ANALELE ŞTIINŢIFICE ALE UNIVERSITĂŢII ”AL.I.

CUZA” IAŞI
Tomul XLIV, s.I.a, Matematică, 1998, f1

GENERALIZED D–METRIC SPACES

AND
MULTI–VALUED CONTRACTION MAPPINGS

BY

B.C. DHAGE

Abstract. In this paper a new structure of a generalized D –metric space is

introduced and illustrated with some examples. Some fixed point theorems bfor multi–
valued contraction and contractive mappings in D –metric spaces are proved which in
turn include the well–known fixed point theorems of Nadler [10] and Smithson [11] in
ordinary metric spaces and D –contraction principle of Dhage [2,4] in D –metric spaces as
the special cases.

1. Introduction. Throughout this paper, let X a non–empty set.

Let (X, d) be an ordinary metric space. Suppose that 2X denotes the class
of non–empty subsets of X and we assume that the members of 2X are
closed and bounded in the topology τd generated by the ordinary metric
d on X. Then by a generalized metric on 2X , we mean a function H on
2X × 2X satisfying the conditions of a ordinary metric on X. Hausdorff [10]
considered the generalized metric as a set–function H = 2X × 2X → IR+
defined by

(1.1) H(A, B) = inf{ > 0 : A ⊂ N (B, ), B ⊂ N (A, )}

(This paper is dedicated in the loving memory of my late beloved mother

Smt.Kasubai C.Dhage who expired on 17–10–1994.)
1991 Mathematics subject classification: 54 H 25
Key–words and phrases: D –metric space, contraction mapping, fixed point etc.
180 B.C. DHAGE 2

Where
N (A, ) = ∪ N (a, ) =
a∈A

= ∪ {x ∈ X|d(a, x) < }
a∈A

The function H satisfies all the properties of a metric on 2X and hence

a generalized metric called Hausdorff metric on 2X . Then the class of 2X
of all non–empty, closed and bounded subsets of (X, d) together with the
Hausdorff metric H becomes a generalized metric space or more specific a
Hausdorff metric space and we denote it by (2X , H). It is a natural desire
to have the generalization of a generalized metric space wherein the set
function is a function of more that two variables of 2X . It is the sole aim of
this paper to introduce the notion of higher dimensional generalized metric
spaces called generalized D–metric spaces, wherein the set–function is a
function of morethan two variables of 2X . We hope that by the introduction
of this new notion of a generalized D–metric space, the question like the
distance between three or more sets, mutually nearest points of three or
more sets etc., could be settled properly. Motivated by these and other like
problems, here we define the new structure of a generalized D–metric space
through the D–metric function D on X, a notion recently developed by the
present author [2] which is as follows.
A function D : X ×X ×X → IR is called a D–metric on X if it satisfies
the following properties :
(M1 ) D(x, y, z) ≥ 0 for all x, y, z ∈ X and the equality holds if and only if
x = y = z, (nonnegativity)
(M2 ) D(x, y, z) = D(x, z, y) = . . ., (symmetry)
(M3 ) D(x, y, z) ≤ D(x, y, a) + D(x, a, z) + D(a, y, z) for all x, y, z, a ∈ X.
(tetrahedral inequality)

The non–empty set X together with a D–metric D is called a D–metric

space and it is denoted by (X, D). The details of a D–metric space and its
generalizations are given in Dhage [2,3], and the topological properties of a
D–metric space are given in Dhage [3]. Below we give two examples of D–
metric spaces which we shalll need in the sequel.
Example 1.1. Let (X, d) be an ordinary metric space and define a
function D1 on X × X × X by

(1.2) D1 (x, y, z) = max{d(x, y), d(y, z), d(z, x)}

for x, y, z ∈ X. Then the function D1 is a D–metric on X and (X, D1 ) is a

D–metric space.
3 GENERALIZED D–METRIC SPACES 181

Example 1.2. Let (X, d) be an ordinary metric space and define a

function D2 on X × X × X by

(1.3) D2 (x, y, z) = d(x, y) + d(y, z) + d(z, x)

for x, y, z ∈ X. Then D2 is a metric on X and hence (X, D2 ) is a D– metric

space.
Generally we consider a D–metric D on X satisfying the one more
additional condition, namely,

(M4 ) D(x, y, y) ≤ D(x, z, z) + D(z, y, y)

for x, y, z ∈ X. There do exist D–metrics D satisfying the condition (M4 ).

In fact the D–metrics given in examples 1.1 and 1.2 satisfy this condition.
In the following section we give the definition of a generalized D–metric
space and give some examples to interpret the underlined ideas.

2. Generalized D–metric spaces.

Definition 2.1. Let (X, D) be a D–metric space and let 2X denote
the class of non–empty, closed and bounded subsets of X w.r.t. the topology
τD generated by the D–metric D on X. Then a function H : 2X × 2X ×
2X → IR is called a generalized D–metric on 2X if it satisfies the properties
(M1 ) − (M3 ) of the D–metric on X. The set 2X together with a generalized
D–metric H is called a generalized D–metric space and it is denoted by
(2X , H). Below in the following we explicity construct some examples of
generalized D–metric on 2X corresponding to a given D–metric D on X.
Let  > 0 be given and a ∈ X arbitrary. Then the neighbourhood
N (a, ) of the point a of radius  is a set in X defined by

(2.1) N (a, ) = {x ∈ X|D(a, x, x) <  and if y, z ∈ N (a, )

are any two points, then D(a, y, z) < }

and so N (A, ) = ∪ N (a, ).
a∈A
For A, B, C ∈ 2X , define a number H(A, B, C) by

(2.2) H(A, B, C) = inf{ > 0 : A ∪ B ⊂ N (C, ),

B ∪ C ⊂ N (A, ) and C ∪ A ⊂ N (B, )}

182 B.C. DHAGE 4

Lemma 2.1. The function H on 2X × 2X × 2X defined by (2.2) is

a generalized D–metric on 2X and hence (2X , H) is a generalized D–metric
space.
Proof: Obviously the first two properties of the generalized D–metric
follow from the definition of the function H. We only show that the rectangle
inequality holds. Let A, B, C, E ∈ 2X . Then by definition of H we have,

H(A, B, C) = inf{ > 0 : A ∪ B ⊂ N (C, ),

B ∪ C ⊂ N (A, ) and C ∪ A ⊂ N (B, )}.

