scieee Open visual document viewer

Fixed points, selections and common fixed points for nonexpansive-type mappings

Espínola García, Rafael; Lorenzo Ramírez, Josefa; Nicolae, Adriana

Abstract

We study the existence of fixed points in the context of uniformly convex geodesic metric spaces, hyperconvex spaces and Banach spaces for single and multivalued mappings satisfying conditions that generalize the concept of nonexpansivity. Besides, we use the fixed point theorems proved here to give common fixed point results for commuting mappings.

Full text

Fixed poin s, selec ions and common ixed poin s o nonexpansi e- ype mappings Ra a Esp´ınola1, Pepa Lo enzo1, Ad iana Nicolae2 1Depa amen o de An´alisis Ma em´a ico, Uni e sidad de Se illa, P.O. Box 1160, 41080 Se illa, Spain 2Depa men o Ma hema ics, Babe¸s-Bolyai Uni e si y, Kog˘alniceanu 1, 400084, Cluj-Napoca, Romania Manusc ip co espondence: Depa amen o de An´alisis Ma em´a ico, Uni e sidad de Se illa, P.O.Box 1160, 41080 Se illa, Spain email: [email p o ec ed] Phone: 0034 954 559 560 Fax: 0034 954 557 972 Email add esses: [email p o ec ed] (R. Esp´ınola), [email p o ec ed] (P. Lo enzo), [email p o ec ed]cluj. o (A. Nicolae) 1 Abs ac We s udy he exis ence o ixed poin s in he con ex o uni o mly con ex geodesic me ic spaces, hype con ex spaces and Banach spaces o single and mul i alued mappings sa is ying condi ions ha gene alize he concep o nonexpansi i y. Be- sides, we use he ixed poin heo ems p o ed he e o gi e common ixed poin esul s o commu ing mappings. Key-wo ds: ixed poin , selec ion o mul i unc ions, gene alized nonexpansi e map- pings, commu ing mappings, me ic space, Banach space. 1 In oduc ion In [26], T. Suzuki ex ends he concep o single alued nonexpansi e mapping in he ol- lowing way: a mapping de ined on a subse Ko a Banach space is said o sa is y con- di ion (C) i o x, y ∈Kwi h (1/2)kx− (x)k≤kx−yk, hen k (x)− (y)k≤kx−yk. T. Suzuki [26] p o es some basic p ope ies and gi es ixed poin heo ems and con- e gence esul s o mappings sa is ying condi ion (C). Following [26], A. Razani and H. Salahi a d [23] s a e pa o T. Suzuki’s [26] esul s in he con ex o a comple e CAT(0) space and gene alize condi ion (C) o he mul i alued case: a mul i alued map- ping Tde ined on subse o a CAT(0) space is said o sa is y condi ion (C) i o each x, y ∈Kand ux∈T(x) wi h (1/2)d(x, ux)≤d(x, y) he e exis s uy∈T(y) such ha d(ux, uy)≤d(x, y). This condi ion is used in [23] o p o e a ixed poin heo em o mul i alued mappings and some common ixed poin esul s. Mo i a ed by he esul s in [26], J. Ga c´ıa-False , E. Llo ens-Fus e and T. Suzuki conside in [7] wo gene aliza- ions in he single alued case o condi ion (C) gi ing examples and es ablishing ixed poin esul s. The pu pose o his pape is o s udy condi ion (C) o mul i alued mappings in he con ex o geodesic me ic spaces (wi h special a en ion o he case o R- ees) and Ba- nach spaces, and condi ion (C) o single alued mappings in he con ex o hype con ex spaces. A e some p elimina y con en s in Sec ion 2, we begin Sec ion 3 by s udying he mul i alued case in geodesic spaces. We assume condi ion (C) o mul i alued mappings as in [23] whe e di e en esul s in his di ec ion we e ob ained o CAT(0) spaces. In ou wo k, we de i e a echnical lemma (Lemma 3.2) which is a mul i alued e sion o he key ac which is behind he main esul s in [7, 26]. Ou esul s a e i s ob ained o as gene al as comple e uni o mly con ex geodesic spaces and hen pa icula ized o mo e p ecise geome ies. Since CAT(0) spaces a e a pa icula class o uni o mly con ex geodesic spaces, we ob ain mo e gene al esul s han hose om [23]. Mo eo e , hanks mainly o Lemma 3.2, we ill in a gap in he p oo o he main mul i alued esul in [23]. We con inue Sec ion 3 by in oducing a new condi ion o mul i alued mappings in he spi i o (C). We gi e examples showing ha his condi ion is ac ually weake han con- di ion (C) and p o e a selec ion heo em in R- ees o mappings sa is ying