ISSN: 0163-0563 (Print) 1532-2467 (Online) Journal homepage: www.tandfonline.

com/journals/lnfa20

Proximinal Retracts and Best Proximity Pair

Theorems

W. A. Kirk, Simeon Reich & P. Veeramani

To cite this article: W. A. Kirk, Simeon Reich & P. Veeramani (2003) Proximinal Retracts and
Best Proximity Pair Theorems, , 24:7-8, 851-862, DOI: 10.1081/NFA-120026380

To link to this article: https://doi.org/10.1081/NFA-120026380

Published online: 07 Feb 2007.

Submit your article to this journal

Article views: 769

View related articles

Citing articles: 23 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=lnfa20
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

NUMERICAL FUNCTIONAL ANALYSIS AND OPTIMIZATION

Vol. 24, Nos. 7 & 8, pp. 851–862, 2003

Proximinal Retracts and Best Proximity Pair Theorems

W. A. Kirk,1,* Simeon Reich,2 and P. Veeramani3

1
Department of Mathematics, University of Iowa,
Iowa City, Iowa, USA
2
Department of Mathematics, The Technion–Israel Institute of
Technology, Haifa, Israel
3
Department of Mathematics, Indian Institute of Technology,
Chennai, India

ABSTRACT

This note is concerned with proximinality and best proximity pair theorems in
hyperconvex metric spaces and in Hilbert spaces. Given two subsets A and B of a
metric space and a mapping T : A ! 2B , best proximity pair theorems provide
sufficient conditions that ensure the existence of an x 2 A such that

distðx, TðxÞÞ ¼ distðA, BÞ:

Thus such theorems provide optimal approximate solutions in the case that the
mapping T does not have fixed points.

Key Words: Best proximity pairs; Hyperconvex spaces; Nonexpansive mappings;

Proximinal sets.

AMS Subject Classification: 47H09, 41A65, 46B20.

*Correspondence: W. A. Kirk, Department of Mathematics, University of Iowa, Iowa City,

IA 52242-1419, USA; E-mail: [email protected].

851

DOI: 10.1081/NFA-120026380 0163-0563 (Print); 1532-2467 (Online)

Copyright & 2003 by Marcel Dekker, Inc. www.dekker.com
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

852 Kirk, Reich, and Veeramani

1. INTRODUCTION

In this note we are concerned with proximinality and best proximity pair
solutions for mappings defined in hyperconvex metric spaces (which include the
classical L1 spaces) and in Hilbert spaces. A subset E of a metric space M is said
to be proximinal if given any x 2 M there exists px 2 E such that dðx, px Þ ¼
distðx, E Þ def
¼ inffdðx, yÞ : y 2 Eg:

Definition 1.1. Let X be a metric space and let A and B be nonempty subsets
of X: Let

A0 ¼ fx 2 A : dðx, yÞ ¼ dist ðA, BÞ for some y 2 Bg;

B0 ¼ fx 2 B : dðx, yÞ ¼ dist ðA, BÞ for some y 2 Ag:

A pair ðx, yÞ 2 A0
B0 for which dðx, yÞ ¼ distðA, BÞ is called a best proximity pair
for A and B.
In terms of the preceding definition, the basic questions we consider are
these. Given subsets A and B of a metric space M: (I) When are A and B
mutually proximinal in the sense that there exists a best proximity pair
ða, bÞ 2 A0
B0 ? (II) Given that A and B are mutually proximinal (and
disjoint) and given a mapping T : A ! 2B , when does the mapping T have a best
proximity pair solution, that is, when does there exist a best proximity pair
ðx, yÞ 2 A0
B0 such that y 2 TðxÞ? In this case one also has dðx, yÞ ¼ inffdistðx,
TðxÞÞ : x 2 Ag: Obviously such results reduce to fixed point theorems when
A \ B 6¼ ;:
Fixed point theory is an important tool for solving equations TðxÞ ¼ x for
mappings T : D ! X (and especially mappings T : D ! D) where D is a subset of
a metric or Banach space. However, if T does not have fixed points, then one often
tries to find an element x which is in some sense closest to TðxÞ: A classical result in
this direction is a best approximation theorem due to Ky Fan (1969) which states
that if K is a nonempty compact convex subset of a normed linear space (or more
generally, a locally convex Hausdorff topological vector space) X and if T : K ! X
is a single-valued continuous mapping, then there exists an element x 2 K such that
kx  TðxÞk ¼ distðTðxÞ, KÞ: There have been many subsequent extensions and
variants of Fan’s theorem, including those by Reich (1978), Sehgal and Singh
(1988, 1989), and Prolla (1982–1983). A unified approach to such results can be
found in Vetrivel et al. (1992).
On the other hand, even though best approximation theorems guarantee
the existence of approximate solutions, such results need not give optimal solutions.
In contrast to this, best proximity pair theorems yield approximate solutions
that are optimal. Sufficient conditions that ensure the existence of an
element x 2 A such that distðx, TðxÞÞ ¼ distðA, BÞ where T : A ! 2B is a
multifunction defined on suitable subsets A and B of a normed linear space
(or a locally convex Hausdorff topological vector space) can be found in
Sadiq Basha and Veeramani (1997, 2000), Srinivasan (2001), and Sadiq Basha
et al. (2001).
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Proximinal Retracts and Best Proximity Pair Theorems 853

