Indag. Mathem., N.S.

, 20 (3), 329379 September, 2009

Structure of Cesro function spaces

by Sergei V. Astashkin
a
and Lech Maligranda
b
a
Department of Mathematics and Mechanics, Samara State University, Acad. Pavlova 1,
443011 Samara, Russia
b
Department of Mathematics, Lule University of Technology, SE-971 87 Lule, Sweden
Communicated by Prof. H.W. Broer
ABSTRACT
The structure of the Cesro function spaces Ces
p
on both [0, 1] and [0, ) for 1 <p is investigated.
We nd their dual spaces, which equivalent norms have different description on [0, 1] and [0, ). The
spaces Ces
p
for 1 < p < are not reexive but strictly convex. They are not isomorphic to any L
q
space with 1 q . They have near zero complemented subspaces isomorphic to l
p
and in the
middle contain an asymptotically isometric copy of l
1
and also a copy of L
1
[0, 1]. They do not have
DunfordPettis property but they do have the weak BanachSaks property. Cesro function spaces on
[0, 1] and [0, ) are isomorphic for 1 < p . Moreover, we give characterizations in terms of p and
q when Ces
p
[0, 1] contains an isomorphic copy of l
q
.
MSC: 46E30, 46B20, 46B42
Key words and phrases: Cesro sequence spaces, Cesro function spaces, Kthe dual, Associated space,
Dual space, L
p
spaces, Copies of l
p
, Weak BanachSaks property, DunfordPettis property, Rademacher
functions, Type, Cotype, Isomorphism, Subspaces, Complemented subspaces

S.V. Astashkin was partially supported by the Ministry of Education and Science of the Russian
Federation grant 3341. L. Maligranda was partially supported by the Swedish Research Council (VR)
grant 621-2008-5058.
E-mails: astashkn@ssu.samara.ru (S.V. Astashkin), lech@sm.luth.se (L. Maligranda).
329
1. INTRODUCTION AND PRELIMINARIES
Let 1 p . The Cesro sequence space ces
p
is dened as the set of all real
sequences x ={x
k
} such that
x
c(p)
=
_

n=1
_
1
n
n

k=1
|x
k
|
_
p
_
1/p
< when 1 p <
and
x
c()
=sup
nN
1
n
n

k=1
|x
k
| < when p =.
The Cesro function spaces Ces
p
= Ces
p
(I) are the classes of Lebesgue measur-
able real functions f on I =[0, 1] or I =[0, ) such that
f
C(p)
=
_
_
I
_
1
x
x
_
0
|f (t )| dt
_
p
dx
_
1/p
< for 1 p <
and
f
C()
= sup
xI,x>0
1
x
x
_
0
|f (t )| dt < for p =.
The Cesro sequence spaces ces
p
and ces

appeared in 1968 in connection

with the problem of the Dutch Mathematical Society to nd their duals. Some
investigations of ces
p
were done by Shiue [50] in 1970. Then Leibowitz [36] and
Jagers [26] proved that ces
1
= {0}, ces
p
are separable reexive Banach spaces for
1 < p < and the l
p
spaces are continuously and strictly embedded into ces
p
for
1 < p . More precisely, x
c(p)
p

x
p
for all x l
p
with p

=
p
p1
when
1 <p <and p

=1 when p =. Moreover, if 1 <p <q , then ces

p
ces
q
with continuous strict embedding. Bennett [8] proved that ces
p
for 1 < p < are
not isomorphic to any l
q
space with 1 q (see also [45] for another proof).
Several geometric properties of the Cesro sequence spaces ces
p
were studied in
the last years by many mathematicians (see e.g. [1016,34]). Some more results on
ces
p
can be found in two books [8,39].
In 19992000 it was proved by Cui and Hudzik [11], Cui, Hudzik and Li [14] and
Cui, Meng and Puciennik [16] that the Cesro sequence spaces ces
p
for 1 <p <
have the xed point property (cf. also [10, Part 9]). Maligranda, Petrot and Suantai
[45] proved that the Cesro sequence spaces ces
p
for 1 <p < are not uniformly
non-square, that is, there are sequences {x
n
} and {y
n
} on the unit sphere such that
lim
n
min(x
n
+y
n

c(p)
, x
n
y
n

c(p)
) =2. They even proved that these spaces
are not B-convex.
The Cesro function spaces Ces
p
[0, ) for 1 p were considered by Shiue
[51], Hassard and Hussein [25] and Sy, Zhang and Lee [54]. The space Ces

