Bulletin of the Iranian Mathematical Society Vol. 39 No. 6 (2013), pp 1261-1272.
FIXED POINTS FOR E-ASYMPTOTIC
CONTRACTIONS AND BOYD-WONG TYPE
E-CONTRACTIONS IN UNIFORM SPACES
A. AGHANIANS, K. FALLAHI AND K. NOUROUZI∗
Communicated by Gholam Hossein Esslamzadeh
Abstract. In this paper we discuss the fixed points of asymptotic
contractions and Boyd-Wong type contractions in uniform spaces
equipped with an E-distance. A new version of Kirk’s fixed point
theorem is given for asymptotic contractions and Boyd-Wong type
contractions is investigated in uniform spaces.
1. Introduction and preliminaries
In 2003, Kirk [5] discussed the existence of fixed points for (not necessarily continuous) asymptotic contractions in complete metric spaces.
Jachymski and Jóźwik [4] constructed an example to show that continuity of the self-mapping is essential in Kirk’s theorem. They also
established a fixed point result for uniformly continuous asymptotic ϕcontractions in complete metric spaces.
MSC(2010): Primary: 47H10; Secondary: 54E15, 47H09.
Keywords: Separated uniform space, E-asymptotic contraction, Boyd-Wong type Econtraction, fixed point.
Received: 1 June 2012, Accepted: 1 December 2012.
∗Corresponding author
c 2013 Iranian Mathematical Society.
1261
1262
Aghanians, Fallahi and Nourouzi
Motivated by [5, Theorem 2.1] and [4, Example 1], we aim to give a
more general form of [5, Theorem 2.1] in uniform spaces where the selfmappings are assumed to be continuous. We also generalize the BoydWong fixed point theorem [3, Theorem 1] to uniform spaces equipped
with an E-distance.
We begin with some basics in uniform spaces which are needed in this
paper. The reader can find an in-depth discussion in, e.g., [6].
A uniformity on a nonempty set X is a nonempty collection U of
subsets of X × X (called the entourages of X) satisfying the following
conditions:
(1) Each entourage of X contains the diagonal {(x, x) : x ∈ X};
(2) U is closed under finite intersections;
(3) For each entourage U in U, the set {(x, y) : (y, x) ∈ U } is in U;
(4) For each U ∈ U, there exists an entourage V such that (x, y), (y, z)
∈ V implies (x, z) ∈ U for all x, y, z ∈ X;
(5) U contains the supersets of its elements.
If U is a uniformity on X, then (X, U) (shortly denoted by X) is called
a uniform space.
If d is a metric on a nonempty set X, then it induces a uniformity,
called the uniformity induced by the metric d, in which the entourages
of X are all the supersets of the sets
(x, y) ∈ X × X : d(x, y) < ε ,
where ε > 0.
It is well-known that a uniformity U on a nonempty set X is separating
if the intersection of all entourages of X coincides with the diagonal
{(x, x) : x ∈ X}. In this case, X is called a separated uniform space.
We next recall some basic concepts about E-distances. For more
details and examples, the reader is referred to [1].
Definition 1.1. [1] Let X be a uniform space. A function p : X × X →
R≥0 is called an E-distance on X if
(1) for each entourage U in U, there exists a δ > 0 such that p(z, x) ≤
δ and p(z, y) ≤ δ imply (x, y) ∈ U for all x, y, z ∈ X;
(2) p satisfies the triangular inequality, i.e.,
p(x, y) ≤ p(x, z) + p(z, y)
(x, y, z ∈ X).
If p is an E-distance on a uniform space X, then a sequence {xn }
p
in X is said to be p-convergent to a point x ∈ X, denoted by xn −→
x, whenever p(xn , x) → 0 as n → ∞, and X is p-Cauchy whenever
E-asymptotic contractions and Boyd-Wong type E-contractions
1263
p(xm , xn ) → 0 as m, n → ∞. The uniform space X is called p-complete
if every p-Cauchy sequence in X is p-convergent to some point of X.
The next lemma contains an important property of E-distances on
separated uniform spaces. The proof is straightforward and it is omitted
here.
