Posts by Cristian Alecsa

Abstract


In this article, a study of the fixed point problem for Ciric type multi-valued operators is presented. More precisely,  some variants of Ciric’s contraction principle for multi-valued operators, as well as a strict fixed point principle for Ciric type multi- valued will be given.

Authors

 Adrian Petruşel
Department of Mathematics, Babeş-Bolyai University, Academy of Romanian Scientists Independenţei Street Cluj-Napoca,

Keywords

metric spaces; fixed point; Ciric type generalized contraction; multi-valued weakly Picard operator; data dependence;Ulam-Hyers stability; strict fixed point; Ostrowski property; topological properties

Paper coordinates

C.D. Alecsa and A. Petrusel, Some variants of Ciric’s multi-valued contraction principle, Analele Universitatii de Vest, Timisoara Seria Matematica –Informatica LVII, 1, (2019), pp. 23–42, https://doi.org/10.2478/awutm-2019-0004

PDF

About this paper

Journal

Analele Universitatii de Vest, Timisoara Seria Matematica –Informatica

Publisher Name
Print ISSN
Online ISSN

google scholar link

[1] J. Andres and L. G´orniewicz, On the Banach contraction principle for multivalued mappings, Approximation, Optimization and Mathematical Economics (M. Lassonde-ed.), Physica-Verlag, Heidelberg, (2001), 1-23
[2] Boriceanu M., Fixed point theory for multivalued generalized contraction on a set with two b-metrics, Stud. Univ.Babe¸s-Bolyai, 54 (3), (2009)
[3] L. Ciric, Fixed points for generalized multi-valued contractions, Mat. Vesnik., 9, (1972), 265–272
[4] N.V. Dung and A. Petru¸sel, On iterated functions systems consisting of Kannan maps, Reich maps, Chatterjea type maps, and related results, J. Fixed Point Theory
Appl., 9, (2017), 2271–2285
[5] G. Chifu and G. Petru¸sel, Fixed point results for multivalued Hardy-Rogers contractions in b-metric spaces, Filomat, 31 (8), (2017), 2499–2507
[6] A.A. Harandi, Endpoints of set-valued contractions in metric spaces, Nonlinear Anal., 72, (2010), 132–134
[7] A.D. Rogers and G.E. Hardy, A generalization of fixed point theorem of Reich, Canad. Math. Bull., 16, (1973), 201–208
[8] N. Hussain, A.A. Harandi, and Y.J. Cho, Approximate endpoints for set-valued contractions in metric spaces, Fixed Point Theory Appl., 2010:614867, (2010), 1–13
[9] J.R. Jachymski, Caristi’s fixed point theorem and selections of set-valued contractions, J. Math. Anal. Appl., 227, (1998), 55–67
[10] T.A. Lazar, A. Petru¸sel, and N. Shahzad, Fixed points for non-self operators and domain invariance theorems, Nonlinear Anal., 70, (2009), 117–125
[11] T. Lazar, D. O’Regan, and A. Petru¸sel, Fixed points and homotopy results for
Ciric-type multivalued operators on a set with two metrics, ´ Bull. Korean Math. Soc., 45 (1), (2008), 67–73
[12] T. Lazar, G. Mot¸, G. Petru¸sel, and S. Szentesi, The theory of Reich’s fixed point theorem for multivalued operators, Fixed Point Theory Appl., 10, (2010)
[13] V.L. Lazar, Fixed point theory for multivalued ϕ-contractions, Fixed Point Theory Appl., 50, (2011)
[14] S.B. Nadler Jr, Multi-valued contraction mappings, Pacific J. Math., 30, (1969), 475–488
[15] T.P. Petru, A. Petru¸sel, and J.C. Yao, Ulam-Hyers stability of operatorial equations and inclusions via nonself operators, Taiwanese J. Math., 15 (5), (2011), 2195–
2212
[16] A. Petrusel, Some variants of the contraction principle for multi-valued operators, generalizations and applications, to appear
[17] A. Petru¸sel, Ciric type fixed point theorems, Stud. Univ. Babes-Bolyai, 59 (2), (2014), 233–245
[18] A. Petru¸sel and I.A. Rus, The theory of a metric fixed point theorem for multivalued operators, Fixed Point Theory and its Applications, Proc. Ninth International
Conference on Fixed Point Theory and Applications, Changhua, Taiwan, (L.J. Lin, A. Petru¸sel, H.K. Xu – Eds.), Yokohama Publ., (2010), 161–175
[19] A. Petrusel, I.A. Rus, and J.C. Yao, Well-posedness in the generalized sense of the fixed point problems for multivalued operators, Taiwanese J. Math., 11 (3),
(2007), 903–914
[20] A. Petru¸sel and G. Petru¸sel, Selection theorems for multivalued generalized contractions, Math. Morav., 9, (2005), 43–52
[21] A. Petru¸sel, I.A. Rus, and M.A. Serban, Basic problems of the metric fixed point theory and the relevance of a metric fixed point theorem for a multivalued
operator, J. Nonlinear Convex Anal., 15 (3), (2014), 493–513
[22] B. Prasad and R. Sahni, Endpoints of multivalued contraction operators, ISRN Mathematical Anal., 2013, (2013), Article ID 542302, 7 pages
[23] J.S. Raymond, Multivalued contractions, Set-Valued Anal., 2, (1994), 559–571
[24] S. Reich, Fixed points of contractive functions, Boll. Un. Math. Ital., 4 (5), (1972), 26–42

Paper (preprint) in HTML form

Some variants of Ćirić’s multi-valued contraction principle

Cristian Daniel Alecsa and Adrian Petruşel
Dedicated to the memory of Professor Ştefan Măruşter
Sažetak

In this article, a study of the fixed point problem for Ćirić type multi-valued operators is presented. More precisely, some variants of Ćirić’s contraction principle for multi-valued operators, as well as a strict fixed point principle for Ćirić type multivalued will be given.

DOI: 10.2478/awutm-2019-0004

Analele Universităţii de Vest, Timişoara
Seria Matematică - Informatică
LVII, 1, (2019), 23-42

AMS Subject Classification (2000). 47H10; 54H25
Keywords. metric spaces; fixed point; Ćirić type generalized contraction; multi-valued weakly Picard operator; data dependence; Ulam-Hyers stability; strict fixed point; Ostrowski property; topological properties

1 Introduction

The aim of this paper is to present a study on Ćirić type multi-valued operators. Following the approach given in [16], where the author considered some variants of the multivalued contraction principle given by Nadler [14], respectively a so-called strict multi-valued contraction principle, we will consider here the case of Ćirić type multi-valued operators, see [3].

We also notice that in [24] Reich developed some fixed point theorems for multi-valued generalized contractions. A fully comprehensive study on Reich operators was made in [12] by T. Lazăr et al. Also, qualitative properties, namely data dependence, Ulam-Hyers stability and so on, were studied
for the case of multi-valued φ\varphi-contractions by V.L. Lazăr in [13]. Moreover, C. Chifu and G. Petruşel in [5] studied qualitative properties concerning Hardy-Rogers multi-valued operators (see [7] for the single-valued case) in the framework of b-metric spaces, while T. Lazăr, D. O’Regan et al. [11] studied the case of multi-valued operators of Ćirić type defined on a set endowed with two metrics. Finally, we point out that in [2], M. Boriceanu studied existence and uniqueness of the fixed point and data dependence for multi-valued Ćirić type operators in the context of b-metric spaces. At the same time, Ćirić type multi-valued operators were are studied in [17] and [19].

Regarding terminology and basic concepts for fixed point problems related to multi-valued operators, we will follow the works [1],[9], [18] and [23]. Furthermore, for the approximation of strict fixed points (also called end-points) of multi-valued mappings, we refer to [6], [8] and [22]. Finally, regarding data dependence, multi-valued fractal operators, selections and qualitative properties for the fixed point inclusion and for multi-valued fractals, we will refer to [4], [10] and [20].

Led ( X,dX,d ) be a metric space. Denote by P(X)P(X) the family of all nonempty subsets of XX. Also, Pb(X)P_{b}(X) stands for the family of nonempty, bounded subsets of XX and Pcl(X)P_{cl}(X) the family of nonempty, closed subsets of XX. In a similar manner, by Pcp(X)P_{cp}(X) we refer to the family of nonempty, compact subsets of XX. From now on, B¯(x0,r)\bar{B}\left(x_{0},r\right) means the closure in (X,d)(X,d) of the ball B(x0,r)B\left(x_{0},r\right), where B(x0,r):={xXd(x0,x)<r}B\left(x_{0},r\right):=\left\{x\in X\mid d\left(x_{0},x\right)<r\right\} is the open ball with radius r>0r>0 and the center x0Xx_{0}\in X. By B~(x0;r):={xXd(x0,x)r}\widetilde{B}\left(x_{0};r\right):=\left\{x\in X\mid d\left(x_{0},x\right)\leq r\right\} we denote the closed ball centered in x0x_{0} with radius rr. We recall now some important functionals which will be used through the paper:

  • the gap functional D:P(X)×P(X)+,D(A,B):=infaA,bB{d(a,b)}D:P(X)\times P(X)\rightarrow\mathbb{R}_{+},D(A,B):=\inf_{a\in A,b\in B}\{d(a,b)\}.

  • the generalized Pompeiu-Haussdorf functional H:P(X)×P(X)+{+}H:P(X)\times P(X)\rightarrow\mathbb{R}_{+}\cup\{+\infty\}, where H(A,B)=max{supaAD(a,B),supbBD(b,A)}H(A,B)=\max\left\{\sup_{a\in A}D(a,B),\sup_{b\in B}D(b,A)\right\}.
    Furthermore, if T:XP(X)T:X\rightarrow P(X) is a multi-valued operator, then an element xXx\in X is a fixed point for TT if and only if xT(x)x\in T(x). We denote by FTF_{T} the set of all fixed points of the operator TT and by (SF)T(SF)_{T} the set of all strict fixed points of TT, where xXx\in X is a strict fixed point of TT (or an endpoint, or a stationary point) if and only if {x}=Tx\{x\}=Tx.
    For a multi-valued operator T:XP(Y)T:X\rightarrow P(Y) we can also define the following useful notions. The graph of the operator TT, defined by Graph(T):={(x,y)X×YyT(x)}\operatorname{Graph}(T):=\{(x,y)\in X\times Y\mid y\in T(x)\}, and the image of the set YP(X)Y\in P(X) will be denoted by T(Y):=xYT(x)T(Y):=\bigcup_{x\in Y}T(x). A single-valued mapping t:XYt:X\rightarrow Y is called a selection of TT if for each xXx\in X, we have that t(x)T(x)t(x)\in T(x).

We present now an important concept, which appears naturally by Nadler’s contraction principle. By [21], we recall here the notion of multi-valued weakly Picard operator.

Definition 1.1. Let ( X,dX,d ) be a metric space.
Consider T:XP(X)T:X\rightarrow P(X) be a multi-valued operator. By definition, TT is a multi-valued weakly Picard operator (briefly MWP operator) if for each xXx\in X and for each yT(x)y\in T(x), there exists a sequence (xn)n\left(x_{n}\right)_{n\in\mathbb{N}}, satisfying the following
(i) x0=xx_{0}=x and x1=yx_{1}=y,
(ii) xn+1T(xn)x_{n+1}\in T\left(x_{n}\right), for each nn\in\mathbb{N},
(iii) the sequence (xn)n\left(x_{n}\right)_{n\in\mathbb{N}} is convergent to a fixed point of TT.

Remark 1.1. A sequence (xn)n\left(x_{n}\right)_{n\in\mathbb{N}} satisfying conditions (i)(i) and (ii)(ii) is called a sequence of successive approximations of TT starting from (x,y)Graph(T)(x,y)\in\operatorname{Graph}(T). If T:XP(X)T:X\rightarrow P(X) is a MWP operator, then we define the operator T:Graph(T)P(FT)T^{\infty}:\operatorname{Graph}(T)\rightarrow P\left(F_{T}\right), by T(x,y):={zFTT^{\infty}(x,y):=\left\{z\in F_{T}\mid\right. there exists a sequence of successive approximations of T starting from ( x,y\mathrm{x},\mathrm{y} ) that converges to z}\}.

