See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/323880629

Set-Valued map and Nadler,s Principal (Masters Project)

Method · November 2014

CITATIONS READS

0 676

3 authors, including:

Mohd Shoaib Khan

M. S. Ramaiah University of Applied Sciences
18 PUBLICATIONS 128 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mohd Shoaib Khan on 20 March 2018.

The user has requested enhancement of the downloaded file.

Contents

Preface (ii)
Notations and Abbreviations (iv)

1. Involved Concepts and Core Results ..........................................................1-14

1.1 Introduction ......................................................................................................1
1.2 Some Examples of Set-Valued Maps ....................................................................2
1.3 Algebraic and Topological Properties of Set-Valued Maps ...................................3
1.4 Certain Results on Set-Valued Maps ...................................................................7
1.5 Continuity of Set-Valued Maps ..........................................................................10
1.6 Relevant Weak Continuity Conditions ...............................................................12

2. Certain Results involving Hausdorff Metric ............................................16-27

2.1 Introduction .....................................................................................................16
2.2 Hausdorff Metric ..............................................................................................16
2.3 H-Continuity of Set-Valued Map ......................................................................20
2.4 Lipschitz Set-Valued Map ................................................................................23
2.5 Nadler Contraction Principle ...........................................................................25

3. Hybrid Fixed Point Theorems ...................................................................28-40

3.1 Introduction ....................................................................................................28
3.2 Some Fixed Point Theorems ............................................................................29
3.3 Some Hybrid Fixed Point Theorems ...............................................................33

References .............................................................................................................41

i
Preface

The aim of this dissertation is broadly two fold. The first half of this dissertation
is devoted to the relevant theory of set-valued maps which are later utilized to prove
Nadler’s contraction principle as well as hybrid fixed point theorems.
Kuratowski realized the importance of set-valued maps and devoted considerable space
in his ever famous book on Topology. Some other eminent mathematicians such as:
Painlevé, Hausdorff and Bouligand have also visualized the vital role of set-valued maps
as one often encounters with such concrete situations and objects in real daily life.
The theory of set-valued maps is a beautiful mixture of analysis, topology and geometry.
Over the last 30 years or so there has been an intense interest in this area of research.
This is partly due to the rich and fruitful applications in various diverse fields such
as: Biological Sciences, Control Theory and Optimization, Economics, Game Theory,
Physics and similar others.
With a view to have a deeper insight, we have chosen a large number of examples which
allow us to realize some intuition, draw conclusions and ideas involved in various proofs.
Our secondary aim of this dissertation is to develop a preparatory ground to undertake a
general fixed point theory. To achieve this goal, we have adopted the following strategy.
Often, we have shown that if we want to extend the result from single-valued fixed point
theory to set-valued fixed point theory, what kind of extra conditions are needed to be
imposed on set-valued maps. Via Remarks 2.2.1, 2.2.2 and 2.2.3, we have shown that
the metric d defined on a non-empty set X does not form a metric on the power set
P (X). These instances often give an idea that why various special types of set-valued
maps are introduced and studied. In this dissertation, we endeavor to appreciate various
notions such as: H-continuity of set-valued map, Lipschitz set-valued map and some
others via examples which also help us to gain some ground as far as set-valued maps
are concerned.
We have visited many “easy to see” remarks and observations via examples. If these
observations are demonstrated via examples, then we are bound to think over them and
may retain them for later use. Further, we have also included many theorems in the
project.
This dissertation is divided into 3 chapters. Chapter 1 consists of 6 sections. Section 1
gives an introduction to set-valued maps. Section 2 is based on definitions and examples
of set-valued maps. Further sections of the chapter deal with algebraic and topological
properties of set-valued maps.

ii
 iii

In Chapter 2, we discuss a useful metric known as Hausdorff metric H defined on the

collection of non-empty bounded and closed subsets of a metric space. Thereafter, we
discuss about the continuity of maps under H often refereed as H-continuity. Finally,
we prove some theorems related to this metric which also include Nadler Contraction
Principle.
The last and final chapter is based on two research papers wherein we present some
hybrid fixed point theorems besides discussing illustrative examples.
In the end, a bibliography is also given which by no means is an exhaustive one but
lists only those books and papers which have been referred to in this explanation.
Notations and Abbreviations

R the set of real numbers

C the set of complex numbers
N the set of natural numbers
+
R the set of non-negative real numbers
R2 the set of all points in the coordinate plane
inf infimum
sup supremum
max maximum
min minimum
∀ for all
∃ there exists
∅ empty set
∞ infinity
⇒ imply
⇔ logical equivalence
|x| absolute value of x
(x, y) open interval on real line
[x, y] closed interval on real line
P (X) power set of X
B(X) bounded set of X
τ topology
f :X⇒Y set-valued mapping
dom(f ) domain of set-valued map f
Im(f ) image set-valued map f
Ex(f ) exterior set-valued map f
fr(f ) frontier set-valued map f
bd(f ) boundary set-valued map f
cl(f ) closure set-valued map f
conv H(f ) convex hull set-valued map f
f◦ interior set-valued map f

iv
 v

lsc lower semi continuous

usc upper semi continuous
 Chapter 1
SET-VALUED MAP AND NADLERS PRINCIPLE

1.1 Set valued fixed point

We say that f is a set-valued map (or point to set map or multivalued function or
simply multifunction) from a set X to another set Y , denoted by f : X ⇒ Y , if for
every x ∈ X, f (x) is a subset of Y . Set-valued maps allow us to get away from the
restriction that a map is bijective when we want to solve an equation. The first natural
instance when set-valued maps occur is the inverse f −1 of a single valued map f from
a set X to another set Y which we treat as one-one. We can always define f −1 as a
set-valued map which associates with any data y the (possibly empty) set of solutions

f −1 (y) = {x ∈ X | f (x) = y}

to the equation f (x) = y.

All single-valued maps in classical analysis can be seen as set-valued maps, while many
problems in applied mathematics are set-valued in nature. For instance, problems of
stability (parametric optimization) and controllability are often best treated with set-
valued maps, while gradients of (differentiable) functions, tangents and normal of sets
(with a structure of differentiable manifold) have natural set-valued generalizations
in the non-smooth case, by means of variational analysis techniques. The inclusion
y ∈ f (x) is the heart of modern variational analysis. The First natural instance when
set-valued maps occur is the inverse of a single-valued map. Kuratowski realized the
importance of set-valued maps (also referred as multivalued maps or point-to-set maps
or multifunctions) and devoted considerable space in his famous book on topology. In
this chapter we give some definitions, examples and some results on set-valued maps.
We begin by providing examples of natural problems involving set-valued maps before
giving a rough description of the results presented in this dissertation.
Consider a real-valued function f : R ⇒ R such that

f (x) = x1/2 , ∀x ∈ R.

Here, we see that for any x in domain of f , we have two distinct images in R.
Indeed assume 4 ∈ R,
f (4) = (4)1/2 = ±2.
Here, we see that the image of 4 under f is a set {+2, −2}. So, f can be treated as a
set valued map.
1
2

Example 1. Consider amn = ma nb where m, n ∈ I and a, b ∈ (0, 1) obviosly we can

easly seen that for each m, n the Definition 1.1.1. Let X and Y be two non-empty
sets and let f : X ⇒ Y be a map such that for any x ∈ X, we have f (x) ⊆ Y , i.e., f (x)
is a set image of x under f , then f is said to be a set-valued map.

1.2. Some Examples of Set-Valued Maps

Example 1.2.1. Consider g : X ⇒ Y as depicted in the following venn diagram, there
corresponds a subset of Y, i.e., every point in X is associated to a subset of Y.
Clearly, g(a) = {1, 2, 3} ⊆ Y and g(b) = {4, 5} ⊆ Y.

Figure 1.2.1
Example 1.2.2. Consider a map from X to its power set, i.e., f : X ⇒ P (X). Then
f is a set-valued map.
Example 1.2.3. If we are able to define a function f : X ⇒ (X, τ ), where (X, τ ) is a
topological space, then f is a set-valued map.
Example 1.2.4. Consider a map f : R ⇒ R such that

{0}, x<0


f (x) = [1, 2], x=0


{3}, x>0


is a set-valued map.
Example 1.2.5. Consider a map f : X ⇒ Y which is not one-one. Then inverse image
of f, i.e., f −1 : Y ⇒ X is treated as a set-valued map.
 3

Example 1.2.6. Let f : R ⇒ R such that

f (x) = [x, x + 1], ∀x ∈ R.

Example 1.2.7. Let f : R ⇒ [0, ∞] such that

f (x) = (x, log x), ∀x ∈ R.

Then f is a set-valued map.

Example 1.2.8. Let f : [0, 1] ⇒ [0, 1] be defined by

f (x) = {y : 0 ≤ y ≤ x}.

Then f is a set-valued map.

1.3. Algebraic and Topological Properties of Set-Valued Maps

Definition 1.3.1. (i) If f1 and f2 are two set-valued maps from X to Y , then the
S
union of f1 and f2 is also a set-valued map (f1 f2 : X ⇒ Y ) defined by
[ [
(f1 f2 )(x) = f1 (x) f2 (x), ∀x ∈ X.

(ii) If f1 and f2 are two set-valued maps from X to Y , then the intersection of f1 and
T
f2 is also a set-valued map (f1 f2 : X ⇒ Y ) defined by
\ \
(f1 f2 )(x) = f1 (x) f2 (x), ∀x ∈ X.

(iii) The cartesian product of f1 and f2 is a set-valued map (f1 × f2 : X ⇒ Y × Y )

defined by
(f1 × f2 )(x) = f1 (x) × f2 (x), ∀x ∈ X.

(iv) If Y is a linear space the sum and difference of f1 and f2 can be considered as

(f1 ± f2 )(x) = f1 (x) ± f2 (x), ∀x ∈ X.

(v) If f1 is a set-valued map from X to Y and f2 is another set-valued map from Y

to Z, then the composition product of f2 by f1 is a set-valued map (f2 ◦ f1 : X ⇒ Z)
defined by
(f2 ◦ f1 )(x) = f2 (f1 (x)), ∀x ∈ X.

Definition 1.3.2. The domain of a set-valued map f : X ⇒ Y is defined as

Dom(f ) = {x ∈ X : f (x) 6= ∅}.

