (5.4.1), we get

√
kx0 kL2 ≤ π(1 − a − b)−1 kgkL1 = r2 . (5.4.3)

Now, from (E 5.4.3) and the assumption (2), we get

|x00 (t)| ≤ c|x0 (t)|2 + h(t),

so x ∈ D(L), and by (5.4.3) it follows that

kx00 kL1 ≤ cπ(1 − a − b)−2 kgk2L1 + khkL1 = r3 . (5.4.4)

On the other hand, by the boundary condition, there exists si ∈ [0, π] such
that x0i (si ) = 0 for 1 ≤ i ≤ n and thus
Z t
|x0i (t)| ≤ | x00i (s)ds| ≤ kx00 kL1 ,
si

which together with (5.4.4) imply that

kx0 k0 ≤ r3 . (5.4.5)

Now, we take r = 1 + r1 + r3 . Then kxkC 1 ≤ r for all λ ∈ (0, 1) and every

possible solution x of (E 5.4.3). Put Ω = B(0, r) ⊂ C01 ([0, π], Rn ). Then we
have Lx 6= tN x for all (t, x) ∈ (0, 1) × ∂Ω. If Lx = N x for some x ∈ ∂Ω, we
know that the problem (E 5.4.1) has a solution. Otherwise, we have

Lx 6= tN x for all (t, x) ∈ [0, 1] × ∂Ω.

Thus DJ (L − N, Ω, 0) = DJ (L, Ω, 0) 6= 0, consequently, it follows that Lx =

N x has a solution in Ω. Thus the problem (E 5.4.1) has a solution. This
completes the proof.
In the following, let X = C([0, 1], Rn ), Y = L1 ([0, 1], Rn ), Q : Y → Y be a
R1
mapping such that Qy(·) = 0 y(s)ds and L : D(L) ⊂ C([0, 1], Rn ) → Y be a
mapping defined by

Lx(t) = x0 (t) for all x(·) ∈ D(L),

Copyright 2006 by Taylor & Francis Group, LLC

 122 Topological Degree Theory and Applications

where

D(L) = {x(·) is absolutely continuous on [0, 1] : x(0) = x(1)}.

Then L is a Fredholm mapping of index zero.

Lemma 5.4.2. Let r > 0 and v(·) ∈ C 1 (Rn , R) be such that v 0 (x) 6= 0
for |x| = r, where v 0 is the gradient of v, and let V : X → Z be defined by
V (x(·))(t) = v 0 (x(t)) for all t ∈ [0, 1]. Then we have

DJ (L − V, B(0, r) ∩ D(L), 0) = DJ (L − QV, B(0, r) ∩ D(L), 0).

Proof. Consider the homotopy T (α, x)) = αV (x(·)) − (1 − α)QV (x(·)) for
all α ∈ [0, 1] and x(·) ∈ X. We claim that Lx(·) 6= T (α, x(·)) for all α ∈ [0, 1]
and x(·) ∈ X. If not, there exist α ∈ [0, 1] and x(·) ∈ X such that

Lx(t) 6= T (α, x(t)) for all t ∈ [0, 1]. (5.4.6)

Multiply both sides of (5.4.6) by x0 (t) and integrate over [0, 1], one gets
Z 1
|x0 (s)|2 ds = 0.
0

Thus x(t) is a constant, |x(t)| = r and v(x(0)) = 0, which is a contradiction.

Thus it follows from Theorem 5.2.2 that

DJ (L − V, B(0, r) ∩ D(L), 0) = DJ (L − QV, B(0, r) ∩ D(L), 0).

This completes the proof.

Theorem 5.4.3. Suppose that the following conditions are satisfied:

(1) There exist v(·) ∈ C 1 ([0, 1], Rn ) such that lim|x|→∞ v(x) = +∞ and
β(·) ∈ L1 ([0, 1], [0, +∞)) such that (v 0 (x), f (t, x)) ≤ β(t) for all x ∈ Rn
and almost all t ∈ [0, 1].

(2) There exist r > 0 and w(·) ∈ C 1 (Rn , R) such that (v 0 (x), w(x)) > 0 for
R1
all x with |x| ≥ r and 0 (w0 (x(s)), f (s, x(s))ds) ≤ 0 for all x(·) ∈ D(L)
satisfying mint∈[0,1] |x(t)| ≥ r.

Then the following equation:

(
x0 (t) = f (t, x(t)), t ∈ [0, 1],
(E 5.4.4)
x(0) = x(1),

has a solution.

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 123

Proof. First, we claim that there exists r0 > 0 such that the solution xλ (·)
of the following equation:
(
x0 (t) = −(1 − λ)v 0 (x(t)) + λf (t, x(t)), t ∈ [0, 1],
(E 5.4.5)
x(0) = x(1),

satisfies kx(·)k0 ≤ r0 . If not, there exist a sequence (λn ) ⊂ [0, 1] and a

sequence (xλn (·)) of solutions such that kxλn k0 ≥ n. By the assumption (1),
we have
d
v(xλn (t)) = −(1 − λn )|v 0 (xλn (t))|2 + λn (v 0 (xλn (t)), f (t, xλn (t))) ≤ β(t).
dt
Extend xλn (·) and β(·) to R with period 1 and then we have
Z t
0
v(xλn (t)) ≤ v (xλn (s)) + β(s)ds, s ∈ R for all t ∈ [s, s + 1].
s

Therefore, we have

max v(xλn (t)) ≤ min v(xλn (t)) + kβk1 . (5.4.7)

t∈[0,1] t∈[0,1]

Now, from kxλn k0 ≥ n and (5.4.7), we deduce mint∈[0,1] v(xλn (t)) → ∞, which
implies that mint∈[0,1] |xλn (t)| → ∞. Thus there exists N > 0 such that, for
n ≥ N,
min |xλn (t)| ≥ r. (5.4.8)
t∈[0,1]

Now, we have
d
w(xλn (t)) = −(1 − λn )(v 0 (xλn (t)), w0 (xλn (t)))
dt
+λn (w0 (xλn (t)), f (t, xλn (t))).

From (5.4.9) and the assumption (2), one deduces that

Z 1
d
0= w(xλn (t))dt < 0,
0 dt

which is a contradiction. Thus the claim is true. Choose r1 > max{r, r0 }. By

Lemma 5.4.2, we have

|DJ (L − V, B(0, r) ∩ D(L), 0)| = |DJ (L − QV, B(0, r) ∩ D(L), 0)|.

By Proposition 5.2.4 and Theorem 1.2.15, we get

|DJ (L − V, B(0, r) ∩ D(L), 0)| = 1.

Thus the problem (E 5.4.4) has a solution.

Copyright 2006 by Taylor & Francis Group, LLC

 124 Topological Degree Theory and Applications

5.5 Exercises
1. Let X be a Banach space, T : X → X be a linear bounded Fredholm
operator and K : X → X be a linear continuous compact mapping.
Show that T + K is a Fredholm mapping.

2. Assume that the conditions of Exercise 1 hold and Ind(T ) = 0. Show

that
Ind(T + K) = 0.

3. Let L : D(L) ⊂ C([0, 1], Rn ) → L1 ([0, 1]; Rn ) × Rn be defined by

Lx(t) = (x0 (t), M x(0) + N x(1)) for all x(·) ∈ D(L),

where M, N are n×n real matrices. Show that L is a Fredholm mapping

of index zero.

4. Let L be defined as in Exercise 3. Construct the projections P, Q such

that Im(P ) = Ker(L) and Im(L) = Ker(Q).

5. Let X, Y be real normed spaces, L : D(L) ⊂ X → Y be a Fredholm

mapping of index zero and A : X → Y be a linear mapping such that A is
L-compact on any bounded subset of A. Assume that Ker(L−A) = {0}.
Show that
(L − A)(D(L)) = Y,
(L − A)−1 = (I − P − JQA − KP Q A)−1 (JQA + KP Q A).

6. Let T : Ω → Y be L-compact. Assume that L − T is one to one on

D(L) ∩ Ω. Show that, for all z ∈ (L − T )(D(L) ∩ Ω),

|DJ (L − T − z, D(L) ∩ Ω, )| = 1.

7. Let T : Ω → Y be L-compact. Suppose the following conditions hold:

(1) kLx − T xk2 ≥ kT xk2 − kLxk2 for all x ∈ (D(L) \ Ker(L)) ∩ ∂Ω;
(2) T x ∈
/ Im(L) = 0 for all x ∈ Ker(L) ∩ ∂Ω;
(3) deg(QTKer(L) , Ω ∩ Ker(L), 0) 6= 0, where Q : Y → Y is the projec-
tion such that Ker(Q) = Im(L).

Show that Lx = T x has a solution in D(L) ∩ Ω.

8. Let H be a Hilbert space, T : Ω → H L-compact. Suppose the following

conditions hold:

(1) (Lx − T x, Lx) ≥ 0 for all x ∈ (D(L) \ Ker(L)) ∩ ∂Ω;

Copyright 2006 by Taylor & Francis Group, LLC

 COINCIDENCE DEGREE THEORY 125

(2) T x ∈
/ Im(L) = 0 for all x ∈ Ker(L) ∩ ∂Ω;
(3) deg(QTKer(L) , Ω ∩ Ker(L), 0) 6= 0, where Q : H → H is the pro-
jection such that Ker(Q) = Im(L).
Show that Lx = T x has a solution in D(L) ∩ Ω.
9. Let f : [0, π] × Rn × Rn → Rn be a function satisfying the Carathéodory
condition. Assume that the following condtions hold:
(1) There exist a, b ∈ R such that a + b < 1 and
(x − u, f (t, x, y) − f (t, u, v)) ≤ a|x − u|2 + b|x − u||y − u|
for all x, y, u, v ∈ Rn and a.e. t ∈ [0, π].
(2) There exist c ≥ 0 and h ∈ L1 ([0, π], [0, +∞)] such that, for all
x ∈ Rn with |x| ≤ π(1 − a − b)−1 kf (t, 0, 0)kL1 ,
|f (t, x, y)| ≤ c|y|2 + h(t) for all y ∈ Rn , and almost all t ∈ [0, π].

Show that the following equation:

(
−x00 (t) = f (t, x(t), x0 (t)), t ∈ [0, π],
x(0) = x(π) = 0

has a unique solution.

10. Let f : [0, π] × Rn × Rn → Rn be a function satisfying the Carathéodory
condition. Assume that there exist a, b ∈ R such that a + b < 1 and
|f (t, x, y) − f (t, u, v)| ≤ a|x − u| + b|y − u|
for all x, y, u, v ∈ Rn and almost all t ∈ [0, π]. Show that the following
equation: (
−x00 (t) = f (t, x(t), x0 (t)), t ∈ [0, π],
x(0) = x(π) = 0
has a unique solution.
11. Let f : [0, π]×Rn → Rn be a function satisfying the Carathéodory condi-
tion. Assume that there exist a number a < 1 and g ∈ L1 ([0, π], [0, +∞)])
such that
(x, f (t, x)) ≤ a|x|2 + g(t)|x| for all x ∈ Rn , and alomst allt ∈ [0, π].
Show that the following equation:
(
−x00 (t) = f (t, x(t)), t ∈ [0, π],
x(0) = x(π) = 0

has a solution.

Copyright 2006 by Taylor & Francis Group, LLC

 126 Topological Degree Theory and Applications

12. Let f : [0, 1] × Rn × Rn → Rn be a function satisfying the Carathéodory

condition and v ∈ C 1 (Rn , R). Suppose that the following conditions are
satisfied:
(1) There exists r > 0 such that (v 0 (x), f (t, x)) ≤ 0 for all x ∈ Rn with
|x| ≥ r and almost all t ∈ [0, 1];
(2) v 0 (x) 6= 0 for all x ∈ Rn with |x| ≥ r and lim|x|→∞ v(x) = +∞.
Show that the following equation:
(
x0 (t) = f (t, x(t)), t ∈ [0, 1],
x(0) = x(1)

has a solution.
13. Let L : D(L) ⊂ X → Y be a Fredholm mapping of index zero, Ω ⊂
X be an open bounded subset, D(L) ∩ Ω 6= ∅ and T : Ω → 2Y be
a mapping with closed convex values. Assume that QT and KP Q T
are upper semicontinuous mapping such that QT (Ω) and KP Q T (Ω) are
relatively compact and Lx ∈/ T x for all x ∈ ∂Ω ∩ D(L). Construct the
coincidence degree for L and T on Ω ∩ D(L).
14. Let L : D(L) ⊂ X → Y be a Fredholm mapping of index zero, Ω ⊂ X
be an open bounded subset, D(L) ∩ Ω 6= ∅ and T : Ω → Y be a mapping
such that QT and KP Q T are continuous countably condensing mapping
and Lx 6= T x for all x ∈ ∂Ω ∩ D(L). Construct the coincidence degree
for L and T on Ω ∩ D(L).

Copyright 2006 by Taylor & Francis Group, LLC

 Chapter 6
DEGREE THEORY FOR
MONOTONE-TYPE MAPS

Monotone-type mappings are a class of mappings without continuous and

compact conditions. The concept of monotone mapping was introduced by
Kachurovski, Vainberg, Zarantonello in 1960, and it plays a very impor-
tant role in studying the weak solution of the partial differential equations
in divergence form and variational inequality problems. Minty, Browder,
Brézis, Rockafellar, Crandall, Gossez, etc., made significant contributions to
monotone operator theory. It was shown by Skrypnik, Browder, Berkovitz,
Mustonen, Kartsatos, and others that it is possible to construct the degree
theory for monotone-type mappings.

The goal of this chapter is to introduce the degree theory for monotone-type
mapping. Chapter 6 has seven sections.
In Section 6.1, we introduce some basic geometric properties of Banach
spaces and various types of monotone and pseudomonotone maps and also
(S+ ), (S+ )0,L and L-(S+ )-mappings. Many examples and properties of these
maps are presented in Section 6.1.
Section 6.2 presents the degree theory for monotone mappings of class (S+ ).
In Section 6.3, using the results of Section 6.2, we present the degree theory
for perturbations of maximal monotone mappings and various properties are
also presented.
In Section 6.4, using the results of chapters 2, 3, we present the topological
degree for multivalued mappings of class (S+ )0,L . Some properties of this
degree are presented in theorems 6.4.4, 6.4.5, and 6.4.6.
A degree for multivalued mappings of class L-(S+ ) type is presented in
Section 6.5 (here L is a Fredholm mapping of index zero type). Various
properties are presented in Theorems 6.5.2, 6.5.3 and 6.5.5. The coincidence
degree of L and a pseudomonotone mapping is also presented in this section.
Section 6.6 presents various results concerning the computation of the topo-
logical degree for a variety of mappings.
Section 6.7 gives various existence results for the partial differential equa-
tions and evolution equations.

127

Copyright 2006 by Taylor & Francis Group, LLC

 128 Topological Degree Theory and Applications

6.1 Monotone Type-Mappings in Reflexive Banach Spaces

In this section, we introduce some monotone-type mappings and discuss
their properties. We first recall some geometrical properties of Banach spaces.
Definition 6.1.1. Let X be a normed space, X ∗∗ be the second dual space
of X and K : X → X ∗∗ be defined by Kx(f ) = f (x). If KX = X ∗∗ , then X
is said to be reflexive.
Theorem 6.1.2. Let X be a normed space and X ∗ be the dual space of
X. Then the following are equivalent:
(1) X is reflexive;
(2) X ∗ is reflexive;
(3) The closed unit ball of X is sequentially weak compact, i.e., each (xn )∞
n=1
in the closed unit ball has a weakly convergent sequence;
(4) For all f ∈ X ∗ \ {0}, there exists x ∈ X with kxk = 1 such that
f (x) = kf k.

Definition 6.1.3. Let X be a Banach space. X is said to be strictly convex

if, for any x, y ∈ X, kxk = kyk = 1 and kx + yk = 2 imply that x = y.
The following proposition follows directly from Definition 6.1.3.
Proposition 6.1.4. Let X be a Banach space. The following statements
are equivalent:
(1) X is strictly convex;
(2) If, for any x, y ∈ X, kxk = kyk = 1 and x 6= y, then kx + yk < 1;
(3) Every point on the unit sphere is an extreme point;
(4) If f ∈ X ∗ is nonzero and kxk = kyk = 1 such that f (x) = f (y) = kf k,
then x = y.

Definition 6.1.5. A Banach space X is said to be locally uniform convex

if, for any x ∈ X with kxk = 1 and  ∈ (0, 2], there exists δ(x) > 0 such that,
for any y ∈ X with ky − xk ≥ , we have kx + yk < 2 − δ(x).
Definition 6.1.6. A Banach space X is said to be uniformly convex if, for
any  ∈ (0, 2], there exists δ > 0 such that, for any x, y ∈ X with ky − xk ≥ ,
we have kx + yk < 2 − δ.
Some well-known uniformly convex spaces are Hilbert spaces, lp , Lp , and
the Sobolev space W m,p , where p > 1 and m > 0 is an integer.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 129

Proposition 6.1.7. A uniformly convex Banach space is reflexive.

In the sequel, let E be a real reflexive Banach space and E ∗ be the dual
space of E, and * represents the weak convergence. By [287], we may assume
that both E and E ∗ are locally uniformly convex. Let J : E → E ∗∗ be the
duality mapping, i.e.,
kJxk = kxk, (Jx, x) = kxk2 .

Proposition 6.1.8. Let E be a real reflexive Banach space and J : E →

E ∗∗ be the duality mapping. Then we have the following:
(1) If E ∗ is locally uniform convex, then J is single valued and continous;
(2) If E is locally uniform convex, (xn )∞
n=1 ⊂ E is a sequence converging
weakly to x0 as n → ∞, kxn k → kx0 k and Jxn (x0 ) → kx0 k2 , then
xn → x0 .

Proof. (1) For all x 6= 0, if f1 , f2 ∈ Jx, then we have (f1 + f2 )(x) = 2kxk2
and thus kf1 + f2 k = 2kxk. In addition, kf1 k = kf2 k = kxk, and so it follows
from the locally uniform convexity of E ∗ that f1 = f2 .
Next, assume that xn → x0 in E and we may assume also that (Jxn )) has
a weakly convergent sequence (Jxnk ) with Jxnk * f0 by reflexivity of E ∗ .
Then we have
Jxnk (xnk ) → kx0 k2 , Jxnk (x0 ) → f0 (x0 ).
In addition, we have
kJ(xnk )(xnk ) − J(xnk )(x0 )k = 0.
Therefore, we get f0 (x0 ) = kx0 k2 and, from this, we deduce that
kf0 k = kx0 k, kJ(xnk )k → kx0 k, kJ(xnk ) + f0 k → kx0 k = kf0 k.
∗
From the locally uniform convexity of E , we deduce J(xnk ) → Jx0 .
(2) For simplicity, we may assume that Jxn * f0 by taking a subsequence
since E ∗ is reflexive. By assumption, we have
Jxn (xn + x0 ) = kxn k2 + Jxn (x0 ) → 2kx0 k2
and so kxn + x0 k → kx0 k. From the locally uniformly convexity of E, we
deduce xn → x0 . This completes the proof.
The following is the well-known Mazur’s separation theorem for convex sets:
Theorem 6.1.9. Let X be a Banach space, C1 be an compact convex set
and C2 be a closed convex set such that C1 ∩ C2 = ∅. Then there exists a
f ∈ X ∗ \ {0} such that
sup f (x) < inf f (x).
x∈C1 x∈C2

Copyright 2006 by Taylor & Francis Group, LLC

 130 Topological Degree Theory and Applications

Next, we recall some terminology as follows:

Definition 6.1.10. Let E be a real reflexive Banach space and E ∗ be the
dual space of E.
∗
(1) An operator T : D(T ) ⊆ E → 2E is said to be monotone if

(f − g, x − y) ≥ 0

for all x, y ∈ D(T ), f ∈ T x and g ∈ T y;

(2) T is said to be maximal monotone if T is monotone and does not have
a proper monotone extension.

Note that when E is a Hilbert space the mapping T is said to be pseudo-

contractive if I − T is maximal monotone (see [51]).
As a consequence of Zorn’s Lemma, every monotone mapping has a maximal
monotone extension. We leave the details to the reader.
Definition 6.1.11. Let E be a reflexive Banach space. A multi-valued
∗
operator T : D(T ) ⊆ E → 2E is said to be a mapping of class (S+ ) if it
satisfies the following conditions:
(1) T x is bounded closed and convex for each x ∈ D(T );
(2) T is weakly upper semicontinuous in each finite dimensional space, i.e.,
for each finite dimensional space F with F ∩ D(T ) 6= ∅, T : F ∩ D(T ) →
∗
2E is upper semicontinuous in the weak topology;
(3) if {xn } ⊂ D(T ) and {xn } converges weakly to x0 in E such that

lim sup(fn , xn − x0 ) ≤ 0 for some fn ∈ T xn ,

n→∞

then xn → x0 ∈ D(T ) and {fn } has a subsequence which converges

weakly to f0 ∈ T x0 in E ∗ .

When E is a Hilbert space, we say that T is pseudocompact if I − T is a

mapping of class (S+ ) (see [51]). One may easily see that a compact mapping
is pseudocompact.
Definition 6.1.12. Let E be a reflexive Banach space. A family of oper-
∗
ators {Tt : D(Tt ) ⊆ E → 2E }t∈[0,1] is said to be a homotopy of mappings of
class (S+ ) if Tt satisfies the conditions (1), (2) in Definition 6.1.11 for each
t ∈ [0, T ] and the following condition:
(3)’ If tn → t0 and xn ∈ D(Ttn ), {xn } converges weakly to x0 in E such that

lim sup(fn , xn − x0 ) ≤ 0 for some fn ∈ Ttn xn ,

n→∞

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 131

then xn → x0 ∈ D(Tt0 ) and {fn } has a subsequence converging weakly

to some f0 ∈ Tt0 x0 in E ∗ .
∗
Definition 6.1.13. Let T : D(T ) ⊆ E → 2E be a mapping satisfying the
conditions (1), (2) in Definition 6.1.11. Let (xj ) ⊂ D(T ), xj * x0 ∈ D(T )
and fj ∈ T xj .
(1) If lim supj→∞ (fj , xj − x0 ) ≤ 0 implies that

(f0 , x0 − v) ≤ lim inf (fj , xj − v) for all v ∈ D(T ), f0 ∈ T x0 ,

j→∞

then T is said to be a pseudomonotone mapping;

Let (xj ) ⊂ D(T ), xj * x0 ∈ D(T ), fj ∈ T xj and fj * f0 .
(2) If lim supj→∞ (fj , xj − x0 ) ≤ 0 implies that

f0 ∈ T x0 , (f0 , x0 ) = lim (fj , xj ),

j→∞

then T is said to be a generalized pseudomonotone mapping.

Lemma 6.1.14. If J : E → E ∗ is the duality mapping, then J is a

continuous mapping of class (S+ ) and J is also monotone.
Proof. Let xn → x0 . We may assume that Jxn * f0 ∈ E ∗ . Since
limn→∞ (Jxn , xn − x0 ) = 0, we have (f0 , x0 ) = kx0 k2 . Therefore, we get
kJxn k → kx0 k = kf0 k. The local uniform convexity of E ∗ implies that
Jxn → Jx0 , and so J is continuous and, furthermore, J monotone is obvious.
Next, we prove that J is a mapping of class (S+ ). Let

xn * x0 , lim sup(Jxn , xn − x0 ) ≤ 0.
n→∞

We may assume that Jxn * f0 ∈ E ∗ . Since

kx0 k2 ≤ lim inf (Jxn , xn ) ≤ (f0 , x0 ),

n→∞

we have kxn k → kx0 k, kf0 k = kx0 k. Thus by Proposition 6.1.8, xn → x0 and

Jxn → Jx0 . This completes the proof.
Definition 6.1.15 [163] Let E be a separable reflexive Banach space and
L be a dense subspace of E. A mapping T : D(T ) ⊂ E → E ∗ is said to be
a mapping of class (S+ )L if, for any sequence of finite dimensional subspaces
∗ ∞
Fj of L with ∪∞
j=1 Fj = E, h ∈ E , {uj }j=1 ⊂ D(T ) with uj * u0 and

lim sup(T uj − h, uj ) ≤ 0, lim (T uj − h, v) = 0

j→∞ j→∞

for all v ∈ ∪∞
j=1 Fj , we have uj → u0 , and u0 ∈ D(T ), T u0 = h. If h = 0, then
we call T a mapping of class (S+ )0,L .

Copyright 2006 by Taylor & Francis Group, LLC

 132 Topological Degree Theory and Applications

Definition 6.1.16. Let E be a reflexive Banach space and L be a subspace

∗
of E. A multi-valued mapping T : D(T ) ⊆ E → 2E is said to be a mapping
of class (S+ )L if it satisfies the following conditions:

(i) T x is bounded closed and convex for each x ∈ D(T );

(ii) T is weakly upper semicontinuous in each finite dimensional space, i.e.,

for each finite dimensional space F of L with F ∩ D(T ) 6= ∅, T : F ∩
∗
D(T ) → 2E is upper semicontinuous in the weak topology;

(iii) if, for any sequence of finite dimensional subspaces Fj of L with L ⊆

∗ ∞
∪∞j=1 Fj , h ∈ E , {xj }j=1 ⊂ D(T ) ∩ L and xj * x0 such that

lim sup(fj − h, xj ) ≤ 0, lim (fj − h, v) = 0

j→∞ j→∞

for all v ∈ ∪∞ j=1 Fj and some fj ∈ T xj , then xj → x0 ∈ D(T ) and

h ∈ T x0 . If h = 0, then we call T a mapping of class (S+ )0,L .

Remark. If T is a mapping of class (S+ )L , then, for any p ∈ E ∗ , T − p is

a mapping of class (S+ )0,L .
When E is a Hilbert space, we say that T is L-pseudocompact if I − T is a
mapping of class (S+ )L , and T is L0 -pseudocompact if I − T is a mapping of
class (S+ )0,L .
Definition 6.1.17. Let E be a reflexive Banach space and let L be a
∗
subspace of E. A family of mappings {Tt : D(Tt ) ⊆ E → 2E }t∈[0,1] is
called a homotopy of mappings of class (S+ )L if the conditions (i) and (ii) in
Definition 6.1.16 hold for each t ∈ [0, T ] and the following condition holds:

(iii) If, for any sequence of finite dimensional subspaces Fj of L with L ⊆

∗
∪∞j=1 Fj , h ∈ E , tj → t0 , xj ∈ D(Ttj ) ∩ L and xj * x0 such that

lim sup(fj − h, xj ) ≤ 0, lim (fj − h, v) = 0,

j→∞ j→∞

for all v ∈ ∪∞ j=1 Fj and some fj ∈ Ttj xj , then xj → x0 ∈ D(Tt0 ),

h ∈ Tt0 x0 . If h = 0, then we call {Tt : D(Tt )}t∈[0,1] a homotopy of
mappings of class (S+ )0,L .

Proposition 6.1.18. If L is dense in E, then the duality mapping J is a

mapping of class (S+ )L .
Proof. Let (Fj )∞
j=1 be a sequence of finite dimensional subspaces of L with
∪∞ = E, h ∈ E ∗ , xj ∈ L and xj * x0 such that
j=1 Fj

lim suphJxj − h, xj i ≤ 0, lim (Jxj − h, v) = 0

j→∞ j→∞

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 133

for all v ∈ ∪∞
j=1 Fj . Without loss of generality, we may assume that Jxj * f0
in E ∗ . Then we have

kx0 k2 ≤ lim sup(Jxj , xj ) ≤ (h, x0 ), (f0 − h, v) = 0

j→∞

for all v ∈ ∪∞ ∞
j=1 Fj . But ∪j=1 Fj is dense in E, so we have f0 = h. Therefore
it follows that

kx0 k2 ≤ lim sup(Jxj , xj ) ≤ (h, x0 ) ≤ kx0 k2 .

j→∞

Thus kxj k → kx0 k as j → ∞. This and the local uniform convexity of E

together with xj * x0 imply that xj → x0 and Jx0 = h. Therefore, J is a
mapping of class (S+ )L .
Now, assume that L is a Fredholm mapping of index zero type. Then there
exist two linear continuous projections P : H → H and Q : H → H such that

Im(P ) = Ker(L), Ker(Q) = Im(L).

Also, we have

H = Ker(L) ⊕ Ker(P ), H = Im(L) ⊕ Im(Q)

as the topological direct sums.

Obviously, the restriction of LP of L to D(L) ∩ Ker(P ) is one to one and
onto Im(L), so its inverse KP : Im(L) → D(L) ∩ Ker(P ) is defined. We
denote by KP Q : H → D(L) ∩ Ker(P ) the generalized inverse of L defined by
KP Q = KP (I − Q). Let J : Im(Q) → Ker(L) be a linear homeomorphism.
Definition 6.1.19. Let H be real Hilbert space, L : D(L) ⊆ H → H
be a Fredholm mapping of index zero type and T : D(T ) ⊆ H → 2H be a
set-valued mapping.

(1) If I − P − (JQ + KP Q )T is maximal monotone mapping, then T is said

to be L-maximal monotone;

(2) if I − P − (JQ + KP Q )T is pseudomonotone (respectively, generalized

pseudo monotone), then T is L-pseudomonotone (respectively, general-
ized L-pseudomonotone). If I − P − (JQ + KP Q )T is also bounded, then
T is said to be bounded L-pseudomonotone;

(3) If I − P − (JQ + KP Q )T is a mapping of class (S+ ), then T is called a

mapping of class L-(S+ ).