his newly in oduced condi ion om whe e a s onge ixed poin esul o mul i alued mappings ollows. This selec ion esul esembles a e y impo an one, see o ins ance [12, 25], o hype con ex spaces (no ice, see [14], ha comple e R- ees a e hype con ex) al hough he app oach he e is comple ely di e en as he p oo elies on e y pa icula p ope ies o R- ees a he han on hype con exi y. I is wo hwhile o poin ou ha R- ees ind 2 a lo o applica ions in di e en a eas as, o ins ance, he indexing o in o ma ion o phylogene ics. We close Sec ion 3 wi h an appendix whe e we s udy he exis ence o ixed poin s o single alued mappings wi h p ope y (C) in hype con ex me ic spaces. I is e y well-known (see [17, Chap e 13]) ha nonexpansi e sel -mappings de ined on nonemp y bounded and closed hype con ex spaces ha e ixed poin s. The e o e i is na u al o wonde abou his p oblem o mappings wi h condi ion (C). We i s s udy he compac case p o iding a posi i e answe . Fo he mo e gene al case we need o in oduce a new condi ion on he mapping unde conside a ion. In pa icula i is shown ha a 2-lipschi zian sel -mapping wi h condi ion (C) de ined on a nonemp y closed and bounded hype con ex space has a ixed poin . This esul is signi ican among he class o known esul s o mappings wi h condi ion (C) since i is he i s one wi hou com- pac ness condi ions o which nei he he uniqueness o asymp o ic cen e s no any hing simila o he Opial p ope y is equi ed (see Sec ions 2 and 4 o de ini ions). The e o e, his esul ollows h ough a comple ely new app oach compa ed o hose in [7, 23, 26] and implies new esul s e en, o ins ance, in injec i e Banach spaces. In Sec ion 4 we e isi he classical heo y o nonexpansi e mul i alued mappings on Banach spaces o s udy i unde condi ion (C). We show he exis ence o ixed poin s o such a mapping in a Banach space wi h he Opial p ope y. The me hod o asymp o ic cen e s allows us o es ablish he same esul in a uni o mly con ex in e e y di ec ion (UCED) Banach space. Mo eo e , i we also assume he con inui y o he mapping we can p o e he exis ence o ixed poin s in a Banach space o which he asymp o ic cen e o a bounded sequence wi h espec o a bounded closed con ex subse is nonemp y and compac , ha is, a coun e pa o he Ki k-Massa heo em. Finally, in Sec ion 5, we appeal o he ixed poin heo ems p o ed in his pape in o de o gi e some common ixed poin esul s o commu ing mappings. 2 P elimina ies Le (X, d) be a me ic space. A geodesic pa h om x o yis a mapping c: [0, l]⊆R→X wi h c(0) = x, c(l) = yand d(c( ), c( 0)) = | − 0| o e e y , 0∈[0, l]. The image c([0, l]) o c o ms a geodesic segmen which joins xand yand is no necessa ily unique. I no con usion a ises, we will use [x, y] o deno e a geodesic segmen joining xand y. (X, d) is a (uniquely) geodesic space i e e y wo poin s x, y ∈Xcan be joined by a (unique) geodesic pa h. A poin z∈Xbelongs o he geodesic segmen [x, y] i and only i he e exis s ∈[0,1] such ha d(z, x) = d(x, y) and d(z, y) = (1 − )d(x, y), and we will w i e z= (1 − )x+ y o simplici y. A subse Ko Xis con ex i i con ains any geodesic segmen ha joins e e y wo poin s o i . In a geodesic space (X, d), he me ic d:X×X→Ris con ex i o any x, y, z ∈X one has d(x, (1 − )y+ z)≤(1 − )d(x, y) + d(x, z) o all ∈[0,1]. A geodesic space which me ic is con ex will be e e ed as a space wi h con ex me ic. A i ial example o a uniquely geodesic space wi h con ex me ic is a s ic ly con ex Banach space. Fo mo e de ails abou geodesic me ic spaces one may check [2]. A geodesic space (X, d) is uni o mly con ex i o any > 0 and ∈(0,2] he e exis s 3 δ∈(0,1] such ha i a, x, y ∈Xwi h d(x, a)≤ ,d(y, a)≤ and d(x, y)≥ hen d(1 2x+1 2y, a)≤(1 −δ) . F om he de ini ion, i is easy o see ha uni o mly con ex me ic spaces a e uniquely geodesic. A mapping δ: (0,∞)×(0,2] →(0,1] p o iding such a δ=δ( , ) o a gi en > 0 and ∈(0,2] is called a modulus o uni o m con exi y. The mapping δis mono one ( esp. lowe semi-con inuous om he igh ) i o e e y ixed i dec eases ( esp. is