4

In case if f (x) = ∅, for some x ∈ X, then f is called improper otherwise proper set-valued
map. And image of f is defined as
[
Im(f ) = f (x).
x∈X

Definition 1.3.3. Suppose f : X ⇒ Y is a set-valued map then graph of f is a subset

of X × Y and is defined as

graph (f ) = {(x, y) : y ∈ f (x)}.

Clearly, any y ∈ f (x) ⇔ (x, y) ∈ graph(f ).

Definition 1.3.4. A set-valued map f : X ⇒ Y is said to be strict if

dom(f ) = X.

Notice that, here for any x ∈ X, f (x) 6= ∅.

Definition 1.3.5. Let K be any non-empty subset of X and f is a strict set-valued
map from K to Y, i.e., dom(f ) = K. Then we can extend this set-valued map f from
K to Y to a set-valued map fK from X to Y such that

f (x), x ∈ K
fK (x) =
∅, x∈/ K.

Here, fK is the extension of f. Obviously, by the definition of domain of set-valued map

dom (fK ) = K.
In the same manner, when f is set-valued map from X to Y and K ⊆ X, then we
define the retraction f |K of f.
Let f : X ⇒ Y be a set-valued map. Then we can restrict this map f to a non-empty
set K such that
f |K (x) = f (x), ∀x ∈ K.
Here, we call this map f as retraction.
Definition 1.3.6. Let f : X ⇒ Y be a set-valued map. Then the interior set-valued
map related to f : X ⇒ Y is defined as f ◦ : X ⇒ Y such that

f ◦ (x) = (f (x))◦ , ∀x ∈ X.

Example 1.3.1. Consider a set-valued map f : R ⇒ R such that

f (x) = [x, x + 1], ∀x ∈ R.

Then the interior set-valued map related to f is defined as f ◦ : R ⇒ R such that

f ◦ (x) = ([x, x + 1])◦ = (x, x + 1), ∀x ∈ R.

 5

Definition 1.3.7. Let f : X ⇒ Y be a set-valued map. Then exterior set-valued map

related to f is defined as Ex(f ) : X ⇒ Y such that
Ex(f )(x) = Ex(f (x)), ∀x ∈ X.

Example 1.3.2. Consider a set-valued map f : R2 ⇒ R2 such that

f (a, b) = {(x, y) : |x − a| + |y − b| < 1, ∀(x, y) ∈ R2 }, ∀(a, b) ∈ R2 .
Then exterior set-valued map associated with f is such that ex(f ) : R2 ⇒ R2
and ex(f )(a, b) = ex(f (a, b))
= ex{(x, y) : |x − a| + |y − b| < 1, ∀(x, y) ∈ R2 }, ∀(a, b) ∈ R2
= {(x, y) : |x − a| + |y − b| > 1, ∀(x, y) ∈ R2 }, ∀(a, b) ∈ R2 .
Example 1.3.3. Consider another set-valued map f : R ⇒ R such that
f (x) = [x, x + 1], ∀x ∈ R,
then ex(f ) is defined as ex(f ) : R ⇒ R such that
ex(f )(x) = ex(f (x))
= ex([x, x + 1])
S
= (−∞, x) (x, x + 1), ∀x ∈ R.
Definition 1.3.8. Let f : X ⇒ Y be a set-valued map. Then frontier set-valued map
related to f is defined as f r(f ) : X ⇒ Y such that
f r(f )(x) = f r(f (x)), ∀x ∈ X.

Example 1.3.4. Consider a set-valued map f : R ⇒ R such that f (x) = [x, x + 1].
Then f r(f ) is defined as f r(f ) : R ⇒ R such that
f r(f )(x) = f r(f (x))
= f r([x, x + 1])
= {x, x + 1}, ∀x ∈ R.
Definition 1.3.9. Let f : X ⇒ Y be a set-valued map. Then boundary set-valued map
related to f is defined as bd(f ) : X ⇒ Y such that
bd(f )(x) = bd(f (x)), ∀x ∈ X.

Example 1.3.5. Consider a set-valued map f : R ⇒ R such that

f (x) = [x, x + 1], ∀x ∈ R,
then boundary set-valued map is defined as bd(f ) : R ⇒ R such that
bd(f )(x) = bd(f (x))
6

= bd([x, x + 1])
= {x, x + 1}, ∀x ∈ R.
Remark 1.3.1. Observe that Range(bd(f )) ⊆ Range(f r(f )).
Definition 1.3.10. Let f : X ⇒ Y be a set-valued map. Then the closure set-valued
map is defined as cl(f ) : X ⇒ Y such that

cl(f )(x) = cl(f (x)), ∀x ∈ X.

Example 1.3.6. Let f : R ⇒ R be a set-valued map such that

f (x) = (x, x + 1), ∀x ∈ R,

then closure of set-valued map related to f is defined as cl(f ) : R ⇒ R such that

cl(f )(x) = cl(f (x))

= cl{(x, x + 1)}, ∀x ∈ R
= [x, x + 1], ∀x ∈ R.
Definition 1.3.11. Let f : X ⇒ Y be a set-valued map. Then convex-hull set-valued
map, convH(f )(X), is defined as

convH(f )(x) = convH(f (x)), ∀x ∈ X.

Example 1.3.7. Consider set-valued map f : R2 ⇒ R2 such that

f (a, b) = {(x, y) : (x − a)2 + (y − b)2 = 1, ∀x, y ∈ R2 }, ∀(a, b) ∈ R2 ,

then convex hull set-valued map convH(f ) : R2 ⇒ R2 is defined as

convH(f )(a, b) = convH(f (x, y))
= {(x, y) : (x − a)2 + (y − b)2 ≤ 1, ∀(x, y) ∈ R2 }, ∀(a, b) ∈ R2
Remark 1.3.2. A set-valued map f : X ⇒ Y is closed, open, bounded, compact,
convex-valued map if f (x), ∀x ∈ X, is so in Y.
Definition 1.3.12. A set-valued map f : X ⇒ Y is said to be quasi-surjective if
f (X) = Y.
Example 1.3.8. Consider f : R ⇒ R such that

f (x) = (−x, x), ∀x ∈ R.

S
Notice that f (R) = (−x, x) = R.
x∈R
Thus, f is a quasi-surjective set-valued map.
 7

Definition 1.3.13. A set-valued map f : X ⇒ Y is said to be hyperunivocal or semi

single-valued set-valued map if for any x, y ∈ X we have
\
f (x) f (y) 6= ∅ ⇒ f (x) = f (y).
Thus, f is hyperunivocal.
Example 1.3.9. Consider f : R+ ⇒ R such that
√ √
f (x) = {− x, x}.
Obviously, for different x and y in R+ , we have f (x)
T
f (y) = ∅ and for x = y,
T
f (x) f (y) 6= ∅ ⇒ f (x) = f (y).
Thus, f is hyperunivocal.
Definition 1.3.14. A set-valued map f : X ⇒ Y is said to be hyperinjective if for any
x, y ∈ X we have \
f (x) f (y) 6= ∅ ⇒ x = y.

Example 1.3.10. Consider a set-valued map f : R ⇒ R such that

f (x) = {x}, ∀x ∈ R.
Obviously, f is hyperinjective.

1.4. Certain Results on Set-Valued Maps

Lemma 1.4.1. Let f : X ⇒ Y be a set-valued map and {Gi }i∈V be a family of
non-empty subsets of X. Then following conditions hold:
(a) G1 ⊆ G2 ⇒ f (G1 ) ⊆ f (G2 ),
T T
(b) f ( Gi ) ⊂ f (Gi ),
i∈∧ i∈∧
S S
(c) f ( Gi ) = f (Gi ),
i∈∧ i∈∧
(d) f (X \ G1 ) ⊃ f (X) \ f (G1 ).
Proof.(a) Verification:- Consider the set-valued map f : R ⇒ R such that
f (x) = [x, x + 1].
Let G1 , G2 ⊆ R such that
G1 = {1, 2, 3} and

G2 = {1, 2, 3, 4}.
Obviously, G1 ⊆ G2 . We have to verify that f (G1 ) ⊆ f (G2 ). Now,
S
f (G1 ) = f (x)
i∈G1
S S
= f (1) f (2) f (3)
8
S S
= [1, 2] [2, 3] [3, 4]
= [1, 4]
S
f (G2 ) = f (x)
i∈G2
S S S
= f (1) f (2) f (3) f (4)
S S S
= [1, 2] [2, 3] [3, 4] [4, 5]
= [1, 5].
Notice that f (G1 ) ⊆ f (G2 ).
If G1 ⊆ G2 ⇒ f (G1 ) ⊆ f (G2 ).
Let f (x) ⊆ f (G1 )
⇒ x ∈ G1
⇒ x ∈ G2 (∵ G1 ⊆ G2 )
⇒ f (x) ⊆ f (G2 )
∴ f (G1 ) ⊆ f (G2 ).
T
(b) Let f (x) ⊆ f ( Gi )
i
T
⇒x∈ (Gi )
i
⇒ x ∈ Gi , for each i
⇒ f (x) ⊆ f (Gi )
T
⇒ f (x) ⊆ f (Gi )
i
T T
∴ f ( Gi ) ⊆ f (Gi ).
i i
(c) Verification:- Consider set-valued f : R ⇒ R such that f (x) = [exp−x , 1].
+

Further, consider a sequence of sets {Gi } in R+ such that Gi = {1, 2, 3, . . . . . . i}.

