Skip to main content

Fixed point theorems for contracting mappings in partial metric spaces

Abstract

The paper deals with a class of contracting mappings that includes the Boyd-Wong and the Matkowski classes. Some of the new fixed point theorems obtained here unify and extend well-known results also in the case of metric spaces.

MSC:47H09, 47H10, 54H25, 54E99.

Introduction

In the simplest case, the condition with a contracting mapping has the form d(f(y),f(x))φ(d(y,x)), where (X,d) is a metric space, f is a mapping on X, and φ:[0,)[0,) is such that φ(α)<α, α>0. Boyd and Wong [1] assumed that φ is upper semicontinuous from the right (i.e., lim sup β α + φ(β)φ(α), α0); Matkowski in [2] assumed that φ is nondecreasing and such that lim n φ n (α)=0 for each α>0. We assume that all sequences ( a n ) n N such that a n + 1 φ( a n ), nN, converge to zero. It appears that the classes of Boyd-Wong’s and Matkowski’s mappings are included in this new class (the problem of φ(0) for Boyd-Wong mappings is meaningless for contractions). The main results are Theorems 21, 22, 23, and theorems extending the well-known classical results: Theorem 30 (Matkowski’s theorem) and Theorem 31 (covering the theorems of Romaguera and Boyd-Wong).

Let us recall the notions of a partial metric space due to Matthews [[3], Definition 3.1] and of a dualistic partial metric due to Oltra and Valero [4] and O’Neill [5].

Definition 1 A dualistic partial metric is a mapping p:X×XR such that

y=xiffp(y,y)=p(y,x)=p(x,x),x,yX,
(1)
p(y,y)p(y,x),x,yX,
(2)
p(y,x)=p(x,y),x,yX,
(3)
p(z,x)p(z,y)+p(y,x)p(y,y),x,y,zX.
(4)

If p is nonnegative, then it is a partial metric.

If p is a dualistic partial metric on X, then q:X×X[0,) defined by

q(y,x)=p(y,x)p(y,y),x,yX
(5)

is a quasi-metric [[3], Theorem 4.1] (y=x iff q(y,x)=q(x,y)=0, q(z,x)q(z,y)+q(y,x)).

An open ball for xX, ϵ>0 is defined by

B(x,ϵ)= { y X : q ( x , y ) < ϵ } = { y X : p ( x , y ) < p ( x , x ) + ϵ } .
(6)

The family of open balls generates topology T q on X. It is accepted that the (dualistic) partial metric space (X,p) is equipped with the topology T q .

It is known (see, e.g., [4]) that a metric d can be defined by a dualistic partial metric p as follows:

d ( y , x ) = max { q ( y , x ) , q ( x , y ) } = max { p ( y , x ) p ( y , y ) , p ( x , y ) p ( x , x ) } , x , y X .
(7)

In this paper it is understood that q, d are defined by (5), (7) respectively for a (dualistic) partial metric p.

For dualistic partial metric spaces, it is accepted (see, e.g., [[4], p.19]) that ( x n ) n N is called a Cauchy sequence in (X,p) if lim m , n p( x n , x m )=αR, and (X,p) is complete if for every Cauchy sequence ( x n ) n N in (X,p), there exists x lim n x n such that lim m , n p( x n , x m )=p(x,x).

The following notions are useful.

Definition 2 [[6], Definition 2.1]

Let p:X×XR be a mapping. The kernel of p is the set Kerp={xX:p(x,x)=0}.

Definition 3 (cp. [[7], Definition 2.1])

A dualistic partial metric space (X,p) is 0-complete if for every sequence ( x n ) n N in X such that lim m , n p( x n , x m )=0, there exists x lim n x n Kerp in (X,p).

In fact Definition 3 is too abstract. The condition x lim n x n Kerp in (X,p) means that lim n p(x, x n )=p(x,x)=0 (see (6), (2)). If, in addition, lim m , n p( x n , x m )=0, then x= lim n x n in (X,d) (see [[4], Lemma 2.2] or [[6], Proposition 1.4]).

Corollary 4 A dualistic partial metric space (X,p) is 0-complete iff every sequence ( x n ) n N such that lim m , n p( x n , x m )=0 converges in (X,d) to a point xKerp. If p is a metric, then 0-completeness is identical with completeness.

There exist 0-complete partial metric spaces which are not complete [8]. Some criterions of 0-completeness can be found in [[9], Section 4].

Proposition 5 Let (X,p) be a partial metric space. Then ( x n ) n N converges in (X,d) to xKerp iff lim n p(x, x n )=0 and iff we have x lim n x n Kerp in (X,p).

Proof A sequence ( x n ) n N converges to x in (X,d) iff lim m , n p( x n , x m )= lim n p(x, x n )=p(x,x) (see [[4], Lemma 2.2] or [[6], Proposition 1.4]). Assume lim n p(x, x n )=0. Then, for nonnegative p, condition (2) yields p(x,x)=0 and lim n p(x, x n )=p(x,x), i.e., xKerp and x lim n x n in (X,p). We also have

0p( x n , x m )p( x n ,x)+p(x, x m )p(x,x)=p(x, x n )+p(x, x m )

and consequently, lim m , n p( x n , x m )= lim n p(x, x n )=p(x,x)=0, i.e., xKerp and ( x n ) n N converges in (X,d) to x. □

Let 2 Y be the family of all subsets of Y. We say that F:X 2 Y is a (multivalued) mapping if F(x) for all xX.

Now, let us investigate the concept of a 0-closed graph.

Definition 6 Let (X,p) be a dualistic partial metric space. A mapping F:X 2 X has a 0-closed graph if for all sequences ( x n ) n N , ( y n ) n N in X the following condition is satisfied:

lim n p(x, x n )= lim n p(x, y n )=0, y n F( x n ),nN, and xKerp yield xF(x).
(8)

Proposition 5 and [[6], Proposition 1.4] yield the following.