2. HYPERCONVEX SPACES

In this section we consider hyperconvex metric spaces and properties of several

well defined subspaces of such spaces. For a historical discussion and a detailed
survey of the general properties of hyperconvex spaces we refer to Espı́nola and
Khamsi (2001). We begin with three basic definitions.

Definition 2.1. A metric space M is said to be hyperconvex if given any family fx
g
of points of M and any family fr
g of real numbers satisfying
dðx
, x Þ  r
þ r
it is the case that \
Bðx
; r
Þ 6¼ ;:

Definition 2.2. A subset A of a metric space M is said to be admissible if it is an

intersection of closed balls of M: Thus A is admissible if A ¼ \i2I Bðxi ; ri Þ where
xi 2 M and ri  0, i 2 I:

Definition 2.3. A subset E of a metric space M is said to be externally hyperconvex

(relative to M) if given any family fx
g of points of M and any family fr
g of real
numbers satisfying for each
, ,
dðx
, x Þ  r
þ r and dist ðx
, E Þ  r
,
it follows that \
Bðx
; r
Þ \ E 6¼ ;:

Among other things, it is shown in Khamsi et al. (2000) that an externally

hyperconvex subset of a hyperconvex space M is always proximinal. It is also
shown in Khamsi et al. (2000) that if M is hyperconvex and if a set-valued mapping
T  from a set S into M takes values in the family of bounded nonempty externally
hyperconvex subsets of M, then T  always has a single-valued selection T which
satisfies
dðTðxÞ, Tð yÞÞ  dH ðT  ðxÞ, T  ð yÞÞ for all x, y 2 S,
where dH denotes the usual Hausdorff metric on the family of nonempty bounded
closed subsets of M: Thus T is continuous if T  is continuous, and if T  is lipschit-
zian, then T is as well and has the same Lipschitz constant.
If a set E in M is proximinal, then there exists a (possibly set-valued) proximinal
retraction of M onto E: One of the nicest properties such a retract can have is the
following one.

Definition 2.4. A subset E of a metric space M is a proximinal nonexpansive retract of

M if there exists a nonexpansive retraction R of M onto E for which
dðx, RðxÞÞ ¼ distðx, E Þ
for each x 2 M: Thus dðRðxÞ, RðyÞÞ  dðx, yÞ for each x, y 2 M:

In an effort to characterize those subsets of a hyperconvex metric space which

are proximinal nonexpansive retracts, the following definition was introduced in
Espı́nola et al. (2000).
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

854 Kirk, Reich, and Veeramani

Definition 2.5. A subset E of a metric space M is said to be weakly externally

hyperconvex (relative to M) if E is externally hyperconvex relative to E [ fzg for
each z 2 M: Precisely, given any family fx
g of points of M all but at most one of
which lies in E, and any family fr
g of real numbers satisfying
dðx
, x Þ  r
þ r , with distðx
, E Þ  r
if x
62 E,
it follows that \
Bðx
; r
Þ \ E 6¼ ;:

The following implications are well-known. Only the first requires some justifi-
cation; the remaining two are immediate from the definitions (see, e.g., Espı́nola and
Khamsi, 2001, p. 398). Given a subset A of a hyperconvex metric space M, then

A is admissible ) A is externally hyperconvex

) A is weakly externally hyperconvex
) A is hyperconvex.