lowe semi-con inuous om he igh ) wi h espec o (see also [5], [18]). CAT(0) spaces in he sense o G omo (see [2]) a e uni o mly con ex me ic spaces wi h con ex me ic. Le (X, d) be a me ic space and le (xn)n∈Nbe a bounded sequence in X. Fo x∈X, de ine (x, (xn)) = lim supn→∞ d(x, xn). The asymp o ic adius o (xn)n∈Nis gi en by ((xn)) = in { (x, (xn)) : x∈X}, and he asymp o ic cen e o (xn)n∈Nis he se A((xn)) = {x∈X: (x, (xn)) = ((xn))}. Th oughou his pape we will deno e a uni o mly con ex me ic space wi h mono one (o lowe semi-con inuous om he igh ) modulus o uni o m con exi y as a UC space. In [5], he au ho s p o e ha e e y bounded sequence in a UC space has a unique asymp o ic cen e . A bounded sequence (xn)n∈Nin a comple e UC space is egula i ((xn)) = ((xnk)) o e e y subsequence (xnk)k∈No (xn)n∈N. I is known ha in a Banach space e e y bounded sequence con ains a egula subsequence (see, o ins ance, [17], Chap e 2, Lemma 5.2). Since he p oo has a me ic na u e we can conclude ha e e y bounded sequence (xn)n∈Nin a comple e UC space has a egula subsequence (xnk)k∈Nand hus e e y subsequence o (xnk)k∈Nhas he same asymp o ic cen e as (xnk)k∈N. Le (X, d) be a me ic space. Taking z∈Xand > 0 we deno e he closed ball cen e ed a zwi h adius by e B(z, ). Gi en Ya nonemp y subse o X, we de ine he dis ance o a poin z∈X o Yby dis (z, Y ) = in y∈Yd(z, y).The me ic p ojec ion (o nea es poin mapping)PYon o Yis he mapping PY(z) = {y∈Y:d(z, y) = dis (z, Y )}, o e e y z∈X. I Yis addi ionally bounded, he diame e o Yis gi en by diamY= supx,y∈Yd(x, y). In his pape we also conside he ollowing amilies o se s: P(X) = {Y⊆X:Yis nonemp y}, Pb(X) = {Y⊆X:Yis nonemp y and bounded}, Pb,c (X) = {Y⊆X:Yis nonemp y, bounded and con ex}, Pcl,c (X) = {Y⊆X:Yis nonemp y, closed and con ex}, Pb,cl,c (X) = {Y⊆X:Yis nonemp y, bounded, closed and con ex}, 4 Pcp(X) = {Y⊆X:Yis nonemp y and compac }, Pcp,c (X) = {Y⊆X:Yis nonemp y, compac and con ex}. A me ic space (X, d) is me ically con ex i o any wo dis inc poin s x, y ∈X and any α, β > 0 such ha d(x, y) = α+β he e exis s z∈Xwi h d(x, z) = αand d(y, z) = β.Xhas he bina y in e sec ion p ope y i Ti∈Ie Bi6=∅ o e e y collec ion o balls ( e Bi)i∈Isuch ha any wo o hese balls in e sec . A me ic space (X, d) is hype con ex i Ti∈Ie B(xi, i)6=∅ o e e y collec ion o poin s (xi)i∈Iin Xand posi i e numbe s ( i)i∈Isuch ha d(xi, xj)≤ i+ j o any i, j ∈ I. Hype con exi y is equi alen o he bina y in e sec ion p ope y and he me ic con exi y. Mo e abou hype con ex spaces can be ound in [1, 12, 25] o in Chap e 13 o [17]. Gi en (X, d) a me ic space and A⊆X, he numbe x(A) = supy∈Ad(x, y) is called he adius o A ela i e o x∈X. The adius o Ais (A) = in x∈X x(A), he cen e o Ais he se C(A) = {x∈X: x(A) = (A)}and he admissible co e o Ais de ined by co (A) = T{e B:e Bis a closed ball and A⊆e B}. The se Ais said o be admissible i A= co (A). Fo Xa hype con ex space and A⊆X, co (A) = Tx∈Xe B(x, x(A)) and diam(A)=2 (A) ( o de ails see Chap e 13 o [17]). An R- ee is a uniquely geodesic me ic space Xsuch ha i [y, x]∩[x, z] = {x} hen [y, x]∪[x, z]=[y, z] o each x, y, z ∈X. F om he de ini ion i immedia ely ollows ha i x, y, z ∈X, hen [x, y]∩[x, z]=[x, w] o some w∈X. Likewise, i Kis a closed and con ex subse o an R- ee X, hen o e e y x∈X,PK(x) is a single on and o any y∈K,d(x, y) = d(x, PK(x)) + d(PK(x), y). A s anda d example o an R- ee is R2 endowed wi h he so-called i e me ic. Fo x= (x1, x2), y = (y1, y2)∈R2, he i e me ic (deno ed by ρ) is de ined by ρ(x, y) = |x2−y2|i x1=y1, |x2|+|y2|+|x1−y1|o he wise. I is known ha R- ees a e CAT(0) spaces and ha a me ic space is a comple e R- ee i and only i i is hype con ex and has unique geodesic segmen s (see [14]). Mo e abou he ixed poin heo y in R- ees can be ound in [4, 15, 21, 22]. In [26], T. Suzuki conside ed he ollowing gene alized amily o nonexpansi e map- pings in he se ing o a Banach space. We will use in he sequel he no m no a ion, bu he same de ini ions also hold when wo king in he me ic se ing (na u ally, he no m will be eplaced by he dis ance). De ini ion 2.1. Le Xbe a Banach space, K∈P(X)and :K→X. Then sa is ies condi ion (C)i 1 2kx− (x)k ≤ kx−yk=⇒ k (x)− (y)k ≤ kx−yk, o all x, y ∈K. Ob iously, e e y nonexpansi e mapping mee s condi ion (C). We nex summa ize some o he basic p ope ies p o ed in [26] in ela ion o hese mappings. The p oo s o hese esul s a e me ic in na u e so he p ope ies also apply in he me ic case. Th oughou his pape we deno e he se o ixed poin s o a mapping by Fix( ). 