Corollary 7 For all sequences ( x n ) n N , ( y n ) n N with lim m , n p( x n , x m )= lim m , n p( y n , y m )=0, condition (8) can be replaced by

lim n x n = lim n y n =xin (X,d), y n F( x n ),nN, and xKerp yield xF(x).
(9)

If p is nonnegative, then (8) is equivalent to (9).

Clearly, if F:X 2 X has a closed graph in (X,d)×(X,d), then (9) is satisfied as Kerp with topology induced by p is a closed metric subspace of (X,d) [[6], Lemma 2.2].

For a partial metric space (X,p), a nonempty set AX, and zX, let us adopt

p(A,z)=inf { p ( x , z ) : x A }

and

E(A,r)= { z X : p ( A , z ) < r } .

The family {E(A,ϵ):A 2 X {},ϵ>0} generates a topology , and we get a topological space ( 2 X {},P). Assume that p is nonnegative and

for every  ϵ > 0 ,  there exists  δ > 0  such that for each  u X , p ( x , u ) < δ  yields  F ( u ) E ( F ( x ) , ϵ )
(10)

holds. Then, for all sequences ( x n ) n N , ( y n ) n N such that lim n p(x, x n )= lim n p(x, y n )=0 and y n F( x n ), nN, there exist z n F(x), nN with lim n p( y n , z n )=0. We have

p(x, z n )p(x, y n )+p( y n , z n )

and consequently, x= lim n z n in (X,d) (Proposition 5). If F(x)Kerp= F ( x ) ¯ Kerp in (X,d), then xF(x) holds.

For a partial metric space (X,p) and nonempty A,CX, let us adopt

P(A,C)=inf { r > 0 : A E ( C , r )  and  C E ( A , r ) } .
(11)

If p is a metric, then P is the Hausdorff metric of the metric p, whenever A and C are nonempty closed and bounded subsets of X. In general, P is not a partial metric (see [[10], Proposition 2.2(i), Proposition 2.3(h3)]).

Clearly, p(x,u)p(x,x)=p(x,u) for xKerp and if (10) holds, then F:(X,p)( 2 X {},P) is continuous on Kerp.

Corollary 8 Let (X,p) be a partial metric space. Assume that for each xKerp, the mapping F:X 2 X satisfies (10) and F(x)Kerp= F ( x ) ¯ Kerp in (X,d). Then F has a 0-closed graph (and conditions (8), (9) are equivalent).

Proposition 9 Let (X,p) be a dualistic partial metric space. If F:X 2 X has a 0-closed graph, then FixFKerp is closed in (X,d).

Proof If x n FixFKerp, nN, then for y n = x n , nN and each xKerp such that lim n p(x, x n )=0, condition (8) yields xF(x), i.e., FixFKerp is closed in Kerp (which with topology induced by p is a closed metric subspace of (X,d) [[6], Lemma 2.2]). □

Now, let us investigate a ‘contraction’ condition.

Let Φ be a class of mappings φ:[0,)[0,) such that φ(α)<α, α>0; φ Φ 0 iff φΦ and φ(0)=0.

Proposition 10 Let (X,p) be a partial metric space. Assume that G,H:X 2 X satisfy

p ( H ( y ) , y ) φ ( p ( y , x ) ) ,yG(x),xX
(12)

for φ Φ 0 . Then FixGKerp; if H has a 0-closed graph, then FixGFixH holds.

Proof For y=xG(x) and nonnegative p, conditions (2), (12) yield

0p(x,x)inf { p ( z , x ) : z H ( x ) } =p ( H ( x ) , x ) φ ( p ( x , x ) ) ,

i.e., p(x,x)=0 and xKerp. What is more, from our inequality it follows that there exist y n H(x) such that lim n p( y n ,x)=0. Now, (8) for x n =x, F=H yields xH(x). □

Propositions 9, 10 yield the following.

Theorem 11 Let (X,p) be a 0-complete partial metric space, and let be a family of mappings X 2 X with 0-closed graphs. Assume that φ Φ 0 ; some and at least all different G,HF satisfy (12). Then all members of have the same set of fixed points; this set is closed in (X,d) and contained in Kerp.

The previous result becomes a little more interesting if a mapping FF has a fixed point.

Let us assume that the following condition is satisfied for a partial metric space (X,p), P defined by (11), and G,H:X 2 X :

P ( H ( y ) , G ( x ) ) φ ( p ( y , x ) ) ,x,yX.
(13)

Then, for any yG(x), we have

p ( H ( y ) , y ) sup { p ( H ( y ) , u ) : u G ( x ) } max { sup { p ( H ( y ) , u ) : u G ( x ) } , sup { p ( G ( x ) , v ) : v H ( y ) } } = P ( H ( y ) , G ( x ) ) φ ( p ( y , x ) ) ,

i.e., condition (13) yields condition (12). The subsequent two propositions enable us to strengthen condition (12).

Proposition 12 Let (X,p) be a partial metric space, and let CX be compact in (X,d). Then, for any yX, there exists xC such that p(x,y)=p(C,y).

Proof Let ( x n ) n N be a sequence in C such that

lim n p( x n ,y)=inf { p ( z , y ) : z C } =p(C,y)=α.

There exists a subsequence ( x k n ) n N of ( x n ) n N and xC such that x= lim n x k n in (X,d), i.e., [[4], Lemma 2.2]

lim m , n p( x k n , x k m )= lim n p(x, x k n )=p(x,x).

Now, from

αp(x,y)p(x, x k n )+p( x k n ,y)p( x k n , x k n )

we get

αp(x,y)p(x,x)+αp(x,x)=α.