Since

inf { > 0 : A ∪ B ⊂ N (C, ) }

≤ inf { > 0 : A ∪ E ⊂ N (C, ), B ∪ E ⊂ N (C, ) },
inf { > 0 : B ∪ C ⊂ N (A, ) }
≤ inf { > 0 : B ∪ C ⊂ N (A, ), C ∪ E ⊂ N (A, ) }
and inf { > 0 : C ∪ A ⊂ N (B, ) }
≤ inf { > 0 : C ∪ E ⊂ N (B, ), A ∪ E ⊂ N (B, ) }

we have

H(A, B, C) ≤ inf { > 0 : A ∪ E ⊂ N (C, ), B ∪ E ⊂ N (C, ),

C ∪ E ⊂ N (B, ), A ∪ E ⊂ N (B, ),
B ∪ E ⊂ N (A, ), C ∪ E ⊂ N (A, ) }
≤ inf { > 0 : A ∪ B ⊂ N (E, ), B ∪ E ⊂ N (A, )
and E ∪ A ⊂ N (B, ) }
+ inf { > 0 : A ∪ E ⊂ N (C, ), E ∪ C ⊂ N (A, )
and C ∪ A ⊂ N (E, ) }
+ inf { > 0 : E ∪ B ⊂ N (C, ), B ∪ C ⊂ N (A, )
and C ∪ A ⊂ N (E, ) }
= H(A, B, E) + H(A, E, C) + H(E, B, C).

Thus H is a generalized D–metric on 2X and hence (2X , H) is a gen-

eralized D–metric space and the proof of the lemma is complete.
5 GENERALIZED D–METRIC SPACES 183

Example 2.1. Let (X, d) and (X, D1 ) denote the metric and D–
metric space respectively, where d and D1 are related by (1.2). Define a
function H1 on 2X × 2X × 2X by

(2.3) H1 (A, B, C) = max{H(A, B), H(B, C), H(C, A)}

where H is a Hausdorff metric on 2X corresponding to the metric d. Now

a closed and bounded subset of X w.r.t. d implies that it is also bounded
and closed w.r.t. D1 , so the number H1 is well defined. We show that H1
satisfies the properties of a generalized D–metric on 2X . Obviously the first
two properties are satisfied. We only prove the rectangle inequality. For, let
A, B, C, E ∈ 2X , then

H1 (A, B, C) = max{H(A, B), H(B, C), H(C, A)}

≤ max{H(A, B), H(B, C), H(C, A),
H(A, E), H(B, E), H(C, E)}
≤ max{H(A, B), H(B, E), H(E, A)}
+ max{H(A, E), H(E, C), H(C, A)}
+ max{H(E, B), H(B, C), H(C, E)}
= H1 (A, B, E) + H1 (A, E, C) + H1 (E, B, C)

Thus H1 is a generalized D–metric on 2X and hence (2X , H1 ) is a

generalized D–metric space.
Remark 2.1. We note that

(2.4) H1 (A, B, B) = H(A, B)

for A, B ∈ 2X .
Example 2.2. Let (X, d) be an ordinary metric space and let (X, D2 )
be the associated D–metric space, where D2 is defined by (1.3). Then a
subset of X is closed and bounded w.r.t. D2 . Define a function H2 on
2X × 2X × 2X by

(2.5) H2 (A, B, C) = H(A, B) + H(B, C) + H(C, A)

for A, B, C ∈ 2X , where H is a Hausdorff metric on 2X corresponding to

d. It is easy to see that H2 satisfies first two properties of a generalized
184 B.C. DHAGE 6

D–metric on 2X . To show the rectangle inequality, let A, B, C, E ∈ 2X ,

then we have

H2 (A, B, C) = H(A, B) + H(B, C) + H(C, A)

≤ H(A, B) + H(B, E) + H(E, A)
+ H(A, E) + H(E, C) + H(C, A)
+ H(E, B) + H(B, C) + H(C, E)
= H2 (A, B, E) + H2 (A, E, C) + H2 (E, B, C).

Thus H2 is a generalized D–metric on 2X and hence (2X , H2 ) is a

generalized D–metric space.
Remark 2.2. For A, B ∈ 2X , we have

(2.6) H2 (A, B, B) = 2H(A, B).

