J. Math. Anal. Appl.

337 (2008) 123–129

www.elsevier.com/locate/jmaa

The characteristic of convexity of a Banach

space and normal structure
Satit Saejung
Department of Mathematics, Khon Kaen University, Khon Kaen 40002, Thailand
Received 16 January 2007
Available online 3 April 2007
Submitted by B. Sims

Abstract
We present some sufficient conditions for normal structure of Banach spaces and their dual spaces in
terms of the characteristic of convexity, the James constant, and the coefficient of weak orthogonality.
Many known results are improved and strengthened. We also show that some of our results are sharp.
© 2007 Elsevier Inc. All rights reserved.

Keywords: Uniform normal structure; Ultrapower; Characteristic of convexity; James constant; Coefficient of weak
orthogonality

1. Introduction

A Banach space X is said to have (weak) normal structure (see [2]) if for every (weakly
compact) closed bounded convex subset K in X that contains more than one point, there exists a
point x0 ∈ K such that
   
sup x0 − y: y ∈ K < sup x − y: x, y ∈ K .

In reflexive spaces, normal structure and weak normal structure are the same. It is well known
(see [8]) that if X fails to have weak normal structure, then there exist a weakly compact convex
subset C ⊂ X and a sequence (xn ) ⊂ C such that dist(xn+1 , co{xk }nk=1 ) → diam C = 1. A Ba-
nach space X is said to have uniform normal structure if there exists 0 < c < 1 such that for any

E-mail address: [email protected].

0022-247X/$ – see front matter © 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jmaa.2007.03.076
124 S. Saejung / J. Math. Anal. Appl. 337 (2008) 123–129

closed bounded convex subset K of X that contains more than one point, there exists x0 ∈ K
such that
   
sup x0 − y: y ∈ K < c sup x − y: x, y ∈ K .
It was proved by W.A. Kirk that every reflexive Banach space with normal structure has the fixed
point property (see [14]).
In this paper, we investigate some sufficient conditions for (uniform) normal structure of Ba-
nach spaces and their dual spaces in terms of many parameters namely the characteristic of
convexity, the James constant, and the coefficient of weak orthogonality. We also show that some
results are sharp.
First we recall some basic facts about ultrapowers. Let F be a filter on N and let X be a
Banach space. A sequence {xn } in X converges to x with respect to F , denoted by limF xi = x,
if for each neighborhood U of x, {i ∈ N: xi ∈ U } ∈ F . A filter U on N is called an ultrafilter
if it is maximal with respect to set inclusion. An ultrafilter is called trivial if it is of the form
{A ⊂ N: i0 ∈ A} for some fixed i0 ∈ N, otherwise, it is called nontrivial. Let l∞ (X) denote the
subspace of the product space n∈N X equipped with the norm
 
(xn ) := sup xn  < ∞.
n∈N
Let U be an
 ultrafilter on N and let 
NU = (xn ) ∈ l∞ (X): lim xn  = 0 .
U
The ultrapower of X, denoted by X, is the quotient space l∞ (X)/NU equipped with the quotient
norm. Write (xn )U to denote the elements of the ultrapower. It follows from the definition of the
quotient norm that
 
(xn )U  = lim xn .
U
Note that if U is nontrivial, then X can be embedded into X  isometrically. For more details
∗  ∗ , then X has uniform normal
see [18]. We also note that if X is super-reflexive, that is X = (X)

structure if and only if X has normal structure (see [12]).
In what follows, we let BX , SX and X ∗ stand for the closed unit ball, the unit sphere, and the
dual space of a Banach space X, respectively.

Lemma 1. (See [17, Lemma 2].) If X is a super-reflexive Banach space and fails to have normal
 and f˜1 , f˜2 , f˜3 ∈ S(X)
structure, then there are x1 , x2 , x3 ∈ SX  ∗ such that

(a) xi − xj  = 1 and f˜i (xj ) = 0 for all i = j ,


(b) ˜ xi ) = 1 for i = 1, 2, 3, and
fi (
(c) 
x3 − ( x2 + x1 )  
x2 + x1 .