 □

Corollary 13 Let (X,p) be a partial metric space, and let G,H:X 2 X be mappings with H compact valued in (X,d). Then condition (12) is equivalent to

for each  x X , y G ( x ) ,  there exists  z H ( y ) such that  p ( z , y ) = p ( H ( y ) , y ) φ ( p ( y , x ) ) .
(14)

Condition (14) extends the idea of α-step mappings [[11], Definition 17].

The next example shows that even a ‘good’ mapping φ in condition (14) does not guarantee the existence of a fixed point.

Example 14 Let us consider a continuous mapping φ Φ 0 defined by φ(α)=α/( α 2 +1) for α[0,1], and φ(α)=1/2 for α>1. An easy computation proves that φ is increasing on [0,1] and, therefore, φ is nondecreasing on [0,). Let us consider X=[1,) and f(x)=x+1/x, xX. For p(y,x)=d(y,x)=|yx|, we have

d ( f 2 ( x ) , f ( x ) ) =x+1/x+1/(x+1/x)x1/x=x/ ( 1 + x 2 ) =(1/x)/ ( ( 1 / x ) 2 + 1 ) .

Now, for α=d(f(x),x)=1/x1, we obtain

d ( f 2 ( x ) , f ( x ) ) =φ(α)=φ ( d ( f ( x ) , x ) ) ,

i.e., (14) is satisfied for H=f, y=f(x), z= f 2 (x). Still it is clear that f:XX has no fixed point.

It is a good idea suggested by [12] to gather together the properties of φ. Let us recall that Φ is a class of mappings φ:[0,)[0,) such that φ(α)<α, α>0; and φ Φ 0 iff φΦ and φ(0)=0.

Proposition 15 Assume that φ Φ 0 . Then every sequence ( a n ) n N such that a n + 1 φ( a n ), nN (in particular ( φ n ( α ) ) n N , α0) is nonincreasing; if, in addition, φ is nondecreasing, then a n + 1 φ n ( a 1 ), nN holds.

Proof From φ(α)α, α0, it follows that a n + 1 φ( a n ) a n . Similarly, for nondecreasing φ, we get a 2 φ( a 1 ), a 3 φ( a 2 ) φ 2 ( a 1 ),, a n + 1 φ n ( a 1 ). □

Let us present some subclasses of Φ 0 .

Let Φ P consist of mappings φ:[0,)[0,) for which every sequence ( a n ) n N such that a n + 1 φ( a n ), nN converges to zero.

Proposition 16 We have Φ P Φ 0 . For φ Φ P , the sequence ( φ n ( α ) ) n N is nonincreasing and it converges to 0, α0. If a mapping φ Φ 0 satisfies

lim sup β α + φ(β)<α,α>0,
(15)

then φ Φ P .

Proof Assume that φ Φ P . Suppose that αφ(α) for α>0. Then a n =α, nN, is a good counterexample (the sequence does not converge to 0). Therefore Φ P Φ holds. Suppose φ(0)=a>0. Then, for a 2 n 1 =a, a 2 n =0, nN, we obtain a divergent sequence ( a n ) n N such that a n + 1 φ( a n ), nN. Consequently, every φ Φ P satisfies φ(0)=0. The sequence ( φ n ( α ) ) n N is nonincreasing, φ n + 1 (α)=φ( φ n (α)) φ n (α), nN and it converges to 0 as φ Φ P . Assume φ Φ 0 . Then any sequence ( a n ) n N such that a n + 1 φ( a n ) ( a n ), nN is nonincreasing and therefore it converges, say, to γ0. Suppose γ>0. Then (15) yields

0= lim n [ a n + 1 a n ] lim sup n φ( a n )γ<γγ=0,

a contradiction. □

Let Φ B W Φ 0 consist of mappings φ upper semicontinuous from the right, call them Boyd-Wong mappings. Proposition 16 yields Φ B W Φ P .

In turn, let Φ M consist of nondecreasing mappings φ:[0,)[0,) such that lim n φ n (α)=0, α>0 (Matkowski mappings). It is well known [[13], Lemma] that Φ M Φ 0 . Moreover, a n + 1 φ( a n ), nN, yields a n + 1 φ n ( a 1 ) (Proposition 15) and hence lim n a n =0. Consequently, Φ M Φ P holds. Let us note that φ from Example 14 belongs to Φ M .

The following is a kind of the reverse condition.

Proposition 17 Let ( a n ) n N be a sequence convergent to zero and such that a n + 1 φ( a n ), nN, for a φ Φ 0 . Then there exists a mapping ψ Φ M Φ B W such that φ( a n )ψ( a n ) a n , and ψ n (α) a n , nN, α0.

Proof Let us adopt ψ(0)=0 and ψ(α)=sup{φ( a n ): a n α,nN}, α>0. Clearly, ψ is nondecreasing and continuous from the right as lim n φ( a n )=0. Therefore, ψ is also upper semicontinuous. The sequence ( a n ) n N is nonincreasing as a n + 1 φ( a n ) a n , nN. Let us adopt a 0 =. For 0< a n α< a n 1 , we have

φ( a n )sup { φ ( a n 1 + k ) : k N } =ψ(α)<sup{ a n 1 + k :kN}= a n α,

and the case of a n =0 is trivial as then ψ(α)=0<α. Consequently, we get ψ(α)< a 1 (if a 1 >0), ψ 2 (α)< a 2 (if a 2 >0), and so on. Finally, we obtain ψ n (α) a n , nN (the possible case of some a k =0). □

The following modification of φ is useful.

Proposition 18 Assume that φ:[0,)[0,) is a mapping. Then ψ defined by

ψ(α)= [ φ ( α ) + α ] /2,α0,
(16)

belongs to Φ iff φΦ; φ Φ 0 iff ψ Φ 0 ; φ satisfies (15) iff ψ satisfies (15). In addition, if φ is nondecreasing, then ψ is increasing.