Remark. (1) If T is a mapping of class (S+ ), L = 0, P = I, Q = I, and

J = −I, then T is a mapping of class 0-(S+ ).

Copyright 2006 by Taylor & Francis Group, LLC

 134 Topological Degree Theory and Applications

(2) If I−T is a mapping of class (S+ ), in this case, T is called pseudocompact

in [51], and L = I, P = 0, Q = 0, J = I, then T is a mapping of class I-(S+ ).
As a special consequence, if T is upper semicontinuous compact mapping with
bounded closed convex values, then I − T is a mapping of class (S+ ), so T is
a mapping of class I-(S+ ).
(3) One may also easily see that, if T is L-compact, then T is a mapping of
class L-(S+ ).
Proposition 6.1.20. Let H be real Hilbert space, L : D(L) ⊂ H → H be
a generalized Fredholm mapping of index zero type and Ti : D(Ti ) ⊆ H → 2H
be a mapping of class L-(S+ ) for i = 1, 2. Then tT1 + (1 − t)T2 is a mapping
of class L-(S+ ) on D(T1 ) ∩ D(T2 ) for all t ∈ [0, T ].
Now, we give some examples of monotone-type mappings.
Example 6.1.21. Let Ω ⊂ Rn be an open bounded subset with smooth
boundary. Assume that ai , bi : R → [0, +∞) are continuous functions for
i = 1, 2, · · · , n. Suppose the following conditions are satisfied:
(1) c1 ≤ ai (x) ≤ c2 for all (t, x) ∈ R2 , where c1 , c2 > 0 are constants;
(2) Σi [bi (xi ) − bi (yi )](xi − yi ) ≥ 0, where x = (xi ), y = (yi ) ∈ RN ;
(3) |bi (x)| ≤ β|x| + γ for all x ∈ R and i = 1, 2, · · · , n;
(4) Σi bi (xi )xi ≥ α|x|2 − c0 for all x = (xi ) ∈ Rn .
Let A : L2 (Ω) × H01 (Ω) → H ∗ be defined as follows:
Z
(A(u, v), w) = [Σni=1 ai (u)bi (Di v)Di w]dx
Ω

for all u, w ∈ H01 (Ω). Then we have the following:

(a) for each v ∈ L2 (Ω),
sZ sZ
(A(v, u), w) ≤ c2 β (Σni=1 |Di u|2 )dx (Σni=1 |Di w|2 )dx
Ω Ω

for all u, w ∈ H01 (Ω),

(b) (A(v, u1 ) − A(v, u2 ), u1 − u2 ) ≥ 0 for t ∈ R, u1 , u2 , ∈ H01 (Ω) and v ∈
L2 (Ω) and A(v, ·) is continuous and monotone for all v ∈ R × L2 (Ω) and
so it is also pseudomonotone.
A special case of Example 6.1.21 is that, if ai (x) = 1 and bi (x) = x for all
x ∈ R, then we get the Laplace operator
Z
(−∆u, v) = (∇u, ∇v)dx for all u, v ∈ H01 (Ω).
Ω

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 135

Example 6.1.22. Let Ω be as in Example 6.1.21 and let f (x, y) : R2 → R

be a continuous function satisfying

|f (x, y)| ≤ M |y| + g(x) for all (x, y) ∈ R2 ,

where M > 0 is a constant and g(·) ∈ L2 (Ω). Then the following mapping A
defined by
Z
(Au, v) = [(∇u, ∇v) + f (x, u)v]dx for all u, v ∈ H01 (Ω)
Ω

is a mapping of class (S+ ) and it is also pseudomonotone.

Example 6.1.23. Let Ω be as in Example 6.1.17, p > 1, and aα : Ω×Rn →
R be a Caratheodory function, i.e., aα (x, ξ) is measurable in x and continuous
in ξ, where α = (α1 , α2 , · · · , αN ) is a multi-index. Assume that

|aα (x, ξ)| ≤ C|ξ|p−1 + g(x)

for almost everywhere x ∈ Ω, where ξ = (ξα : |α| ≤ m}, C > 0 is a constant,

g(·) ∈ Lq and p−1 + q −1 = 1.
We consider the following partial differential equation:
(
Σ|α|≤m (−1)|α| Dα aα (x, u, Du, · · · , Dm u) + f (x, u) = 0, x ∈ Ω,
u(x) = 0, x ∈ ∂Ω,

and we define a mapping A : H0m (Ω) → (H0m (Ω)∗ by

Z
(Au, v) = [Σ|α|≤m aα (x, u, Du, · · · , Dm u)Dα v + f (x, u)v]dx
Ω

for all u, v ∈ H0m (Ω).

By imposing suitable conditions on aα , the above
mapping A will be maximal monotone, pseudomonotone or of class (S+ ).
Example 6.1.24. Consider the following wave equation with discontinuity:


 utt (t, x) − uxx (t, x) + g(u(t, x)) + h(u(t, x)) = f (t, x),

t ∈ (0, 2π), x ∈ (0, π),


 u(t, 0) = u(t, π) = 0, t ∈ (0, 2π),
u(0, x) = u(2π, x), x ∈ (0, π),


where g : R → R is a nondecreasing function with

|g(u)| ≤ α|u| + β for all u ∈ R,

h : R → R is a continuous function satisfying

|h(u)| ≤ δ|u| + γ

Copyright 2006 by Taylor & Francis Group, LLC

 136 Topological Degree Theory and Applications

and f (·) ∈ L2 ((0, 2π) × (0, π)), where α > 0, β > 0, δ > 0, γ > 0 are constants.
Set
g− (u) = lim inf g(s), g+ (u) = lim sup g(s),
s→u s→u

Gu = [g− (u), g+ (u)].

R
Then G : R → 2 is a maximal monotone mapping. Let a mapping N :
L2 ((0, 2π) × (0, π)) → L2 ((0, 2π) × (0, π)) be defined by

N (u(t, x)) = {v(t, x) ∈ L2 ((0, 2π) × (0, π)) : v(t, x) ∈ Gu(t, x)}

for all u(t, x) ∈ L2 ((0, 2π) × (0, π)).

From our assumption on g, we know that D(N ) = L2 ((0, 2π) × (0, π)) and
N is also maximal monotone, so N is upper semicontinuous from the strong
topology of L2 ((0, 2π) × (0, π)) to the weak topology of L2 ((0, 2π) × (0, π)).
Let L : D(L) ⊂ L2 ((0, 2π) × (0, π)) → L2 ((0, 2π) × (0, π)) be the wave
operator Lu = utt − uxx . Then it is well known that L is self-adjoint, densely
defined and closed and Ker(L) is infinite dimensional with Ker(L)⊥ = Im(L).
Thus L is a Fredholm mapping of zero index type and also the right inverse
of L denoted by L−1 : Im(L) → Im(L) is compact.
Let P : L2 ((0, 2π) × (0, π) → Ker(L) be a projection. We assume that
lim inf i→∞ (P (fj + h(uj )), uj − u0 ) ≥ 0 for any uj * u0 in L2 ((0, 2π) × (0, π))
and fj ∈ N uj (for example, take g(x) = x for x < 0, g(x) = x + 1 for x ≥ 0
and h(x) = xsinx for x ∈ R). Then N (u) + h(u) is L-pseudomonotone. In
fact, if uj * u0 in L2 ((0, 2π) × (0, π)), fj ∈ N uj with fj * f0 , h(uj ) * h0
and

lim sup((I − P )uj + P (fj + h(uj )) + L−1 (I − P )(fj + h(uj )), uj − u0 ) ≤ 0,

j→∞

then, since
lim sup(P (fj + h(uj )), uj − u0 ) ≥ 0
j→∞

and L−1 is compact, we have

lim sup(L−1 (I − P )(fj + h(uj )), uj − u0 ) = 0

j→∞

and so lim supj→∞ ((I − P )uj , uj − u0 ) ≤ 0. Thus (I − P )uj → (I − P )u0 .

Consequently, we get

lim ((I − P )uj + P (fj + h(uj )) + L−1 (I − P )(fj + h(uj )), uj )

j→∞

= ((I − P )u0 + P (f0 + h0 ) + L−1 (I − P )(f0 + h0 ), u0 ).

Thus N + h is L-pseudomonotone.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 137
∗
Lemma 6.1.25. Let X be a real Banach space and T : D(T ) ⊂ E → 2E
be a monotone mapping. Then T is locally bounded on the interior of D(T ).
Proof. Let x0 ∈ int(D(T )). Without loss of generality, we assume that
x0 = 0. There exists r > 0 such that Br (0) ⊂ D(T ). By the monotonicity of
T , we have
(f − f0 , x − z) ≥ 0
where f0 ∈ T z is given and x ∈ D(T ), f ∈ T x. For each n ≥ 1, let

Mn = {z ∈ Br (0) : (f, x − z) ≥ −n, x ∈ Br (0), f ∈ T x}.

Then we have
Br (0) = ∪∞
n=1 Mn .

By the Baire’s category theorem, there exists n0 such that Mn0 has nonempty
interior and so there exist z0 ∈ Br (0), r0 > 0 such that Br0 (z0 ) ⊂ Mn0 . Since
−z0 ∈ Br (0), there exists m > 0 such that

(f, x + z0 ) ≥ m for all x ∈ Br (0), f ∈ T x.

Therefore, we have

(f, 2x − y) ≥ −(n0 + m0 ) for all y ∈ Br0 (0).

Now, for all x ∈ B r4 (0), f ∈ T x and z ∈ B r20 (0), we have

(f, z) = (f, 2x − (2x − z)) ≥ −(n0 + m0 )

and so
2
kf k ≤ (n0 + m0 ).
r
This completes the proof.
∗
Proposition 6.1.26. Let P : E → 2E be a bounded pseudomonotone
mapping and {xn } ⊂ E be such that xn * x0 . If fn ∈ P xn such that fn * f0
and lim supn→∞ (fn , xn − x0 ) ≤ 0, then f0 ∈ P x0 and (fn , xn ) → (f0 , x0 ).
Proof. Since lim supn→∞ (fn , xn − x0 ) ≤ 0, we have

(f, x0 − v) ≤ lim inf (fn , xn − v) for all v ∈ E, f ∈ P x0 . (6.1.1)

n→∞

Putting v = x0 in (6.1.1), it follows that

lim inf (fn , xn − x0 ) ≥ 0

n→∞

and so we have (fn , xn ) → (f0 , x0 ). Now, (6.1.1) becomes

(f, x0 − v) ≤ (f0 , x0 − v) for all v ∈ E, f ∈ P x0 . (6.1.2)

Copyright 2006 by Taylor & Francis Group, LLC

 138 Topological Degree Theory and Applications

But since P x0 is bounded, closed and convex, by Mazur’s separation theorem

of convex subsets, we get f0 ∈ P x0 . This completes the proof.
Lemma 6.1.27. Let E be a real reflexive Banach space, T : D(T ) ⊂ E →
∗
2E be a maximal monotone mapping and let P : E → E ∗ be a bounded,
coercive and demicontinuous pseudomonotone mapping. Then there exists
x0 ∈ C such that

(f + P x0 , x − x0 ) ≥ 0 for all x ∈ D(T ), f ∈ T x.

Proof. We first prove that, for any finite dimensional subspace F of E

such that F ∩ D(T ) 6= ∅, there exists xF ∈ F such that

(f + P xF , x − xF ) ≥ 0 for all x ∈ F ∩ D(T ), f ∈ T x. (6.1.3)

Since P is coercive, we notice that if the above conclusion is true, xF must

be in a bounded ball B(0, R). Therefore, we may first prove that the above
conclusion is true in the case that D(T ) is bounded. For the unbounded case,
we find xn ∈ D(T ) such that

(f + P xn , x − xn ) ≥ 0

for all x ∈ F ∩ D(T ) ∩ B(0, n) and f ∈ T x. But since {xn } is bounded, we

may assume that xn → x0 as n → ∞ and then we can use the demicontinuity
of P to conclude that x0 is the desired point.
In the following, we assume that D(T ) is bounded and suppose that the
conclusion is not true. Then, for any x ∈ F ∩ D(T ), there exist z ∈ F ∩ D(T )
and fz ∈ T z such that
(fz + P x, z − x) < 0. (6.1.4)
Now, if we take a compact set C such that F ∩ D(T ) ⊂ C, then C =
∪z∈D(T ) Uz , where Uz = {x ∈ F : (fz + P x, z − x) < 0 for some fz ∈ T z}.
Notice that each Uz is open and so there exist finite many zi ∈ F ∩ D(T ) for
i = 1, 2, · · · , N such that
C = ∪Ni=1 Uzi .

Let {α1 , α2 , · · · , αN } be a partition of unity subordinated to the covering

{Uzi }. Then we define a mapping K : C → C by

Kx = ΣN
i=1 αi (x)zi for all x ∈ C.

Then K has a fixed point x0 ∈ C, i.e., x0 = ΣN i=1 αi (x0 )zi .

We may assume that αi (x0 ) > 0 for i = 1, 2, · · · , N . Otherwise, we exclude
it. Now, we have

0 = (Kx0 − x0 , ΣN i=1 αi (x)fzi + P x0 )

= Σni,j=1 αi (x0 )αj (x0 )(fzj + P x0 , zj − x0 )
< 0,

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 139

which is a contradiction. Thus (6.1.3) is true.

Finally, for each finite dimensional subspace F of E, we put

WF = {(xF , GxF ) : (xF , GxF ) satisfies (6.1.3)}.

Obviously, we have
∩ni=1 WFi 6= ∅.
∗
If we denote by WF the weak closure of WF in E, then we have
∗
∩F ⊂E,dimF <∞ WF 6= ∅.
∗
Take (x0 , g0 ) ∈ ∩F ⊂E,dimF <∞ WF and, for any x ∈ D(T ) ∩ C, a finite dimen-
sional subspace F of E such that x, x0 ∈ F . Then there exists {(xj , P xj )} ⊂
WF such that
xj * x0 , P xj * g0 as j → ∞.
Therefore, we have

(f + P xj , x − xj ) ≥ 0 for all x ∈ F ∩ D(T ), f ∈ T x.

Since P is pseudomonotone and C is closed and convex, it follows from Propo-

sition 6.1.21 that
lim sup(P xj , xj ) ≥ (g0 , x0 )
j→∞

and so we have

(f + g0 , x − x0 ) ≥ 0 for all x ∈ F ∩ D(T ), f ∈ T x.

But since F is arbitrary, (f + g0 , x − x0 ) ≥ 0 for all x ∈ D(T ) and f ∈ T x.

Hence x0 ∈ D(T ), which implies that

lim sup(P xj , xj − x0 ) ≤ 0.
j→∞

Thus we have (P xj , xj ) → (P x0 , x0 ) and so

(f + P x0 , x − x0 ) ≥ 0 for all x ∈ D(T ), f ∈ T x.

This completes the proof.

∗
Let T : D(T ) ⊆ E → 2E be a maximal monotone operator. Let Tλ =
(λJ −1 + T −1 )−1 denote the Yosida approximation and Rλ = I − λJ −1 Tλ the
resolvent with respect to Tλ , respectively.
As a direct consequence of Lemma 6.1.27, we get the following:
∗
Proposition 6.1.28. If T : D(T ) → 2E is a monotone mapping, then T
is maximal monotone if and only if T + J is surjective for any  > 0.

Copyright 2006 by Taylor & Francis Group, LLC

 140 Topological Degree Theory and Applications

Proof. We assume that both the spaces E and E ∗ are locally uniform
convex. If T is maximal monotone, then, for any p∗ ∈ E ∗ , J − p∗ is a
continuous bounded coercive monotone mapping and, thus, pseudomonotone.
By Lemma 6.1.27, there exists x0 ∈ E such that

(f + Jx0 − p∗ , x − x0 ) ≥ 0 for all x ∈ D(T ), f ∈ T x.

So x0 ∈ T x0 and −Jx0 + p∗ ∈ T x0 , i.e., p∗ ∈ T x0 + Jx0 .

Conversely, suppose that (f − g, x − x0 ) ≥ 0 for all x ∈ D(T ) and f ∈ T x.
Then there exist y0 ∈ D(T ) and f0 ∈ T y0 such that f0 + Jy0 = Jx0 + g.
Therefore, it follows that

(Jx0 − Jy0 , y0 − x0 ) ≥ 0.

So we must have x0 = y0 . This completes the proof.

Lemma 6.1.29. Let T : D(T ) ⊆ E → E ∗ be a maximal monotone map-
ping. If x ∈ D(T ), then

lim Rλ x = x, lim Tλ x = f,
λ→0+ λ→0+

where f ∈ T x and kf k = min{kgk, g ∈ T x}.

Proof. By the monotonicity of T , it follows that

kRλ xk2 ≤ −λ(Rλ x − y, g) − (x − y, Tλ x) (6.1.5)

for all x, y ∈ D(T ) and g ∈ T y. Put y = x in (6.1.5) and then we have

kTλ xk ≤ min{kgk : g ∈ T x} = kf k. (6.1.6)

By letting λ → 0+ , it immediately yields that Rλ x → x. Without loss of

generality, we may assume that Tλ x * f0 as λ → 0+ . Then the maximal
monotonicity of T yields that f0 ∈ T x, and so from (6.1.6) and the locally
uniform convexity of E ∗ we infer that limλ→0+ Tλ x = f0 and f0 = f . This
completes the proof.
Proposition 6.1.30. The following conclusions hold:

lim Tλ x = Tλ0 x, lim Rλ x = Rλ0 x

λ→λ0 λ→λ0

where λ0 > 0 and x ∈ D(T ).

Proof. Since

(J(Rλ x − x) − J(Rλ0 x − x), Rλ x − Rλ0 x)

λ − λ0
≤ (J(Rλ0 x − x), Rλ x − Rλ0 x)
λ0

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 141

and

(kRλ x − xk − kRλ0 x − xk)2 ≤ (J(Rλ x − x) − J(Rλ0 x) − x), Rλ x − Rλ0 x),

we have
lim kRλ x − xk = kRλ0 x − xk,
λ→0+

lim (J(Rλ0 x − x), Rλ x) − x) = kRλ0 x − xk2 .

λ→0+

Thus the locally uniform convexity of E and E ∗ imply that

lim Tλ x = Tλ0 x, lim Rλ x = Rλ0 x.

λ→λ0 λ→λ0

This completes the proof.

∗
Proposition 6.1.31. Let A : D(A) ⊆ E → 2E be a multi-valued mapping
of class (S+ ). Then we have the following:
∗
(1) If T : D(A) → 2E is an upper semicontinuous operator with closed
convex values and T maps each bounded subset of D(A) into a relatively
compact subset of E ∗ , then T + A is a multi-valued mapping of class
(S+ );
∗
(2) If M : D(M ) ⊆ E → 2E is a maximal monotone operator, then Mλ +A
is a multi-valued mapping of class (S+ );
∗
(3) If P : E → 2E is a pseudomonotone mapping, then P + A is a mapping
of class (S+ );
∗
(4) If P : E → 2E is a bounded generalized pseudomonotone mapping,
then P + A is a mapping of class (S+ ).

The proof of Proposition 6.1.31 is left to the reader as an exercise.

Proposition 6.1.32. If T1 and T2 are two bounded mappings of class (S+ ),
then {tT1 + (1 − t)T2 : t ∈ [0, 1]} is a homotopy of mappings of class (S+ ).
Proof. To show that tT1 + (1 − t)T2 satisfies the conditions (1) and (2) of
Definition 6.1.12 is trivial.
Now, suppose that tj → t0 , xj * x0 and fji ∈ Ti xj such that

lim sup(tj fj1 + (1 − tj )fj2 , xj − x0 ) ≤ 0. (6.1.7)

j→∞

Since T1 and T2 are mappings of class (S+ ), we have

lim inf(fj1 , xj − x0 ) ≥ 0, lim inf(fj2 , xj − x0 ) ≥ 0.

Copyright 2006 by Taylor & Francis Group, LLC

 142 Topological Degree Theory and Applications

By virtue of (6.1.7), we get limj→∞ (fj1 , xj − x0 ) = 0 or limj→∞ (fj2 , xj − x0 ) =

0. Hence we have xj → x0 and
lim (fj1 , xj − x0 ) = 0, lim (fj2 , xj − x0 ) = 0.
j→∞ j→∞

Therefore, {fj1 } and {fj2 } have subsequences {fj1k } and {fj2k } that converge
weakly to f 1 ∈ T1 x0 and f 2 ∈ T2 x0 , respectively. Therefore, we have
tjk fj1k + (1 − tjk )fj2k * t0 f 1 + (1 − t)f 2 ∈ t0 T1 x0 + (1 − t0 )T2 x0 .
This completes the proof.

6.2 Degree Theory for Mappings of Class (S+ )

In this section, we present a degree theory for multi-valued mappings of
class (S+ ). In the single valued case it was constructed by Browder [35], and
for a degree theory for other monotone-type mapping, see [51]. For some
notations, we refer to Section 6.1.
Lemma 6.2.1. Let F be a finite dimensional subspace, Ω ⊂ F be an open
∗
bounded subset and 0 ∈ Ω. Let T : Ω → 2F be an upper semicontinuous
mapping with compact convex values, F0 be a proper subspace of F , ΩF0 =
∗
Ω ∩ F0 6= φ and TF0 = jF∗ 0 T : ΩF0 → 2F0 be the Galerkin approximation of
T , where jF∗ 0 is the adjoint mapping of natural inclusion jF0 : F0 → F . If
deg(T, Ω, 0) 6= deg(TF0 , ΩF0 , 0). Then there exist x ∈ ∂Ω and f ∈ T x such
that
(f, x) ≤ 0, (f, v) = 0 for all v ∈ F0 ,
where deg(·, ·, ·) is the topological degree for upper semicontinuous mappings
with compact convex values in finite dimensional spaces.
Proof. Since F is finite dimensional, we may assume that F is a Hilbert
space and hence F ∗ = F and jF∗ 0 is the projection P : F → F0 . Let F1 =
F F0 and then F = F0 ⊕ F1 . Let B1 be the open unit ball of F1 and then
Ω1 = ΩF0 ⊕ B1 is an open subset of F .
Now, we define a mapping T1 : Ω1 → 2F by
T1 (u + v) = TF0 u + v for all u ∈ ΩF0 , v ∈ B1 .
Then we have
deg(T1 , Ω1 , 0) = deg(TF0 , ΩF0 , 0) 6= deg(T, Ω1 , 0).
Again, we define a mapping T ∗ : Ω → 2F by
T ∗ u = P T u + (I − P )u for all u ∈ Ω.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 143

One can easily see that

deg(T1 , Ω1 , 0) = deg(T1 , Ω ∩ Ω1 , 0),

deg(T ∗ , Ω, 0) = deg(T ∗ , Ω ∩ Ω1 , 0).

Consider the homotopy class {Tt : 0 ≤ t ≤ 1}, where Tt : Ω ∩ Ω1 → 2F is
defined by Tt u = tT1 u + (1 − t)T ∗ for all u ∈ Ω ∩ Ω1 and 0 ≤ t ≤ 1. It is easy
to see that, if 0 ∈ Tt0 u0 , then we have

u 0 ∈ F0 ∩ Ω ∩ Ω 1

and so

deg(T ∗ , Ω, 0) = deg(T ∗ , Ω ∩ Ω1 , 0)
= deg(T1 , Ω ∩ Ω1 , 0)
= deg(T, Ω1 , 0)
= deg(T0 , Ω0 , 0).

Therefore, we get
deg(T ∗ , Ω, 0) 6= deg(T, Ω, 0). (6.2.1)
Let

Ht u = (tI + (1 − t)P )T u + (1 − t)(I − P )u for all (t, u) ∈ [0, 1] × Ω.

It follows from (6.2.1) that there exist t1 ∈ [0, 1] and u1 ∈ ∂Ω such that
0 ∈ Ht1 u1 . Hence there exists f1 ∈ T u1 such that

tf1 + (1 − t1 )P f1 + (1 − t)(I − P )u1 = 0. (6.2.2)

Multiplying (6.2.2) by v ∈ F0 and u1 , respectively, we obtain

(f1 , v) = 0, (f1 , u1 ) = 0.

This completes the proof.

Lemma 6.2.2. Let {Tt }t∈[0,1] be a homotopy of mappings of class (S+ ).
If 0 ∈
/ Tt (∂Ω) for all t ∈ [0, 1], then there exists a finite dimensional subspace
F0 such that 0 ∈ / Tt,F (∂Ω) for all F with F0 ⊂ F and dimF < ∞, where
Tt,F = jF∗ Tt for all t ∈ [0, 1].
Proof. Suppose that the conclusion is not true. For finite dimensional
subspaces F0 and F with F0 ⊂ F , we define a set WF as follows:

WF = {(t, x) ∈ [0, 1] × ∂Ω : there exists f ∈ Tt x

such that (f, x) ≤ 0 and (f, v) = 0 for all v ∈ F }.

Copyright 2006 by Taylor & Francis Group, LLC

 144 Topological Degree Theory and Applications

Then WF is nonempty. Let WF be the closure of WF in [0, 1] × E with E

endowed with weak topology.
Consider the following family of sets:

F = {WF : F0 ⊂ F, dim(F ) ≤ ∞}.

It is easy to show that ∩F ∈F WF 6= φ. Let (t0 , x0 ) ∈ ∩F ∈F WF . For all v ∈ E,

if we take a finite dimensional subspace F such that F0 ⊂ F , v ∈ F and
x0 ∈ F , then there exist (tvj , xvj ) ∈ WF and fjv ∈ Ttvj xvj such that

tvj → t0 , xvj * x0 , (fjv , xvj ) ≤ 0, (fjv , v) = 0

for j = 0, 1, 2, · · · . Hence we have lim supj→∞ (fjv , xvj − x0 ) ≤ 0. But since

{Tt : t ∈ [0, 1]} is a homotopy of mappings of class (S+ ), it follows that
xvj → x0 ∈ ∂Ω and (fjv ) has a subsequence (fjvk ) that converges weakly to
f0v ∈ Tt0 x0 . Therefore, we have (f0v , v) = 0 for all v ∈ E. By Mazur’s
separation theorem, we get 0 ∈ Tt0 x0 , which is a contradiction. This completes
the proof.
Lemma 6.2.3. Let {Tt }t∈[0,1] be a homotopy of mappings of class (S+ ).
If 0 ∈
/ Tt (∂Ω) for all t ∈ [0, 1], then there exists a finite dimensional subspace
F0 such that the topological degree deg(Tt,F , ΩF , 0) is well defined and does
not depend on t ∈ [0, 1] and each finite dimensional subspace F with F0 ⊂ F .
Proof. By Lemma 6.2.2, there exists a finite dimensional subspace F 0 such
that 0 ∈/ Tt,F (∂Ω) for all t ∈ [0, 1] and F with F 0 ⊂ F and dimF < ∞. It
∗
is easy to see Tt,F : ΩF → 2F is upper semicontinuous with compact convex
values. Thus, by Definition 6.3.12, the topological degree deg(Tt,F , ΩF , 0) is
well defined.
Now, we show that there exists a finite dimensional subspace F0 such that
F 0 ⊂ F0 , deg(Tt,F , Ω, 0) does not depend on t ∈ [0, 1] and F with F0 ⊂ F and
dimF < ∞. In fact, if this is not true, then, as in the proof of Lemma 6.2.2,
we define

WF = {(t, x) ∈ [0, 1] × ∂Ω : there exists f ∈ Tt x

such that (f, x) ≤ 0 and (f, v) = 0 for all v ∈ F }.

Then WF is nonempty by Lemma 6.2.1. Let WF be the closure of WF in

[0, 1] × E with E endowed with the weak topology.
Consider again the following family of sets:

F = {WF : F0 ⊂ F, dim(F ) ≤ ∞}.

It is easy to show that ∩F ∈F WF 6= φ. Let (t0 , x0 ) ∈ ∩F ∈F WF . Then, for all

v ∈ E, if we take a finite dimensional subspace F such that F0 ⊂ F , v ∈ F
and x0 ∈ F , then there exist (tvj , xvj ) ∈ WF and fjv ∈ Ttvj xvj such that

tvj → t0 , xvj * x0 , (fjv , xvj ) ≤ 0, (fjv , v) = 0

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 145

for j = 0, 1, 2, · · · . Hence we have limj→∞ (fjv , xvj − x0 ) ≤ 0. But since {Tt :

t ∈ [0, 1]} is a homotopy of mappings of class (S+ ), we have xvj → x0 ∈ ∂Ω
and (fjv ) has a subsequence (fjvk ) which converges weakly to f0v ∈ Tt0 x0 .
Therefore, we have (f0v , v) = 0 for all v ∈ E and so, by Mazur’s separation
theorem, 0 ∈ Tt0 x0 , which is a contradiction. This completes the proof.
∗
Now, let T : Ω → 2E be a mapping of class (S+ ) with 0 6∈ ∂Ω. By Lemma
6.2.3, there exists a finite dimensional subspace F0 such that 0 ∈
/ TF (∂Ω) for
all F with F0 ⊂ F and dimF < ∞, and the topological degree deg(TF , ΩF , 0)
is well defined and does not depend on F .
Now, we define the topological degree deg(T, Ω, 0) as the common value of
deg(TF , ΩF , 0).
Theorem 6.2.4. Let E be a reflexive Banach space, Ω ⊂ E be an open
∗
bounded subset and T : Ω → 2E be a mapping of class (S+ ). If 0 6∈ T (∂Ω),
then the topological degree deg(T, Ω, 0) defined as above has the following
properties:

(1) deg(J, Ω, 0) = 1 if 0 ∈ J(Ω);

(2) If deg(T, Ω, 0) 6= 0, then 0 ∈ T x has a solution in Ω;

(3) If Ω1 and Ω2 ⊂ Ω are two open subsets with Ω = Ω1 ∪Ω2 and Ω1 ∩Ω2 = φ,
then
deg(T, Ω, 0) = deg(T, Ω1 , 0) + deg(T, Ω2 , 0);

(4) If {Tt }t∈[0,1] is a homotopy of mappings of class (S+ ), Ω ⊂ D(Tt ) and

0 ∈/ Tt (∂Ω) for all t ∈ [0, 1], then deg(Tt , Ω, 0) does not depend on
t ∈ [0, 1].

