 Research
 Open Access
 Published:
Some applications of Caristi’s fixed point theorem in metric spaces
Fixed Point Theory and Applications volume 2016, Article number: 16 (2016)
Abstract
In this work, partial answers to Reich, Mizoguchi and Takahashi’s and AminiHarandi’s conjectures are presented via a light version of Caristi’s fixed point theorem. Moreover, we introduce the idea that many of known fixed point theorems can easily be derived from the Caristi theorem. Finally, the existence of bounded solutions of a functional equation is studied.
Introduction and preliminaries
In the literature, the Caristi fixed point theorem is known as one of the very interesting and useful generalizations of the Banach fixed point theorem for selfmappings on a complete metric space. In fact, the Caristi fixed point theorem is a modification of the εvariational principle of Ekeland ([1, 2]), which is a crucial tool in nonlinear analysis, in particular, optimization, variational inequalities, differential equations, and control theory. Furthermore, in 1977 Western [3] proved that the conclusion of Caristi’s theorem is equivalent to metric completeness. In the last decades, Caristi’s fixed point theorem has been generalized and extended in several directions (see e.g., [4, 5] and the related references therein).
The Caristi’s fixed point theorem asserts the following.
Theorem 1.1
([6])
Let \((X,d)\) be a complete metric space and let \(T:X\to X\) be a mapping such that
for all \(x\in X\), where \(\varphi:X\to[0,+\infty)\) is a lower semicontinuous mapping. Then T has at least a fixed point.
Let us recall some basic notations, definitions, and wellknown results needed in this paper. Throughout this paper, we denote by \(\mathbb {N}\) and \(\mathbb {R}\) the sets of positive integers and real numbers, respectively. Let \((X,d)\) be a metric space. Denote by \(\mathcal{CB}(X)\) the family of all nonempty, closed, and bounded subsets of X. A function \(\mathcal{H}:\mathcal{CB}(X)\times \mathcal{CB}(X)\rightarrow{}[0,\infty)\) defined by
is said to be the Hausdorff metric on \(\mathcal{CB}(X)\) induced by the metric d on X. A point v in X is a fixed point of a map T if \(v=Tv\) (when \(T:X\rightarrow X\) is a singlevalued map) or \(v\in Tv\) (when \(T:X\rightarrow\mathcal{CB}(X)\) is a multivalued map).
Let \((X,d)\) be a complete metric space and \(T :X\to X\) a map. Suppose there exists a function \(\phi: [0,+\infty) \to[0,+\infty)\) satisfying \(\phi (0) = 0\), \(\phi(s) < s\) for \(s > 0\), and suppose that ϕ is right upper semicontinuous such that
BoydWong [7] showed that T has a unique fixed point.
In 1972, Reich [8] introduced the following open problem.
Problem 1.1
Let \((X,d)\) be a complete metric space and let \(T:X\to\mathcal{CB}(X)\) be a multivalued mapping such that
for all \(x,y \in X\), where \(\mu: \mathbb {R}^{+}\rightarrow \mathbb {R}^{+}\) is continuous and increasing map such that \(\mu(t)< t\), for all \(t>0\). Does T have a fixed point?
Some partial answers to Problem 1.1 were given by Daffer et al. (1996) [9] and Jachymski (1998) [10]. In these works, the authors consider additional conditions on the mapping μ to find a fixed point.

Daffer et al. assumed that \(\mu: \mathbb {R}^{+}\rightarrow \mathbb {R}^{+}\)

is upper right semicontinuous,

\(\mu(t)< t\) for all \(t>0\), and

\(\mu(t)\leq tat^{b}\), where \(a>0\), \(1< b<2\) on some interval \([0,s]\), \(s>0\).


Jachymski assumed that \(\mu: \mathbb {R}^{+}\rightarrow \mathbb {R}^{+}\)

is superadditive, i.e., \(\mu(x+y)>\mu(x)+\mu(y)\), for all \(x,y\in \mathbb {R}^{+}\), and

\(t\mapsto t\mu(t)\) is nondecreasing.