[
⇒ Gi = N.
i

Now,
S S S S S
f ( Gi ) = f (x) = f (1) f (2) f (3) . . .
i x∈N
= [exp−1 , 1] [exp−2 , 1] . . . [0, 1]
S S S

= [0, 1].
On the other hand,
S S S S
f (Gi ) = {f (1)
f (2) . . . f (i)}
i
Si
= { [exp−1 , 1] exp−2 , 1] . . . [exp−1 , 1] }
S S

Si
= [exp−i , 1] = [0, 1] (as i −→ ∞)
i
 9
S S
∴ f( Gi ) = f (Gi ).
S i i
Let f (x) ⊆ f ( Gi )
i
S
⇒x∈ Gi
i
⇒ x ∈ Gi , for some i
⇒ f (x) ⊆ f (Gi )
S
⇒ f (x) ⊆ f (Gi )
i
[ [
⇒ f ( Gi ) ⊆ f (Gi ). (1.4.1)
i i
S
Again consider f (x) ⊆ f (Gi )
i
⇒ f (x) ⊆ f (Gi ), for some i
⇒ x ∈ Gi
S
⇒ x ∈ Gi
i
S
⇒ f (x) ⊆ f ( Gi )
i
[ [
⇒ f (Gi ) ⊆ f ( Gi ). (1.4.2)
i i

From Equations 1.4.1 and 1.4.2,

[ [
f (Gi ) = f ( Gi ).
i i

(d) Let f (x) ⊆ f (X) \ f (G1 )

⇒ f (x) ⊆ f (X) and f (x) * f (G1 )

⇒ x ∈ X and x ∈
/ G1
⇒ x ∈ X \ G1
⇒ f (x) ⊆ f (X \ G1 )
⇒ f (X) \ f (G1 ) ⊆ f (X \ G1 ).
T
Theorem 1.4.1. If f, g : X ⇒ Y be two hyperunivocal set-valued maps, then f g is
also hyperunivocal.
Proof. Since f and g are hyperunivocal maps, therefore for any x, y ∈ X, we have
\
f (x) f (y) 6= ∅ ⇒ f (x) = f (y). (1.4.3)

Similarly,
\
g(x) g(y) 6= ∅ ⇒ g(x) = g(y). (1.4.4)
10

Now, consider
\ \ \
(f g)(x) (f g)(y) 6= ∅
\ \ \
⇒ (f (x) g(x)) (f (y) g(y)) 6= ∅
\ \ \
⇒ (f (x) f (y)) (g(x) g(y) 6= ∅
T T
⇒ f (x) f (y) 6= ∅ and g(x) g(y)) 6= ∅.
From Equations 1.4.3 and 1.4.4,

f (x) = f (y) and g(x) = g(y)

T T
⇒ f (x) g(x) = f (y) g(y)
T T
⇒ (f g)(x) = (f g)(y)
T
∴ f g is hyperunivocal.

1.5. Continuity of Set-Valued Maps

There are two distinct ways to extend the concept of continuity to the set-valued map.
For defining the continuity of set-valued maps, we first introduce the concepts of lower
inverse and upper inverse of set-valued maps. The concept of two kinds of semicontinuity
of a set-valued map was introduced by G. Bouligand and K. Kuratowski.
Definition 1.5.1. Let f : X ⇒ Y be a set-valued map. Then lower inverse image of
any set V of Y under f is denoted as f − and defined as
\
f − (V ) = {x ∈ X : f (x) V 6= ∅}.

Definition 1.5.2. Let f : X ⇒ Y be a set-valued map then upper inverse image of

any set V of Y under f is denoted as f + and defined as

f + (V ) = {x ∈ X : f (x) ⊂ V }.

Definition 1.5.3. Consider a set-valued map f : X ⇒ Y. Then f is said to be upper

semi continuous at x◦ ∈ X if and only if for every open set V in Y such that f (x◦ ) ⊂ V, ∃
a neighborhood U of x◦ in X such that

f (x) ⊂ V, ∀x ∈ U.

Example 1.5.1. Consider a set-valued map f : R ⇒ R such that


[−1, 1], x = 1
f (x) =
{1}, x=6 1.

Then f is upper semi continuous at 1.

 11

Figure 1.5.1

Consider any open set V in R such that f (1) ⊂ V. Then we have neighborhood U of 1
in R such that f (U ) = {1}. In either case f (U ) ⊂ [−1, 1].
Obviously, f (U ) ⊂ V. (∵ f (1) ⊂ V, i.e., [−1, 1] ⊂ V ).
Thus, f is upper semi continuous at 1.
Example 1.5.2. Consider a set-valued map f : R ⇒ R such that


{0}, x=1
f (x) =
[−1, 1], x 6= 1.

Then, f is not upper semi continuous at 1.

Here, for any open set V in R such that f (1) ⊂ V, there does not exist a neighborhood
U of 1 such that f (x) ⊂ V, ∀x ∈ U. Thus, f is not upper semi continuous at 1.
Definition 1.5.4. Consider a set-valued map f : X ⇒ Y then f is said to be lower semi
T
continuous at x◦ ∈ X if and only if for any open set V in Y such that f (x◦ ) V 6= ∅, ∃
a neighborhood U of x◦ in X such that

\
f (x) V 6= ∅, ∀x ∈ U.

Example 1.5.3. Consider a set-valued map f : R ⇒ R such that


{0}, x=0
f (x) =
[−1, 1], x 6= 0.

Then f is lower semi continuous at zero.

12

Figure 1.5.2

Definition 1.5.5. A set-valued map f : X ⇒ Y is said to be continuous if it is upper

semi continuous as well as lower semi continuous.

1.6. Relevant Weak Continuity Conditions

(i) Let X, Y and Z be metric spaces. If f1 : X ⇒ Y and f2 : Y ⇒ Z are lsc [usc]
set-valued maps then the composition product f = f2 ◦ f1 : X ⇒ Z is also lsc [usc]
set-valued map.
S
(ii) Let X and Y be metric spaces. The arbitrary union f = fα of a family of lsc
α∈∧
Sn
set-valued maps fα : X ⇒ Y is also lsc. But only finite union f = fi of usc set-valued
i=1
maps fi : X ⇒ Y is usc.
T
(iii) Let X and Y be metric spaces. The intersection f = fα of a family of usc
α∈∧
set-valued maps fα : X ⇒ Y is usc.
(iv) Let X and Yi be metric spaces. Let fi : X ⇒ Yi be finite number of lsc [usc]
n
Q
set-valued maps. Then the cartesian product f = fi : X ⇒ Y is lsc [usc] set-valued
i=1
n
Q
map, where Y = Yi .
i=1

Theorem 1.6.1. Let X and Y be metric spaces. A set-valued map f : X ⇒ Y is lower

semi continuous if and only if f + (G) is open for each open subset G of Y.
Proof. Let f : X ⇒ Y be a set-valued map.
Suppose f is lower semi continuous and G is an open subset of Y.
If f + (G) = ∅, then it is open.
If f + (G) 6= ∅, let x1 ∈ f + (G)
⇒ f (x1 ) ⊆ G
T
⇒ f (x1 ) G 6= ∅.
Let y1 ∈ f (x1 ) ∩ G
⇒ y1 ∈ G.
∵ G is open.
⇒ G is the neighborhood of each of its points.
⇒ G is the neighborhood of y.
Thus, ∃ a δ > 0 such that ∀x ∈ BX (x1 , δ),
\
f (x) G 6= 0
 13

⇒ x ∈ f + (G).
∴ BX (x1 , δ) ⊆ f + (G)
⇒ f + (G) is open.
Conversely, suppose f + (G) is open for each open subset G of Y such that
\
f (x1 ) G 6= ∅ ⇒ x1 ∈ f + (G)
⇒ f + (G) is a neighborhood of x1 . (∵ f + (G) is open)
∴ ∀x1 ∈ f + (G), ∃ a δ > 0 such that
BX (x1 , δ) ⊆ f + (G).
Now, ∀x ∈ BX (x1 , δ), x ∈ f + (G).
T
⇒ f (x) G 6= ∅
⇒ f is lower semi continuous.
Theorem 1.6.2. Let X and Y be metric spaces and f : X ⇒ Y be a set-valued map
such that f (x) is compact. Then for each x ∈ X, f is upper semi continuous if and only
if for each open subset G of Y the set
f − (G) = {x ∈ X : f (x) ⊆ G}
is open.
Proof. Let f : X ⇒ Y be a set-valued map.
Suppose f is upper semi continuous.
Given : f (x) is compact, ∀x ∈ X.
For each open subset G of Y f − (G) is given by
f − (G) = {x ∈ X : f (x) ⊆ G}.
Suppose f is upper semi continuous.
If f − (G) = ∅, then it is open.
If f − (G) 6= ∅, then let x◦ ∈ f + (G)
⇒ f (x◦ ) ⊆ G.
G is the neighborhood of f (x◦ ) (∵ G is open).
∵ f is upper semi continuous.
∴ ∃ a δ > 0 such that ∀x ∈ BX (x◦ , δ), f (x) ⊆ G
⇒ x ∈ f − (G)
⇒ BX (x◦ , δ) ⊆ f − (G).
∴ f − (G) is open.
Conversely, suppose for each open subset G of Y, f − (G) is open.
Given : f (x) is compact, ∀x ∈ X.
14

Let x◦ ∈ X and G be an open set containing f (x◦ ).

Since f (x◦ ) is compact, G is a neighborhood of f (x◦ ).
∴ f − (G) is a neighborhood of x◦ and we have
∀x◦ ∈ f + (G), ∃ δ > 0 such that BX (x◦ , δ) ⊆ f − (G).
∴ ∀x ∈ BX (x◦ , δ), x ∈ f − (G).
⇒ f (x) ⊆ G, ∀x ∈ BX (x◦ , δ).
∴ f is upper semi continuous.
Theorem 1.6.3. Let X and Y be metric spaces and f : X ⇒ Y be a set-valued map
which is upper semi continuous such that for all x ∈ X, f (x) is compact. Then the
image f (K) of a compact subset K of X is compact.
Proof. Let {Gα }α∈∧ be an open cover of f (K).
Since for every x ∈ K, f (x) is compact, f (x) can be covered by a finite number of G0α s.
Let Gx be the union of the sets in such a finite family. Then {f − (G) : x ∈ K} is an
open covering of K.
∴ It contains a finite covering f − (Gx1 ), f − (Gx2 ), . . . , f − (Gxn ).
The sets Gx1 , Gx2 , . . . , Gxn cover f (K).
∴ f (K) can be covered by a finite number of G0α s.
Definition 1.6.1. A set-valued map f : X ⇒ Y is closed if whenever x◦ ∈ X,
y◦ ∈ Y , y◦ ∈
/ f (x◦ ) there exist two neighborhoods U (x◦ ) and V (y◦ ) such that x ∈ U (x◦ )
⇒ f (x) ∩ V (y◦ ) = ∅.
Example 1.6.1. Let f : R ⇒ R be a set-valued map such that
f (x) = {x}.
Hence, f is a closed set-valued map.
Theorem 1.6.4.. Let X and Y be metric spaces and f : X ⇒ Y be a closed set-valued
map. If K is a compact subset of X, then the set f (K) is closed.
Proof. Let yn ∈ f (K), for all n = 1, 2, 3, . . . and yn −→ y.
Suppose xn ∈ f + (yn ) K.
T

Without loss of generality, we may assume that {xn } converges to some point x.
Thus, y ∈ f (x) and therefore, y ∈ f (K).
 CHAPTER 2
CERTAIN RESULTS INVOLING HAUSDORFF METRIC

2.1. Introduction
The key to the classical Banach fixed point theorem is that one is working in a
complete metric space. We realize that the metric which is defined on X may not define
metric on P (X).
If we put it on P (X) such that

d(M, N ) = sup {d(x, N ) : x ∈ M }, ∀ M, N ∈ P (X), (2.1.1)

x∈M

then d does not form a metric on P (X) but if we restrict our idea about the set chosen
M, N, i.e., if we choose M, N in P (X) such that the two sets M, N both as non-empty,
closed, bounded and d is symmetric in P (X), i.e., d(M, N ) = d(M, N ) then d define a
metric on P (X). In this chapter we introduce a new metric defined on a set known as
the Hausdorff metric and then we discuss its uses in set-valued maps such as Lipschitz,
contraction and non-expansive set-valued map. We denote the collection of subsets of
metric space X in the following manner :
2X = set of all subsets of X
2X
cl = set of all non-empty closed and bounded subsets of X
2X
q = set of all non-empty compact subsets of X.