Furthermore, if ( X,dX,d ) is a metric space and T:XP(X)T:X\rightarrow P(X) a multi-valued operator, then TT is said to be closed if Graph(T)\operatorname{Graph}(T) is a closed set in X×XX\times X. By T1(x):=T(x),,Tn(x):=T(Tn1(x))T^{1}(x):=T(x),\ldots,T^{n}(x):=T\left(T^{n-1}(x)\right) we denote the iterates of the multi-valued mapping TT, while the set V0(Y;ε):={xXD(x,Y)<ε}V^{0}(Y;\varepsilon):=\{x\in X\mid D(x,Y)<\varepsilon\} is called the (open) ε\varepsilon-neighborhood of YP(X)Y\in P(X).

From [14], we shall recall some important lemmas that are used throughout the article.

Lemma 1.1. Let AA and BB from P(X)P(X) and q>1q>1. Then, for each aAa\in A, there exists bBb\in B, such that d(a,b)qH(A,B)d(a,b)\leq qH(A,B).

Lemma 1.2. Let AA and BB from P(X)P(X). Also, consider η>0\eta>0, such that
(i) for each aAa\in A, there exists bBb\in B, with d(a,b)ηd(a,b)\leq\eta,
(ii) for each bBb\in B, there exists aAa\in A, with d(a,b)ηd(a,b)\leq\eta.

Then H(A,B)ηH(A,B)\leq\eta.
Now, we recall the basic concepts for the qualitative properties of the fixed point inclusion and of the fixed point iteration. The first two definitions are related to well-posedness of the fixed point problem. For the concept of well-posedness, we let the reader follow [12] and [19].

Definition 1.2. Let ( X,dX,d ) be a metric space and T:YPcl(X)T:Y\rightarrow P_{cl}(X) be a multi-valued operator. Then the fixed point problem is well-posed for TT with
respect to the gap functional DD if and only if:
(i) FT={x}F_{T}=\left\{x^{*}\right\};
(ii) if (xn)X\left(x_{n}\right)\in X has the property that D(xn,T(xn))0D\left(x_{n},T\left(x_{n}\right)\right)\rightarrow 0, then xnxx_{n}\rightarrow x^{*}.

Definition 1.3. Let (X,d)(X,d) be a metric space, YP(X)Y\in P(X) and T:YPcl(X)T:Y\rightarrow P_{cl}(X) be a multi-valued operator. Then the fixed point problem is well-posed for TT with respect to the Pompeiu-Haussdorf functional HH if and only if:
(i) (SF)T={x}(SF)_{T}=\left\{x^{*}\right\};
(ii) if (xn)X\left(x_{n}\right)\in X is a sequence such that H(xn,Txn)0H\left(x_{n},Tx_{n}\right)\rightarrow 0, then xnxx_{n}\rightarrow x^{*}.

Now, the second important concept related to the fixed point problem is limit shadowing or Ostrowski property, which can be found in [12] and [13].

Definition 1.4. Let ( X,dX,d ) be a metric space and T:XP(X)T:X\rightarrow P(X) be a multi-valued operator. By definition, the multi-valued operator TT has the Ostrowski property, if FT={x}F_{T}=\left\{x^{*}\right\} and for any sequence (yn)nX\left(y_{n}\right)_{n\in\mathbb{N}}\subset X, such that D(yn+1,Tyn)0D\left(y_{n+1},Ty_{n}\right)\rightarrow 0, we have (yn)nx\left(y_{n}\right)_{n\in\mathbb{N}}\rightarrow x^{*}, as nn\rightarrow\infty.

We introduce now the notions of ψ\psi-MWP operator and of generalized Ulam-Hyers stabilites. For the study of generalized Ulam-Hyers stability we refer to [15].

Definition 1.5. Let ( X,dX,d ) be a metric space and T:XP(X)T:X\rightarrow P(X) be a MWP operator. Let ψ:++\psi:\mathbb{R}_{+}\rightarrow\mathbb{R}_{+}be continuous in 0 , increasing, such that ψ(0)=0\psi(0)=0. By definition, TT is ψ\psi-MWP operator, if there exists a selection t:Graph(T)FTt^{\infty}:\operatorname{Graph}(T)\rightarrow F_{T} of TT^{\infty}, such that d(x,t(x,y))ψ(d(x,y))d\left(x,t^{\infty}(x,y)\right)\leq\psi(d(x,y)), for each (x,y)Graph(T)(x,y)\in\operatorname{Graph}(T).

Definition 1.6. Let ( X,dX,d ) be a metric space and T:XP(X)T:X\rightarrow P(X). By definition, the fixed point inclusion

xT(x)x\in T(x) (1.1)

is called generalized Ulam-Hyers stable if and only if there exists an increasing, continuous in 0 function ψ:++\psi:\mathbb{R}_{+}\rightarrow\mathbb{R}_{+}, with ψ(0)=0\psi(0)=0, such that for every ε>0\varepsilon>0 and for each yXy^{*}\in X for which D(y,T(y))εD(y,T(y))\leq\varepsilon, there exists a solution xx^{*} a solution of the fixed point inclusion (1.1), such that d(y,x)ψ(ε)d\left(y^{*},x^{*}\right)\leq\psi(\varepsilon).

Definition 1.7. Let ( X,dX,d ) be a metric space and T:YP(X)T:Y\rightarrow P(X). By definition, the strict fixed point inclusion

{x}=T(x)\{x\}=T(x) (1.2)

is called generalized Ulam-Hyers stable if and only if there exists an increasing, continuous in 0 function ψ:++\psi:\mathbb{R}_{+}\rightarrow\mathbb{R}_{+}, with ψ(0)=0\psi(0)=0, such that for
every ε>0\varepsilon>0 and for each yXy^{*}\in X for which H(y,T(y))εH(y,T(y))\leq\varepsilon, there exists a solution xx^{*} a solution of the strict fixed point inclusion (1.2), such that d(y,x)ψ(ε)d\left(y^{*},x^{*}\right)\leq\psi(\varepsilon).

Finally, following [6], [8] and [22], we recall the last important concepts.
Definition 1.8. Let XX\neq\emptyset and T:XP(X)T:X\rightarrow P(X) be a multi-valued operator. Then, TT has the approximate endpoint property if infxXsupyTxd(x,y)=0\inf_{x\in X}\sup_{y\in Tx}d(x,y)=0.

2 Main results

The aim of this paper is to extend to the case of Ćirić type multi-valued generalized contractions, the results given in [16], where the author studied extended properties for the fixed point problem related to Nadler’s multivalued contractions through relevant metrical and topological properties. In the present section some variants of the multi-valued Ćirić principle are given. We shall enhance the classical result of Ćirić [3] with additional metrical and topological conclusions with respect to the fixed point problem.
Theorem 2.1 (An extended version of the Ćirić’s multi-valued contraction principle). Let ( X,dX,d ) be a complete metric space and T:XPcl(X)T:X\rightarrow P_{cl}(X) be a multi-valued α\alpha-Ćirić type operator, i.e., there exists α(0,1)\alpha\in(0,1), such that

H(T(x),T(y))αM(x,y), for each x,yXH(T(x),T(y))\leq\alpha\cdot M(x,y),\text{ for each }x,y\in X

where

M(x,y):={d(x,y),D(x,T(x)),D(y,T(y)),12[D(x,T(y))+D(y,T(x))]}M(x,y):=\left\{d(x,y),D(x,T(x)),D(y,T(y)),\frac{1}{2}[D(x,T(y))+D(y,T(x))]\right\}

Then, the following conclusions hold:
(a) there exists xFTx^{*}\in F_{T};
(b) for each (x,y)Graph(T)(x,y)\in\operatorname{Graph}(T), there exists a sequence (xn)n\left(x_{n}\right)_{n\in\mathbb{N}} of successive approximations for TT starting from (x,y)(x,y), convergent to a fixed point of TT;
(c) there exists a selection t:Graph(T)FTt^{\infty}:\operatorname{Graph}(T)\rightarrow F_{T} of TT^{\infty}, such that

d(x,t(x,y))11αd(x,y),(x,y)Graph(T)d\left(x,t^{\infty}(x,y)\right)\leq\frac{1}{1-\alpha}d(x,y),\forall(x,y)\in Graph(T)

(d) FTF_{T} is closed in ( X,dX,d );
(e) if (xn)n\left(x_{n}\right)_{n\in\mathbb{N}} is a sequence of successive approximations for TT, starting from a pair (x,y)G(x,y)\in G raph (T)(T), which converges to a fixed point x(x,y)x^{*}(x,y) of TT, then

d(xn,x)αn1αd(x,y),n;d\left(x_{n},x^{*}\right)\leq\frac{\alpha^{n}}{1-\alpha}d(x,y),\forall n\in\mathbb{N}^{*};

(f) if G:XPcl(X)G:X\rightarrow P_{cl}(X) is a Ćirić-type multi-valued operator with coefficient β\beta, and there exists η>0\eta>0, such that H(T(x),G(x))ηH(T(x),G(x))\leq\eta, for all xXx\in X, then H(FT,FG)ηmax{11α,11β};H\left(F_{T},F_{G}\right)\leq\eta\cdot\max\left\{\frac{1}{1-\alpha},\frac{1}{1-\beta}\right\};
(g) if Tn:XPcl(X)T_{n}:X\rightarrow P_{cl}(X) is a sequence of multi-valued α\alpha-Ćirić-type operators, with Tn(x)𝐻T(x)T_{n}(x)\xrightarrow{H}T(x) as nn\rightarrow\infty, uniformly with respect to xXx\in X, then

limnH(FTn,FT)=0\lim_{n\rightarrow\infty}H\left(F_{T_{n}},F_{T}\right)=0

(h) if there exists x0Xx_{0}\in X and r>0r>0, such that D(x0,T(x0))<(1α)rD\left(x_{0},T\left(x_{0}\right)\right)<(1-\alpha)r, then there exists xFTB(x0,r)x^{*}\in F_{T}\cap B\left(x_{0},r\right);
(i) if there exists x0Xx_{0}\in X and r>0r>0 such that δ(x0,T(x0))<(1α)r\delta\left(x_{0},T\left(x_{0}\right)\right)<(1-\alpha)r, then T:B~(x0,r)P(B~(x0,11αr))T:\tilde{B}\left(x_{0},r\right)\rightarrow P\left(\tilde{B}\left(x_{0},\frac{1}{1-\alpha}r\right)\right) and there exists xFTB(x0,r)x^{*}\in F_{T}\cap B\left(x_{0},r\right);
(j) if XX is a Banach space, UU an open subset of XX and T:UPcl(X)T:U\rightarrow P_{cl}(X) is a Ćirić multi-valued operator, then the associated multi-valued field G:UP(X),G(x):=xT(x)G:U\rightarrow P(X),G(x):=x-T(x) is open;
(k) there exists a Caristi selection of TT;
(m) if, additionally, T:XPcp(X)T:X\rightarrow P_{cp}(X), then the fixed point inclusion xT(x)x\in T(x) is generalized Ulam-Hyers stable;
(n) the multi-valued operator TT has the approximate fixed point property;
(o) if the multi-valued operator TT is lower semicontinuous, then it has the approximate endpoint property if and only if it has a unique strict fixed point;
(p) if α<12\alpha<\frac{1}{2}, then the fixed point set FTF_{T} is compact.
(q) if T:XPb,cl(X)T:X\rightarrow P_{b,cl}(X), then for each p>0p>0, one has H(Fp,FT)p1αH\left(F_{p}^{*},F_{T}\right)\leq\frac{p}{1-\alpha}, where Fp:={xXD(x,T(x))<p}F_{p}^{*}:=\{x\in X\mid D(x,T(x))<p\}.