In 1983, Reich [11], introduced another problem as follows.
Problem 1.2
Let \((X,d)\) be a complete metric space and let \(T:X\to\mathcal{CB}(X)\) be a mapping such that
for all \(x,y \in X\), where \(\eta:(0,+\infty)\to[0,1)\) is a mapping such that \(\limsup_{r\to t^{+}}\eta(r)<1\), for all \(r\in(0,+\infty)\). Does T have a fixed point?
In 1989, Mizoguchi and Takahashi [12], gave a partial answer to Problem 1.2 as follows.
Theorem 1.2
Let \((X,d)\) be a complete metric space and let \(T:X\to\mathcal{CB}(X)\) be a mapping such that
for all \(x,y \in X\), where \(\eta:(0,+\infty)\to[0,1)\) is a mapping such that \(\limsup_{r\to t^{+}}\eta(r)<1\), for all \(r\in[0,+\infty)\). Then T has a fixed point.
Another analogous open problem was raised in 2010 by AminiHarandi [13], which we state after the following notations.
In the following, \(\gamma:[0,+\infty)\to[0,+\infty)\) is subadditive, i.e. \(\gamma(x+y)\leq\gamma(x)+\gamma(y)\), for each \(x,y\in [0,+\infty)\), a nondecreasing continuous map such that \(\gamma^{1}(\{0\})=\{0\}\), and let Γ consist of all such functions. Also, let \(\mathcal{A}\) be the class of all maps \(\theta:[0,+\infty)\to[0,+\infty)\) for which there exists an \(\epsilon _{0} > 0\) such that
where \(\gamma\in\Gamma\).
Problem 1.3
Assume that \(T:X\to\mathcal{CB}(X)\) is a weakly contractive setvalued map on a complete metric space \((X,d)\), i.e.,
for all \(x,y\in X\), where \(\theta\in\mathcal{A}\). Does T have a fixed point?
The answer is yes if Tx is compact for every x (AminiHarandi [13], Theorem 3.3).
In this work, we show that many of the known Banach contraction generalizations can be deduced and generalized by Caristi’s fixed point theorem and its consequences. Also, partial answers to the mentioned open problems are given via our main results. For more details as regards a fixed point generalization of multivalued mappings we refer to [14].
Main result
In this section, we show that many of the known fixed point results can be deduced from the following light version of Caristi’s theorem.
Corollary 2.1
Let \((X,d)\) be a complete metric space, and let \(T:X\to X\) be a mapping such that
for all \(x,y\in X\), where \(\varphi:X\times X\to[0,\infty)\) is lower semicontinuous with respect to the first variable. Then T has a unique fixed point.
Proof
For each \(x\in X\), let \(y=Tx\) and \(\psi(x)=\varphi(x,Tx)\). Then for each \(x\in X\)
and ψ is a lower semicontinuous mapping. Thus, applying Theorem 1.1 leads us to conclude the desired result. To see the uniqueness of the fixed point suppose that u, v are two distinct fixed points for T. Then
Thus, \(u=v\). □
Corollary 2.2
([15], Banach contraction principle)
Let \((X,d)\) be a complete metric space and let \(T:X\to X\) be a mapping such that for some \(\alpha\in[0,1)\)
for all \(x,y\in X\). Then T has a unique fixed point.
Proof
Define \(\varphi(x,y)=\frac{d(x,y)}{1\alpha}\). Then (6) shows that
It means that
and so
and so by applying Corollary 2.1, one can conclude that T has a unique fixed point. □
Corollary 2.3
Let \((X,d)\) be a complete metric space and let \(T:X\to X\) be a mapping such that
where \(\eta:[0,+\infty)\to[0,\infty)\) is a lower semicontinuous mapping such that \(\eta(t)< t\), for each \(t>0\), and \(\frac{\eta(t)}{t}\) is a nondecreasing map. Then T has a unique fixed point.
Proof