Among other things, it is shown in Espı́nola et al. (2000) that a compact subset D
of a hyperconvex metric space M is weakly externally hyperconvex if and only if it is
a proximinal nonexpansive retract of M:

Definition 2.6 (Sine, 1989a). Let S be a subset of a metric space M: A mapping

R : S ! M is said to be "-constant if dðx, RxÞ  " for each x 2 S:

For any subset A of a metric space, let

N" ðAÞ ¼ [ fBða; "Þ : a 2 Ag:

We will need the following fact.

Lemma 2.7 (Sine, 1989a). Let D ¼ \Bðx
; r
Þ be a nonempty admissible set in a

hyperconvex metric space M: Then for any " > 0 the set
N" ðDÞ ¼ \ Bðx
; r
þ "Þ:

Hence N" ðDÞ is admissible. Moreover, there is an "-constant nonexpansive retraction

of N" ðDÞ onto D:

Proposition 2.8. Let M be a hyperconvex metric space and let A and B be a nonempty
admissible subsets of M. Then A0 and B0 are nonempty and hyperconvex.

Proof. Since B is admissible, there exists a proximinal nonexpansive retraction PB of

M onto B (Sine, 1989b, Corollary 11). Let pB denote the restriction of this retraction
to A: Let d ¼ distðA, BÞ and let "n ¼ ðd þ 1=nÞ: Consider the set
An :¼ fx 2 A : dðx, pB ðxÞÞ  "n g:
Obviously An is nonempty. We claim that An is hyperconvex. To see this, let fx
g be
points of An and let fr
g  Rþ be such that
dðx
, x Þ  r
þ r :
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Proximinal Retracts and Best Proximity Pair Theorems 855

Let D be the admissible set

D :¼ \Bðx
; r
Þ \ A:
Clearly D is nonempty since A is hyperconvex. By Lemma 2.7 there exists an
"n -constant retraction  that maps N"n ðDÞ onto D: If x 2 D, then pB ðxÞ 2 N"n ðDÞ
because
dð pB ðxÞ, x
Þ  dð pB ðxÞ, pB ðx
ÞÞ þ dð pB ðx
Þ, x
Þ
 dðx, x
Þ þ "n  r
þ "n :
Now define R : D ! D by setting R ¼   pB : Since R is nonexpansive, it has a fixed
point x0 2 D: But  is "n -constant; so
dðx0 , pB ðx0 ÞÞ ¼ dð  pB ðx0 Þ, pB ðx0 ÞÞ  "n :

Therefore x0 2 \Bðx
; r
Þ \ An , and this proves that An is hyperconvex.

We now have a descending sequence fAn g of nonempty bounded hyperconvex
subsets of A, so by Baillon’s theorem (Baillon, 1988), \1 n¼1 An is nonempty and
hyperconvex. Clearly u 2 \1 A
n¼1 n , dðu, p B ðuÞÞ ¼ d ¼ distðA, BÞ , u 2 A0 and
pB ðuÞ 2 B0 : g

Notice that the preceding proof also gives the following result.

Lemma 2.9. Let M be a hyperconvex metric space, let A be a compact weakly exter-
nally hyperconvex subset of M, and let B be a nonempty admissible subset of M: Then
A0 and B0 are nonempty and hyperconvex.

Theorem 2.10. Let A and B be two admissible subsets of a hyperconvex metric space
ðM, dÞ, and suppose T  : A ! 2B is such that:

(i) for each x 2 A, T  x is a nonempty admissible (more generally, externally

hyperconvex) subset of B;
(ii) T  : ðA, dÞ ! ð2B , dH Þ is nonexpansive;
(iii) T  ðA0 Þ  B0 :

Then there exists x0 2 A such that

distðx0 , T  ðx0 ÞÞ ¼ distðA, BÞ ¼ inffdðx, T  ðxÞÞ : x 2 Ag:

Proof. First observe that A0 6¼ ; by Proposition 2.8. By Theorem 1 of Khamsi et al.