Lemma 1.2. [1] Let {xn } and {yn } be two arbitrary sequences in a
p
separated uniform space X equipped with an E-distance p. If xn −→ x
p
and xn −→ y, then x = y. In particular, p(z, x) = p(z, y) = 0 for some
z ∈ X implies x = y.
Using E-distances, p-boundedness and p-continuity are defined in uniform spaces.
Definition 1.3. [1] Let p be an E-distance on a uniform space X. Then
(1) X is called p-bounded if
δp (X) = sup p(x, y) : x, y ∈ X < ∞.
p
(2) A mapping T : X → X is called p-continuous on X if xn −→ x
p
implies T xn −→ T x for all sequences {xn } and all points x in
X.
2. E-asymptotic contractions
In this section, we denote by Φ the class of all functions ϕ : R≥0 →
R≥0 with the following properties:
• ϕ is continuous on R≥0 ;
• ϕ(t) < t for all t > 0.
It is worth mentioning that if ϕ ∈ Φ, then
0 ≤ ϕ(0) = lim ϕ(t) ≤ lim t = 0,
t→0+
t→0+
that is, ϕ(0) = 0.
Following [5, Definition 2.1], we define E-asymptotic contractions.
Definition 2.1. Let p be an E-distance on a uniform space X. We say
that a mapping T : X → X is an E-asymptotic contraction if
(2.1)
p(T n x, T n y) ≤ ϕn p(x, y) for all x, y ∈ X and n ≥ 1,
where {ϕn } is a sequence of nonnegative functions on R≥0 converging
uniformly to some ϕ ∈ Φ on the range of p.
1264
Aghanians, Fallahi and Nourouzi
If (X, d) is a metric space, then replacing the E-distance p by the metric d in Definition 2.1, we get the concept of an asymptotic contraction
introduced by Kirk [5, Definition 2.1]. So each asymptotic contraction
on a metric space is an E-asymptotic contraction on the uniform space
induced by the metric. But in the next example, we see that the converse
is not generally true.
Example 2.2. Uniformize the set X = [0, 1] with the uniformity induced
from the Euclidean metric and put p(x, y) = y for all x, y ∈ X. It is
easily verified that p is an E-distance on X. Define T : X → X and
ϕ1 : R≥0 → R≥0 by
1
0
0
≤
x
<
1
16 0 ≤ t < 1
and ϕ1 (t) =
Tx =
1
1
x=1
t≥1
8
8
for all x ∈ X and all t ≥ 0, and set ϕn = ϕ for n ≥ 2, where ϕ is any
arbitrary fixed function in Φ. Clearly, ϕn → ϕ uniformly on R≥0 and
T n = 0 for all n ≥ 2. To see that T is an E-asymptotic contraction on
X, it suffices to check (2.1) for n = 1. To this end, given x, y ∈ [0, 1], if
y = 1, then we have
1
p(T x, T 1) = T 1 = = ϕ1 (1) = ϕ1 p(x, 1) ,
8
and for 0 ≤ y < 1, we have
1
p(T x, T y) = T y = 0 ≤
= ϕ1 (y) = ϕ1 p(x, y) .
16
But T fails to be an asymptotic contraction on the metric space X with
the functions ϕn since
1
1
1
1
1
.
= ϕ1
= ϕ1 1 −
T1 − T = >
2
8
16
2
2
In the next example, we see that an E-asymptotic contraction need
not be p-continuous.
Example 2.3. Let X and p be as in Example 2.2. Define a mapping
T : X → X by T x = 0 if 0 < x ≤ 1 and T 0 = 1. Note that T is fixed
point free. Now, let ϕ1 be the constant function 1 and ϕ2 = ϕ3 = · · · = ϕ,
where ϕ is an arbitrary fixed function in Φ. Then T satisfies (2.1) and
since T 0 6= 0, it follows that T fails to be p-continuous on X.