Proof. (a), (b), (c) and (e) (In fact (a) and (b) means that TT is a MWP operator, while (a), (b) and (c) can be concise represented by saying that TT is a ψ\psi-MWP operator, with ψ(t)=11αt\psi(t)=\frac{1}{1-\alpha}t ).
Let x0Xx_{0}\in X and x1T(x0)x_{1}\in T\left(x_{0}\right) be arbitrary elements. Then H(T(x0),T(x1))αM(x0,x1)H\left(T\left(x_{0}\right),T\left(x_{1}\right)\right)\leq\alpha M\left(x_{0},x_{1}\right). Furthermore, consider q(1,1α)q\in\left(1,\frac{1}{\alpha}\right).
Now, for x1x_{1}, there exists x2T(x1)x_{2}\in T\left(x_{1}\right), such that d(x1,x2)qH(T(x0),T(x1))d\left(x_{1},x_{2}\right)\leq qH\left(T\left(x_{0}\right),T\left(x_{1}\right)\right), so d(x1,x2)qαM(x0,x1)d\left(x_{1},x_{2}\right)\leq q\alpha M\left(x_{0},x_{1}\right). We consider the following cases:
If M(x0,x1)=d(x0,x1)M\left(x_{0},x_{1}\right)=d\left(x_{0},x_{1}\right), then d(x1,x2)(qα)d(x0,x1)d\left(x_{1},x_{2}\right)\leq(q\alpha)d\left(x_{0},x_{1}\right).
If M(x0,x1)=D(x0,T(x0))d(x0),x1M\left(x_{0},x_{1}\right)=D\left(x_{0},T\left(x_{0}\right)\right)\leq d\left(x_{0}\right),x_{1}, then d(x1,x2)(qα)d(x0,x1)d\left(x_{1},x_{2}\right)\leq(q\alpha)d\left(x_{0},x_{1}\right).
If M(x0,x1)=D(x1,T(x1))d(x1,x2)M\left(x_{0},x_{1}\right)=D\left(x_{1},T\left(x_{1}\right)\right)\leq d\left(x_{1},x_{2}\right), then d(x1,x2)(qα)d(x1,x2)d\left(x_{1},x_{2}\right)\leq(q\alpha)d\left(x_{1},x_{2}\right), which is a contradiction, So M(x0,x1)M\left(x_{0},x_{1}\right) can not be d(x1,x2)d\left(x_{1},x_{2}\right).

Finally, if M(x0,x1)=12[D(x1,T(x0))+D(x0,T(x1))]M\left(x_{0},x_{1}\right)=\frac{1}{2}\left[D\left(x_{1},T\left(x_{0}\right)\right)+D\left(x_{0},T\left(x_{1}\right)\right)\right], then by using the fact that D(x1,T(x0))d(x1,x1)=0D\left(x_{1},T\left(x_{0}\right)\right)\leq d\left(x_{1},x_{1}\right)=0 and the fact that D(x0,T(x1))d(x0,x2)d(x0,x1)+d(x1,x2)D\left(x_{0},T\left(x_{1}\right)\right)\leq d\left(x_{0},x_{2}\right)\leq d\left(x_{0},x_{1}\right)+d\left(x_{1},x_{2}\right), it follows that M(x0,x1)12d(x0,x1)+12d(x1,x2)M\left(x_{0},x_{1}\right)\leq\frac{1}{2}d\left(x_{0},x_{1}\right)+\frac{1}{2}d\left(x_{1},x_{2}\right). So d(x1,x2)qα2d(x0,x1)+qα2d(x1,x2)d\left(x_{1},x_{2}\right)\leq\frac{q\alpha}{2}d\left(x_{0},x_{1}\right)+\frac{q\alpha}{2}d\left(x_{1},x_{2}\right). Then d(x1,x2)qα2qαd(x0,x1)d\left(x_{1},x_{2}\right)\leq\frac{q\alpha}{2-q\alpha}d\left(x_{0},x_{1}\right).
Since q(1,1α)q\in\left(1,\frac{1}{\alpha}\right), we get that d(x1,x2)(qα)d(x0,x1)d\left(x_{1},x_{2}\right)\leq(q\alpha)d\left(x_{0},x_{1}\right).
Let us denote by λ:=qα\lambda:=q\alpha. Then, by all the cases d(x1,x2)λd(x0,x1)d\left(x_{1},x_{2}\right)\leq\lambda d\left(x_{0},x_{1}\right).
Also, denote by ρn:=d(xn,xn+1)\rho_{n}:=d\left(x_{n},x_{n+1}\right), for each nn\in\mathbb{N}. By induction, we can construct a sequence (xn)n\left(x_{n}\right)_{n\in\mathbb{N}}, such that for xnT(xn1)x_{n}\in T\left(x_{n-1}\right), there exists xn+1T(xn)x_{n+1}\in T\left(x_{n}\right), for which ρnλρn1\rho_{n}\leq\lambda\rho_{n-1}, for each nn\in\mathbb{N}. Then ρnλnρ0\rho_{n}\leq\lambda^{n}\rho_{0}, so by triangle inequality d(xn,xn+p)λn1λp1λρ0d\left(x_{n},x_{n+p}\right)\leq\lambda^{n}\frac{1-\lambda^{p}}{1-\lambda}\rho_{0}. Taking nn\rightarrow\infty, it follows up that the sequence (xn)n\left(x_{n}\right)_{n\in\mathbb{N}} is Cauchy, so there exists xXx^{*}\in X, such that xnxx_{n}\rightarrow x^{*}. Furthermore, in the estimate d(xn,xn+p)λn1λp1λρ0d\left(x_{n},x_{n+p}\right)\leq\lambda^{n}\frac{1-\lambda^{p}}{1-\lambda}\rho_{0}. taking pp\rightarrow\infty, it follows that d(xn,x)λn1λd(x0,x1)d\left(x_{n},x^{*}\right)\leq\frac{\lambda^{n}}{1-\lambda}d\left(x_{0},x_{1}\right). Taking n=0n=0 and making q1q\searrow 1, it follows the estimate d(x,x)=d(x,t(x,y))11αd(x,y)d\left(x,x^{*}\right)=d\left(x,t^{\infty}(x,y)\right)\leq\frac{1}{1-\alpha}d(x,y), with yT(x)y\in T(x). Here, we denoted by x:=x0x:=x_{0} and y:=x1T(x0)y:=x_{1}\in T\left(x_{0}\right).
The final step is to show that xFTx^{*}\in F_{T}, i.e., to prove that D(x,T(x))=0D\left(x^{*},T\left(x^{*}\right)\right)=0. We have the following estimation:

D(x,T(x))d(x,xn+1)+H(T(xn),T(x))d(x,xn+1)+αM(xn,x).D\left(x^{*},T\left(x^{*}\right)\right)\leq d\left(x^{*},x_{n+1}\right)+H\left(T\left(x_{n}\right),T\left(x^{*}\right)\right)\leq d\left(x^{*},x_{n+1}\right)+\alpha M\left(x_{n},x^{*}\right).

Moreover, since

M(xn,x)max{d(xn,x),d(xn,xn+1),D(x,T(x))12[d(x,xn)+d(x,xn+1)+D(x,T(x))]}\begin{gathered}M\left(x_{n},x^{*}\right)\leq\max\left\{d\left(x_{n},x^{*}\right),d\left(x_{n},x_{n+1}\right),D\left(x^{*},T\left(x^{*}\right)\right)\right.\\ \left.\frac{1}{2}\left[d\left(x^{*},x_{n}\right)+d\left(x^{*},x_{n+1}\right)+D\left(x^{*},T\left(x^{*}\right)\right)\right]\right\}\end{gathered}

by letting nn\rightarrow\infty, we obtain that limnM(xn,x)D(x,T(x))\lim_{n\rightarrow\infty}M\left(x_{n},x^{*}\right)\leq D\left(x^{*},T\left(x^{*}\right)\right). This means that

D(x,T(x))αmax{D(x,T(x)),12D(x,T(x))}<D(x,T(x))D\left(x^{*},T\left(x^{*}\right)\right)\leq\alpha\max\left\{D\left(x^{*},T\left(x^{*}\right)\right),\frac{1}{2}D\left(x^{*},T\left(x^{*}\right)\right)\right\}<D\left(x^{*},T\left(x^{*}\right)\right)

and the conclusion follows.
(d) We know that FTP(X)F_{T}\in P(X). We shall show that FTF_{T} is closed in (X,d)(X,d). For this, let xnFTx_{n}\in F_{T}, such that xnxx_{n}\rightarrow x^{*}. So, for each n,D(xn,T(xn))=0n\in\mathbb{N},D\left(x_{n},T\left(x_{n}\right)\right)=0.

We shall show that xFTx^{*}\in F_{T}, i.e. xT(x)x^{*}\in T\left(x^{*}\right). Also, since the operator TT has closed values, then it is enough to show that D(x,T(x))=0D\left(x^{*},T\left(x^{*}\right)\right)=0. We have the following inequalities :

D(x,T(x))\displaystyle D\left(x^{*},T\left(x^{*}\right)\right) d(x,xn)+D(xn,T(x))d(x,xn)+H(T(xn),T(x))\displaystyle\leq d\left(x^{*},x_{n}\right)+D\left(x_{n},T\left(x^{*}\right)\right)\leq d\left(x^{*},x_{n}\right)+H\left(T\left(x_{n}\right),T\left(x^{*}\right)\right)
d(x,xn)+αM(x,xn)\displaystyle\leq d\left(x^{*},x_{n}\right)+\alpha M\left(x^{*},x_{n}\right)

We have the following cases:
If M(x,xn)=d(x,xn)M\left(x^{*},x_{n}\right)=d\left(x^{*},x_{n}\right), then D(x,T(x))(1+α)d(xn,x)0D\left(x^{*},T\left(x^{*}\right)\right)\leq(1+\alpha)d\left(x_{n},x^{*}\right)\rightarrow 0.
Furthermore, if M(x,xn)=D(xn,T(xn))=0M\left(x^{*},x_{n}\right)=D\left(x_{n},T\left(x_{n}\right)\right)=0, then D(x,T(x))d(x,xn)0D\left(x^{*},T\left(x^{*}\right)\right)\leq d\left(x^{*},x_{n}\right)\rightarrow 0. Moreover, if M(x,xn)=D(x,T(x))M\left(x^{*},x_{n}\right)=D\left(x^{*},T\left(x^{*}\right)\right), then we obtain that

D(x,T(x))11αd(x,xn)0D\left(x^{*},T\left(x^{*}\right)\right)\leq\frac{1}{1-\alpha}d\left(x^{*},x_{n}\right)\rightarrow 0

Finally, if M(x,xn)=12[D(xn,T(x))+D(x,T(xn))]12D(xn,T(x))+12d(x,xn)M\left(x^{*},x_{n}\right)=\frac{1}{2}\left[D\left(x_{n},T\left(x^{*}\right)\right)+D\left(x^{*},T\left(x_{n}\right)\right)\right]\leq\frac{1}{2}D\left(x_{n},T\left(x^{*}\right)\right)+\frac{1}{2}d\left(x^{*},x_{n}\right).
Also, D(xn,T(x))H(T(xn),T(x))αM(xn,x)α2D(xn,T(x))+α2d(x,xn)D\left(x_{n},T\left(x^{*}\right)\right)\leq H\left(T\left(x_{n}\right),T\left(x^{*}\right)\right)\leq\alpha M\left(x_{n},x^{*}\right)\leq\frac{\alpha}{2}D\left(x_{n},T\left(x^{*}\right)\right)+\frac{\alpha}{2}d\left(x^{*},x_{n}\right), so D(xn,T(x))α2αd(xn,x)D\left(x_{n},T\left(x^{*}\right)\right)\leq\frac{\alpha}{2-\alpha}d\left(x_{n},x^{*}\right). This implies that

D(x,T(x))d(x,xn)+α2αd(xn,x)=22αd(xn,x)0,nD\left(x^{*},T\left(x^{*}\right)\right)\leq d\left(x^{*},x_{n}\right)+\frac{\alpha}{2-\alpha}d\left(x_{n},x^{*}\right)=\frac{2}{2-\alpha}d\left(x_{n},x^{*}\right)\rightarrow 0,n\rightarrow\infty

Thus, by all cases xFTx^{*}\in F_{T}, so FTF_{T} is closed.
(f) By (a),(b),(c) and (e), we have that d(x,x)11αd(x,y)d\left(x,x^{*}\right)\leq\frac{1}{1-\alpha}d(x,y), where xx is an arbitrary element of XX and yT(x)y\in T(x), where xFTx^{*}\in F_{T}. Taking x=yFGx=y^{*}\in F_{G}, then we obtain that d(x,y)11αd(y,y)d\left(x^{*},y^{*}\right)\leq\frac{1}{1-\alpha}d\left(y,y^{*}\right), where yT(y)y\in T\left(y^{*}\right). Furthermore, since yT(y)y\in T\left(y^{*}\right) is arbitrary, we can make the following assertion: for yFGy^{*}\in F_{G}, there exists yT(y)y\in T\left(y^{*}\right), such that d(y,y)H(G(y),T(y))ηd\left(y,y^{*}\right)\leq H\left(G\left(y^{*}\right),T\left(y^{*}\right)\right)\leq\eta, so d(x,y)η1αd\left(x^{*},y^{*}\right)\leq\frac{\eta}{1-\alpha}.
Now, also from the global principle of the existence of the fixed point of GG, we get that d(x,x)11βd(x,y)d\left(x,x^{*}\right)\leq\frac{1}{1-\beta}d(x,y), with xFG,xx^{*}\in F_{G},x is an arbitrary element of XX and yGxy\in Gx.
Taking x=yFTx=y^{*}\in F_{T}, then we obtain that d(x,y)11βd(y,y)d\left(x^{*},y^{*}\right)\leq\frac{1}{1-\beta}d\left(y,y^{*}\right), where yG(y)y\in G\left(y^{*}\right).
As in the first case, since yG(y)y\in G\left(y^{*}\right) is arbitrary, then for yFTy^{*}\in F_{T}, there exists

