Nonlinear Analysis 52 (2003) 961 – 969

www.elsevier.com/locate/na

Fixed point theorems for a broad class

of multimaps
S.V.R. Naidu
Department of Applied Mathematics, Andhra University, Visakhapatnam 530003, India

Received 8 December 2000; accepted 28 January 2002

Abstract
Multimaps from a metric space into the collection of all its nonempty closed subsets are
considered and -xed point theorems for such maps have been obtained under new contractive
conditions not involving the extended Hausdor0 metric.
? 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Multimap; Fixed point; Orbital completeness; Orbital lower semi-continuity

Recently Zhong et al. [5] obtained -xed point theorems for multi-valued maps
on a metric space into the collection of all its nonempty closed bounded subsets.
One of their theorems (Theorem 2.1) is a generalization of Nadler’s theorem [2] for
multi-valued contraction mappings which in turn is a generalization of the famous
Banach contraction principle which plays an important role in several branches of
mathematics. The governing inequality for the multimap considered by Zhao et al. re-
duces, in a certain sense, to a local contraction inequality even though the coe>cient
function considered by them is not a constant. Moreover, they restrict to multimaps
with bounded images and the proof they have given is not constructive. In this pa-
per, we overcome these di>culties by developing a technical lemma which is used in
proving the Cauchy nature of a specially chosen sequence of iterates.
Throughout this paper, unless otherwise stated, (X; d) is a metric space, x0 is a
-xed element of X; N (X ) is the collection of all nonempty subsets of X; C(X ) is
the collection of all nonempty closed subsets of X; CB(X ) is the collection of all
nonempty closed bounded subsets of X; R+ is the set of all nonnegative real numbers,

E-mail address: [email protected] (S.V.R. Naidu).

0362-546X/02/$ - see front matter ? 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 6 2 - 5 4 6 X ( 0 2 ) 0 0 1 4 6 - 3
962 S.V.R. Naidu / Nonlinear Analysis 52 (2003) 961 – 969

N is the set of all positive integers, ’ : R+ → [0; 1) and for A; B ∈ CB(X ); H+ (A; B) =
sup {d(y; A) : y ∈ B}.

De nition 1. We say that ’ has property

∞ A if there exists an M ∈ (0; ∞) such that ’ is
monotonically increasing on [M; ∞); M [1−’(t)] dt=+ ∞ and sup{’(t) : r 6 t 6 M }
¡ 1 ∀r ∈ (0; M ).

De nition 2. We say that ’ has property B if sup{’(t) : r 6 t ¡ ∞} ¡ 1 ∀r ∈ (0; ∞).

Lemma 1. Let {un }∞

n=0 be a sequence in X such that

d(un ; un+1 ) 6 ’(d(x0 ; un−1 ))d(un−1 ; un ) (0.1)

for all n ∈ N. Let  = lim inf n→∞ d(x0 ; un ). Then the following statements are true.
1. If ’ is monotonically increasing on R+ and
 ∞
[1 − ’(t) dt] ¿ d(u0 ; u1 )
d(x0 ;u0 )

then {un }∞
n=0 is Cauchy.
2. If  ¿ 0 and ’ has either property A or property B, then {un }∞
n=0 is Cauchy.
3. If  = 0 and ’ has either property A or property B, then {d(un ; un+1 )}∞ n=0
converges to zero.

Proof. Let tn =d(un ; un+1 ); n =d(x0 ; un ) and sn =’(n ) (n=0; 1; 2; : : :). From inequality
(0.1) we have

tk 6 sk−1 tk−1 ∀k ∈ N:
n n n−1
Hence k=1 tk 6 k=1 sk−1 tk−1 = s0 t0 + k=1 sk tk ∀n ∈ N. Hence
n−1

tn + (1 − sk )tk 6 s0 t0 ∀n ∈ N: (0.2)
k=1

Since ’(t) ¡ 1 ∀t ∈ R+ ; sk ¡ 1 ∀k ∈ N. Hence 1− sk ¿ 0 ∀k ∈ N. Clearly tk ¿ 0 ∀k

∞
∈ N. Hence from inequality (0.2) it follows that k=1 (1 − sk )tk is convergent and that
∞