2. Results

2.1. The characteristic of convexity and the James constant

The modulus of convexity [4] of a Banach space X is the function δX : [0, 2] → [0, 1] defined
by
1
δX (ε) = inf 1 − x + y: x, y ∈ SX , x − y = ε .
2
 S. Saejung / J. Math. Anal. Appl. 337 (2008) 123–129 125

The function δ is continuous on [0, 2), nondecreasing on [0, ε0 (X)] and strictly increasing on
[ε0 (X), 2]. Here ε0 (X) = sup{ε: δX (ε) = 0} is the characteristic of convexity of X. We say that
X is uniformly convex provided ε0 (X) = 0. Also, X is uniformly nonsquare provided ε0 (X) < 2.
The James constant of X is defined as
   
J (X) = sup min x + y, x − y : x, y ∈ SX .
It follows easily that X is uniformly nonsquare if and only if J (X) < 2. Moreover, it is proved
in [7] that J (X) = sup{ε: ε  2(1 − δX (ε))}.
The search for an upper bound of the James constant for which (uniform) normal structure
of a Banach space is guaranteed√
is widely studied. The best known result is that X has uniform
normal structure if J (X) < 1+2 5 [6]. It is also well known that ε0 (X) < 1 implies both X and X ∗
have uniform normal structure (see also [8, Theorems 6.1 and 7.2]). We now give a new sufficient
condition in terms of both constants.

Theorem 2. Let X be a Banach space.

(1) If J (X)ε0 (X ∗ ) < 2, then X has uniform normal structure.

(2) If J (X ∗ )ε0 (X) < 2, then X ∗ has uniform normal structure.
√
Proof. To prove (1), we assume that J (X)ε0 (X ∗ ) < 2. Since J (X)  2, the condition
J (X)ε0 (X ∗ ) < 2 implies X is super-reflexive. Suppose next that X fails weak normal structure.
 and f˜1 , f˜2 , f˜3 ∈ S(X)
Then there are elements x1 , x2 , x3 ∈ SX  ∗ satisfying conditions in Lemma 1.
˜ ˜ ˜ ˜
Hence f2 − f1   (f2 − f1 )( ˜ ˜
x2 − x1 ) = 2, so f2 − f1  = 2. Since the characteristics of con-
 ∗ are the same,
vexity of X ∗ and of (X)
   
f˜2 + f˜1  = f˜2 − −f˜1   ε0 X ∗ ,

and similarly, f˜3 + f˜1   ε0 (X ∗ ). Now, we obtain a contradiction since

 
J (X)  min  x3 − x2 + x1 , 
x3 − x2 − x1 
 
 min  x3 − x2 + x1 , 
x2 + x1 
 ˜   ˜ 
f3 + f˜1 f2 + f˜1
 min x3 − x2 + x1 ),
( x2 + x1 )
(
ε0 (X ∗ ) ε0 (X ∗ )
2
= .
ε0 (X ∗ )
We have proved that X has weak normal structure if J (X)ε0 (X ∗ ) < 2. The conclusion that X
has uniform normal structure follows since the condition J (X)ε0 (X ∗ ) < 2 is invariant under the
taking of ultrapowers.
(2) By (1), X ∗ has uniform normal structure whenever J (X ∗ )ε0 (X ∗∗ ) < 2 which is equivalent
to J (X ∗ )ε0 (X) < 2. 2

Corollary 3. If ε0 (X) = 1, then X ∗ has uniform normal structure.

Proof. It is clear that the condition ε0 (X) = 1 implies X is uniformly nonsquare, and so is X ∗ .
Consequently, ε0 (X)J (X ∗ ) = J (X ∗ ) < 2 and the assertion follows. 2
126 S. Saejung / J. Math. Anal. Appl. 337 (2008) 123–129

Remark 4. Corollary 3 gives new information. In particular, if ε0 (X) = 1, then X ∗ has the
fixed point property for mappings of asymptotically nonexpansive type [15] (cf. [13]). The upper
bound 1 in Corollary 3 is best possible, that is, for any 0 < η < 1 there is a Banach space X
such that X ∗ fails to have uniform normal structure and ε0 (X) = 1 + η. In fact, we consider the
Bynum space p,1 (see [3]) which is the p space equipped with the norm
xp,1 = x + p + x − p
where x + and x − are positive and negative parts of x ∈ p , i.e., (x + )n = max{xn , 0} and (x − )n =
max{−xn , 0}. It is proved that ε0 (p,1 ) = 21/p and q,∞ = ∗p,1 , where p1 + q1 = 1, does not have
even normal structure.