[0, 1]
330
appeared already in 1948 and it is known as the Korenblyum, Krein and Levin space
K (see [31] and [59]).
Recently, we proved in the paper [4] that, in contrast to Cesro sequence spaces,
the Cesro function spaces Ces
p
(I) on both I =[0, 1] and I =[0, ) for 1 < p <
are not reexive and they do not have the xed point property. In other paper [5]
we investigated Rademacher sums in Ces
p
[0, 1] for 1 p . The description is
different for 1 p < and p =.
We recall some notions and denitions which we will need later on. By L
0
=
L
0
(I) we denote the set of all equivalence classes of real-valued Lebesgue mea-
surable functions dened on I =[0, 1] or I =[0, ). A normed function lattice or
normed ideal space X = (X, ) (on I ) is understood to be a normed space in
L
0
(I), which satises the so-called ideal property: if |f | |g| a.e. on I and g X,
then f X and f g. If, in addition, X is a complete space, then we say that
X is a Banach function lattice or a Banach ideal space (on I ). Sometimes we write

X
to be sure in which space the norm is taken.
For two normed ideal spaces X and Y on I the symbol X Y means that X Y
and the imbedding is continuous, and the symbol X
C
Y means that X Y with
the inequality x
Y
Cx
X
for all x X. Moreover, notation X Y means that
these two spaces are isomorphic.
For a normed ideal space X =(X, ) on I and 1 p <the p-convexication
X
(p)
of X is the space of all f L
0
(I) such that |f |
p
X with the norm
f
X
(p) :=
_
_
|f |
p
_
_
1/p
X
.
X
(p)
is also a normed ideal space on I .
For a normed ideal space X =(X, ) on I the Kthe dual (or associated space)
X

is the space of all f L

0
(I) such that the associate norm
f

:= sup
gX,g
X
1
_
I
|f (x)g(x)| dx
is nite. The Kthe dual X

=(X

) is a Banach ideal space. Moreover, X X

with f f

for all f X and we have equality X =X

with f =f

if
and only if the norm in X has the Fatou property, that is, if 0 f
n
f a.e. on I
and sup
nN
f
n
<, then f X and f
n
f .
For a normed ideal space X =(X, ) on I with the Kthe dual X

we have the
following Hlder type inequality: if f X and g X

, then fg is integrable and

_
I
|f (x)g(x)| dx f
X
g
X
.
A function f in a normed ideal space X on I is said to have absolutely continuous
norm in X if, for any decreasing sequence of Lebesgue measurable sets A
n
I with
331
empty intersection, we have that f
A
n
0 as n . The set of all functions in
X with absolutely continuous norm is denoted by X
a
. If X
a
=X, then the space X
itself is said to have absolutely continuous norm. For a normed ideal space X with
absolutely continuous norm, the Kthe dual X

and the dual space X

coincide.
Moreover, a Banach ideal space X is reexive if and only if both X and its associate
space X

have absolutely continuous norms.

For general properties of normed and Banach ideal spaces we refer to the books
Krein, Petunin and Semenov [32], Kantorovich and Akilov [28], Bennett and
Sharpley [9], Lindenstrauss and Tzafriri [38] and Maligranda [43].
The paper is organized as follows: In Section 1 some necessary denitions and
notation are collected. In Section 2 some simple results on Cesro function spaces
are presented. In particular, we can see that the Cesro function spaces Ces
p
(I) are
not reexive but strictly convex for all 1 <p <.
Sections 3 and 4 contain results on the dual and Kthe dual of Cesro function
spaces. There is a big difference between the cases on [0, ) and on [0, 1], as
we can see from Theorems 2 and 3. This was also the reason why we put these
investigations into two parts. Important in our investigations were earlier results on
the Kthe dual (ces
p
)

and remark on the Kthe dual (Ces

p
[0, ))

due to Bennett
[8]. This remark was recently proved, even for more general spaces, by Kerman,
Milman and Sinnamon [30]. Luxemburg and Zaanen [42] gave a description of the
Kthe dual (Ces