5 Lemma 2.2. Le Xbe a Banach space and K∈P(X). Assume ha he mapping :K→Xsa is ies condi ion (C). Then o each x, y ∈K, (i) i z∈Fix( ), hen kz− (x)k≤kz−xk, ha is, is quasinonexpansi e; (ii) k (x)− (y)k ≤ kx−yko k 2(x)− (y)k≤k (x)−yk; (iii) kx− (y)k ≤ 3k (x)−xk+kx−yk. Using hese p ope ies, T. Suzuki [26] p o es ixed poin heo ems o mappings sa is ying condi ion (C). In [7], he au ho s s udy wo gene aliza ions o condi ion (C) gi ing examples and es ablishing ixed poin esul s. One o hese condi ions is he ollowing. De ini ion 2.3. Le Xbe a Banach space, K∈P(X), :K→Xand µ≥1. The mapping sa is ies condi ion (Eµ)i o all x, y ∈K, kx− (y)k ≤ µk (x)−xk+kx−yk. Lemma 2.2, (iii) yields ha condi ion (C) implies (E3), bu Example 3 o [7] shows ha (E3) does no imply (C). O he examples o di e en alues o µa e s udied in [7]. In he nex sec ions we will make use o he lemma below which is a special case o P oposi ion 2 in [9]. Lemma 2.4. Le Xbe a geodesic me ic space wi h con ex me ic, α∈(0,1) and (xn)n∈Nand (yn)n∈Nbounded sequences in Xsuch ha xn+1 = (1 −α)xn+αynand d(yn+1, yn)≤d(xn+1, xn) o e e y n∈N. Then limn→∞ d(xn, yn)=0. The ollowing wo heo ems we e p o ed in [23], bu in he se ing o a comple e CAT(0) space. I is easy o see ha hese esul s hold in mo e gene al con ex s. We will o mula e he i s esul in he amewo k o a uniquely geodesic me ic space. Theo em 2.5. Le Xbe a uniquely geodesic me ic space and K∈Pcl,c (X). Suppose :K→Ksa is ies condi ion (C)and Fix( )6=∅. Then Fix( )is closed and con ex. The p oo o he second heo em only equi es he uniqueness o he asymp o ic cen e and he con exi y o he me ic. This is why we s a e his esul unde he hypo hesis o a comple e UC space wi h con ex me ic. Theo em 2.6. Le Xbe a comple e UC space wi h con ex me ic and suppose K∈ Pb,cl,c (X). I :K→Ksa is ies condi ion (C) hen Fix( )is nonemp y, closed and con ex. In [23], he au ho s also ex end Suzuki’s [26] condi ion (C) o he mul i alued case in he ollowing way. De ini ion 2.7. Le Xbe a me ic space and K∈P(X). A mapping T:K→P(X)is said o sa is y condi ion (C)i o each x, y ∈Kand ux∈T(x)such ha 1 2d(x, ux)≤d(x, y), he e exis s uy∈T(y)such ha d(ux, uy)≤d(x, y). 6 The abo e condi ion is used in [23] o gi e a ixed poin heo em o mul i alued mappings and some common ixed poin esul s. In he es o his pape we use condi ion (C) o bo h single and mul i alued map- pings wi h he con ex dis inguishing be ween he wo cases. The same also holds o o he condi ions we make use o . 3 Fixed poin s and selec ions in geodesic spaces In his sec ion we s udy he mul i alued e sion o mappings wi h condi ion (C) in geodesic me ic spaces. Following he single alued case, we in oduce he nex condi ion and p o e ha o µ= 3 i is a gene aliza ion o condi ion (C). De ini ion 3.1. Le Xbe a me ic space, K∈P(X),T:K→P(X)and µ≥1. The mapping Tsa is ies condi ion (Eµ)i o each x, y ∈Kand ux∈T(x) he e exis s uy∈T(y)such ha d(x, uy)≤µd(x, ux) + d(x, y). We p o e nex ha a mul i alued mapping which sa is ies condi ion (C) also sa is ies (E3). This p ope y will cons i u e a key ool in p o ing ou esul s. Lemma 3.2. Le Xbe a me ic space, K∈P(X)and le T:K→P(K)sa is y condi ion (C). Then Tsa is ies condi ion (E3). P oo . Le x, y ∈Kand ux∈T(x). Because (1/2)d(x, ux)≤d(x, ux) he e exis s x∈T(ux) such ha d(ux, x)≤d(x, ux).