Obviously, 2X X X
q ⊆ 2cl ⊆ 2 .

2.2. Hausdorff Metric

Definition 2.2.1. Let (X, d) be a metric space. Then for M, N ∈ 2X , we define the
number d(M, N ) as follows :

d(x, N ) = inf {d(x, y) : y ∈ N }

y∈N

d(M, N ) = sup {d(x, N ) : x ∈ M }

x∈M

= sup inf d(x, y).

x∈M y∈N

Remark 2.2.1.(A) inf ∅ = ∞.

(B) d(M, N ) being small thus means that each point of M is close to some point of N.
The value of d(M, N ) will certainly be finite, if M and N are bounded.
Remark 2.2.2. The above definition of d does not define a metric because in respect
to the following fact:
16
 17

(A) Consider two sets (−∞, 0), [0, ∞). Then the distance between these two sets under
d is defined as:

d((−∞, 0), [0, ∞)) = sup{d(x, [0, ∞)) : x ∈ (−∞, 0)} = ∞.

This is not possible in the definition of metric. Thus, we conclude that the two sets
which we have taken earlier should be bounded.
(B) d([0, ∞), {0}) = sup{d(x, {0}) : x ∈ [0, ∞)} = ∞.
This is not possible in the definition of metric. Thus, we conclude that the two sets
which we have taken earlier should be bounded.
Also, d(∅, {0}) = ∞. Thus, both the sets must be non-empty.
Remark 2.2.3. In general, d(M, N ) 6= d(N, M ).
Consider N = {0}, M = (0, 1],
then d(M, N ) = sup {d(x, {0}) : ∀x ∈ (0, 1]}
x∈M
= 1,
but d(N, M ) = sup{d({0}, (0, 1])}
= 0.
Thus, in general, d(M, N ) = d(N, M ) need not hold.

Theorem 2.2.1. Let L, M, N ∈ 2X

cl . Then following hold :

(1) d(M, N ) = 0 ⇔ M ⊆ N.
(2) d(M, N ) ≤ d(M, L) + d(L, N ).
Proof :(1) Consider d(M, N ) = 0
⇔ sup {d(x, N ) : x ∈ M } = 0 (by Definition 2.1.1)
x∈M
⇔ d(x, N ) = 0, ∀x ∈ M
⇔ x ∈ N, ∀x ∈ M (∵ d(x, A) = 0 ⇔ x ∈ A)
⇔ M ⊆ N.
(2) By the triangular inequality in the metric space (X, d) we can write
d(m, n) ≤ d(m, l) + d(l, n), ∀m ∈ M, n ∈ N, l ∈ L
⇒ inf d(m, n) ≤ inf {d(m, l) + d(l, n)}
n∈N n∈N
⇒ inf d(m, n) ≤ d(m, l) + inf d(l, n).
n∈N n∈N
This inequality holds for all l ∈ L.

This implies
inf d(m, n) = d(m, N ) ≤ d(m, l) + d(l, N ).
x∈N

On taking sup we obtain,

l∈L
18

d(m, N ) ≤ d(m, L) + d(L, N ).

On taking sup we obtain,
m∈M
d(M, N ) ≤ d(M, L) + d(L, N ).
Theorem 2.2.2. Let (X, d) be a metric space. Let A, B, C ∈ 2X
cl . Then we have the
following:
(1) B ⊂ C ⇒ d(A, C) ≤ d(A, B).
S
(2) d(A B) = max{d(A, C), d(B, C)}.
Proof.(1) B ⊂ C ⇒ d(x, C) ≤ d(x, B), ∀x ∈ A.
Taking supremum over M ,
sup d(x, C) ≤ sup d(x, B)
x∈A x∈A
d(A, C) ≤ d(A, B)
S S
(2) d(A B) = max{d}(x, y), x, y ∈ A B}
= max{d(A, C), d(B, C)}
Case 1 : Let d(A, C) ≤ d(B, C)

max(d(A, C), d(B, C)) = d(B, C)

= sup{d(x, C), ∀x ∈ B}
S
= sup{d(x, C), ∀x ∈ A B}
S
= d(A B)
Case 2 : Let d(B, C) ≤ d(A, C)

max(d(A, C), d(B, C)) = d(A, C)

= sup{d(y, C), ∀y ∈ A}
S
= sup{d(x, C), ∀y ∈ A B}
S
= d(A B)
In all, we see that d does not form a metric over a set. Thus, we have the following
characterization known as the Hausdorff metric.
Definition 2.2.2. From the above examination we define a new metric H called the
Hausdorff Metric on the family of all non-empty closed and bounded subsets of metric
space (X, d) as
H(M, N ) = max{d(M, N ), d(N, M )}.

Theorem 2.2.3. Show that H define a metric where

H(M, N ) = max{d(M, N ), d(N, M )}, ∀M, N ∈ 2X

cl .
 19

Proof. (i) H(M, N ) > 0, it is obvious by the definition.

(ii) H(M, N ) = 0 ⇔ M = N.
Let M = N = K (say). Then

H(K, K) = max{d(K, K), d(K, K)}

= d(K, K)
= 0.
Conversely, suppose H(M, N ) = 0
⇒ sup d(x, N ) = 0
m∈M
⇒ d(x, N ) = 0, ∀x ∈ M
⇒x∈N =N (∵ N, M ∈ 2X
cl )

⇒ M ⊆ N. (2.2.1)

Similarly,
N ⊆ M. (2.2.2)
From Equations 2.2.1 and 2.2.2,
M = N.
Thus, the result holds.
(iii) H(M, N ) = H(N, M ) (obvious from the definition).
(iv) Now we have to prove the triangular inequality,

H(M, N ) ≤ H(M, Q) + H(Q, N ), ∀M, N, Q ∈ 2X

cl .

By Theorem 2.2.1. we have

d(M, N ) ≤ d(M, Q) + d(Q, N ).

Now,
H(M, N ) = max{d(M, N ), d(N, M )}
≤ max{d(M, Q) + d(Q, N ), d(N, Q) + d(Q, M )}
≤ max{d(M, Q), d(Q, M )}+max{d(Q, N ), d(N, Q)}
= H(M, Q) + H(Q, N ).
Theorem 2.2.4. Let (X, d) be a metric space, M, N ∈ 2X q and r > 0 a given number.
Then M ⊆ Sr (N ) and N ⊆ Sr (M ) if and only if H(M, N ) ≤ r.
Proof. First, we show that M ⊆ Sr (N ) if and only if d(M, N ) ≤ r.
Let d(M, N ) ≤ r
⇒ sup {d(x, N )} ≤ r
x∈M
20

⇒ d(x, N )} ≤ r, ∀x ∈ M.
Since N is compact we have y ∈ N such that
d(x, y) ≤ r, y∈N
⇒ x ∈ Sr (y), y∈N
⇒ x ∈ Sr (N )
⇒ M ⊆ Sr (N ).
Conversely, suppose that M ⊆ Sr (N ) then for all x ∈ M, ∃y ∈ N such that d(x, y) ≤ r

⇒ ∀x ∈ M, d(x, N ) ≤ d(x, y) ≤ r

taking sup over M we have

sup d(x, N ) ≤ r
x∈M
⇒ d(M, N ) ≤ r.
To show second one replace the role of M and N in the above proof.
Now, H(M, N ) = max{d(M, N ), d(N, M )} ≤ r if and only if d(M, N ) ≤ r.
Hence, H(M, N ) ≤ r if and only if M ⊆ Sr (N ) and N ⊆ Sr (M ).
Theorem 2.2.5. If (X, d) is a complete metric space, then (2X
q , H) is also a complete
metric space.

2.3. H-Continuity of Set-Valued Map

Now, we define the H-continuity of a set-valued map:
Definition 2.3.1. Let (X, d1 ) and (Y, d2 ) be two metric spaces. Then a set-valued map
f : X ⇒ 2X
q is said to be H-continuous if for any ε > 0, ∃ δ > 0 such that

H(f (x), f (y)) < ε whenever d1 (x, y) < δ, ∀x, y ∈ X.

Remark 2.3.1. If a set-valued map f is only upper semi continuous or lower semi
continuous, then we cannot say anything about the H-continuity of f.
Example 2.3.1. Consider X = [1, 2] be a metric space with usual metric and consider
a set-valued map f : X ⇒ X such that

{2}, x < 3/2


f (x) = [1, 2], x = 3/2


{1}, x > 3/2.


Then f is obviously upper semi continuous but f is not H-continuous,

for y 6= 0, we have

H(f (3/2), f (y)) = max{d(f (3/2), f (y)), d(f (y), f (3/2))}

 21

= 1.
This shows that f is not H-continuous.
Example 2.3.2. Consider a metric space X = [1, 2] with usual metric and f : X ⇒ 2X
q
such that 
[1, 2], x 6= 1
f (x) =
{1}, x = 1.
In this example, f is lower semi continuous but not H-continuous as

∀x ∈ X, H(f (1), f (x)) = 1.

This shows that f is not H-continuous.

Example 2.3.3. Let X = R be metric space with usual metric d1 and Y = R2 be
another metric space with the metric d2 defined by
d1 (x, y)
d2 (x, y) = ,
1 + d1 (x, y)
where d1 is usual metric on R2 .
2
Further, consider a set-valued map f : R ⇒ 2Rcl defined as

∀s ∈ R, f (s) = {(s, y) : y ∈ R}.

