Fixed Point Theory, Volume 4, No.

1, 2003, 79-89
http://www.math.ubbcluj.ro/∼ nodeacj/journal.htm

FIXED POINTS FOR MAPPINGS SATISFYING

CYCLICAL CONTRACTIVE CONDITIONS

W. A. KIRK∗ , P. S. SRINIVASAN∗∗ , P. VEERAMANI∗∗

∗
Department of Mathematics, University of Iowa
Iowa City, IA 52242-1419 USA
E-mail address: [email protected]
∗∗
Department of Mathematics, Indian Institute of Technology
Madras, Chennai 600 036 India
E-mail address: [email protected]

Abstract. This paper deals with mappings of the type f : Ai → Ai+1 , i = 1, 2, · · ·, p + 1,

with Ap+1 = Ap , where the contractive assumptions are restricted to pairs (x, y) ∈ Ai ×Ai+1 .
Extensions of Banach’s theorem are considered as well as an extension of Caristi’s theorem.
The paper concludes with a result about nonexpansive mappings in a Banach space setting.
Keywords: fixed points, contractive conditions, Caristi’s theorem, nonexpansive mappings
AMS Subject Classification: 54H25, 47H10

1. Introduction

It is well known and easy to prove that if X is a complete metric space and
if F : X → X is continuous and satisfies

d F (x) , F 2 (x) ≤ kd (x, F (x))

∀x ∈ X, where k ∈ (0, 1) ,
then F has a fixed point in X. The condition on F ensures that {F n (x)} is a
Cauchy sequence for each x ∈ X, and continuity does the rest.

On the other hand, suppose there exist two nonempty closed subsets A and
B of X such that the mapping F : A ∪ B → A ∪ B satisfies:
(1) F (A) ⊆ B and F (B) ⊆ A.
(2) d (F (x) , F (y)) ≤ kd (x, y) ∀x ∈ A and y ∈ B, where k ∈ (0, 1) .
79
80 W. A. KIRK, P. S. SRINIVASAN, P. VEERAMANI

Then it readily follows that for any x ∈ A ∪ B

d F (x) , F 2 (x) ≤ kd (x, F (x))

and this again implies that {F n (x)} is a Cauchy sequence. Consequently

{F n (x)} converges to some point z ∈ X. However in view of (2) an infinite
number of terms of the sequence {F n (x)} lie in A and an infinite number
of terms lie in B. Therefore z ∈ A ∩ B, so A ∩ B 6= ∅. Now (1) implies
F : A ∩ B → A ∩ B and (2) implies that F restricted to A ∩ B is a contraction
mapping. Since Banach’s contraction mapping principle applies to F on A ∩ B
we have the following result.

Theorem 1.1. Let A and B be two non-empty closed subsets of a complete

metric space X, and suppose F : X → X satisfies (1) and (2) above. Then F
has a unique fixed point in A ∩ B.

An interesting feature about the above observation is that continuity of F

is no longer needed. Indeed, simple examples can be constructed showing that
discontinuous mappings can satisfy all the assumptions. Also, it is possible to
reformulate this result as a common fixed point theorem for two mappings.

Corollary 1.2. Let A and B be two non-empty closed subsets of a complete

metric space X. Let f : A → B and g : B → A be two functions such that

(*) d (f (x) , g (y)) ≤ kd (x, y) ∀x ∈ A and y ∈ B,

where k ∈ (0, 1) . Then there exists a unique x0 ∈ A ∩ B such that

f (x0 ) = g (x0 ) = x0 .

Proof. Apply Theorem 1.1 to the mapping F : A ∪ B → A ∪ B defined by

setting
(
f (x) if x ∈ A;
F (x) =
g (x) if x ∈ B.
Observe that F is well defined since (*) implies f (x) = g (x) if x ∈ A ∩ B. 

Obviously the reasoning of Theorem 1.1 can be extended to a collection of

finite sets.
 FIXED POINTS FOR MAPPINGS 81

Theorem 1.3. Let {Ai }pi=1 be nonempty closed subsets of a complete metric
space, and suppose F : ∪pi=1 Ai → ∪pi=1 Ai satisfies the following conditions
(where Ap+1 = A1 ):

(1) F (Ai ) ⊆ Ai+1 for 1 ≤ i ≤ p;

(2) ∃ k ∈ (0, 1) such that d (F (x) , F (y)) ≤ kd (x, y) ∀x ∈ Ai , y ∈ Ai+1 for

1 ≤ i ≤ p.

Then F has a unique fixed point.