A generalized Dm –metric space with the generalized D–metric Hm as

a function of m variables may be defined via Dm –metric Dm on X. The
details of a Dm –metric Dm on X are given in Dhage [2]. Let (X, Dm ) be a
Dm –metric space and let 2X denote the class of closed and bounded subsets
of the Dm –metric space X. Let denote the set of natural numbers, m ∈ IN
 IN m
X
and M = {1, 2, . . . , m}. By 2
 = 2X × 2X × . . . (m times), which is
the set of all system A = (Ai )i∈M = (A1 , . . . , Am ) of elements of 2X called
the components  of A.Let p denotethe permutations  on M such that for
m  m
X X
A = (Ai )i∈M ∈ 2  , p(A) = Ap(Ai )  ∈ 2  . By AAi →B ,
 m i∈M
X th
we denote an element 2  in which i component Ai is replaced by B.
 m
Then a generalized Dm –metric Hm is a function on 2X  satisfying the
properties of a Dm –metric Dm on X,
mnamely
(P1 ) Hm (A) ≥ 0 for all A ∈ 2X  , and the equality holds if and only
if all components of A are equal, i.e. A1 = A2 = . . . = Am .
(P2 ) Hm (A) = H(p(A))
P (symmetry)
(P3 ) Hm (A) ≤ Hm (AAi →B ) (polyhedral inequality)
i∈M
X
The set 2 together with a generalized Dm –metric Hm is called a
generalized Dm –metric space and it is denoted by (2X , Hm ). Below we
7 GENERALIZED D–METRIC SPACES 185
 m
construct a generalized Dm –metric on 2X . Let A = (A1 , . . . , Am ) ∈ 2X 
 m
be arbitrary. Define a number Hm on 2X  by

(2.7) Hm (A) = Hm (A1 , . . . , Am ) =

m
= inf { > 0 : ∪ Ai ⊂ N (Aj , )}
i=1
i6=j
 m
Lemma 2.2. The function Hm on 2X  defined by (2.7) is a
generalized Dm –metric on 2X and hence (2X , Hm ) is a generalized Dm –
metric space.
Proof. The proof is similar to Lemma 2.1. and we omit the details.
The definitions of generalized D–metrics given in (2.2), (2.3), (2.5) and
(2.7) are natural and considered to be standard, however,  other
mdefinitions
X
may also be given. For example, define a function H on 2  by

H(A, B, C) = if any two of A, B, C ∈ 2X are distinct,
1
0
if A = B = C
 m
Similarly, define a function Hm on 2X  by

 1 if any two of A1 , . . . , Am ∈ 2X
Hm (A) = Hm (A1 , . . . , Am ) = are distinct,
0 if A1 = A2 = . . . = Am .


It is easy to prove that the functions H and Hm are respectively gen-

eralized D–metric and generalized Dm –metric on 2X .
In this paper, we do not deal with the generalized Dm –metric spaces,
but restrict our discussion to generalized D–metric spaces, wherein the gen-
eralized D–metric is a function of three variables only.

3. Topology of (2X , H)
Let A ∈ 2X be arbitrary and let  > 0 be given. Consider the balls
∗
B (A, ) and B(A, ) in a generalized D–metric space (2X , H) defined by

(3.1) B ∗ (A, ) = {B ∈ 2X |H(A, B, B) < } and

(3.2) B(A, ) = {b ∈ B ∗ (A, )| if B, C ∈ B(A, ) are any two elements,

then H(A, B, C) < }
186 B.C. DHAGE 8

It is clear that B(A, ) ⊆ B ∗ (A, ) and for a given  > 0 we can find
a σ > 0 such that B ∗ (A, σ) ⊆ B(A, ). We assume that the generalized
D–metric H satisfies the property (M4 ),

(3.3) H(A, B, B) ≤ H(A, E, E) + H(E, B, B)

for A, B, E ∈ 2X , it is easy to verify that the generalized D–metric H defined

by (2.2), (2.3) and (2.5) satisfy this property. Now it can be shown on the
lines of a D–metric space (see Dhage [3]) that the collections {B ∗ (A, ) :
A ∈ 2X } and {B(A, ) : A ∈ 2X } of all bals in 2X form the bases for a
given topology on 2X which we denote by τ ∗ and τ respectively. It is not
difficult to prove that τ ∗ and τ are equivalent topologies on 2X . Throughout
the subsequent part of this paper, we assume that 2X is equipped with
the topology τ and in this case we say 2X is a H–metrizable topological
space. The topological properties of a generalized D–metric space (2X , H)
are similar to that of a D–metric space. Now we will consider some basic
definitions in a generalized D–metric space which we shall use in the sequel.
Definition 3.1. A sequence {An } of points of a generalized D–
metric space 2X is said to be H–convergent and H–converges to a member
A of 2X if for  > 0 there exists an n0 ∈ IN such that H(Am , An , A) <  for
all m, n ≥ n0 .
Definition 3.2. A sequence {An } in a generalized D–metric space
2X is said to be H–Cauchy if for  > 0 there exist an n0 ∈ IN such that
H(Am , An , Ap ) <  for all m > n, p ≥ n0 .
Definition 3.3. A complete generalized D–metric spaces (2X , H) is
one in which every H–Cauchy sequence H–converges to a point in 2X .
Remark 3.1. We note that the topology τ generated by the collection
{B(A, ) : A ∈ 2X } of all balls is same as the topology of the H–metric
convergence, i.e. the convergence of a sequence {An } in τ implies the H–
convergence given in the sense of definition 3.1. and vice–versa.
Preposition 3.1. A generalized D–metric H is a continuous func-
tion on 2X × 2X × 2X in the topology of the H–metric convergence.
Proof. The proof is similar to the continuity of a D–metric function
D on X 3 and is given in Dhage [3].
Remark 3.2. Let (X, d) be an ordinary metric space and let (X, D)
be a associated D–metric space. Let H be the Hausdorff metric on 2X
corresponding to ordinary metric d. Then the generalized D–metrics H, H1
9 GENERALIZED D–METRIC SPACES 187

and H2 defined respectively by (2.2), (2.3) and (2.5) generate the equivalent
topologies on 2X .
A generalized D–metric space (2X , H) is said to be H– bounded if there
exists a constant M > 0 such that H(A, B, C) ≤ M for all A, B, C ∈ 2X . In
this case M is called a H–bound of 2X .
Example 3.1. Let (2X , HJ ) be a generalized D–metric space and
consider a generalized D–metric space (2X , H), where H is defined by