Then
H(f (s), f (t)) ≤ 2|s − t|, ∀s, t ∈ R,
then if we take |s − t| < δ, then

H(f (s), f (t)) < , ∀s, t ∈ R where  = 2δ.

This shows that f is H-continuous.

Further, consider an open set U in R2 such that

U = {(x, y) : x|y| < 1 or x = 0}.

But f −1 (U ) = {0} which is not open in R, i.e., inverse image of open set is not open
and so f is not upper semi continuous.

In the mutual relation of these continuities such as lower semi continuity, upper semi
continuity and H-continuity we have the following theorem.
Theorem 2.3.1. Let X and Y be metric spaces. A mapping f : X ⇒ 2X q is H-
continuous if and only if the set-valued map f : X ⇒ Y is compact-valued upper semi
continuous as well as lower semi continuous.
22

Proof. Let f be H-continuous and U be an open subset of Y.

We first show that f is upper semi continuous by proving that the set
f − (U ) = {x ∈ X : f (x) ⊆ G}
is open in X.
Let x◦ ∈ f − (U )
⇒ f (x◦ ) ⊆ U.
Since f (x◦ ) is compact, ∃ ε > 0 such that Sε (f (x◦ )) ⊆ U.
Since f is H-continuous, we have a δ > 0 such that
H(f (x◦ ), f (x)) < ε
⇒ f (x) ⊆ Sε (f (x◦ )) ⊆ U
Sδ (x◦ ) ⊆ f − (U )
⇒ f − (U ) is open in X.
This shows that f is upper semi continuous.
Now, we show that f is lower semi continuous. For this we have to show that
\
f − (U ) = {x ∈ X : f (x) U 6= ∅}
is open in X.
Let x◦ ∈ f + (U ) \
⇒ f (x◦ ) U 6= ∅
\
⇒ ∃ y◦ ∈ f (x◦ ) U
⇒ y◦ ∈ f (x◦ ) and y◦ ∈ U
and since U is open and y◦ ∈ U, ∃ ε > 0 such that Sε (y◦ ) ⊆ G.
Since f is H-continuous, we can find δ > 0 such that
H(f (x◦ ), f (x)) < ε/2, ∀x ∈ Sδ (x◦ ) (2.3.1)
T
Claim : f (x) Sε (y◦ ) 6= ∅.
T
Suppose f (x) Sε (y◦ ) = ∅.
Also from Equation 2.3.1 we have
f (x◦ ) ⊂ Sε/2 (f (x))
⇒ y◦ ∈ Sε/2 (f (x))
⇒ ∃ z◦ ∈ f (x) such that d(z◦ , y◦ ) < ε/2
⇒ z◦ ∈ Sε (y◦ ) which is a contradiction.
T
⇒ f (x) Sε (y◦ ) 6= ∅
⇒ f is lower semi continuous.
Conversely, assume f is upper semi continuous as well as lower semi continuous.
Let U = Sε (f (x◦ )), ε > 0 and x◦ ∈ X.
 23

By upper and lower semi continuity the sets f + (U ) and f − (U ) both are open and
x◦ ∈ f + (U ) f − (U ).
T

Let V = f + (U ) f − (U ).
T

Then obviously V is an open neighborhood of x◦ such that

Sδ (x◦ ) ⊂ V and f (x◦ ) ⊂ Sε (f (x)), ∀x ∈ Sδ (x◦ ).

To find such δ we cover the compact set by n open spheres Sε (yi ), (1 ≤ i ≤ n).
n
S
⇒ f (x◦ ) = Sε (yi ) ⊂ Sε/2 (f (x◦ )).
i=1
Since f is lower semi continuous there is an open sphere Sδi (x◦ ⊂ V such that
\
f (x) Sε/2 (yi ) 6= ∅, ∀x ∈ Sδi (x◦ ).

Let δ = min{δ1 , δ1 , . . . , δn }.
Then Sδ (x◦ ) ⊂ V
and for any y ∈ f (x◦ ) we have y ∈ Sε/2 (yi ), for some i.
Furthermore, we know that for x ∈ Sδ (x◦ ) we have
\
f (x) Sε/2 (yi ) 6= ∅, ∀ i = 1, 2, . . . , n.

Thus, for every x ∈ Sδ (x◦ ) any y ∈ f (x◦ ), i = 1, 2, . . . , n such that

ρ(y, f (x)) ≤ d(y, yi ) + ρ(yi , f (xi ))

< ε/2 + ε/2

=ε
⇒ ∀x ∈ Sδ (x◦ ), we have
ρ(x◦ ) ⊂ Sε (f (x◦ )).
Example 2.3.4. Consider a set-valued map f : R ⇒ R such that

f (x) = [−1, 1], ∀x ∈ R.

2.4. Lipschitz Set-Valued Map

Definition 2.4.1. Let (X, d1 ) and (Y, d2 ) be metric spaces and f : X ⇒ 2Ycl be a set-
valued map. Then f is said to satisfy Lipschitz condition or f is said to be a set-valued
Lipschitz Map if ∃ a constant k > 0 such that

H(f (x), f (y)) ≤ kd1 (x, y), ∀x, y ∈ X.

The constant k is called Lipschitz Constant for set-valued map f.

Definition 2.4.2. If we take Lipschitz constant k ∈ (0, 1) in Definition 2.4.1, i.e.,

H(f (x), f (y)) ≤ kd1 (x, y), ∀x, y ∈ X and k ∈ (0, 1),
24

then f is said to be Set-Valued Contraction Map.

Definition 2.4.3. If we take Lipschitz constant k = 1 in Definition 2.4.1, i.e.,

H(f (x), f (y)) ≤ d1 (x, y), ∀x, y ∈ X,

then f is said to be Set-Valued Nonexpansive Map.

Remark 2.4.1. Union of two set-valued Lipschitz maps is again a set-valued Lipschitz
map, i.e., if f : X ⇒ 2Ycl is a set-valued Lipschitz map with Lipschitz constant k1
and g : X ⇒ 2Ycl is another set-valued Lipschitz map with Lipschitz constant k2 then
f g : X ⇒ 2Ycl such that
S
[ [
(f g)(x) = f (x) g(x), ∀x ∈ X

is a set-valued Lipschitz map with Lipschitz constant k = max{k1 , k2 }.

Remark 2.4.2. Intersection of two set-valued contraction maps need not be a set-
valued contraction map.
Example 2.4.1. Consider A = {(x, y) : 1 ≤ x ≤ 2, 1 ≤ y ≤ 2} and let f : A × A ⇒
2A×A
cl such that
f (x, y) = line segment in A×A joining the points (x/2, 1) and (x/2, 2), ∀(x, y) ∈ A×A
and let g : A × A ⇒ 2A×A
cl such that
g(x, y) = line segment in A×A joining the points (x/2, 1) and (x/3, 2), ∀(x, y) ∈ A×A,
then obviously, f and g are set-valued contraction maps.
Now, 
\ (x/2, 0), x 6= 0
(f g)(x) =
{(x, y) ∈ A × A}, x = 0

is not a set-valued contraction map.

Definition 2.4.4. Let f : X ⇒ X be a set-valued map. A point x ∈ X is said to be
fixed point of f if x ∈ f (x).
Example 2.4.2. Consider a set-valued map f : R ⇒ R such that

f (x) = [−x, x], ∀x ∈ R,

then ∀a ∈ R, we have a ∈ f (a).

Thus, every real number is a fixed point of f.
Example 2.4.3. Consider f : R ⇒ R such that

f (x) = (−x, x), ∀x ∈ R.

Obviously, for any a ∈ R, we can see that a ∈

/ f (a).
Thus, f does not have any fixed point.
 25

Example 2.4.4. Let f : R ⇒ R such that


{0}, x=0


f (x) = (−∞, 0), x > 0


(0, ∞), x < 0,


then we can see that 0 ∈ R is only point such that 0 ∈ f (0) = {0}. Thus, f has only
zero as fixed point.
Example 2.4.5. Let f : R ⇒ R such that

f (x) = {log x},

then there does not exist a ∈ f (a) = {log a}.

Thus, f does not have any fixed point.

2.5. Nadler Contraction Principle

In this section we extend some fixed point theorems from single-valued map to set-
valued map. We begin this process from the core fixed point theorem known as Banach
Fixed Point Theorem. First of all, we give the following definition.
Definition 2.5.1. If f : X −→ X is a single-valued map then x ∈ X is a fixed point
of f if x = f (x).

There is a famous fixed point theorem for single-valued maps known as Banach-Contraction
Theorem which has been extended for multivalued maps.
Theorem 2.5.1. (Banach-Contraction Theorem) Let f : X −→ X be a contrac-
tion single-valued map from a complete metric space (X, d) into itself. Then f has a
unique fixed point.

We can extend the existence part of Banach-Contraction Theorem for single-valued

map to set-valued contraction mappings. In 1969 Nadler established the following ana-
logue of Banach-Contraction Theorem for set-valued maps. Generalization of Banach
Contraction Theorem in Set-Valued Map.
Theorem 2.5.2. (Nadler’s Theorem) Let (X, d) be a complete metric space. If
f : X ⇒ 2X
cl is a set-valued contraction map then f has a fixed point.

Proof. Let α be a Lipschitz constant for f and let x◦ ∈ X.

Choose x1 ∈ f (x◦ ).
Since f (x◦ ), f (x1 ) ∈ 2X
cl and x1 ∈ f (x◦ ), there is a point x2 ∈ f (x1 ) such that

d(x1 , x2 ) ≤ H(f (x◦ ), f (x1 )) + α.

26

Since f (x1 ), f (x2 ) ∈ 2X

cl and x2 ∈ f (x1 ), there is a point x3 ∈ f (x2 ) such that

d(x2 , x3 ) ≤ H(f (x1 ), f (x2 )) + α2 .

Continuing this way, we have a sequence {xn }∞
n=1 of points of X such that

xn+1 ∈ f (xn )
and
d(xn , xn+1 ) ≤ H(f (xn−1 ), f (xn )) + αn , ∀n ≥ 1.
Note that
d(xn , xn+1 ) ≤ H(f (xn−1 ), f (xn )) + αn
≤ αd(xn−1 , xn ) + αn
≤ α[H(f (xn−2 ), f (xn−1 )) + αn−1 ] + αn
≤ α2 d ( x n−2 , x n−1 ) + 2αn
..
.
≤ αn d(x◦ , x1 ) + nαn , ∀ n ≥ 1.
This implies that for any n and m,
d(xn , xn+m ) ≤ d(xn , xn+1 ) + d(xn+1 , xn+2 ) + . . . + d(xn+m−1 , xn+m )
≤ αn d(x◦ , x1 ) + nαn + αn+1 d(x◦ , x1 ) + (n + 1)αn+1 +
. . . + αn+m−1 d(x◦ , x1 ) + (n + m − 1)αn+m−1
n+m−1 n+m−1
αk d(x◦ , x1 ) + kαk .
P P
=
k=n k=n

It follows that the sequence {xn } is a Cauchy sequence.