6.3 Degree for Perturbations of Monotone-Type Map-

pings
In this section, based on the results in Sections 6.1 and 6.2, we establish
various topological degree theories for monotone type mappings. In the sequel,
∗
T : Ω → 2E is a bounded mapping of class (S+ ), where Ω is an open bounded
subset in E.
∗
Lemma 6.3.1. Let M : D(M ) ⊆ E → 2E be a maximal monotone
mapping with D(M ) ∩ Ω 6= ∅. If 0 ∈
/ (M + T )(∂Ω ∩ D(M )), then there exists
λ0 > 0 such that

0∈
/ (Mλ + T )(∂Ω) for all λ < λ0 ,

Copyright 2006 by Taylor & Francis Group, LLC

 146 Topological Degree Theory and Applications

where Mλ is the Yosida approximation of M .

Proof. Suppose this is not true. Then there exist xj ∈ ∂Ω with xj * x0 ,
λj → 0+ , fj ∈ T xj with fj * f0 such that

Mλj xj + fj = 0.

We have (Mλj xj + fj , xj − x0 ) = 0. Since T is a mapping of class (S+ ), we

have lim inf j→∞ (fj , xj − x0 ) ≥ 0. Thus we get

lim sup(Mλj xj , xj − x0 ) ≤ 0.
j→∞

Hence
lim sup(Mλj xj , xj ) ≤ (−f0 , x0 ). (6.3.1)
j→∞

On the other hand, it follows that (m − Mλj xj , x − Rλj xj ) ≥ 0 for all

x ∈ D(M ) and m ∈ M x, i.e.,

(m − Mλj xj , x − xj + λj J −1 Mλj xj ≥ 0. (6.3.2)

By (6.3.1) and (6.3.2), we get

(m, x − x0 ) − (−f0 , x) ≥ −(−f0 , x0 ), x ∈ D(M ), m ∈ T x.

Thus x0 ∈ D(M ), and −f0 ∈ M x0 . Notice that (Mλj xj −Mλj x0 , xj −x0 ) ≥ 0.

By Lemma 6.1.29, we get

lim inf (Mλj xj , xj − x0 ) ≥ 0

j→∞

and so we have
lim sup(fj , xj − x0 ) ≤ 0.
j→∞

Therefore, it follows that xj → x0 ∈ ∂Ω∩D(M ), f0 ∈ T x0 and 0 ∈ M x0 +T x0 ,

which is a contradiction.
Now, suppose that T and M satisfy the conditions of Lemma 6.3.1. By
Lemma 6.3.1, there exists λ0 > 0 such that

0 6∈ (Mλ + T )(∂Ω ∩ D(M )).

Since Mλ + T is a mapping of class of (S+ ), deg(Mλ + T, Ω, 0) is well defined

for any λ < λ0 . By Proposition 6.1.30, it is easy to check that

{Mtλ1 +(1−t)λ2 + T : t ∈ [0, 1]}

is a homotopy of mappings of class (S+ ). Therefore, deg(Mλ + T, Ω, 0) does

not depend on λ ∈ (0, λ0 ).

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 147

Now, we define

deg(M + T, Ω ∩ D(T ), 0) = lim+ deg(Mλ + T, Ω, 0). (6.3.3)

λ→0

Theorem 6.3.2. The topological degree defined by (6.3.3) has the follow-
ing properties:

(1) deg(J, Ω, 0) = 1 if and only if 0 ∈ J(Ω);

(2) If deg(M + T, Ω, 0) 6= 0, then 0 ∈ (M + T )x has a solution in Ω;

(3) If Ω1 and Ω2 ⊂ Ω are two open subsets with Ω = Ω1 ∪Ω2 and Ω1 ∩Ω2 = φ,
then

deg(M + T, Ω, 0) = deg(M + T, Ω1 , 0) + deg(M + T, Ω2 , 0);

(4) If {Tt }t∈[0,1] is a homotopy of mappings of class (S+ ), M is a maximal

monotone mapping, Ω ∩ D(M ) 6= ∅, Ω ⊂ D(Tt ) and 0 ∈ / (M + Tt )(∂Ω ∩
D(M )) for any t ∈ [0, 1], then

deg(M + T0 , Ω, 0) = deg(M + T1 , Ω, 0).

Proof. The proof is straightforward.

Remark. The condition (4) of Theorem 6.3.2 is not in its most general
form. In applications one may find different homotopy properties in which M
may depend on t.
∗
Lemma 6.3.3. Let M : D(M ) ⊆ E → 2E be a maximal monotone
mapping. If xj ∈ D(T ) ∩ Ω, xj * x0 , (j ) is a positive sequence with j → 0+
and 0 ∈ (M + j T )xj , then xj → x0 ∈ D(M ) and 0 ∈ M x0 .
Proof. By our assumptions, there exists fj ∈ T xj such that

(j fj − i fi , xj − xi ) ≤ 0 for i, j = 1, 2, · · · .

By letting i → ∞, we get

j (fj , xj − x0 ) ≤ 0.

But since T is a mapping of a class (S+ ), and so xj → x0 , we have x0 ∈ D(M )

and 0 ∈ M x0 . This completes the proof.
Lemma 6.3.4. Let T1 , T2 be two mappings of class (S+ ), i > 0 and
λi > 0 for i = 1, 2. Then {Mtλ1 +(1−t)λ2 + t1 T1 + (1 − t)2 T2 : t ∈ [0, 1]} is a
homotopy of mappings of class (S+ ).

Copyright 2006 by Taylor & Francis Group, LLC

 148 Topological Degree Theory and Applications

Proof. Let tj → t0 and xj * x0 be such that

lim (Mtj λ1 +(1−tj )λ2 xj + tj j fj1 + (1 − tj )fj2 , xj − x0 ) ≤ 0.
j→∞

By the monotonicity of M and Proposition 6.1.30, we have

limj→∞ (Mtj λ1 +(1−tj )λ2 xj , xj − x0 ) ≥ 0.
Consequently, we have
lim (tj j fj1 + (1 − tj )fj2 , xj − x0 ) ≤ 0.
j→∞

But, since {t1 T1 + (1 − t)2 T2 : t ∈ [0, 1]} is a homotopy of mappings of class

(S+ ), it follows that xj → x0 , (fj1 ) and (fj2 ) have subsequences (fj1k ) and (fj2k )
such that fj1k * f 1 ∈ T1 x0 and fj2k * f 2 ∈ T 2 x0 , respectively. Moreover, we
have
lim (Mtj λ1 +(1−tj )λ2 xj = Mt0 λ1 +(1−t0 )λ2 x0 .
j→∞

This completes the proof.

Now, suppose that 0 6∈ M (∂Ω ∩ D(M )). By Lemma 6.3.3, there exists
0 > 0 such that
0∈
/ (M + T )(∂Ω ∩ D(M )) for all  ∈ (0, 0 )
and so deg(M +T, Ω∩D(M ), 0) is well defined for any  ∈ (0, 0 ). By Lemma
6.3.4, deg(M + T, Ω ∩ D(M ), 0) does not depend on  ∈ (0, 0 ).
Now, we define
deg(M, Ω ∩ D(M ), 0) = lim deg(M + T, Ω ∩ D(M ), 0). (6.3.4)
→0

Theorem 6.3.5. The topological degree defined by (6.3.4) has the follow-
ing properties:
(1) deg(J, Ω, 0) = 1 if and only if 0 ∈ J(Ω);
(2) If deg(M, Ω, 0) 6= 0, then 0 ∈ M x has a solution in Ω;
(3) If Ω1 and Ω2 ⊂ Ω are two open subsets with Ω = Ω1 ∪Ω2 and Ω1 ∩Ω2 = ∅,
then
deg(M, Ω, 0) = deg(M, Ω1 , 0) + deg(M, Ω2 , 0);
(4) If M1 , M2 are two maximal monotone mappings, Ω∩D(M1 )∩D(M2 ) 6= ∅
and
0∈
/ (tM1 +(1−t)M2 )(∂Ω∩D(M1 )∩D(M2 ))∪(tM1,λ +(1−t)M2,λ )(∂Ω)
for all t ∈ [0, 1] and λ ∈ (0, λ0 ), then
deg(M1 , Ω, 0) = deg(M2 , Ω, 0).

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 149

Proof. The proof follows from (6.3.4), (6.3.3), and Theorem 6.2.4.
Remark. A degree theory can also be developed for pseudomonotone map-
pings and generalized pseudomonotone mappings and their perturbations with
a maximal monotone mappings by using Proposition 6.1.31, which is a method
similar to the one employed above, and so we leave it to the reader as an ex-
ercise.

6.4 Degree Theory for Mappings of Class (S+ )L

In this section, we construct a topological degree theory for multi-valued
mappings of class (S+ )0,L , and the topological degree for multi-valued map-
pings of class (S+ )L follows from definition 6.1.16.
In the following, we assume that E is also separable, L ⊂ E is a dense
subspace of E and Ω ⊂ E is a nonempty open bounded subset. Let {Tt }t∈[0,1]
be a homotopy of mappings of class (S+ )0,L which has a common domain D
and ΩF = Ω ∩ D ∩ F for each finite dimensional subspace F of L is open in
F . We may choose a sequence of finite dimensional subspaces Fj of L such
that Fj ⊆ Fj+1 and ∪∞ j=1 Fj = E. This is possible because E is separable and
L is dense in E.
∗
Lemma 6.4.1. Let T : Ω ∩ D(T ) → 2E be a mapping of (S+ )0,L . Suppose
that 0 ∈
/ T (∂Ω ∩ D(T )). Then there exists an integer N > 0 such that

0∈
/ Tn (∂Ω ∩ D(T ) ∩ Fn ) for all n > N,

where Tn = jF∗ n T .
Proof. Suppose that the conclusion is not true. Then there exists xnk ∈
∂Ω ∩ D(T ) ∩ Fnk such that 0 ∈ Tnk xnk , i.e., there exists fnk ∈ T xnk such
that 0 = jF∗ n fnk for k = 1, 2, · · · . Without loss of generality, we may assume
k
that xnk * x0 . Now we have (fnk , x) = 0 for all x ∈ Fnk , k = 1, 2, · · · .
Thus (fnk , xnk ) = 0 and limk→∞ (fnk , v) = 0 for all v ∈ ∪∞ j=1 Fj . Since T is
a mapping of class (S+ )0,L , it follows that xnk → x0 ∈ ∂Ω ∩ D and 0 ∈ T x0 ,
which is a contradiction. This completes the proof.
Under the conditions of Lemma 6.4.1, we know from Section 2.3 that the
topological degree deg(Tn , Ω ∩ D(T ) ∩ Fn , 0) is well defined for sufficiently
large n, and we have the following:
Lemma 6.4.2. Let T be the same as in Lemma 6.4.1. Then there exists
an integer N > 0 such that the topological degree deg(Tn , Ω ∩ D(T ) ∩ Fn , 0)
does not depend on n > N , where Tn = jF∗ n T .

Copyright 2006 by Taylor & Francis Group, LLC

 150 Topological Degree Theory and Applications

Proof. Suppose that the conclusion is not true. By Lemma 6.2.1, there
exist xnk ∈ ∂Ω∩D∩Fnk , fnk ∈ T xnk such that (fnk , xnk ) ≤ 0 and (fnk , x) = 0
for all x ∈ Fnk , k = 1, 2, · · · . We may assume that xnk * x0 . By the same
proof as in Lemma 6.4.2, we get xnk → x0 ∈ ∂Ω ∩ D and 0 ∈ T x0 , which is a
contradiction. This completes the proof.
Suppose that {Ej }∞
j=1 is another sequence of finite dimensional subspaces
∞
of L and ∪j=1 Ej = E. Then we have the following:
Lemma 6.4.3. Let T be the same as in Lemma 6.4.1. Then there exists
an integer N > 0 such that

deg(Tn , Ω ∩ D(T ) ∩ Fn , 0) = deg(Tn , Ω ∩ D(T ) ∩ En , 0) for all n > N.

Proof. Put Kn = En ∪ Fn . By using the same proof as in Lemma 6.4.2,

there exist two integers N1 > 0 and N2 > 0 such that

deg(Tn , Ω ∩ D(T ) ∩ Fn , 0) = deg(Tn , Ω ∩ D(T ) ∩ Kn , 0) for all n > N1

and

deg(Tn , Ω ∩ D(T ) ∩ En , 0) = deg(Tn , Ω ∩ D(T ) ∩ Kn , 0) for all n > N2 .

Therefore, the conclusion of Lemma 6.4.3 is true. This completes the proof.
Now, let L be a dense subspace of E, Ω ⊂ E be a nonempty open bounded
∗
subset and T : D(T ) ⊂ E → 2E be a mapping of class (S+ )0,L . Assume
that Ω ∩ D(T ) ∩ F is open in F for each finite dimensional subspace F of L.
Suppose that 0 6∈ T (∂Ω ∩ D(T )). In view of lemmas 6.4.1 to 6.4.3, we may
define the topological degree by

Deg(T, Ω ∩ D(T ), 0) = lim deg(Tn , Ω ∩ D(T ) ∩ Fn , 0).

n→∞

In general, if T is a mapping of class (S+ )L and p 6∈ T (∂Ω ∩ D(T )), then

we can define the topological degree by

Deg(T, Ω ∩ D(T ), p) = Deg(T − p, Ω ∩ D(T ), 0).

The topological degree defined above has the following properties:

Theorem 6.4.4. Let E be a reflexive Banach space, L be a dense subspace
∗
of E, Ω ⊂ E be an open bounded subset and T : D(T ) ∩ Ω → 2E be a
mapping of class (S+ )0,L . If 0 6∈ T (∂Ω ∩ D(T )), then the topological degree
Deg(T, Ω ∩ D(T ), 0) defined above has the following properties:
(1) Deg(J, Ω, 0) = 1 if 0 ∈ J(Ω);
(2) If Deg(T, Ω ∩ D(T ), 0) 6= 0, then 0 ∈ T x has a solution in Ω ∩ D(T );

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 151

(3) If Ω1 and Ω2 ⊂ Ω are two open subsets with Ω = Ω1 ∪Ω2 and Ω1 ∩Ω2 = ∅,
then

Deg(T, Ω ∩ D(T ), 0) = Deg(T, Ω1 ∩ D(T ), 0) + Deg(T, Ω2 ∩ D(T ), 0);

(4) If {Tt }t∈[0,1] is a homotopy of mappings of class (S+ )0,L with D(Tt ) = D
and 0 ∈ / Tt (∂Ω ∩ D) for all t ∈ [0, 1], then Deg(Tt , Ω ∩ D, 0) does not
depend on t ∈ [0, 1].

Proof. (1) to (3) follow easily from the definition and the properties of
degree theory in finite dimensional spaces.
We only need to prove (4). Assume that (Fj )∞ j=1 is a sequence of finite
dimensional subspaces of L with ∪∞ F
j=1 j = E. We claim that there exists an
integer N > 0 such that

0∈
/ Tt,n (∂Ω ∩ D ∩ Fn ) for all n > N, t ∈ [0, 1],

where Tt,n = jF∗ n Tt . If not, then there exist tnk → t0 , xnk ∈ ∂Ω ∩ D ∩ Fnk
with xnk * x0 , fnk ∈ Ttnk xnk such that 0 = jF∗ n fnk , which implies that
k

(fnk , xnk ) = 0, (fnk , v) = 0 for all v ∈ Fnk .

Since {Tt }t∈[0,1] is a homotopy of mappings of class (S+ )0,L , we get xnk →
x0 ∈ ∂Ω ∩ D and 0 ∈ Tt0 x0 , which is a contradiction. Thus the claim is true.
Now, for all n > N , we know from Section 2.3 that deg(Tt,n , Ω∩D∩L∩Fn , 0)
is a constant for t ∈ [0, 1], where Tt,n = jF∗ n Tt . In view of Lemma 6.4.3, we
see that the conclusion of (4) is true. This completes the proof.
∗
Theorem 6.4.5. Let T : D(T ) ⊂ E → 2E be a mapping of class (S+ )0,L
and Ω ⊂ E be an open bounded subset such that Ω ∩ D(T ) ∩ F is open in F
for each finite dimensional subspace F of L. If 0 ∈ Ω ∩ D(T ) and (f, x) > 0
for all x ∈ ∂Ω ∩ D(T ) and f ∈ T x, then

Deg(T, Ω ∩ D(T ), 0) = 1.

Proof. Assume that (Fj )∞ j=1 is a sequence of finite dimensional subspaces

of L with ∪∞
j=1 Fj = E. It is easy to check that

0 6∈ (tjF∗ n T + (1 − t)jF∗ n J)(∂Ω ∩ D(T ) ∩ Fn )

for all t ∈ [0, 1] and n = 1, 2, · · · . Therefore, we have

deg(jF∗ n T, Ω ∩ D(T ) ∩ Fn , 0) = deg(jF∗ n J, Ω ∩ D(T ) ∩ Fn , 0) = 1

for all n = 1, 2, · · · such that Fn ∩ Ω ∩ D(T ) 6= ∅. By the definition of

topological degree Deg(T, Ω ∩ D(T ), 0), we get

Deg(T, Ω ∩ D(T ), 0) = 1.

Copyright 2006 by Taylor & Francis Group, LLC

 152 Topological Degree Theory and Applications

This completes the proof.

Theorem 6.4.6. Let H be a real separable Hilbert space, L be a dense
subspace of H, Ω ⊂ H be an open bounded subset such that Ω ∩ D(T ) ∩ F is
open in F for each finite dimensional subpace F of L and T : D(T ) → 2H be
a L0 -pseudocompact mapping. Suppose that 0 ∈ D(T ) ∩ Ω and

(T x, x) ≤ kxk2 for all x ∈ ∂Ω ∩ D(T ).

Then T has a fixed point in Ω ∩ D(T ).

Proof. We may assume that x 6∈ T x for all x ∈ ∂Ω ∩ D(T ). Assume that
(Fj )∞ ∞
j=1 is a sequence of finite dimensional subspaces of L with ∪j=1 Fj = H.
We claim that there exists an integer N > 0 such that

0∈
/ [(tPn (I − T ) + (1 − t)Pn ](∂Ω ∩ D(T ) ∩ Fn

for all n > N and t ∈ [0, 1], where Pn : H → Fn is the projection. Assume
this is not true. Then there exist tj → t0 , xj ∈ ∂Ω∩D ∩Fnj with xj * x0 and
fj ∈ T xj such that PFnj (xj −tj fj ) = 0 for j = 1, 2, · · · , where PFnj : H → Fnj
is the projection. Thus it follows that (xj − tj fj , xj ) = 0. This and our
assumption imply that tj = 1, so we have

(xj − fj , xj ) = 0, (xj − fj , v) = 0 for all v ∈ Fnj , j = 1, 2, · · · .

From which we get xj → x0 ∈ ∂Ω ∩ D(T ) and x0 ∈ T x0 , which is a contra-

diction. Thus we get

deg(Pn (I − T ), Ω ∩ D(T ) ∩ Fn , 0) = deg(Pn , Ω ∩ Fn , 0) = 1.

Thus it follows that Deg(I − T, Ω ∩ D(T ), 0) = 1. By Theorem 6.4.4, T has a

fixed point in Ω ∩ D(T ). This completes the proof.

6.5 Coincidence Degree for Mappings of Class L-(S+ )

In this section, we construct a topological degree theory for multi-valued
mappings of class L-(S+ ).
Lemma 6.5.1. Let L be a Fredholm mapping of index zero type, Ω ⊂ H
be an open bounded subset and T : Ω → 2H be a mapping. If 0 ∈ / (L −
T )(∂Ω ∩ D(L)), then 0 ∈
/ [I − P − (JQ + KP Q )T ](∂Ω).
Proof. Suppose the contrary. Then there exists x0 ∈ ∂Ω such that 0 ∈
x0 − P x0 − (JQ + KP Q )T x0 , and so there exist f0 ∈ T x0 such that

x0 − P x0 − JQf0 − KP Q f0 = 0.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 153

Since

JQf0 ∈ Ker(L) = Im(P ), x0 − P x0 ∈ Ker(P ), KpQ f0 ∈ D(L) ∩ Ker(P ),

we must have
JQf0 = 0, x0 − P x0 − KP Q f0 = 0.
Therefore, we have

Qf0 = 0, x0 − P x0 − KP f0 = 0, i.e., Lx0 − f0 = 0,

which is a contradiction to 0 ∈
/ (L − T )(∂Ω ∩ D(L)). This completes the proof.
Now, let L be a Fredholm mapping of index zero type, Ω ⊂ H be an open
bounded subset and T : Ω → 2H be a mapping of class L-(S+ ). Suppose that
0 ∈
/ (L − T )(∂Ω ∩ D(L)). By Lemma 6.5.1, we have 0 ∈ / [I − P − (JQ +
KP Q )T ](∂Ω). As a result, deg(I − P − (JQ + KP Q )T, Ω, 0) is well defined.
We define

degJ (L − T, Ω, 0) = deg(I − P − (JQ + KP Q )T, Ω, 0),

which is called the coincidence degree of L and T on Ω.

Notice that the projections P, Q in Hilbert spaces are uniquely determined,
so degJ (L − T, Ω, 0) is well defined.
The following result follows directly from Theorem 6.2.4:
Theorem 6.5.2. The coincidence degree of L and T on Ω has the following
properties:

(1) If Ω1 and Ω2 are disjoint open subsets of Ω such that 0 ∈

/ (L−T )(D(L)∩
Ω \ (Ω1 ∪ Ω2 ), then

degJ (L − T, Ω, 0) = degJ (L − T, Ω1 ) + degJ (L − T, Ω2 , 0);

(2) If H(t, x) : [0, 1] × Ω → Y is a mapping of class L-(S+ ) on [0, 1] × Ω and,

if 0 6= Lx − H(t, x) for all (t, x) ∈ [0, 1] × ∂Ω, then degJ (L − H(t, ·), Ω, 0)
does not depend on t ∈ [0, 1];

(3) If degJ (L − T, Ω, 0) 6= 0, then 0 ∈ (L − T )(D(L) ∩ Ω);

(4) If L : D(L) ⊆ H → H is a linear mapping such that L−1 : H → D(L)

is continuous, then degJ (L, Ω, 0) = 1 if 0 ∈ Ω.

Theorem 6.5.3. Let L be a Fredholm mapping of index zero type with

Ker(L) = {0}, 0 ∈ Ω ⊂ H be an open bounded subset and T : Ω → 2H be a
mapping of class L-(S+ ). Suppose that Lx ∈ / tT x for all x ∈ ∂Ω ∩ D(L) and
t ∈ [0, 1]. Then degJ (L − T, Ω ∩ D(T ), 0) = 1.

Copyright 2006 by Taylor & Francis Group, LLC

 154 Topological Degree Theory and Applications

Proof. Since Ker(L) = {0}, we take P = Q = 0 and J = I. Now, I −L−1 T

is mapping of class (S+ ) and {I − tL−1 T ]}t∈[0,1] is a homotopy of mappings
/ (x − tL−1 T x)
of class (S+ ). From our assumption, one can easily see that 0 ∈
for all x ∈ ∂Ω, t ∈ [0, 1]. From Theorem 6.2.4, we have

deg(I − L−1 T, Ω, 0) = deg(I, Ω, 0) = 1,

i.e., degJ (L − T, Ω ∩ D(T ), 0) = 1. This completes the proof.

Corollary 6.5.4. Let L be a Fredholm mapping of index zero type with
Ker(L) = {0} and T : H → 2H be a mapping of class L-(S+ ). Then one of
the following conditions hold:
(1) Lx ∈ T x has a solution in D(T );
(2) {x : Lx ∈ λT x for some λ ∈ (0, 1)} is unbounded.

Theorem 6.5.5. Let L be a Fredholm mapping of index zero type, Ω ⊂ H

an open bounded subset and Ti : Ω → 2H be a mapping of class L-(S+ ) for
i = 1, 2. Suppose that T1 x = T2 x for all x ∈ ∂Ω ∩ D(L). Then we have the
following:
degJ (L − T1 , Ω, 0) = degJ (L − T2 , Ω, 0).

Proof. Since T1 x = T2 x for all x ∈ ∂Ω, we have

0∈
/ (I − P − (JQ + KP Q (tT1 + (1 − t)T2 )))x for all x ∈ ∂Ω.

Moreover, J is linear, so

I − P − (JQ + KP Q )(tT1 + (1 − t)T2 )

= I − P − t(JQ + KP Q )T1 − (1 − t)(JQ + KP Q )T2

is a homotopy of mappings of class (S+ ). Therefore, it follows from Theorem

6.2.4 that

deg(I − P − (JQ + KP Q )T1 , Ω, 0) = deg(I − P − (JQ + KP Q )T2 , Ω, 0)

and so
degJ (L − T1 , Ω, 0) = degJ (L − T2 , Ω, 0).
This completes the proof.
Suppose that L is a Fredholm mapping of index zero type, Ω ⊂ H be an
open bounded subset, Ω∩D(L)∩D(T ) 6= ∅ and T : D(T ) → 2H is a L-maximal
monotone mapping. Also, suppose that 0 ∈
/ (L−T )(∂Ω∩D(L)∩D(T )). Then,
by Lemma 6.5.1, we have

0∈
/ [I − P − (JQ + KP Q )T ](∂Ω ∩ D(T )).

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 155

Since I − P − (JQ + KP Q )T is maximal monotone, the degree deg(I − P −

(JQ + KP Q )T, Ω, 0) is well defined, and we may define a degree

degJ (L − T, Ω, 0) = deg(I − P − (JQ + KP Q )T, Ω, 0),

and one may deduce results similar to Theorem 6.5.3.

Lemma 6.5.6. Let T : D(T ) ⊆ H → 2H be a bounded multi-valued
mapping and 0 ∈
/ (L − T )(∂Ω ∩ D(L) ∩ D(T )). Then there exists 0 > 0 such
that

0∈
/ [I − P − (JQ + KP Q )T + I](∂Ω ∩ D(T )) for all  ∈ (0, 0 ).

Proof. Assume that the conclusion is false. Then there exist n → 0 and
xn ∈ ∂Ω ∩ D(T ) such that

0 ∈ (I − P − (JQ + KP Q )T + n )xn .

Thus there exist fn ∈ T xn such that 0 = xn − P xn − (JQ + KP Q )fn + n xn .

Therefore, we have

(1 + n )(I − P )xn + (n P xn + JQfn ) + KP Q fn = 0.

Consequently, it follows that

JQfn = −n P xn , (1 + n )(I − P )xn + KP Q fn = 0

and so we have (1 + n )Lxn − fn − n J −1 P xn = 0. By the boundedness of T ,

we get Lxn − fn → 0, which is a contradiction. This completes the proof.
Now, we assume that T : Ω → H is a bounded mapping and T is L-
pseudomonotone. Suppose that 0 ∈ / (L − T )(∂Ω ∩ D(L)). Then by Lemma
6.5.6, there exists 0 > 0 such that

0∈
/ [I − P − (JQ + KP Q )T + I](∂Ω) for all  ∈ [0, 0 ).

Since I − P − (JQ + KP Q )T is pseudomonotone, I − P − (JQ + KP Q )T + I is

a mapping of class (S+ ) and the degree deg(I − P − (JQ + KP Q )T + I, Ω, 0)
is well defined for all  ∈ (0, 0 ). We define

degJ (L − T, Ω, 0) = lim deg(I − P − (JQ + KP Q )T + I, Ω, 0),

→0+

which is easily seen to be well defined if we note that {I − P − (JQ+ KP Q )T +

[t1 + (1 − t)2 ]I}t∈[0,1] is a homotopy of mappings of class (S+ ).
Theorem 6.5.7. The coincidence degree of L and the pseudomonotone
mapping T on Ω has the following properties:

Copyright 2006 by Taylor & Francis Group, LLC

 156 Topological Degree Theory and Applications

(1) If Ω1 and Ω2 are open subsets of Ω such that

Ω1 ∩ Ω2 = ∅, / (L − T )(D(L) ∩ Ω \ (Ω1 ∪ Ω2 ),
0∈

then we have

degJ (L − T, Ω, 0) = degJ (L − T, Ω1 ) + degJ (L − T, Ω2 , 0);

(2) If Ti : Ω → H, i = 1, 2, are bounded mapping such that tT1 + (1 − t)T2

is L-pseudomonotone on [0, 1] × Ω and

0∈
/ ∪t∈[0,1] [L − tT1 + (1 − t)T2 ](∂Ω ∩ D(L)),

then we have

degJ (L − T1 , Ω, 0) = degJ (L − T2 , Ω, 0);

(3) If degJ (L − T, Ω, 0) 6= 0, then 0 ∈ (L − T )(D(L) ∩ Ω).