M H(A, B, C)
(3.4) H= , M > 0, k > 0,
k + H(A, B, C)

for A, B, C ∈ 2X . It can be easily shown that the function H is a generalized

D–metric on 2X and (2X , H) is a generalized D–metric space. Obviuosly,
(2X , H) is a H–bounded generalized D–metric space with H–bound M .
A generalized D–metric space (2X , H) inherits some properties of the
associated D–metric space X, D) which are given in the following.
Theorem 3.1. If (X, D) is a bounded D–metric space, then the
associated generalized D–metric space 2X , H) is H–bounded.
Proof. The proof follows from the definition of the generalized D–
metric H on 2X .
Theorem 3.2. If (X, D) is complete, then the associated generalized
D– metric space (2X , H) is complete.
Proof. Let {An } be a H–Cauchy sequence in 2X . Then for  > 0,
there exists an n0 ∈ IN such that

(3.5) H(Am , An , Ap ) < 

for m > n, p ≥ n0 .
Consider the sequence {xn } in X defined by

(3.6) xn ∈ An , n = 1, 2, . . .

Then by (3.5), we get

1
(3.7) D(xm , xn , xp ) < + αm
2m
for some α, 0 < α < 1.
Letting m → ∞ in (3.7), one gets D(xm , xn , xp ) → 0. This shows
that {xn } is a D–Cauchy sequence in X. Since X is complete, there is a
188 B.C. DHAGE 10

point x∗ ∈ X such that lim x = x∗ . Let A be the set of all such points
n→∞
in X. Clearly A is closed and bounded of the D–metric space X and hence
A ∈ 2X . Then by definition of A, for  > 0, there exists σ > 0, σ <  such
that An ⊆ N (A, σ) for all n ≥ n0 . Obviously, A ⊂ N (A, σ) for all n ≥ n0 ,
since each An is closed and bounded subset of X. Again as  is arbitrary,
we have lim An = A. Hence (2X , H) is a complete generalized D–metric
n→∞
space. This completes the proof.
Theorem 3.3. If X, D) is a precompact, i.e. totally bounded D–
metric space, then so is also (2X , H).
Proof. Given  > 0, choose σ <  and let S be a finite subset of X
such that the collection {BD (x, σ) : x ∈ S} convers X. Let S be the class
of all non–empty subsets of S. Then obviously, {BH (A, ) : A ∈ S} convers
2X . Hence (2X , H) is a totally bounded generalized D–metric space.
Theorem 3.4. If (X, D) is a compact D–metric space, then (2X , H)
is a compact generalized D–metric space.
Proof. The proof follows from Theorems 3.2. and 3.3. We omit the
details.

4. Multi–valued contraction mappings.

Nadler [10] first proved the fixed point theorems for multi–valued con-
traction mappings in complete ordinary metric spaces and since then sev-
eral authors, see for examples, Ciric [1], Smithson [11] and Yanagi [12] and
the the references given therein, have proved the fixed point theorems for
the multi–valued mappings in complete metric spaces satisfying more gen-
eral contraction type conditions. In this section we prove some fixed point
theorems for multi–valued mappings in D–metric spaces satisfying certain
contraction conditions.
A correspondence F : (X, D) → (2X , H) is called a multi–valued map-
ping or multi–mapping (in short multi–map) on X. A point x∗ ∈ X is called
a fixed point of the multi–map F if x∗ ∈ F x∗ .
Definition 4.1. A multi–map F : (X, D) → (2X , H) is said to be
Lipschitzian with Lipschitz constant α if
(4.1) H(F x, F y, F z) ≤ α D(x, y, z)
for all x, y, z ∈ X.
Lemma 4.1. Let F : (X, D) → (2X , H) be a multi–map and define
a multi–map Fb : (2X , H) → (2X , H) by Fb(A) = ∪ F x, A ∈ 2X . It F is a
x∈A
11 GENERALIZED D–METRIC SPACES 189

Lipschitzian with Lipschitz constant α, then so is Fb with the same Lipschitz

constant, i.e. for any A, B, C ∈ 2X , we have
 
(4.2.) H Fb(A), Fb(B), Fb(C) ≤ α H(A, B, C)

Proof. The proof follows from the definition of the generalized D–

metric H on 2X .
Definition 4.2. A multi–map F : (X, D) → (2X , H) is called a
multi–valued contraction if it is a Lipschitzian with the Lipschitz constant
α < 1. The constant α is called the contraction constant for F .
Lemma 4.2. If F : (X, D) → (2X , H) is a multi–valued contraction
mapping and if Fb : (2F , H) → (2X , H) is a associated multi–map, then
Fb is also a multi–valued contraction mapping with the same contraction
constant.
Theorem 4.1. Let X be a complete and bounded D–metric space
and let Fb : (2X , H) → (2X , H) be a multi–valued contraction mapping.
Then Fb has a unique fixed point, i.e. there is a member A∗ ∈ 2X such that
Fb(A∗ ) = A∗ .
Proof. Since X is complete and bounded D–metric space, by Theo-
rems 3.1. and 3.2., (2X , H) is a complete and bounded generalized D–metric
space. Now the conclusion of the theorem follows by an application of The-
orem 2.1. of Dhage [2].
Corollary 4.1 (Nadler [10]). Let (X, d) be a complete and bounded
metric spaces, and let Fb : (2X , H) → (2X , H) be a multi–mapping satisfying

(4.3) H(Fb(A), Fb(B)) ≤ α H(A, B)

for all A, B ∈ 2X and 0 ≤ α < 1. Then Fb has a unique fixed point.

