Nonlinear Analysis, Theory, Melhods & Applrrurrons. Vol. 5, No 3, pp. 287-292 0362.546X:X1.‘030287A,6 SOS.

OO/O
0 Pergamon Press Ltd. 1981. Printed in Great Britain

FIXED POINT THEOREMS IN COMPLETE METRIC SPACES

G.EMMANUELE
Seminar of Mathematics, University of Catania, Catania. Italy

(Rrcriued 4 February 1980; in revised form 30 June 1980)

Key wmls: Generalized contractions, fixed point theorems. convergence of sequence of iterates.

1. INTRODUCTION

THE STUDY in fixed point theory has generally developed in three main directions: generalization
of conditions which ensure existence, and, if possible, uniqueness, of fixed points; investigation
of the character of the sequence of iterates (T”x)~~~, where T: X + X, X a complete metric
space, is the map under consideration; study of the topoligical properties of the set of fixed points,
whenever T has more than one fixed point. This note treats only some aspects of the first and
second question, along a line followed by many other authors. We mention, in particular, De
Blasi [3], Kannan [4], Opial [7], Reich [S-lo].
More precisely we consider maps T: X -+ X, which satisfy conditions of the type

d(Tx, TY) d 44x, Y)) + @(TX, 4) + x@(Ty, ~1) for each x, Y E X,

and for these mappings we prove, under suitable hypotheses, existence and uniqueness of fixed
points.
The paper consists of four sections. In Section 1 we prove a fundamental lemma. In Section 2
we use this in order to establish the main fixed point theorems. In Section 3 we obtain some well-
known results as corollaries of our theorems. Further results are presented in Section 4.

1. Through all the paper, X denotes a complete metric space and 7: T:X + X, an asymptotically
regular mapping (see 121); l.e, a function satisfying lim d(T”x, T”+‘x) = 0 for each x E X.

Furthermore, we suppose that there exist three functionsncp, tj, x, from [0, + co[ into [0, + rx;[,
which satisfy the assumptions:

(I,) q(r) < Y ifr > 0,

(I,) there exists lim q(r) d cp(r) for each ? E [0, + co[
*+F+
(I,) $(O) = Z(O) = 0.
Moreover, we suppose that 7: cp, $, x satisfy the inequality

(I,) d(Tx, TY) d &4x, ~1) + W(Tx, 4) + x@(Ty, Y)) for each x, Y E X.

LEMMA. Under the above assumptions on X and T and if, in addition, Ic/ and x are continuous
at r = 0, then, for each x E X, there exists z E X such that (Tnx}~zo converges to z.
287
288 G. EMMANUELE

Proof: Suppose that there exists XE X such that the sequence of iterates is not a Cauchy
sequence. Then, following [9] there exist E > 0, {m(j)};=,, {n(j)}J?E, which satisfy the conditions

m(j) > n(j) foreach,jEN (1.1)

lim n(j) = + cc (1.2)

d(Tm(j)x, T”(h) 3 & (1.3)

d(Tm(j)- lx, 7-n")x) < E. (1.4)
Then, we have

& 6 d(T”‘j’ X, Tn(j)x) < d(Tm(jjx, Tm(j) - lx) + d(Tm(j) - lx, Tn(Ax) < s + d(TWAx, 7-W) - lx)

which implies

lim d(Tm(j)x, Tn(jjx) = E. (1.5)

j
On the other hand

d(Tm(jJx, p(jJx) < d(pjJx, y(j) + lx) + d(yOx, y(j) + lx) + d(Tm(j) + lx,

T”(j) + ‘x) < d( T”(j)x, T”(j) + lx) + d( Tn(jjx, T”(j) + lx) + cp(d( yUx, TWX))

+ $(d(T”(j)+ *x, T"(~)x)) + X(d(T"(j)+lx, y(j)x))

that is

d(Tm(j)x, Tn(j)x) - cp(d(Tm(jJx,Tn(j)x)) d d(Tm(jjx, -p(j)+ lx) + d(Tn(jjx,

T”(j) + lx) + $(d( T”(j) + lx, Tm(j)x)) + X(d( T”(j) + lx, TWX))

and, letting ,j -+ + co

E - lim cp(d(T”%, Tn(Jx)) < 0.

Hence, by (1.3) and (1.5) it follows that E = 0, a contradiction. Since X is a complete metric space,
the proof is complete.

2. In this section we shall prove two fixed point theorems: the lirst for noncontinuous, the
second for continuous mappings.

THEOREM 1. Let X, 7; cp, $, x be as in the Lemma. Furthermore, we suppose that x(r) < r if r > 0.
Then T has a unique fixed point.