Define \(\varphi(x,y)=\frac{d(x,y)}{1\frac{\eta(d(x,y))}{d(x,y)}}\), if \(x\neq y\) and otherwise \(\varphi(x,x)=0\). Then (10) shows that
It means that
Since \(\frac{\eta(t)}{t}\) is nondecreasing and \(d(Tx,Ty) < d(x,y)\),
and so by applying Corollary 2.1, one can conclude that T has a unique fixed point. □
The following results are the main results of this paper and play a crucial role to find the partial answers for Problem 1.1, Problem 1.2, and Problem 1.3. Comparing the partial answers for Reich’s problems, our answers include simple conditions. Also, the compactness condition on Tx is not needed.
Theorem 2.1
Let \((X,d)\) be a complete metric space, and let \(T:X\to\mathcal {CB}(X)\) be a nonexpansive mapping such that, for each \(x\in X\), and for all \(y\in Tx\), there exists \(z\in Ty\) such that
where \(\varphi:X\times X\to[0,\infty)\) is lower semicontinuous with respect to the first variable. Then T has a fixed point.
Proof
Let \(x_{0}\in X\) and let \(x_{1}\in Tx_{0}\). If \(x_{0}=x_{1}\) then \(x_{0}\) is a fixed point and we are through. Otherwise, let \(x_{1}\neq x_{0}\). By assumption there exists \(x_{2}\in Tx_{1}\) such that
Alternatively, one can choose \(x_{n}\in Tx_{n1}\) such that \(x_{n}\neq x_{n1}\) and find \(x_{n+1}\in Tx_{n}\) such that
which means that \(\{\varphi(x_{n1},x_{n})\}_{n}\) is a nonincreasing sequence, bounded below, so it converges to some \(r\geq0\). By taking the limit on both sides of (15) we have \(\lim_{n\to\infty }d(x_{n1},x_{n})=0\). Also, for all \(m,n\in \mathbb {N}\) with \(m>n\),
Therefore, by taking the limsup on both sides of (16) we have
It means that \(\{x_{n}\}\) is a Cauchy sequence and so it converges to \(u\in X\). Now we show that u is a fixed point of T. We have
By taking the limit on both sides of (17), we get \(d(x,Tx)=0\) and this means that \(x\in Tx\). □
The following theorem is a partial answer to Problem 1.1.
Theorem 2.2
Let \((X,d)\) be a complete metric space, and let \(T:X\to\mathcal{CB}(X)\) be a multivalued function such that
for all \(x,y \in X\), where \(\eta:[0,\infty)\rightarrow[0,\infty)\) is a lower semicontinuous map such that \(\eta(t)< t\), for all \(t\in (0,+\infty)\), and \(\frac{\eta(t)}{t}\) is nondecreasing. Then T has a fixed point.
Proof
Let \(x\in X\) and \(y\in Tx\). If \(y=x\) then T has a fixed point and the proof is complete, so we suppose that \(y\neq x\). Define
We have \({\mathcal{H}}(Tx,Ty)\leq\eta(d(x,y))<\theta(d(x,y))<d(x,y)\). Thus there exists \(\epsilon_{0}>0\) such that \(\theta(d(x,y))={\mathcal{H}}(Tx,Ty)+\epsilon_{0}\). So there exists \(z\in Ty\) such that
We again suppose that \(y\neq z\); therefore \(d(x,y)\theta(d(x,y))\leq d(x,y)d(y,z)\) or equivalently
since \(\frac{\theta(t)}{t}\) is also a nondecreasing function and \(d(y,z)< d(x,y)\) we get
Define \(\Phi(x,y)=\frac{d(x,y)}{1\frac{\theta(d(x,y))}{d(x,y)}}\) if \(x\neq y\), otherwise 0 for all \(x,y\in X\). It means that
Therefore, T satisfies (14) of Theorem 2.1 and so we conclude that T has a unique fixed point u and the proof is completed. □
The following theorem is a partial answer to Problem 1.2.
Corollary 2.4
([12], MizoguchiTakahashi’s type)
Let \((X,d)\) be a complete metric space and let \(T:X\to\mathcal{CB}(X)\) be a multivalued mapping such that
for all \(x,y\in X\), where \(\eta:[0,+\infty)\to[0,1)\) is a lower semicontinuous and nondecreasing mapping. Then T has a fixed point.