(1) We p o e ha ei he 1 2d(x, ux)≤d(x, y) (2) o 1 2d(ux, x)≤d(ux, y) (3) holds. Suppose (1/2)d(x, ux)> d(x, y) and (1/2)d(ux, x)> d(ux, y). Then, using (1) we ob ain he ollowing con adic ion d(x, ux)≤d(x, y) + d(y, ux)<1 2d(x, ux) + 1 2d(ux, x)≤d(x, ux). Hence, i (2) holds, hen he e exis s uy∈T(y) such ha d(ux, uy)≤d(x, y), so d(x, uy)≤d(x, ux) + d(ux, uy)≤d(x, ux) + d(x, y). I (3) holds, hen he e exis s uy∈T(y) such ha d( x, uy)≤d(ux, y). Using again (1) we ha e ha d(x, uy)≤d(x, ux) + d(ux, x) + d( x, uy)≤2d(x, ux) + d(ux, y)≤3d(x, ux) + d(x, y). Thus, he inequali y holds in each o he wo cases and we a e done. 7 De ini ion 3.3. Le Xbe a me ic space, K∈P(X)and T:K→P(X). We say ha (xn)n∈N⊆Kis an app oxima e ixed poin sequence o he mapping Ti o each n∈N he e exis s yn∈T(xn)such ha limn→∞ d(xn, yn)=0. The nex esul p o ides an app oxima e ixed poin sequence o a mul i alued mapping sa is ying condi ion (C). We use his esul in he es o he pape because many o ou p oo s ely on i . P oposi ion 3.4. Le Xbe geodesic me ic space wi h con ex me ic, K∈Pb,c (X) and T:K→P(K). I Tsa is ies condi ion (C), hen Thas an app oxima e ixed poin sequence. P oo . Le x1∈K,y1∈T(x1) and ake x2= (1/2)x1+ (1/2)y1. Then (1/2)d(x1, y1) = d(x1, x2) so, by condi ion (C), he e exis s y2∈T(x2) such ha d(y1, y2)≤d(x1, x2). Con inuing in his ein, we can build he sequences (xn)n∈Nand (yn)n∈Nsuch ha yn∈T(xn), xn+1 = (1/2)xn+ (1/2)ynand d(yn+1, yn)≤d(xn+1, xn) o e e y n∈N. Using Lemma 2.4 we ob ain ha limn→∞ d(xn, yn) = 0. Ou i s ixed poin esul o mul i alued mappings is gi en o sel -mappings on a compac se . Theo em 3.5. Le Xbe a geodesic space wi h con ex me ic and K∈Pcp,c (X). Sup- pose T:K→Pcl(K)sa is ies condi ion (C). Then Fix(T)6=∅. P oo . By P oposi ion 3.4, he e exis wo sequences (xn)n∈Nand (yn)n∈Nin Ksuch ha yn∈T(xn) and limn→∞ d(xn, yn) = 0. Since Kis compac , we can ind a subsequence (xnk)k∈No (xn)n∈Nsuch ha (xnk)k∈Ncon e ges o some x∈K. Using Lemma 3.2, we ha e ha o all k∈N dis (xnk, T(x)) ≤3d(xnk, ynk) + d(xnk, x). Taking he limi as k→ ∞ we ob ain ha dis (x, T (x)) = 0. Since T(x) is closed i ollows ha x∈T(x). In he ollowing heo em we mo e he compac ness condi ion om he domain o he images o he mapping. This heo em is ac ually an ex ension o Theo em 3.2 o [23] in he con ex o a comple e UC space wi h con ex me ic. We also emo e he con exi y condi ion on he image se s o he mapping. Mo eo e , we ob ain ou esul s in a simple way as a consequence o Lemma 3.2 which a oids o go h ough a delica e poin in he p oo o Theo em 3.2 o [23]. Theo em 3.6. Le Xbe a comple e UC space wi h con ex me ic and K∈Pb,cl,c (X). Suppose T:K→Pcp(K)sa is ies condi ion (C). Then Fix(T)6=∅. P oo . By P oposi ion 3.4, we can ind he sequences (xn)n∈Nand (yn)n∈Nin Ksuch ha yn∈T(xn) and limn→∞ d(xn, yn) = 0. As explained in Sec ion 2, we may suppose ha (xn)n∈Nis egula (o he wise choose a egula subsequence o i ). Deno e he unique asymp o ic cen e o (xn)n∈Nby x. Le n∈N. Applying Lemma 3.2 o xn, x and yn espec i ely i ollows ha he e exis s zn∈T(x) such ha d(xn, zn)≤3d(xn, yn) + d(xn, x). 8 Le (znk)k∈Nbe a subsequence o (zn)n∈N ha con e ges o some z∈T(x). Then, o each k∈N, d(xnk, z)≤d(xnk, znk) + d(znk, z)≤3d(xnk, ynk) + d(xnk, x) + d(znk, z). Taking he supe io limi as k→ ∞ and knowing ha he asymp o ic cen e o (xnk)k∈N is p ecisely xwe ob ain ha x=z∈T(x). Hence, he p oo is comple e. Rema k 3.7. F om he abo e p oo i is immedia e ha in Theo em 3.6 we can d op he con exi y o he me ic and assume ins ead ha he mapping admi s an app oxima e ixed poin sequence. In he nex esul we will conside he ollowing new condi ion o mul i alued map- pings which will be shown o be weake han condi ion (C). De ini ion 3.8. Le Xbe a me ic space, K∈P(X)and T:K→P(X). The mapping Tsa is ies condi ion (C0)i o each x, y ∈Kand ux∈T(x)wi h d(x, ux) = dis (x, T(x)) and 1 2d(x, ux)≤d(x, y), he e exis s uy∈T(y)such ha d(ux, uy)≤d(x, y). We p o e nex a selec ion heo em in R- ees o mul i alued mappings sa is ying condi ion (C0) and analyze a e wa ds he ela ion o (C0) o (C) and (E3) espec i ely. Theo em 3.9. Le Xbe an R- ee, K∈P(X)and T:K→Pcl,c (X)a mapping which sa is ies (C0). Then he mapping :K→Xde ined by (x) = PT(x)(x) o each x∈K is a selec ion o T ha sa is ies condi ion (C). P oo . No ice ha he p ope ies o R- ees (see Sec ion 2) gua an ee ha is well- de ined. Le x, y ∈Ksuch ha (x)6= (y) and (1/2)d(x, (x)) ≤d(x, y). Conside p(x) = PT(y)( (x)) and p(y) = PT(x)( (y)). Fi s , suppose p(x)6= (y) and p(y)6= (x). Since p(x) is he p ojec ion o (x) on o T(y) i ollows ha d( (x), (y)) = d( (x), p(x)) + d(p(x), (y)), i.e., p(x)∈[ (x), (y)]. Since T(y) is con ex, [p(x), (y)] ⊆T(y). This implies [ (x), (y)] ∩[ (y), y] = { (y)}because o he wise he minimali y o (y) would be con adic ed. Thus, (y)∈[ (x), y]. Simila ly, (x)∈[ (y), x]. Then (x), (y)∈[x, y] (o he wise supposing o example ha z∈[x, (y)]∩[ (y), y] wi h z6= (y) we ha e ha (x)∈[z, (y)] and (y)∈[z, (x)] which is alse). The e o e, d( (x), (y)) ≤d(x, y). In ac , d( (x), (y)) = d(x, y)−dis (x, T(x)) −dis (y, T(y)). Now assume p(x) = (y). Then d( (x), (y)) = dis ( (x), T(y)) and so, by condi ion (C0), d( (x), (y)) = dis ( (x), T(y)) ≤d(x, y). 9 Recall ha Xis said o be uni o mly con ex in e e y di ec ion (UCED, in sho ) i δz()>0 o all  > 0 and z∈Xwi h kzk= 1, whe e δz() is he modulus o con exi y o Xin he di ec ion zde ined by δz() = in 1−1 2kx+yk:kxk ≤ 1,kyk ≤ 1, x −y=z. Ob iously, uni o mly con ex Banach spaces a e UCED. I is known ha in a UCED Banach space, he asymp o ic cen e o a sequence wi h espec o a weakly compac con ex se is a single on. Hence, e e y egula sequence wi h espec o such a se is asymp o ically uni o m. Theo em 4.6. Le Kbe a weakly compac and con ex subse o a UCED Banach space X. Suppose T:K→Pcp(K)is a mapping sa is ying condi ion (C). Then Fix(T)6=∅. P oo . Le (xn)n∈Nand (yn)n∈Nbe wo sequences in Ksuch ha yn∈T(xn) and limn→∞ kxn−ynk= 0. Wi hou loss o gene ali y, me may assume ha (xn)n∈Nis egula wi h espec o K. Le zbe he unique poin in he asymp o ic cen e o (xn)n∈N in K. By Lemma 3.2, o each n∈N he e exis s n∈T(z) such ha kxn− nk ≤ 3kxn−ynk+kxn−zk. F om he compac ness o T(z) we can assume ha ( n)n∈Ncon e ges o a poin ∈T(z). I ollows ha lim sup n→∞ kxn− k ≤ lim sup n→∞ kxn−zk. Since (xn)n∈Nis egula we conclude ha =z∈T(z). Dhompongsa e al. [3] ha e ecen ly p o ed he Tin a iance o he asymp o ic cen e in Ko an app oxima e ixed poin sequence o T, when Tis a single alued mapping sa is ying condi ion (C). We now s a e a esul which can be seen as an adap a ion o his ac o he mul i alued case. P oposi ion 4.7. Le Kbe a weakly compac subse o a Banach space X. Suppose T:K→Pcp(K)sa is ies condi ion (C)and (xn)n∈Nis an app oxima e ixed poin sequence o T. Then, he e exis s a subsequence (zn)n∈No (xn)n∈Nsuch ha T(x)∩A6=∅, o all x∈A:= A(K, (zn)). P oo . Since Tis a sel -mapping we can build a subsequence (zn)n∈No (xn)n∈Nwhich is egula and asymp o ically uni o m wi h espec o K. Deno e (K, (zn)) by . Taking any x∈Aand ollowing he same a gumen as in he p oo o he abo e heo em we ob ain a sequence ( n)n∈N⊆T(x) no m con e gen o a poin ∈T(x) such ha lim sup n→∞ kxn− k ≤ lim sup n→∞ kxn−xk= . This shows ha ∈A, and so T(x)∩A6=∅. Now we a e eady o p o e an analogous esul o he Ki k-Massa heo em [16] o mappings sa is ying condi ion (C). 16 Theo em 4.8. Le Kbe a bounded, closed and con ex subse o a Banach space X and T:K→Pcp,c (K)be a con inuous mapping wi h espec o he Pompeiu-Hausdo dis ance sa is ying condi ion (C). Suppose ha each sequence in Khas a nonemp y and compac asymp o ic cen e ela i e o K. Then Fix(T)6=∅. P oo . Acco ding o he p e ious p oposi ion we can ake a sequence (xn)n∈Nin Ksuch ha T(x)∩A6=∅, o all x∈A:= A(K, (xn)). Now we de ine he mapping ˜ T:A→Pcp,c (A) by ˜ T(x) = T(x)∩A. Since Tis con inuous, om P oposi ion 2.45 in [11] we know ha he mapping ˜ Tis uppe semi-con inuous. Since T(x)∩Ais a compac con ex se we can apply he Kaku ani-Bohnenblus -Ka lin heo em (see [10]) o ob ain a ixed poin o ˜ Tand hence o T. Rema k 4.9. Recall ha a mul i alued mapping T:K→Pb(X) is said o be nonex- pansi