(1 − sk )tk 6 s0 t0 : (0.3)
k=1

∞
1. Suppose that ’ is monotonically increasing on R+ and d(x0 ;u0 ) [1 − ’(t) dt] ¿
d(u0 ; u1 ).
If possible, suppose that {n }∞
n=0 is not bounded. Let n1 = 0. Suppose that for a
positive integer k a nonnegative integer nk has been chosen. Let

nk+1 = min{ j ∈ N : j ¿ nk and j ¿ nk }:

 S.V.R. Naidu / Nonlinear Analysis 52 (2003) 961 – 969 963

Then {nk }∞k=1 is a strictly increasing sequence of nonnegative real numbers increasing
to + ∞. We have j 6 nk if nk 6 j ¡ nk+1 . Since ’ is monotonically increasing on
R+ , we have sj 6 snk and hence 1 − snk 6 1 − sj if nk 6 j ¡ nk+1 . We have
 n
k+1
[1 − ’(t)] dt 6 [1 − ’(nk )][nk+1 − nk ]
nk

6 [1 − snk ]d(unk ; unk+1 )

nk+1 −1

6 [1 − snk ] tj
j=nk

nk+1 −1

6 (1 − sj )tj :
j=nk

Hence we have
 ∞ ∞ 
 nk+1
[1 − ’(t)] dt = [1 − ’(t)] dt
0 k=1 nk

∞

6 (1 − sj )tj
j=0

∞

= (1 − s0 )t0 + (1 − sj )tj :
j=1

∞
Hence from inequality (0.3) we have 0 [1 − ’(t)] dt 6 t0 . This is a contradiction to
the hypothesis. Hence {n }∞ n=0 is bounded.
Let =sup{k : k =0; 1; 2; : : :} and =’(). Since 0 6 ’(t) ¡ 1 ∀t ∈ R+ ; 0 6  ¡ 1.
Since ’ is monotonically increasing on R+ ; sk 6  ∀k = 0; 1; 2; : : : . Hence from in-
equality (0.1) we have tn 6  tn−1 ∀n ∈ N. Hence tn 6 n t0 ∀n ∈ N. Hence for positive
integers m; n with m ¿ n we have
m−1
m−1 
  n
d(un ; um ) 6 tk 6  k t0 6 t0 → 0 as n → ∞:
1−
k=n k=n

Hence {un }∞ n=0 is Cauchy.

2. Suppose that  ¿ 0. Choose a positive real number  ¡ . Then there exists a posi-
tive integer N such that n ¿  ∀n ¿ N . Let  = sup{’(t): t ¿ }. Since 0 6 ’(t) ¡ 1
∀t ∈ R+ ; 0 6  6 1. From inequality (0.1) we have tn 6  tn−1 ∀n ¿ N + 1. Hence
tn 6 n−N tN ∀n ¿ N . Hence if  ¡ 1, it can be seen that {un }∞
n=0 is Cauchy. If ’ has
property B, then necessarily  ¡ 1 and we are through.
Suppose now that ’ has property A and that  = 1. We may assume that  ¡ M .
Let 0 = sup{’(t):  6 t 6 M }. Then 0 ¡ 1. Since  = 1; ∃ M0 ∈ (M; ∞) such that
964 S.V.R. Naidu / Nonlinear Analysis 52 (2003) 961 – 969

0 ¡ ’(M0 ). De-ne on [; ∞) as


’(M0 ) if  6 t 6 M0 ;
(t) =
’(t) if t ¿ M0 :
Then

∞ is monotonically increasing on [; ∞); ’(t) 6 (t) ¡ 1 ∀t ∈ [; ∞) and

[1 − (t)] dt = + ∞. From inequality (0.1) we have tn 6 (n−1 ) tn−1 ∀n ¿ N + 1.
Let sn = 
∞ n (n = 0; 1; 2; : : :). Then we have tk 6 sk−1 tk−1 ∀k = N + 1; N + 2; : : : .
( )
Hence k=N +1 (1 − sk ) tk is a convergent series of nonnegative terms bounded above
by sN tN .
If possible, suppose that {n }∞
n=0 is not bounded. Let m1 = N + 1. Suppose that for
a positive integer k a nonnegative integer mk has been chosen. Let
mk+1 = min{j ∈ N : j ¿ mk and j ¿ mk }:
Then {mk }∞k=1 is a strictly increasing sequence of nonnegative real numbers increasing
to + ∞. We have  6 j 6 mk if mk 6 j ¡ mk+1 . Since is monotonically increasing
on [; + ∞), we have sj 6 sm 
k

and hence 1 − sm k
6 1 − sj if mk 6 j ¡ mk+1 . We have
 m
k+1
[1 − (t)] dt 6 [1 − (mk )][mk+1 − mk ]
mk