Proof. One only need to observe that given x ∈ ∪pi=1 Ai , infinitely many terms
of the Cauchy sequence {F n (x)} lie in each Ai . Thus ∩pi=1 Ai 6= ∅, and the
restriction of F to this intersection is a contraction mapping. 

The objective of this note is to extend the above reasoning to more general
classes of mappings.

2. Contractive extensions

We first take up the question of whether Edelstein’s classical result for

contractive mappings can be similarly extended. Recall that a mapping
F : M → M is said to be contractive if d (F (x) , F (y)) < d (x, y) whenever
x, y ∈ M, x 6= y. Edelstein’s result ([3]) asserts that a contractive mapping
defined on a complete metric space has a unique fixed point if some Picard
sequence {F n (x)} , x ∈ M, has a convergent subsequence. This result extends
as follows. (Again, no continuity assumption is needed.)

Theorem 2.1. Let {Ai }pi=1 be nonempty closed subsets of a complete metric
space, at least one of which is compact, and suppose F : ∪pi=1 Ai → ∪pi=1 Ai
satisfies the following conditions (where Ap+1 = A1 ):

(1) F (Ai ) ⊆ Ai+1 for 1 ≤ i ≤ p;

(2) d (F (x) , F (y)) < d (x, y) whenever x ∈ Ai , y ∈ Ai+1 and x 6= y, (1 ≤ i ≤

p).

Then F has a unique fixed point.

82 W. A. KIRK, P. S. SRINIVASAN, P. VEERAMANI

Proof. Assume A1 is compact, and let

d = dist (A1 , Ap ) := inf {d (x, y) : x ∈ A1 , y ∈ Ap } .
By compactness there exists x0 ∈ A1 and a sequence {un } ⊂ Ap such that
limn d (x0 , un ) = d. Assume d > 0. Then
d F p+1 (x0 ) , F p+1 (un ) < · · · < d (F (x0 ) , F (un )) < d (x0 , un ) .


(2.1)
∞
Since the sequence F p+1 (un ) n=1 ⊂ A1 and A1 is compact, this sequence


has a subsequence that converges to some z ∈ A1 . By (2.1) and continuity of

the distance function it must be the case that
d z, F p+1 (x0 ) ≤ d.

However this implies

d F p−1 (z) , F 2p (x0 ) < d

and since F p−1 (z) ∈ Ap and F 2p (x0 ) ∈ A1 we have a contradiction. We

conclude therefore that d = 0 and A1 ∩ Ap 6= ∅. Thus by (1), A1 ∩ A2 6= ∅.
We now consider the sets A01 = A1 ∩ A2 , A02 = A2 ∩ A3 , · · ·, A0P = AP ∩ A1 .
In view of condition (1) these sets are all nonempty (and closed) and A01 is
compact. Thus conditions (1) and (2) of the theorem hold for F and the family
{A0i }pi=1 , and by repeating the argument just given we conclude
A01 ∩ A0p 6= ∅.
This in turn implies A1 ∩ A2 ∩ A3 6= ∅. Continuing step-by-step we conclude
A := ∩pi=1 Ai 6= ∅.
Since A is compact and the restriction of F to A is contractive, we conclude
that F has a unique fixed point in A. Uniqueness follows from the fact that
any fixed point of F necessarily lies in A by condition (1). 

We now take up the question of whether condition (2) of Theorem 1.3 can
be replaced by contractive conditions which typically arise in extensions of
Banach’s theorem. The answer is affirmative, but the arguments now become
a little more subtle. We begin with a condition introduced by Geraghty [4].
Let S denote the class of those function α : R+ → [0, 1) that satisfy the simple
condition:
α (tn ) → 1 ⇒ tn → 0.
 FIXED POINTS FOR MAPPINGS 83

Theorem 2.2. ([4]) Let X be a complete metric space, let f : X → X and

suppose there exists α ∈ S such that

d (f (x) , f (y)) ≤ α (d (x, y)) d (x, y) ∀x, y ∈ X.

Then f has a unique fixed point z ∈ X, and {f n (x)} converges to z for each
x ∈ X.

Theorem 2.3. Let {Ai }pi=1 be nonempty closed subsets of a complete metric
space, let α ∈ S, and suppose f : ∪pi=1 Ai → ∪pi=1 Ai satisfies the following
conditions (where Ap+1 = A1 )

(1) f (Ai ) ⊆ Ai+1 for 1 ≤ i ≤ p;

(2) d (f (x) , f (y)) ≤ α (d (x, y)) d (x, y) ∀x ∈ Ai , y ∈ Ai+1 for 1 ≤ i ≤ p.

Then f has a unique fixed point.