Proof It is clear that φ(α)<ψ(α)<α holds iff ψ(α)<α. If φ is nondecreasing, then for 0α<β we obtain

ψ(β)ψ(α)= [ φ ( β ) φ ( α ) ] /2+(βα)/2(βα)/2>0,

i.e., ψ is increasing. The remaining part of the proof is also trivial. □

The next two propositions can be helpful in proving fixed point theorems.

Proposition 19 Assume that (X,p) is a partial metric space, and let φ Φ 0 . If mappings G,H:X 2 X satisfy (12) and H has a 0-closed graph, then for φ=ψ, where ψ is defined by (16), condition (14) holds.

Proof For p(y,x)>0, there exist ϵ>0 and zH(y) such that p(z,y)φ(p(y,x))+ϵψ(p(y,x)). If p(y,x)=0, then (2) yields x=y, i.e., x is a fixed point of G, and by Proposition 10, z=xH(x). □

Proposition 20 Let (X,p) be a partial metric space, let be a nonempty family of at most two mappings X 2 X with 0-closed graphs, and let φ Φ P . Assume that all members of either satisfy (12) and φ has property (15) or satisfy (14). Then there exists a sequence ( x n ) n N such that x 2 n H( x 2 n 1 ), x 2 n + 1 G( x 2 n ), nN, and lim n p( x n + 1 , x n )=0.

Proof If φ Φ 0 satisfies (15) and (12) holds, then for ψ as in (16) and φ=ψ, condition (14) is satisfied (Proposition 19), and φ Φ P (Propositions 18, 16). Thus, it is sufficient to consider the case of φ Φ P and condition (14). Let x 0 X be arbitrary, x 1 G( x 0 ), and let x 2 H( x 1 ) be such that p( x 2 , x 1 )φ(p( x 1 , x 0 )). If x 2 k H( x 2 k 1 ) is defined, then x 2 k + 1 G( x 2 k ) is such that p( x 2 k + 1 , x 2 k )φ(p( x 2 k , x 2 k 1 )), and, similarly, x 2 k + 2 H( x 2 k + 1 ) satisfies p( x 2 k + 2 , x 2 k + 1 )φ(p( x 2 k + 1 , x 2 k )). Thus, for a n =p( x n + 1 , x n ), we have a n + 1 φ( a n ), nN, and ( a n ) n N converges to zero as φ Φ P . □

As was shown in Example 14, conditions (12), (14) are too weak to guarantee the existence of a fixed point, even for φ Φ M . The next two theorems, with stronger assumptions, are general results.

Theorem 21 Let (X,p) be a 0-complete partial metric space, and let F:X 2 X be a mapping with 0-closed graph. Assume that for a φ Φ P there exists a sequence ( x n ) n N in X such that

sup k N p( x n + 1 + k , x n + 1 ) a n + 1 φ( a n ),nN, for a sequence  ( a n ) n N ,
(17)

and lim n p(F( x n ), x n + 1 )=0. Then x= lim n x n in (X,d) is a fixed point of F, and xKerp.

Proof Clearly, ( a n ) n N converges to zero as φ Φ P . Therefore, ( x n ) n N is a Cauchy sequence (see (2)) and it converges in (X,d), say, to xKerp (X is 0-complete). There exist y n F( x n ) such that lim n p( y n , x n + 1 )=0. Now, from

p( y n , x n )p( y n , x n + 1 )+p( x n + 1 , x n )

and condition (8) it follows that xF(x). □

Theorem 21, with φ(α)=kα, α0 (for k<1), is an extension of the Nadler theorem on multivalued contractions [[14], Theorem 5].

Now, Theorem 21 and Theorem 11 yield the following.

Theorem 22 Let (X,p) be a 0-complete partial metric space, and let be a family of mappings X 2 X with 0-closed graphs. Assume that φ Φ P ; some and at least all different G,HF satisfy (12). If for an FF there exists a sequence ( x n ) n N such that lim n p(F( x n ), x n + 1 )=0 and (17) holds, then all members of have the same nonempty set of fixed points; this set is closed in (X,d) and contained in Kerp.

A simple consequence of Theorem 21 is the following one.

Theorem 23 Let (X,p) be a 0-complete partial metric space, and let f:XX be a mapping with a 0-closed graph (e.g., f:(X,d)(X,d) is continuous). Assume that for a φ Φ P and x 0 the following condition is satisfied:

sup k N p ( f n + 1 + k ( x 0 ) , f n + 1 ( x 0 ) ) a n + 1 φ ( a n ) , n N , for a sequence  ( a n ) n N .
(18)

Then x= lim n f n ( x 0 ) in (X,d) is a fixed point of f, and xKerp.

Proof For x n = f n ( x 0 ) (and x n + 1 = f n + 1 ( x 0 )), we have (see (2))

0 lim n p ( f ( x n ) , x n + 1 ) lim n p( x n + 1 , x n )=0,

and condition (17) holds. □

The next proposition shows that Theorem 23 is related to the well-known theorem of Matkowski [[2], Theorem 1.2, p.8].

Proposition 24 Let (X,p) be a partial metric space, and let f:XX be a bounded mapping satisfying

p ( f ( y ) , f ( x ) ) φ ( p ( y , x ) ) ,x,yX
(19)

for a nondecreasing mapping φ:[0,)[0,). Then condition (18) holds.

Proof Let us adopt x n = f n ( x 0 ) and a n = sup k N p( x n + k , x n ), nN. For each k,nN, we have

p ( x n + 1 + k , x n + 1 ) = p ( f ( x n + k ) , f ( x n ) ) φ ( p ( x n + k , x n ) ) φ ( sup k N p ( x n + k , x n ) ) = φ ( a n )

as φ is nondecreasing, and we get (18). □

Let (X,p) be a partial metric space, and let f:XX be a mapping. Let us recall the conditions used by Romaguera in [15]:

p ( f ( y ) , f ( x ) ) φ ( m f ( y , x ) ) ,x,yX
(20)

for

m f (y,x)=max { p ( y , x ) , p ( f ( y ) , y ) , p ( f ( x ) , x ) } ,

and

p ( f ( y ) , f ( x ) ) φ ( p f ( y , x ) ) ,x,yX
(21)

for

p f (y,x)=max { p ( y , x ) , p ( f ( y ) , y ) , p ( f ( x ) , x ) , [ p ( f ( y ) , x ) + p ( f ( x ) , y ) ] / 2 } .