e i H(T(x), T(y)) ≤ kx−yk o all x, y ∈K, whe e Hdeno es he Pompeiu-Hausdo dis ance. I is wo h poin ing ou ha ano he na u al ex ension o he Suzuki’s condi ion (C) o a mul i alued mapping T:K→P(X) is he ollowing: o all x, y ∈K 1 2dis (x, T(x)) ≤ kx−yk=⇒H(T(x), T(y)) ≤ kx−yk. Ob iously, a nonexpansi e mapping mee s he abo e condi ion. Howe e , i is no clea i a mapping sa is ying he abo e condi ion also sa is ies (C). S ill, i T akes compac alues is easy o see ha his new condi ion implies condi ion (C). Since in ou heo ems Tis assumed o be compac alued, such esul s gene alize classical ixed poin heo ems o mul i alued mappings (see [16],[19],[20]). 5 Common ixed poin s In ou las sec ion we will apply some o he ixed poin heo ems s a ed in p e ious sec ions o ob ain esul s on he exis ence o common ixed poin . De ini ion 5.1. Le Xbe a me ic space and K∈P(X). Suppose :K→Kand T:K→P(K). Then and Ta e commu ing mappings i (y)∈T( (x)) o all x∈K and y∈T(x). We s a by gi ing a lemma ha will cons i u e a main ool in p o ing ou esul s. Lemma 5.2. Le Xbe a me ic space, K∈P(X), :K→Ksa is ying condi ion (C) and wi h Fix( )6=∅. Supppose T:K→P(K)is such ha o e e y x, y ∈Fix( ), he se PT(y)(x)is a single on. I and Tcommu e, hen PT(y)(x)∈Fix( ) o all x, y ∈Fix( ). P oo . Le x, y ∈Fix( ) and deno e PT(y)(x) by u. Because mee s condi ion (C) and 0 = (1/2)d(x, (x)) ≤d(x, u) we ob ain ha d(x, (u)) = d( (x), (u)) ≤d(x, u) = dis (x, T(y)). Bu (u)∈T(y) because and Tcommu e, y∈Fix( ) and u∈T(y). Hence, (u) = uand he conclusion ollows. 17 The ollowing heo em is an ex ension o Theo em 4.2 o [23] in he se ing o a UC space wi h con ex me ic. No ice ha ou app oach is di e en in he second hal o he p oo om ha o [23]. In pa icula , ou s ills a gap in he p oo o [23]. No ice also ha his heo em ex ends some o he esul s in he heo y, see, o ins ance, [6, 24]. Theo em 5.3. Le Xbe a comple e UC space wi h con ex me ic and K∈Pb,cl,c (X). Suppose :K→Kand T:K→Pcp,c (K)sa is y condi ion (C). I and Tcommu e, hen he e exis s z∈Ksuch ha z= (z)∈T(z). P oo . Using Theo em 2.6, i ollows ha Fix( ) is nonemp y, closed and con ex. Since he se ing we wo k in is a UC space, he p ojec ion on o each compac and con ex se is a single on. By Lemma 5.2, PT(x)(x)∈T(x)∩Fix( ) o each x∈Fix( ) and so we can conside he mapping T(·)∩Fix( ) : Fix( )→Pcp(Fix( )). We show ha his mapping sa is ies condi ion (C). Le x, y ∈Fix( ), ux∈T(x)∩Fix( ) such ha (1/2)d(x, ux)≤ d(x, y). Since T ul ills (C), he e exis s y∈T(y) such ha d(ux, y)≤d(x, y). Le uy s and o PT(y)(ux). Acco ding o Lemma 5.2, uy∈T(y)∩Fix( ). I is also clea ha d(ux, uy)≤d(ux, y)≤d(x, y). Thus, he mapping T(·)∩Fix( ) : Fix( )→Pcp(Fix( )) sa is ies (C) which means, using Theo em 3.6, ha he e exis s z∈Ksuch ha z= (z)∈T(z). Likewise, one can p o e he ollowing esul in he amewo k o R- ees. Theo em 5.4. Le Xbe a bounded comple e R- ee. Suppose :X→Xand T:X→ Pcl,c (X)sa is y condi ions (C)and (C0) espec i ely. I and Tcommu e, hen he e exis s z∈Ksuch ha z= (z)∈T(z). P oo . Acco ding o Theo em 2.6, i ollows ha Fix( ) is nonemp y, closed and con ex (and so also hype con ex). This means ha Fix( ) is in i s own u n a comple e R- ee. Since in an R- ee he p ojec ion on o each closed and con ex se is a single on we can apply Lemma 5.2 and so T(x)∩Fix( )6=∅ o each x∈Fix( ). Now conside he mapping T(·)∩Fix( ) : Fix( )→Pcl,c (Fix( )). We show ha his mapping sa is ies condi ion (C0). Le x, y ∈Fix( ), ux∈T(x)∩Fix( ) such ha d(x, ux) = dis (x, T(x)∩ Fix( )) and (1/2)d(x, ux)≤d(x, y). Applying Lemma 5.2, PT(x)(x)∈T(x)∩Fix( ) which implies ha dis (x, T (x)) = dis (x, T(x)∩Fix( )), so d(x, ux) = dis (x, T(x)). Because Tsa is ies (C0), he e exis s y∈T(y) such ha d(ux, y)≤d(x, y). Le uy s and o PT(y)(ux). Acco ding o Lemma 5.2, uy∈T(y)∩Fix( ). I is also clea ha d(ux, uy)≤d(ux, y)≤d(x, y). Thus, he mapping T(·)∩Fix( ) : Fix( )→ Pcl,c (Fix( )) sa is ies (C0) which means, using Co olla y 3.11, ha he e exis s z∈K such ha z= (z)∈T(z). 