6 [1 − sm k
]d(xmk ; xmk+1 )
mk+1 −1


6 [1 − sm k
] tj
j=mk

mk+1 −1

6 (1 − sj )tj :
j=mk

Hence we have
 ∞ ∞ 
 mk+1
[1 − (t)] dt = [1 − (t)] dt
m1 k=1 mk

∞

6 (1 − sj )tj
j=m1

∞

= (1 − sj )tj
j=N +1

6 sN tN ¡ + ∞:

∞
Hence  [1 − (t)] dt ¡ + ∞. This is a contradiction. Hence {n }∞ n=0 is bounded.
Let 0 = sup {k : k ¿ N }. Let 0 = sup{’(t):  6 t 6 0 }. Then 0 ¡ 1 since
sup{’(t):  6 t 6 M } ¡ 1; ’ is monotonically increasing on [M; ∞) and ’(t) ¡ 1 ∀t ∈
R+ . From inequality (0.1) we have tn 6 0 tn−1 ∀n ¿ N +1. Hence {un }∞n=0 is Cauchy.
 S.V.R. Naidu / Nonlinear Analysis 52 (2003) 961 – 969 965

3. Suppose that =0. Let E denote the set of all subsequential limits of the sequence
{n }∞ +
n=0 . Then 0 ∈ E ⊆ [0; ∞]. Since ’(t) ¡ 1 ∀t ∈ R , from inequality (0.1) it follows
that {tn }∞
n=0 is a monotonically decreasing sequence of nonnegative real numbers.
Case (i) : E ∩ (0; ∞) = .
Let ! ∈ (0; ∞). Choose a real number " ¿ !+t0 . Since E ∩(0; ∞)= ; {n ∈ N: ! 6 n
6 "} is -nite. Hence there exists an N ∈ N such that n ∈ [!; "] for any n ¿ N . Since
=0; ∃ a positive integer m ¿ N such that m ¡ !. We have m+1 6 m +tm ¡ !+t0 ¡ ".
Hence m+1 ¡ !. Arguing like this it can be seen that n ¡ ! ∀n ¿ m. Hence {n }∞ n=0
converges to zero. Hence {tn }∞ n=0 converges to zero.
Case (ii) : E ∩ (0; ∞) = .
Let # ∈ E ∩ (0; ∞). Choose real numbers a and b such that 0 ¡ a ¡ # ¡ b. Let
 = sup{’(t): a 6 t 6 b}. Then 0 6  6 1. Suppose that ’ has either property A or
property B. Then  ¡ 1. Since # is a subsequential limit of {tn }∞ n=0 , there exists a
subsequence {nk }∞ k=1 of the sequence of positive integers such that a ¡ tnk ¡ b ∀k ∈ N.
From inequality (0.1) we have
n−1
tn 6 (&l=0 ’(l )) t0 (0.4)

for all n ∈ N. Since ’(t) ¡ 1 ∀t ∈ R+ ; ’(t) 6  ∀t ∈ [a; b] and nl ∈ [a; b] for all l ∈ N,
it follows from inequality (0.4) that tnk +1 6 k t0 ∀ k ∈ N. Hence {tnk +1 }∞k=1 converges
to zero. Since {tn }∞n=0 is monotonically decreasing, it follows that it converges to
zero.

De nition 3. Let F : X → N (X ) and v0 ∈ X . By an orbit of F with respect to v0

we mean a sequence {vn }∞ n=0 in X such that vn ∈ Fvn−1 ∀n ∈ N. We say that X is
F-orbitally complete if any Cauchy subsequence of any orbit of F is convergent in
X . A real valued function f on X is said to be F-orbitally lower semi-continuous
at a cluster point z of an orbit of F w.r.t. v0 if f(z) 6 lim inf k→∞ f(vnk ) whenever
{vnk }∞
k=1 is a convergent subsequence of an orbit of F w.r.t. v0 and lim k→∞ vnk = z.

Notation
∞ 1. Let F : X → N (X ) and v0 ∈ X . ThenO(F; v0 ) stands for the set {v0 } ∪
( n=1 F n v0 ); where for a subset A of X; F(A) = x∈A Fx; F 1 v0 = Fv0 and F n+1 v0 =
F(F n v0 ) (n = 1; 2; : : :).