For nondecreasing φ, Ćirić’s condition (21) is more general than (20), and (20) is more general than condition (19). All these conditions are used to prove that f has a fixed point.

Let us note that φ(0) for conditions (19), (20), (21) can be arbitrary as p(f(x),x)=0 on the right-hand side of any of these inequalities (for y=f(x)) means that x is a fixed point of f.

Many fixed point theorems use more sophisticated conditions than (19), (20), (21) (see, e.g., [16], [17]) or the spaces under consideration have a richer structure (see, e.g., [9], [18]). We are interested in extending the most classical results.

The subsequent two lemmas are proved for condition (21), and the reasonings for (20) and (19) as well can be easily deduced. The next lemma (for condition (21)) has much in common with Romaguera’s Lemma 1 and Lemma 2 [15].

Lemma 25 Let (X,p) be a partial metric space, and let f:XX be a mapping satisfying condition (19), (20), or (21) for a φ Φ P . Then, for any xX, the condition

p ( f 2 ( x ) , f ( x ) ) φ ( p ( f ( x ) , x ) )

is satisfied, and lim n p( f n + 1 (x), f n (x))=0.

Proof For notational simplicity, let us adopt x n = f n (x), nN. For y= x 1 , condition (21) has the following form:

p( x 2 , x 1 )φ ( p f ( x 1 , x ) ) =φ ( max { p ( x 1 , x ) , p ( x 2 , x 1 ) , [ p ( x 2 , x ) + p ( x 1 , x 1 ) ] / 2 } ) .

We have

p ( x 2 , x ) + p ( x 1 , x 1 ) p ( x 2 , x 1 ) + p ( x 1 , x ) p ( x 1 , x 1 ) + p ( x 1 , x 1 ) = p ( x 2 , x 1 ) + p ( x 1 , x ) ,

and therefore,

max { p ( x 2 , x 1 ) , p ( x 1 , x ) } m f ( x 1 ,x) p f ( x 1 ,x)max { p ( x 2 , x 1 ) , p ( x 1 , x ) }

holds, i.e., p f (f(x),x)=max{p( f 2 (x),f(x)),p(f(x),x)} (Lemma 1 [15]). This last equality and condition (21) for p( x 1 ,x)<p( x 2 , x 1 ) yield

0<p( x 2 , x 1 )φ ( p f ( x 1 , x ) ) =φ ( p ( x 2 , x 1 ) ) ,

i.e., p( x 2 , x 1 )=p( f 2 (x),f(x))=0 (φΦ). This contradiction proves that p( x 2 , x 1 )p( x 1 ,x) must hold, and then condition (21) yields p( f 2 (x),f(x))φ(p(f(x),x)) (see Romaguera’s Lemma 2 [15]). Now, it is clear that for arbitrary x 0 X and x n = f n ( x 0 ) the sequence ( a n ) n N , where a n =p( x n + 1 , x n ), nN, converges to zero as φ Φ P and a n + 1 φ( a n ) holds. □

The next lemma is also helpful in proving fixed point theorems.

Lemma 26 Let (X,p) be a 0-complete partial metric space, and let f:XX be a mapping satisfying condition (19), (20), or (21) for a φΦ. If for x n = f n ( x 0 ), lim m , n p( x n , x m )=0 holds, then ( x n ) n N converges in (X,d) to a unique fixed point of f, and this point belongs to Kerp.

Proof From lim m , n p( x n , x m )=0 it follows that ( x n ) n N converges to a point xKerp in (X,d) (Corollary 4). We get

p ( f ( x ) , x ) p ( x n + 1 , f ( x ) ) + p ( x n + 1 , x ) φ ( p f ( x n , x ) ) + p ( x n + 1 , x ) = φ ( max { p ( x n , x ) , p ( x n + 1 , x n ) , p ( f ( x ) , x ) , [ p ( x n + 1 , x ) + p ( f ( x ) , x n ) ] / 2 } ) + p ( x n + 1 , x ) .

Suppose 0<p(f(x),x). Then from

p ( f ( x ) , x n ) /2 [ p ( f ( x ) , x ) + p ( x , x n ) ] /2,

and lim n p(x, x n )= lim n p(x, x n + 1 )=0 it follows that

max { p ( x n , x ) , p ( x n + 1 , x n ) , p ( f ( x ) , x ) , [ p ( x n + 1 , x ) + p ( f ( x ) , x n ) ] / 2 } =p ( f ( x ) , x )

for large n. Now we get

0<p ( f ( x ) , x ) φ ( p ( f ( x ) , x ) ) ,

a contradiction. Clearly, p(f(x),x)=0 means that p(x,x)=p(f(x),f(x))=p(f(x),x)=0 (see (2)), and f(x)=x (see (7)). If x, y are fixed points of f, then p(y,y)p(y,x)φ(p(y,x)) (see (2), (21)) means that x=y. □

The next result extends Romaguera’s Theorem 4 [15], and consequently, an earlier celebrated result due to Matkowski [[2], Theorem 1.2, p.8].

Theorem 27 Let (X,p) be a 0-complete partial metric space, and let f:XX be a mapping satisfying condition (19) or (20) for a φ Φ P such that

lim sup β α φ(β)<α,α>0
(22)

holds (e.g., if φ is nondecreasing). Then f has a unique fixed point; if x=f(x), then xKerp and x= lim n f n ( x 0 ) in (X,d), x 0 X.