Proof. The strategy is to prove that ∩pi=1 Ai 6= ∅ and apply Geraghty’s theorem
to f restricted to ∩pi=1 Ai 6= ∅. For convenience of notation, if j > p, define
Aj = Ai where i = j mod p and 1 ≤ i ≤ p.
The argument we give a slight modification of the proof of Geraghty’s the-
orem given in [6]. Let x0 ∈ A1 and let xn = f n (x0 ) , n = 1, 2, · · ·.

Step 1.limn→∞ d (xn , xn+1 ) = 0.

Proof. Condition (2) implies that {d (xn , xn+1 )} is monotone decreasing

and bounded below. Thus limn→∞ d (xn , xn+1 ) = r ≥ 0. Assume r > 0. Then
again by condition (2)
d (xn+1 , xn+2 )
≤ α (d (xn , xn+1 )) , n = 1, 2, · · ·.
d (xn , xn+1 )
Letting n → ∞ we see that α (d (xn , xn+1 )) → 1. But since α ∈ S, this in turn
implies d (xn , xn+1 ) → 0. Thus it must be the case that r = 0.

Step 2. {xn } is a Cauchy sequence.

Proof. Suppose there exists ρ > 0 such that given any N ∈ N there exist
n > m ≥ N with n − m = 1 mod p such that d (xn , xm ) ≥ ρ > 0. By the
triangle inequality

d (xn , xm ) ≤ d (xn , xn+1 ) + d (xn+1 , xm+1 ) + d (xm+1 , xm ) .

84 W. A. KIRK, P. S. SRINIVASAN, P. VEERAMANI

Since n − m = 1 mod p xm and xn lie in different adjacently labelled sets Ai

and Ai+1 for some 1 ≤ i ≤ p, so by the contractive condition

[1 − α (d (xn , xm ))] ρ < [1 − α (d (xn , xm ))] d (xn , xm )

≤ d (xn , xn+1 ) + d (xm , xm+1 ) .

Letting n, m → ∞ with n − m = 1 mod p we conclude α (d (xn , xm )) → 1. But

since α ∈ S this implies d (xn , xm ) → 0, which is a contradiction. Therefore,
given any ε > 0 there exists N ∈ N such that if n, m ≥ N and n − m = 1
mod p, d (xn , xm ) ≤ ε/p. By Step 1 it is possible to choose N1 ∈ N so that
d (xn , xn+1 ) ≤ ε/p if n ≥ N1 . Now let n, m ≥ max {N, N1 } with m > n.
Then there exists k ∈ {1, 2, · · ·, p} such that and n − m = k mod p. Thus
n − m + j = 1 mod p, where j = p − k + 1, so

d (xn , xm ) ≤ d (xm , xn+j ) + d (xn+j , xn+j−1 ) + · · · + d (xn+1 , xn ) ≤ ε.

This proves that {xn } is a Cauchy sequence, and consequently that ∩pi=1 Ai 6=
∅. By Geraghty’s theorem f has a unique fixed point in ∩pi=1 Ai , and by con-
dition (1) any fixed point of f must lie in this intersection. 

Next we look at the well known Boyd-Wong condition [1]. A statement of

the Boyd-Wong theorem can be obtained by taking Ai = Aj for all i, j in the
following.

Theorem 2.4. Let {Ai }pi=1 be nonempty closed subsets of a complete metric
space and suppose f : ∪pi=1 Ai → ∪pi=1 Ai satisfies the following conditions
(where Ap+1 = A1 ):

(1) f (Ai ) ⊆ Ai+1 for 1 ≤ i ≤ p;

(2) d (f (x) , f (y)) ≤ ψ (d (x, y)) ∀x ∈ Ai , y ∈ Ai+1 for 1 ≤ i ≤ p, where

ψ : R+ → [0, ∞) is upper semi-continuous from the right and satisfies 0 ≤
ψ (t) < t for t > 0.

Then f has a unique fixed point.

Proof. We follow the same strategy as in the preceding argument. For j > p
define Aj = Ai if j = i mod p. Let x0 ∈ A1 and let xn = f n (x0 ) , n = 1, 2, · · ·.

Step 1. limn→∞ d (xn , xn+1 ) = 0.

 FIXED POINTS FOR MAPPINGS 85

Proof. By (2) the sequence {d (xn , xn+1 )} is monotone decreasing and

bounded below, so limn→∞ d (xn xn+1 ) = r ≥ 0. Thus

d (xn+1 , xn+2 ) ≤ ψ (d (xn , xn+1 )) ⇒ r ≤ ψ (r) ⇒ r = 0.