[0, 1])

.
Section 5 deals with the p-concavity and cotype of Cesro sequence spaces ces
p
and Cesro function spaces Ces
p
(I). It is shown, in Theorem 4, that they are p-
concave for 1 <p < with constant one and, thus, they have cotype max(p, 2).
In Section 6 it is proved, in Theorem 6, that the Cesro function spaces Ces
p
(I)
contain an order isomorphic and complemented copy of l
p
. Therefore, they do
not have the DunfordPettis property. This result and cotype property imply that
Ces
p
(I) are not isomorphic to any L
q
(I) space for 1 q (Theorem 7).
The authors proved in [4] that in the middle Cesro function spaces Ces
p
(I)
contain an asymptotically isometric copy of l
1
and consequently they are not
reexive and do not have the xed point property. This is a big difference with
Cesro sequence spaces ces
p
, which for 1 < p < are reexive and which have
the xed point property.
Section 7 contains the proof that the Cesro function spaces Ces
p
[0, 1] for 1
p < have the weak BanachSaks property. Important role in the proof will be
played by the description of the dual space given in Section 4.
In Section 8 we present a construction showing that the Cesro function spaces
Ces
p
[0, ) and Ces
p
[0, 1] for 1 < p are isomorphic. The isomorpisms are
different in the cases 1 <p < and p =.
In Section 9 it is proved that Ces
p
[0, 1] contains an isomorphic copy of l
q
if and
only if q [1, 2] for the case 1 p 2 and in the case when p >2 this can happen
when either q [1, 2] or q = p. This result is, in fact, different from the one for
L
p
[0, 1] space.
332
2. PRELIMINARY PROPERTIES OF CESRO FUNCTION SPACES
The Cesro function spaces Ces
p
[0, ) for 1 p were considered by Shiue
[51], Hassard and Hussein [25] and Sy, Zhang and Lee [54]. The space Ces

[0, 1]
appeared in 1948 and it is known as the Korenblyum, Krein and Levin space K (see
[31] and [59, p. 26 and 61]).
We collect some known or clear properties of Ces
p
(I) for both I = [0, 1] and
I =[0, ) in one place.
Theorem 1.
(a) If 1 < p , then Ces
p
(I) are Banach spaces, Ces
1
[0, 1] = L
1
w
with the
weight w(t ) =ln
1
t
, t (0, 1] and Ces
1
[0, ) ={0}.
(b) The spaces Ces
p
(I) are separable for 1 < p < and Ces

(I) is non-
separable.
(c) If 1 < p , then L
p
(I)
p

Ces
p
(I), where p

=
p
p1
and the embedding is
strict.
(d) If 1 < p < , then Ces
p
[0, 1]
|[0,a]
L
1
[0, a] for any a (0, 1) but not for
a = 1 and Ces
p
[0, )
|[0,a]
L
1
[0, a] for any 0 < a < but not for a = ,
that is, Ces
p
[0, ) L
1
[0, ). Moreover, Ces

[0, 1]
1
L
1
[0, 1].
(e) If 1 <p <q , then Ces
q
[0, 1]
1
Ces
p
[0, 1] and the embedding is strict.
(f) The spaces Ces
p
(I) are not rearrangement invariant.
(g) The spaces Ces
p
(I) are not reexive.
(h) The spaces Ces
p
(I) for 1 < p < are strictly convex, that is, if f
C(p)
=
g
C(p)
=1 and f =g, then
f +g
2

C(p)
<1.
Proof. (a), (b) Shiue [51] and Hassard and Hussein [25] proved that Ces
p
(I) are
separable Banach spaces for 1 < p < and non-separable ones for p = . We
only show here that Ces
1
[0, 1] is a weighted L
1
w
[0, 1] space with the weight w(t ) =
ln
1
t
for 0 <t 1 and Ces
1
[0, ) ={0}. In fact,
1
_
0
_
1
x
x
_
0
|f (t )| dt
_
dx =
1
_
0
_
1
_
t
1
x
dx
_
|f (t )| dt =
1
_
0
|f (t )| ln
1
t
dt. (1)
Moreover, if f L
0
[0, ) and |f (x)| >0 for x A with 0 <m(A) <, then there
exists sufciently large a >0 such that =
_
a
0
|f (t )| dt >0. Therefore, for b >a, it
yields that
b
_
0
_
1
x
x
_
0
|f (t )| dt
_
dx
b
_
a
_
1
x
x
_
0
|f (t )| dt
_
dx