Proof In view of Lemmas 25, 26 it is sufficient to prove that ( x n ) n N is a Cauchy sequence for x n = f n ( x 0 ), nN. Suppose that there are infinitely many k,nN such that p( f n + 1 + k ( x 0 ), f k ( x 0 ))ϵ>0. Let n=n(k)>0 be the smallest numbers satisfying this inequality. For simplicity let us adopt x= f k ( x 0 ) and x n = f n (x), nN. We have

ϵp( x n + 1 ,x)p( x n + 1 , x n )+p( x n ,x)<p( x n + 1 , x n )+ϵ,

which for n=n(k) means that

lim k p( x n + 1 ,x)= lim k p( x n ,x)=ϵ

as we have lim k p( x n + 1 , x n )= lim k p( x 1 ,x)=0. Now, for y= x n , condition (20) yields

ϵ p ( x n + 1 , x ) p ( x n + 1 , x 1 ) + p ( x 1 , x ) φ ( m f ( x n , x ) ) + p ( x 1 , x ) = φ ( max { p ( x n , x ) , p ( x n + 1 , x n ) , p ( x 1 , x ) } ) + p ( x 1 , x ) ,

and we obtain (from (19) as well)

ϵφ ( p ( x n , x ) ) +p( x 1 ,x)

for large k. Now, p( x n ,x)<ϵ, lim k p( x n ,x)=ϵ, and condition (22) yield

ϵ lim sup k φ ( p ( x n , x ) ) <ϵ,

a contradiction. Therefore, ( x n ) n N is a Cauchy sequence. □

Let φ 1 , φ 2 Φ 0 be continuous mappings. Then φ Φ 0 such that φ= φ 1 on Q[0,) and φ= φ 2 on [0,)Q is a member of Φ P and φ satisfies conditions (15), (22); clearly, φ does not necessarily belong to Φ M or to Φ B W .

Paesano and Vetro [9] have proved some theorems on coincidences and common fixed points. Maybe Theorem 27 can be extended to that case.

The next result extends Romaguera’s Theorem 3 [15], and consequently, an earlier celebrated result due to Boyd-Wong [[1], Theorem 1]. Let us recall that φ(0)=0 does not spoil the generality of condition (21). The proof is a modification of the one presented by Romaguera.

Theorem 28 Let (X,p) be a 0-complete partial metric space, and let f:XX be a mapping satisfying condition (21) for a mapping φ Φ 0 and satisfying (15) (e.g., with φ upper semicontinuous from the right). Then f has a unique fixed point; if x=f(x), then xKerp and x= lim n f n ( x 0 ) in (X,d), x 0 X.

Proof We follow the initial part of the proof of Theorem 27 preceding the sentence with condition (20) (Proposition 16 yields φ Φ P ). For large kN and n=n(k), we obtain

ϵ p ( x n + 1 , x ) p f ( x n + 1 , x ) = max { p ( x n + 1 , x ) , [ p ( x n + 2 , x ) + p ( x n , x 1 ) ] / 2 } max { p ( x n + 1 , x ) , [ p ( x n + 2 , x n + 1 ) + p ( x n + 1 , x ) + p ( x n , x ) + p ( x 1 , x ) ] / 2 } ,

and lim k p f ( x n + 1 ,x)=ϵ. Now, condition (21) yields

ϵ p ( x n + 1 , x ) p ( x n + 1 , x n + 2 ) + p ( x n + 2 , x 1 ) + p ( x 1 , x ) p ( x n + 1 , x n + 2 ) + φ ( p f ( x n + 1 , x ) ) + p ( x 1 , x ) ,

and we obtain (see (15))

ϵ lim sup k φ ( p f ( x n + 1 , x ) ) <ϵ,

a contradiction. Therefore, ( x n ) n N is a Cauchy sequence. □

The next lemma enables us to extend the preceding two theorems.

Lemma 29 Let f:XX be a mapping such that f t for tN has a unique fixed point, say, x. Then x is the unique fixed point of f. If, in addition, x lim n ( f t ) n ( x 0 ), x 0 X, then x lim n f n ( x 0 ), x 0 X holds.

Proof If x is a fixed point of f t , then f t (f(x))=f( f t (x))=f(x) means that f(x) is a fixed point of f t and the uniqueness yields f(x)=x. If x,yX are fixed points of f, then we get x= f t (x), y= f t (y), and x=y as f t has a unique fixed point. If x lim n ( f t ) n ( x 0 ) holds for each x 0 X, then we also obtain x lim n ( f t ) n (f( x 0 )) lim n ( f t ) n ( f t 1 ( x 0 )), which means that x lim n f n ( x 0 ). □

Theorem 30 Let (X,p) be a 0-complete partial metric space, and let f:XX be a mapping satisfying condition (19) or (20) with f replaced by f t for a tN, and a φ Φ P having property (22) (e.g., with φ nondecreasing). Then f has a unique fixed point; if x=f(x), then xKerp and x= lim n f n ( x 0 ) in (X,d), x 0 X.

Proof Clearly, all the assumptions of Theorem 27 are satisfied for f replaced by f t . Now, we apply Lemma 29. □

The next theorem is a consequence of Theorem 28 and of Lemma 29.

Theorem 31 Let (X,p) be a 0-complete partial metric space, and let f:XX be a mapping satisfying the condition

p ( f t ( y ) , f t ( x ) ) max { p ( y , x ) , p ( f t ( y ) , y ) , p ( f t ( x ) , x ) , [ p ( f t ( y ) , x ) + p ( f t ( x ) , y ) ] / 2 }
(23)

for tN, and a mapping φ Φ 0 satisfying (15) (e.g., with φ upper semicontinuous from the right). Then f has a unique fixed point; if x=f(x), then xKerp and x= lim n f n ( x 0 ) in (X,d), x 0 X.