Proof. Define a generalized D–metric H on 2X by (2.3). Since
(2 , H) is complete and bounded, (2X , H) is also complete and bounded.
X

Let A, B, C ∈ 2X be arbitrary. Then by (4.3), we have,


 H(Fb(A), Fb(B)) ≤ αH(A, B)
(4.4) H(Fb(B), Fb(C)) ≤ αH(B, C) and

H(Fb(C), Fb(A)) ≤ αH(C, A).
190 B.C. DHAGE 12

The inequalities in (4.4) imply that

H(Fb(A), Fb(B), Fb(C)) =

= max{H(Fb(A), Fb(B)), H(Fb(B), Fb(C)), H(Fb(C), Fb(A))}
≤ α max{H(A, B), H(B, C), H(C, A)}
= α H(A, B, C).

This shows that Fb is a multi–valued contraction in (2X , H). Hence by

an application of Theorem 4.1. yields that Fb has a unique fixed point.
Remark 4.1. The example given in Nadler [10] shows that the fixed
points of multi–valued contraction mapping Fb need not be same as the fixed
point set of F .
We need the following important lemma, the proof of which directly
follows from the definition of generalized D–metric H on 2X .
Lemma 4.3. Let A, B, C ∈ 2X and a ∈ A. then for each positive
integer η > 0, there exist points b ∈ B and c ∈ C such that

(4.5) D(a, b, c, ) ≤ H(A, B, C) + η

Remark 4.2. For a given a ∈ A, there may not exist the points b ∈ B
and c ∈ C such that
D(a, b, c) ≤ H(A, B, C).
Theorem 4.1. Let (X, D) be a complete and bounded D–metric
space and let F : (X, D) → (2X , H) be a multi–valued contraction mapping.
Then F has a fixed point.
Proof. Let x ∈ X be arbitrary and consider the seuence {xn } in X
defined by

(4.6) x0 = x, xn+1 ∈ F xn , n = 0, 1, 2, . . .

Let α < 1 be the multi–valued contraction constant for F , (we may

assume α > 0). Choose η > 0, then there are points x1 ∈ F x0 and xp+1 ∈
F xp such that

D(x1 , x2 xp+1 ) ≤ H(F x0 , F x1 , F xp ) + η

(4.7) ≤ αD(x0 , x1 , xp ) + η
≤ αM + η
13 GENERALIZED D–METRIC SPACES 191

where M is the D–bound of X. Similarly for η 2 > 0, there are points

x3 ∈ F x2 and xp+2 ∈ F xp+1 such that

D(x2 , x3 , xp+2 ) ≤ H(F x1 , F x2 , F xp+1 ) + η 2

(4.8) ≤ αD(x1 , x2 , xp+1 ) + η 2
≤ α(αM + η) + η 2
= α2 M + αη + η 2 .

Proceeding in this way, by induction, we obtain

(4.9) D(xn , xn+1 , xn+p ) ≤ αn M + αn−1 η + . . . + αη n−1 + η n .

Similarly

(4.10) D(xn , xn+1 , xn+p+i ) ≤ αn M + αn−1 η + . . . + αη n−1 + η n

Letting η = α in (4.9) and (4.10), we get

(4.11) D(xn , xn+1 , xn+p ) ≤ αn M + nαn and

(4.12) D(xn , xn+1 , xn+p+i ) ≤ αn M + nαn .

Therefore, for p, t ∈ IN , we have

192 B.C. DHAGE 14

(4.13) D(xn , xn+p , xn+p+i ) ≤

≤ D(xn , xn+1 , xn+p+i ) + D(xn , xn+1 , xn+p )+
+ D(xn+1 , xn+2 , xn+p+i ) + D(xn+1 , xn+2 , xn+p )+
+ ...
≤ (αn M + nαn ) + (αn M + nαn )+
+ (αn+1 M + (n + 1)αn+1 ) + (αn+1 M + (n + 1)αn+1 )+
+ ...
n+p+i
X
n n+1 n+p+i
= 2M (α + α + ... + α )+2 iαi =
i=n
n+p+i
X n+p+i
X
= 2M ( αi + iαi ) → 0 as n → ∞.
i=n i=n

This shows that {xn } is a D–Cauchy sequence in X. Since X is a com-

plete D–metric space, there is a point x∗ ∈ X such that limn→∞ xn = x∗ .
Now every Lipschitzian multi–mapping is continuous, so the sequence {F xn }
converges to F x∗ . But xn+1 ∈ F xn for all n, n = 0, 1, 2, . . . , therefore, it
follows that x∗ ∈ F x∗ . This completes the proof.
As a consequence of Theorem 4.2., we obtain the following fixed point
theorem proved by Nadler [10] as a corollary.
Corollary 4.1 (Nadler [10]). Let (X, d) be a complete and bounded
ordinary metric space and let F : (X, d) → (2X , H) be a multi–mapping
satisfying

(4.14) H(F x, F y) ≤ αd(x, y)

for all x, y ∈ X, 0 ≤ α < 1. Then F has a fixed point.