b
_
a
_
1
x
a
_
0
|f (t )| dt
_
dx
333
= ln
b
a
as b .
Thus f / Ces
1
[0, ).
(c) Considering the Hardy operator Hf (x) =
1
x
_
x
0
f (t ) dt and using the Hardy
inequality (cf. [24, Theorem 327] and [33, Theorem 2]) we obtain that
f
C(p)
=H(|f |)
p
p

f
p
for all f L
p
(I), which means that the L
p
(I)
p

Ces
p
(I) for 1 <p .
The embeddings are strict. For example, f =

n=1
1
n
1/p

[n
2
1,n
2
)
Ces
p
(I) \
L
p
(I) for I =[0, ) and 1 <p <.
(d) If 0 <a <1 and suppf [0, a], then
f
C(p)

_
1
_
a
_
1
x
x
_
0
|f (t )| dt
_
p
dx
_
1/p

_
1
_
a
_
1
x
a
_
0
|f (t )| dt
_
p
dx
_
1/p
=
a
_
0
|f (t )| dt
_
1 a
1p
p 1
_
1/p
.
For a =1 this is not the case. In fact, consider function f (x) =
1
1x
for x [0, 1).
Then
1
x
_
x
0
f (t ) dt =
1
x
ln
1
1x
and
f
p
C(p)
=
1
_
0
_
1
x
ln
1
1 x
_
p
dx =

_
1
_
t lnt
t 1
_
p
dt
t
2
c +

_
2
(2lnt )
p
t
2
dt <
and, hence, f Ces
p
[0, 1] for any 1 p < but clearly, f / L
1
[0, 1].
In the case of Ces
p
[0, ) we will have for 0 < a < with suppf [0, a] and
p (1, ),
f
C(p)

_

_
a
_
1
x
x
_
0
|f (t )| dt
_
p
dx
_
1/p

_

_
a
_
1
x
a
_
0
|f (t )| dt
_
p
dx
_
1/p
=
a
_
0
|f (t )| dt
1
(p 1)a
11/p
.
334
For the function f (x) =
1
x

[1,)
(x), x (0, ) we have
1
x
_
x
0
f (t ) dt =
1
x
lnx (x
1) and
f
p
C(p)
=

_
1
_
lnx
x
_
p
dx <.
Thus, f Ces
p
[0, ) for any 1 <p <, but clearly f / L
1
[0, ).
(e) If 1 <p <q , then L
q
[0, 1]
1
L
p
[0, 1] and the embedding is strict, and,
thus,
f
C(p)
=H(|f |)
p
H(|f |)
q
=f
C(q)
for all f Ces
q
[0, 1], that is, Ces
q
[0, 1]
1
Ces
p
[0, 1] and the embedding is strict
since for positive decreasing functions the norms of Ces
p
and L
p
are equivalent.
The last statement follows from the fact that for a positive decreasing function f on
I we have f (x)
1
x
_
x
0
f (t ) dt for x I and so
f
p
Hf
p
=f
C(p)
p

f
p
for any 0 f L
p
(I).
(f) Consider f (x) =
1
1x
for x [0, 1). Then, as it was shown in (d), f
Ces
p
[0, 1] for any 1 p <. However, its non-increasing rearrangement f

(t ) =
t
1
(0 <x 1) does not belong to Ces
p
[0, 1] for any 1 p and therefore the
space Ces
p
[0, 1] is not rearrangement invariant for 1 p < . In the case when
p = we can take the function g(x) =
1