E-asymptotic contractions and Boyd-Wong type E-contractions
1265
Theorem 2.4. Let p be an E-distance on a separated uniform space X
such that X is p-complete and let T : X → X be a p-continuous Easymptotic contraction for which the functions ϕn in Definition 2.1 are
all continuous on R≥0 for large indices n. Then T has a unique fixed
p
point u ∈ X, and T n x −→ u for all x ∈ X.
Proof. We divide the proof into three steps.
Step 1: p(T n x, T n y) → 0 as n → ∞ for all x, y ∈ X.
Let x, y ∈ X be given. Letting n → ∞ in (2.1), we get
0 ≤ lim sup p(T n x, T n y) ≤ lim ϕn p(x, y) = ϕ p(x, y) ≤ p(x, y) < ∞.
n→∞
n→∞
Now, if
lim sup p(T n x, T n y) = ε > 0,
n→∞
then there exists a strictly increasing sequence {nk } of positive integers
such that p(T nk x, T nk y) → ε, and so by the continuity of ϕ, one obtains
ϕ p(T nk x, T nk y) → ϕ(ε) < ε.
Therefore, there is an integer k0 ≥ 1 such that ϕ(p(T nk0 x, T nk0 y)) < ε.
So (2.1) yields
ε = lim sup p(T n x, T n y)
n→∞
= lim sup p T n (T nk0 x), T n (T nk0 y)
n→∞
≤ lim ϕn p(T nk0 x, T nk0 y)
n→∞
= ϕ p(T nk0 x, T nk0 y) < ε,
which is a contradiction. Hence
lim sup p(T n x, T n y) = 0.
n→∞
Consequently,
0 ≤ lim inf p(T n x, T n y) ≤ lim sup p(T n x, T n y) = 0,
n→∞
that is,
p(T n x, T n y)
n→∞
→ 0.
Step 2: The sequence {T n x} is p-Cauchy for all x ∈ X.
Suppose that x ∈ X is arbitrary. If {T n x} is not p-Cauchy, then there
exist ε > 0 and positive integers mk and nk such that
mk > nk ≥ k
and
p(T mk x, T nk x) ≥ ε
k = 1, 2, . . . .
1266
Aghanians, Fallahi and Nourouzi
Keeping the integer nk fixed for sufficiently large k, say k ≥ k0 , and
using Step 1, we may assume without loss of generality that mk > nk is
the smallest integer with p(T mk x, T nk x) ≥ ε, that is,
p(T mk −1 x, T nk x) < ε.
Hence for each k ≥ k0 , we have
ε ≤ p(T mk x, T nk x)
≤ p(T mk x, T mk −1 x) + p(T mk −1 x, T nk x)
< p(T mk x, T mk −1 x) + ε.
If k → ∞, since p(T mk x, T mk −1 x) → 0, it follows that p(T mk x, T nk x) →
ε.
We next show by induction that
(2.2)
lim sup p(T mk +i x, T nk +i x) ≥ ε,
i = 1, 2, . . . .
k→∞
To this end, note first that from Step 1,
ε = lim p(T mk x, T nk x) = lim sup p(T mk x, T nk x)
k→∞
k→∞
h
≤ lim sup p(T mk x, T mk +1 x) + p(T mk +1 x, T nk +1 x)
i
k→∞
+ p(T nk +1 x, T nk x)
≤ lim sup p(T mk x, T mk +1 x) + lim sup p(T mk +1 x, T nk +1 x)
k→∞
+ lim sup p(T nk +1 x, T nk x)
k→∞
k→∞
= lim sup p(T mk +1 x, T nk +1 x),
k→∞
that is, (2.2) holds for i = 1. If (2.2) is true for some i, then
ε ≤ lim sup p(T mk +i x, T nk +i x)
k→∞ h
≤ lim sup p(T mk +i x, T mk +i+1 x) + p(T mk +i+1 x, T nk +i+1 x)
i
k→∞
+ p(T nk +i+1 x, T nk +i x)
≤ lim sup p(T mk +i+1 x, T nk +i+1 x).