Proof: Uniqueness is obvious by virtue of hypotheses (I,), (I,) and (I,). So let us show existence.
From the Lemma there is a z E X such that T”x + z, as II -+ + co, for each x E X. Since

d(z, Tz) < d(z, T”x) + d(T”x, T”+‘x) + d(T”+ lx, Tz) < d(z, T”x) + d(T”x, T”+‘x)

+ cp(d(T”x, z,) + X(d(Tz, z)) + t,b(d(T”x, T”+ lx))

 Fixed point theorems in complete metric spaces 289

we have

d(z, TX) - x(d(z, Tz)) < 2d(z, T”x) + d(T”x, T”+‘x) + i++(T”x, T”+lx))
and, letting n -+ + co, we have

0 d d(z, Tz) - #(z, Tz)) < 0;

thus, z = Tz follows.

Remark 1. Obviously, the role of the functions $ and x can be reversed.

For continuous functions, we can prove the following result.

THEOREM2. Let X, IT;q, $,x be as in the Lemma. If, in addition, T is continuous, then it has a
unique fixed point.

Proof: Uniqueness follows as in Theorem 1. Let z E X be such that T”x -+ z, as n + + co.

From continuity of T we obtain

lim T"+ lx = Tz

and, since T is asymptotically regular, we have z = Tz.

Remark 2. If X is a closed, non-void, subset of a Banach space B, Theorem 2 is still true under the
assumption: “T is a strongly-weakly continuous mapping from X into X”.

Remark 3. From the Lemma, Theorem 1 and Theorem 2, there follows that the sequence of
iterates { T”x>,“_o converges to the unique fixed point of ‘I for each x E X.

Remark 4. In [3] the author puts the following question: “Let T be an asymptotically regular,
continuous mapping and let S be a non-void, weakly closed subset of a Hilbert space X. If
IT;T:S + S, satisfies the condition

llTx- Tylld pllx- Y II+ q(IITX - xl/ + //TY - YI(), P’ + q2= 4 p, q Z 0,

for each x, y E S, does the sequence { T”x),“,, converge to the unique fixed point of 17:if it exists?”
By using Theorem 2, with q(r) = pr, $(r) = X(r) = qr for each r E [0, + co[, we answer, posi-
tively, to this question. Furthermore, if we use Theorem 1, we can dispense with the continuity
of lY

3. In this section we obtain some well-known results, as corollaries of our previous theorems.

COROLLARY 1 [6]. Let T be a function from X into X. We suppose that there exists a map X f:
[0, + co[ -+ [0, + co[, continuous from the right for each r E [0, + co[, such that

d(Tx, Ty) d f(d(x, y)), for each x, y E X.

If f(r) < r, for r > 0, then the sequence { T”x},“, o converges to the unique fixed point of r
290 G.EMMANUELE

Proof: We use Theorem 2. In this theorem let q(r) = f(r), Ii/(r) = X(r) = 0 be, for each r E [0,
+ a[. Moreover, if v E N exists such that TV+rx = I”+, for each x E X, T is a asymptotically
regular mapping. On the contrary, one has, for each x E X such that this v doesn’t exist,
d( T”+ lx, T”x) d f(d( T”x, T”- ‘x)) < d( T”x, T”- ‘x) for each II E N.
Put d = lim d(T”+l x, T”x), we have d d f(d) < d and so d = 0.
”

COROLLARY 2 [ 11. Let X, T be as in Corollary 1. Suppose that 1; from [0, + cc[ into [0, + co[
is a non-decreasing function, continuous from the right such that
d(Tx, Ty) 6 f(d(x, y)), for each x, y E X.
If f(r) < r (r > 0) and if X is bounded, then T has a unique fixed point.

CORALLARY 3 [3].Let T be an asymptotically regular and continuous function, such that T

maps S into S, with S a non-void weakly closed subset of a Hilbert space H. Also, we suppose
that T satisfies
\/TX - TyIj < IITx - XII + [[Ty - y[l, foreachx,yEX.

Then, for each x E S, the sequence {T”x},“,O converges to the unique fixed point of T.

Proo& In Theorem 2, we take cp(r) = 0, Ii/(r) = x (r) = r foreach r 2 0.

C~ROLLARY~ [3].LetT be, T:S + S, S a non-void weakly closed subset of a Hilbert space H,
an asymptotically regular and continuous mapping, such that

\lTx - TYII G PIJX - YII + 4l)Tx - XII + IITY - YII), foreachx,yEX,

with p* + q2 < 1. Then, the thesis of Corollary 3 is true.

Proof. In Theorem 1, we put cp(r) = pr, $(r) = x(r) = qr, for each r 2 0. We observe that, by
using Theorem 1, we can dispense with the continuity of IT:

COROLLARY 5 [8].Let (X, d) be a complete metric space and let u, b, c be nonnegative numbers,
with a + b + c < 1. Furthermore, suppose that T: X --t X satisfies
d( Tx, Ty) < ad(x, y) + bd( Tx, x) + cd( Ty, y), for each x, y E X.
Then, T has a unique fixed point.