Vol. LVII (2019) Some variants of Ćirić’s multi-valued contraction principle 31
yG(y)y\in G\left(y^{*}\right), such that d(y,y)H(T(y),G(y))ηd\left(y^{*},y\right)\leq H\left(T\left(y^{*}\right),G\left(y^{*}\right)\right)\leq\eta. So d(x,y)η1βd\left(x^{*},y^{*}\right)\leq\frac{\eta}{1-\beta}. From the first case we get that for yFGy^{*}\in F_{G}, there exists xFTx^{*}\in F_{T}, such that

d(x,y)ηmax{11α,11β}d\left(x^{*},y^{*}\right)\leq\eta\cdot\max\left\{\frac{1}{1-\alpha},\frac{1}{1-\beta}\right\}

while from the second case we infer that for yFTy^{*}\in F_{T}, there exists xFGx^{*}\in F_{G}, such that

d(x,y)ηmax{11α,11β}d\left(x^{*},y^{*}\right)\leq\eta\cdot\max\left\{\frac{1}{1-\alpha},\frac{1}{1-\beta}\right\}

By Lemma 1.2, we get the conclusion H(FT,FG)ηmax{11α,11β}H\left(F_{T},F_{G}\right)\leq\eta\cdot\max\left\{\frac{1}{1-\alpha},\frac{1}{1-\beta}\right\}.
(g)(\mathrm{g}) Let ε>0\varepsilon>0 be an arbitrary fixed element. Since Tn(x)𝐻T(x)T_{n}(x)\xrightarrow{H}T(x) as nn\rightarrow\infty, uniformly for each xXx\in X, then for all xXx\in X, we have that limnH(Tnx,Tx)=0\lim_{n\rightarrow\infty}H\left(T_{n}x,Tx\right)=0.
This means that for ε>0\varepsilon>0, there exists N(ε)N(\varepsilon)\in\mathbb{N}, such that for each nN(ε)n\geq N(\varepsilon), we have that supxXH(Tn(x),T(x))<ε\sup_{x\in X}H\left(T_{n}(x),T(x)\right)<\varepsilon. From the conclusion (f) of data dependence, we have that for ε>0\varepsilon>0, there exists N(ε)N(\varepsilon)\in\mathbb{N}, such that for all nN(ε)n\geq N(\varepsilon), one has H(FTn,FT)<11αεH\left(F_{T_{n}},F_{T}\right)<\frac{1}{1-\alpha}\cdot\varepsilon. So, the conclusion is valid.
(h) Let s(0,r)s\in(0,r), such that B~(x0,s)B(x0,r)\tilde{B}\left(x_{0},s\right)\subset B\left(x_{0},r\right), with D(x0,T(x0))<(1α)s<(1α)rD\left(x_{0},T\left(x_{0}\right)\right)<(1-\alpha)s<(1-\alpha)r. Since D(x0,T(x0))<(1α)sD\left(x_{0},T\left(x_{0}\right)\right)<(1-\alpha)s, then there exists x1T(x0)x_{1}\in T\left(x_{0}\right), such that d(x0,x1)<(1α)s<sd\left(x_{0},x_{1}\right)<(1-\alpha)s<s, so x1B(x0,s)B~(x0,s)x_{1}\in B\left(x_{0},s\right)\subset\tilde{B}\left(x_{0},s\right). From the hypothesis, we have that H(T(x0),T(x1))αM(x0,x1)H\left(T\left(x_{0}\right),T\left(x_{1}\right)\right)\leq\alpha M\left(x_{0},x_{1}\right), where:

M(x0,x1)=\displaystyle M\left(x_{0},x_{1}\right)=
=max{d(x0,x1),D(x0,T(x0)),D(x1,T(x1)),12[D(x0,T(x1))+D(x1,T(x0))]}\displaystyle=\max\left\{d\left(x_{0},x_{1}\right),D\left(x_{0},T\left(x_{0}\right)\right),D\left(x_{1},T\left(x_{1}\right)\right),\frac{1}{2}\left[D\left(x_{0},T\left(x_{1}\right)\right)+D\left(x_{1},T\left(x_{0}\right)\right)\right]\right\}
max{d(x0,x1),D(x0,T(x0)),D(x1,T(x1)),12[d(x0,x1)+D(x1,T(x1))]}\displaystyle\leq\max\left\{d\left(x_{0},x_{1}\right),D\left(x_{0},T\left(x_{0}\right)\right),D\left(x_{1},T\left(x_{1}\right)\right),\frac{1}{2}\left[d\left(x_{0},x_{1}\right)+D\left(x_{1},T\left(x_{1}\right)\right)\right]\right\}
=max{d(x0,x1),D(x0,T(x0)),D(x1,T(x1))}\displaystyle=\max\left\{d\left(x_{0},x_{1}\right),D\left(x_{0},T\left(x_{0}\right)\right),D\left(x_{1},T\left(x_{1}\right)\right)\right\}
max{d(x0,x1),H(T(x0),T(x1))}\displaystyle\leq\max\left\{d\left(x_{0},x_{1}\right),H\left(T\left(x_{0}\right),T\left(x_{1}\right)\right)\right\}

We consider the following cases:
If the maximum is d(x0,x1)d\left(x_{0},x_{1}\right), then H(T(x0),T(x1))αd(x0,x1)H\left(T\left(x_{0}\right),T\left(x_{1}\right)\right)\leq\alpha d\left(x_{0},x_{1}\right).
If the maximum is H(T(x0),T(x1))H\left(T\left(x_{0}\right),T\left(x_{1}\right)\right), then, since α<1\alpha<1, we obtain a contradiction.
From the above cases, it follows that H(T(x0),T(x1))αd(x0,x1)H\left(T\left(x_{0}\right),T\left(x_{1}\right)\right)\leq\alpha d\left(x_{0},x_{1}\right). Since D(x1,T(x1))H(T(x0),T(x1))αd(x0,x1)<α(1α)sD\left(x_{1},T\left(x_{1}\right)\right)\leq H\left(T\left(x_{0}\right),T\left(x_{1}\right)\right)\leq\alpha d\left(x_{0},x_{1}\right)<\alpha(1-\alpha)s, then there exists x2T(x1)x_{2}\in T\left(x_{1}\right) for which d(x1,x2)<α(1α)sd\left(x_{1},x_{2}\right)<\alpha(1-\alpha)s.

Furthermore, by triangle inequality one can obtain d(x0,x2)d(x0,x1)+d(x1,x2)<(1α)s+α(1α)s=(1α2)s<sd\left(x_{0},x_{2}\right)\leq d\left(x_{0},x_{1}\right)+d\left(x_{1},x_{2}\right)<(1-\alpha)s+\alpha(1-\alpha)s=\left(1-\alpha^{2}\right)s<s, so x2T(x1)B~(x0,s)x_{2}\in T\left(x_{1}\right)\cap\tilde{B}\left(x_{0},s\right).
By induction, we can construct a sequence (xn)n\left(x_{n}\right)_{n\in\mathbb{N}}, with xnx_{n} from T(xn1)B~(x0,s)T\left(x_{n-1}\right)\cap\tilde{B}\left(x_{0},s\right), such that :
{xn+1T(xn), for each n,d(xn1,xn)αn1(1α)s, for each n,d(x0,xn)(1αn)s, for all n.\left\{\begin{array}[]{l}x_{n+1}\in T\left(x_{n}\right),\text{ for each }n\in\mathbb{N},\\ d\left(x_{n-1},x_{n}\right)\leq\alpha^{n-1}(1-\alpha)s,\text{ for each }n\in\mathbb{N}^{*},\\ d\left(x_{0},x_{n}\right)\leq\left(1-\alpha^{n}\right)s,\text{ for all }n\in\mathbb{N}.\end{array}\right.
It follows that the sequence (xn)n\left(x_{n}\right)_{n\in\mathbb{N}} is Cauchy, so there exists xB~(x0,s)x^{*}\in\tilde{B}\left(x_{0},s\right), such that xnxx_{n}\rightarrow x^{*}.
As in the proof of (a),(b),(c) and (e), one can show that xT(x)x^{*}\in T\left(x^{*}\right). Moreover, since xnB~(x0,s)x_{n}\in\tilde{B}\left(x_{0},s\right) and B~(x0,s)\tilde{B}\left(x_{0},s\right) is closed in XX, then xB~(x0,s)B(x0,r)x^{*}\in\tilde{B}\left(x_{0},s\right)\subset B\left(x_{0},r\right).
(i) Let uT(x0)u\in T\left(x_{0}\right). Then, by applying the triangle inequality, we get that d(z,x0)d(z,u)+d(u,x0)d\left(z,x_{0}\right)\leq d(z,u)+d\left(u,x_{0}\right), so d(z,x0)d(z,u)+δ(x0,T(x0))d\left(z,x_{0}\right)\leq d(z,u)+\delta\left(x_{0},T\left(x_{0}\right)\right). Now, taking infuT(x0)\inf_{u\in T\left(x_{0}\right)}, it follows that d(z,x0)D(z,T(x0))+δ(x0,T(x0))d\left(z,x_{0}\right)\leq D\left(z,T\left(x_{0}\right)\right)+\delta\left(x_{0},T\left(x_{0}\right)\right).
We first show that T(B~(x0,r))B~(x0,11αr)T\left(\tilde{B}\left(x_{0},r\right)\right)\subset\tilde{B}\left(x_{0},\frac{1}{1-\alpha}r\right).
Let yB~(x0,r)y\in\tilde{B}\left(x_{0},r\right). We will show that T(y)B~(x0,11αr)T(y)\subset\tilde{B}\left(x_{0},\frac{1}{1-\alpha}r\right).
So, take zT(y)z\in T(y). The aim is to show that zB~(x0,11αr)z\in\tilde{B}\left(x_{0},\frac{1}{1-\alpha}r\right), i.e., d(z,x0)11αrd\left(z,x_{0}\right)\leq\frac{1}{1-\alpha}r.
Then d(z,x0)D(z,T(x0))+δ(x0,T(x0))<H(T(y),T(x0))+(1α)rd\left(z,x_{0}\right)\leq D\left(z,T\left(x_{0}\right)\right)+\delta\left(x_{0},T\left(x_{0}\right)\right)<H\left(T(y),T\left(x_{0}\right)\right)+(1-\alpha)r. So d(z,x0)<αM(y,x0)+(1α)rd\left(z,x_{0}\right)<\alpha M\left(y,x_{0}\right)+(1-\alpha)r.
We know that:

M(y,x0)=max\displaystyle M\left(y,x_{0}\right)=\max {d(y,x0),D(x0,T(x0)),D(y,T(y)),\displaystyle\left\{d\left(y,x_{0}\right),D\left(x_{0},T\left(x_{0}\right)\right),D(y,T(y)),\right.
12[D(y,T(x0))+D(x0,T(y))]}.\displaystyle\left.\frac{1}{2}\left[D\left(y,T\left(x_{0}\right)\right)+D\left(x_{0},T(y)\right)\right]\right\}.