De nition 4. Let F : X → N (X ). We say that

∞’ has property C w.r.t. (F; x0 ) if (i) it
is monotonically increasing on R+ and (ii) 0 [1 − ’(t)] dt ¿ d(x0 ; Fx0 ).

Theorem 1. Let F : X → C(X ). Suppose that (X; d) is F-orbitally complete; the func-
tion f de6ned on X as f(x) = d(x; Fx) is F-orbitally lower semi-continuous at any
cluster point of any orbit of F w.r.t. x0 ; the function ’ has property A or B or
property C w.r.t. (F; x0 ) and that

d(y; Fy) 6 ’(d(x0 ; x)) d(x; y) (0.5)

whenever x ∈ O(F; x0 ); y ∈ Fx and x ∈ Fx. Then there exists an orbit {x n }∞

n=0 of F
w.r.t. x0 which converges to a 6xed point of F.
966 S.V.R. Naidu / Nonlinear Analysis 52 (2003) 961 – 969

Proof. Choose - ∈ (0; 1). When ’ has property C w.r.t.
∞ (F; x0 ) the selection of -
is subject to the additional condition: d(x0 ; Fx0 ) ¡ - 0 [1 − ’(t)] dt. De-ne on
R+ as (t) = (1 − -) + -’(t) ∀t ∈ R+ . Since 0 6 ’(t) ¡ 1 ∀t ∈ R+ ; is nonnegative
and ’(t) ¡ (t) ¡ 1 ∀t ∈ R+ . It is evident that has property C w.r.t. (F; x0 ) or
property A or property B according as ’ has property C w.r.t. (F; x0 ) or property A or
property B.
Choose x1 ∈ Fx0 subject to the condition that x1 = x0 if x0 ∈ Fx0 . If
’ has property
∞
C w.r.t. (F; x0 ) and x0 ∈ Fx0 ; x1 ∈ Fx0 is so chosen that d(x0 ; x1 ) ¡ 0 [1 − (t)] dt.
If x0 ∈ Fx0 , then from inequality (0.5) we have d(x1 ; Fx1 ) 6 ’(0) d(x0 ; x1 ) ¡ (0)
d(x0 ; x1 ) so that we can choose x2 ∈ Fx1 such that
d(x1 ; x2 ) ¡ (0) d(x0 ; x1 )
subject to the condition that x2 = x1 if x1 ∈ Fx1 . If x0 ∈ Fx0 , then x1 = x0 and we take
x2 = x1 .
Suppose that for an integer n ¿ 2 elements x n−2 ; x n−1 ; x n of O(F; x0 ) have been
chosen such that x n ∈ Fx n−1 ,
d(x n−1 ; x n ) 6 (d(x0 ; x n−2 )) d(x n−2 ; x n−1 )
and subject to the condition that x n = x n−1 if x n−1 ∈ Fx n−1 .
If x n−1 ∈ Fx n−1 , then from inequality (0.5) we have
d(x n ; Fx n ) 6 ’(d(x0 ; x n−1 )) d(x n−1 ; x n )

¡ (d(x0 ; x n−1 )) d(x n−1 ; x n )