Since 12 ε0 (X ∗ ) = ρX (0) := limt→0 ρX (t)

t , where
x + ty + x − ty
ρX (t) = sup − 1: x, y ∈ SX ,
2
Corollary 3 also strengthens Theorem 2.1 of Prus [16] and Proposition 3 of Turett [21]. More
precisely, if ρX (0) = 12 or ε0 (X ∗ ) = 1, then X has uniform normal structure.

2.2. The coefficient of weak orthogonality

The WORTH property was introduced by B. Sims in [19] as follows: a Banach space X has
the WORTH property if
 
lim xn + x − xn − x = 0
n→∞

for all x ∈ X and all weakly null sequences (xn ). In [20], the author defined the coefficient of
weak orthogonality, which measures the “degree of WORTH-whileness.” As in [9], we prefer to
use its reciprocal, μ(X), which is defined as the infimum of the set of real numbers r > 0 such
that
lim sup xn + x  r lim sup xn − x
n→∞ n→∞

for all x ∈ X and for all weakly null sequence (xn ) in X. It is known that X has the WORTH
property if and only if μ(X) = 1. We also note here that μ(X) = μ(X ∗ ) if X is reflexive [10,
Theorem 3].
Before going to the result, we start with the following lemma which is a refinement of [7,
Lemma 2.3].

Lemma 5. Let X be a Banach space for which BX∗ is w*-sequentially compact (for example,
X is reflexive or separable, or has an equivalent smooth norm). Suppose that X fails to have
weak normal structure. Then, for any ε > 0, there exist z1 , z2 ∈ SX and g1 , g2 ∈ SX∗ such that
the following conditions are satisfied:

(a) |z1 − z2  − 1| < ε and |gi (zj )| < ε for all i = j ,

(b) gi (zi ) = 1 for i = 1, 2, and
(c) z2 + z1   μ(X) + ε.

Proof. By the assumptions, there exist sequences (xn ) ⊂ X and (fn ) ⊂ SX∗ such that
 S. Saejung / J. Math. Anal. Appl. 337 (2008) 123–129 127

w
(1) xn → 0,
(2) diam{xn }∞ ∞
n=1 = 1 = limn→∞ xn − x for all x ∈ co{xn }n=1 ,
(3) fn (xn ) = xn  for all n ∈ N, and
w∗
(4) fn → f for some f ∈ BX∗ .

Observe that 0 is in the weakly closed convex hull of {xn }∞ n=1 which equals the norm closed
convex hull co{xn }∞n=1 . This implies that limn→∞ x n  = 1.
Let ε ∈ (0, 1) be given. Pick η = 3ε . We first choose a natural number n1 so that
 
f (xn ) < η and 1 − η  xn1   1.
1
2
By the definition of μ(X), we have
lim sup xn + xn1   μ(X) lim xn − xn1  = μ(X).
n→∞ n→∞

Next, we choose n2 > n1 so that

xn2 + xn1   μ(X) + η,

1 − η  xn2   1, 1 − η  xn2 − xn1   1,
   
fn (xn ) < η, and (fn − f )(xn ) < η .
1 2 2 1
2
This implies that
     
fn (xn )  (fn − f )(xn ) + f (xn ) < η.
2 1 2 1 1

Set
xn1 xn2
z1 := , z2 := , g1 := fn1 , g2 := fn2 .
xn1  xn2 
We now prove that (a), (b) and (c) are satisfied. Clearly, (b) holds. Moreover, for i = j ,
  |f (x )|
gi (zj ) = ni nj < η < 2η < ε.
xnj  1−η
Next, we observe that
 