Proof. Define a associated D–metric D and generalized D–metric
H by (1.2) and (2.3) respectively. Since (X, d) is complete and bounded,
(X, D) is a complete and bounded D–metric space (see Dhage [3]). Now
the condition (4.14) implies that F : (X, D) → (2X , H) is a multi–valued
contraction mapping. Thus all conditions of Theorem 4.2 are satisfied and
an application of it yields the desired result.
15 GENERALIZED D–METRIC SPACES 193

Corollary 4.2 (Dhage [2]). Let X be a complete and bounded

D–metric space and let f : X → X be a mapping satisfying
(4.15) D(f x, f y, f z) ≤ αD(x, y, z)
for all x, y, z ∈ X and 0 ≤ α < 1. Then f has a unique fixed point.
Proof. Letting F x = {f x} = a single–valued mapping, in Theorem
4.2, it follows that f has a fixed point. The uniqueness of fixed point follows
from condition (4.15).
Next we define two functions ρ and D on 2X × 2X × 2X by
ρ(A, B, C) = sup{D(a, b, c)|a ∈ A, b ∈ B, c ∈ C}
and
D(A, B, C) = inf{D(a, b, c)|a ∈ A, b ∈ B, c ∈ C}
Theorem 4.3. Let (X, D) be a complete and bounded D–metric
space and let F : X → 2X be a multi–map satisfying
(4.16) ρ(F x, F y, F z) ≤ αD(x, y, z)
for all x, y, x ∈ X and 0 ≤ α < 1. Then F has a unique fixed point, i.e.
there is a point x∗ ∈ X such that F x∗ = {x∗ }.
Proof. Let x ∈ X be arbitrary and consider the sequence {xn } in
X defined by (4.6). We show that {xn } is a D–Cauchy sequence. Let
x = x0 , y = x1 and z = x2 , then by (4.15), we get
D(x1 , x2 , x3 ) ≤ ρ(F x0 , F x1 , F x2 )
≤ αD(x0 , x1 , x2 ).
Proceeding in this way, by induction, we obtain
(4.17)
D(xn , xn+1 , xm ) ≤ α D(xn−1 , xn , xm−1 ) ≤ αn D(x0 , x1 , xm−n ) ≤ αn K
for all m, n = 1, 2, . . . and m > n.
Now by a Lemma 2.1. of Dhage [5], {xn } is a D–Cauchy sequence in
X. Since X is complete, there is a point x∗ ∈ X such that lim xn = x∗ .
n→∞
We show that x∗ is a fixed point of F . For, we have
D(x∗ , F x∗ , F x∗ ) = lim D(x∗n+1 , F x∗0 , F x∗ )
n→∞
≤ lim ρ(F xn , F x∗ , F x∗ )
n→∞
≤ α lim D(xn , x∗ , x∗ )
n→∞
= 0.
194 B.C. DHAGE 16

This proves that F x∗ = {xx }. To prove the uniqueness, let y ∗ (6= x∗ )

be another fixed point of F , then

D(x∗ , y ∗ , y ∗ ) ≤ ρ(F x∗ , F y ∗ , F y ∗ )
≤ αD(x∗ , y ∗ , y ∗ )

which is not possible since α < 1. Hence x∗ = y ∗ and the proof is complete.
Corollary 4.3. Let (X, d) be a complete and bounded D–metric
space and let F : X → 2X be a multi–map satisfying

(4.18) ρd (F x, F y) ≤ α d(x, y)

for all x, y ∈ X, 0 ≤ α < 1 and where ρd is a function on 2X × 2X defined

by µ̂d (A, B) = sup{d(a, b) : a ∈ A, b ∈ B}. Then F has a unique fixed point
x∗ with F x∗ = {x∗ }.
Proof. Define a D–metric D on X by (1.2). Since X is complete and
bounded w.r.t. d, so is w.r.t. D. Now for x, y, z ∈ X, we have

ρ(F x,F y, F z)
= sup{D(a, b, c)|a ∈ F x, b ∈ F y, c ∈ F z}
= sup [max{d(a, b), d(b, c), d(c, a)}|a ∈ F x, b ∈ F y,
c ∈ F z}]
= max [sup{d(a, b)|a ∈ F x, b ∈ F y},
sup{d(b, c)|b ∈ F y, c ∈ F z},
sup{d(c, a)|c ∈ F z, a ∈ F x}]
= max{ρd (F x, F y), ρd (F y, F z), ρd (F z, F x)}
≤ α max{d(x, y), d(y, z), d(z, x)}
= α D(x, y, z)(α < 1).

This shows that F satisfies the condition (4.16) of Theorem 4.3. Also
the boundedness and completness of (X, d) implies the boundedness and
completness of (X, D). Now the desired result follows by an application of
Theorem (4.3).
17 GENERALIZED D–METRIC SPACES 195

5. Contractive multi–mappings.
In this section we prove some fixed point theorems for the contractive
multi– maps, a class of multi–mappings larger than the class of multi–valued
contraction mappings and under the weaker conditions of the bopundedness
and the completness of the D–metric space under consideration. As a conse-
quence of our main results, we obtain the fixed point theorems of Smithson
[11], and the present author [4] as the special cases.
Definition 5.1. A multi–map Fb : (2X , H) → (2X , H) is said to be
contractive if
(5.1) H(Fb(A), Fb(B), Fb(C)) < H(A, B, C)
for all A, B, C ∈ 2X , H(A, B, C) 6= 0.
Theorem 5.1. Let Fb : (2X , H) → (2X , H) be a contractive multi–
map. If the sequence {An } in 2X defined by An+1 = Fb(An ) has a subse-
quence converging to a number A∗ of 2X , then A∗ is a unique fixed point of
Fb.
Proof. The proof follows by a direct application of Theorem 2.1. of
Dhage [4].
Corollary 5.1. Let X be a compact D–metric space, by Theorem
3.4, (2X , H) is also compact. As every sequence in a compact generalized
D- -metric space has a H–convergent subsequence, the conclusion of the
theorem follows by an application of Theorem 5.1.
Corollary 5.2. Let Fb : (2X , H) → (2X , H) be a multi– map satis-
fying
(5.2) H(Fb(A), Fb(B)) < H(A, B)
for all A, B ∈ 2X , H(A, B) 6= 0. If the sequence {An } in 2X defined by
An+1 = Fb(An ) has a subsequence converging to a point A∗ , then A∗ is a
unique fixed point of Fb.
Proof. Define a generalized D–metric H on 2X by (2.3). It can be
shown as in the proof of Corollary 4.1 that the condition (5.2) implies the
multi–map Fb is contractive on 2X . Also the convergence w.r.t. H implies
the convergence w.r.t. H. So the desired conclusion follows by an application
of Theorem 5.1.
Smithson [11] extended the Edelstein’s [8] fixed point theorem of con-
tractive multi–maps in ordinary metric spaces. Here now we extend the fixed
point theorem of Dhage [4] for contractive mappings in D–metric spaces to
multi–valued case. For, we need the following definition in the sequel.
196 B.C. DHAGE 18