Since (X, d) is complete, the sequence {xn } converges to some point x ∈ X.
Since f is H-continuous,
lim H(f (xn ), f (x)) = 0.
n→∞
Since xn ∈ f (xn−1 ), we have
lim ρ(xn , f (x)) = lim inf{d(xn , y) : y ∈ f (x)} = 0.
n→∞ n→∞

This implies that

ρ(x, f (x)) = inf{d(x, y) : y ∈ f (x)} = 0,
and since f (x) is closed, x ∈ f (x).

The following theorem is established by Dancs , Hegedũs and Medvegyev. We call it

DHM Theorem.

Theorem 2.5.3. (DHM Theorem) Let (X, d) be a complete metric space and f :
X ⇒ X be a set-valued map such that the following conditions hold:
 27

(1) For all x ∈ X, f (x) is closed.

(2) For all x ∈ X, x ∈ f (x).
(3) For all x, y ∈ X, y ∈ f (x) implies f (y) ⊆ f (x).
(4) For any sequence {xn } in X such that
x2 ∈ f (x1 ), x3 ∈ f (x2 ), . . . , xn ∈ f (xn−1 ), . . . ,
the distance d(xn , xn+1 ) tends to zero as n −→ +∞.
Then the set-valued map f has a fixed point x ∈ X in the sense that f (x) = {x}.
Proof. Since d satisfies condition (4), the equivalent metric d0 = (1+d)d
also satisfies
condition (4). So, we suppose that d is bounded on X.
Let δ(A) be the diameter of a subset A of X.
By (2), f (x) 6= ∅, ∀x ∈ X and we construct a generalized Picard-iteration such that
x1 = x
e,
xn ∈ f (xn−1 )
and
δ(f (xn−1 )) 1
d(xn , xn−1 ) ≥ − n−1 .
2 2

Hence, by (3) and (4), we have

f (xn+1 ) ⊇ f (xn ) and the diameter of f (xn ), δ(f (xn )) −→ 0 as n −→ +∞.
Since X is complete, {f (xn )} is a decreasing sequence of non-empty closed sets such
that
δ(f (xn )) −→ 0 as n −→ +∞,
by Cantor’s intersection theorem
∞
\
f (xn ) = {e
x}.
n=1

The point x e ∈ f (xn ) and condition (3) imply

e is a fixed point, since on the other hand x
∞
T
x) ⊆
f (e f (xn ) = {ex}.
n=1
{e
x} ⊆ f (e
x) (by(2)).
Therefore,
f (e
x) = x
e.
 CHAPTER 3
HYBRID FIXED POINT THEOREMS

3.1. Introduction
In earlier chapters we study fixed point for the set-valued map. In this chapter, we
collect a very interesting concept on the theory of fixed point known as hybrid pair of
mapping in which we hybridized one set-valued map with a single-valued map. In
this chapter, we study fixed point theorem for hybrid pair if mappings, i.e., one set-
valued and one single-valued mapping. The concept of hybrid pair of mapping is very
consequential for the theory of fixed point and it has an important role in game theory,
differential equation and optimization. We divide this chapter into three sections. In
Section 1 we collect some basic definitions and examples.
Let (X, d) be a metric space and B(X) be the collection of all bounded subsets of X.
Recall that
δ(A, B) = sup{d(a, b) : a ∈ A, b ∈ B}.

Remark 3.1.1. When A and B both are singleton, i.e., A = {a} and B = {b}, then

d(a, b) = δ(a, b).

Theorem 3.1.1. For any A, B, C ∈ B(X), we have

(i) δ(A, B) = δ(B, A) ≥ 0.
(ii) δ(A, B) ≤ δ(A, C) + δ(C, B).
(iii) δ(A, A) = sup{d(a, b) : a, b ∈ A} = diameter of A.
(iv) δ(A, B) = 0 ⇒ A = B{a}.
Proof.(i) δ(A, B) = sup{d(a, b) : a ∈ A, b ∈ B}
= sup{d(b, a) : b ∈ B, a ∈ A}
= δ(B, A) ≥ 0.
(ii) δ(A, B) = sup{d(a, b) : a ∈ A, b ∈ B}
≤ sup{d(a, c) + d(c, b), a ∈ A, b ∈ B, c ∈ C}
= sup{d(a, c), a ∈ A, c ∈ C} + sup{d(c, b), c ∈ C, b ∈ B}
= δ(A, C) + δ(C, B)
⇒ δ(A, B) ≤ δ(A, C) + δ(C, B).
(iii) δ(A, A) = sup{d(a, b) : a, b ∈ A}
The right hand side of the above equation denotes the diameter of the bounded set A.
28
 29

(iv) Suppose δ(A, B) = 0

⇒ sup{d(a, b) : a ∈ A, b ∈ B} = 0

⇒ d(a, b) = 0, ∀a ∈ A, b ∈ B
⇒ a = b, ∀a ∈ A, b ∈ B
⇒ A = B = {a}.
Definition 3.1.1. If {An } is a sequence in B(X), we say that {An } converges to
A ⊆ X, and write An −→ A if and only if for
(i) a ∈ A implies that an −→ a for some sequence an with an ∈ An , for some n ∈ N.
(ii) For any ε > 0, ∃ m ∈ N such that

An ⊆ Aε = {x ∈ X : d(x, a) < ε for some a ∈ A} for n > m.

Definition 3.1.2. Let f, g : X −→ X be single-valued maps. Then the pair (f, g) is

said to be weakly commuting if for all x ∈ X, we have

d(f gx, gf x) ≤ d(f x, gx).

Now, we define different types of commutativity in hybrid mapping as follows:

Definition 3.1.3. Let f : X ⇒ B(X) be a set-valued map and g : X −→ X be a
single-valued map, then we say that the pair (f, g) is
(i) commuting on X if gf x = f gx.
(ii) weakly commuting on X if δ(f gx, gf x) ≤ max{δ(gx, f x), diam(gf x)}, for any
x ∈ X.
(iii) quasi-commuting on X if gf x ⊆ f gx.
(iv) slightly commuting on X if δ(f gx, gf x) ≤ max{δ(gx, f x), diam(gfx)}, for any
x ∈ X.
Clearly, two commuting mappings satisfy above definition but the converse may not be
true, i.e., two commuting mappings must be weakly commuting, quasi commuting and
slightly commuting but the two mappings satisfying all these three conditions may not
be commuting.

3.2. Some Fixed Point Theorems

In this section we provide some core theorems on hybrid pair of mapping. We begin
with this lemma.
Lemma 3.2.1. Let X be a complete metric space and {An } and {Bn } be sequences in
B(X). If An −→ A ∈ B(X) and Bn −→ B ∈ B(X), then δ(An , Bn ) −→ δ(A, B).
30

Lemma 3.2.2. If {An } is a sequence of non-empty bounded subsets in the complete

metric space X. If δ(An , y) −→ 0, for some y ∈ X, then An −→ y.
Now, we recall that the set-valued map f : X ⇒ B(X) has a fixed point x if x ∈ f (x).
Definition 3.2.1. Let f : X ⇒ B(X) be a set-valued map. Then f is said to be
continuous at x ∈ X if for any sequence {xn } in X such that xn −→ x ∈ X, then we
have the sequence {f (xn )} in B(X) such that f (xn ) −→ f (x) ∈ B(X).
Further let us denote ψ, the set of all real functions χ such that

ψ = {χ : [0, ∞) −→ [0, ∞) | χ is non − decreasing, u. s. c. and χ(t) < t, t > 0}.

Throughout this chapter, we denote R+ the set of all non-negative reals, Φ the family
of all mappings φ : (R+ )5 −→ R+ such that φ is upper semi continuous, non-decreasing
in each coordinate variable and for any t > 0,

χ(t) = φ(t, t, a1 t, a2 t, a3 t) < t,

where χ : R+ −→ R+ and a1 + a2 + a3 = 8.
To accomplish our purpose we prove the following lemma.
Lemma 3.2.3. For every t > 0, we have χ(t) < t if and only if lim χn (t) = 0, where
n−→∞
χn denote the nth composition of χ.
Proof. Necessary Part : Since φ is upper semi continuous, then χ is upper semi
continuous.
Assume that
lim χn (t) = t, t > 0
n−→∞

⇒ t = lim χn+1 (t)

n−→∞
= χ lim χn (t)
n−→∞
= χ(t) < t (by definition)
⇒ t < t, which is a contradiction.
⇒t=0
∴ lim χn (t) = 0.
n−→∞
Sufficient Part : Since φ is upper semi continuous, then χ is non-decreasing such that

lim χn (t) = 0.
n−→∞

Then we have to show that χ(t) < t for t > 0.

We prove this by the method of contradiction.
Case 1. χ(t) > t, for some t > 0.

⇒ χn (t) > t, for some t > 0 and n = 1, 2, . . .