so that we can choose x n+1 ∈ Fx n such that d(x n ; x n+1 ) ¡ (d(x0 ; x n−1 )) d(x n−1 ; x n )
subject to the condition that x n+1 = x n if x n ∈ Fx n . If x n−1 ∈ Fx n−1 , then x n = x n−1 and
we take x n+1 = x n .
Thus inductively we select a sequence {x n }∞ n=1 in O(F; x0 ) such that x n+1 ∈ Fx n (n =
0; 1; 2; : : :),
d(x n ; x n+1 ) 6 (d(x0 ; x n−1 )) d(x n−1 ; x n )
for all n ∈ N and for any n ∈ N; x n =x n−1 if x n−1 ∈ Fx n−1 . Let =lim inf n→∞ d(x0 ; x n ).
From Lemma 1 it follows that {x n }∞ n=0 is Cauchy if ’ has property C w.r.t. (F; x0 ) or
if  ¿ 0 and ’ has either property A or property B. Again from Lemma 1 it follows
that {d(x n ; x n+1 )}∞
n=0 converges to zero if  = 0 and ’ has either property A or prop-
erty B. Suppose that  = 0. Then there exists a subsequence {nk }∞ k=1 of the sequence
of positive integers such that {d(x0 ; x nk )}∞ k=1 converges to zero. Hence {x nk }∞
k=1 con-
∞
verges to x0 . Since {x n }n=0 is an orbit of F w.r.t. x0 , by hypothesis it follows that f
is F-orbitally lower semi-continuous at x0 . Since {d(x n ; x n+1 )}∞ n=0 converges to zero,
{f(x nk )}∞
k=1 converges to zero. From the F-orbital lower semi-continuity of f at x0 we
have f(x0 ) 6 lim inf k→∞ f(x nk )=0. Hence f(x0 )=0. Since Fx0 is closed, x0 ∈ Fx0 . By
the selection of the sequence {x n }∞ n=0 , for any n ∈ N we have x n =x n−1 if x n−1 ∈ Fx n−1 .
Hence x n = x0 ∀n ∈ N. Thus irrespective of whether ’ has property C w.r.t. (F; x0 ) or
property A or property B or whether  is positive or not we always have the conclu-
sion that {x n }∞ ∞
n=0 is Cauchy. Since X is F-orbitally complete, {x n }n=0 converges to an
element z of X .
 S.V.R. Naidu / Nonlinear Analysis 52 (2003) 961 – 969 967

From the F-orbital lower semi-continuity of f at z we have f(z) = 0. Hence

z ∈ Fz.

Corollary 1. (Theorem 2:1 of Zhong et al: [5]) Let F : X → CB(X ) be an upper semi-
continuous multi-valued mapping; . ∈ (0; 1] be a constant

∞ and h : [0; + ∞) → [0; + ∞)
be a continuous nondecreasing function such that 0 dt=(1 + h(t)) = + ∞. Suppose
that (X; d) is complete and
.
H+ (Fy; Fx) 6 1 − d(x; y) (0.6)
1 + h(d(x0 ; x))

whenever x ∈ X; y ∈ Fx and x ∈ Fx. Then F has a 6xed point.

Proof. De-ne ’ on R+ as ’(t)=1−.=(1+h(t))

∞ ∀t ∈ R+ . Then 0 6 ’(t) ¡ 1 ∀t ∈ R+ ; ’
is monotonically increasing on R+ ; 0 [1 − ’(t)] dt = + ∞ and inequality (0.5) is
true whenever x ∈ X; y ∈ Fx and x ∈ Fx. As observed by Zhong et al. [5] the upper
semi-continuity of F implies the lower semi-continuity of the function f de-ned on
X as f(x) = d(x; Fx). The corollary is now evident from Theorem 1.

Remark 1.
Theorem 1 shows that in Corollary 1 the
∞continuity of h is redundant; the
∞
condition 0 dt=(1+h(t))=+ ∞ can be relaxed as 0 dt=(1+h(t)) ¿ (1=.) d(x0 ; Fx0 )
and that inequality (0.6) can be replaced by the less stringent inequality obtained by
replacing its LHS with d(y; Fy). Moreover the values of F need not be restricted
to CB(X ); they can as well be in C(X ); provided the upper semi-continuity of F is
replaced with the lower semi-continuity of the function d(x; Fx) which is less stringent.

Remark 2. Examples 1 and 2 show that Theorem 1 is a proper generalization of

Corollary 1.

Example 1. Let X = {0} ∪ { 1n : n ∈ N} with the usual metric. De-ne T : X → X as

Tx = x=(1 + x) ∀x ∈ X . Clearly T is continuous on X . De-ne ’ : R+ → [0; 1) as


 0 if t = 0

1
’(t) = 1+2t if 0 ¡ t 6 1


1
3 if t ¿ 1:

Then ’ has property A as well as property B and

|Tx − T 2 x| = ’(|x|) |x − Tx| (0.7)

for all x ∈ X . Since ’( 1n ) = n=(n + 2) → 1 as n → + ∞; there does not exist a

monotonically increasing function : R+ → [0; 1) such that

|Tx − T 2 x| 6 (|x|) |x − Tx| (0.8)

for all x ∈ X . Hence for x0 = 0 Corollary 1 cannot be applied.