6.6 Computation of Topological Degree

In this section, we compute the topological degree of some monotone-type
mappings under certain boundary conditions.
∗
Theorem 6.6.1. Let T : Ω ⊂ E → 2E be a mapping of class (S+ ). If
0 ∈ Ω and (f, x) > 0 for all x ∈ ∂Ω and f ∈ T x, then deg(T, Ω, 0) = 1.
Proof. Consider the family {tT + (1 − t)J : t ∈ [0, 1]} of mappings. By our
assumptions, we know that 0 6∈ (tT + (1 − t)J)(∂Ω) for all t ∈ [0, 1], and so a
direct proof shows that {tT + (1 − t)J : t ∈ [0, 1]} is a homotopy of mappings
of class (S+ ). By Theorem 6.2.4, we have

deg(T, Ω, 0) = d(J, Ω, 0) = 1.

This completes the proof.

∗
Theorem 6.6.2. Let T : Ω ⊂ E → 2E be a mapping of class (S+ ) and
∗
M : D(M ) ⊆ E → 2E be a maximal monotone mapping. If x0 ∈ Ω ∩ D(M )
and
(f, x − x0 ) > −kf kkx − x0 k
for all x ∈ ∂Ω ∩ D(M ) and f ∈ (M + T )x, then deg(M + T, Ω, 0) = 1.
Proof. It is easy to see that

0 6∈ (tM + (1 − t)T + (1 − t)J(· − x0 ))(∂Ω ∩ D(M )) for all t ∈ [0, 1].

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 157

Now, we prove that there exists λ0 > 0 such that

0 6∈ (tMλ + (1 − t)T + (1 − t)J(· − x0 ))(∂Ω) for all t ∈ [0, 1].

In fact, if this is not true, then there exist tj → t0 , xj ∈ ∂Ω, xj * y0 , fj ∈ T xj

and λj → 0+ such that

0 = tj Mλj xj + (1 − tj )fj + (1 − tj )J(xj − x0 ). (6.6.1)

Multiplying (6.6.1) by xj − y0 , we obtain

(tj Mλj xj + (1 − tj )fj + (1 − tj )J(xj − x0 ), xj − y0 ) = 0. (6.6.2)

We consider the following two cases:

Case (i) t0 = 0. Multiplying (6.6.1) by xj − x0 , we obtain

(tj Mλj xj + (1 − tj )fj + (1 − tj )J(xj − x0 ), xj − x0 ) = 0.

By Lemma 6.1.29, we have tj Mλj x0 → 0 and so we obtain

lim sup((1 − tj )fj + (1 − tj )J(xj − x0 ), xj − x0 ) = 0.

j→∞

It follows that xj → x0 , which is a contradiction.

Case (ii) t0 6= 0. Since T and J(·−x0 ) are mappings of class (S+ ), it follows
that
lim inf ((1 − tj )fj + (1 − tj )J(xj − x0 ), xj − y0 ) ≥ 0.
j→∞

By (6.6.2), we have

lim sup(tj Mλj xj , xj − y0 ) ≤ 0. (6.6.3)

j→∞

Without loss of generality, we may assume that Mλj xj * m0 . By (6.6.3), we

have
lim sup(Mλj xj , xj ≤ (m0 , y0 ). (6.6.4)
j→∞

The monotonicity of M implies that

(Mλj xj − m, xj − λj J −1 Mλj xj − x) ≥ 0 (6.6.5)

for all x ∈ D(M ) and m ∈ M x. By letting j → ∞ in (6.6.5), we get

lim sup(Mλj xj , xj ) ≥ (m, y0 − x) + (m0 , x) (6.6.6)

j→∞

for all x ∈ D(M ) and m ∈ M x. Thus, from (6.6.4) and (6.6.6), it follows that

(m0 − m, y0 − x) ≥ 0 for all x ∈ D(M ), m ∈ M x,

Copyright 2006 by Taylor & Francis Group, LLC

 158 Topological Degree Theory and Applications

which implies that y0 ∈ D(M ). Therefore, (6.6.3) becomes the following:

lim (tj Mλj xj , xj − y0 ) = 0

j→∞

and hence we have

lim ((1 − tj )fj + (1 − tj )J(xj − x0 ), xj − y0 ) = 0.

j→∞

Therefore, xj → y0 , and (6.6.1) implies that

0 ∈ (t0 M + (1 − t0 )T + (1 − t0 )J(· − x0 ))(∂Ω ∩ D(M ),

which is a contradiction. By Theorem 6.2.4, we conclude that

deg(Mλ + T, Ω, 0) = deg(J(· − x0 ), Ω, 0) = 1 for all λ ∈ (0, λ0 ).

By the definition of deg(T + M, Ω, 0), we finally obtain

deg(T + M, Ω, 0) = 1.

This completes the proof.

Theorem 6.6.3. Let f : D(f ) ⊆ E → R ∪ {+∞, −∞} be a proper lower
semicontinuous convex function. If lim inf kxk→∞ f (x) = +∞, then

lim deg(∂f, B(0, R) ∩ D(∂f ), 0) = 1,

R→∞

where B(0, R) is the open ball with radius R in E.

Proof. Since f is a lower semicontinuous convex function, ∂f is maximal
monotone. By our assumptions, we know that there exists x0 ∈ D(∂f ) such
that 0 ∈ ∂f (x0 ). Hence we have

(g, x − x0 ) ≥ 0 for any g ∈ ∂f (x), x ∈ D(∂f ).

If we take R sufficiently large such that 0 6∈ (D(∂f ) ∩ B(0, R)), then, in

addition, we have

(g + J(x − x0 ), x − x0 ) > 0 for all x ∈ D(∂f ) ∩ B(0, R), g ∈ ∂f.

By Theorem 6.5.2, we conclude that

deg(∂f + J(· − x0 ), B(0, R) ∩ D(∂f ), 0) = 1.

By the definition of deg(∂f, B(0, R) ∩ D(∂f ), 0), we get

deg(∂f, B(0, R) ∩ D(∂f ), 0) = 1.

This completes the proof.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 159

Theorem 6.6.4. Let H be real Hilbert space, 0 ∈ Ω ⊂ H be an open

bounded subset and T : Ω → 2H be a pseudocompact mapping. Suppose that

λx 6∈ T x for all x ∈ ∂Ω, λ > 1.

Then T has a fixed point in Ω.

Proof. We may assume that x 6∈ T x for all x ∈ ∂Ω and λ > 1. Put
H(t, x) = t(I − T )x + (1 − t)x for (t, x) ∈ [0, 1] × Ω. Then H(t, ·) is a
homotopy mapping of a class (S+ ). Moreover, we have 0 6∈ H(t, x) for all
(t, x) ∈ [0, 1] × Ω. Therefore, we have

deg(I − T, Ω, 0) = deg(I, Ω, 0) = 1.

Hence T has a fixed point in Ω. This completes the proof.

Theorem 6.6.5. Let H be real Hilbert space, 0 ∈ Ω ⊂ H be an open
bounded subset and T : Ω → 2H be a pseudo-contractive mapping. Suppose
that
λx 6∈ T x for all x ∈ ∂Ω, λ > 1.
Then T has a fixed point in Ω.
Proof. We may assume that x 6∈ T x for all x ∈ ∂Ω and λ > 1. Put
H(t, x) = t(I − T )x + (1 − t)x for all (t, x) ∈ [0, 1] × Ω. Then we can obtain the
conclusion by using the same argument as in Theorem 6.6.4. This completes
the proof.

6.7 Applications to PDEs and Evolution Equations

In this section, we give some applications of our degree theory in the previ-
ous sections to periodic nonlinear evolution equations and existence problems
of partial differential equations.
In the following, let E be a separable reflexive Banach space which is densely
embedded in a real Hilbert space H and E ∗ be the dual space of E. Let k · k
and k · k∗ be the norms in E and E ∗ , respectively. We always assume that
both E and E ∗ are locally uniformly convex. The following function spaces
will be used in the sequel:
n Z T o
Lp (0, T ; E) = f (t) : [0, T ] → E, kf (s)kp ds < ∞
0
RT 1
with the norm kf (·)kp = ( 0 kf (s)kp ds) p and Lq (0, T ; E ∗ ) is the dual space
RT 1
of Lp (0, T ; E) with the norm kg(·)k∗,q = ( 0 kg(s)kq ds) q , where p1 + 1q = 1.

Copyright 2006 by Taylor & Francis Group, LLC

 160 Topological Degree Theory and Applications

For all u(·) ∈ Lp (0, T ; E) and v(·) ∈ Lq (0, T ; E ∗ ), let

Z T
hu, vi = (u(s), v(s))ds
0

and denote

W 1,p (0, T ; E) = {u(·) ∈ Lp (0, T ; E) : u0 (·) ∈ Lq (0, T ; E ∗ ), u(0) = u(T )}

with the norm kukw = kukp + ku0 k∗,q , where u0 is the generalized derivative
of u and k · kp , k · k∗,q are norms of the spaces Lp (0, T ; E) and Lq (0, T ; E ∗ ),
respectively. Let W ∗ be the dual space of W 1,p (0, T ; E).
∗
Theorem 6.7.1. Let A(t) : E → 2E be an operator of class (S+ ) for all
t ∈ R. Suppose the following conditions are satisfied:
(1) For all u(t) ∈ W 1,p (0, T ; E), A(t)u(t) is E ∗ -measurable on [0, T ];
(2) There exist a constant C > 0 and C1 (·) ∈ Lq (0, T ) such that

kf k∗ ≤ Ckxkp−1 + C1 (t)

for all x ∈ E and f (t) ∈ A(t)x for almost everywhere t ∈ [0, T ];

(3) There exist a constant α > 0 and a function γ(·) ∈ L(0, T ) such that

(f, x) ≥ αkxkp − γ(t)

for all x ∈ E and f (t) ∈ A(t)x for almost everywhere t ∈ [0, T ].

Then we have the following:
(
x0 (t) ∈ −A(t)x(t), a.e. t ∈ (0, T ),
(E 6.7.1)
x(0) = x(T )

has a solution x(t) ∈ W 1,p (0, T ; E).

First, we define a multi-valued operator A : W 1,p (0, T ; E) → W ∗ by

Au(·) = {f (·) ∈ W ∗ : f (t) ∈ A(t)u(t)} for a.e. t ∈ [0, T ]}.

By the condition (1) and Theorem 2.3.9, A is well defined.

Next, for each n ≥ 1, we define a multi-valued operator Sn : W 1,p (0, T ; E) →
W ∗ as follows:
For all u ∈ W 1,p (0, T ; E) and g ∈ Sn u, there exists f ∈ Au such that
1
(g, v) = (u0 (·), v(·)) + (f, v(·)) + (Ju, v) for all v ∈ W 1,p (0, T ; E), (6.7.1)
n
where J : W 1,p (0, T ; E) → W ∗ is the duality mapping.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 161

Lemma 6.7.2. Sn defined by (6.7.1) is an operator of class (S+ ) for each

positive integer n.
Proof. The bounded closed convexity of Sn u is obvious. To prove the finite
dimensional weak upper semicontinuity of Sn , we show that Sn is actually
upper semicontinous from the strong topology in W 1,p (0, T ; E) to the weak
topology in W ∗ . Note that the condition (2) implies that Sn is bounded and
so it is relatively weakly compact in W ∗ . Hence, we only need to show that
Sn is weakly closed (see [12]). Let un → u0 in W 1,p (0, T ; E), gn ∈ Sun and
gn * g0 in W ∗ .
Now, we prove that g0 ∈ Sn u0 . In fact, there exists fn ∈ Aun such that
1
fn * g0 − u00 − Ju0
n
and there exists a subsequence (unk ) such that unk (t) → u0 (t) for almost
everywhere t ∈ [0, T ]. Therefore, it follows from the condition (2) that

lim (fnk (t), unk (t) − u0 (t)) = 0

nk →∞

for almost everywhere t ∈ [0, T ] and A(t) is a mapping of class (S+ ) and so
(fnk (t)) has a subsequence converging weakly to f0 (t) ∈ A(t)u0 (t) for almost
everywhere t ∈ [0, T ]. Hence we have g0 ∈ Sn u0 .
Let uj ∈ W 1,p (0, T ; E) with uj * u0 in W 1,p (0, T ; E) and gj ∈ Sn uj such
that
limj→∞ (gj , uj − u0 ) ≤ 0.
Then, by (6.7.1), there exist fj ∈ Auj such that
Z T h 1 i
limj→∞ (fj , uj − u0 ) + (Juj , uj − u0 ) dt ≤ 0. (6.7.2)
0 n

Now, using the fact that A(t) is a mapping of class (S+ ), the conditions (2)
and (3), we have
limj→∞ (fj (t), uj (t) − u0 (t)) ≥ 0 (6.7.3)
for almost everywhere t ∈ [0, T ] and thus uj → u0 in W 1,p . There is a
subsequence (ujk ) such that ujk (t) → u0 (t) for almost everywhere t ∈ [0, T ].
Thus, condition (2) implies that

lim (fjk (t), ujk (t) − u0 (t)) = 0

j→∞

for almost everywhere t ∈ [0, T ] and A(t) is a mapping of class (S+ ) and so
(fjk ) has a subsequence (fjk0 ) such that fjk0 (t) * f0 (t) ∈ A(t)u0 (t). Hence,
(gjk0 ) converges weakly to g0 ∈ Sn u0 , where g0 satisfies

1
(g0 , v) = (u00 , v) + (f0 , v) + (Ju0 , v) for all v ∈ W 1,p (0, T ; E).
n

Copyright 2006 by Taylor & Francis Group, LLC

 162 Topological Degree Theory and Applications

This completes the proof.

Proof of Theorem 6.7.1. First, by the condition (3), we have
Z T Z T
1 T
Z
(g, u) ≥ α kukp dt − γ(t)dt + [kukp + ku0 kq∗ ]dt
0 0 n 0

for all g ∈ Sn u and so there exists r0n > 0 such that

(g, u) > 0 for all u ∈ W 1,p (0, T ; E), kukw = r0n , g ∈ Sn u.

It is easy to prove that tSn + (1 − t)J is a homotopy of operators of class (S+ )

for all t ∈ [0, 1]. By Theorem 6.2.4, we get

deg(Sn , B(0, r0n ), 0) = deg(J, B(0, r0n ), 0) = 1,

where B(0, r0n ) is the open ball with radius r0n in W 1,p (0, T ; E). Therefore,
0 ∈ Sn u has a solution un ∈ B(0, r0n ), i.e., there exists fn (t) ∈ A(t)un (t) for
almost everywhere t ∈ [0, T ] such that
1
(u0n , v) + (fn , v)dt + (Jun , v) = 0 for all v ∈ W 1,p (0, T ; E). (6.7.4)
n
Put v = un in (6.7.4). Then, by the condition (3), we know that there exists
N > 0 such that
1
kun kp ≤ N, √ ku0n k∗,q ≤ N for all n ≥ 1.
n
By the condition (2), u0n is bounded in Lq (0, T ; E ∗ ).
Now, we may assume that un * u0 in W 1,p (0, T ; E). Again, by the condi-
tion (2), we know that (fn ) is bounded in Lq (0, T ; E ∗ ), and so we may assume
that fn * f0 in Lq (0, T ; E ∗ ). Let v = un − u0 . It follows from (6.7.4) that
Z T
lim (fn (s), un (s) − u0 (s))ds = 0.
n→∞ 0

Since A(t) is a mapping of class (S+ ), it is easy to show that (fn (s), un (s) −
u0 (s)) → 0 in measure and hence there exists a subsequence ((fnk , unk − u0 ))
such that (fnk (s), unk (s) − u0 (s)) → 0 for almost everywhere s ∈ [0, T ]. Thus
unk (s) → u0 (s) for almost everywhere s ∈ [0, T ] and fnk converges weakly to
f0 (s) ∈ A(s)u(s). By letting nk → ∞ in (6.7.4), we get

u00 (s) ∈ A(s)u(s) for almost everywhere s ∈ (0, T ).

It is obvious that u0 (0) = u0 (T ). This completes the proof.

Example 6.7.3. Let Ω ⊂ RN be a bounded domain with a smooth bound-
ary ∂Ω. Let ai (x, u, ξ) : Ω × R × RN → R be continuous with respect to u, ξ
and measurable with respect to x for 1 ≤ i ≤ N . Suppose that

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 163

(1) |ai (x, u, ξ)| ≤ L(|u|m1 + kξkm−1 ) + M (x) for 1 ≤ i ≤ mn, where L > 0,
m1 < N N −m , 2 ≤ m < N are constants and M (x) ∈ L
m−1 (Ω);

(2) ρ is continuous on R and

 Z u r
0 ≤ ρ(u) ≤ µ k ρ(s)dsk + 1 for all u ∈ R,
0

N
where µ > 0 and 0 ≤ r < N −2 are constants;
(3) Σni=1 [ai (x, u, ξ) − ai (x, u, η)](ξi − ηi ) ≥ kkξ − ηkm ;
(4) f (x, z) : Ω × R → R satisfy the following conditions:
(4a) There exists Ω0 ⊂ Ω with mes(Ω0 ) = 0 such that
[
Df = {z ∈ R : f (x, .) is discontinuous at z}
x6∈Ω0

has measure zero;

(4b) x → f (x, z) is measurable for all z 6∈ Df ;
(4c) There exist k > 0 and l > 0 such that

|f (x, z)| ≤ k|z|m−1 + l for all z ∈ Dfc , a.e.x ∈ Ω.

We consider the boundary value problem with discontinuous nonlinearity:

(
∂ ∂u ∂u
ΣN 2
i=1 ∂xi [ρ (u) ∂xi + ai (x, u, ∂x )] + f (x, u) = 0, a.e.x ∈ Ω.
(E 6.7.2)
u(x) = 0, x ∈ ∂Ω.

Choose Ω1 ⊂ Ω such that mesΩ1 = 0 and (4c) holds for all x ∈ Ω1 . By

(4a), there exists a countable subset D ⊂ Dfc satisfying D = R. For all
(x, z) ∈ Ω × R, we define
(T
−1 , z + k −1 ] ∩ D)), x 6∈ Ω ,
k∈N co(f (x, [z − k 1
F (x, z) =
R, otherwise.

It is known that F (x, z) is nonempty closed convex and compact for all x ∈ Ωc1
(see [58]). Moreover, F (x, z) is measurable in x and z → F (x, z) is upper
semicontinuous for almost all x ∈ Ω.
q
Next, we define a mapping Ψ : W01,m (Ω) → 2L (Ω) by

Ψ(u) = {v ∈ Lq (Ω) : v(x) ∈ F (x, u(x)), a.e.x ∈ Ω}

for all u ∈ W01,m (Ω), where m1

+ 1q = 1. By Chen, Cho, and Yang [58], we
know that Ψ(u) is well defined and closed convex and the following properties
hold:

Copyright 2006 by Taylor & Francis Group, LLC

 164 Topological Degree Theory and Applications

(i) un → u ∈ W01,p (Ω) and vn ∈ Ψt (un ), vn * v in Lq (Ω) imply that

v ∈ Ψ(t, u);
(ii) Ψ maps bounded subsets of W01,p (Ω) to weakly relatively compact sub-
sets of Lq (Ω).
Finally, we define a mapping A : W01,m (Ω) → W −1,m by
Z n h ∂u  ∂u i ∂v(x) o
(Au, v) = Σni=1 ρ2 (u) + ai x, u, dx
Ω ∂xi ∂x ∂xi

for all u ∈ D(A) and v ∈ W01,m (Ω), where

n ∂u m
o
D(A) = u ∈ W01,m (Ω) : ρ2 (u) ∈ L m−1 .
∂x
Since F (x, z) = {f (x, z)} for all x ∈ Ωc1 , z ∈ Dfc , the equation (E 6.7.2) is
equivalent to the following equation:

0 ∈ Au + Ψu, u ∈ W01,m (Ω).

Theorem 6.7.4. A + Ψ is a mapping of class (S+ )0,L , where L = C0∞ (Ω).

1,m
Proof. Let (Fj )∞
j=1 be a sequence of finite dimensional subspace of W0 (Ω)
∞
such that L ⊂ ∪j=1 Fj . Let (uj ) ⊂ L satisfy

lim sup(gj , uj ) ≤ 0, lim (gj , v) = 0 for all v ∈ ∪∞

j=1 Fj ,
j→∞ j→∞

where gj ∈ Auj + Ψuj . Let gj = Auj + fj , where fj ∈ Ψuj . We have

∂(uj − v) m
Z
k | | dx
Ω ∂x
N Z
X ∂uj ∂v ∂uj ∂v
≤ [ai (x, uj , ) − ai (x, uj , )]( − )dx
i=1 Ω
∂x ∂x ∂x ∂x (6.7.5)
= (gj , uj − v) − (fj , uj − v)
N Z
X ∂uj ∂v ∂uj ∂v
− [ρ2 (uj ) + ai (x, uj , )]( − )dx.
i=1 Ω
∂x ∂x ∂x ∂x

Put v(x) ≡ 0. Then, by (1) and (4c), there exists a constant M > 0 such that
Z
∂uj 2
lim sup ρ2 (uj )| | dx ≤ M. (6.7.6)
j→∞ Ω ∂x

By (6.7.6), we may also assume that

Z
∂uj 2
lim ρ2 (uj )| | dx = M1 (6.7.7)
j→∞ Ω ∂x

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 165

by taking a subsequence, where M1 is a constant. Set

Z u
ρ̃(u) = ρ(s)ds, u˜j (x) = ρ̃(uj (x)).
0

Then u˜j ∈ W01,2 (Ω). We may assume that (u˜j ) converges strongly in Lp (Ω),
p < N2N−2 , to some u˜0 . Otherwise, take a subsequence. From this, we obtain
that u˜j converges in measure to u˜0 and ρ(u˜ 0 ). Consequently, we have

u˜0 (x) = ρ(u0˜)(x). (6.7.8)

2N
By the assumption (2) and the boundedness of (u˜j ) in L 2N −2 (Ω), we get
2N
(ρ(uj )) is bounded in L 2N −2 (Ω). Since ρ(uj ) converges in measure, we obtain
2
ρ(uj ) → ρ(u0 ) in L (Ω).
Now, we can pass to the limit in (6.7.5) for all v ∈ ∪∞ j=1 Fj to obtain

∂(uj − v) m
Z
k | | dx
Ω ∂x
∂u0 ∂v
≤ −M1 − (f0 , u0 − v) + ΣN 2
i=1 ρ (u0 ) dx
∂xi ∂xi
N Z (6.7.9)
X ∂v ∂(u0 − v)
− ai (x, u0 , ) dx ≤ −(f0 , u0 − v)
i=1 Ω
∂x ∂xi
N Z
X ∂u0 ∂v ∂(u0 − v)
− [ρ2 (u0 ) + ai (x, u0 , )] dx,
i=1 Ω
∂x i ∂x ∂xi

where f0 ∈ Ψu0 . By taking limits in (6.7.9), we know that (6.7.9) is also true
for all v ∈ W01,q (Ω) for q = 2[1 − NN−2 r]−1 , where r is the same as in the
assumption (2).
Now, consider the following functional on W01,q (Ω) defined by
Z
∂u0 ∂φ
h(φ) = ΣN
i=1 ρ2 (u0 ) dx. (6.7.10)
Ω ∂xi ∂xi

It follows from (6.7.9) that h is bounded on W01,q (Ω). Consequently, h can be

extended to a linear functional on W01,m (Ω) and we denote this extension by
0
h̃. Notice that the Laplace operator ∆ : W01,m (Ω) → (W01,m (Ω))∗ , m0 = m−1
m
,
0
is a homeomorphism and there exists u0 ∈ W01,m (Ω) such that
∂u0 ∂φ
Z
h̃(φ) = ΣN
i=1 dx for all φ ∈ W01,m (Ω). (6.7.11)
Ω ∂xi ∂xi

From (6.7.10) and (6.7.11), we get

∂(u00 (x) − u0 (x)) ∂φ
Z
ΣNi=1 dx = 0 for all φ ∈ W01,m (Ω), (6.7.12)
Ω ∂xi ∂xi

Copyright 2006 by Taylor & Francis Group, LLC

 166 Topological Degree Theory and Applications
R u0 (x)
where u00 (x) = 0
ρ2 (s)ds. Thus u00 (x) = u0 (x) and, consequently, we have

u0 (x) u0 (x) 0
ρ2 (u0 ) = ∈ Lm (Ω). (6.7.13)
∂xi ∂xi

Therefore u0 ∈ D(A). Notice that (6.6.9) is also true for all v ∈ W01,m (Ω). Put
v(x) = u0 (x) + tw(x) for all t > 0 and w ∈ W01,m (Ω) in (6.7.9). Then, divide
the resulting inequality by t and pass to the limit. We obtain 0 ∈ Au0 + Ψu0 .
Finally, we prove that uj → u0 ∈ W01,m (Ω). We may assume that

uj (x) 0
ai (x, uj (x), ) * bi (x) in Lm (Ω).
∂x
1,m
Take u0j ∈ ∪∞ 0
j=1 Fj such that uj → u0 in W0 (Ω). By using the assumption
(3), we have

∂(uj − u0j ) m
Z
k | | dx
Ω ∂x
∂u0j ∂uj ∂u0j
Z
∂uj
≤ ΣNi=1 [ai (x, u j , ) − ai (x, u j , )]( − )dx
Ω ∂x ∂x ∂x ∂x
Z
∂uj 2
= (gj , uj ) − ρ2 (uj )| | dx − (fj , uj ) (6.7.14)
Ω ∂x
∂uj ∂u0j
Z
− ΣN i=1 ai (x, uj , ) dx
Ω ∂x ∂x
∂u0j ∂(uj − u0j )
Z
− ΣN i=1 ai (x, uj , ) dx.
Ω ∂x ∂x

By letting j → ∞ in (6.7.14), we get

∂(uj − u0j ) m
Z
k | | dx
Ω ∂x
Z Z (6.7.15)
2 ∂u0 2 N ∂u0 (x)
≤ −(f0 , u0 ) − ρ (u0 )| | dx − Σi=1 hi (x) dx.
Ω ∂x Ω ∂xi

Now, using the fact limj→∞ (Auj + fj , v) = 0 for all v ∈ ∪∞

j=1 Fj , we obtain
Z
∂u0 (x) ∂v(x)
(f0 , v) + ΣN
i=1 [ρ2 (u0 ) + hi (x)] dx = 0 (6.7.16)
Ω ∂xi ∂xi

for all v ∈ ∪∞
j=1 Fj . From (6.7.16) and (6.7.15), it follows that uj → u0 ∈
W01,m (Ω). This completes the proof.
From Theorem 6.7.4, we know that, if for some open bounded subset U ⊂
W01,m (Ω) and 0 ∈/ Au+Ψu for all u ∈ ∂U ∩D(A), then deg(A+Ψ, U ∩D(A), 0)
is well defined.

Copyright 2006 by Taylor & Francis Group, LLC

 DEGREE THEORY FOR MONOTONE-TYPE MAPS 167

Corollary 6.7.5. If deg(A + Ψ, U ∩ D(A), 0) 6= 0, then the problem (E

6.7.2) has a solution.
Remark. Some results on degree theory in this chapter can also be given
in locally convex space.

6.8 Exercises
∗
1. Let A : D(A) ⊆ E → 2E be a set-valued mapping of class (S+ ) and
∗
P : E → 2E be a bounded pseudomonotone mapping. Show that P +A
is a mapping of class (S+ ).
∗
2. Let A : D(A) ⊆ E → 2E be a multi-valued mapping of class (S+ ),
∗
T : D(A) → 2E be an upper semicontinuous operator with closed
convex values and T maps each bounded subset of D(A) into a relatively
compact subset of E ∗ . Show that T + A is a multi-valued mapping of
class (S+ ).
∗
3. Let A : D(A) ⊆ E → 2E be a multi-valued mapping of class (S+ ) and
∗
M : D(M ) ⊆ E → 2E be a maximal monotone operator. Show that
Mλ + A is a multi-valued mapping of class (S+ ).
∗
4. Let M : D(M ) ⊆ E → 2E be a maximal monotone operator and
∗
P : E → 2E be a bounded pseudomonotone mapping. Suppose that

(g, x − x0 )
lim = +∞ for some x0 ∈ D(M ) and all g ∈ P x.
kxk→∞ kxk

Show that (M + P )(D(M )) = E ∗ .

5. Let φ : D(φ) ⊆ E → R = R ∪ {+∞} be a proper lower semicontinuous

convex function and

∂φ(x) = {f ∈ E ∗ : φ(y) ≥ φ(x) + f (y − x) for all y ∈ D(φ)}.

Show that ∂φ is maximal monotone.

6. Let φ : D(φ) ⊆ E → R = R ∪ {+∞} be a proper lower semicontinuous

convex function and
1
φλ (x) = inf ( kx − yk2 + φ(y)).
y∈D(φ) 2λ

Show that (∂φ)λ = ∂φλ and φλ (x) → φ(x).

Copyright 2006 by Taylor & Francis Group, LLC

 168 Topological Degree Theory and Applications

7. Let φ : D(φ) ⊆ E → R = R ∪ {+∞} be a proper lower semicontinuous

convex function. Show that ∂φ is surjective and (∂φ)−1 is bounded if
and only if limkxk→∞ φ(x)
kxk = +∞.
∗
8. Let M : D(M ) ⊆ E → 2E be a linear monotone operator. Show that
M is single valued if D(M )) = E, and M is maximal monotone if and
only if Graph(M ) is closed and M ∗ is monotone.
∗
9. Let M : D(M ) ⊆ E → 2E be a maximal monotone mapping. Show
that M is surjective if and only if M −1 is locally bounded.
10. Construct the degree theory for a bounded pseudomonotone mapping.
11. Construct the degree theory for a bounded generalized pseudomonotone
mapping.
12. Construct the degree theory for the sum of a bounded pseudomontone
mapping and a maximal monotone mapping.
13. Construct the degree theory for the sum of a bounded generalized pseudo-

monotone mapping and a maximal monotone mapping.