Now, we present the respective versions of conditions (19), (20), (21) for a multivalued mapping F:X 2 X , where P is defined by (11).

P ( F ( y ) , F ( x ) ) φ ( p ( y , x ) ) ,x,yX,
(24)
P ( F ( y ) , F ( x ) ) φ ( m F ( y , x ) ) ,x,yX
(25)

for

m F (y,x)=max { p ( y , x ) , p ( F ( y ) , y ) , p ( F ( x ) , x ) } ,

and

P ( F ( y ) , F ( x ) ) φ ( p F ( y , x ) ) ,x,yX
(26)

for

p F (y,x)=max { p ( y , x ) , p ( F ( y ) , y ) , p ( F ( x ) , x ) , [ p ( F ( y ) , x ) + p ( F ( x ) , y ) ] / 2 } .

For nondecreasing φ, condition (26) is more general than (25), and (25) is more general than condition (24).

The subsequent two lemmas are proved for condition (26), and the reasonings for (25) and (24) as well can be easily deduced.

At first, let us present the following extension of Lemma 25.

Lemma 32 Let (X,p) be a partial metric space, and let F:X 2 X be a mapping. Then

m F (y,x)= p F (y,x)=max { p ( y , x ) , p ( F ( y ) , y ) } ,xX,yF(x)
(27)

holds, and condition (26) for any φΦ yields p F (y,x)=p(y,x), xX, yF(x) ( m F in place of p F for (25)), and

p ( F ( y ) , y ) φ ( p ( y , x ) ) ,xX,yF(x).
(28)

If φ Φ P and

for each  y F ( x ) ,  there exists  z F ( y ) such that  p ( z , y ) φ ( p ( y , x ) ) , x X
(29)

holds (e.g., F satisfies (28) and is compact valued), then there exists x n + 1 F( x n ), nN, such that

p( x n + 2 , x n + 1 )φ ( p ( x n + 1 , x n ) ) ,nN,and  lim n p( x n + 1 , x n )=0.

Proof For yF(x), we have (see (2))

p ( F ( y ) , x ) + p ( F ( x ) , y ) = inf { p ( z , x ) : z F ( y ) } + inf { p ( z , y ) : z F ( x ) } inf { p ( z , y ) + p ( y , x ) p ( y , y ) : z F ( y ) } + p ( y , y ) = p ( F ( y ) , y ) + p ( y , x ) .

Consequently, we get

m F ( y , x ) p F ( y , x ) max { p ( y , x ) , p ( F ( y ) , y ) , p ( F ( x ) , x ) , [ p ( F ( y ) , y ) + p ( y , x ) ] / 2 } max { p ( y , x ) , p ( F ( y ) , y ) }

as p(F(x),x)p(y,x), yF(x) holds. Clearly,

max { p ( y , x ) , p ( F ( y ) , y ) } m F (y,x) p F (y,x)

is satisfied and hence we get (27). Suppose p(y,x)<p(F(y),y) for a yF(x). Then (see (26))

p ( F ( y ) , y ) P ( F ( y ) , F ( x ) ) φ ( p F ( y , x ) ) =φ ( p ( F ( y ) , y ) )

means that p(F(y),y)=0 (as φΦ), and p(y,x)=p(F(y),y). This contradiction proves (28).

Let x 1 X, x 2 F( x 1 ) be arbitrary. If x n + 1 , x n are known, then let x n + 2 F( x n + 1 ) be such that (see (29) or (28) and Proposition 12) p( x n + 2 , x n + 1 )φ(p( x n + 1 , x n )). If p Φ P , then for a n =p( x n + 1 , x n ) we get lim n a n =0. □

Now, let us prove an analog of Lemma 26.

Lemma 33 Let (X,p) be a 0-complete partial metric space, and let F:X 2 X be a mapping with 0-closed graph satisfying condition (24), (25), or (26) for a φΦ. If lim n p(F( x n ), x n + 1 )=0 and lim m , n p( x n , x m )=0, then ( x n ) n N converges in (X,d) to a fixed point of F, and this point belongs to Kerp.

Proof From lim m , n p( x n , x m )=0 it follows that ( x n ) n N converges to a point xKerp in (X,d) (Corollary 4). We get

p ( F ( x ) , x ) p ( F ( x ) , x n + 1 ) + p ( x n + 1 , x ) P ( F ( x ) , F ( x n ) ) + p ( x n + 1 , x ) φ ( p F ( x n , x ) ) + p ( x n + 1 , x ) = φ ( max { p ( x n , x ) , p ( x n + 1 , x n ) , p ( F ( x ) , x ) , [ p ( x n + 1 , x ) + p ( F ( x ) , x n ) ] / 2 } ) + p ( x n + 1 , x ) .

Suppose 0<p(F(x),x). Then from

p ( F ( x ) , x n ) /2 [ p ( F ( x ) , x ) + p ( x , x n ) ] /2,

and lim n p(x, x n )= lim n p(x, x n + 1 )=0 it follows that

max { p ( x n , x ) , p ( x n + 1 , x n ) , p ( F ( x ) , x ) , [ p ( x n + 1 , x ) + p ( F ( x ) , x n ) ] / 2 } =p ( F ( x ) , x )

for large n. Now we get

0<p ( F ( x ) , x ) φ ( p ( F ( x ) , x ) ) ,

a contradiction. Clearly, p(F(x),x)=0 means that xKerp (see (2)). There exist y n F( x n ) such that lim n p( y n , x n + 1 )=0, and now,

p( y n , x n )p( y n , x n + 1 )+p( x n + 1 , x n )

together with condition (8) yield xF(x). □

It should be noted that a partial metric p defines metric δ in the following way: δ(x,y)=0 iff x=y, and δ(x,y)=p(x,y) for xy. The topology of (X,δ) is clearly larger than the topology of (X,d) (see (7)). Moreover, (X,p) is 0-complete iff (X,δ) is complete [19], [[20], Proposition 2.1].