We also have d(y,x0)rd\left(y,x_{0}\right)\leq r and D(x0,T(x0))δ(x0,T(x0))<(1α)rrD\left(x_{0},T\left(x_{0}\right)\right)\leq\delta\left(x_{0},T\left(x_{0}\right)\right)<(1-\alpha)r\leq r. So, we obtain:

M(y,x0)max{r,D(y,T(y)),12[D(y,T(x0))+D(x0,T(y))]}M\left(y,x_{0}\right)\leq\max\left\{r,D(y,T(y)),\frac{1}{2}\left[D\left(y,T\left(x_{0}\right)\right)+D\left(x_{0},T(y)\right)\right]\right\}

We employ an analysis on the following cases :
If the maximum from the right hand side is rr, then d(z,x0)<αr+(1α)r=r<11αrd\left(z,x_{0}\right)<\alpha r+(1-\alpha)r=r<\frac{1}{1-\alpha}r.
If the maximum is D(y,T(y))D(y,T(y)), then d(z,x0)<αD(y,T(y))+(1α)rd\left(z,x_{0}\right)<\alpha D(y,T(y))+(1-\alpha)r. So
d(z,x0)<(1α)r+αd(y,z)(1α)r+αd(y,x0)+αd(z,x0)d\left(z,x_{0}\right)<(1-\alpha)r+\alpha d(y,z)\leq(1-\alpha)r+\alpha d\left(y,x_{0}\right)+\alpha d\left(z,x_{0}\right). This means that d(z,x0)<11αrd\left(z,x_{0}\right)<\frac{1}{1-\alpha}r.
Finally, if the maximum is 12[D(y,T(x0))+D(x0,T(y))]\frac{1}{2}\left[D\left(y,T\left(x_{0}\right)\right)+D\left(x_{0},T(y)\right)\right], then d(z,x0)<α2D(y,T(x0))+α2d(x0,z)+(1α)rd\left(z,x_{0}\right)<\frac{\alpha}{2}D\left(y,T\left(x_{0}\right)\right)+\frac{\alpha}{2}d\left(x_{0},z\right)+(1-\alpha)r. This implies that (2α)d(z,x0)<αd(y,x0)+αδ(x0,Tx0)+2(1α)r(2-\alpha)d\left(z,x_{0}\right)<\alpha d\left(y,x_{0}\right)+\alpha\delta\left(x_{0},Tx_{0}\right)+2(1-\alpha)r and thus d(z,x0)2α22αrd\left(z,x_{0}\right)\leq\frac{2-\alpha^{2}}{2-\alpha}r.
From all the cases, it follows that d(z,x0)max{11αr,2α22αr}=11αrd\left(z,x_{0}\right)\leq\max\left\{\frac{1}{1-\alpha}r,\frac{2-\alpha^{2}}{2-\alpha}r\right\}=\frac{1}{1-\alpha}r. This means that T(B~(x0,r))B~(x0,11αr)T\left(\tilde{B}\left(x_{0},r\right)\right)\subset\tilde{B}\left(x_{0},\frac{1}{1-\alpha}r\right). We have that

D(x0,T(x0))δ(x0,T(x0))<(1α)r.D\left(x_{0},T\left(x_{0}\right)\right)\leq\delta\left(x_{0},T\left(x_{0}\right)\right)<(1-\alpha)r.

Taking X:=B~(x0,11αr)X:=\tilde{B}\left(x_{0},\frac{1}{1-\alpha}r\right) and T:B~(x0,r)Pcl(X)T:\tilde{B}\left(x_{0},r\right)\rightarrow P_{cl}(X), we apply the conclusion (i) for local version of the fixed point problem for the Ćirić operator on the closed ball. We mention that we have used the fact that B~(x0,11αr)\tilde{B}\left(x_{0},\frac{1}{1-\alpha}r\right) is closed in the complete metric space (X,d)(X,d). Then, there exists xFTB~(x0,r)x^{*}\in F_{T}\cap\tilde{B}\left(x_{0},r\right). Using the fact that d(x0,x)rd\left(x_{0},x^{*}\right)\leq r, we can show that d(x0,x)<rd\left(x_{0},x^{*}\right)<r.
Suppose to the contrary that r=d(x0,x)r=d\left(x_{0},x^{*}\right). Then, we have the following inequalities: r=d(x,x0)H(T(x),T(x0))+δ(x0,T(x0))<αM(x,x0)+(1α)rr=d\left(x^{*},x_{0}\right)\leq H\left(T\left(x^{*}\right),T\left(x_{0}\right)\right)+\delta\left(x_{0},T\left(x_{0}\right)\right)<\alpha M\left(x^{*},x_{0}\right)+(1-\alpha)r, where

M(x,x0)=max\displaystyle M\left(x^{*},x_{0}\right)=\max {d(x,x0),D(x,T(x)),D(x0,T(x0)),\displaystyle\left\{d\left(x^{*},x_{0}\right),D\left(x^{*},T\left(x^{*}\right)\right),D\left(x_{0},T\left(x_{0}\right)\right),\right.
12[D(x,T(x0))+D(x0,T(x))]}.\displaystyle\left.\frac{1}{2}\left[D\left(x^{*},T\left(x_{0}\right)\right)+D\left(x_{0},T\left(x^{*}\right)\right)\right]\right\}.

Notice that D(x,T(x))=0,D(x0,T(x0))δ(x0,T(x0))(1α)r<rD\left(x^{*},T\left(x^{*}\right)\right)=0,D\left(x_{0},T\left(x_{0}\right)\right)\leq\delta\left(x_{0},T\left(x_{0}\right)\right)\leq(1-\alpha)r<r, D(x0,T(x))d(x0,x)D\left(x_{0},T\left(x^{*}\right)\right)\leq d\left(x_{0},x^{*}\right) and D(x,T(x0))d(x,x0)+D(x0,T(x0))<d(x,x0)+(1α)rD\left(x^{*},T\left(x_{0}\right)\right)\leq d\left(x^{*},x_{0}\right)+D\left(x_{0},T\left(x_{0}\right)\right)<d\left(x^{*},x_{0}\right)+(1-\alpha)r.
Then, we get the following cases :
If the maximum from the right hand side is d(x,x0)d\left(x^{*},x_{0}\right), then r<αd(x,x0)+(1α)r=αr+(1α)r=rr<\alpha d\left(x^{*},x_{0}\right)+(1-\alpha)r=\alpha r+(1-\alpha)r=r, which is false.
If the maximum is D(x,T(x))D\left(x^{*},T\left(x^{*}\right)\right), then r<(1α)r<rr<(1-\alpha)r<r, which is also false.
If the maximum is D(x0,T(x0))<(1α)rD\left(x_{0},T\left(x_{0}\right)\right)<(1-\alpha)r, then we get r<(1α)r+α(1α)r=(1α2)r<rr<(1-\alpha)r+\alpha(1-\alpha)r=\left(1-\alpha^{2}\right)r<r, also false.
For the last case, if the maximum from the right hand side is 12[D(x,T(x0))\frac{1}{2}\left[D\left(x^{*},T\left(x_{0}\right)\right)\right.
+D(x0,T(x))]\left.+D\left(x_{0},T\left(x^{*}\right)\right)\right], then M(x,x0)12H(T(x),T(x0))+(1α)rM\left(x^{*},x_{0}\right)\leq\frac{1}{2}H\left(T\left(x^{*}\right),T\left(x_{0}\right)\right)+(1-\alpha)r. Furthermore, by the condition that the operator is of Ćirić-type, we have that H(T(x),T(x0))αM(x,x0)H\left(T\left(x^{*}\right),T\left(x_{0}\right)\right)\leq\alpha M\left(x^{*},x_{0}\right). So H(T(x),T(x0))α(1α)r2αH\left(T\left(x^{*}\right),T\left(x_{0}\right)\right)\leq\frac{\alpha(1-\alpha)r}{2-\alpha}.
It follows that r=d(x,x0)<H(T(x),T(x0))+(1α)rr=d\left(x^{*},x_{0}\right)<H\left(T\left(x^{*}\right),T\left(x_{0}\right)\right)+(1-\alpha)r, so r<(1α2)r<rr<\left(1-\alpha^{2}\right)r<r, which is false.
From all the cases from above, it follows that d(x,x0)<rd\left(x^{*},x_{0}\right)<r.
(j) We prove that, if VV is an open subset of UU, then G(V)G(V) is open in XX. This means that for x0Ux_{0}\in U and r0>r>0r_{0}>r>0, with B(x0,r)UB\left(x_{0},r\right)\subset U, then V0(G(x0),(1α)r)G(B(x0,r))V^{0}\left(G\left(x_{0}\right),(1-\alpha)r\right)\subset G\left(B\left(x_{0},r\right)\right).
So, let yV0(G(x0),(1α)r)y\in V^{0}\left(G\left(x_{0}\right),(1-\alpha)r\right), i.e. D(y,G(x0))<(1α)rD\left(y,G\left(x_{0}\right)\right)<(1-\alpha)r. We shall show that yG(B(x0,r))y\in G\left(B\left(x_{0},r\right)\right). In other words, we shall show that there exists xB(x0,r)x^{*}\in B\left(x_{0},r\right), such that yG(x)y\in G\left(x^{*}\right), i.e., yxT(x)y\in x^{*}-T\left(x^{*}\right).
Let us consider the multi-valued operator F:B(x0,r)Pcl(X)F:B\left(x_{0},r\right)\rightarrow P_{cl}(X), defined by F(x):=y+T(x)F(x):=y+T(x).
If FF has a fixed point xx^{*}, then xy+T(x)x^{*}\in y+T\left(x^{*}\right) or yxT(x)y\in x^{*}-T\left(x^{*}\right). Now, for each x,zB(x0,r)x,z\in B\left(x_{0},r\right), we have that :
H(F(x),F(z))=H(y+T(x),y+T(z))H(T(x),T(z))αM(x,z)H(F(x),F(z))=H(y+T(x),y+T(z))\leq H(T(x),T(z))\leq\alpha M(x,z). Moreover, D(x0,F(x0))=D(x0,y+T(x0))=D(y,x0T(x0))=D(y,G(x0))<(1α)rD\left(x_{0},F\left(x_{0}\right)\right)=D\left(x_{0},y+T\left(x_{0}\right)\right)=D\left(y,x_{0}-T\left(x_{0}\right)\right)=D\left(y,G\left(x_{0}\right)\right)<(1-\alpha)r. Then FF is a Cirić operator defined on the open ball B(x0,r)B\left(x_{0},r\right), where D(x0,F(x0))<(1α)rD\left(x_{0},F\left(x_{0}\right)\right)<(1-\alpha)r. Applying the conclusion (h), i.e. the local version involving an open ball, it follows easily that GG is open.
(k) For the proof of the Caristi selection of the multi-valued Ćirić operator TT, we refer to the work of A. Petruşel and G. Petruşel [20].
(m) Let ε>0\varepsilon>0 and consider yXy^{*}\in X that satisfies D(y,T(y))εD\left(y^{*},T\left(y^{*}\right)\right)\leq\varepsilon. Then, for each (x,y)Graph(T)(x,y)\in\operatorname{Graph}(T), we have that d(x,t(x,y))11αd(x,y)d\left(x,t^{\infty}(x,y)\right)\leq\frac{1}{1-\alpha}d(x,y).
Now, since there exists (y,u)=D(y,T(y))\left(y^{*},u^{*}\right)=D\left(y^{*},T\left(y^{*}\right)\right), we take x:=t(y,u)x^{*}:=t^{\infty}\left(y^{*},u^{*}\right).
This implies that d(y,x)=d(y,t(y,u))11αd(y,u)=ψ(ε)d\left(y^{*},x^{*}\right)=d\left(y^{*},t^{\infty}\left(y^{*},u^{*}\right)\right)\leq\frac{1}{1-\alpha}d\left(y^{*},u^{*}\right)=\psi(\varepsilon), where ψ(t)=t1α\psi(t)=\frac{t}{1-\alpha}.
(n) For the proof of this, we refer to [2].
(o) Let ε>0\varepsilon>0. Let’s denote Eε(T):={xXsupzTxd(x,z)ε}E_{\varepsilon}(T):=\left\{x\in X\mid\sup_{z\in Tx}d(x,z)\leq\varepsilon\right\}. Since TT is lower semicontinuous, by Lemma 3.3 from [8], we get that for each ε>0\varepsilon>0, the set Eε(T)E_{\varepsilon}(T) is nonempty. Now, let x,yEε(T)x,y\in E_{\varepsilon}(T). It follows that:

d(x,y)H({x},T(x))+H(T(x),T(y))+H({y},T(y))d(x,y)\leq H(\{x\},T(x))+H(T(x),T(y))+H(\{y\},T(y))

Now, since x,yEε(T)x,y\in E_{\varepsilon}(T), then H({x},T(x))εH(\{x\},T(x))\leq\varepsilon and H({y},T(y))εH(\{y\},T(y))\leq\varepsilon. So,
we get that

d(x,y)2ε+αM(x,y)d(x,y)\leq 2\varepsilon+\alpha M(x,y)