(2000), the mapping T  has a nonexpansive selection which we denote by T: Since A
is admissible, it is a proximinal nonexpansive retract of X (Sine, 1989b). Let r be a
nonexpansive proximinal retraction of X onto A and let x 2 A0 : Since r  TðxÞ 2 A is
a best approximation to TðxÞ, it follows that TðxÞ 2 B0 : Therefore there exists a 2 A
such that
dðTðxÞ, aÞ ¼ distðA, BÞ:
On the other hand, dðTðxÞ, r  TðxÞÞ  dðTðxÞ, aÞ: Therefore
dðTðxÞ, r  TðxÞÞ ¼ distðA, BÞ:
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

856 Kirk, Reich, and Veeramani

Thus r  T : A0 ! A0 : Since r  T is nonexpansive and A0 is hyperconvex, there

exists x0 2 A0 such that

r  Tðx0 Þ ¼ x0 :

It follows that ðx0 , Tðx0 ÞÞ is a best proximity pair for T; hence also for T  : g

The following version assumes only continuity of T  in exchange for a

compactness assumption on A:

Theorem 2.11. Let A and B be two closed convex subsets of a hyperconvex metric space
M. Suppose A is compact and weakly externally hyperconvex, and that B is admissible.
Suppose T  : A ! 2B is such that:

(i) for each x 2 A, T  x is a nonempty externally hyperconvex subset of B;

(ii) T  : ðA, dÞ ! ð2B , dH Þ is continuous;
(iii) T  ðA0 Þ  B0 :

Then there exists a best proximity pair for T  , i.e., there exists x0 2 A such that

distðx0 , T  ðx0 ÞÞ ¼ distðA, BÞ ¼ inffdðx, T  ðxÞÞ : x 2 Ag:

Proof. First observe that A0 6¼ ; by Lemma 2.9. By Theorem 1 of Khamsi et al.

(2000), the mapping T  has a continuous selection which we denote by T: Since A is
compact and weakly externally hyperconvex, it is a proximinal nonexpansive retract
of X (Espı́nola et al., 2000). Let r be a nonexpansive proximinal retraction of X onto
A: Let x 2 A0 : Since r  TðxÞ 2 A is a best approximation to TðxÞ, it follows that
TðxÞ 2 B0 : Therefore there exists a 2 A such that

dðTðxÞ, aÞ ¼ distðA, BÞ:

On the other hand, dðTðxÞ, r  TðxÞÞ  dðTðxÞ, aÞ: Therefore

dðTðxÞ, r  TðxÞÞ ¼ distðA, BÞ:

Thus r  T : A0 ! A0 : Since A0 is compact and hyperconvex and r  T is continuous,

there exists (by Khamsi (1996) and Kirk and Shin (1997)) x0 2 A0 such that

r  Tðx0 Þ ¼ x0 :

It follows that ðx0 , Tðx0 ÞÞ is a best proximity pair for T; hence also for T  : g

A natural question to ask is whether it is possible to replace the assumption that

the sets A and B are admissible in Theorem 2.10 with the weaker assumption that these
sets are either externally hyperconvex or weakly externally hyperconvex. At this point
the answer is unknown. However, it is known that if D is an externally hyperconvex
(respectively, a weakly externally hyperconvex) subset of a hyperconvex metric space,
then N" ðDÞ is also externally hyperconvex (respectively, weakly externally hypercon-
vex) (Khamsi et al. (2000), respectively Espı́nola et al. (2000)). This fact and the
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Proximinal Retracts and Best Proximity Pair Theorems 857

following observations suggest that such a weakening might be possible, and indeed
this is the case at least in more concrete spaces.

Lemma 2.12. Let M be hyperconvex and let D  M be (weakly) externally

hyperconvex: Then for any " > 0 the set N" ðDÞ is weakly externally hyperconvex.
Moreover, there is an "-constant nonexpansive retraction of N" ðDÞ onto D:

Proof. See Espı́nola et al. (2000). Indeed, the existence of such a retraction char-
acterizes the weakly externally hyperconvex subsets of a hyperconvex (Espı́nola and
Khamsi 2001, Theorem 4.15). g

In Khamsi et al. (2000, Lemma 2) it is proved that the intersection of an

externally hyperconvex set and an admissible set in a hyperconvex space is
again externally hyperconvex. The proof of that result carries over to establish the
following fact.