 xn1 xn2 
z1 − z2  =  −
 x  x  

n1 n2
   
 xn1   xn2 
  
− xn1  + xn1 − xn2  + xn2 − 
xn1  xn2  
   
= 1 − xn1  + xn1 − xn2  + 1 − xn2 
< 1 + 2η < 1 + ε,
and
z1 − z2   g1 (z1 − z2 ) = g1 (z1 ) − g1 (z2 )  1 − η > 1 − ε,
that is (a) is satisfied. Moreover, (c) is satisfied, since
128 S. Saejung / J. Math. Anal. Appl. 337 (2008) 123–129

 
 xn1 xn2 

z1 + z2  =  + 
xn1  xn2  
   
 xn1   xn2 
  
− xn1  + xn1 + xn2  + xn2 − 
xn1  xn2  
   
= 1 − xn1  + xn1 + xn2  + 1 − xn2 
< 1 + 3η = 1 + ε.
This completes the proof. 2

Lemma 5 can be rewritten in ultrapower language as:

Lemma 6. If a super-reflexive Banach space X fails normal structure, then there are x1 , x2 ∈ SX

and f˜1 , f˜2 ∈ S(X)
 ∗ such that

(a) x1 − x2  = 1 and f˜i (xj ) = 0 for all i = j ,

(b) f˜i (
xi ) = 1 for i = 1, 2,
(c) x1 + x2   μ(X).

We now give sufficient conditions for normal structure of a space and its dual in terms of the
characteristic of convexity and the coefficient of weak orthogonality.

Theorem 7.

(1) If ε0 (X ∗ )μ(X) < 2, then X has normal structure.

(2) If ε0 (X)μ(X) < 2, then X ∗ has normal structure.

Proof. (1) Since μ(X)  1 always, the condition ε0 (X ∗ )μ(X) < 2 implies that X is super-
 and f˜1 , f˜2 ∈
reflexive. If X does not have normal structure, then there are vectors x1 , x2 ∈ SX
˜ ˜
 ∗ satisfying properties in Lemma 6. Clearly, f2 − f1  = 2. Hence
S(X)
 
  x2 + x1 2 2
ε0 (X ∗ )  f˜2 + f˜1   f˜2 + f˜1 =  ,

x2 + x1  
x2 + x1  μ(X)
or ε0 (X ∗ )μ(X)  2 which is a contradiction.
(2) We assume that ε0 (X)μ(X) < 2. This implies that X is super-reflexive. In particular,
μ(X ∗ ) = μ(X) and ε0 (X ∗∗ ) = ε0 (X). It then follows from ε0 (X ∗∗ )μ(X ∗ ) = ε0 (X)μ(X) < 2
and (1) that X ∗ has normal structure. 2

Remark 8. (1) Recently, it is proved in [10, Theorems 1 and 2] that if CNJ (X) < 1 + 1
(μ(X))2
then X has normal structure. Here CNJ (X) = sup{ x+y +x−y
2 2
or J (X) < 1 + 1
μ(X) , 2(x2 +y2 )
: x, y ∈
∗ 2
X and (x, y) = (0, 0)}. Since CNJ (X) = CNJ (X ∗ )  1 + (ε0 (X4 )) , the class of spaces with
CNJ (X) < 1 + (μ(X)) 1 ∗
2 is included in the wider class of spaces satisfying ε0 (X )μ(X) < 2. Fur-
thermore, let Xp be the p-direct sum 2,1 ⊕p 2,1 where 1 < p  2. It follows that ε0 (Xp∗ ) = 1
√
(see [5, Theorem 9]), CNJ (Xp )  22/p−1 , J (Xp )  21/p and μ(Xp ) = 2 (see [11, Theorem 7]).
If p is sufficiently closed to 1, then neither CNJ (Xp ) < 1+ (μ(X1 ))2 nor J (Xp ) < 1+ μ(X
1
p)
holds.
p
 S. Saejung / J. Math. Anal. Appl. 337 (2008) 123–129 129

However, the condition ε0 ((Xp )∗ )μ(Xp ) < 2 still holds true. We also note here that sometimes
calculating the characteristic of convexity of the dual space is much easier than calculating the
James or von Neumann–Jordan constants of the space (see [1, Example IV.7] and [10, Theo-
rem 4]). √
(2) It is easy to see that ε0 (2,1 ) = μ(2,∞ ) = 2 and 2,∞ fails normal structure. Hence we
conclude from Theorem 7 that our results are sharp.