Then, we have the following cases:
If M(x,y)=d(x,y)M(x,y)=d(x,y), then d(x,y)2ε+αd(x,y)d(x,y)\leq 2\varepsilon+\alpha d(x,y), so d(x,y)2ε1αd(x,y)\leq\frac{2\varepsilon}{1-\alpha}.
If M(x,y)=D(x,T(x))=D({x},T(x))H({x},T(x))εM(x,y)=D(x,T(x))=D(\{x\},T(x))\leq H(\{x\},T(x))\leq\varepsilon, then d(x,y)2ε+αε=ε(2+α)d(x,y)\leq 2\varepsilon+\alpha\varepsilon=\varepsilon(2+\alpha).
Similarly, if M(x,y)=D(y,T(y))=D({y},T(y))H({y},T(y))εM(x,y)=D(y,T(y))=D(\{y\},T(y))\leq H(\{y\},T(y))\leq\varepsilon, then d(x,y)2ε+αε=ε(2+α)d(x,y)\leq 2\varepsilon+\alpha\varepsilon=\varepsilon(2+\alpha).
Finally, if M(x,y)=12[D(x,T(y))+D(y,T(x))]M(x,y)=\frac{1}{2}[D(x,T(y))+D(y,T(x))], then we infer that:

D(x,T(y))\displaystyle D(x,T(y)) d(x,y)+D(y,T(y))\displaystyle\leq d(x,y)+D(y,T(y))
=d(x,y)+infzT(y)d(y,z)d(x,y)+supzT(y)d(y,z)\displaystyle=d(x,y)+\inf_{z\in T(y)}d(y,z)\leq d(x,y)+\sup_{z\in T(y)}d(y,z)
d(x,y)+ε\displaystyle\leq d(x,y)+\varepsilon

since yEε(T)y\in E_{\varepsilon}(T).
Furthermore,

D(y,T(x))\displaystyle D(y,T(x)) d(x,y)+D(x,T(x))\displaystyle\leq d(x,y)+D(x,T(x))
=d(x,y)+infzT(x)d(x,z)d(x,y)+supzT(x)d(x,z)\displaystyle=d(x,y)+\inf_{z\in T(x)}d(x,z)\leq d(x,y)+\sup_{z\in T(x)}d(x,z)
d(x,y)+ε\displaystyle\leq d(x,y)+\varepsilon

since xEε(T)x\in E_{\varepsilon}(T). Thus, d(x,y)2ε+αε+αd(x,y)d(x,y)\leq 2\varepsilon+\alpha\varepsilon+\alpha d(x,y).
From all the cases, it follows that

d(x,y)εmax{2+α1α,21α,2+α}=2+α1αεd(x,y)\leq\varepsilon\max\left\{\frac{2+\alpha}{1-\alpha},\frac{2}{1-\alpha},2+\alpha\right\}=\frac{2+\alpha}{1-\alpha}\varepsilon

Now, if the multi-valued Ćirić-type operator TT has a strict fixed point, then TT has the approximate endpoint property. Let us suppose now that the multi-valued operator TT has the approximate endpoint property. We define Cn:=E1n(T)={xX|supyT(x)d(x,y)1n}C_{n}:=E_{\frac{1}{n}}(T)=\left\{x\in X\left\lvert\,\sup_{y\in T(x)}d(x,y)\leq\frac{1}{n}\right.\right\}. Then, by our hypothesis, for each n,Cnn\in\mathbb{N},C_{n} is nonempty. Furthermore, for all n,Cn+1Cnn\in\mathbb{N},C_{n+1}\subseteq C_{n}.
Also, since TT is lower semicontinuous, then CnC_{n} are closed, for each nn\in\mathbb{N}. Also, we observe that :
δ(Cn)=δ(E1n(T))2+α1α1n\delta\left(C_{n}\right)=\delta\left(E_{\frac{1}{n}}(T)\right)\leq\frac{2+\alpha}{1-\alpha}\cdot\frac{1}{n}, so limnδ(Cn)=0\lim_{n\rightarrow\infty}\delta\left(C_{n}\right)=0.
Then, by Cantor’s intersection theorem, it follows that nCn={x0}\bigcap_{n\in\mathbb{N}}C_{n}=\left\{x_{0}\right\}, so the conclusion follows easily.
(p) By (d), we have that FTF_{T} is closed in ( X,dX,d ). Since ( X,dX,d ) is complete, then FTF_{T} is complete with respect to dd. Furthermore, let’s suppose that FTF_{T} is not compact. Then FTF_{T} is not precompact. This means that there exist δ>0\delta>0 and (xk)kFT\left(x_{k}\right)_{k\in\mathbb{N}}\subset F_{T}, such that d(xi,xj)δd\left(x_{i},x_{j}\right)\geq\delta, for all iji\neq j.
Denote ρ:=inf{RaX\rho:=\inf\left\{R\mid\exists a\in X\right., such that B(a,R)B(a,R) contains an infinity of xks}\left.x_{k}^{\prime}s\right\}. It is obvious that ρδ2\rho\geq\frac{\delta}{2}, because for each aX,B(a,δ2)a\in X,B\left(a,\frac{\delta}{2}\right) contains at most one xkx_{k}.
Furthermore, consider 0<ε<(12α)ρ0<\varepsilon<(1-2\alpha)\rho and take aXa\in X, such that the set J:={kxkB(a,ρ+ε)}J:=\left\{k\mid x_{k}\in B(a,\rho+\varepsilon)\right\} is infinite. Then, for each kJk\in J, we have

D(xk,T(a))H(T(xk),T(a))αM(xk,a)D\left(x_{k},T(a)\right)\leq H\left(T\left(x_{k}\right),T(a)\right)\leq\alpha M\left(x_{k},a\right)

Now, we have the following cases:
If M(xk,a)=d(xk,a)M\left(x_{k},a\right)=d\left(x_{k},a\right), then D(xk,T(a))αd(xk,a)α(ρ+ε)D\left(x_{k},T(a)\right)\leq\alpha d\left(x_{k},a\right)\leq\alpha(\rho+\varepsilon).
Also, if M(xk,a)=D(a,T(a))M\left(x_{k},a\right)=D(a,T(a)), then D(xk,T(a))αd(a,y)D\left(x_{k},T(a)\right)\leq\alpha d(a,y), for yTay\in Ta.
Now, if M(xk,a)=12D(xk,T(a))+12D(a,T(xk))M\left(x_{k},a\right)=\frac{1}{2}D\left(x_{k},T(a)\right)+\frac{1}{2}D\left(a,T\left(x_{k}\right)\right), then

D(xk,T(a))α2D(xk,T(a))+α2d(a,xk)D\left(x_{k},T(a)\right)\leq\frac{\alpha}{2}D\left(x_{k},T(a)\right)+\frac{\alpha}{2}d\left(a,x_{k}\right)

so D(xk,T(a))α2αd(a,xk)D\left(x_{k},T(a)\right)\leq\frac{\alpha}{2-\alpha}d\left(a,x_{k}\right). It implies that D(xk,T(a))α(ρ+ε)D\left(x_{k},T(a)\right)\leq\alpha(\rho+\varepsilon).
So, all the cases from above imply that D(xk,a)max{α(ρ+ε),αd(a,y)}D\left(x_{k},a\right)\leq\max\{\alpha(\rho+\varepsilon),\alpha d(a,y)\}, where yT(a)y\in T(a). From all of this, we have two cases to consider :
In the first case, by D(xk,T(a))αd(a,y)D\left(x_{k},T(a)\right)\leq\alpha d(a,y), with yT(a)y\in T(a), we obtain that D(xk,T(a))αd(a,xk)+αd(xk,y)D\left(x_{k},T(a)\right)\leq\alpha d\left(a,x_{k}\right)+\alpha d\left(x_{k},y\right). Taking infyT(a)\inf_{y\in T(a)}, we get, for each kJk\in J, that D(xk,T(a))α1α(ρ+ε)D\left(x_{k},T(a)\right)\leq\frac{\alpha}{1-\alpha}\cdot(\rho+\varepsilon).
Now, the second case is for D(xk,T(a))α(ρ+ε)D\left(x_{k},T(a)\right)\leq\alpha(\rho+\varepsilon). From these two cases, one can get D(xk,T(a))max{α,α1α}(ρ+ε)=α1α(ρ+ε)D\left(x_{k},T(a)\right)\leq\max\left\{\alpha,\frac{\alpha}{1-\alpha}\right\}\cdot(\rho+\varepsilon)=\frac{\alpha}{1-\alpha}\cdot(\rho+\varepsilon). Then D(xk,T(a))α1α(ρ+ε)D\left(x_{k},T(a)\right)\leq\frac{\alpha}{1-\alpha}\cdot(\rho+\varepsilon), so since T(a)T(a) is compact, there exists ykT(a)y_{k}\in T(a), such that d(xk,yk)α1α(ρ+ε)d\left(x_{k},y_{k}\right)\leq\frac{\alpha}{1-\alpha}(\rho+\varepsilon), for each kJk\in J.
Moreover, since T(a)T(a) is compact, then there exists bT(a)b\in T(a), for which the set J:={kJd(yk,b)<ε}J^{\prime}:=\left\{k\in J\mid d\left(y_{k},b\right)<\varepsilon\right\} is infinite. This means that for each kJk\in J^{\prime} (since α<12\alpha<\frac{1}{2} and ε\varepsilon was chosen such that ε<ρ(12α)\varepsilon<\rho\cdot(1-2\alpha) ), we have that

d(xk,b)d(xk,yk)+d(yk,b)<α1α(ρ+ε)+ε<ρd\left(x_{k},b\right)\leq d\left(x_{k},y_{k}\right)+d\left(y_{k},b\right)<\frac{\alpha}{1-\alpha}(\rho+\varepsilon)+\varepsilon<\rho

Vol. LVII (2019) Some variants of Ćirić’s multi-valued contraction principle 37

This contradicts the fact that the ball B(b,R)B(b,R) contains an infinite number of elements xksx_{k}^{\prime}s, where R=α1αρ+ε(1+α1α)R=\frac{\alpha}{1-\alpha}\rho+\varepsilon\left(1+\frac{\alpha}{1-\alpha}\right).
(q) Let Fp:={xXD(x,T(x))<p}F_{p}^{*}:=\{x\in X\mid D(x,T(x))<p\}, for each p>0p>0. Notice that if xFTx\in F_{T}, then D(x,T(x))=0<pD(x,T(x))=0<p, for each p>0p>0. So FTFpF_{T}\subseteq F_{p}^{*}. This implies that H(Fp,FT)=ρ(Fp,FT):=supxFpD(x,FT)H\left(F_{p}^{*},F_{T}\right)=\rho\left(F_{p}^{*},F_{T}\right):=\sup_{x\in F_{p}^{*}}D\left(x,F_{T}\right), for all p>0p>0, where ρ\rho denotes the excess functional.
Moreover, let xFpx\in F_{p}^{*} and ε>0\varepsilon>0. Because xFpx\in F_{p}^{*}, then D(x,T(x))<pD(x,T(x))<p. So, for xFpx\in F_{p}^{*} there exists x1T(x)x_{1}\in T(x), for which d(x,x1)<(1+ε)pd\left(x,x_{1}\right)<(1+\varepsilon)p.
For x0=xx_{0}=x and x1T(x)=T(x0)x_{1}\in T(x)=T\left(x_{0}\right), following (b) there exists a sequence of successive approximations (xn)n\left(x_{n}\right)_{n\in\mathbb{N}}, starting from (x0,x1)Graph(T)\left(x_{0},x_{1}\right)\in\operatorname{Graph}(T), such that d(xn,x)Ln(q)1L(q)d(x0,x1)d\left(x_{n},x^{*}\right)\leq\frac{L^{n}(q)}{1-L(q)}d\left(x_{0},x_{1}\right), for each nn\in\mathbb{N}, where L(q):=qαL(q):=q\alpha, with q(1,1α)q\in\left(1,\frac{1}{\alpha}\right) and with the property that xnxFTx_{n}\rightarrow x^{*}\in F_{T} as nn\rightarrow\infty.
Taking n=0n=0, we obtain d(x0,x)11L(q)d(x0,x1)(1+ε)p1L(q)d\left(x_{0},x^{*}\right)\leq\frac{1}{1-L(q)}d\left(x_{0},x_{1}\right)\leq\frac{(1+\varepsilon)p}{1-L(q)}. So d(x0,x)(1+ε)p1qαd\left(x_{0},x^{*}\right)\leq\frac{(1+\varepsilon)p}{1-q\alpha}. Taking q1q\searrow 1, respectively ε0\varepsilon\searrow 0, it follows that d(x0,x)p1αd\left(x_{0},x^{*}\right)\leq\frac{p}{1-\alpha}. So, the conclusion follows easily from this inequality.

We will present now the second result of this article, which is an extended version of strict fixed point principle for multi-valued Ćirić operators. Since all the conclusion from Theorem 2.1 are valid even in the particular case when (SF)T(SF)_{T}\neq\emptyset, for this case we shall present only the metrical conclusions that are new.