Proof
Let \(\theta(t)=\eta(t)t\), \(\theta(t)< t\) for all \(t\in R_{+}\), and \(\frac{\theta(t)}{t}=\eta(t)\) is a nondecreasing mapping. By the assumption \(d(Tx,Ty)\leq\eta(d(x,y))d(x,y)=\theta(d(x,y))\) for all \(x,y\in X\), therefore by Theorem 2.2 T has a fixed point. □
Note that if \(\eta:[0,\infty)\rightarrow[0,1)\) is a nondecreasing map then for all \(s\in[0,+\infty)\)
It means that Corollary 2.4 comes from the MizoguchiTakahashi’s results directly and here we deduce it from our results [12].
The following theorem is a partial answer to Problem 1.3.
Corollary 2.5
Let \((X,d)\) be a complete metric space, and let \(T:X\to\mathcal{CB}(X)\) be a multivalued function such that
for all \(x,y \in X\), where \(\theta:(0,\infty)\rightarrow(0,\infty)\) is an upper semicontinuous map such that, for all \(t\in(0,+\infty)\), \(\frac{\theta(t)}{t}\) is nonincreasing. Then T has a fixed point.
Proof
Let \(\eta(t)=t\theta(t)\), for each \(t>0\). Then \(\eta(t)< t\), for each \(t>0\), and \(\frac{\eta(t)}{t}=1\frac{\theta(t)}{t}\) is nondecreasing. Thus, the desired result is obtained by Theorem 2.2. □
Existence of bounded solutions of functional equations
Mathematical optimization is one of the fields in which the methods of fixed point theory are widely used. It is well known that dynamic programming provides useful tools for mathematical optimization and computer programming. In this setting, the problem of dynamic programming related to a multistage process reduces to solving the functional equation
where \(\eta: {\mathcal{Z}}\times {\mathcal{T}}\to {\mathcal{Z}}\), \(f:{\mathcal{Z}}\times {\mathcal{T}}\to \mathbb {R}\), and \(\Im: {\mathcal{Z}}\times {\mathcal{T}}\times \mathbb {R}\to \mathbb {R}\). We assume that \({\mathcal{M}}\) and \({\mathcal{N}}\) are Banach spaces, \({\mathcal{Z}}\subset {\mathcal{M}}\) is a state space, and \({\mathcal{T}}\subset {\mathcal{N}}\) is a decision space. The studied process consists of a state space, which is the set of the initial state, actions, and a transition model of the process, and a decision space, which is the set of possible actions that are allowed for the process.
Here, we study the existence of the bounded solution of the functional equation (20). Let \({\mathcal{B}}({\mathcal{Z}})\) denote the set of all bounded realvalued functions on W and, for an arbitrary \(h\in {\mathcal{B}}({\mathcal{Z}})\), define \(\Vert h\Vert = \sup_{x\in {\mathcal{Z}}} \vert h(x)\vert \). Clearly, \(({\mathcal{B}}(W),\Vert \cdot \Vert )\) endowed with the metric d defined by
for all \(h,k\in {\mathcal{B}}({\mathcal{Z}})\), is a Banach space. Indeed, the convergence in the space \({\mathcal{B}}({\mathcal{Z}})\) with respect to \(\Vert \cdot \Vert \) is uniform. Thus, if we consider a Cauchy sequence \(\{h_{n}\}\) in \({\mathcal{B}}({\mathcal{Z}})\), then \(\{h_{n}\}\) converges uniformly to a function, say \(h^{*}\), that is bounded and so \(h\in {\mathcal{B}}({\mathcal{Z}})\).
We also define \(S:{\mathcal{B}}({\mathcal{Z}})\to {\mathcal{B}}({\mathcal{Z}})\) by
for all \(h\in {\mathcal{B}}({\mathcal{Z}})\) and \(x\in {\mathcal{Z}}\).
We will prove the following theorem.
Theorem 3.1
Let \(S : {\mathcal{B}}({\mathcal{Z}})\to {\mathcal{B}}({\mathcal{Z}})\) be an upper semicontinuous operator defined by (22) and assume that the following conditions are satisfied:

(i)
\(f:{\mathcal{Z}}\times {\mathcal{T}}\to \mathbb {R}\) and \(\Im: {\mathcal{Z}}\times {\mathcal{T}}\times \mathbb {R}\to \mathbb {R}\) are continuous and bounded;

(ii)
for all \(h,k\in {\mathcal{B}}({\mathcal{Z}})\), if
$$ \begin{aligned} & 0< d(h,k)< 1 \quad \textit{implies}\quad \bigl\vert \Im\bigl(x,y,h(x)\bigr)\Im\bigl(x,y,k(x)\bigr)\bigr\vert \leq \frac {1}{2}d^{2}(h,k), \\ &d(h,k)\geq1 \quad \textit{implies}\quad \bigl\vert \Im\bigl(x,y,h(x)\bigr)\Im \bigl(x,y,k(x)\bigr)\bigr\vert \leq\frac{2}{3}d(h,k), \end{aligned} $$(23)where \(x\in {\mathcal{Z}}\) and \(y\in {\mathcal{T}}\). Then the functional equation (20) has a bounded solution.
Proof
Note that \(({\mathcal{B}}({\mathcal{Z}}),d)\) is a complete metric space, where d is the metric given by (21). Let μ be an arbitrary positive number, \(x\in {\mathcal{Z}}\), and \(h_{1}, h_{2} \in {\mathcal{B}}({\mathcal{Z}})\), then there exist \(y_{1}, y_{2} \in {\mathcal{T}}\) such that
Let \(\varrho:[0,\infty)\rightarrow[0,\infty)\) be defined by
Then we can say that (23) is equivalent to
for all \(h,k\in {\mathcal{B}}({\mathcal{Z}})\). It is easy to see that \(\varrho(t)< t\), for all \(t>0\), and \(\frac{\varrho(t)}{t}\) is a nondecreasing function.
Therefore, by using (24), (27), and (28), it follows that
Then we get
Analogously, by using (25) and (26), we have
Hence, from (29) and (30) we obtain
that is,
Since the above inequality does not depend on \(x\in {\mathcal{Z}}\) and \(\mu> 0\) is taken arbitrary, we conclude immediately that
so we deduce that the operator S is a ϱcontraction. Thus, due to the continuity of S, Theorem 2.2 applies to the operator S, which has a fixed point \(h^{*}\in {\mathcal{B}}({\mathcal{Z}})\), that is, \(h^{*}\) is a bounded solution of the functional equation (20). □
References
 1.
Ekeland, I: On the variational principle. J. Math. Anal. Appl. 47(2), 324353 (1974)
 2.
Ekeland, I: Nonconvex minimization problems. Bull. Am. Math. Soc. 1, 443474 (1979)
 3.
Weston, JD: A characterization of metric completeness. Proc. Am. Math. Soc. 64(1), 186188 (1977)
 4.
Agarwal, RP, Khamsi, MA: Extension of Caristi’s fixed point theorem to vector valued metric space. Nonlinear Anal. TMA 74, 141145 (2011). doi:10.1016/j.na.2010.08.025
 5.
Khamsi, MA: Remarks on Caristi’s fixed point theorem. Nonlinear Anal. TMA 71, 227231 (2009)
 6.
Caristi, J: Fixed point theorems for mappings satisfying inwardness conditions. Trans. Am. Math. Soc. 215, 241251 (1976)
 7.
Boyd, DW, Wong, JSW: On nonlinear contractions. Proc. Am. Math. Soc. 20, 458464 (1969)
 8.
Reich, S: Fixed points of contractive functions. Boll. Unione Mat. Ital. (4) 5, 2642 (1972)
 9.
Daffer, PZ, Kaneko, H, Li, W: On a conjecture of S. Reich. Proc. Am. Math. Soc. 124, 31593162 (1996)
 10.
Jachymski, JR: Caristi’s fixed point theorem and selections of setvalued contractions. J. Math. Anal. Appl. 227, 5567 (1998)
 11.
Reich, S: Some problem and results in fixed point theory. Contemp. Math. 21, 179187 (1983)
 12.
Mizoguchi, N, Takahashi, W: Fixed point theorems for multivalued mappings on complete metric spaces. J. Math. Anal. Appl. 141, 177188 (1989)
 13.
AminiHarandi, A: Some generalizations of Caristi’s fixed point theorem with applications to the fixed point theory of weakly contractive setvalued maps and the minimization problem. Nonlinear Anal. TMA 72, 46614665 (2010)
 14.
Ahmad, J, Hussain, N, Rahim Khan, A, Azam, A: Fixed point results for generalized multivalued contractions. J. Nonlinear Sci. Appl. 8, 909918 (2015)
 15.
Banach, S: Sur les opérations dans les ensembles abstraits et leur application aux équations intégrales. Fundam. Math. 3, 133181 (1922)
Acknowledgements
The first author would like to thank Prof. S. Mansour Vaezpour for valuable suggestions.
Author information
Additional information
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All of the authors have made equal contributions.
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided 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.
About this article
Received
Accepted
Published
DOI
MSC
 47H10
 54E05
Keywords
 Caristi’s fixed point theorem
 Hausdorff metric
 MizoguchiTakahashi
 Reich’s problem
 Boyd and Wong’s contraction