Definition 5.2. A multi–map F : (X, D) → (2X , H) is said to be

contractive if
(5.3) H(F x, F y, F z) < D(x, y, z)
for all x, y, z ∈ X with D(x, y, z) 6= 0.
Let x = x0 ∈ X be any point. Then an orbit o(x) of a multi–map
F : (X, d) → (2X , H) at a point x = x0 is a set in X defined by
(5.4) o(x) = {xn+1 : xn+1 ∈ F xn , n = 0, 1, 2, . . .}
The orbit o(x) of a multi–map F : (X, d) → (2X , H) at some point
x ∈ X is said to be regular if and only if
(5.5) d(xn+1 , xn+2 ) ≤ d(xn , xn+1 ) and
d(xn+1 , xn+2 ) ≤ H(F xn , F xn+1 )
for all n, n = 0, 1, 2, . . .
Similarly the orbit o(x) of the multi–map F : (X, D) → (2X , H) at a
point x ∈ X, is said to be regular if and only if

D (xn+1 , xn+2 , xn+3 ) ≤ D (xn , xn+1 , xn+2 ) and
(5.6)
D (xn+1 , xn+2 , xn+3 ) ≤ H (F xn , F xn+1 , F xn+2 )
for all n, n = 0, 1, 2, . . . We note that for a given point x ∈ X, the multi–map
F : X → 2X may have several orbits, some of which may or may not be
regular.
Remark 5.1. If F : (X, D) → (2X , H) is a contractive multi–map,
then from the definition of H, it follows that if y1 ∈ F (x1 ), then there are
elements y2 ∈ F (x2 ) and y3 ∈ F (x3 ) such that
(5.7) D(y1 , y2 , y3 ) < D(x1 , x2 , x3 )

Remark 5.2. Let F : (X, D) → (2X , H) be a compact multi– map,

i.e. F (x) is compact for each x ∈ X, which is also contractive. Define the
orbit o(x) be choosing xn ∈ F (xn−1 ), xn+1 ∈ F (xn ) and xn+2 ∈ F (xn+1 )
such that
D(xn−1 , xn , xn+1 ) = D(xn−1 , F (xn−1 ), F (xn ))
= inf{D(xn−1 , y, z)|y ∈ F xn−1 , z ∈ F xn }
Since F is contractive and that
D(F xn−2 , F xn−1 , F xn ) ≤ H(F xn−2 , F xn−1 , F xn ).
the orbit o(x) is regular.
19 GENERALIZED D–METRIC SPACES 197

Theorem 5.2. Let F : (X, D) → (2X , H) be a contractive multi–

map. If there is a regular orbit o(x) of F which contains a subsequence {xnk }
converging to y0 such that xnk +1 → y1 and xnk +2 → y2 , then y0 = y1 = y2 ,
i.e. F has a fixed point.
Proof. Suppose that the orbit o(x) given by (5.4) is regular, then we
have

D(xn , xn+1 , xn+2 ) ≤ D(xn−1 , xn , xn+1 )
(5.8)
D(xn , xn+1 , xn+2 ) ≤ H(F xn−1 , F xn , F xn+1 )
for all n ∈ IN . The contractivity of the multi–map F implies that
(5.9) D(xn , xn+1 , xn+1 ) < D(xn−1 , xn , xn+1 )
for all n ∈ IN . This shows that the sequence {cn }, cn = D(xn , xn+1 ,
xn+2 ) is strictly decreasing sequence of positive real numbers which is
bounded below. Hence {cn } is convergent and converges to a point, say
c, i.e. limn→∞ cn = c. Further every subsequence of {cn } is strictly decreas-
ing and converges to the same limit point. This we have
(5.10) lim cnk = c = lim cnk +1
k→∞ k→∞

Now suppose that lim xnk = y0 , lim xnk +1 = y1 , lim xnk +2 = y2

k→∞ k→∞ k→∞
and lim xnk +3 = y3 . Then from (5.10), it follows that
k→∞

lim D (xnk , xnk +1 , xnk +2 ) = lim D (xnk +1 , xnk +2 , xnk +3 )

k→∞ k→∞

(5.11) → D(y0 , y1 , y2 ) = D(y1 , y2 , y3 )

If D(y0 , y1 , y2 ) 6= 0, then from (5.8), we obtain
D(y1 , y2 , y3 ) ≤ H (F xnk , F xnk +1 , F xnk +2 )
< lim D (xnk , xnk +1 , xnk +2 )
k→∞
= D(y0 , y1 , y2 )
which is a contradiction to (5.9). Hence D(y0 , y1 , y2 ) = 0, i.e. y0 = y1 = y2 .
But y0 = y1 = lim xnk +1 ∈ F xnk , and F is continuous in view of condition
k→∞  
(5.3), so we obtain y0 ∈ F  lim xnk 
 = F y0 . Thus F has a fixed point
k→∞
and the proof of the theorem is complete.
As a consequence of theorem 5.2, we obtain the following corollaries.
198 B.C. DHAGE 20