968 S.V.R. Naidu / Nonlinear Analysis 52 (2003) 961 – 969

Example 2. Let X = [0; + ∞) with the usual metric. De-ne T : X → X as Tx = x2 =(1 +

x) ∀x ∈ X . Clearly T is continuous on X . De-ne ’ on R+ as ’(t) = (t + t 2 )=(1 +
t + t 2 ) ∀t ∈ R+ . Then ’
is a strictly increasing

∞ nonnegative continuous

∞ function on
∞
R+ ; ’(t) ¡ 1 ∀t ∈ R+ ; 0 [1 − ’(t)] dt = 0 dt=(1 + t + t 2 ) = 1 du=( 34 + u2 ) =
√ 2
22=3 3 ¿ 0 = |0 − T 0| and equation (0.7) is true for all x ∈ X . If : R+ → [0; 1)
is monotonically increasing and inequality
(0.8) is true for all

∞x ∈ X ; then we should
∞
have ’(t) 6 (t) ∀t ∈ (0; ∞); so that 0 6 0 [1 − (t)] dt 6 0 [1 − ’(t)] dt ¡ + ∞.
Hence for x0 = 0 Corollary 1 cannot be applied here.

Remark 3. Theorem 2.1 of Zhong et al. [5] is based on Lemma 1.2 of their paper.
They derived it from their Lemma 1.1 which was -rst proved by Zhong [4]. It was
observed by the referee that Lemma 1.2 of [5] is not new and that it is equivalent to
Theorem 2.1 of Ray and Anita [3].

∞
Remark 4. Example 3 shows that in Theorem 1 the condition 0 [1 − ’(t)] dt ¿ d(x0 ;
Fx0 ) cannot be dropped when ’ is monotonically increasing on R+ . We note that when
+

’∞is monotonically increasing on R ; property A as well as property B implies that
0
[1 − ’(t)] dt = + ∞.
n
Example 3. Let an = k=1 k1 and bn = an − 1 (n ∈ N). Let X = {an : n ∈ N} with the
usual metric. Let x0 = a1 . De-ne T : X → X as Tan = an+1 ∀n ∈ N. De-ne ’ on R+ as
n+1 bn+1 − t n+2 t − bn
’(t) = +
n+2 bn+1 − bn n+3 bn+1 − bn
if bn 6 t 6 bn+1 (n ∈ N):
Then
’ is a strictly increasing nonnegative continuous function on R+ ; ’(t) ¡ 1 ∀t ∈
∞
R+ ; 0 [1 − ’(t)] dt = 11 1
24 ¡ 2 = |x0 − Tx0 | and
|Tx − T 2 x| = ’(|x0 − x|) |x − Tx|
for all x ∈ X . Clearly T has no -xed point in X .

A perusal of the proof of Theorem 1 yields the following.

Theorem 2. Let F : X → C(X ). Suppose that (X; d) is F-orbitally complete; the func-
tion ’ has property C w.r.t. (F; x0 ); and that inequality (0.5) is true whenever
x ∈ O(F; x0 ); y ∈ Fx and x ∈ Fx. Then there exists an orbit {x n }∞
n=0 of F w.r.t. x0
which converges to an element z of X and z is a 6xed point of F if and only if the
function f de6ned on X as f(x) = d(x; Fx) is F-orbitally lower semi-continuous at z.

Remark 5. Theorem 2 is a partial generalization of the theorem of Hicks and

Rhoades [1].

The author expresses his heart felt thanks to Prof. Jiang Zhu for providing the
electronic version of the paper cited in [5] and to Prof. B.E. Rhoades for providing
the reprints of many of his papers including the one cited in [1].
 S.V.R. Naidu / Nonlinear Analysis 52 (2003) 961 – 969 969

References

[1] T.L. Hicks, B.E. Rhoades, A Banach type -xed point theorem, Math. Japon. 24 (1979) 327–330.
[2] S.B. Nadler Jr., Multivalued contraction mappings, Paci-c J. Math. 30 (1969) 475–487.
[3] W.O. Ray, A.M. Walker, Mapping theorems for Gâteaux di0erentiable and accretive operators, Nonlinear
Anal. 6 (5) (1992) 423–433.
[4] C.K. Zhong, A generalization of Ekeland’s variational principle and applications to the study of the
relation between the weak P.S. condition and coercivity, Nonlinear Anal. 29 (1997) 1421–1431.
[5] C.K. Zhong, J. Zhu, P.H. Zhao, An extension of multi-valued contraction mappings and -xed points,
Proc. Amer. Math. Soc. 128 (2000) 2439–2444.