Acknowledgment

The author would like to acknowledge suggestions by the referee(s) which led to improvements in the manuscript.

References

[1] A. Ayerbe-Toledano, T. Domínguez-Benavides, G. López-Acedo, Measures of noncompactness in metric fixed point

theory, in: Oper. Theory Adv. Appl., vol. 99, Birkhäuser, Basel, 1997.
[2] M.S. Brodskiı̆, D.P. Mil’man, On the center of convex sets, Dokl. Akad. Nauk SSSR (N.S.) 59 (1948) 837–840 (in
Russian).
[3] W.L. Bynum, A class of spaces lacking normal structure, Compos. Math. 25 (1972) 233–236.
[4] J.A. Clarkson, Uniformly convex spaces, Trans. Amer. Math. Soc. 40 (1936) 396–414.
[5] D.J. Downing, B. Turett, Some properties of the characteristic of convexity relating to fixed point theory, Pacific J.
Math. 104 (1983) 343–350.
[6] S. Dhompongsa, A. Kaewkhao, S. Tesana, On a generalized James constant, J. Math. Anal. Appl. 285 (2003) 419–
435.
[7] J. Gao, K.-S. Lau, On two classes of Banach spaces with uniform normal structure, Studia Math. 99 (1) (1991)
41–56.
[8] K. Goebel, W.A. Kirk, Topics in Metric Fixed Point Theory, Cambridge Studies in Advanced Mathematics, Cam-
bridge University Press, Cambridge, 1990.
[9] A. Jiménez-Melado, E. Llorens-Fuster, The fixed point property for some uniformly nonsquare Banach spaces, Boll.
Un. Mat. Ital. (7) 10-A (1996) 587–595.
[10] A. Jiménez-Melado, E. Llorens-Fuster, S. Saejung, The von Neumann–Jordan constant, weak orthogonality and
normal structure in Banach spaces, Proc. Amer. Math. Soc. 134 (2006) 355–364.
[11] M. Kato, T. Tamura, Weakly nearly uniform smoothness and WORTH property of ψ -direct sums of Banach spaces,
Comment. Math. Prace Mat. 46 (2006) 113–129.
[12] M.A. Khamsi, Uniform smoothness implies super-normal structure property, Nonlinear Anal. 19 (1992) 1063–1069.
[13] T.W. Kim, W.A. Kirk, Fixed point theorems for Lipschitzian mappings in Banach spaces, Nonlinear Anal. Forum 1
(1995) 53–62.
[14] W.A. Kirk, A fixed point theorem for mappings which do not increase distances, Amer. Math. Monthly 72 (1965)
1004–1006.
[15] G. Li, B. Sims, Fixed point theorems for mappings of asymptotically nonexpansive type, Nonlinear Anal. 50 (2002)
1085–1091.
[16] S. Prus, Some estimates for the normal structure coefficient in Banach spaces, Rend. Circ. Mat. Palermo 40 (1991)
128–135.
[17] S. Saejung, Sufficient conditions for uniform normal structure of Banach spaces and their duals, J. Math. Anal.
Appl. 330 (2007) 597–604.
[18] B. Sims, Ultra-techniques in Banach space theory, in: Queen’s Papers in Pure and Applied Mathematics, vol. 60,
Queen’s University, Kingston, 1982.
[19] B. Sims, Orthogonality and fixed points of nonexpansive maps, in: Proc. Centre Math. Anal. Austral. Nat. Univ.,
Canberra, 1988, pp. 178–186.
[20] B. Sims, A class of spaces with weak normal structure, Bull. Austral. Math. Soc. 50 (1994) 523–528.
[21] B. Turett, A dual view of a theorem of Baillon, in: Nonlinear Analysis and Applications (St. Johns, Newfoundland,
1981), in: Lect. Notes Pure Appl. Math., vol. 80, Dekker, New York, 1982, pp. 279–286.