14. Construct the degree theory for the sum of a maximal monotone map-
ping and an upper semicontinuous compact mapping with closed convex
values.
15. Let H be a real Hilbert space and L : D(L) ⊆ H → H be a linear, one
to one maximal monotone mapping. Compute deg(L, D(L) ∩ B(0, r), 0)
for all r > 0.
16. Let H be a real Hilbert space, L : D(L) = H → H be a linear one-to-
one maximal monotone mapping and φ : D(φ) ⊆ H → R be a proper
lower semicontinuous convex function such that φ(0) = 0 ≤ φ(x) for all
x ∈ D(φ). Compute deg(L + ∂φ, B(0, r), 0) for all r > 0.

Copyright 2006 by Taylor & Francis Group, LLC

 Chapter 7
FIXED POINT INDEX THEORY

The study of non-negative solutions to the nonlinear equations, especially

the ordinary or partial differential equations and integral equations, is a very
important problem in nonlinear functional analysis. The non-negativity con-
dition can be described by a closed convex subset P in a Banach space which
satisfies λP ⊂ P for all λ ≥ 0 and P ∩ −P = {0}. We are interested in solving
the equation T x = y in P . The fixed point index theory has proved to be a
useful tool in studying such an equation.
It is our goal to introduce this theory in this chapter. Chapter 7 has six
main sections.
Some introductory material on cones (normal, minihedral, etc.) is presented
in Section 7.1.
In Section 7.2, we present a fixed point index for countably condensing
mappings based on Dugundji’s extension theorem and the Leray Schauder
degree. Various properties of this index are given in Theorem 7.2.2.
Section 7.3 presents a variety of fixed point theorems for mappings defined
in cones of Banach spaces (see, in particular, theorems 7.3.1, 7.3.3, 7.3.7, and
7.3.13).
The results in Section 7.3 are used in Section 7.4 to prove various fixed
point results for perturbations of condensing maps (see Theorem 7.4.1 and
7.4.3).
In Section 7.5, we present a fixed point index for continuous generalized
inward mappings.
Finally, some applications to integral and differential equations are given in
Section 7.6.

7.1 Cone in Normed Spaces

We begin this section by first introducing the concept of partial order.
Definition 7.1.1. Let X be a nonempty set. There is an equivalent relation
≤ such that
(1) If x ≤ y, y ≤ z for some x, y, z ∈ X, then x ≤ z;

169

Copyright 2006 by Taylor & Francis Group, LLC

 170 Topological Degree Theory and Applications

(2) x ≤ x for all x ∈ X;

(3) If x ≤ y and y ≤ x for some x, y ∈ X, then x = y.

Then ≤ is said to be a partial order on X and (X, ≤) is said to be a partially

ordered set.
Let (X, ≤) be a partially ordered set and M ⊂ X. For any x, y ∈ M , if
x ≤ y or y ≤ x holds, then we say M is a well-ordered set. If x∗ ∈ X satisfies
y ≤ x∗ (respectively, x∗ ≤ y) for all y ∈ D, then x∗ is said to be an upper
bound (respectively, lower bound) of M . If x0 ∈ X satisfies x0 ≤ y for some
y ∈ X and y = x0 , then x0 is said to be a maximal element of X. Similarly,
if x0 ≥ y for some y ∈ X and x0 = y, then x0 is said to be a minimal element
of X.
Theorem 7.1.2. (Zorn Lemma) If every well-ordered set in X has an
upper bound (lower bound), then X has a maximal element (minimal ele-
ment).
A cone is a very useful concept that can be used to generate a partial order
in a linear space. Usually, this method is easy to manipulate and has been
widely used in searching for positive solutions of nonlinear equations (see [8],
[176]). Now, we recall this concept as follows:
Definition 7.1.3. Let E be a linear vector space and P be a nonempty
convex subset of E. Then P is called a cone if

(1) λx ∈ P for all x ∈ P and λ > 0;

(2) P ∩ (−P ) = {0}.

If E is a linear space and P ⊂ E is a cone, we define an order ≤ on E as

follows:
x ≤ y if and only if y − x ∈ P.
Note that ≤ is a well-defined partial order on X.
Example 7.1.4. Let P ⊂ Rn be given by P = {(x1 , x2 , · · · , xn ) ∈ Rn :
xi ≥ 0, i = 1, 2, · · · , n}. Then P is a cone.
Example 7.1.5. Let Ω be a nonempty Lebesgue measurable set in Rn and
R0 < m(Ω) < ∞. Suppose that Lp (Ω) = {f (·) : Ω → R is measurable and
p
Ω
|f (x)| dx < ∞}, where 0 < p < 1. Put

P = {f ∈ Lp (Ω) : f (x) ≥ 0 for almost everywhere x ∈ Ω}.

Then P is a cone in Lp (Ω).

The following well-known order principle is due to Brezis and Browder [30]:

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 171

Theorem 7.1.6. Let (X, ≤) be a partially ordered set and S(x) = {y ∈

X : x ≤ y}. Suppose that ψ : X → R is a function satisfying the following
conditions:
(1) For x ≤ y with x 6= y, ψ(x) ≤ ψ(y);
(2) For x1 ≤ x2 ≤ · · · ≤ xn ≤ · · · , if {ψ(xn )} bounded, then there exists
y ∈ X such that xn ≤ y;
(3) ψ(S(x)) is bounded for all x ∈ X.
Then, for all x ∈ X, there exists z ∈ S(x) such that z is maximal in X.
Proof. For each x ∈ X, we prove that sup S(x) exists and it belongs to
S(x). To see this, let F ⊂ S(x) be a well-ordered set. Set α = supx∈F φ(x).
Then there exists a sequence (yn ) ⊂ F such that φ(y1 ) ≤ φ(y2 ) ≤ · · · → α.
By the assumption (1), we must have y1 ≤ y2 ≤ y3 ≤ · · · . By assumption
(2), there exists y ∈ X such that yn ≤ y. Notice that x ≤ yn , so we have
x ≤ y. Thus y ∈ S(x) and φ(y) ≥ α, so y ≥ z for all z ∈ F . Therefore, y is
an upper bound of F in S(x). By the Zorn Lemma, sup S(x) exists in S(x).
Assume that z ∈ X such that sup S(x) ≤ z. Then z ∈ S(x), so we must have
z = sup S(x). Thus sup S(x) is a maximal element. This completes the proof.
Definition 7.1.7. Let X be a normed space, P ⊂ X be a cone and ≤ be
the partial ordering defined by P . Then
(1) P is said to be normal if inf{x + yk : x, y ∈ P, kxk = kyk = 1} > 0;
(2) P is said to be quasinormal if there exist y ∈ P with y 6= 0 and σ(y) > 0
such that kx + yk ≥ σ(y)kxk for any x ∈ P .

We set σ(y) = inf{ kx+yk

kxk : x ∈ P, x 6= 0} for y ∈ P with y 6= 0 and define
σ = sup{σ(y) : y ∈ P, y 6= 0}, which is called the quasinormality of P . Then
we have 21 ≤ σ ≤ 1 (see [180], [246]).
Let P ⊂ E be a cone and D ⊂ E be a nonempty set. If y, z ∈ E satisfies
the following:
(i) x ≤ y for all x ∈ D;
(ii) x ≤ z for all x ∈ D and so y ≤ z;
then y is said to be an supremum of D and we denote it by sup D.
Similarly, if (i) x ≥ y for all x ∈ D, (ii) x ≥ z for all x ∈ D and so y ≥ z,
then y is said to an infimum of D and we denote it by inf D.
Definition 7.1.8. Let X be a normed space, P ⊂ X be a cone and ≤ be
the partial ordering defined by P . Then
(1) P is said to be reproducing if P − P = X and total if P − P = X;

Copyright 2006 by Taylor & Francis Group, LLC

 172 Topological Degree Theory and Applications

(2) P is said to be regular if every increasing sequence which is bounded

from above is already convergent, and is fully regular if every bounded
increasing sequence is convergent;
(3) P is said to be lastering if there exist a cone P1 and a constant α > 0
such that B(x, αkxk) ⊂ P1 for all x ∈ P \ {0};
(4) P is said to be minihedral if sup{x, y} exists for all x, y ∈ X and strongly
minihedral if every set which is bounded from above has a supremum.

Definition 7.1.9. Let X be a normed space, P ⊂ X be a cone and ≤ be

the partial ordering defined by P . Then
(1) The norm k · k is said to be monotonic if 0 ≤ x ≤ y implies that
kxk ≤ kyk;
(2) The norm k · k is said to be semimonotonic if 0 ≤ x ≤ y implies that
kxk ≤ αkyk for some α > 0.

Proposition 7.1.10. Let X be a Banach space and P ⊂ X be a cone.

Then we have
(1) P is quasinormal;
(2) P is normal if and only if k · k is semimonotonic;
(3) If P 0 6= ∅, then P is reproducing.

Proof. (1) Suppose that P is not quasinormal. Then, for any y ∈ P with
kyk = 1, there exist xn ∈ P such that

kxn + yk < n−1 kxn k for n = 1, 2, · · · .

Therefore, if kxn k is bounded, then xn + y → 0, i.e., xn → −y ∈ P , which is a

contradiction. If kxn k is unbounded, then kxn k−1 xn → 0, which is impossible.
Therefore, P is quasinormal.
(2) If k · k is semimonotonic, then, for any x, y ∈ P with kxk = kyk = 1,
we have kx + yk ≥ α−1 kyk = α−1 , and so P is normal. On the other hand, if
P is normal and k · k is not semimonotonic, then there exist xn and yn such
that 0 < xn ≤ yn and kxn k > nkyn k for n = 1, 2, · · · . Put

vn = kxn k−1 xn , wn = kyn k−1 yn , zn = kn−1 wn − vn k−1 (n−1 wn − vn ).

Then we have zn ∈ P and kvn + zn k → 0 as n → ∞, i.e., P is not normal,

which is a contradiction.
(3) Take x0 ∈ P 0 and r > 0 such that B(x0 , 2r) ⊂ P . For any x ∈ X with
x 6= 0, we have x0 + rkxk−1 x ∈ P . Moreover,

x = kxkr−1 (x0 + rkxk−1 x) − kxkr−1 x0 ∈ P − P,

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 173

thus P is reproducing. This completes the proof.

By Proposition 7.1.10, if P is normal, then we set

σ(P ) = min{N : if 0 ≤ x ≤ y, then kxk ≤ N kyk},

which is called the normality of P .

The following result gives a relation between various cones:
Proposition 7.1.11. Let X be a Banach space and P ⊂ X be a cone.
Then we have the implications: P allows plastering ⇒ P is fully regular ⇒
P is regular ⇒ P is normal.
Proof. (1) Assume that P1 is a plastering cone of P . Take x0 ∈ / P1 ; then
there exists r0 > 0 such that B(x0 , r0 ) ∩ P1 = ∅. By Mazur’s separation
theorem for convex sets, there exist f ∈ X ∗ and β ∈ R such that f (y) ≥ β for
all y ∈ P1 and f (y) < β for all y ∈ B(x0 , r0 ). This and λx ∈ P1 for all λ > 0
and x ∈ P1 imply that f (y) > 0 for all y ∈ P1 . But B(x, αkxk) ⊂ P1 for all
x ∈ P , so we have f (x) ≥ αkxkkf k.
Assume that x1 ≤ x2 ≤ · · · ≤ xn · · · and (xn )∞ n=1 is bounded. Then we
have
f (xn+m − xn ) ≥ αkxn+m − xn kkf k.
Thus (xn )∞ ∞
n=1 is a Cauchy sequence and, consequently, (xn )n=1 is convergent.
(2) To show that P is fully regular implies that P is regular, we first show
that P is fully regular implies that P is normal. Assuming this is false, there
exist xn , yn ∈ P ∩ ∂B(0, 1) such that kxn + yn k ≤ 21n for n = 1, 2, · · · . Put

z2n = Σ2n
i=1 (xi + yi ), z2n+1 = z2n + x2n+1 ,

then (zn ) is increasing and bounded, so (zn ) is convergent, which contradicts

kz2n+1 − z2n k = 1.
Now, suppose that x1 ≤ x2 ≤ · · · ≤ xn ≤ y. Then y −xn ≤ y −x1 for n > 1,
so we have ky − xn k ≤ αky − x1 k for some α > 0. Thus (xn ) is convergent.
(3) Assume that P is regular, but P is not normal. Then there exist xn , yn ∈
P ∩ ∂B(0, 1) such that kxn + yn k ≤ 21n for n = 1, 2, · · · . Then we have

zn = Σni=1 xi ≤ Σ∞
i=1 (xi + yi ) ∈ P

and (zn ) is increasing, so (zn ) is convergent, which contradicts kzn − zn−1 k =

kxn k = 1. Thus P is regular ⇒ P is normal. This completes the proof.
Proposition. 7.1.12. Let X be a normed space and P ⊂ E be a regular
cone. Then we have the following:
(1) If D ⊂ E is a well-ordered set and has an upper bound, then sup D
exists.
(2) If D ⊂ E is a well-ordered set and has a lower bound, then inf D exists.

Copyright 2006 by Taylor & Francis Group, LLC

 174 Topological Degree Theory and Applications

Proof. (2) follows from (1) by replacing D by −D.

Now, we prove (1). By Proposition 7.1.11, E is normal. Let x0 be an upper
bound of D. Define a function φ : D → [0, +∞) as follows:

φ(x) = sup{kz − yk : x ≤ y ≤ z, y, z ∈ D}.

Note that z−y ≤ x0 −x and we have φ(x) ≤ σ(P )kx0 −xk. Obviously, if x ≤ y,
then φ(x) ≥ φ(y). We claim that inf x∈D φ(x) = 0. If not, inf x∈D φ(x) = δ >
0, there exists x1 ∈ D such that φ(x1 ) > δ, so there are x1 ≤ y1 ≤ z1 such
that kz1 − y1 k > δ. Again, since φ(z1 ) > δ, there are z1 ≤ y2 ≤ z2 such that
kz2 − y2 k > δ. Repeat the above process, we get a sequence (yn ) in D such
that
y1 ≤ z1 ≤ y2 ≤ z2 ≤ · · · ≤ x0
with kzi − yi k > δ for i = 1, 2, · · · , so it does not converge, which contradicts
the fact that P is regular. Thus there exist (xn )∞ n=1 ⊂ D such that φ(xn ) → 0.
Since D is well ordered, put yn = max{x1 , x2 , · · · , xn }. Then y1 ≤ y2 ≤ · · · ≤
x0 and so φ(yn ) → 0. Since P is regular, there exists y0 ∈ E such that
yn → y0 . It is easy to see that y0 = sup{yi : i = 1, 2, · · · }.
Next, we prove that y0 = sup D. To see this, we only need to prove that y0
is an upper bound of D. For any x ∈ D, we have two cases: (1) x ≤ yn for
some n and (2) yn ≤ x.
In case (1), we have x ≤ y0 .
In case (2), we have y0 ≤ x and 0 ≤ x − y0 ≤ x − yn and thus

kx − y0 k ≤ N kyn − xk ≤ N φ(yn ) → 0 as n → ∞.

Thus x = y0 . Combine (1) and (2), then we have x ≤ y0 , so y0 is an upper

bound of D. Thus y0 = sup D. This completes the proof.
Proposition 7.1.13. Let P ⊂ E be a cone. If P is regular and minihedral,
then P is strongly minihedral.
Proof. Let D ⊂ E and D has an upper bound x0 . Set

U = {x : x is an upper bound of D},

then x0 ∈ U . Consider any well-ordered set F of U . Obviously, any element

in D is a lower bound for F . By Proposition 7.1.12, we know that f = inf F
exists and x ≤ f for all x ∈ D, and thus f ∈ U . By the Zorn Lemma, U has
a minimal element x∗ .
Now, we prove that x∗ = sup D. To see this, we only need to prove x∗ ≤ y
for all y ∈ U . For any y ∈ U , put z = inf{x∗ , y}. We have z ≤ x∗ , z ≤ y, but
x ≤ x∗ , x ≤ y. Thus we must have x ≤ z and so z ∈ U . Since x∗ is minimal,
we must have z = x∗ and hence x∗ ≤ y, as desired. This completes the proof.
Example 7.1.14. Let E = l∞ and P = {(xi ) ∈ l∞ : xi ≥ 0, i = 1, 2, · · · }.
Then P is normal. For any set D with a upper bound, it is easy to see

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 175

that sup D = z = (z1 , z2 , · · · ), where zi = supx∈D x(i) and x(i) is the i-th
coordinate of x for i = 1, 2 · · · . Thus, P is strongly minihedral. For any
integer n ≥ 1, let xn ∈ l∞ be defined by
(
n 1, if i ≤ n,
x (i) =
0, if i > n.

Then x1 ≤ x2 ≤ · · · ≤ (1, 1, 1, · · · , 1, · · · ). However, (xn ) does not converge

in l∞ . Thus, P is not regular.
Proposition 7.1.15. Let E be a Banach space, P ⊂ E be a cone and
B = {x : kxk ≤ 1}. The the following conclusions are equivalent:

(1) P is reproducing;

(2) There exists r > 0 such that, for any x ∈ E, there exist y, z ∈ P with
kyk ≤ rkxk, kzk ≤ rkxk and x = y − z;

(3) There exists α > 0 such that αB ⊂ B ∩ P − B ∩ P ;

(4) There exists β > 0 such that βB ⊂ B ⊂ B ∩ P − B ∩ P .

Proof. It is obvious that (2) is equivalent to (3), (2) implies (1), and (3)
implies (4), so we only need to prove (1) implies (2) and (4) implies (3).
(1) ⇒ (2) Since P is reproducing, we have E = ∪∞ n=1 En , where

En = {x ∈ E : there exists y ∈ P such that x ≤ y, kyk ≤ nkxk}

for n = 1, 2, · · · . By Baire’s Theorem, there exist an integer n0 > 0, x0 ∈ E,

r > s > 0 such that

T = {x ∈ E : s < kx − x0 k < r} ⊂ En0 .

Let y0 , z0 ∈ P be such that −x0 = y0 − z0 . Take an integer n1 > 0 such that

ky0 k ≤ n1 kx0 k. Set T0 = {x ∈ E : s ≤ kxk < r}. Take an integer n2 such
that
1
n2 > n0 + (n0 + n1 )kx0 k.
s
We prove that T0 ⊂ En3 . In fact, for any x ∈ T0 , we have y + x0 + x ∈ T ,
so there exist xi ∈ En0 , i = 1, 2, · · · , such that xi → y. We also assume
that xi ∈ T . There exist yi ∈ P such that xi ≤ yi and kyi k ≤ n0 kxi k for
i = 1, 2, · · · . We also have xi − x0 ≤ yi + y0 ∈ P and

kyi − y0 k ≤ n0 kxi k + n1 kx0 k ≤ (n0 + n1 )kx0 k + n0 kxi − x0 k

kx0 k
≤ [(n0 + n1 ) + n0 ]kxi − x0 k ≤ n2 kxi − x0 k.
s

Copyright 2006 by Taylor & Francis Group, LLC

 176 Topological Degree Theory and Applications

Therefore, xi − x0 ∈ En2 for i = 1, 2, · · · . Obviously, xi − x0 → y − x0 = x as

i → ∞, so x ∈ En2 .
Finally, we prove that E = E3n2 . For any x ∈ E \ {0}, take x1 ∈ En2 such
that
1
kx − x1 k < kxk.
2
There exist y1 ∈ P such that x1 ≤ y1 and ky1 k ≤ n2 kx1 k. Similarly, there
exist x2 ∈ En3 and y2 ∈ P such that
1
kx − x1 − x2 k < kxk, x2 ≤ y2 , ky2 k ≤ n2 kx2 k.
22
Repeating this process, we get two sequences (xk )∞ ∞
k=1 ⊂ En2 , (yk )k=1 ∈ P
such that
1
kx − x1 − x2 − · · · − xk k < kxk, xk ≤ yk , kyk k ≤ n2 kxk k
2k
3kxk
for k = 1, 2, · · · . Obviously, x = Σ∞i=1 xi and kxk k < 2k for k = 1, 2, · · · .
∞ ∞
Thus, Σi=1 kyi k ≤ 3n2 kxk. Put y = Σi=1 yi ; then x ≤ y and kyk ≤ 3n2 kxk.
Thus, x ∈ E3n2 .
(4) ⇒ (3) Set C = B ∩P −B ∩P . Let β > 0 be such that βB ⊂ C. We prove
that β2 B ⊂ C. For any y ∈ β2 B, there exists 2y1 ∈ C such that k2y−2y1 k < β2 .
Again, there exists 22 y2 ∈ C such that k22 y − 22 y1 − 22 y2 k < β2 . Repeat this
process, we get a sequence (yn )∞ n
n=1 ⊂ E such that 2 yn ∈ C and

β
ky − y1 − y2 − · · · − yn k < for n = 1, 2, · · · .
2n+1
Thus, we have

∞ 1 1
y = σi=1 yi , yn = xn − zn , xn , zn ∈ P, kxn k ≤ n
, kzn k ≤ n
2 2
for n = 1, 2, · · · . Put x = Σ∞ ∞
i=1 xi , z = Σi=1 zi . Then x, z ∈ P , kx|| ≤ 1,
β
kzk ≤ 1 and x ∈ C. Thus, 2 B ⊂ C. This completes the proof.

7.2 Fixed Point Index Theory

In this section, let E be a Banach space and P ⊂ E be a cone.
Lemma 7.2.1. [291] Let B ⊂ P be a bounded closed subset and T :
B → P a countably condensing mapping. Set C1 = conv(T B), Cn+1 =
conv(T (Cn ∩ B)) for n ≥ 1 and C = ∩∞ n=1 Cn . If M ⊂ E and M \ Cn is finite
for n = 1, 2, · · · , then M is relatively compact. In particular, C is compact.

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 177

Proof. Let F be the family of all subsets M such that M \ Cn is finite

for n = 1, 2, · · · and FB be the family of all countable subsets M ∈ F with
M ⊂ B.
Step 1 We prove that there exists B ∗ ∈ FB such that α(K) ≤ α(B ∗ )
for all K ∈ FB . In fact, since α(K) ≤ α(B) for all K ∈ FB , we have
s = supK∈FB α(K) < +∞, and let Kn ∈ FB such that α(Kn ) → s as n → ∞.
Put B ∗ = ∪∞ ∗ ∗
n=1 Kn , then B is countable and α(B ) = s ≥ α(K) for all
K ∈ FB .
Step 2 If M ∈ F, and xn ∈ M for n = 1, 2, · · · and no xn appears infinitely
many times, then there exist A ∈ FB and yn ∈ conv(T A) such that kxn −
yn k → 0 as n → ∞. To see this, observe that M \ C1 is finite and it contains
at most finitely many xn , and so we may assume that xn ∈ C1 for all n ≥ 1.
For any given integer n, let kn be the largest integer such that xn ∈ Ckn . If
no such integer exists, put kn = n. For any integer k, M \ Ck is finite, thus
{n : xn ∈
/ Ck } is finite. One can easily see that

Ik = {n : kn ≤ k} ⊂ Ik+1 ∪ {1, 2, · · · , k}

by virtue of C1 ⊃ C2 ⊃ · · · and thus Ik is finite for all k. We have

xn ∈ Ckn = conv(T (Ckn −1 ∩ B)) for any n.

Hence there exists yn ∈ conv((T (Ckn −1 ∩ B)) such that kxn − yn k < n−1 . In
particular, we can find some finite An ⊂ Ckn −1 ∩B such that yn ∈ conv(T An ).
Now, if we put A = ∪∞ n=1 An , then A is the required subset. To see this,
we have to check that A ∈ FB . Since C1 ⊃ C2 ⊃ · · · , we have An ⊂ Ci for
i ≤ kn − 1, and thus we have

A \ Cn = ∪∞
i=1 (Ai \ Cn ) = ∪i;n>ki −1 (Ai \ Cn ) ⊂ ∪i∈In Ai ,

but the last set is finite, so we get the desired result.

Step 3 We prove that any K ∈ FB is finite. For any K ∈ FB , if K is finite,
we are done. So we assume that K is infinite. Replace K by K ∪ B ∗ , we get
α(K) = s. The countability of K implies that K = {xn : n ≥ 1}.
By Step 2, there exist A ∈ FB and yn ∈ conv(T A) such that kxn − yn k → 0
as n → ∞. Obviously, K ∪ A ∈ FB and thus α(A ∪ K) = α(K) = s. Put

Kn = {x1 , x2 , · · · , xn , yn+1 , yn+2 }, K0 = {y1 , y2 , · · · }.

We have α(K0 ) = α(K) = s. On the other hand,

s = α(K0 ) ≤ α(conv(T A)) ≤ α(T (A ∪ K)).

But T is countably compact, so we must have s = 0. Consequently, K is

relatively compact.

Copyright 2006 by Taylor & Francis Group, LLC

 178 Topological Degree Theory and Applications

Step 4 We prove that any F ∈ F is relatively compact. For any sequence

(xn ) ⊂ F , if xn appears infinitely many times for some n, we are done already.
Otherwise, by Step 2, we may take yn and A as in Step 2. By Step 3, A is
relatively compact, so conv(T (A) is relatively compact. Thus {yi : i ≥ 1} is
relatively compact and so F is relatively compact. This completes the proof.
Construction of Index Theory. Let Ω ⊂ E be an open bounded subset
and Ω ∩ P 6= ∅. Let T : Ω ∩ P → P be a countably condensing and continuous
mapping. Put

C0 = conv(T (Ω ∩ P )), Cn = conv(T (Cn−1 ∩ Ω) for n ≥ 1,

then, by Lemma 7.2.1, C = ∩∞ n=1 Cn is compact and T : C → C is a mapping.

Now, assume also that T x 6= x for all x ∈ ∂Ω ∩ P . Then we have the following
two cases:
Case (1) If C = ∅, we define ind(T, Ω ∩ P ) = 0.
Case (2) If C 6= ∅ and r : E → C is a retraction, then r−1 (Ω ∩ C) is open
in E. It is easy to see that T rx 6= x for all x ∈ ∂(Ω ∩ r−1 (Ω)) and a mapping
T r : Ω ∩ r−1 (Ω) → C, so the Leray Schauder degree deg(I −T r, Ω∩r−1 (Ω), 0)
is well defined. We define

ind(T, Ω ∩ P ) = deg(I − T r, Ω ∩ r−1 (Ω), 0),

which is called the fixed point index of T .

A slight modification of the argument in Section 2 of Chapter 3 guarantees
that deg(I − T r, Ω ∩ r−1 (Ω), 0) does not depend on r, so ind(T, Ω ∩ P ) is well
defined.
Theorem 7.2.2. The fixed point index ind(T, Ω ∩ P ) satisfying the follow-
ing properties:

(1) ind(x0 , Ω ∩ P ) = 1 if x0 ∈ Ω ∩ P ;

(2) If ind(T, Ω ∩ P ) 6= 0, then x = T x has a solution in Ω;

(3) If Ωi ⊂ Ω for i = 1, 2, Ω1 ∩ Ω2 = ∅ and 0 ∈

/ (I − T )[(Ω \ (Ω1 ∪ Ω2 )) ∩ P ],
then
ind(T, Ω ∩ P ) = ind(T, Ω1 ∩ P ) + ind(T, Ω2 ∩ P );

(4) If H(t, x) : [0, 1] × Ω ∩ P → P is a continuous mapping satisfying

α(H([0, 1]×B)) < α(B) for any countable subset B of Ω∩P with α(B) 6=
0 and x 6= H(t, x) for all (t, x) ∈ [0, 1] × ∂Ω, then ind(H(t, ·), Ω ∩ P )
does not depend on t ∈ [0, 1].

Proof. It is similar to the proof of Theorem 3.2.1.

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 179

Note that, in the case that T : Ω ∩ P → P is continuous and compact and

x 6= T x for all x ∈ ∂Ω ∩ P , one can define the fixed point index by

ind(T, Ω ∩ P ) = deg(I − T r, r−1 (Ω ∩ P ) ∩ Ω, 0),

where r : E → P is a retraction. One can check easily this definition does not
depend on r and we have the following properties:
Theorem 7.2.3. The fixed point index ind(T, Ω∩P ) satisfies the following
properties:
(1) ind(x0 , Ω ∩ P ) = 1 if x0 ∈ Ω ∩ P ;
(2) If ind(T, Ω ∩ P ) 6= 0, then x = T x has a solution in Ω;
(3) If Ωi ⊂ Ω for i = 1, 2, Ω1 ∩ Ω2 = ∅ and 0 ∈
/ (I − T )[(Ω \ (Ω1 ∪ Ω2 )) ∩ P ],
then
ind(T, Ω ∩ P ) = ind(T, Ω1 ∩ P ) + ind(T, Ω2 ∩ P );

(4) If H(t, x) : [0, 1] × Ω ∩ P → P is a continuous compact such that x 6=

H(t, x) for all (t, x) ∈ [0, 1] × ∂Ω, then ind(H(t, ·), Ω ∩ P ) does not
depend on t ∈ [0, 1].