Corollary 5.3. Let X be a compact D–metric space and let F :

(X, D) → (2X , H) be a contractive multi–map. Then F has a fixed point.
Proof. Since a closed and bounded subset of a compact set is compact
and F (x) is closed and bounded for each x ∈ X, the desired result follows
by an application of theorem 5.2.
Corollary 5.4 (Dhage [4]). Let f : (X, D) → (X, D) be a mapping
satisfying

(5.12) D(f x, f y, f z) < D(x, y, z)

for all x, y, z ∈ X, D(x, y, z) 6= 0. If a sequence {xn } defined by xn+1 =

f xn , n = 0, 1, 2, . . . , for some x0 = x ∈ X, has a subsequence converging to
a point u ∈ X, then u is a unique fixed point of f .
Corollary 5.5 (Smithson [11]). Let f : (X, d) → (2X , H) be a
multi–map satisfying

(5.13) H(F x, F y) < d(x, y)

for all x, y ∈ X, x 6= y. If there is a regular orbit o(x) which contains a

subsequence {xnk } converging to y0 and xnk +1 → y1 , then y0 = y1 , is F has
a fixed point.
Proof. Define a D–metric D and a generalized D–metric H by (1.2)
and (2.3) respectively. Then convergence w.r.t. D and the regularity of
an orbit o(x) w.r.t. H implies the regularity w.r.t. H. Also the condition
(5.13) implies the condition (5.4.). Now the desired conclusion follows by
an application of theorem 5.2. The proof is complete.
Theorem 5.3. Let F : (X, D) → (2X , H) be a multi–map satisfying

(5.14) ρ(F x, F y, F z) < D(x, y, z)

for all x, y, z ∈ X with D(x, y, z) 6= 0. If the orbit o(x) of F , for some

x ∈ X, has a cluster point x∗ ∈ X, then it is a unique fixed point of F , such
that F x∗ = {x∗ }.
Proof. We assume that xn 6= xn+1 for all n ∈ IN , otherwise the result
follows very easily then for x = xn−1 , y = xn and z = xn+1 , from (5.14), we
obtain

D(xn , xn+1 , xn+2 ) ≤ ρ(F xn−1 , F xn , F xn+1 )

(5.15)
< D(xn−1 , xn , xn+1 )
21 GENERALIZED D–METRIC SPACES 199

for every n, n = 1, 2, . . .
Letting cn + D(xn , xn+1 , xn+2 ), we find that {cn } is a strictly decreas-
ing sequence of positive real numbers. The rest of the proof is similar to
theorems 4.3 and 5.2 in view of

H(F x, F y, F z) ≤ ρ(F x, F y, F z)

for all x, y, z ∈ X. This completes the proof.

Corollary 5.6. Let X be a compact D–metric space and F :
(X, D) → (2X , H) be a multi–map satisfying the condition (5.14). Then
F has a unique fixed point x∗ ∈ X such that {x∗ } = F x∗ .

6. Discussion.
From the foregoing discussion it is clear that the generalized D–metric
space is a natural generalization of the notion of generalized metric space
or Hausdorff metric space, and the results of this paper are the interesting
extensions of some basic results in Hausdorff metric spaces. This is a pre-
liminary work regarding the higher dimensional generalized metric spaces,
and so several interesting results concerning (2X , H) or (2X , Hm ) may be
obtained in the analogous way to a Hausdorff metric space (2X , H). These
problems constitute the open problems and the further scope for the research
work in the field of higher dimensional generalized metric spaces. Again The-
orems 4.1 and 5.1 are fundamental in the theory of fixed point theorems for
multi–valued contraction mappings in generalized D–metric spaces. These
results may be extended and generalized in several ways for different classes
of multi–valued contraction mappings, which is again a new area for research
work. Some of the results in this direction will be reported elsewhere. Also
the results prooved in this paper may have wide range of applications to
other areas of mathematics such as approximation and optimization theory
etc. which are yet to be investigated.

REFERENCES

1. C i r i c, Lj.B. – Fixed points for generalized multi–valued contraction, Math. Vesnik,

9 (24) (1972), 265–272.
2. D h a g e, B.C.– Generalized metric spaces and mappings with fixed point, Bull. Cal.
Math. Soc. 84 (4) (1992), 329–336.
3. D h a g e, B.C.– Generalized metric spaces and topological structure I, (Submitted.)
4. D h a g e, B.C.– On contractive mappings in D – metric spaces, J. Math. Phy. Sci.
(Accepted.)
200 B.C. DHAGE 22

5. D h a g e, B.C.– On some contractive mappings in D –metric spaces, (Submitted.)

6. D h a g e, B.C.– On continuity mappings in D – metric spaces, Bull. Cal. Mat. Soc.
86 (6) (1994), 27–32.
7. D u g u n d j i, J.J.– Topology, Prentice Hall of India Pvt. Ltd.
8. E d e l s t e i n, M.– On fixed and periodic points under contractive mappings, J.
Lond. Math. Soc. 37 (1962), 74–79.
9. I s t r ă ţ e s c u, V.I.– Fixed Point Theory, D.Reidel Publishing comp., 1981.
10. N a d l e r, S.B.– Multi–valued Contraction mappings, Pacific Jour. Math. 30 (2)
(1969), 475–488.
11. S m i t h s o n, E.– Fixed points of contractive multi–functions, Proc. Amer. Math.
Soc. 27 (1971), 192–194.
12. Y a n a g i, K.– On some fixed point theorems for multi–valued mappings, Pacific J.
Math. 87 (1980), 233–240.

Received: 20.IV.1996 Gurukul Colony,

Ahmedpur–413515.
Dist: Latur (M.S.)
INDIA