Step 2. {xn } is a Cauchy sequence.

Proof. Suppose not. Then there exists ε > 0 such that for any k ∈ N, there
exists mk > nk ≥ k such that

d (xmk , xnk ) ≥ ε.

Furthermore it may be assumed that for each k, mk is chosen to be the smallest

number greater that nk for which the above is true. In view of Step 1,

lim d (xmk , xmk −1 ) = 0.

k→∞

Since

ε ≤ d (xmk , xnk ) ≤ d (xmk , xmk −1 ) + d (xmk −1 , xnk )

≤ d (xmk , xmk −1 ) + ε

and we conclude limk→∞ d (xmk , xnk ) = ε. Also, since

d (xmk , xnk )−d (xmk +1, xmk ) ≤ d (xmk +1 , xnk ) ≤ d (xmk +1 , xmk )+d (xmk , xnk ) ,

we conclude also that limk→∞ d (xmk +1 , xnk ) = ε. There exists j, 0 ≤ j ≤ p−1,

such that mk − nk + j = 1 mod p for infinitely many k. If j = 0 we have, for
such k,

d (xmk , xnk ) ≤ d (xmk , xmk +1 ) + d (xmk +1 , xnk +1 ) + d (xnk +1 , xnk )

≤ d (xmk , xmk +1 ) + ψ (d (xmk , xnk )) + d (xnk +1 , xnk ) .

Letting k → ∞ gives ε ≤ ψ (ε) , which contradicts ψ (t) < t for t > 0. The
case j 6= 0 similar. 

The preceding two theorems are just examples to illustrate the methodology.
Other extensions of Banach’s theorem can be similarly recast. It would be nice
to find a more abstract formulation that would unify all of these results. Of
course it would even be nicer to have applications.
86 W. A. KIRK, P. S. SRINIVASAN, P. VEERAMANI

3. Caristi’s theorem

The preceding ideas lead also to an analogous extension of Caristi’s theorem

[2], which is simply the following theorem when Ai = Aj for all i, j.

Theorem 3.1. Let A1 , A2 , · · ·, Ap , Ap+1 = A1 be nonempty closed subsets of

a complete metric space X, and suppose f : X → X satisfies the following
conditions.

(1) f (Ai ) ⊆ Ai+1 for 1 ≤ i ≤ p.

(2) d (x, f (x)) ≤ ϕi (x) − ϕi+1 (f (x)) ∀x ∈ Ai (1 ≤ i ≤ p), where each ϕi :

Ai → R is lower semicontinuous and bounded below.

Then f has a fixed point.

Proof. Let x1 ∈ A1 and xn = f n−1 (x1 ) . By condition (2)

ϕ1 (x1 ) ≥ · · · ≥ ϕn (xn ) ≥ · · ·, n = 1, 2, · · ·,

where of course ϕi = ϕj if i = j mod p. Therefore limi→∞ ϕi (xi ) = r. Now fix

xn ∈ An , let k ∈ N, and let m > n. Then

d (xn , xm ) ≤ d (xn , f (xn )) + d (f (xn ) , f (xn+1 )) + · · · + d (f (xm−2 ) , xm )

≤ ϕn (xn ) − ϕn+1 (f (xn )) + ϕn+1 (f (xn )) − ϕn+2 (f (xn+1 ))
+ · · · + ϕm−1 (f (xm−2 )) − ϕm (xm )
= ϕn (xn ) − ϕm (xm ) .

This proves that {xn } is a Cauchy sequence, and in turn that A := ∩pi=1 Ai 6= ∅.
We now have the following situation. f : A → A and

d (x, f (x)) ≤ min [ϕi (x) − ϕi+1 (f (x))] for all x ∈ A.

1≤i≤p

Thus

pd (x, f (x)) ≤ ϕ1 (x) − ϕ2 (f (x)) + ϕ2 (x) − ϕ3 (f (x)) + · · ·

+ ϕp (x) − ϕ1 (f (x))
p
X
= [ϕi (x) − ϕi (f (x))] .
i=1
 FIXED POINTS FOR MAPPINGS 87

Now define Φ : A → R by taking

p
X
−1
Φ (x) = p ϕi (x) , x ∈ A.
i=1

Then Φ is lower semicontinuous and bounded below, and moreover

d (x, f (x)) ≤ Φ (x) − Φ (f (x)) for each x ∈ A.

The conclusion now follows from Caristi’s theorem. 