7.3 Fixed Point Theorems in Cones

In this section, we derive some fixed point theorems by using fixed point
index theory.
Theorem 7.3.1. Let E be a Banach space, P be a cone in E, Ω be
an open bounded subset of E, 0 ∈ Ω1 ⊂ Ω be an open subset of Ω, and
T : Ω \ Ω1 ∩ P → P be a completely continuous mapping. Suppose that one
of the following conditions is satisfied:
(1) T x ≥ x for all x ∈ ∂Ω ∩ P and T x ≤ x for all x ∈ ∂Ω1 ∩ P ;
(2) T x ≥ x for all x ∈ ∂Ω1 ∩ P and T x ≤ x for all x ∈ ∂Ω ∩ P .
Then T has a fixed point in Ω \ Ω1 ∩ P .
To prove Theorem 7.3.1, we first prove the following result:
Lemma 7.3.2. Suppose that C ⊂ P is a compact set such that 0 6∈ C.
Then 0 6∈ Conv(C).
Proof. Since 0 ∈
/ C and C is compact, there exists 0 > 0 such that

d(0, C) ≥ 0 .

Copyright 2006 by Taylor & Francis Group, LLC

 180 Topological Degree Theory and Applications

We also have C ⊂ ∪x∈C Bx (0 ), so there exist finitely many xi for i =

1, 2, · · · , n such that
C ⊂ ∪ni=1 Bxi (0 ).
Therefore, it follows that

Conv(C) = Conv(∪ni=1 Bxi (0 ) ∩ C).

If 0 ∈ Conv(C), then there exists yi ∈ Bxi (0 ) ∩ C, αi ∈ (0, 1) for i =

1, 2, · · · , n and Σni=1 αi = 1 such that

0 = Σni=1 αi yi ,

which contradicts yi > 0 for i = 1, 2, · · · , n. This completes the proof.

Proof of Theorem 7.3.1. We may assume that (1) holds (the proof is
similar when (2) holds). By our assumption, we have 0 6∈ T (∂Ω ∩ P ). In view
of Lemma 7.3.2, it follows that

0 6∈ Conv(T ∂Ω ∩ P )).

Hence, we can find a completely continuous mapping T1 : E → conv(T ∂Ω ∩ P ))

such that T1 x = T x for all x ∈ ∂Ω ∩ P .
Now, for any k > 1, we have tT x + (1 − t)kT1 x 6= x for all (t, x) ∈ [0, 1] ×
∂Ω ∩ P , and so by Theorem 7.2.3

ind(kT1 , Ω ∩ P ) = ind(T, Ω ∩ P ) for all k > 1.

Therefore, ind(T, Ω ∩ P ) must equal to 0 (otherwise, Ω is unbounded, which

is a contradiction). It is easy to see that ind(T, Ω1 ∩ P ) = 1, and so we have

ind(T, Ω \ Ω1 ∩ P ) = 1.

Therefore, T has a fixed point in Ω \ Ω1 ∩ P . This completes the proof.

By the same proof as in Theorem 7.3.1, we get the following more general
result:
Theorem 7.3.3. Let P be a cone in a Banach space E, Ω be an open
bounded subset of E, Ω1 ⊂ Ω be an open subset of Ω with 0 ∈ Ω1 , and
T : Ω \ Ω1 ∩ P → P be a completely continuous mapping. Suppose that one
of the following conditions is satisfied:

(1) T x 6= λx for all λ > 1 and x ∈ ∂Ω ∩ P ;

(2) T x 6= αx for all x ∈ ∂Ω1 ∩ P and α < 1;

(3) inf{kT xk : x ∈ ∂Ω1 ∩ P } > 0.

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 181

Then T has a fixed point in Ω \ Ω1 ∩ P .

From Theorem 7.3.3, we have the following:
Corollary 7.3.4. Let P be a cone in a Banach space E, Ω be an open
bounded subset of E, Ω1 ⊂ Ω be an open subset of Ω with 0 ∈ Ω1 and
T : Ω \ Ω1 ∩ P → P be a completely continuous mapping. Suppose that one
of the following conditions is satisfied:

(1) kT xk ≤ kxk for all x ∈ ∂Ω ∩ P ;

(2) kT xk ≥ kxk for all x ∈ ∂Ω1 ∩ P .

Then T has a fixed point in Ω \ Ω1 ∩ P .

Another variant form of Theorem 7.3.3 is as follows:
Theorem 7.3.5. Let P be a cone in a Banach space E, Ω be an open
bounded subset of E, Ω1 ⊂ Ω be an open subset of Ω with 0 ∈ Ω1 and
T : Ω \ Ω1 ∩ P → P be a completely continuous mapping. Suppose that one
of the following conditions is satisfied:

(1) T x 6≥ x for all x ∈ ∂Ω ∩ P ;

(2) T x 6≤ x for all x ∈ ∂Ω1 ∩ P ;

(3) inf{kT xk : x ∈ ∂Ω1 ∩ P } > 0.

Then T has a fixed point in Ω \ Ω1 ∩ P .

Theorem 7.3.3 fails if we assume that T is a k-set contraction with k < 1.
This can be seen with the following example:
Example 7.3.6. Let a mapping T : l2 → l2 be defined by
1
T (x1 , x2 , · · · ) = (0, x1 , x2 , · · · ) for all (x1 , x2 , · · · ) ∈ l2 .
3
Now, T is obviously 13 -set contraction. Take P = {(xi ) : xi ≥ 0 for i =
1, 2, · · · } and Ω = B(0, 2), Ω1 = B(0, 1). Then T has no fixed point in
Ω \ Ω1 ∩ P .
In the following, we provide sufficient conditions such that the conclusion
of Theorem 7.3.3 still holds in the case of k-set contractions.
Theorem 7.3.7. Let X be a real Banach space,Ω be an open bounded
subset of X with 0 ∈ Ω and P a quasinormal cone in X. If T : Ω ∩ P → P
is a continuous countably condensing mapping and the following condition is
satisfied: n kT xk o 1
inf : x ∈ ∂Ω ∩ P > ,
kxk σ

Copyright 2006 by Taylor & Francis Group, LLC

 182 Topological Degree Theory and Applications

where σ is the quasinormality constant of P , then ind(T, Ω ∩ P ) = 0.

Proof. For any  > 0, by the definition of the quasinormality constant of
P , there exists y ∈ P with y 6= 0 such that

kx + λy k ≥ (σ − )kxk for all x ∈ P.

By assumption, we may fix  > 0 such that

n kT xk o 1
inf : x ∈ ∂Ω ∩ P > . (7.3.1)
kxk σ−

For any m > 0, we claim that x − T x 6= tmy for all x ∈ ∂Ω ∩ P and t ∈ [0, 1].
If not, then there exist x0 ∈ ∂Ω ∩ P , t0 ∈ [0, 1] such that x0 = T x0 + t0 my ,
and so we have

kx0 k = kT x0 + t0 my k ≥ (σ − )kT x0 k.

Hence we get
kT x0 k 1
≤ ,
kx0 k σ−
which contradicts (7.3.1). By the homotopy property (4) of Theorem 7.2.3,
we get
ind(T, Ω ∩ P ) = ind(T + m0 y , Ω ∩ P )
and thus, we must have

ind(T + my , Ω ∩ P ) = 0

(since Ω is bounded and T (E) is bounded for any countably bounded subset
E of Ω ∩ P ). Consequently, ind(T, Ω ∩ P ) = 0. This completes the proof.
Corollary 7.3.8. Let X be a real Banach space, Ω be an open bounded
subset of X with 0 ∈ Ω and P be a quasinormal cone in X. If T : Ω ∩ P → P
is a continuous countably k-set contraction and the following conditions are
satisfied:

(1) T x 6= αx for all x ∈ ∂Ω ∩ P and α ∈ [0, 1];

n o
(2) inf kµT xk
kxk : x ∈ ∂Ω ∩ P > σ1 , where 1 ≤ µ < 1
k and σ is the quasi-
normality constant of P ;

then ind(T, Ω ∩ P ) = 0.
Proof. If we put T1 x = µT x for all x ∈ Ω ∩ P, then T1 is countably
condensing and T1 satisfies all the conditions of Theorem 7.3.7. Thus, we
have
ind(T1 , Ω ∩ P ) = 0.

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 183

From the assumption (1) and the homotopy property (4) of Theorem 7.2.3,
we know that
ind(T, Ω ∩ P ) = ind(µT, Ω ∩ P ).
Therefore, we have ind(T, Ω ∩ P ) = 0. This completes the proof.
Corollary 7.3.9. Let X be a real Banach space, Ω be an open bounded
subset of X with 0 ∈ Ω and P be a quasinormal cone in X. If T : Ω ∩ P →
P is a continuous countably k-set contraction with k < σ, where σ is the
quasinormality constant of P and the following conditions are satisfied:

(1) T x 6= x for all x ∈ ∂Ω ∩ P ;

(2) kT xk ≥ kxk for all x ∈ ∂Ω ∩ P ;

then ind(T, Ω ∩ P ) = 0.
1
Proof. We take 0 > 0 such that k + 0 < min{1, σ} and put T1 = k+ 0
T.
k
Then T1 is a countably k+0 -contraction, and we know, by the assumption
(2), that
n kT xk o 1 1
1
inf : x ∈ ∂Ω ∩ P ≥ > for all x ∈ ∂Ω ∩ P.
kxk k + 0 σ

We also have T1 x 6= αx for all x ∈ ∂Ω ∩ P and α ∈ [0, 1]. Therefore, by

Theorem 7.3.7, we have ind(T1 , Ω ∩ P ) = 0. This completes the proof.
Theorem 7.3.10. Let X be a real Banach space, Ω be an open bounded
subset of X with 0 ∈ Ω and P be a cone in X. Let T : Ω ∩ P → P be a
continuous countably condensing mapping. Suppose that there exists y0 ∈
P \ {0} such that x − T x 6= λy0 for all x ∈ ∂Ω ∩ P and λ ≥ 0. Then
ind(T, Ω ∩ P ) = 0.
Proof. For any m > 0, by the homotopy property (4) of Theorem 7.2.3,
we have
ind(T, Ω ∩ P ) = ind(T + my0 , Ω ∩ P )
and thus it follows that ind(T, Ω ∩ P ) = 0.
Corollary 7.3.11. Let X be a real Banach space, Ω1 , Ω2 be two open
bounded subsets of X and P be a cone in X. If 0 ∈ Ω1 ⊂ Ω2 , T : Ω2 ∩ P → P
is a continuous countably condensing mapping and the following conditions
are satisfied:

(1) T x 6> x for all x ∈ ∂Ω1 ∩ P ;

(2) T x 6< x for all x ∈ ∂Ω2 ∩ P ;

then T has a fixed point in Ω2 \ Ω1 ∩ P .

Copyright 2006 by Taylor & Francis Group, LLC

 184 Topological Degree Theory and Applications

Proof. We may assume that x 6= T x for all x ∈ ∂Ωi ∩ P , i = 1, 2. By the

assumption (1) and the homotopy property (4) of Theorem 7.2.2, we have

ind(T, Ω1 ∩ P ) = 1.

On the other hand, for any y0 ∈ P with y0 6= 0, we know, by the assumption

(2), that x 6= T x + λy0 for all x ∈ ∂Ω2 ∩ P and λ ≥ 0. Therefore, it follows
from Theorem 7.3.10 that ind(T, Ω2 ∩ P ) = 0 and so

ind(T, (Ω2 \ Ω1 ) ∩ P ) = −1.

Therefore, T has a fixed point in (Ω2 \ Ω1 ) ∩ P . This completes the proof.

Corollary 7.3.12. Let X be a real Banach space, Ω1 , Ω2 be two open
bounded subsets of X and P be a quasinormal cone in X. If 0 ∈ Ω1 ⊂ Ω2 ,
T : Ω2 ∩ P → P is a continuous countably k-set contraction with k < σ,
where σ is the quasinormality constant of P and the following conditions are
satisfied:
(1) kT xk ≤ kxk for all x ∈ ∂Ω1 ∩ P ;
(2) kT xk ≥ kxk for all x ∈ ∂Ω2 ∩ P ;
then T has a fixed point in Ω2 \ Ω1 ∩ P .
Proof. We may assume that x 6= T x for all x ∈ ∂Ωi ∩ P , i = 1, 2. By the
assumption (1) and the homotopy property (4) of Theorem 7.2.2, we have

ind(T, Ω1 ∩ P ) = 1.

On the other hand, we have, by the assumption (2) and Corollary 7.3.9,
that ind(T, Ω2 ∩ P ) = 0 and so

ind(T, (Ω2 \ Ω1 ) ∩ P ) = −1.

Therefore, T has a fixed point in (Ω2 \ Ω1 ) ∩ P . This completes the proof.

Theorem 7.3.13. Let X be a real Banach space, Ω be an open bounded
subset of X and P be a cone in X. If T : Ω∩P → P is a continuous countably
condensing mapping, S : ∂Ω ∩ P → P is a continuous compact mapping and
the following conditions are satisfied:
(1) inf{kSxk : x ∈ ∂Ω ∩ P } > 0;
(2) x 6= T x + λSx for all x ∈ ∂Ω ∩ P and λ > 0;
then ind(T, Ω ∩ P ) = 0.
/ S(∂Ω ∩ P ). and thus, it
Proof. By the assumption (1), we know that 0 ∈
follows from the well-known argument, (see Lemma 7.3.2) that

0∈
/ Conv(S(∂Ω ∩ P )).

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 185

By Dugundji’s extension theorem, there exists a continuous mapping S ∗ :

Ω ∩ P → Conv(S(∂Ω ∩ P )) such that S ∗ x = Sx for all x ∈ ∂Ω ∩ P and so S ∗
is compact.
From assumption (2) and the homotopy property (4) of Theorem 7.2.2, we
have
ind(T, Ω ∩ P ) = ind(T + mS ∗ , Ω ∩ P ) for all m > 0
and thus ind(T, Ω ∩ P ) = 0, for, otherwise, x = T x + mS ∗ x has a solution in
Ω ∩ P for any m > 0, which is a contradiction (note that Ω is bounded and
inf x∈Ω∩P kS ∗ xk = a > 0). This completes the proof.
Theorem 7.3.14. Let X be a real Banach space, Ω1 , Ω2 be two open
bounded subsets of X such that 0 ∈ Ω1 ⊂ Ω2 and P is a cone in X. If T :
Ω2 ∩ P → P is a continuous countably condensing mapping, S : ∂Ω2 ∩ P → P
is a continuous compact mapping and the following conditions are satisfied:

(1) inf{kSxk : x ∈ ∂Ω2 ∩ P } > 0;

(2) T x 6= αx for all x ∈ ∂Ω1 ∩ P , α ≥ 1;

then one of the following conclusion holds:

(I) T has a fixed point in (Ω2 \ Ω1 ) ∩ P ;

(II) x = T x + λSx has a solution in ∂Ω2 ∩ P for some λ > 0.

Proof. Assume that (II) is not true. Then, by Theorem 7.3.13, we have
ind(T, Ω2 ∩ P ) = 0. On the other hand, by the assumption (2), we have
ind(T, Ω1 ∩ P ) = 1 and thus,

ind(T, (Ω2 \ Ω1 ) ∩ P ) = −1.

Therefore, T has a fixed point in (Ω2 \ Ω1 ) ∩ P .

Theorem 7.3.15. Let X be a real Banach space, Ωi , i = 1, 2, 3, be open
bounded subsets of X such that 0 ∈ Ω1 , Ωi ⊂ Ωi+1 , i = 1, 2, and P is a
cone in X. If T : Ω3 ∩ P → P is a continuous countably condensing mapping
satisfying the following conditions:

(1) kT xk ≤ kxk, for all x ∈ ∂Ω1 ∩ P ;

n o
(2) inf kT xk
kxk : x ∈ ∂Ω 2 ∩ P > σ1 , where σ is the quasinormality constant
of P ;

(3) kT xk ≤ kxk, for all x ∈ ∂Ω3 ∩ P .

Then T has two positive fixed points in (Ω3 \ Ω1 ) ∩ P .

Copyright 2006 by Taylor & Francis Group, LLC

 186 Topological Degree Theory and Applications

Proof. We first prove that T has a fixed point in (Ω2 \ Ω1 )) ∩ P . We may

also assume that T x 6= x for all x ∈ ∂Ω1 ∩ P . By the assumption (1), we then
have x 6= tT x for all x ∈ ∂Ω1 ∩ P and t ∈ [0, 1]. Thus,

ind(T, Ω1 ∩ P ) = 1. (7.3.2)

By the assumption (2) and Theorem 7.3.7, we have

ind(T, Ω2 ∩ P ) = 0. (7.3.3)

From (7.3.2) and (7.3.3), we deduce ind(T, (Ω2 \ Ω1 ) ∩ P ) = −1 and, conse-

quently, T has a fixed point in (Ω2 \ Ω1 ) ∩ P . To see that T has a fixed point
in (Ω3 \ Ω2 )) ∩ P , we may also assume that T x 6= x for all x ∈ ∂Ω3 ∩ P . By
the assumption (3), we then have x 6= tT x for all x ∈ ∂Ω3 ∩ P and t ∈ [0, 1].
Thus,
ind(T, Ω3 ∩ P ) = 1. (7.3.4)
From (7.3.3) and (7.3.4), we deduce ind(T, (Ω3 \ Ω2 ) ∩ P ) = 1 and, conse-
quently, T has a fixed point in (Ω3 \ Ω2 ) ∩ P . This completes the proof.
Finally, we give a different version of Theorem 7.3.15 as follows:
Theorem 7.3.16. Let X be a real Banach space, Ωi , i = 1, 2, 3, be open
bounded subsets of X such that 0 ∈ Ω1 , Ωi ⊂ Ωi+1 , i = 1, 2, and P is a
cone in X. If T : Ω3 ∩ P → P is a continuous countably condensing mapping
satisfying the following conditions:

(1) T x 6> x for all x ∈ ∂Ω1 ∩ P ;

(2) T x 6≤ x for all x ∈ ∂Ω2 ∩ P ;

(3) T x 6> x for all x ∈ ∂Ω3 ∩ P ;

then T has two positive fixed points in (Ω3 \ Ω1 ) ∩ P ;

7.4 Perturbations of Condensing Mappings

In this section, let X be a real Banach space, P be a quasinormal cone in
X, T : D(T ) ⊂ X → X be a completely continuous or condensing mapping
and A : D(A) ⊂ X → 2X be an accretive operator. We consider the existence
of solutions to the mapping equation x ∈ −Ax + T x and prove some existence
results by using the index theory in Section 7.2
Theorem 7.4.1. Let P be a cone in a Banach space E, A : D(A) ⊆ P → 2P
be an accretive operator with P = (I + A)(D(A)), 0 ∈ Ω be an open bounded

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 187

subset of E and K : Ω ∩ P → P be a continuous compact mapping. Suppose

that
Kx ≤ x for all x ∈ ∂Ω ∩ P.
Then −A + K has a fixed point in Ω ∩ D(A).
Proof. It is easy to see that x ∈ −Ax+Kx if and only if x = (I +A)−1 Kx.
If x ∈ −Ax + Kx for some x ∈ ∂Ω ∩ D(A), then the conclusion is true and so
we may assume that x ∈ / −Ax + Kx for all x ∈ ∂Ω ∩ D(A).
Put H(t, x) = (I + A)−1 tKx for all (t, x) ∈ [0, 1] × Ω ∩ P . Then we have

x 6= (I + A)−1 tKx for all (t, x) ∈ [0, 1] × ∂Ω ∩ P.

Indeed, if x = (I + A)−1 tKx for some (t, x) ∈ [0, 1] × ∂Ω ∩ P , then t 6= 1 and

tKx ∈ x + Ax. However, Ax ≥ 0 and hence we have tKx ≥ x. Therefore, it
follows that t > 0 and Kx ≥ t−1 x > x, which is a contradiction. Hence we
have
ind(I + A)−1 K, Ω ∩ P ) = deg(0, Ω ∩ P ) = 1.
Therefore, −A + K has a fixed point in Ω ∩ D(A). This completes the proof.
Theorem 7.4.2. Let P be a cone in a Banach space E, A : D(A) ⊆ P → 2P
be an accretive operator with P = (I + A)(D(A)), 0 ∈ Ω be an open bounded
subset of E and K : Ω ∩ P → P be a continuous compact mapping. Suppose
that
kKxk ≤ kxk for all x ∈ ∂Ω ∩ P.
Then −A + K has a fixed point in Ω ∩ D(A).
Proof. We may assume that x ∈ / −Ax + Kx for all x ∈ ∂Ω ∩ D(A). Let
H(t, x) be defined as in Theorem 7.4.1. Then we have

x 6= H(t, x) for all (t, x) ∈ [0, 1] × ∂Ω ∩ P.

Indeed, if x = (I + A)−1 tKx for some (t, x) ∈ [0, 1] × ∂Ω ∩ P , then we have

kxk ≤ k(I + A)−1 tKxk ≤ ktKxk

since (I + A)−1 0 = 0 and (I + A)−1 is nonexpansive. Therefore t = 1, which

contradicts our assumption. Thus, we have

ind(I + A)−1 K, Ω ∩ P ) = ind(0, , Ω ∩ P ) = 1

and so T has a fixed point. This completes the proof.

Theorem 7.4.3. Let X be a real Banach space, P be a quasinormal cone
in X, A : D(A) ⊆ P → 2P be an accretive mapping with P = (I +λA)P for all
λ > 0 and Ωi , i = 1, 2, be two open bounded subsets of X with 0 ∈ Ω1 ⊂ Ω2 .
Let T : Ω2 ∩ P → P be a continuous countably condensing mapping. Suppose
that the following conditions are satisfied:

Copyright 2006 by Taylor & Francis Group, LLC

 188 Topological Degree Theory and Applications

(1) kT xk ≤ kxk for all x ∈ ∂Ω1 ∩ P ;

n o
kT xk 1
(2) inf kxk+kvk : x ∈ ∂Ω ∩ P, v ∈ Ax > σ; where σ is the quasinormality
constant of P .

Then −A + T has a fixed point in Ω2 \ Ω1 ∩ P .

Proof. We may assume that x ∈
/ −Ax + T x for all x ∈ ∂Ωi ∩ P for i = 1, 2.
Set
U1 = (I + A)(Ω1 ∩ P ), U2 = (I + A)(Ω2 ∩ P ).
Since (I + A)−1 is nonexpansive, it follows that U1 , U2 are open subsets of
P and T (I + A)−1 is countably condensing. It is easy to see that −A + T
has a fixed point in Ω2 \ Ω1 ∩ P if and only if T (I + A)−1 has a fixed point
in U2 \ U1 and the set of fixed points of T (I + A)−1 are bounded in U2 \ U1 ,
and so we may simply assume that U1 and U2 are bounded. It is obvious that
H(·, ·) : [0, 1] × Ω1 ∩ P → P defined by

H(t, x) = tT (I + A)−1 x for all (t, x) ∈ [0, 1] × Ω1 ∩ P

is a homotopy of countably condensing mappings.

Now, we claim that x 6= H(t, x) for all (t, x) ∈ [0, 1] × ∂U1 . In fact, if not,
then there exist t0 ∈ [0, 1] and x0 ∈ ∂U1 such that x0 = t0 T (I + A)−1 x0 . Put
z0 = (I + A)−1 x0 , then z0 ∈ ∂Ω1 ∩ P . Take v0 ∈ Az0 such that x0 = z0 + v0
and then we have

kz0 k ≤ kz0 + v0 k = kx0 k = t0 kT z0 k.

From the assumption (1), we must have t0 = 1, which contradicts the fact
that −A + T has no fixed point on ∂Ω1 ∩ P . Thus we have

ind(T (I + A)−1 , U1 ) = ind(0, U1 ) = 1

since 0 ∈ A0 ⊂ U1 . By the definition of quasinormality constant of P and

the assumption (2), we may take  > 0 and y0 6= 0 ∈ P such that kx + y0 k ≥
(σ − )kxk for all x ∈ P and
n kT xk o 1
inf : x ∈ ∂Ω ∩ P, v ∈ Ax > . (7.4.1)
kxk + kvk σ−

Next, we claim that x 6= T (I + A)−1 x + λy0 for all x ∈ ∂U2 and λ ≥ 0. If

not, then there exist x0 ∈ ∂U2 and λ0 > 0 such that x0 = T (I + A)−1 x0 + λy0 .
Put z0 = (I + A)−1 x0 , and then z0 ∈ ∂Ω2 ∩ P . There exists v0 ∈ Az0 such
that x0 = z0 + v0 , and so we have

kz0 + v0 k = kx0 k = kT z0 + λy0 k ≥ (σ − )kT z0 k,

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 189

which contradicts (7.4.1). By the homotopy property (4) of Theorem 7.2.2,

we get

ind(T (I + A)−1 , U2 ) = ind(T (I + A)−1 + λy0 ) for all λ > 0

and so ind(T (I + A)−1 , U2 ) = 0. Therefore, we have

ind(T (I + A)−1 , U2 \ U1 ) = −1

and so T (I + A)−1 has a fixed point in U2 \ U1 , i.e., −A + T has a fixed point

in (Ω2 \ Ω1 ) ∩ P . This completes the proof.
Remark. Nonexpansive mappings with dissipative perturbations in cones
have been studied by Chang et al. [53].

7.5 Index Theory for Nonself Mappings

In this section, first, let K ⊂ Rn be a nonempty cone, Ω ⊂ Rn be a
nonempty bounded subset with Ω ∩ K 6= ∅ and f : Ω ∩ K → Rn be a
continuous mapping. We recall the following definition from [177].
Definition 7.5.1. [182] A mapping f : Ω∩K → Rn is said to be generalized
inward if d(x, f (x)) 6= d(f (x), K) for all x ∈ Ω ∩ K with f (x) ∈
/ K, where
d(f (x), K) = inf y∈K d(f (x), y).
Now, assume that f : Ω ∩ K → Rn is a continuous generalized inward
mapping. Let r : Rn → K be a metric projection, i.e., d(x, r(x)) = d(x, K)
for all x ∈ Rn . Such a mapping always exists and is unique. We know that
rf : Ω ∩ K → K is continuous and rf (x) 6= x for all x ∈ ∂Ω ∩ K. Otherwise,
if rf (x) = x for some x ∈ ∂Ω ∩ K, then d(x, f (x)) = d(rf (x), f (x)) =
d(f (x), K), which is a contradiction.
Now, we define
ind(f, Ω ∩ K) = ind(rf, Ω ∩ K), (7.5.1)
where ind(rf, Ω ∩ K) is the fixed point index in Section 7.2, and this is called
the fixed point index for generalized inward mapping.
Theorem 7.5.1. The fixed point index for generalized inward mapping
has the following properties:
(1) ind(x0 , Ω ∩ K) = 1 if x0 ∈ Ω ∩ K;
(2) If ind(f, Ω ∩ K) 6= 0, then f (x) = x has a solution in Ω ∩ K;
(3) If Ω1 , Ω2 are two open bounded disjoint subsets, then ind(f, (Ω1 ∪ Ω2 ) ∩
K) = ind(f, Ω1 ∩ K) + ind(f, Ω2 ∩ K);

Copyright 2006 by Taylor & Francis Group, LLC

 190 Topological Degree Theory and Applications

(4) Let H(t, x) : [0, 1] × Ω ∩ K → Rn be a continuous mapping satisfy-

ing H(t, x) 6= x for all (t, x) ∈ [0, 1] × ∂Ω ∩ K and d(f (x), H(t, x)) 6=
d(f (x), x) for all (t, x) ∈ Ω ∩ K with H(t, x) ∈
/ K. Then ind(H(t, ·), Ω ∩
K) does not depend on t ∈ [0, 1].

Now, we assume that E is a real Banach space, Ω ⊂ E is open and bounded

and P ⊂ E is a cone with Ω ∩ P 6= ∅. For all x ∈ P , IP (x) = {x + λ(y − x) :
λ ≥ 1, y ∈ P } is called the inward set of x relative to P .
Definition 7.5.2. Let T : Ω ∩ P → E be a mapping.
(1) If d(T x, P ) 6= d(x, T x) for all T x ∈
/ P , then T is said to be a generalized
inward mapping.
(2) If T x ∈ IP (x) for all x ∈ Ω ∩ P , then T is said to an inward mapping.
(3) If T x ∈ IP (x) for each x ∈ Ω ∩ P , then T is said to a weakly inward
mapping.

It is obvious that an inward mapping is weakly inward.