 31

⇒ lim χn (t) 9 0, which is a contradiction.

n−→∞
⇒ χ(t) ≤ t for t > 0.
Case 2. χ(t) = t, for some t > 0
⇒ χn (t) = t, for some t > 0 and n = 1, 2, . . .
⇒ lim χn (t) 9 0, which is a contradiction.
n−→∞
Thus, in all we have
χ(t) < t, for t > 0.
This completes the proof.
Lemma 3.2.4. If dn = δ(An , Bn ), then lim dn = 0.
n−→∞
Proof. First, we show that the sequence {dn } is a decreasing sequence.
On contrary, suppose that it is increasing, i.e.,
d2n+1 > d2n
1
⇒ d2n+1 ≤ (φ(d2p 2p 2p 2p 2p
2n+1 , d2n+1 , 4d2n+1 , 2d2n+1 , 2d2n+1 ))
2p

1
≤ (χ(d2p
2n+1 ))
2p

≤ d2n+1
which is a contradiction. Therefore, we have
d2n+1 ≤ d2n .
Similarly, we can show that
d2n+2 ≤ d2n+1
This shows that the sequence {dn } is a decreasing sequence.
Now, we have
d2p 2p 2p 2p 2p 2p
2 ≤ (φ(d1 , d1 , 4d1 , 2d1 , 2d1 )

≤ χn (d2p
1 )
By induction, we have
d2p n 2p
n+1 ≤ χ (d1 )

If d1 > 0, then by Lemma 3.2.3, we have lim dn = 0.

n−→∞
If d1 = 0, then obviously dn = 0, for n = 1, 2, . . . .
Lemma 3.2.5. {yn } is a Cauchy sequence in X.
Proof. To show that {yn } is a Cauchy sequence, it is enough to show that {y2n } is a
Cauchy sequence.
Suppose on contrary that {y2n } is not a Cauchy sequence, then for ε > 0 such that for
an even integer 2r, ∃ even integers
2m > 2n > 2r
32

such that
d(y2n , y2m ) > ε (3.2.1)
For every integer 2r, let 2m be the least positive integer exceeding 2n satisfying Equation
3.2.1 such that
d(y2n , y2m−2 ) < ε. (3.2.2)
Now,
ε ≤ d(y2n , y2m )
≤ d(y2n , y2m−2 ) + d(y2m−2 , y2m−1 ) + d(y2m−1 , y2m ).
Then by Equation 3.2.1 and Equation 3.2.2 it follows that
lim d(y2n , y2m ) = ε. (3.2.3)
r−→∞

Also, by triangular inequality we have

|d(y2n , y2m−1 ) − d(y2n , y2m )| < d(y2m−1 , y2m ).
From Equation 3.2.3, we have
d(y2n , y2m−1 ) −→ ε as r −→ ∞.
Now, by Equation 3.2.2, we get
d(y2n , y2m ) ≤ d(y2n , y2n+1 ) + δ(f x2n , gx2m )
≤ d(y2n , y2n+1 ) + {φ(d2p (y2n , y2m ),
0
dq (y2r , y2r+1 )dq (y2m−2 , y2m−1 ),
[dr (y2n , y2m−1 ) + dr (y2m−2 , y2m−1 )]×
0 0
[dr (y2m−1 , y2n ) + dr (y2n , y2n+1 )],
0 0
ds (y2n , y2n+1 )[ds (y2m−1 , y2n ) + ds (y2n , y2n+1 )],
0 1
[dl (y2m−1 , y2n )+dl (y2m−2 , y2m−1 )]dl (y2m−2 , y2m−1 ))} 2p .
taking r −→ ∞, we get
1
ε ≤ {φ(ε2p , 0, ε2p , 0, 0)} 2p
1
≤ (ε2p ) 2p
<ε
which is a contradiction. Thus, {yn } is a Cauchy sequence in X.
Definition 3.2.2. Consider the mappings f, g : X ⇒ X, then we say that f and g are
weakly commuting on X if for any x ∈ X
δ(f gx, gf x) ≤ max{δ(gx, f x), diam(gf x)} (3.2.4)

Remark 3.2.1. If two mappings f and g are commuting, then it will be weakly
commuting but converse does not hold in general.
 33

Example 3.2.1. Let X = [0, 2] and define f, g : X ⇒ B(X) such that

x x
g(x) = and f (x) = [0, ], ∀x ∈ X.
20 x + 20
Then
x
f gx = [0, ]
x + 400
and
x
gf x = [0, ]
20x + 400
This shows that f and g do not commute.
But
x x
δ(f gx, gf x) = d([0, ], [0, ])
x + 400 20x + 400
x
=
x + 400
and
x x 20x
max{δ(g(x, f x), diam(gf x))} = max{δ([ , [0, ]), diam([0, ])}
20 x + 20 x + 400
x
= max{ , 0}
20
x
= .
20
Obviously,
x x
< , for x > 0.
x + 400 20
Thus, f and g commute weakly.

3.3. Some Hybrid Fixed Point Theorems

Here, we consider
f (x) ⊂ g(x), (3.3.1)
i.e., range of f contained in range of g.
Further, let x◦ ∈ X and y1 be any element in f (x◦ ), ∃ x1 ∈ X such that g(x1 ) = y1
In general, if we choose xn−1 ∈ X and for any yn in f (xn−1 ), ∃ xn ∈ X such that
g(xn ) = yn , then we define

δf (x◦ ) = sup{δ(f xn , f x1 ) : n = 0, 1, 2, . . .} < ∞. (3.3.2)

As the consequence of the above we have the following theorem.

Theorem 3.3.1. Let f : X ⇒ B(X) and g : X −→ X satisfying Equation 3.2.4 and
Equation 3.3.1 such that

δ(f x, f y) ≤ χ(max{δ(gx, f x), δ(gy, f y), δ(gx, f y), δ(gy, f x), d(gx, gy)}) (3.3.3)
34

where χ ∈ Ψ and x, y ∈ X.
If there exists a point x◦ ∈ X satisfying Equation 3.3.1 and if one of f and g is contin-
uous, then f and g have a unique common fixed point {z} = g(z) = f (z).
Proof. Let z be a common fixed point of f and g.
Obviously, z ∈ f (z) and g(z) = z.
By Equation 3.3.2, we have
δ(z, f z) ≤ δ(f z, f z)

≤ χ(max{δ(gz, f z), d(gz, gz)})

= χ(δ(z, f z))
Being χ(t) < t, for t > 0.
The above inequality implies
δ(z, f z) = 0.

⇒ f (z) = z.
If w is another fixed point of f and g, such that w 6= z, then in view of Equation 3.3.2,

d(w, z) = δ(f w, f z)

≤ χ(max{δ(gw, f z), δ(gz, f w), d(w, z)})

≤ χ(d(w, z))
< d(w, z)
which is a contradiction and hence, w = z.
Now, we prove the existence of a fixed point z,
Case (i) Suppose that δ(f x◦ , f x1 ) = 0.
Thus, we obtain f x◦ = y1 = gx1 = f x1 = y2 .
Being diam(gf x1 ) = 0 because f x1 is a singleton, from Equation 3.2.4:

gf x1 = f gx1 = f f x1 . (3.3.4)

From Equation 3.3.3, we have

δ(f f x1 , f x1 ) ≤ χ(max{δ(gf x1 , f f x1 ), δ(gx1 , f x1 ), δ(gf x1 , f x1 ),

δ(gx1 , f f x1 ), d(gf x1 , f x1 )})

= χ(max{0, 0, δ(f f x1 , f x1 ), 0, δ(f x1 , f x1 )})
= χ(δ(f f x1 , f x1 ))
< δ(f f x1 , f x1 )
which gives f f x1 = f x1 . Thus, f x1 is a fixed point of f and since Equation (3.3.4)
holds, f x1 is also a fixed point of g.
Thus, the theorem is proved.
 35

Case (ii) Suppose that δ(f x◦ , f x1 ) > 0.

Since
δ(f xr , f xs ) ≤ δ(f xr , f x1 ) + δ(f xs , f x1 )
≤ 2δf (x◦ ),
we have from Equation 3.3.2 that
M = sup{δ(f xr , f xs ) : r, s = 0, 1, 2, . . .}
is finite. Being M ≥ δ(f x◦ , f x1 ) > 0, by Lemma 3.2.3, let p be an integer such that
χp (M ) < ε, for arbitrary ε > 0.
By applying Equation 3.3.3 to the term δ(f xm , f xn ), being χ non-decreasing, we have
for m, n > p
δ(f xm , f xn ) ≤ χ(max{δ(ym , f xm ), δ(yn , f xn ), δ(ym , f xn ),

δ(yn , f xm ), d(ym , yn )}) (3.3.5)

≤ χ(max{δ(f xm−1 , f xm ), δ(f xn−1 , f xn ), δ(f xm−1 , f xn ),
δ(f xn−1 , f xm ), δ(f xm−1 , f xn−1 )})
because ym = gxm ∈ f xm−1 , yn = gxn ∈ f xn−1 .
By iterating Equation 3.3.5 p times, we deduce for m, n > p
δ(f xm , f xn ) ≤ χ(max{δ(f xr , f xs ), δ(f xr , f xh ), δ(f xs , f xk ) :
m − 1 ≤ r; h ≤ m; n − 1 ≤ s; k ≤ n})
≤ χ(max{δ(f xr , f xs ), δ(f xr , f xh ), δ(f xs , f xk ) :
m − 2 ≤ r; h ≤ m; n − 2 ≤ s; k ≤ n})
..
.
≤ χp (max{δ(f xr , f xs ), δ(f xr , f xh ), δ(f xs , f xk ) :
m − p ≤ r; h ≤ m; n − p ≤ s; k ≤ n})
≤ χp (M ) < ε.
If zn is chosen arbitrarily in f xn for n = 1, 2, . . . , we obtain that
d(zm , zn ) ≤ δ(zm , f xn ) ≤ δ(f xm , f xn ) < ε, (3.3.6)
for m, n > p. Therefore, {zn } is a Cauchy sequence and by completeness of X, converges
to some point z ∈ X.
Let us now assume that g is continuous and then the sequence {gzn } converges to gz.
On the other hand, being g continuous, in correspondence of ε, we can certainly find
σ > 0 such that d(gzm , gzn ) < ε whenever d(zm , zn ) < σ. Then there exists an integer q
such that d(zm , zn ) < σ, for every m, n > q. So, for m, n > q,
d(gzm , gzn ) < ε. (3.3.7)
36

The inequality holds for arbitrary zn ∈ f xn . Therefore, Equation 3.3.7 implies

d(gzm , gf xn ) ≤ ε, (3.3.8)
for m, n > q. If we put zn = yn+1 = gxn+1 ∈ f xn , we have that the sequence {gxn } =
{yn } converges to z. Consequently, the sequence {g 2 xn } = {gyn } converges to gz and
from Equations 3.3.6 and 3.3.8 we have
δ(ym , f xn ) < ε and δ(gym , gf xn ) ≤ ε, 3.3.9
for m, n > max{p, q}.
On using Equations 3.2.4, 3.3.3 and 3.3.9, we obtain the following inequalities for any
n > max{p, q}
d(gyn+1 , yn+1 ) = d(g 2 xn+1 , gxn+1 ) ≤ δ(gf xn , gxn+1 )
≤ δ(f gxn , gf xn+1 ) + δ(f gxn , gxn+1 )
≤ max{δ(yn , f xn ), diam(gf xn )} + δ(f gxn , f xn )
≤ max{ε, 2δ(gyn , gf xn )} + χ(max{δ(gyn , f gxn ),
δ(yn , f xn ), δ(gyn , f xn ), δ(yn , f gxn ), d(gyn , yn )})
≤ max{ε, 2ε} + χ(max{δ(gyn , gf xn ) + δ(gf xn , f gxn ),
ε, d(gyn , yn ) + δ(yn , f xn ), d(gyn , yn ) + δ(gyn , f gxn ),
d(gyn , yn )})
≤ 2ε + χ(max{3ε, ε, d(gyn , yn ) + ε, d(gyn , yn ) + 2ε,
d(gyn , yn )})
= 2ε + χ(max{3ε, d(gyn , yn ) + 2ε}).
By letting n −→ ∞, in account of the right continuity of χ, we have
d(gz, z) ≤ 2ε + lim sup χ(max{3ε, d(gyn , yn ) + 2ε})
n−→∞