Lemma 2.13. Let M be hyperconvex, suppose D is an admissible subset of M, and

suppose E is a weakly externally hyperconvex subset of M: Then D \ E is weakly
externally hyperconvex.

We are now in a position to prove the following extension of Proposition 2.8.

Proposition 2.14. Let M be a hyperconvex metric space, let A be a nonempty weakly

externally hyperconvex subset of M, and let B be a nonempty proximinal nonexpansive
retract of M. Then A0 and B0 are nonempty and A0 is hyperconvex.

Proof. By assumption there exists a proximinal nonexpansive retraction PB of M

onto B. Let pB denote the restriction of this retraction to A: Let d ¼distðA, BÞ and
as before let "n ¼ ðd þ 1=nÞ: Consider the set

An :¼ fx 2 A : dðx, pB ðxÞÞ  "n g:

Obviously An is nonempty. As before, we claim that An is hyperconvex. To see this,

let fx
g be points of An and let fr
g  Rþ be such that

dðx
, x Þ  r
þ r :

Let D ¼ \Bðx
; r
Þ \ A: Clearly D is nonempty since A is hyperconvex. Also, since D

is the intersection of an admissible set and a weakly externally hyperconvex set,
D is weakly externally hyperconvex by Lemma 2.13. Therefore, by Lemma 2.12,
there exists an "n -constant retraction  that maps N"n ðDÞ onto D: By following the
argument of Lemma 2.8 we conclude that A0 and B0 are nonempty, and that A0 is
hyperconvex. g

It is noted in Espı́nola et al. (2000) that a weakly externally hyperconvex

subspace D of any hyperconvex space is a proximinal nonexpansive retract of the
superspace M if D is compact in any topology  relative to which the distance function
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

858 Kirk, Reich, and Veeramani

is lower semicontinuous. This means that if fx g and fy g are two nets in D that
-converge, respectively to x, y 2 D, then
dðx, yÞ  lim inf dðx , y Þ:

In particular, this observation applies when d is the norm and  is the weak topology
in a dual Banach space. Suppose ðY,
, Þ is a -finite measure space. The dual of
L1 ðY,
, Þ can be identified with L1 ðY,
, Þ: Since this latter space is a hyperconvex
Banach spacea we can apply Proposition 2.14 to obtain the following result.

Theorem 2.15. Let A and B be nonempty, weak -compact, and weakly externally
hyperconvex subsets of an L1 ðY,
, Þ space. Suppose T  : A ! 2B is such that:

(i) for each x 2 A, T  x is a nonempty admissible (more generally, externally

hyperconvex) subset of B;
(ii) T  : ðA, k  kÞ ! ð2B , dH Þ is nonexpansive;
(iii) T  ðA0 Þ  B0 :

Then there exists x0 2 A such that

distðx0 , T  ðx0 ÞÞ ¼ distðA, BÞ ¼ inffdistðx, T  ðxÞÞ : x 2 Ag:
Because of the Alaoglu theorem, the weak -compactness assumption in the preced-
ing theorem is redundant if the sets A and B are actually admissible. In this case
Theorem 2.15 reduces to a special case of Theorem 2.10.

3. HILBERT SPACES

In this section we apply a classical fixed point theorem of T.-C. Lim (1974) to
prove a best proximity pair theorem for Hilbert spaces. Given a nonempty closed
convex subset A of a Hilbert space X, PA will always denote the nearest point
projection of X onto A. We will use the well-known fact that PA is nonexpansive.
As before, dH denotes the Hausdorff metric.
We begin with two straightforward lemmas.