Proposition 7.5.3. If T : Ω ∩ P → E is weakly inward, then T is gener-
alized inward.
Proof. For all x ∈ Ω ∩ P with T x ∈ / P , we have d(T x, IP (x)) = 0. Thus,
there exists y ∈ IP (x) such that d(T x, y) < d(x, T x). On the other hand,
there exists z ∈ P such that z = ty + (1 − t)x for some t ∈ (0, 1). Thus, we
have
d(T x, P ) < d(T x, z) = d(T x, ty + (1 − t)x)
≤ td(T x, y) + (1 − t)d(x, T x) = d(x, T x).
Therefore, T is generalized inward on Ω ∩ P . This completes the proof.
In the following, we assume that there exists a continuous metric projection
r : E → P . Assume that T x 6= x for all x ∈ ∂Ω∩P and then we have rT x 6= x
for all x ∈ ∂Ω ∩ P . Otherwise, we have rT x = x for some x ∈ ∂Ω ∩ P .
Then d(x, T x) = d(rT x, T x) = d(T x, P ), which is a contradiction. Thus,
ind(rT, Ω ∩ P ) is well defined. Now, we define
ind(T, Ω ∩ P ) = ind(rT, Ω ∩ P ). (7.5.2)
If r1 , r2 : E → P are two continuous metric projections, then {(tr1 + (1 −
t)r2 }t∈[0,1] is a family of continuous metric projections. One can easily see
that (tr1 +(1−t)r2 )T x 6= x for all x ∈ ∂Ω∩P , thus ind([tr1 +(1−t)r2 ]T, Ω∩P )
does not depend on t ∈ [0, 1] by Theorem 7.2.3. Therefore, ind(rT, Ω ∩ P )
does not depend on r, and so ind(T, Ω ∩ P ) is well defined.
Theorem 7.5.4. The fixed point index defined by (7.5.2) has the following
properties:

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 191

(1) ind(x0 , Ω ∩ P ) = 1 if x0 ∈ Ω ∩ P ;
(2) If ind(T, Ω ∩ K) 6= 0, then T x = x has a solution in Ω ∩ K;
(3) If Ω1 , Ω2 are two open bounded disjoint subsets, then ind(T, (Ω1 ∪ Ω2 ) ∩
P ) = ind(T, Ω1 ∩ P ) + ind(T, Ω2 ∩ P );
(4) Let H(t, x) : [0, 1] × Ω ∩ P → E be a continuous compact mapping
satisfying H(t, x) 6= x for all (t, x) ∈ [0, 1] × ∂Ω ∩ P and d(x, H(t, x)) 6=
d(H(t, x), P ) for all (t, x) ∈ Ω∩P with H(t, x) ∈ / P . Then ind(H(t, ·), Ω∩
P ) does not depend on t ∈ [0, 1].

Remark. For more details about the results in this section, we refer the
reader to [182].

7.6 Applications to Integral and Differential

Equations
In this section, we give some applications to the integral and differential
equations by using the results of the previous sections.
Example 7.6.1. Let K(t, s) : [a, b] × [a, b] → [0, +∞) be a continous
function and f (t, x) : [a, b) × R → [0, +∞) be a continuous function. Suppose
the following conditions are satisfied:

(1) For each t ∈ [a, b], f (t, x) is increasing in x;

Rb
(2) a f (s, c)ds < cM −1 , where M = max{K(t, s) : (t, s) ∈ [a, b] × [a, b]}
and c > 0 is a constant;
(3) f (t, s) ≥ αsγ for all s ∈ [0, 0 ), where 0 > 0, α > 0 and 0 < γ < 1 are
constants;
(4) K(t, s) 6≡ 0 for all (t, s) ∈ [a, b] × [a, b].

Then the integral equation:

Z b
x(t) = K(t, s)f (t, x(s))ds (E 7.6.1)
a

has a nontrivial non-negative solution in C([a, b]).

Proof. Define a mapping T : C([a, b]) → C([a, b]) by
Z b
T x(t) = K(t, s)f (t, x(s))ds for all x(·) ∈ C([a, b]). (7.6.1)
a

Copyright 2006 by Taylor & Francis Group, LLC

 192 Topological Degree Theory and Applications

Obviously, T is continuous and compact and finding a solution of (E 7.6.1) is

equivalent to finding a fixed point of T . Put
P = {x(·) ∈ C([a, b]) : x(t) ≥ 0, t ∈ [a, b]}.
Then P is a cone in C([a, b]). By the assumption (1) and (2), we have
kT x(·)k < c, kx(·)k = c, x(·) ∈ P. (7.6.2)
By the assumption (4), there exist (t0 , s0 ) ∈ [a, b] × [a, b], δ > 0 and β > 0
such that K(t, s) ≥ β for all (t, s) ∈ [t0 − δ, t0 + δ] × [s0 − δ, s0 + δ]. Thus we
have Z s0 +δ
kT x(·)k ≥ βf (s, x(s))ds.
s0 −δ
By the assumption (3), if kx(·)k < min{1, 0 }, we have
kT x(·)k ≥ 2βδkx(·)kγ . (7.6.3)
Therefore, we have kT x(·)k > kx(·)k for kx(·)k = r and x(·) ∈ P with r
sufficiently small. We take Ω = B(0, c) and Ω0 = B(0, r). By Corollary 7.3.4,
we know that T has a fixed point in Ω \ Ω0 ∩ P , i.e., (E 7.6.1) has a nontrivial
non-negative solution.
Example 7.6.2. Consider the boundary value problem:
(
x00 (t) + f (t, x(t)) = 0, t ∈ [0, 1],
(E 7.6.2)
x(0) = x(1), x0 (0) = −x0 (1).
Assume that f is continuous and satisfies the following conditions:
(1) For all t ∈ [0, 1], f (t, x) is increasing in x;
R1
(2) 0 f (s, c)ds < 3c4 ;

(3) f (t, s) ≥ αsγ for all s ∈ [0, 0 ), where 0 > 0, α > 0 and 0 < γ < 1 are
constants.
Then (E 7.6.2) has a nontrivial non-negative C 2 solution.
Proof. It is well known that (E 7.6.2) is equivalent to the following integral
equation: Z 1
x(t) = G(t, s)f (s, x(s))ds, (E 7.6.3)
0
where G(t, s) is the Green function defined by
(
1
(t + 1)(2 − s), t ≤ s,
G(t, s) = 13
3 (s + 1)(2 − t), t > s.
One may easily see that M = max{G(t, s) : (t, s) ∈ [0, 1]×[0, 1]} < 34 and f, G
satisfy the conditions of Example 7.6.1. Therefore, (E 7.6.3) has a nontrivial
non-negative solution in C([0, 1]), i.e., (E 7.6.1) has a nontrivial non-negative
C 2 solution.

Copyright 2006 by Taylor & Francis Group, LLC

 FIXED POINT INDEX THEORY 193

7.7 Exercises
1. Let Ω ⊂ Rn be measurable subset with m(Ω) < +∞, 1 ≤ p ≤ ∞ and
P ⊂ Lp (Ω) be given by P = {f (·) ∈ Lp (Ω) : f (x) ≥ 0, almost all x ∈
Ω}. Show that P is fully regular.

2. Let P ⊂ c0 be given by P = {(xi ) ∈ c0 : xi ≥ 0, i = 1, 2, · · · }. Show

that P is regular, but not fully regular.

3. Let E be a reflexive Banach space and P ⊂ E be a cone. Prove that

the following conclusions are equivalent:
(a) P is normal;
(b) P is regular;
(c) P is fully regular.

4. Let E be a normed space and P ⊂ E be a cone. If P has the nonempty

interior, then show that P is reproducing.

5. Let E be a normed space, P ⊂ E be a cone and P ∗ ⊂ E ∗ be defined by

P ∗ = {f ∈ E ∗ : f (x) ≥ 0 for all x ∈ P }. If P is reproducing, then show
that P ∗ is a cone in E ∗ .

6. Let P ⊂ L1 (Ω) be defined by P = {f (·) ∈ L1 (Ω) : f (x) ≥ 0, almost all x ∈

Ω}, where m(Ω) < ∞. Show that P allows plastering.

7. Let E be a Banach space, P ⊂ be a normal cone, [a, b] = {x ∈ E :

a ≤ x ≤ b} and T : [a, b] → [a, b] be a continuous condensing mapping
satisfying Ax ≤ Ay for all x, y ∈ [a, b] with x ≤ y. Show that T has a
fixed point in [a, b].

8. Let E be a locally convex space, P ⊂ E be a cone, Ω ⊂ E be an open

subset with Ω ∩ P 6= ∅ and T : Ω ∩ P → P be a continuous mapping
such that T (Ω ∩ P ) is relatively compact in E. Assume that x 6= T x for
all x ∈ ∂Ω ∩ P . Construct the fixed point index theory for T on Ω ∩ P .

9. Let E be a locally convex space, P ⊂ E be a cone, Ω ⊂ E be an open

subset with 0 ∈ Ω and Ω ∩ P 6= ∅ and T : Ω ∩ P → P be a continuous
mapping such that T (Ω ∩ P ) is relatively compact in E. Assume that
x 6= tT x for all x ∈ ∂Ω ∩ P and t ∈ [0, 1). Show that T has a fixed point
in Ω ∩ P .

10. Let E be a Banach space, P ⊂ E be a cone and A : P → E be a

continuous accretive operator. Assume that, for all x ∈ P , there exists
α(x) > 0 such that Ax ≤ α(x)x. Show that (λI + A)(P ) ⊇ P for all
λ > 0.

Copyright 2006 by Taylor & Francis Group, LLC

 194 Topological Degree Theory and Applications

11. Let a(s, t) : [0, 1] × [0, 1] → [0, +∞) be a continuous function satisfying
the following conditions:
(1) [a(s, t) − a(s, r)](t − r) ≥ 0;
(2) a(s, t) ≤ ct for some constant c > 0 and all (s, t) ∈ [0, 1] × [0, 1].
Let k(s, t) : [0, 1] × [0, 1] → [0, +∞) be a continuous function such that
k(s, t) ≤ αt + β for all (s, t) ∈ [0, 1] × [0, 1], where α ∈ (0, 1), β > 0 are
constant. Show that the following integral inequation:
Z 1
x(t) + a(t, x(t)) − k(t, x(s))ds = 0 for all t ∈ [0, 1]
0

has a solution x(·) ∈ C([0, 1]).

12. Let a, b, c, d be non-negative numbers such that e = ac + bc + ad > 0
and f (x, y) : [0, 1] × R → R be defined by f (x, y) = Σm i=1 ai (x)y
αi
for
all (x, y) ∈ [0, 1] × R, where ai (t) : [0, 1] → [0, +∞) is continuous for
i = 1, 2 · · · , m and αi > 0 for i = 1, 2 · · · , m. Suppose that there exist
1 ≤ j, k ≤ m such that αj (t) < 1, αk > 1, aj (t)ak (t) > 0 for all
R1 −1
t ∈ [0, 1] and Σm i=1 0 ai (t)dt < f , where f = e−1 (4ac)−1 if ac 6= 0,
−1 −1
f = e (bc + bd) if a = 0 and f = e (ad + bd) if c = 0. Show that the
following equation:
(
x00 (t) = −f (t, x(t)), t ∈ [0, 1],
ax(0) − bx0 (0) = 0, cx(1) + dx0 (1) = 0

has two non-negative nontrivial solutions in C 2 ([0, 1]) by using the Green
function G(t, s) defined as
(
e−1 (at + b)[c(1 − s) + d], t ≤ s,
G(t, s) = −1
e (as + b)[(c(1 − t) + d], t > s.

Copyright 2006 by Taylor & Francis Group, LLC

 Chapter 8
REFERENCES

1. A. Addou and B. Mermri, Topological degree and application to a par-

abolic variational inequality problem, Inter. J. Math. Sci. 25(2001)
273–287.
2. R. P. Agarwal, M. Meehan and D. O’Regan, Fixed Point Theory and
Application, Cambridge University Press, New York, 2001.
3. R. Agarwal and D. O’Regan, An index theory for P -concentrative J-
maps, Appl. Math. Lett. 16(2003), 1265–1271.
4. S. Aizicovici and Y. Q. Chen, Note on the topological degree of the
subdifferential of a lower semicontinuous convex function, Proc. Amer.
Math. Soc. 126(1998), 2905–2908.
5. H. Alex, S. Hahn and L. Kaniok, The fixed point index for noncompact
mappings in non-locally convex topological vector spaces, Comment.
Math. Univ. Carolinae 32(1994), 249–257.
6. M. Altman, A fixed theorem for completely continuous operators in
Banach spaces, Bull. Acad. Polon. Sci. Sér. Sci. Math. Astonom.
Phys. 3(1955), 490–513.
7. H. Amann, On the number of solutions of nonlinear equations in Banach
spaces, J. Funct. Anal. 11(1972), 346–384.
8. H. Amann, Fixed point equations and nonlinear eigenvalue problems in
ordered Banach spaces, SIAM Review 18(1976), 620–709.
9. H. Amann, A note on degree theory for gradient mappings, Proc. Amer.
Math. Soc. 85(1982), 591–595.
10. H. Amann and S. Weiss, On the uniqueness of the topological degree,
Math. Z. 130(1973), 39–54.
11. J. Appell, M. Väth and A. Vignoli, Compactness and existence results
for ordinary differential equations in Banach spaces, Z. Anal. Anw.
18(1988), 469–484.
12. J. P. Aubin and A. Cellina, Differential Inclusions, Springer-Verlag,
Berlin, 1984.

195

Copyright 2006 by Taylor & Francis Group, LLC

 196 Topological Degree Theory and Applications

13. J. P. Aubin and I. Ekeland, Applied Nonlinear Analysis, Wiley, New

York, 1984.

14. J. P. Aubin and H. Frankowska, Set-Valued Analysis, Birkhauser, Boston,

1990.

15. R. Bader, A topological fixed-point index theory for evolution inclusions,

Z. Anal. Anw. 20 (2001) 3–16.

16. S. Banach, Theory of linear operations (English translation), North-

Holland, 1987.

17. V. Barbu, Nonlinear Semigroups and Differential Equations in Banach

Spaces, Noordhoff, Leyden, 1976.

18. R. I. Becker, Periodic solutions of semilinear equations of evolution of

compact type, J. Math. Anal. Appl. 82(1982 ), 33–48.

19. V. Benci, A geometrical index for the group S 1 and some applications to
the study of periodic solutions of ordinary differential equations, Comm.
Pure Appl. Math. 34(1981), 393–432.

20. J. Berkovits and V. Mustonen, On the degree for mappings of monotone

type, Nonlinear Anal. 12(1986), 1373–1383.

21. J. Berkovits and V. Mustonen, Topological degree for perturbation of

linear maximal monotone mappings and applications to a class of par-
abolic problems, Rend. Mat. VII(1992), 597–621.

22. U. G. Borisovic, B. D. Gelman, A. D. Myskis and V. V. Obuhowskii,

Topological methods in fixed point theory of multivalued mappings, Us-
piehi Mat. Nauk. 1(1980), 59–126.

23. K. Borsuk, Drei Sätze über die n-dimensional Euklidische Sphäre, Fund.
Math. 20(1933), 177–190.

24. D. Bothe, Upper semicontinuous perturbations of m-accretive operators

and differential inclusions with dissipative right-hand side, Proc. Work-
shop “Topological Methods in Differential Inclusions”, Warsaw, 1994.

25. J. Bourgain, H. Brézis and P. Mironescu, Lifting, degree, and distribu-

tional Jacobian, Comm. Pure and Appl. Math. Vol. 58, 2005, 529–551.

26. H. Brézis, Operateurs maxinaux monotones, North-Holland, 1973.

27. H. Brézis, Degree theory: old and new, in Topological Nonlinear Analy-
sis II: Degree, Singularity and Variations (M. Matzeu and A. Vignoli,
Ed.) Birkhauser, 1997, 87–108.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 197

28. H. Brézis, Y. Li, P. Mironescu and L. Nirenberg, Degree and Sobolev

spaces, Top. Meth. Nonlinear Anal. 13(1999), 181–190.
29. H. Brézis, L. Nirenberg, Degree theory and BMO, Part I: compact man-
ifolds without boundaries, Selecta Math. 1(1995), 197–263.
30. H. Brézis and F. E. Browder, A general ordering principle in nonlinear
functional analysis, Advances in Math. 21(1976), 355–364.
31. H. Brézis, M. G. Crandall and A. Pazy, Perturbations of nonlinear max-
imal monotone sets, Comm. Pure Appl. Math. 23(1970), 123–144.
32. L. E. J. Brouwer, Uber Abbildung der Mannigfaltigkeiten, Math. Ann.
70(1912), 97–115.
33. F. E. Browder, The fixed point theory of multi-valued mappings in topo-
logical vector spaces, Math. Ann. 177(1968), 283–301.
34. F. E. Browder, Nonlinear Operators and Nonlinear Equations of Evo-
lution in Banach Spaces, Proc. Symp. Pure Math., Amer. Math. Soc.
18, Part 2, 1976.
35. F. E. Browder, Fixed point theory and nonlinear problems, Bull. Amer.
Math. Soc. 1(1983 ), 1–39.
36. F. E. Browder, Degree theory for nonlinear mappings, Proc. Symp.
Pure Math. Soc. 45(1986), 203–226.
37. F. E. Browder and R. D. Nussbaum, The topological degree for non-
compact nonlinear mappings in Banach spaces, Bull. Amer. Math.
Soc. 74(1968), 671–676.
38. F. E. Browder and W. V. Petryshyn, Approximation methods and the
generalized topological degree for nonlinear mappings in Banach spaces,
J. Funct. Anal. 3(1969), 217–274.
39. R. F. Brown, An Introduction to Nonlinear Analysis, Birkhauser, 1993.
40. A. Buica, Contributions to the coincidence degree theory of homogenous
operators, Pure Math. Appl. 11(2000), 153–159.
41. A. Buica and Llibre, Averaging methods for finding periodic orbits via
Brouwer degree, Bull. Sci. Math. 128(2004), 7–22.
42. G. L. Cain and M. Z. Nashed, Fixed points and stability for a sum of two
operators in locally convex spaces, Pacific J. Math. 39 (1971), 581–592.
43. A. Capietto, M. Henrard, J. Mawhin and F. Zanolin, A continuation ap-
proach to forced superlinear Sturm-Liouville BVP’s, Topological Meth-
ods in Nonlinear Anal. 3(1994), 81–100.

Copyright 2006 by Taylor & Francis Group, LLC

 198 Topological Degree Theory and Applications

44. A. Capietto, J. Mawhin and F. Zanolin, A continuation approach to

superlinear periodic BVP’s, J. Diff. Equat. 88(1990), 347–395.

45. A. Capietto, J. Mawhin and F. Zanolin, Continuation theorems for pe-

riodic perturbations of autonomous systems, Trans. Amer. Math. Soc.
329(1992), 41–72.

46. R. Cauty, Solution du probléme de point fixe de Schauder, Fund. Math.

170 (2001), 231–246.

47. A. Cellina and A. Lasota, A new approach to the definition of topological

degree for multivalued mappings, Atti. Acad. Naz. Lincei Rend. Cl.
Sci. Fis. Mat. Natur. 47(1969), 434–440.

48. K. C. Chang, Variational methods for non-differentiable functionals and

their applications to partial differential equations, J. Math. Anal. Appl.
80(1981), 102–129.

49. K. C. Chang, Critical Point Theory and Its Applications, Shanghai Sci-
ence and Technology Publishing House, Shanghai, 1986.

50. K. C. Chang, Infinite Dimensional Morse Theory and Multiple Solution

Problems, Prog. Non. Different. Equat. Appl. 6, Birkhäuser, Boston,
1993.

51. S. S. Chang and Y. Q. Chen , Degree theory for multivalued (S) type
mappings and fixed point theorems, Appl. Math. Mech. 11 (1990),
441–454.

52. S. S. Chang and Y. Q. Chen, Coincidence index methods in studying

accretive operator equations, Chinese Annals of Math. 5(1993), 579–
583.

53. S. S. Chang, Y. Q. Chen, Y. J. Cho and B. S. Lee, Fixed point the-

orems for nonexpansive operators with disssipative perturbations in
cones, Comment. Math. Univ. Carolinae 39(1998), 49–54.

54. Y. Q. Chen, On accretive operators in cones of Banach spaces, Non.

Anal. 27(1996), 1125–1135.

55. Y. Q. Chen, Periodic solutions for evolution equations in Hilbert spaces,

Yokohama Math. J. 44 (1997), 43–53.

56. Y. Q. Chen, On the semi-monotone operator theory and applications,

J. Math. Anal. Appl. 231(1999), 177–192.

57. Y. Q. Chen and Y. J. Cho, On 1-set contraction perturbations of ac-

cretive operators in cones of Banach spaces, J. Math. Anal. Appl.
201(1996), 966–980.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 199

58. Y. Q. Chen, Y. J. Cho and L. Yang, Periodic solutions for nonlinear

evolution equations, Dynam. Cont. Disc. Impul. Syst. 9(2002), 581–
598.

59. Y. Q. Chen and Y. J. Cho, Anti-periodic solutions for semilinear evolu-

tion equations, J. Concrete and Appl. Math. 1(2003), 113–124.

60. Y. Q. Chen and Y. J. Cho, Nonlinear Operator Theory in Abstract

Spaces and Applications, Nova Science Publishers, New York, 2004.

61. Y. Q. Chen and Y. J. Cho, Topological Degree Theory for Multi-Valued

Mappings of Class (S+ )L , Arch. Math. 84(2005), 325–333.

62. Y. Q. Chen, Y. J. Cho and J. S. Jung, Anti-periodic solutions for evo-

lution equations, Math. Computer Model. 40(2004), 1123–1130.

63. Y. Q. Chen, Y. J. Cho and D. O’Regan, On Positive Fixed Points of

Countably Condensing Mappings, Dynamics Cont. Discrete and Impul.
Systems 12(2005), 519–527.

64. Y. Q. Chen, Y. J. Cho and D. O’Regan, On Perturbations of Accretive

Mappings, Appl. Math. Lett. 18(2005), 775–781.

65. Y. Q. Chen, Y. J. Cho and D. O’Regan, Anti-periodic solutions for

evolution equations with mappings in the class (S+ ), Math. Nachr.
278(2005), 356–362.

66. Y, Q, Chen and J. K. Kim, Anti-periodic solutions for first order semi-
lnear evolution equations, Non. Funct. Anal. Appl. 8(2003), 49–57.

67. Y, Q, Chen and J. K. Kim, Existence of periodic solutions for first-order

evolution equations without coercivity, J. Math. Anal. Appl. 282(2003),
801–815.

68. Y. Q. Chen and D. O’Regan, Coincidence degree theory for mappings

of class L-(S+ ), (submitted).

69. Y. Q. Chen and D. O’Regan, Generalized degree theory for semilinear

operator equations, Glasgow Math. J. 48(2006), 65–73.

70. Y. Q. Chen, X. D. Wang and H. X. Xu, Anti-periodic solutions for

semilinear evolution equations, J. Math. Anal. Appl. 273(2002), 627–
636.

71. S. N. Chow and J. K. Hale, Methods of Bifurcation Theory, Springer,

1982.

72. F. H. Clarke, Optimization and Nonsmooth Analysis, Wiley, New York,

1983.

Copyright 2006 by Taylor & Francis Group, LLC

 200 Topological Degree Theory and Applications

73. C. Conley and E. Zehnder, A Morse type index theory for flows and
periodic solutions to Hamiltonian systems, Comm. Pure Appl. Math.
37(1984), 207–253.
74. M. G. Crandall and T. Liggett, Generation of semigroups of nonlinear
transformations on general Banach spaces, Amer. J. Math. 93(1971),
265–298.
75. M. G. Crandall, H. Ishii and P. L. Lions, User’s guide to viscosity solu-
tions of second order partial differential equations, Bull. Amer. Math.
Soc. 27(1992), 1–67.
76. J. Cronin, Fixed Points and Topological Degree in Nonlinear Analysis,
Math. Surveys, No. 11, Amer. Math. Soc., 1964.
77. S. J. Daher, On a fixed point principle of Sadovskii, Nonlinear Anal.
2(1978), 643–645.
78. S. J. Daher, Fixed point theorems for nonlinear operators with a se-
quential condition, Nonlinear Anal. 3(1979), 59–63.
79. E. N. Dancer, On the indices of fixed points in cones and applications,
J. Math. Anal. Appl. 91(1983), 131–151.
80. E. N. Dancer, A new degree for S 1 -invariant gradient mappings and
applications, Anal. Nonlinear 2(1985), 329–370.
81. E. N. Dancer, Degree theory on convex sets and applications to bifurca-
tion, Calculus of variations and partial differential equations, Topics on
geometrical evolution problems and degree theory, Springer, 185–241,
2000.
82. E. N. Dancer, R. D. Nausbaum and C. A. Stuart, Quasinormal cones in
Banach spaces, Nonlinear Anal. 7(1983), 539–553.
83. E. N. Dancer and J. Toland, Degree theory for orbits of prescribed period
for flows with a first integral, Proc. London Math. Soc. 60(1990), 549–
580.
84. B. Decorogna, Weak Continuity and Weak Lower Semicontinuity of Non-
linear Functionals, Lect. Notes Math. 922, Springer, Berlin, 1982.
85. H. Debrunner and P. Flor, Ein Erweiterungssatz fur monotone Mengen,
Arch. Math. 15(1964), 445–447.
86. K. Deimling, Nonlinear Functional Analysis, Springer-Verlag, Berlin,
1985.
87. K. Deimling, Multivalued Differential Equations, W. De Gruyter, Berlin,
1992.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 201

88. J. Diestel W. M. Ruess and W. Schachermayer, Weak compactness in

L1 (µ, X), Proc. Amer. Math. Soc. 118(1993), 447–453.

89. W. Y. Ding, A generalization of the Poincare-Birkhoff theorem, Proc.

Amer. Math. Soc. 88(1983), 341–346.

90. S. Domachowski, An application of the topological degree and com-

pletely continuous extensions of selectors to nonconvex differential in-
clusions, Nonlinear Anal. 27(1996), 987–997.

91. Y. H. Du, Fixed points of increasing operators in ordered Banach spaces

and applications, Appl. Anal. 38(1990), 1–20.

92. Y. H. Du, A degree theoretic approach to N -species periodic competition

systems on the whole Rn , Proc. Roy. Soc. Edinburgh 129(1999), 295–
318.

93. J. Dugundji and A. Granas, Fixed Point Theory I, Warszawa, 1982.

94. G. Duvaut and J. L. Lions, Les Inequations en Mecanique et en Physique,

Dunod, 1970.

95. Z. Dzedzej, Fixed point index theory for a class of nonacyclic multivalued
maps, Dissert. Math. 253(1985), 1–53.

96. S. Eilenberg and D. Montgomery, Fixed point theorems for multivalued

transformations, Amer. J. Math. 58(1946), 214–222.

97. D. Eisenbud and H. Levine, The Topological degree of a finite C ∞ map

germ, in Structural Stability, the Theory of Catastrophes, and Applica-
tions in the Sciences, Lecture Notes in Math. 525, Springer-Verlag.

98. I. Ekland, On the variational principle, J. Math. Anal. Appl. 47(1974),

324–353.

99. I. Ekland and H. Hofer, Periodic solutions with prescribed minmal

period for convex autonomous Hamiltonian systems, Invent. Math.
81(1985), 155–188.

100. I. Ekland and J. M. Lasry, On the number of periodic trajectories for a

Hamiltonian flow on a convex energy surface, Ann. Math. 112(1980),
283–319.

101. I. Ekland and R. Temam, Convex Analysis and Variational Problems,

North-Holand, 1976.

102. K. D. Elworthy and A. J. Tromba, Degree theory on Banach manifolds,

Proc. Symp. Pure Math. 18, Part I, 86-94, Amer. Math. Soc., RI,
1970.

Copyright 2006 by Taylor & Francis Group, LLC

 202 Topological Degree Theory and Applications

103. L. H. Erbe, W. Krawcewicz and J. Wu, Leray-Schauder degree for semi-

linear Fredholm maps and period BVP’s of neutral equations, Nonlinear
Anal. 15(1990), 747–764.

104. L. H. Erbe, W. Krawcewicz and J. Wu, A composite coincidence degree

with applications to BVP’s of neutral equations, Trans. Amer. Math.
Soc. 335(1993), 459–478.

105. L. C. Evans, On solving certain nonlinear partial differential equations

by accretive operator methods, Israel J. Math. 36(1980), 225–247.

106. E. R. Fadell, Recent results in the fixed point theory of continuous maps,
Bull. Amer. Math. Soc. 76(1970), 10–29.

107. E. R. Fadell and S. Husseini, Relative cohomological index theories,

Advan. Math. 64(1987), 1–31.

108. E. R. Fadell, S. Husseini and P. H. Rabinowitz, Borsuk-Ulam theorems

for arbitrary S 1 actions and applications, Trans. Amer. Math. Soc.
274(1982), 345–360.

109. K. Fan, Fixed-point and minimax theorems in locally convex topological

linear spaces, Proc. Nat. Acad. Sci. USA 38(1952), 121–126.

110. K. Fan, A generalization of Tychonoff’s fixed point theorem, Math. Ann.

142(1961), 305–310.

111. K. Fan, Sur un theoreme minimax, C. R. Acad. Sci. Paris, Groupe 1

142(1962), 3925–3928.

112. K. Fan, Some properties of convex sets related to fixed point theorems,
Math. Ann. 266(1984), 519–537.

113. M. Feckan and R. Kollar, Discontinuous wave equations and a topo-

logical degree for some classes of multi-valued mappings, Appl. Math.
44(1999), 15–32.

114. V. V. Fillipov, Basic Topological Structures of Odinary Differential

Equations, Kluwer Academic Publishers, 1998.

115. P. M. Fitzpatrick, On the structure of the set of solutions of equations

of A-proper mappings, Trans. Amer. Math. Soc. 189(1974), 107–131.

116. M. Furi, M. Martelli and A. Vigonoli, On the solvability of nonlinear

operators in normed spaces, Ann. Mat. Pure Appl. 124(1980), 321–343.

117. P. M. Fitzpatrick, A generalized degree for uniform limits of A-proper

mappings, J. Math. Anal. Appl. 35(1971), 536–552.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 203

118. P. M. Fitzpatrick, On the structure of the set of solutions of equations

involving A-proper mappings, Trans. Amer. Math. Soc. 189(1974),
107–131.