Theorem 2.2 (An extended strict fixed point principle for multi-valued Ćirić operators). Let ( X,dX,d ) be a complete metric space and T:XPcl(X)T:X\rightarrow P_{cl}(X) be a multi-valued α\alpha-Cirić type operator. Suppose that (SF)T(SF)_{T}\neq\emptyset. Then, the following conclusions hold:
(a) (SF)T=FT={x}(SF)_{T}=F_{T}=\left\{x^{*}\right\};
(b) if α<12\alpha<\frac{1}{2}, then TT has the Ostrowski property;
(c) the fixed point inclusion xT(x)x\in T(x) is generalized Ulam-Hyers stable;
(d) the strict fixed point inclusion {x}=T(x)\{x\}=T(x) is generalized Ulam-Hyers stable;
(e) the fixed point problem is well-posed for TT, with respect to DD and, respectively, with respect to HH;
(f) if α<12\alpha<\frac{1}{2}, then H(T(x),x)α1 alpha d(x,x)H\left(T(x),x^{*}\right)\leq\frac{\alpha}{1-\text{ alpha }}d\left(x,x^{*}\right), for each xXx\in X;
(g) d(x,x)11αH(x,T(x))d\left(x,x^{*}\right)\leq\frac{1}{1-\alpha}H(x,T(x)), for each xXx\in X;
(h) if G:XP(X)G:X\rightarrow P(X) is a multi-valued operator with FGF_{G}\neq\emptyset, and there exists η>0\eta>0, such that H(T(x),G(x))ηH(T(x),G(x))\leq\eta, for all xXx\in X, then H(FT,FG)η11αH\left(F_{T},F_{G}\right)\leq\eta\cdot\frac{1}{1-\alpha}.

Proof. (a) Since (SF)T(SF)_{T}\neq\emptyset, then there exists x(SF)TFTx^{*}\in(SF)_{T}\subset F_{T}. Suppose there exists yFTy^{*}\in F_{T}. We show that x=yx^{*}=y^{*}. For this, suppose the contrary that xyx^{*}\neq y^{*}. Then:

d(x,y)=D(T(x),y)H(T(x),T(y))αM(x,y)d\left(x^{*},y^{*}\right)=D\left(T\left(x^{*}\right),y^{*}\right)\leq H\left(T\left(x^{*}\right),T\left(y^{*}\right)\right)\leq\alpha M\left(x^{*},y^{*}\right)

Since D(x,T(x))=D(y,T(y))=0,D(x,T(y))d(x,y)D\left(x^{*},T\left(x^{*}\right)\right)=D\left(y^{*},T\left(y^{*}\right)\right)=0,D\left(x^{*},T\left(y^{*}\right)\right)\leq d\left(x^{*},y^{*}\right) and D(yD\left(y^{*}\right., T(x))=d(x,y)\left.T\left(x^{*}\right)\right)=d\left(x^{*},y^{*}\right), it follows that M(x,y)d(x,y)M\left(x^{*},y^{*}\right)\leq d\left(x^{*},y^{*}\right).
So d(x,y)αd(x,y)<d(x,y)d\left(x^{*},y^{*}\right)\leq\alpha d\left(x^{*},y^{*}\right)<d\left(x^{*},y^{*}\right). This implies that d(x,y)=0d\left(x^{*},y^{*}\right)=0, so we obtain a contradiction.
Finally, x=yx^{*}=y^{*}, so FT={x}=(SF)TF_{T}=\left\{x^{*}\right\}=(SF)_{T}.
(b) Let (yn)n\left(y_{n}\right)_{n\in\mathbb{N}} be a sequence, such that D(yn+1,T(yn))0D\left(y_{n+1},T\left(y_{n}\right)\right)\rightarrow 0. We shall show that d(yn,x)0d\left(y_{n},x^{*}\right)\rightarrow 0. Then, we have d(x,yn+1)H(T(x),T(yn))+D(yn+1,T(yn))αM(x,yn)+D(yn+1,T(yn))d\left(x^{*},y_{n+1}\right)\leq H\left(T\left(x^{*}\right),T\left(y_{n}\right)\right)+D\left(y_{n+1},T\left(y_{n}\right)\right)\leq\alpha M\left(x^{*},y_{n}\right)+D\left(y_{n+1},T\left(y_{n}\right)\right), where

M(x,yn)=\displaystyle M\left(x^{*},y_{n}\right)=
=max{d(x,yn),D(x,T(x)),D(yn,T(yn)),12[D(x,T(yn))+D(yn,T(x))]}\displaystyle\quad=\max\left\{d\left(x^{*},y_{n}\right),D\left(x^{*},T\left(x^{*}\right)\right),D\left(y_{n},T\left(y_{n}\right)\right),\frac{1}{2}\left[D\left(x^{*},T\left(y_{n}\right)\right)+D\left(y_{n},T\left(x^{*}\right)\right)\right]\right\}
max{d(x,yn),D(yn,T(yn)),12[d(yn,x)+D(x,T(yn))]}\displaystyle\quad\leq\max\left\{d\left(x^{*},y_{n}\right),D\left(y_{n},T\left(y_{n}\right)\right),\frac{1}{2}\left[d\left(y_{n},x^{*}\right)+D\left(x^{*},T\left(y_{n}\right)\right)\right]\right\}

Now, we have the following cases :
If the maximum from the right hand side is d(x,yn)d\left(x^{*},y_{n}\right), then d(x,yn+1)D(yn+1,T(yn))+αd(x,yn)d\left(x^{*},y_{n+1}\right)\leq D\left(y_{n+1},T\left(y_{n}\right)\right)+\alpha d\left(x^{*},y_{n}\right).
If the maximum is D(yn,T(yn))d(yn,x)+D(x,T(yn))D\left(y_{n},T\left(y_{n}\right)\right)\leq d\left(y_{n},x^{*}\right)+D\left(x^{*},T\left(y_{n}\right)\right), then we have H(x,T(yn))=H(T(x),T(yn))αM(x,yn)αd(yn,x)+αH(x,T(yn))H\left(x^{*},T\left(y_{n}\right)\right)=H\left(T\left(x^{*}\right),T\left(y_{n}\right)\right)\leq\alpha M\left(x^{*},y_{n}\right)\leq\alpha d\left(y_{n},x^{*}\right)+\alpha H\left(x^{*},T\left(y_{n}\right)\right). So, we get that H(T(x),T(yn))α1αd(yn,x)H\left(T\left(x^{*}\right),T\left(y_{n}\right)\right)\leq\frac{\alpha}{1-\alpha}d\left(y_{n},x^{*}\right).
It implies that d(yn+1,x)D(yn+1,T(yn))+α1αd(yn,x)d\left(y_{n+1},x^{*}\right)\leq D\left(y_{n+1},T\left(y_{n}\right)\right)+\frac{\alpha}{1-\alpha}d\left(y_{n},x^{*}\right).
Consider now the case when the maximum is 12[d(yn,x)+D(x,T(yn))]\frac{1}{2}\left[d\left(y_{n},x^{*}\right)+D\left(x^{*},T\left(y_{n}\right)\right)\right]. Then, we obtain D(x,T(yn))H(T(x),T(yn))αM(x,yn)D\left(x^{*},T\left(y_{n}\right)\right)\leq H\left(T\left(x^{*}\right),T\left(y_{n}\right)\right)\leq\alpha M\left(x^{*},y_{n}\right). Thus H(T(x),T(yn))α2(d(yn,x)+H(T(x),T(yn)))H\left(T\left(x^{*}\right),T\left(y_{n}\right)\right)\leq\frac{\alpha}{2}\left(d\left(y_{n},x^{*}\right)+H\left(T\left(x^{*}\right),T\left(y_{n}\right)\right)\right). This means that

H(T(x),T(yn))α2αd(yn,x)H\left(T\left(x^{*}\right),T\left(y_{n}\right)\right)\leq\frac{\alpha}{2-\alpha}d\left(y_{n},x^{*}\right)

Hence d(yn+1,x)D(yn+1,T(yn))+α2αd(yn,x)d\left(y_{n+1},x^{*}\right)\leq D\left(y_{n+1},T\left(y_{n}\right)\right)+\frac{\alpha}{2-\alpha}d\left(y_{n},x^{*}\right).
Now, since β:=max{α,α1α,α2α}=α1α\beta:=\max\left\{\alpha,\frac{\alpha}{1-\alpha},\frac{\alpha}{2-\alpha}\right\}=\frac{\alpha}{1-\alpha}, then from all the cases from above, it follows that d(yn+1,x)D(yn+1,T(yn))+βd(yn,x)D(yn+1,T(yn))+βD(yn,T(yn1))+β2d(yn1,x)βn+1d(y0,x)+k=0nβnkD(yk+1,T(yk))d\left(y_{n+1},x^{*}\right)\leq D\left(y_{n+1},T\left(y_{n}\right)\right)+\beta d\left(y_{n},x^{*}\right)\leq D\left(y_{n+1},T\left(y_{n}\right)\right)+\beta D\left(y_{n},T\left(y_{n-1}\right)\right)+\beta^{2}d\left(y_{n-1},x^{*}\right)\leq\ldots\leq\beta^{n+1}d\left(y_{0},x^{*}\right)+\sum_{k=0}^{n}\beta^{n-k}D\left(y_{k+1},T\left(y_{k}\right)\right). Now, since β<1\beta<1, using Cauchy’s lemma, we get that d(yn+1,x)0d\left(y_{n+1},x^{*}\right)\rightarrow 0.
(c) By (a) we know that (SF)T=FT={x}(SF)_{T}=F_{T}=\left\{x^{*}\right\}.

Now, let us consider xXx\in X and yT(x)y\in T(x). Then, we have the following:
d(x,x)d(x,y)+H(T(x),T(x))d(x,y)+αM(x,x)d(x,y)+αmax{d(x,x),D(x,T(x)),12d(x,y)+12d(x,x)}d\left(x,x^{*}\right)\leq d(x,y)+H\left(T(x),T\left(x^{*}\right)\right)\leq d(x,y)+\alpha M\left(x,x^{*}\right)\leq d(x,y)+\alpha\max\left\{d\left(x,x^{*}\right),D(x,T(x)),\frac{1}{2}d\left(x^{*},y\right)+\frac{1}{2}d\left(x,x^{*}\right)\right\}. Moreover, we consider the following cases:
If M(x,x)=d(x,x)M\left(x,x^{*}\right)=d\left(x,x^{*}\right), then d(x,x)11αd(x,y)d\left(x,x^{*}\right)\leq\frac{1}{1-\alpha}d(x,y).
If M(x,x)=D(x,T(x))M\left(x,x^{*}\right)=D(x,T(x)), then

d(x,x)d(x,y)+αD(x,T(x))(1+α)d(x,y)d\left(x,x^{*}\right)\leq d(x,y)+\alpha D(x,T(x))\leq(1+\alpha)d(x,y)

Finally, if M(x,x)12d(x,y)+12d(x,x)M\left(x,x^{*}\right)\leq\frac{1}{2}d\left(x^{*},y\right)+\frac{1}{2}d\left(x,x^{*}\right), then we have d(x,x)d(x,y)+α2d(x,y)+α2d(x,x)d\left(x,x^{*}\right)\leq d(x,y)+\frac{\alpha}{2}d\left(x^{*},y\right)+\frac{\alpha}{2}d\left(x,x^{*}\right). So, we get d(x,x)2+α2(1α)d(x,y)d\left(x,x^{*}\right)\leq\frac{2+\alpha}{2(1-\alpha)}d(x,y). From all the cases we obtain that

d(x,x)max{11α,1+α,2+α2(1α)}d(x,y)=2+α2(1α)d(x,y)d\left(x,x^{*}\right)\leq\max\left\{\frac{1}{1-\alpha},1+\alpha,\frac{2+\alpha}{2(1-\alpha)}\right\}d(x,y)=\frac{2+\alpha}{2(1-\alpha)}d(x,y)