6 Acknowledgmen s We would like o hank A apol Kaewkhao o no icing and le ing us know abou a mis ake in a p e ious e sion o his pape . The esea ch o he i s wo au ho s was pa ially suppo ed by DGES, G an MTM2009-10696-C02-01 and Jun a de Andaluc´ıa, G an FQM-127. The hi d au ho was suppo ed by p og ams co- inanced by he Sec o al Ope a ional P og amme Human Resou ces De elopmen , Con ac POS DRU 6/1.5/S/3 - “Doc o al s udies: h ough 18 science owa ds socie y”. She would also like o exp ess he app ecia ion o he Depa - men o Ma hema ical Analysis and o he Ins i u e o Ma hema ics o he Uni e si y o Se ille (IMUS) o hei suppo . Re e ences [1] J.B. Baillon, Nonexpansi e mapping and hype con ex spaces, Con emp. Ma h. 72 (1988) 11-19. [2] M.R. B idson, A. Hae lige , Me ic spaces o non-posi i e cu a u e, Sp inge - Ve lag, Be lin, 1999. [3] S. Dhompongsa, W. In hakon, A. Kaewkhao, Edels ein’s me hod and ixed poin heo ems o some gene alized nonexpansi e mappings, J. Ma h. Anal. Appl. 350 (2009) 12-17. [4] R. Esp´ınola, W.A. Ki k, Fixed poin heo ems in R- ees wi h applica ions o g aph heo y, Topology Appl. 153 (2006) 1046-1055. [5] R. Esp´ınola, A. Fe n´andez-Le´on, B. Pi¸a ek, Fixed poin s o single- and se - alued mappings in uni o mly con ex me ic spaces wi h no me ic con exi y, Fixed Poin Theo y and Appl. 2010 (2010) A icle ID 169837 16 pages. [6] R. Esp´ınola, N. Hussain, Common ixed poin s o mul imaps in me ic spaces, Fixed Poin Theo y and Appl., Volume 2010 (2010), A icle ID 204981, 14 pages [7] J. Ga c´ıa-False , E. Llo ens-Fus e , T. Suzuki, Fixed poin heo y o a class o gene alized nonexpansi e mappings, doi:10.1016/j.jmaa.2010.08.069. [8] K. Goebel, On a ixed poin heo em o mul i alued nonexpansi e mappings, Ann. Uni . Ma ie Cu ie-Sklodowska 29 (1975) 70-72. [9] K. Goebel, W.A. Ki k, I e a ion p ocesses o nonexpansi e mappings, Con emp. Ma h. 21 (1983) 115-123. [10] K. Goebel, W.A. Ki k, Topics in Me ic Fixed Poin Theo y, Camb idge Uni . P ess, 1990. [11] S. Hu, N. Papageo giou, Handbook o Mul i alued Analysis, ol. 1, Kluwe Aca- demic Publishe s, Do d ech , 1997. [12] M.A. Khamsi, W.A. Ki k, C. Ma ´ınez Y´a˜nez, Fixed poin and selec ion heo ems in hype con ex spaces, P oc. Ame . Ma h. Soc. 128 (2000) 3275-3283. [13] W.A. Ki k, Nonexpansi e mappings in p oduc spaces, se - alued mappings and k-uni o mly o undi y, (F. E. B owde , ed.) Ame . Ma h. Soc. Symp. Pu e Ma h. 45 (1986) 51-64. [14] W.A. Ki k, Hype con exi y o R- ees, Fund. Ma h. 156 (1998) 67-72. 19 [15] W.A. Ki k, Some ecen esul s in me ic ixed poin heo y, J. Fixed Poin Theo y Appl. 2 (2007), 195-207. [16] W.A. Ki k, S. Massa, Rema ks on asymp o ic and Chebyshe cen e s, Hous on J. Ma h. 16 (1990) 357-364. [17] W.A. Ki k, B. Sims (eds.), Handbook o Me ic Fixed Poin Theo y, Kluwe Aca- demic Publishe s, Do d ech , 2001. [18] U. Kohlenbach, L. Leu¸s ean, Asymp o ically nonexpansi e mappings in uni o mly con ex hype bolic spaces, J. Eu . Ma h. Soc. 12 (2010) 71-92. [19] E. Lami Dozo, Mul i alued nonexpansi e mappings and Opial’s condi ion, P oc. Ame . Ma h. Soc. 38 (1973) 286-292. [20] T.C. Lim, A ixed poin heo em o mul i alued nonexpansi e mappings in a uni- o mly con ex Banach space, Bull. Ame . Ma h. Soc. 80 (1974) 1123-1126. [21] J.T. Ma kin, Fixed poin s, selec ions and bes app oxima ion o mul i alued map- pings in R- ees, Nonlinea Anal. 67 (2007) 2712–2716. [22] B. Pi¸a ek, Bes app oxima ion o coincidence poin s in me ic ees, Ann. Uni . Ma iae Cu ie-Sk lodowska Sec . A 62 (2008) 113–121. [23] A. Razani, H. Salahi a d, In a ian app oxima ion o CAT(0) spaces, Nonlinea Anal. 72 (2010) 2421-2425. [24] N. Shahzad, Fixed poin esul s o mul imaps in CAT(0) spaces, Topology Appl. 156 (2009), 997-1001. [25] R. Sine, Hype con exi y and nonexpansi e mul i unc ions, T ans. Ame . Ma h. Soc. 315 (1989) 755–767. [26] T. Suzuki, Fixed poin heo ems and con e gence heo ems o some gene alized nonexpansi e mappings, J. Ma h. Anal. Appl. 340 (2008) 1088-1095. 20