k→∞
Consequently, we have
ϕ(ε) = lim ϕ p(T mk x, T nk x)
k→∞
E-asymptotic contractions and Boyd-Wong type E-contractions
= lim lim ϕi p(T mk x, T nk x)
k→∞ i→∞
= lim lim ϕi p(T mk x, T nk x)
i→∞ k→∞
1267
≥ lim sup lim sup p(T mk +i x, T nk +i x)
i→∞
k→∞
≥ ε,
where the first equality holds because ϕ is continuous, the second equality holds because {ϕi } is pointwise convergent to ϕ on the range of p,
the third equality holds because {ϕi } is uniformly convergent to ϕ on
the range of p, and the last two inequalities hold by (2.1) and (2.2), respectively. Hence ϕ(ε) ≥ ε, which is a contradiction. Therefore {T n x}
is p-Cauchy.
Step 3: T has a unique fixed point.
Because X is p-complete, it is concluded from Steps 1 and 2 that the
family {{T n x} : x ∈ X} of Picard iterates of T is p-equiconvergent,
p
that is, there exists u ∈ X such that T n x −→ u for all x ∈ X. In
p
particular, T n u −→ u. We claim that u is the unique fixed point for T .
To this end, first note that since T is p-continuous on X, it follows that
p
T n+1 u −→ T u, and so, by Lemma 1.2, we have u = T u. And if v ∈ X
is a fixed point for T , then
p(u, v) = lim p(T n u, T n v) ≤ lim ϕn p(u, v) = ϕ p(u, v) ,
n→∞
n→∞
which is impossible unless p(u, v) = 0. Similarly p(u, u) = 0 and using
Lemma 1.2 once more, we get v = u.
It is worth mentioning that the boundedness of some orbit of T is not
necessary in Theorem 2.4 unlike [5, Theorem 2.1] or [2, Theorem 4.1.15].
As a consequence of Theorem 2.4, we have the following version of [1,
Theorem 3.1].
Corollary 2.5. Let p be an E-distance on a separated uniform space X
such that X is p-complete and p-bounded and let a mapping T : X → X
satisfy
(2.3)
p(T x, T y) ≤ ϕ p(x, y) for all x, y ∈ X,
where ϕ : R≥0 → R≥0 is nondecreasing and continuous with ϕn (t) → 0
p
for all t > 0. Then T has a unique fixed point u ∈ X, and T n x −→ u
for all x ∈ X.
1268
Aghanians, Fallahi and Nourouzi
Proof. Note first that ϕ(0) = 0; for if 0 < t < ϕ(0) for some t, then the
monotonicity of ϕ implies that 0 < t < ϕ(0) ≤ ϕn (t) for all n ≥ 1, which
contradicts the fact that ϕn (t) → 0.
Next, since ϕ is nondecreasing, it follows that T satisfies
p(T n x, T n y) ≤ ϕn p(x, y) for all x, y ∈ X and n ≥ 1.
Setting ϕn = ϕn for each n ≥ 1 in Definition 2.1, it is seen that {ϕn }
converges pointwise to the constant function 0 on [0, +∞), and since
n
o
sup ϕn p(x, y) : x, y ∈ X = ϕn δp (X) → 0,
it follows that {ϕn } converges uniformly to 0 on the range of p. Because
the constant function 0 belongs to Φ, it is concluded that T is an Easymptotic contraction on X. Moreover, ϕn ’s are all continuous on R≥0
and (2.3) ensures that T is p-continuous on X. Consequently, the result
follows immediately from Theorem 2.4.
The next corollary is a partial modification of Kirk’s theorem [5, Theorem 2.1] in uniform spaces. One can find it with an additional assumption, e.g., in [2, Theorem 4.1.15].
Corollary 2.6. Let X be a complete metric space and let T : X → X
be a continuous asymptotic contraction for which the functions ϕn in
Definition 2.1 are all continuous on R≥0 for large indices n. Then T
has a unique fixed point u ∈ X, and T n x → u for all x ∈ X.
3. Boyd-Wong type E-contractions
In this section, we denote by Ψ the class of all functions ψ : R≥0 →
with the following properties:
• ψ is upper semicontinuous on R≥0 from the right, i.e.,
R≥0
tn ↓ t ≥ 0
implies
lim sup ψ(tn ) ≤ ψ(t);
n→∞
• ψ(t) < t for all t > 0, and ψ(0) = 0.