≤ 2ε + χ(max{3ε, d(gz, z) + 2ε})

and for ε −→ 0+ , this inequality yields
d(gz, z) ≤ lim+ χ(max{3ε, d(gz, z) + 2ε})
ε−→0
≤ χ(d(gz, z))
which gives gz = z because χ(t) < t for t > 0.
Now, we prove that z is also a fixed point of f.
Since zn lies in f xn , by Equations 3.3.3 and 3.3.9, we have for n > max{p, q}
δ(f z, zn ) ≤ δ(f z, f xn )
χ(max{δ(z, f z), δ(yn , f xn ), δ(z, f xn ), δ(yn , f z), d(z, yn )})
≤ χ(max{δ(z, f z), ε, d(z, yn )+ε, δ(yn , f z), f d(z, yn )}).
 37

For n −→ ∞, we deduce
δ(f z, z) ≤ χ(max{δ(z, f z), ε})
being χ right continuous. By letting ε −→ 0+ , finally we have

δ(f z, z) ≤ χ(δ(f z, z))

which gives f z = {z}.

Let’s now suppose that f is continuous. Then the sequence {f xn } converges to f z and
from Equation 3.3.6 we obtain for n > p

δ(z, f zn ) ≤ δ(z, zn ) + δ(zn , f zn )

= d(z, zn ) + δ(zn , f zn )
< d(z, zn ) + ε
and as n −→ ∞, we deduce δ(z, f z) ≤ ε which implies {f z = z} being ε arbitrary.
Since the range of g contains the range of f, let z 0 ∈ X such that

{z} = f z = gz 0 .

Applying Equations 3.3.3, 3.3.6 and remembering zn = gxn+1 ∈ f xn for n = 0, 1, 2, . . . ,

we have for n > p
δ(zn+1 , f z 0 ) ≤ δ(f xn+1 , f z 0 )
≤ χ(max{δ(zn , f xn+1 ), δ(z, f z 0 ), δ(zn , f z 0 ),
δ(z, f xn+1 ), d(zn , z)})
≤ χ(max{ε, δ(z, f z 0 ), δ(zn , f z 0 ), d(z, zn ) + ε,
d(zn , z)})
which yields for n −→ ∞

δ(z, f z 0 ) ≤ χ(max{ε, δ(z, f z 0 )}).

By letting ε −→ 0+ , we also get δ(z, f z 0 ) ≤ χ(δ(z, f z 0 )), i.e.,

f z 0 = {z} = gz 0 .

From Equation 3.2.4, it follows

δ(f gz 0 , gf z 0 ) ≤ max{(δ(gz 0 , f z 0 ), diam(gf z 0 )} = 0

resulting gf z 0 = gz singleton.
So, gz = gf z 0 = f gz 0 = f z = {z},, i.e., z is also a fixed point of g.
This completes the proof.
Example 3.3.1. Let X be a set with any metric d and f, g : X ⇒ X given by

f x = f y = x, gx = y, gy = x.
38

All the assumptions of Theorem 3.3.1 are immediately verified except being
d(f gy, gf y) = d(f x, gx) = d(x, y) > 0 = d(x, x) = d(gy, f y),
but g has no fixed points.
Example 3.3.2. Let X = {x, y, w, z} a finite set with metric d given by
d(x, w) = d(x, z) = d(y, w) = d(y, z) = 2,

d(x, y) = d(w, z) = 4.
Let f, g : X ⇒ X be defined as
f x = f y = f w = y, f z = w,

gx = gy = gw = x, gz = z.
Resulting
d(f gx, gf x) = d(f gy, gf y) = d(f gw, gf w) = 2 = d(gx, f x) = d(gy, f y) = d(gw, f w)

and d(f gz, gf z) = d(x, w) = 1 < 2 = d(z, w) = d(gz, f z),

f and g verify
Moreover, being
d(f x, f y) = d(f x, f w) = d(f y, f w) = d(y, y) = 0
and 
d(f x, f z)


d(y, w) = d(f y, f z)


d(f w, f z)


1 1
d(y, w) = 2 = 4 = d(x, y)
2 2

1
 2 d(gx, f x)


= 1
2
d(gy, f y)


1
2
d(gw, f w),
it suffices to assume χ(t) = 2t for any t ≥ 0, in order to satisfy Equation 3.3.3. Obviously,
f and g are both continuous in X and Equation 3.3.2 holds, therefore all the assumptions
of Theorem 3.3.1 are satisfied except Equation 3.3.1, but f and g have no common fixed
points.
Remark 3.3.1. In the above theorem the Equation 3.3.2, i.e.,
δ(f gx, gf x) ≤ max{δ(gx, f x), diam (gf x)}
is necessary even if f is single-valued.
 39

Example 3.3.3. Let X = {a, b} with metric d and consider the functions f, g : X ⇒ X
such that
f (a) = f (b) = a
and g(a) = b, g(b) = a.
Here, all the assumptions of Theorem 3.3.1 are satisfied except Equation 3.3.2

d(f gy, gf y) = d(f x, gx) = d(x, y) > 0 = d(x, x) = d(gy, f y).

We can see that g does not have fixed points.

Remark 3.3.2. In the above theorem the Equation 3.3.3, i.e.,

f (X) ⊂ g(X)

is also necessary. We justify it by the following example.

Example 3.3.4. Let X = {k, l, m, n} with metric d such that

d(k, m) = d(k, n) = d(l, m) = 1, d(k, l) = d(m, n) = 2.

Consider f, g : X ⇒ X such that

f (k) = f (m) = f (n) = b, f (n) = c,

g(k) = g(l) = g(m) = k, g(n) = n.

Now, we have

d(f gk, gf k) = d(f gl, gf l) = d(f gm, gf m) = 2 = d(gk, f k) = d(gl, f l) = d(gm, f m)

and
d(f gn, gf n) = d(k, m) = 1 < 2 = d(n, m) = d(gn, f n),
then f and g satisfy Equation 3.3.2,
also
d(f k, f l) = d(f k, f m) = d(f l, f n) = d(l, l) = 0
and 
d(f k, f m)


d(l, n) = d(f l, f m)


d(f n, f m)


1 1
d(l, n) = 1 = 2 = d(x, y)
2 2

1
 2 d(gk, f k)


= 12 d(gl, f l)


1
2
d(gn, f n).
40

It is enough to assume that χ(t) = 2t for t > 0 in order to satisfy Equation 3.3.4.
Clearly, f and g both are continuous in X and Equation 3.3.1 holds. hence, all the
assumptions of Theorem 3.3.1 are satisfied except Equation 3.3.3, but f and g do not
have common fixed point.
Remark 3.3.3. The condition that one of f and g are continuous is also necessary in
Theorem 3.3.1.
Example 3.3.5. Let X = [0, 1] and let d be the metric on X and f : X ⇒ B(X), g :
X ⇒ X such that 
1, x=0
f (x) = 2
[0, x ], x 6= 0,
4

1, x=0
g(x) =
x, x 6= 0.
2
Now,
1 1
δ(f g(0), gf (0)) = < = δ(g(0), f (0)
4 2
and
x
δ(f gx, gf x) = = diam (gf x), for x 6= 0,
8
then Equation 3.3.2 holds.
Also, for x = 0 and y ∈ X, δ(f (0), f (y)) ≤ 21 = 12 .1 = 12 δ(g(0), f (y))
and for x 6= 0, y 6= 0, δ(f (x), f (y)) = 41 max{x, y} = 21 max{δ(g(x), f (x)), δ(g(y), f (y))}.
Since, f (X) = [0, 41 ] { 12 } ⊂ [0, 12 ] {1} = g(X),
S S

thus, Equation 3.3.1 is verified also Equation 3.3.2 holds. So, all the conditions of the
Theorem 3.3.1 are satisfied.
Assume χ(t) = 2t , t ≥ 0 except the continuity of f and g, which have no fixed point.
 References

[1] J. P. Aubin and H. Frankowska, Set-Valued Analysis, Modern Birkhauser Classics,

Birkhauser Boston, Inc. Boston, 1990.
[2] Q. H. Ansari, Metric Spaces Including Fixed Point Theory and Set-Valued Maps,
Narosa Publishing House, New Delhi, 2010
[3] E.T. Copson, Metric Spaces, Cambridge University Press, 1990.
[4] V. I. Istratescu, Fixed Point Theory An Introduction, D. Reidel Publishing, 1981.
[5] P. K. Jain and K. Ahmad, Metric Spaces, Narosa Publishing House Pvt. Ltd., New
Delhi, 1993.
[6] M. A. Khami and W. A. Kirk, An Introduction to Metric Spaces and Fixed Point
Theory, John Wiley and Sons, Inc. New York, 2001.
[7] James R. Munkers, Prentice Hall of India Private Limited, New Delhi, 2005
[8] Mursaleen, Elements of Metric Spaces, Anamaya Publishers, New Delhi, 2005.
[9] S. B. Nadler, Multi-Valued Contraction Mappings, Pacific Journal of Mathematics,
1969.
[10] George F. Simmons, Kogakeesha Company Ltd., 1963.
[11] D. R. Smart, Fixed Point Theorems, Cambridge University Press, Cambridge, 1974.
[12] S.Sessa, M. S. Khan and M. Imdad, A Common Fixed Point Theorem With A
Weak Commutativity Condition, Glasnik Mathematicki , 1986.
[13] Duran Turkoglu, Ishak Altun, Brian Fisher, Common Fixed Point Theorems for
Four Mappings With Some Weak Conditions of Commutativity Novi Sad J. Math.,
2006.

41

View publication stats