4. Nonexpansive mappings

We do not know of results analogous to Theorems 2.2 and 2.3 for nonex-
pansive mappings. However assumption (1) can arise in a reasonably natural
way in the study of nonexpansive mappings. We illustrate this by giving a
short proof of the following theorem (which is stated in a more general but
less elegant way in [7]). Recall that if X is a Banach space and K ⊆ X, then
T : K → X is said to be nonexpansive if kT (x) − T (y)k ≤ kx − yk for each
x, y ∈ K.

Theorem 4.1. ([7]) Let K be a nonempty weakly compact convex subset of a

Banach space X, suppose T : K → K is nonexpansive, and suppose for each
x ∈ K there exists a positive integer N (x) and an α (x) ∈ (0, 1) such that

N (x)
T (x) − T N (x) (y) ≤ α (x) kx − yk for each y ∈ K.

Then T has a unique fixed point.

Proof. By weak compactness we may suppose K is minimal with respect to

being nonempty, closed, convex, and T -invariant. It is well known (cf., [5], p.
124) that in this case each point w ∈ K must be diametral, that is,

sup {kw − yk : y ∈ K} = diam (K) .

We shall show that the assumption diam (K) > 0 implies the existence of a
nondiametral point, and thereby to the conclusion that K contains a single
point which is fixed under T.
88 W. A. KIRK, P. S. SRINIVASAN, P. VEERAMANI

Let x ∈ K, let N = N (x) , and let r = α (x) diam (K) . Define

S1 := z ∈ K : z − T iN (x) ≤ r for almost all i ≥ 1 ;


S2 := z ∈ K : z − T iN +1 (x) ≤ r for almost all i ≥ 1 ;

..
.
SN := z ∈ K : z − T (i+1)N −1 (x) ≤ r for almost all i ≥ 1 .


Observe that T N (x) ∈ S1 , so S1 6= ∅. Also, if z ∈ S1 then

T (z) − T iN +1 (x) ≤ z − T iN (x) ;

hence T (z) ∈ S2 . Similarly T (Si ) ⊆ Si+1 , 1 ≤ i ≤ N − 1 and T (SN ) ⊆ S1 .

Also it is easy to see that the sets {Si }N
i=1 are convex. Now let i ≥ j ≥ N, say
j = N + k and i = N + k + s. Then
i
T (x) − T j (x) ≤ T N +s (x) − T N (x) ≤ r.

Thus T i (x) lies in the closed ball B T j (x) ; r for all i ≥ j ≥ N. Therefore the


∞
family B T i (x) ; r ∩ K i=N of closed convex sets has the finite intersection


property, so by weak compactness there exists a point z ∈ K such that

z ∈ ∩∞ i


i=N B T (x) ; r .

Clearly this implies z ∈ S := ∩Ni=1 Si ∩ K. Since S is nonempty closed convex

and T -invariant, by minimality of K it must be the case that S = K. In
particular S 1 = K. Hence if u ∈ K and if ε > 0 is chosen so that r + ε <
diam (K) , then
u − T iN (x) ≤ r + ε < diam (K)

for i sufficiently large. In particular, if {u1 , u2 , · · ·, uk } ⊂ K then

k
∩i=1 B (ui ; r) ∩ K 6= ∅. Thus the family of weakly compact sets

{B (u; r + ε) ∩ K : u ∈ K}

also has the finite intersection property. Let

z ∈ (∩u∈K B (u; r + ε)) ∩ K.

Then kz − uk ≤ r + ε < diam (K) for each u ∈ K. This means that z is a

non-diametral point of K – a contradiction. 
 FIXED POINTS FOR MAPPINGS 89

References
[1] D. W. Boyd and J. S. W. Wong, On nonlinear contractions, Proc. Amer. Math. Soc.
20(1969), 458-464.
[2] J. Caristi, Fixed point theorems for mappings satisfying inwardness conditions, Trans.
Amer. Math. Soc. 215(1976), 241-251.
[3] M. Edelstein, On fixed and periodic points under contractive mappings, J. London Math.
Soc. 37(1962), 74-79.
[4] M. A. Geraghty, On contractive mappings, Proc. Amer. Math. Soc. 40(1973), 604-608.
[5] K. Goebel and W. A. Kirk, Topics in Metric Fixed Point Theory (W. A. Kirk and B.
Sims, eds.), Cambridge Univ. Press, Cambridge, 2000.
[6] W. A. Kirk, Contraction mappings and extensions; in Handbook of Metric Fixed Point
Theory, Kluwer Acad. Publ., Dordrecht, Boston, London, 2001; pp. 1-34.
[7] W. A. Kirk and W. D. Royalty, Fixed-point theorems for certain nonlinear nonexpansive
mappings, Illinois J. Math. 15(1971), 656-663.