Now, let us define ψ(t):=2+α2(1α)t\psi(t):=\frac{2+\alpha}{2(1-\alpha)}t, so d(x,x)ψ(d(x,y))d\left(x,x^{*}\right)\leq\psi(d(x,y)). We notice that ψ\psi is continuous in 0 , increasing and with ψ(0)=0\psi(0)=0.
Then, as in (m) of Theorem 2.1, we have the following:
Let ε>0\varepsilon>0 and consider yXy^{*}\in X that satisfies D(y,T(y))εD\left(y^{*},T\left(y^{*}\right)\right)\leq\varepsilon. Then, for each (x,y)Graph(T)(x,y)\in Graph(T), we have d(x,t(x,y))ψ(d(x,y))d\left(x,t^{\infty}(x,y)\right)\leq\psi(d(x,y)).
Now, since there exists (y,u)=D(y,T(y))\left(y^{*},u^{*}\right)=D\left(y^{*},T\left(y^{*}\right)\right), we take x:=t(y,u)x^{*}:=t^{\infty}\left(y^{*},u^{*}\right). This implies that d(y,x)=d(y,t(y,u))ψ(d(y,u))d\left(y^{*},x^{*}\right)=d\left(y^{*},t^{\infty}\left(y^{*},u^{*}\right)\right)\leq\psi\left(d\left(y^{*},u^{*}\right)\right) and the conclusion follows.
(d) Let ε>0\varepsilon>0 and yXy^{*}\in X, such that H(y,T(y))εH\left(y^{*},T\left(y^{*}\right)\right)\leq\varepsilon. Since TT is a Ćirić multi-valued operator, from (h) we have that d(x,x)11αH(x,T(x))d\left(x,x^{*}\right)\leq\frac{1}{1-\alpha}H(x,T(x)), for each xXx\in X. This implies that d(y,x)11αH(y,T(y))ψ(ε)d\left(y^{*},x^{*}\right)\leq\frac{1}{1-\alpha}H\left(y^{*},T\left(y^{*}\right)\right)\leq\psi(\varepsilon), where ψ(t):=t1α\psi(t):=\frac{t}{1-\alpha} satisfies ψ(0)=0\psi(0)=0 and it is an increasing and continuous
mapping in 0 .
(e) The proof of this conclusion is given in [19].
(f) We know that
H(T(x),T(x))αmax{d(x,x),D(x,T(x)),12[D(x,T(x))+D(x,T(x))]}H\left(T(x),T\left(x^{*}\right)\right)\leq\alpha\max\left\{d\left(x,x^{*}\right),D(x,T(x)),\frac{1}{2}\left[D\left(x,T\left(x^{*}\right)\right)+D\left(x^{*},T(x)\right)\right]\right\}.

We have the following cases:
If the maximum is d(x,x)d\left(x,x^{*}\right), then H(T(x),T(x))αd(x,x)H\left(T(x),T\left(x^{*}\right)\right)\leq\alpha d\left(x,x^{*}\right).
If the maximum is 12[D(x,T(x))+D(x,T(x))]\frac{1}{2}\left[D\left(x,T\left(x^{*}\right)\right)+D\left(x^{*},T(x)\right)\right], then H(T(x),T(x))=α2d(x,x)+α2H(T(x),T(x))H\left(T(x),T\left(x^{*}\right)\right)=\frac{\alpha}{2}d\left(x,x^{*}\right)+\frac{\alpha}{2}H\left(T(x),T\left(x^{*}\right)\right) and so H(T(x),T(x))α2αd(x,x)H\left(T(x),T\left(x^{*}\right)\right)\leq\frac{\alpha}{2-\alpha}d\left(x,x^{*}\right).
If the maximum is D(x,T(x))D(x,T(x)), then we obtain H(T(x),T(x))α1αd(x,x)H\left(T(x),T\left(x^{*}\right)\right)\leq\frac{\alpha}{1-\alpha}d\left(x,x^{*}\right).
Since max{α2α,α,α1α}=α1α,H(T(x),x)=H(T(x),T(x))α1αd(x,x)\max\left\{\frac{\alpha}{2-\alpha},\alpha,\frac{\alpha}{1-\alpha}\right\}=\frac{\alpha}{1-\alpha},H\left(T(x),x^{*}\right)=H\left(T(x),T\left(x^{*}\right)\right)\leq\frac{\alpha}{1-\alpha}d\left(x,x^{*}\right).
(g) We have the following chain of inequalities d(x,x)H(x,T(x))+H(T(x),x)H(x,T(x))+αd(x,x)d\left(x,x^{*}\right)\leq H(x,T(x))+H\left(T(x),x^{*}\right)\leq H(x,T(x))+\alpha d\left(x,x^{*}\right). Thus d(x,x)11αH(x,T(x))d\left(x,x^{*}\right)\leq\frac{1}{1-\alpha}H(x,T(x)).
(h) Let x(SF)Tx^{*}\in(SF)_{T} and yFGy^{*}\in F_{G}. Then, we have
d(x,y)H(G(y),x)H(G(y),T(y))+H(T(y),x)η+αM(y,x)d\left(x^{*},y^{*}\right)\leq H\left(G\left(y^{*}\right),x^{*}\right)\leq H\left(G\left(y^{*}\right),T\left(y^{*}\right)\right)+H\left(T\left(y^{*}\right),x^{*}\right)\leq\eta+\alpha M\left(y^{*},x^{*}\right).

Now, we have the following cases for M(y,x)M\left(y^{*},x^{*}\right) :

  1. 1.

    if M(y,x)=d(y,x)M\left(y^{*},x^{*}\right)=d\left(y^{*},x^{*}\right), then d(y,x)η1αd\left(y^{*},x^{*}\right)\leq\frac{\eta}{1-\alpha}.
    2)2) if M(y,x)=D(x,T(x))M\left(y^{*},x^{*}\right)=D\left(x^{*},T\left(x^{*}\right)\right), then d(y,x)=0d\left(y^{*},x^{*}\right)=0.

  2. 2.

    if M(y,x)=D(y,T(y))H(G(y),T(y))ηM\left(y^{*},x^{*}\right)=D\left(y^{*},T\left(y^{*}\right)\right)\leq H\left(G\left(y^{*}\right),T\left(y^{*}\right)\right)\leq\eta, then d(y,x)(1+α)ηd\left(y^{*},x^{*}\right)\leq(1+\alpha)\eta.

  3. 3.

    finally, if M(y,x)=12D(x,T(y))+12D(y,T(x))M\left(y^{*},x^{*}\right)=\frac{1}{2}D\left(x^{*},T\left(y^{*}\right)\right)+\frac{1}{2}D\left(y^{*},T\left(x^{*}\right)\right), then H(T(x),T(y))αM(y,x)α2H(T(y),T(x))+α2d(x,y)H\left(T\left(x^{*}\right),T\left(y^{*}\right)\right)\leq\alpha M\left(y^{*},x^{*}\right)\leq\frac{\alpha}{2}H\left(T\left(y^{*}\right),T\left(x^{*}\right)\right)+\frac{\alpha}{2}d\left(x^{*},y^{*}\right). Hence, we get that H(T(x),T(y))α2αd(x,y)H\left(T\left(x^{*}\right),T\left(y^{*}\right)\right)\leq\frac{\alpha}{2-\alpha}d\left(x^{*},y^{*}\right). Then d(x,y)η+α2αd(x,y)d\left(x^{*},y^{*}\right)\leq\eta+\frac{\alpha}{2-\alpha}d\left(x^{*},y^{*}\right), which implies that d(y,x)2α2(1α)ηd\left(y^{*},x^{*}\right)\leq\frac{2-\alpha}{2(1-\alpha)}\eta. It follows that d(y,x)ηd\left(y^{*},x^{*}\right)\leq\eta. max{(1+α),11α,2α2(1α)}=11αη\max\left\{(1+\alpha),\frac{1}{1-\alpha},\frac{2-\alpha}{2(1-\alpha)}\right\}=\frac{1}{1-\alpha}\eta. Using Lemma 1.2 the conclusion follows.

Vol. LVII (2019) Some variants of Ćirić’s multi-valued contraction principle 41

References

[1] J. Andres and L. Górniewicz, On the Banach contraction principle for multivalued mappings, Approximation, Optimization and Mathematical Economics (M. Lassonde -ed.), Physica-Verlag, Heidelberg, (2001), 1-23
[2] Boriceanu M., Fixed point theory for multivalued generalized contraction on a set with two b-metrics, Stud. Univ.Babeş-Bolyai, 54 (3), (2009)
[3] L. Ćirić, Fixed points for generalized multi-valued contractions, Mat. Vesnik., 9, (1972), 265-272
[4] N.V. Dung and A. Petruşel, On iterated functions systems consisting of Kannan maps, Reich maps, Chatterjea type maps, and related results, J. Fixed Point Theory Appl., 9, (2017), 2271-2285
[5] G. Chifu and G. Petrusel, Fixed point results for multivalued Hardy-Rogers contractions in b-metric spaces, Filomat, 31 (8), (2017), 2499-2507
[6] A.A. Harandi, Endpoints of set-valued contractions in metric spaces, Nonlinear Anal., 72, (2010), 132-134
[7] A.D. Rogers and G.E. Hardy, A generalization of fixed point theorem of Reich, Canad. Math. Bull., 16, (1973), 201-208
[8] N. Hussain, A.A. Harandi, and Y.J. Cho, Approximate endpoints for set-valued contractions in metric spaces, Fixed Point Theory Appl., 2010:614867, (2010), 1-13
[9] J.R. Jachymski, Caristi’s fixed point theorem and selections of set-valued contractions, J. Math. Anal. Appl., 227, (1998), 55-67
[10] T.A. Lazăr, A. Petruşel, and N. Shahzad, Fixed points for non-self operators and domain invariance theorems, Nonlinear Anal., 70, (2009), 117-125
[11] T. Lazăr, D. O’Regan, and A. Petruşel, Fixed points and homotopy results for Cirić-type multivalued operators on a set with two metrics, Bull. Korean Math. Soc., 45 (1), (2008), 67-73
[12] T. Lazăr, G. Moţ, G. Petruşel, and S. Szentesi, The theory of Reich’s fixed point theorem for multivalued operators, Fixed Point Theory Appl., 10, (2010)
[13] V.L. Lazăr, Fixed point theory for multivalued φ\varphi-contractions, Fixed Point Theory Appl., 50, (2011)
[14] S.B. Nadler Jr, Multi-valued contraction mappings, Pacific J. Math., 30, (1969), 475-488
[15] T.P. Petru, A. Petruşel, and J.C. Yao, Ulam-Hyers stability of operatorial equations and inclusions via nonself operators, Taiwanese J. Math., 15 (5), (2011), 21952212
[16] A. Petruşel, Some variants of the contraction principle for multi-valued operators, generalizations and applications, to appear
[17] A. Petruşel, Ćirić type fixed point theorems, Stud. Univ. Babęs-Bolyai, 59 (2), (2014), 233-245
[18] A. Petruşel and I.A. Rus, The theory of a metric fixed point theorem for multivalued operators, Fixed Point Theory and its Applications, Proc. Ninth International Conference on Fixed Point Theory and Applications, Changhua, Taiwan, (L.J. Lin, A. Petruşel, H.K. Xu - Eds.), Yokohama Publ., (2010), 161-175
[19] A. Petruşel, I.A. Rus, and J.C. Yao, Well-posedness in the generalized sense of the fixed point problems for multivalued operators, Taiwanese J. Math., 11 (3), (2007), 903-914
[20] A. Petruşel and G. Petruşel, Selection theorems for multivalued generalized contractions, Math. Morav., 9, (2005), 43-52
[21][21] A. Petruşel, I.A. Rus, and M.A. Serban, Basic problems of the metric fixed point theory and the relevance of a metric fixed point theorem for a multivalued operator, J. Nonlinear Convex Anal., 15 (3), (2014), 493-513
[22] B. Prasad and R. Sahni, Endpoints of multivalued contraction operators, ISRN Mathematical Anal., 2013, (2013), Article ID 542302, 7 pages
[23] J.S. Raymond, Multivalued contractions, Set-Valued Anal., 2, (1994), 559-571
[24] S. Reich, Fixed points of contractive functions, Boll. Un. Math. Ital., 4 (5), (1972), 26-42

Cristian Daniel Alecsa
Department of Mathematics, Babeş-Bolyai University
M. Kogălniceanu Street, nr. 1

Cluj-Napoca
Romania
E-mail: cristian.alecsa@math.ubbcluj.ro
Tiberiu Popoviciu Institute of Numerical Analysis
Romanian Academy
Fântânele Street nr. 57
Cluj-Napoca
Romania
E-mail: cristian.alecsa@ictp.acad.ro
Adrian Petruşel
Department of Mathematics, Babeş-Bolyai University
M. Kogălniceanu Street, nr. 1

Cluj-Napoca
Romania
Academy of Romanian Scientists
Independenţei Street, nr. 54
Bucharest
Romania
E-mail: petrusel@math.ubbcluj.ro

Related Posts