It might be interesting for the reader to be mentioned that the family
Φ defined and used in Section 2 is contained in the family Ψ but these
two families do not coincide. To see this, consider the function ψ(t) = 0
if 0 ≤ t < 1, and ψ(t) = 12 if t ≥ 1. Then ψ is upper semicontinuous
from the right but it is not continuous on R≥0 . Furthermore, the upper
semicontinuity of ψ on R≥0 from the right and the condition that ψ(t) <
E-asymptotic contractions and Boyd-Wong type E-contractions
1269
t for all t > 0, do not imply that ψ vanishes at zero in general. In fact,
the function ψ : R≥0 → R≥0 defined by the rule
a
t=0
t
0<t<1
ψ(t) =
2
1
t≥1
2t
for all t ≥ 0, where a is an arbitrary positive real number, confirms this
claim.
Theorem 3.1. Let p be an E-distance on a separated uniform space X
such that X is p-complete and let T : X → X satisfy
(3.1)
p(T x, T y) ≤ ψ p(x, y) for all x, y ∈ X,
p
where ψ ∈ Ψ. Then T has a unique fixed point u ∈ X, and T n x −→ u
for all x ∈ X.
Proof. We divide the proof into three steps as Theorem 2.4.
Step 1: p(T n x, T n y) → 0 as n → ∞ for all x, y ∈ X.
Let x, y ∈ X be given. Then for each nonnegative integer n, by the
contractive condition (3.1) we have
(3.2)
p(T n+1 x, T n+1 y) ≤ ψ p(T n x, T n y) ≤ p(T n x, T n y).
Thus, {p(T n x, T n y)} is a nonincreasing sequence of nonnegative numbers and so it converges decreasingly to some α ≥ 0. Letting n → ∞ in
(3.2), by the upper semicontinuity of ψ from the right, we get
α = lim p(T n+1 x, T n+1 y) ≤ lim sup ψ p(T n x, T n y) ≤ ψ(α),
n→∞
n→∞
which is a contradiction unless α = 0. Consequently, p(T n x, T n y) → 0.
Step 2: The sequence {T n x} is p-Cauchy for all x ∈ X.
Let x ∈ X be arbitrary and suppose on the contrary that {T n x} is
not p-Cauchy. Then similar to the proof of Step 2 of Theorem 2.4, it is
seen that there exist an ε > 0 and sequences {mk } and {nk } of positive
integers such that mk > nk for each k and p(T mk x, T nk x) → ε. On the
other hand, for each k, by (3.1) we have
p(T mk x, T nk x) ≤ p(T mk x, T mk +1 x) + p(T mk +1 x, T nk +1 x)
+ p(T nk +1 x, T nk x)
1270
Aghanians, Fallahi and Nourouzi
≤ p(T mk x, T mk +1 x) + ψ p(T mk x, T nk x)
+ p(T nk +1 x, T nk x).
Letting k → ∞ and using Step 1 and the upper semicontinuity of ψ from
the right, we obtain
ε = lim p(T mk x, T nk x) = lim sup p(T mk x, T nk x)
k→∞
k→∞
h
mk
mk +1
≤ lim sup p(T x, T
x) + ψ p(T mk x, T nk x)
i
k→∞
+ p(T nk +1 x, T nk x)
≤ lim sup p(T mk x, T mk +1 x) + lim sup ψ p(T mk x, T nk x)
k→∞
+ lim sup p(T nk +1 x, T nk x)
k→∞
= lim sup ψ p(T mk x, T nk x)
k→∞
k→∞
≤ ψ(ε),
which is a contradiction. Therefore, {T n x} is p-Cauchy.
Step 3: T has a unique fixed point.