119. P. M. Fitzpatrick, Existence results for equations involving noncompact

perturbations of Fredholm mappings with applications to differential
equations, J. Math. Anal. Appl. 66(1978), 151–171.

120. P. M. Fitzpatrick and W. V. Petryshyn, Fixed point theorems and the

fixed point index for multivalued mappings in cones, J. London Math.
Soc. 12(1975), 75–85.

121. P. M. Fitzpatrick and J. Pejsachowicz, Orientation and the Leray-Schau-

der theory for fully nonlinear elliptic bounday value problems. Mem.
Amer. Math. Soc. 101(1993).

122. P. M. Fitzpatrick and J. Pejsachowicz, An extension of the Leray-Schau-

der degree for fully nonlinear elliptc problems, Proc. Symp. Pure Math-
ematics 45(1986), 425–438.

123. P. M. Fitzpatrick and J. Pejsachowicz, The fundamental group of the

space of linear Fredholm operators and global analysis of semilinear
equations, Contemp. Math. 72(1998), 47–87

124. I. Fonseca and W. Gangbo, Degree Theory in Analysis and Applications,

Clarendon Press, Oxford, 1995.

125. M. Frigon and D. O’Regan, Fixed points of cone-compressing and cone-

extending operators in Frechet spaces, Bull. London Math. Soc. 35(2003),
672–680.

126. S. Fucik, M. Kucera and J. Necas, Ranges of nonlinear asymptotically

linear operators, J. Diff. Equat. 17(1975), 375–394.

127. S. Fucik, J. Necas, J. Soucek and V. I. Soucek, Spectral Analysis of

Nonlinear Operators, Lect. Notes in Math. 346, Springer-Verlag, 1973.

128. M. Furi, M. Martelli and M. P. Pera, Cobifurcating branches of solutions

of nonlinear eigenvalue problems in Banach spaces, Ann. Mat. Pura
Appl. 135(1983), 119–131.

129. M. Furi, M. Martelli and M. P. Pera, On the existence of forced os-

cillations for the spherical pendulum, Boll. dell’Unione Mat. Italiana
4-B(1990), 381390

130. M. Furi, M. Martelli and M. P. Pera, A continuation theorem for periodic

solutions of forced motion equations on manifolds and applications to
bifurcation theory, Pacific J. Math. 160(1993), 219–244.

Copyright 2006 by Taylor & Francis Group, LLC

 204 Topological Degree Theory and Applications

131. M Furi, M. Martelli and A. Vigonoli, On the solvability of nonlinear

operators in normed spaces, Ann. Mat. Pure Appl. 124(1980), 321–
343.

132. E. Getzler, Degree theory for Wiener maps, J. Func. Anal. 68 (1986),
388–403.

133. E. Getler, The degree of the Nicolai map, J. Func. Anal. 74(1987),
121–138.

134. R. E. Gaines and J. Mawhin, Coincidence degree and nonlinear differ-

ential equations, Lect. Notes in Math. 568, Springer-Verlag, 1977.

135. D. Gilbarg and N. S. Trudinger, Elliptic Partial Differential Equations

of Second Order, Springer-Verlag, Berlin, 1983.

136. I. Glicksberg, A further generalization of the Kakutani fixed-point the-

orem, with application to Nash equilibrium points, Proc. Amer. Math.
Soc. 3(1952), 170–174.

137. L. Gorniewicz, Topological Fixed Point Theory of Multivalued Map-

pings, Kluwer Academic Publishers, 1999.

138. J. P. Gossez, Opérateutrs monotones non-linéaires dans les espaces de

Banach nonréflexifs, J. Math. Anal. Appl. 44(1973), 71–87.

139. A. Granas, The Leray-Schauder index and the fixed point theory for
arbitrary ANR’s, Bull. Soc. Math. France 100(1972), 209–228.

140. A. Granas, KKM-maps and Their Applications to Nonlinear Problems,

The Scottish Book, R. D. Mauldin, Ed., Birkhauser, 1981, 45–61.

141. D. J. Guo, A new fixed point theorem, Acta Math. Sinica 24(1981),
444–450.

142. D. J. Guo, Nonlinear Functional Analysis, Shangdon Science and Tech-

nology Publishing Press, Shangdon, 1985.

143. D. J. Guo, Partial Order Method in Nonlinear Analysis, Shangdon Sci-

ence and Technology Publishing Press, Shangdon, 1997.

144. D. Guo and V. Lakshmikantham, Nonlinear Problems in Abstract Cones,

Academic Press, 1988.

145. O. Hadzic, On Kakutani’s fixed point theorem in topological vector

spaces, Bull. Acad. Polon. Sci. Sér. Sci. Math. 30(1982), 141–144.

146. B. R. Halpern and G. M. Bergman, A fixed point theorem for inward

and outward maps, Trans. Amer. Math. Soc. 130(1968), 353–358.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 205

147. F. B. Hang and F. H. Lin, Topology of Sobolev mappings II, Acta Math.
191(2003), 55–107.

148. A. Haraux, Anti-periodic solutions of some nonlinear evolution equa-

tions, Manuscripta Math., V.63, 1989, 479–505

149. P. Hartman and G. Stampacchia, On some nonlinear elliptic differential

functional equations, Acta Math. 115(1966), 271–310.

150. M. Hazewinkel and M. van de Vel, On almost fixed point theory, Canad.
J. Math. 30(1978), 673–699.

151. C. J. Himmelberg, Fixed points of compact multifunctions, J. Math.

Anal. Appl. 38(1972), 205–207.

152. H. Hofer, Variational and topological methods in partially ordered Hilbert

spaces, Math. Ann. 261(1982), 493–514.

153. H. Hofer, A note on the topological degree at a critical point of mountain-

pass-type, Proc. Amer. Math. Soc. 90(1984), 309–315.

154. C. D. Horvath and M. Lassonde, Leray-Schauder spaces, Nonlinear Anal.

45(2001), 923–931.

155. S. Hu and Y. Sun, Fixed point index for weakly inward mappings, J.
Math. Anal. Appl. 172(1993), 266–273.

156. A. D. Ioff and V. M. Tihemirov, Theory of Extremal Problems, North-

Holland, 1979.

157. G. Isac, 0-epi families of mappings, topological degree and optimization,

J. Opt. Theory Appl. 42(1984), 51–75

158. S. Ishikawa, Fixed points and iteration of a nonexpansive mapping in a

Banach space, Proc. Amer. Math. Soc. 59(1976), 65–71.

159. I. Ize, I. Massabo and A. Vignoli, Degree theory for equivariant maps,
315(1989), 433–510.

160. L. Janos and M. Martelli, Sequentially condensing maps, Univ. u Novom

Sadu Zb. Rad. Priod. Mat. Fak. Ser. Mat. 10(1986), 85–94.

161. J. Jost and M. Struwe, Morse-Conley theory for minimal surface of

varing topological type, Invent. Math. 102(1990), 465–499.

162. T. Kaczynski and W. Krawcewicz, Solvability of BVP’s for the inclusion

utt −uxx ∈ g(t, x, u) via the theory of multi-valued A-Proper maps, ZAA
7(1988), 337–346.

Copyright 2006 by Taylor & Francis Group, LLC

 206 Topological Degree Theory and Applications

163. A. G. Kartsatos and I. V. Skrypnik, Topological degree theories for

densely defined mappings involving operators of type (S+ ), Advan. in
Different. Equat. 4(1999), 413–456.
164. A. G. Kartsatos and I. V. Skrypnik, The index of a critical point for
nonlinear elliptic operators with strong coefficient growth, J. Math. Soc.
Japan 52(2000), 109–137.
165. A. G. Kartsatos and I. V. Skrypnik, The index of a critical point for
densely defined operators of type (S+ )L in Banach spaces, Trans. Amer.
Math. Soc. 354(2001), 1601–1630.
166. A. G. Kartsatos and I. V. Skrypnik, A new topological degree theory for
densely defined quasibounded (Se+ )-perturbations of multivalued maxi-
mal monotone operators in reflexive Banach spaces, Abstr. Appl. Anal.
2005(2005), 121–158.
167. T. Kato, Nonlinear semigroups and evolution equations, J. Math. Soc.
Japan 19(1967), 508–520.
168. D. Kinderlehrer and G. Stampacchia, An Introduction to Variational
Inequalities and Their Applications, Academic Press, New York, 1980.
169. W. A. Kirk, A fixed point theorem for mappings which do not increase
distance, Amer. Math. Monthly 72(1965), 1004–1006.
170. V. Klee, Leray-Schauder theory without local convexity, Math. Ann.
141(1960), 286–296.
171. W. Klingenberg, Riemannian Geometry, Walter de Gruyter, 1982.
172. J. Kobayashi and M. Otani, Topological degree for (S)+ mappings with
maximal monotone perturbations and its applications to variational in-
equalities, Nonlinear Anal. 59(2004), 147–172.
173. Y. Kobayashi, Difference approximation of Cauchy problems for quasi-
dissipative operators and generation of nonlinear semigroups, J. Math.
Soc. Japan 27(1975), 640–665.
174. Y. Komura, Semigroups of operators in locally convex spaces, J. Funct.
Anal. 2(1968), 258–296.
175. J. Korevaar, On a question of Brezis and Nirenberg concerning the de-
gree of circle maps, Selecta Math. 5(1999), 107–122.
176. M. A. Krasnoselskii, Positive Solutions of Operator Equations, Gronin-
gen, Noordhoff, 1961.
177. M. A. Krasnoselskii, Topological Methods in the Theory of Nonlinear
Integral Equations, Macmillan, New York, 1964.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 207

178. M. A. Krasnoselskii and P. P. Zabreiko, Geometrical Methods of Non-

linear Analysis, Springer-Verlag, Berlin, 1984.
179. K. Kuratowski, C. R. Nardzewski, A general theorem on selections, Bull.
Acad. Polon. Soc. 13(1965), 397–403.
180. E. Lami-Dozo, Quasinormality in cones in Hilbert spaces, Acad. Roy.
Belg. Bull. CI. Sci. 30(1981), 531–541.
181. K. Q. Lan and J. R. L. Webb, A fixed point index for weakly inward
A-proper maps, Nonlinear Anal. 28(1997), 315–325.
182. K. Q. Lan and J. R. L. Webb, A fixed point index for generalized inward
mappings of condensing type, Trans. Amer. Math. Soc. 349(1997),
2175–2186.
183. R. Leggett and L. Williams, Multiple positive fixed points of nonlinear
operators on ordered Banach spaces, Indiana Univ. Math. J. 28(1979),
673–688.
184. J. Leray, Les problemes nonlineaires, Enseign. Math. 30(1936), 141.
185. J. Leray and J. Schauder, Topologie et equations fonctionnelles, Ann.
Sci. Ecole. Norm. Sup. 51(1934), 45–78.
186. G. Z. Li, A new fixed point theorem on semicompact 1-set contraction
mappings, Proc. Amer. Math. Soc. 97(1986), 277–280.
187. S. J. Li and D. X. Feng, Degree theory for set-valued maximal monotone
operators in Hilbert spaces, Acta Math. Sinica 25(1982), 533–541.
188. Y. Y. Li, Degree theory for second order nonlinear elliptic operators and
its applications, Comm. in PDE 14(1989), 1541–1578.
189. T. C. Lim, A fixed point theorem for multivalued nonexpansive map-
pings in a uniformly convex banach space, Bull. Amer. Math. Soc.
80(1974), 1123–1126.
190. C. S. Lin, Topological degree mean field equations on S 2 , Duke Math.
J. 104(2000), 501–536
191. J. Lindenstrauss and L. Tzafriri, Classical Banach Spaces I, Springer-
Verlag, Berlin, 1977.
192. J. L. Lions and E. Magenes, Non-Homogeneous Boundary Value Prob-
lems and Applications, I, II, III, Springer-Verlag, Berlin, 1972, 1973.
193. C. G. Liu and Y. M. Long, Iteration inequalities of the Maslov-type
index theory with applications, J. Different. Equat. 165(2000), 355–
376.

Copyright 2006 by Taylor & Francis Group, LLC

 208 Topological Degree Theory and Applications

194. N. G. Lloyd, Degree Theory, Cambridge University Press, Cambridge,

1978.
195. Y. M. Long, A Maslov-type index theory for symplectic paths., Topo-
logical Methods in Nonlinear Anal. 10(1997), 47–78.
196. G. Lumer, Semi-inner product spaces, Trans. Amer. Math. Soc.
100(1961), 29–43.
197. T. W. Ma, Topological degree for set-valued compact vector fields in
locally convex spaces, Dissertationes Math. 92(1972), 1–43.
198. R. H. Martin, Differential equations on closed subsets of a Banach space,
Trans. Amer. Math. Soc. 179(1973), 399–414.
199. R. H. Martin, Nonlinear Operators and Differential Equations in Banach
Spaces, Wiley, New York, 1976.
200. I. Massabo and C. A. Stuart, Positive eigenvectors for k-set contractions,
Nonlinear Anal. 3(1979), 35–44.
201. J. L. Massera and J. J. Schaffer, Linear Differential Equations and Func-
tion Spaces, Academic Press, New York, 1966.
202. R. D. Mauldin, The Scottish book: Mathematical problems from the
Scottish Cafe, Birkhauser, 1981.
203. J. Mawhin, Topological Degree Methods in Nonlinear Boundary Value
Problems, Amer. Math. Soc. 40, Providence, RI, 1979.
204. J. Mawhin, Degré topologique et solutions périodiques des systémes
différentiels non linéaires, Bull. Soc. Roy. Sci. Liége 38(1969), 308–398.
205. J. Mawhin, Equivalence theorems for nonlinear operator equations and
coincidence degree theory for some mappings in locally convex topolog-
ical spaces, J. Diff. Equat. 12(1972), 610–636.
206. J. Mawhin, A futher invariance property of coincidence degree in convex
spaces, Ann. Soc. Sci. Bruxelles Sér. I 87(1973), 51–57.
207. J. Mawhin, Leray-Schauder continuation theorems in the absence of a
prior bounds, Topological Methods in Nonlinear Anal. 9(1997), 179–
200.
208. J. Mawhin, Leray-Schauder degree: a half century of extensions and
applications, Topological Methods in Nonlinear Anal. 14(1999), 195–
228.
209. J. Mawhin and W. Willem, Critical Point Theory and Hamiltonian Sys-
tems, Springer-Verlag, Berlin, 1989.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 209

210. S. Mazur, Uber konvexe in linearen normierten Raumen, Studia Math.

4(1933), 70–84.

211. R. E. Megginson, An Introduction to Banach Space Theory, Springer-

Verlag, Berlin, 1998.

212. E. Michael, Continuous selections I, Ann. Math. 63(1956), 361–382.

213. M. Mininni, Conicidence degree and solvability of some nonlinear func-

tional equations in normed spaces: a spectral approach, Nonlinear Anal.
2(1978), 597–607.

214. J. Milnor, Analytic proof of the hairy ball theorem and the Brouwer
fixed point theorem, Amer. Math. Monthly 85(1978), 521–524.

215. P. S. Milojevic, A generalization of Leray-Schauder’s theorem and sur-

jectivity results for multivalued A-proper and pseudo A-proper map-
pings, Nonlinear Anal. 1(1977), 263–276.

216. P. S. Milojevic, Fredholm alterative and surjectivity results for multival-

ued A−proper and condensing mappings with applications to nonlinear
integral and differential equations, Czech. Math. J. 30(1980), 387–417.

217. P. S. Milojevic, Solvability of operator equations involving nonlinear

perturbations of Fredholm mappings with nonnegative index and ap-
plications, in Lect. Notes Math., 957, pp. 212–228, Springer-Verlag,
1982.

218. G. J. Minty, Monotone (nonlinear) operators in Hilbert spaces, Duke

Math. J. 29(1962), 341–346.

219. G. J. Minty, On the generalization of a direct method of calculus of

variations, Bull. Amer. Math. Soc. 73(1967), 315–321.

220. H. Mönch, Boundary value problems for nonlinear ordinary differential

equations in Banach spaces, Non. Anal. 4(1980), 985–999.

221. H. Mönch and G. F. von Harten, On the Cauchy problems for ordinary
differential equations in Bnach spaces, Arch. Math. 39(1982), 153–160.

222. C. Morales, Remarks on compact perturbations of m-accretive operator,

Nonlinear Anal. 16(1991), 771–780.

223. S. B. Nadler, Multi-valued contraction mappings, Pacific J. Math. 30(19

69), 475–487.

224. M. Nagumo, A theory of degree based on infinitesimal analysis, Amer.

J. Math. 73(1951), 485–496.

Copyright 2006 by Taylor & Francis Group, LLC

 210 Topological Degree Theory and Applications

225. M. Nagumo, Degree of mappings in convex linear topological spaces,

Amer. J. Math. 73(1951), 497–511.

226. W. M. Ni, Some minimax principles and their applications in nonlinear

elliptic equations, J. Anal. Math. 37(1980), 248–275.

227. L. Nirenberg, Topics in Nonlinear Functional Analysis, Lecture Notes,

Courant Institute, New York, 1974.

228. L. Nirenberg, Variational and topological methods in nonlinear prob-

lems, Bull. Amer. Math. Soc. 4(1981), 267–302.

229. R. D. Nussbaum, Degree theory for local condensing maps, J. Math.

Anal. Appl. 37(1972), 741–766.

230. R. D. Nussbaum, On the uniqueness of the topological degree for k-set

contractions, Math. Z. 137(1974), 1–6.

231. R. D. Nussbaum, Estimates for the number of solutions of operator

equations, Appl. Anal. 1(1971), 183–200.

232. V. V. Obkhovskii, On the topological degree for a class of noncompact

multivalued mappings, Funkt. Anal. 23(1982), 82–93.

233. H. Okochi, On the existence of anti-periodic solutions to a nonlinear evo-

lution equation associated with odd subdifferential operators, J. Funct.
Anal. 91(1990), 246–258.

234. H. Okochi, On the existence of anti-periodic solutions to nonlinear par-

abolic equations in noncylindrical domains, Nonlinear Anal. 14(1990),
771–783.

235. T. Okon, A generalized topological degree in admissible linear spaces,

Zeit. für Anal. Anwend. 14(1995), 469–496.

236. D. O’Regan and R. Precup, Theorems of Leray-Schauder Type and Ap-

plications, Gordon and Breach, 2001.

237. R. S. Palais, Morse theory on Hilbert manifolds, Topology 2(1963), 299–

340.

238. R. S. Palais, Lusternik-Schnirelman theory on Banach manifolds, Topol-

ogy 5(1966), 115–132.

239. R. S. Palais, Critical point theory and the minimax principle, Proc.
Symp. Pure Math. 15(1970), 185–212.

240. R. S. Palais and S. Smale, A generalized Morse theory, Bull. Amer.

Math. Soc. (1964), 165–171.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 211

241. D. Pascali and S. Sburlan, Nonlinear Mappings of Monotone Type, No-

ordhoff, Leyden, 1978.
242. N. Pavel, Nonlinear Evolution Operators and Semigroups, Lect. Notes
Math., 1260, Springer-Verlag, Berlin, 1987.
243. A. Pazy, Semigroups of Linear Operators and Applications to Partial
Differential Equations, Springer-Verlag, Berlin, 1983.
244. W. V. Petryshyn, Antipodes theorems for A-proper mappings of the
modified type (S) or (S)+ and to mappings with the Pm property, J.
Funct. Anal. 71 (1971), 165–211.
245. W. V. Petryshyn and P. M. Fitzpatrick, A degree theory, fixed point the-
orems, and mapping theorems for multivalued noncompact mappings,
Trans. Amer. Math. Soc. 194(1974), 1–25.
246. W. V. Petryshyn, On the solvability of x ∈ T x + λF x in quasinormal
cones with T and F k-set contractive, Nonlinear Anal. 5(1981), 585–591.
247. W. V. Petryshyn, Existence of fixed points of positive k-set contractive
maps as consequences of suitable boundary conditions, J. London Math.
Soc. 38(1988), 503–512
248. W. V. Petryshyn, Generalized Topological Degree and Semilinear Equa-
tions, Cambridge University Press, Cambridge, 1995
249. R. S. Phillips, Integration in convex linear topological spaces, Trans.
Amer. Math. Soc. 47 (1940), 114–145.
250. T. Pruszko, Some applications of the topological degree theory to multi-
valued boundary value problems, Dissert. Math. 229(1984), 1–52.
251. T. Pruszko, Topological degree methods in multi-valued boundary value
problems, Nonlinear Anal. 9(1981), 959–973.
252. P. H. Rabinowitz, Some aspects of nonlinear eigenvalue problems, Rocky
Mountain J. Math. 3(1973), 161–202.
253. P. H. Rabinowitz, A note on topological degree for potential operators,
J. Math. Anal. Appl. 51(1975), 483–492.
254. I. Rachunkova and S. Stanek, Topological degree method in functional
boundary value problems, Nonlinear Anal. 27(1996), 153–166.
255. S. Reich, Some problems and results in fixed point theory, Contemporary
Math. 21(1983), 179–187.
256. T. Riedrich, Die Räume Lp (0, 1) (0 < p < 1) sind sulässig, Wiss. Z.
Techn. Univ. Dresden 12(1963), 1149–1152.

Copyright 2006 by Taylor & Francis Group, LLC

 212 Topological Degree Theory and Applications

257. T. Riedrich, Der Raum S(0, 1) ist zulässig, Wiss Z. TU Dresden 13(1964),
1–6.
258. J. W. Roberts, A compact convex set with no extreme points, Studia
Math. 60(1977), 255–266.
259. R. T. Rockafellar, Characterization of the subdifferential of convex func-
tions, Pacific J. Math. 17(1966), 497–510.
260. R. T. Rockafellar, Convex Analysis, Princeton University Press, Prince-
ton, 1970.
261. R. T. Rockafellar, On the maximal monotonicity of subdifferential map-
pings, Pacific J. Math. 33(1970), 209–216.
262. R. T. Rockafellar, Local boundedness of nonlinear monotone operators,
Michigan Math. J. 16(1970), 397–407.
263. S. Rolewicz, Metric Linear Spaces, Polish Scientific Publishers, Warszawa,
1972.
264. E. H. Rothe, A relation between the type numbers of a critical point and
the index of the corresponding field of gradient vectors, Math. Nachr.
4(1950-51), 12–17.
265. E. H. Rothe, On the Cesari index and the Browder-Petryshyn degree, in
International Symposium on Dynamic Systems, pp. 295–312, Academic
Press, New York, 1977.
266. E. H. Rothe, Introduction to Various Aspects of Degree Theory in Ba-
nach Spaces, Amer. Math. Soc., Providence, RI, 1986.
267. W. Rudin, Functional Analysis, MacGraw-Hill, New York, 1973.
268. K. P. Rybakowski, On the homotopy index for the infinite-dimensional
semiflows, Trans. Amer. Math. Soc. 269(1982), 351–382.
269. K. P. Rybakowski and E. Zehnder, On a Morse equation in Conley’s
index theory for semiflows on metric spaces, Ergodic Theory Dyn. Syst.
5(1985), 123–143.
270. B. Rzepecki, Remarks on Schauder’s fixed point principle and its appli-
cations, Bull. Acad. Polon. Sci. Sér. Sci. Math. 27(1979), 473–480.
271. B. N. Sadovskii, Limit-compact and condensing operators, Uspekhi Mat.
Nauk. 27(1972), 81–146.
272. S. S. Schaefer, Topological Vector Spaces, Macmillan, New York, 1966.
273. J. Schauder, Der Fixpunktsatz in Funktionalraumen, Studia Math. 2(19
30), 171–180.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 213

274. J. T. Schwartz, Nonlinear Functional Analysis, Gordon and Breach, New

York, 1969.

275. J. C. Scovel, A simple intuitive proof of a theorem in degree theory for

gradient mappings, Proc. Amer. Math. Soc. 93(1985), 751–753.

276. V. M. Sehgal and E. Morrison, A fixed point theorem for mutifunctions,

Proc. Amer. Math. Soc. 38(1973), 643–646.

277. H. W. Sieberg, Some historical remarks concerning degree theory, Amer.

Math. Monthly (1981), 125–139.

278. M. Sion, On general minimax theorems, Pacific J. Math. 8(1958), 171–

176.

279. I. V. Skrypnik, Nonlinear Elliptic Equations of Higher Order, Nauk.

Dumka, Kiev, 1973.

280. I. V. Skrypnik, Sovlvability and properties of solutions of nonlinear el-

liptic equations, VINITI, Vol. 9, 131–254, Moscow, 1976.

281. M. Struwe, Variational Methods, 2nd ed., Springer-Verlag, Berlin, 2000.

282. A. Sulkin, Positive solutions to variational inequalities, a degree theoritic

approach, J. Diff. Equat. 57(1985), 90–111.

283. J. X. Sun, A generalization of Guo’s theorem and applications, J. Math.

Anal. Appl. 126(1987), 12–17.

284. Y, Sun and J. X. Sun, Multiple positive fixed points of weakly inward
mappings, J. Math. Anal. Appl. 148(1990), 431–439.

285. W. Takahashi, Nonlinear Functional Analysis, Yokohama Publishers,

Japan, 2000.

286. J. F. Toland, A Leray-Schauder degree calculation leading to nonstan-

dard global bifurcation results, Bull. London Math. Soc. 15(1983),
149–154.

287. S. L. Trojansky, On locally uniformly convex and differentiable norms in

certain nonseparable Banach spaces, Studia Math. 37(1971), 173–180.

288. T. S. Tucker, Leray-Schauder theorem for P-compact operators and its

consequences, J. Math. Anal. Appl. 23(1972), 355–364.

289. A. Tychonoff, Ein fixpunktsatz, Math. Ann. 111(1935), 767–776.

290. M. M. Vainberg, A new principle in the theory of nonlinear equations,

Usphehi Mat. Nauk SSS 15(1960), 243–244.

Copyright 2006 by Taylor & Francis Group, LLC

 214 Topological Degree Theory and Applications

291. M. Vath, Fixed point theorems and fixed point index for countably
condensing maps, Topological Methods in Nonlinear Anal. 13(1999),
341–363.

292. M. Visser, Topological degree for supersymmetric chiral models, Phys.

Rev. D. 32(1985), 510–512.

293. V. A. Volpert and A. I. Volpert, Properness and topological degree for

general elliptic operators, AAA 3(2003), 129–181.

294. A. I. Volpert and V. A. Volpert, Application of the theory of the rota-

tion of vector fields to the investigation of wave solutions of parabolic
equations, Trans. Moscow Math. Soc. 1990 (1990), 59–108.

295. A. I. Volpert and V. A. Volpert, Construction of the Leray-Schauder

degree for elliptic operators in unbounded domains, Ann. Inst. H.
Poincaré Anal. Non Linéaire, 11(1994), 245–273.

296. V. A. Volpert, A. I. Volpert and J. F. Collet, Topological degree for

elliptic operators in unbounded cylinders, Adv. Differential Equations
4(1999), 777–812.

297. Z. Q. Wang, Symmetries and the calculations of degree, Chinese Ann.

Math. Ser. B 10(1989), 520–536.

298. G. F. Webb, Continuous nonlinear perturbations of linear accretive op-

erators, J. Funct. Anal. 10(1972), 191–203.

299. J. R. L. Webb, A-properness and fixed points of weakly inward maps,

J. London. Math. Soc. 27(1983), 141–149.

300. J. R. L. Webb, Topological degree and A-proper operators, Linear Al-

gebra Appl. 84(1986), 227–242.

301. J. Weiberg, A topological degree of locally A-proper maps, J. Math.

Anal. Appl. 65(1978), 66–79.

302. M. Willem, On a result of Rothe about the Cesari index and A-proper
mappings, Boll. Un. Mat. Ital. 17(1980), 178–182.

303. M. Willem, Topology and semilinear equations ar resonance in Hilbert

spaces, Nonlinear Anal. 5(1981), 517–524.

304. V. Williams, Closed Fredholm and semi-Fredholm operators, essential

spectra and perturbations, J. Funct. Anal. 20(1975), 1–25.

305. S. F. Wong, The topological degree of A-proper maps, Canad. J. Math.

23(1971), 403–412.

Copyright 2006 by Taylor & Francis Group, LLC

 REFERENCES 215

306. S. F. Wong, A product formula for the degree of A-proper maps, J.

Funct. Anal. 10(1972), 361–371.
307. J. Z. Xiao and X. H. Zhu, Topological degree theory and fixed point
theorems in fuzzy normed spaces, Fuzzy Sets and Systems 147(2004),
437–452.
308. K. Yosida, Functional Analysis, Springer-Verlag, 1969.
309. T. Yuasa, Differential equations in a locally convex space via the measure
of noncompactness, J. Math. Anal. Appl. 84(1981), 534–554.
310. E. H. Zarantonello, Solving functional equations by contractive averag-
ing, Math. Research Center Report 160, Madison, WI, 1960.
311. E. H. Zarantonello, Projections on convex sets in Hilbert space and spec-
tral theory, Contributions to Nonlinear Functional Analysis, Academic
Press, 1971.
312. E. Zeidler, Nonlinear Functional Analysis and its Applications, Springer-
Verlag, Berlin, 1990.
313. V. V. Zelokova, On a topological degree for Vietoris type multimaps in
locally convex spaces, Sb. Stat. Asp. Stud. Mat. Fak. Vgu. (1999),
Voronezh, 45–51.
314. Y. C. Zhao, On the topological degree for the sum of maximal monotone
operator and generalized pseudomonotone operator, Chinese Ann. Math.
4(1983), 241–253.

Copyright 2006 by Taylor & Francis Group, LLC