Proof. We can take, in Theorem 1, q(r) = ar, t&r) = br, X(r) = cr, for each r c [0, + m[.
Since, for each n EN, we have
d(T”+ lx, T”x) < ad(T”x, T”- ‘x) + bd(T”+ lx, T”x) + cd(T”x, T”- ‘x), x E X,

if follows that

d(T”+’ ’ d( Tx, x), XEX.

 Fixed point theorems in complete metric spaces 291

This fact implies that T is asymptotically regular, being a + b + c =c 1.

Remark 5. Under the assumptions of Corollary 5, we have, by virtue of Remark 3, that the
sequence {T”x},“,~ converges to the unique fixed point of I:

COROLLARY 6 [5]. Let 7: T:X + X, be a mapping satisfying the condition

d(Tx, ,Ty) < (x, y) - A(&, y)), for each x, y E X
where A, A: [0, + co[ + [0, + co[, is a continuous function such that A(r) > 0, if r > 0. Then,
for each x E X, the sequence {T”x},“, o converges to the unique fixed point of T

Proof: Asymptotical regularity follows as in Corollary 1. By putting, in Theorem 1, 4(r) =

I - A(r), $(t) = X(r) = 0 for each r E [0, + co[ one obtains the thesis. We observe that we can
dispense with the continuity of A. In fact it suffices that A is a right continuous or a monotonically
nondecreasing function.

4. In this section, we consider mappings T, T: X + X, satisfying all assumptions of n. 1, with the

exception of (I,), which is replaced with

KJ W’k TY) < &4x, Y)) + P(& Y)) Ic/(Wx,4) + &4x, Y)) XWY, Y)) foreachx, Y E X.
Here p,p:[O, + a[-[0, + co[, is a function such that
(I,) lim p(s) < + co, for each SE [0, + co[
s-r;+
and 4, q:[O, + co[+[O, + co[, is a function such that
(I,) lim q(s) < 1 and lim q(s) < + co, for each S E [0, + co[.
s-o+ s-+s+

Before proving the announced fixed point theorems, we observe that in this case the Lemma is
still true if we suppose that II/,x are continuous functions as r -+ 0 + . Then, with standard argu-
ments, we prove

THEOREM3. If X, T, q, II/,x, p, q satisfy all previous assumptions and, in addition, X(r) d r for
each r E [0, + CQ[,then T has a unique fixed point. Moreover, the sequence { T”x},“,~ converges
to the unique fixed point of 7:

This last Theorem 3 generalizes the following result of [lo].

“Let (X, d) be a complete metric space. If IT:T: X + X, satisfies

&% TY) d a(+, Y))d(x, Y) + (b(x,y)) d(Tx, 9 + c(d(x, Y))d(Ty, Y), (4.1)
where x, y E X, x # y, and a, b, c are monotonically decreasing functions from [0, + co[ into
[0, l[ such that a(s) + b(s) + c(s) < 1, the T has a unique fixed point”.
To this end, we observe, first of all, that T is an asymptotically regular mapping (see [lo]).
Moreover, we can take

0
444 = ’ = O; $(r) = X(r) = r foreachrE[O, +co[; p(s) = 1
a(r)r r # 0
292 G. EMMANUELE

for each s E [0, + cc [; since we may assume c(s) < i (see [9])
s=o
q(s) = +
{ c(s) s # 0.
By using Theorem 3 we obtain that T has a unique fixed point in X and, furthermore, that the
sequence of iterates coverges to this unique fixed point.
Finally, if T is a continuous mapping satisfying assumptions as in the Lemma we may prove a
result as Theorem 2, without assumptions lim p(s) < + co and lim q(s) < 1.
s-o+ s-o+

REFERENCES

1. BROWDER F. E., On the convergence of successive approximation for non linear functional equations, Indug. math.
XXX(l), 27-35 (1968).
2. BROWDER F. E. & PETRTSHYNW. V., The solution by iteration of non-linear functional equations in Banach spaces,
Bull. Am. math. Sot. 72, 571-575 (1966).
3. DE BLASI F. S., Fixed points for Kannan’s mappings in Hilbert spaces, Boll. Un. mat. Ifal. (4)9, 818-823 (1974).
4. KANNAN R., Some results on fixed points, Bull. Calcutra math. Sot. 60, 71-76 (1968).
5. KRASNOSEL’SKUM. A. & STECENKO V. JA., About the theory of equations with concave operators (in Russian),
Sib. mat. Zh. lO(3) 565-572 (1969).
6. MUKHERJEA A., Contractions and completely continuous mappings, Nonlinear Analysis, TMA l(3), 235-247 (1977).
7. OPIAL Z., Weak convergence of the sequence of successive approximation for non expansive mappings, Bull. Am.
math. Sot. 13, 591-597 (1967).
8. REICH S., Some remarks concerning contraction mappings, Can. math. Bull. 14, 121-124 (1971).
9. REICH S., Fixed point of contractive functions, Boll. Un. mat. Ital. (4)5, 2642 (1972).
10. REICH S., Kannan’s fixed point theorem, Boll. Un. mat. Ital. (4)4, I-11 (1971).