Since X is p-complete, it follows from Steps 1 and 2 that the family
{{T n x} : x ∈ X} is p-equiconvergent to some u ∈ X. In particular,
p
T n u −→ u. Since (3.1) implies the p-continuity of T on X, it follows
p
that T n+1 u −→ T u and so, by Lemma 1.2, we have u = T u, that is, u
is a fixed point for T . If v ∈ X is a fixed point for T , then
p(u, v) = p(T u, T v) ≤ ψ p(u, v) ,
which is impossible unless p(u, v) = 0. Similarly p(u, u) = 0. Therefore,
using Lemma 1.2 once more, one gets v = u.
As an immediate consequence of Theorem 3.1, we have the BoydWong’s theorem [3] in metric spaces:
Corollary 3.2. Let X be a complete metric space and let a mapping
T : X → X satisfy
(3.3)
d(T x, T y) ≤ ψ d(x, y) for all x, y ∈ X,
where ψ ∈ Ψ. Then T has a unique fixed point u ∈ X, and T n x → u for
all x ∈ X.
E-asymptotic contractions and Boyd-Wong type E-contractions
1271
In the following example, we see that Theorem 3.1 guarantees the
existence and uniqueness of a fixed point while Corollary 3.2 cannot be
applied.
Example 3.3. Let the set X = [0, 1] be endowed with the uniformity
induced by the Euclidean metric and define a mapping T : X → X by
T x = 0 if 0 ≤ x < 1, and T 1 = 41 . Then T does not satisfy (3.3) for any
ψ ∈ Ψ since it is not continuous on X. In fact, if ψ ∈ Ψ is arbitrary,
then
1
3
1
3
.
=ψ 1−
T1 − T = > ψ
4
4
4
4
Now set p(x, y) = max{x, y}. Then p is an E-distance on X and T
satisfies (3.1) for the function ψ : R≥0 → R≥0 defined by the rule ψ(t) =
t
4 for all t ≥ 0. It is easy to check that this ψ belongs to Ψ, and the
hypotheses of Theorem 3.1 are fulfilled.
Remark 3.4. In Theorem 2.4 (Corollary 2.6), assume that for some
index k the function ϕk belongs to Φ. Then Theorem 3.1 (Corollary 3.2)
p
implies that T k and so T has a unique fixed point u and T kn x −→ u for
all x ∈ X. So, it is concluded by the p-continuity of T that the family
{{T n x} : x ∈ X} is p-equiconvergent to u. Hence the significance of
Theorem 2.4 (Corollary 2.6) is whenever none of ϕn ’s satisfy ϕn (t) < t
for all t > 0, that is, whenever for each n ≥ 1 there exists a tn > 0 such
that ϕn (tn ) ≥ tn .
References
[1] M. Aamri and D. El Moutawakil, Common fixed point theorems for E-contractive
or E-expansive maps in uniform spaces, Acta Math. Acad. Paedagog. Nyházi.
(N.S.) 20 (2004), no. 1, 83–91.
[2] R. P. Agarwal, D. O’Regan and D. R. Sahu, Fixed Point Theory for LipschitzianType Mappings with Applications, Springer, New York, 2009.
[3] D. W. Boyd and J. S. W. Wong, On nonlinear contractions, Proc. Amer. Math.
Soc. 20 (1969) 458–464.
[4] J. Jachymski and I. Jóźwik, On Kirk’s asymptotic contractions, J. Math. Anal.
Appl. 300 (2004), no. 1, 147–159.
[5] W. A. Kirk, Fixed points of asymptotic contractions, J. Math. Anal. Appl. 277
(2003), no. 2, 645–650.
[6] S. Willard, General Topology, Addison-Wesley Publishing Co., Mass.-LondonDon Mills, Ont., 1970.
1272
Aghanians, Fallahi and Nourouzi
Aris Aghanians
Department of Mathematics, K. N. Toosi University of Technology, P.O. Box 163151618, Tehran, Iran
Email: a.aghanians@dena.kntu.ac.ir
Kamal Fallahi
Department of Mathematics, K. N. Toosi University of Technology, P.O. Box 163151618, Tehran, Iran
Email: k fallahi@dena.kntu.ac.ir
Kourosh Nourouzi
Department of Mathematics, K. N. Toosi University of Technology, P.O. Box 163151618, Tehran, Iran
Email: nourouzi@kntu.ac.ir