Lemma 3.1. Let A be a nonempty closed convex subset of a Hilbert space X. If C and D
are nonempty closed and bounded subsets of X, then
dH ðPA ðCÞ, PA ðDÞÞ  dH ðC, DÞ:

Proof. First observe that distðx, DÞ  dH ðC, DÞ for any x 2 C and that for any " > 0
there exists y" 2 D such that
kx  y" k  distðx, DÞ þ ":

a
A Banach space is hyperconvex if and only if it is isometrically isomorphic to a space CðKÞ of
continuous functions on a stonian space K (see, e.g., Schaefer (1974)).
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Proximinal Retracts and Best Proximity Pair Theorems 859

Thus

distðPA ðxÞ, PA ðDÞÞ  kPA ðxÞ  PA ðy" Þk

 kx  y" k
 distðx, DÞ þ "
 dH ðC, DÞ þ ",

and since " > 0 is arbitrary it follows that

distðPA ðxÞ, PA ðDÞÞ  dH ðC, DÞ:

Therefore

sup distðPA ðxÞ, PA ðDÞÞ  dH ðC, DÞ:

x2C

Similarly,

sup distðPA ðyÞ, PA ðCÞÞ  dH ðC, DÞ,

y2D

from which it follows that

dH ðPA ðCÞ, PA ðDÞÞ  dH ðC, DÞ: g

Lemma 3.2. Let X be a reflexive Banach space, let A be a nonempty bounded closed
convex subset of X, and let B be a nonempty closed convex subset of X: Then A0 and
B 0 (as in Definition 1.1) are nonempty and satisfy

PB ðA0 Þ  B0 and PA ðB0 Þ  A0 :

Proof. Since

distðA, BÞ ¼ inffkx  yk : x 2 A, y 2 Bg;

there exist sequences fxn g in A and fyn g in B such that kxn  yn k !distðA, BÞ: Since
A is weakly closed and bounded we may further suppose (passing to a subsequence
if necessary) that fxn g converges weakly, say to x0 2 A: Now

k yn k  kxn  yn k þ kxn k,

so fyn g is also bounded. By again passing to a subsequence we may suppose fyn g

converges weakly to yo 2 B: Therefore

kx0  y0 k  lim kxn  yn k ¼ distðA, BÞ,

proving that A0 and B0 are nonempty. If y 2 PB ðA0 Þ then y ¼ PB ðxÞ for some
x 2 A0 , so clearly, kx  yk ¼ distðA, BÞ: Hence PB ðA0 Þ  B0 : g
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

860 Kirk, Reich, and Veeramani

Theorem 3.3. Let X be a Hilbert space and let A, B be nonempty closed convex subsets
of X with A bounded. Let T : A ! 2B be a multi-valued mapping such that

(i) for each x 2 A, TðxÞ is a nonempty compact subset of B;

(ii) dH ðTðxÞ, TðyÞÞ  kx  yk, for all x, y 2 A;
(iii) TðA0 Þ  B0 :

Then there exists x0 2 A such that

distðx0 , Tðx0 ÞÞ ¼ distðA, BÞ ¼ inffdistðx, TðxÞÞ : x 2 Ag:

Proof. By Lemma 3.2, A0 6¼ ;: Let x 2 A0 and z 2 PA ðTðxÞÞ: Then z ¼ PA ð yÞ for

some y 2 TðxÞ  B0 : Now y 2 B0 implies there exists a 2 A0 such that ka  yk ¼
distðA, BÞ: Also
kz  yk ¼ distð y, AÞ  ka  yk ¼ distðA, BÞ,

so we conclude z 2 A0 : Hence PA ðTðxÞÞ  A0 : From this it follows that

PA  T : A0 ! 2A0 : Also, for x, y 2 A0 ,
dH ðPA ðTðxÞÞ, PA ðTðyÞÞÞ  dH ðTðxÞ, TðyÞÞ  kx  yk:

Moreover, A0 is a nonempty closed bounded convex subset of X, and

for each x 2 A0 , PA ðTðxÞÞ is a compact subset of A0 : Hence by Lim’s theorem
(Lim, 1974) there exists x0 2 A0 such that x0 2 PA ðTðx0 ÞÞ: This implies that
kx0  yk ¼ distðy, AÞ

for some y 2 Tðx0 Þ  B0 , and since y 2 B0 , distðy, AÞ ¼distðA, BÞ: Thus

distðx0 , Tðx0 ÞÞ  kx0  yk ¼ distðA, BÞ:

This implies that distðx0 , Tðx0 ÞÞ ¼distðA, BÞ which also implies

distðx0 ; Tðx0 ÞÞ ¼ inffdistðx, TðxÞÞg: g

Example 3.4. Let

A ¼ fð1, yÞ : 0  y  1g, B ¼ fð2, yÞ : 0  y  1g:

Define T : A ! B by setting Tð1, yÞ ¼ ð2, 1  yÞ: Then A0 ¼ A, and B0 ¼ B:

Moreover ð1, 1=2Þ 2 A satisfies
       
1 1 1 1
d 1; ; T 1; ¼ d 1; ; 2;
2 2 2 2
¼ distðA; BÞ

¼ inf x  T ðxÞ : x 2 A :

Thus in general PB 6¼ T on A0 , although PB ðxÞ ¼ TðxÞ and PA ðxÞ ¼ x whenever

ðx, TðxÞÞ is a best proximity pair in A
B:
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Proximinal Retracts and Best Proximity Pair Theorems 861

ACKNOWLEDGMENT

The work of the second author was partially supported by the Fund for the
Promotion of Research at the Technion and by the Technion VPR Fund.

REFERENCES

Baillon, J.-B. (1988). Nonexpansive mappings and hyperconvex spaces. Contemp.

Math. 72:11–19.
Espı́nola, R., Khamsi, M. A. (2001). Introduction to hyperconvex spaces. In: Kirk,
W. A., Sims, B., eds. Handbook of Metric Fixed Point Theory. Dordrecht:
Kluwer Academic Publishers, pp. 391–435.
Espı́nola, R., Kirk, W. A., López, G. (2000). Nonexpansive retractions in hyper-
convex spaces. J. Math. Anal. Appl. 251:557–570.
Khamsi, M. A. (1996). KKM and Ky Fan theorems in hyperconvex metric spaces.
J. Math. Anal. Appl. 204:298–306.
Khamsi, M. A., Kirk, W. A., Martı́nez Yañez, C. (2000). Fixed point and selection
theorems in hyperconvex spaces. Proc. Amer. Math. Soc. 128:3275–3283.
Kirk, W. A., Shin, S. S. (1997). Fixed point theorems in hyperconvex spaces.
Houston J. Math. 23:175–187.
Ky Fan. (1969). Extensions of two fixed point theorems of F. E. Browder. Math.
Zeit. 112:234–240.
Lim, T.-C. (1974). A fixed point theorem for multivalued nonexpansive
mappings in a uniformly convex Banach space. Bull. Amer. Math. Soc.
80:1123–1126.
Prolla, J. B. (1982–1983). Fixed point theorems for set valued mappings and
existence of best approximations. Numer. Funct. Anal. Optim. 5:449–455.
Reich, S. (1978). Approximate selections, best approximations, fixed points, and
invariant sets. J. Math. Anal. Appl. 62:104–113.
Sine, R. (1989a). Hyperconvexity and approximate fixed points. Nonlinear Anal.
13:863–869.
Sine, R. (1989b). Hyperconvexity and nonexpansive multifunctions. Trans. Amer.
Math. Soc. 315:755–767.
Sadiq Basha, S., Veeramani, P. (1997). Best approximations and best proximity
pairs. Acta Sci. Math. (Szeged) 63:289–300.
Sadiq Basha, S., Veeramani, P. (2000). Best proximity pair theorems for multifunc-
tions with open fibres. J. Approximation Theory 103:119–129.
Sadiq Basha, S., Veeramani, P., Pai, D. V. (2001). Best proximity pair theorems.
Indian J. Pure Appl. Math. 32:1237–1246.
Schaefer, H. H. (1974). Banach Lattices and Positive Operators. Berlin and
New York: Springer-Verlag.
Sehgal, V. M., Singh, S. P. (1988). A generalization to multifunctions of Fan’s best
approximation theorem. Proc. Amer. Math. Soc. 102:534–537.
Sehgal. V. M., Singh, S. P. (1989). A theorem on best approximations. Numer. Funct.
Anal. Optim. 10:181–184.
 MARCEL DEKKER, INC. • 270 MADISON AVENUE • NEW YORK, NY 10016
©2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

862 Kirk, Reich, and Veeramani

Srinivasan, P. S. (2001). Best proximity pair theorems. Acta Sci. Math. (Szeged)
67:421–429.
Vetrivel, V., Veeramani, P., Bhattacharyya, P. (1992). Some extensions of Fan’s
best approximation theorem. Numer. Funct. Anal. Optim. 13:397–402.