If the proof of a theorem is based on p( x n + 1 , x n )0, nN, then it works for (X,δ) and the theorem is an immediate consequence of the respective result (if known) for metric spaces. Numerous examples can be found in [20].

Let us add that Corollary 4 and Proposition 5 show that for a 0-complete partial metric space (X,p), if we prove that lim m , n p( x n , x m )=0, then the remaining part of the proof concerns the metric space (X,d) (see (7)). Let us also recall that Kerp with its partial metric topology is a closed metric subspace of (X,d) [[6], Lemma 2.2].

References

  1. Boyd DW, Wong JSW: On nonlinear contractions. Proc. Am. Math. Soc. 1969, 20: 458–464. 10.1090/S0002-9939-1969-0239559-9

    Article  MathSciNet  MATH  Google Scholar 

  2. Matkowski J: Integrable solutions of functional equations. Diss. Math. 1975, 127: 1–68.

    MathSciNet  MATH  Google Scholar 

  3. Matthews SG: Partial metric topology. Ann. N.Y. Acad. Sci. 1994, 728: 183–197. Proc. 8th Summer Conference on General Topology and Applications. 10.1111/j.1749-6632.1994.tb44144.x

    Article  MathSciNet  MATH  Google Scholar 

  4. Oltra S, Valero O: Banach’s fixed point theorem for partial metric spaces. Rend. Ist. Mat. Univ. Trieste 2004, 36: 17–26.

    MathSciNet  MATH  Google Scholar 

  5. O’Neill SJ: Partial metrics, valuations, and domain theory. Ann. N.Y. Acad. Sci. 1996, 806: 304–315. Proc. 11th Summer Conference on General Topology and Applications. 10.1111/j.1749-6632.1996.tb49177.x

    Article  MathSciNet  MATH  Google Scholar 

  6. Pasicki, L: Partial metric, fixed points, variational principles. Fixed Point Theory. (to appear)

  7. Romaguera S: A Kirk type characterization of completeness for partial metric spaces. Fixed Point Theory Appl. 2010., 2010: Article ID 493298

    Google Scholar 

  8. Paesano D, Vetro P: Suzuki’s type characterizations of completeness for partial metric spaces and fixed points for partially ordered metric spaces. Topol. Appl. 2012, 159: 911–920. 10.1016/j.topol.2011.12.008

    Article  MathSciNet  MATH  Google Scholar 

  9. Paesano D, Vetro P: Common fixed points in a partially ordered partial metric space. Int. J. Anal. 2013., 2013: Article ID 428561

    Google Scholar 

  10. Aydi H, Abbas M, Vetro C: Partial Hausdorff metric and Nadler’s fixed point theorem on partial metric spaces. Topol. Appl. 2012, 159: 3234–3242. 10.1016/j.topol.2012.06.012

    Article  MathSciNet  MATH  Google Scholar 

  11. Pasicki L: Towards lim. Topol. Appl. 2011, 158: 479–483. 10.1016/j.topol.2010.11.024

    Article  MathSciNet  MATH  Google Scholar 

  12. Ćirić L, Samet B, Aydi H, Vetro C: Common fixed points of generalized contractions on partial metric spaces. Appl. Math. Comput. 2011, 218: 2398–2406. 10.1016/j.amc.2011.07.005

    Article  MathSciNet  MATH  Google Scholar 

  13. Matkowski J: Fixed point theorems for mappings with a contractive iterate at a point. Proc. Am. Math. Soc. 1977, 62: 344–348. 10.1090/S0002-9939-1977-0436113-5

    Article  MathSciNet  MATH  Google Scholar 

  14. Nadler SB: Multi-valued contraction mappings. Pac. J. Math. 1969, 30: 475–488. 10.2140/pjm.1969.30.475

    Article  MathSciNet  MATH  Google Scholar 

  15. Romaguera S: Fixed point theorems for generalized contractions on partial metric spaces. Topol. Appl. 2012, 159: 194–199. 10.1016/j.topol.2011.08.026

    Article  MathSciNet  MATH  Google Scholar 

  16. Altun I, Acar Ö: Fixed point theorems for weak contractions in the sense of Berinde on partial metric spaces. Topol. Appl. 2012, 159: 2642–2648. 10.1016/j.topol.2012.04.004

    Article  MathSciNet  MATH  Google Scholar 

  17. Alghamdi MA, Shahzad N, Valero O: On fixed point theory in partial metric spaces. Fixed Point Theory Appl. 2012., 2012: Article ID 175

    Google Scholar 

  18. Alghamdi MA, Shahzad N, Valero O: Fixed point theorems in general metric spaces with applications to computer science. Fixed Point Theory Appl. 2013., 2013: Article ID 118

    Google Scholar 

  19. Hitzler P, Seda A Semantics, Studies in Informatic Series. In Mathematical Aspects of Logic Programming. Chapman & Hall, CRC Press, Boca Raton; 2011.

    Google Scholar 

  20. Haghi RH, Rezapour S, Shahzad N: Be careful on partial metric fixed point results. Topol. Appl. 2013, 160: 450–454. 10.1016/j.topol.2012.11.004

    Article  MathSciNet  MATH  Google Scholar 

Download references

Acknowledgements

The work has been supported by the Polish Ministry of Science and Higher Education.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Lech Pasicki.

Additional information

Competing interests

The author declares that they have no competing interests.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0), which permits use, duplication, adaptation, distribution, and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Pasicki, L. Fixed point theorems for contracting mappings in partial metric spaces. Fixed Point Theory Appl 2014, 185 (2014). https://doi.org/10.1186/1687-1812-2014-185

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/1687-1812-2014-185

Keywords