1x
, x [0, 1) for which
1
x
_
x
0
g(t ) dt =
2
x
(1

1 x) =
2
1+

1x
and so g
C()
= 2 and for its rearrangement g

(t ) =
t
1/2
, t (0, 1) we have g

C()
=sup
t (0,1)
2t
1/2
=, that is, g

/ Ces

[0, 1]
and the space Ces

[0, 1] is not rearrangement invariant. Similarly, we can consider

the case when I =[0, ).
(g) If 1 < p < , then Ces
p
(I) contains a copy of L
1
(I) (cf. [4], Lemma 1
for I = [0, 1] and Theorem 2 for I = [0, )) and therefore, in particular, these
spaces cannot be reexive. Of course, Ces
1
[0, 1] = L
1
(ln1/t ) is not reexive and
the space Ces

(I) does not have absolutely continuous norm and therefore is also
not reexive.
(h) Assume that f
C(p)
=g
C(p)
=1 and f +g
C(p)
=2; then H(|f |)
L
p =
H(|g|)
L
p =1 and
2 =f +g
C(p)
=H(|f +g|)
L
p
H(|f |) +H(|g|)
L
p H(|f |)
L
p +H(|g|)
L
p
=f
C(p)
+g
C(p)
=2.
Thus H(|f |) +H(|g|)
L
p =2 and by the strict convexity of L
p
(I) for 1 <p <
and the above estimates we obtain that H(|f |)(x) =H(|g|)(x) for almost all x in I .
Therefore, |f (x)| = |g(x)| for almost all x I . We want to show that this implies
that f (x) =g(x) for almost all x I . Assume on the contrary that f =g on I , that
335
is, there exists a set A I of positive measure m(A) > 0 such that f (x) = g(x)
for all x A. Then f (x) =g(x) and |f (x)| > 0 for x A. Moreover, if B ={x
I: m([0, x] (I \ A)) <x}, then m(B) >0 and
x
_
0

f (t ) +g(t )
2

dt =
_
[0,x](I\A)
|f (t )| dt <
x
_
0
|f (t )| dt
for all x B. Therefore,
1 =
_
_
_
_
f +g
2
_
_
_
_
p
C(p)
=
_
I
_
1
x
x
_
0

f (t ) +g(t )
2

dt
_
p
dx
<
_
I
_
1
x
x
_
0
|f (t )| dt
_
p
dx =f
p
C(p)
=1,
which is a contradiction and the proof is complete. P
3. THE DUAL SPACES OF THE CESRO FUNCTION SPACES Ces
p
[0, )
We describe the dual and Kthe dual spaces of Ces
p
(I) for 1 < p < in the
case I = [0, ). The description appeared as remark in Bennett [8] paper but
it was proved recently, even for more general spaces, by Kerman, Milman and
Sinnamon [30, Theorem D] and they used in the proof some of Sinnamon results
[53, Theorem 2.1] and [52, Proposition 2.1 and Lemma 3.2].
We present here another proof following the Bennetts idea for Cesro sequence
spaces together with factorization theorems which are of independent interest. Since
the case I =[0, 1] is essentially different it will be considered in the next section.
Theorem 2. Let I =[0, ). If 1 <p <, then
(Ces
p
)

=(Ces
p
)

=D
_
p

_
, p

=
p
p 1
, (2)
with f
C(p)
p

f
D(p

)
8p

f
C(p)
, where the norm in D(p

) is given by
formula
f
D(p

)
=

f
L
p
with

f (x) =ess sup
t [x,)
|f (t )|. (3)
We need the denition of the G(p) space for 1 p < , which is the p-
convexication of Ces

[0, ), that is, its norm is given by the functional

f
G(p)
=
_
_
|f |
p
_
_
1/p
C()
=sup
x>0
_
1
x
x
_
0
|f (t )|
p
dt
_
1/p
.
336
Proposition 1. If 1 <p <, then
Ces
p
=L
p
G
_
p

_
, (4)
that is, f Ces
p
if and only if f =gh with g L
p
, h G(p

) and
f
C(p)
inf g
p
h
G(p

)
, (5)
where inmum is taken over all factorizations f =gh with g L
p
, h G(p

).
Proof. Imbedding . For f Ces
p
, f 0 let
k(x) =

_
x
u
p
_
u
_
0
|f (t )| dt
_
p1
du, x >0.
Then k(x) >0, k is decreasing and by the HlderRogers inequality
k(x) =

_
x
u
1
_
1
u
u
_
0
|f (t )| dt
_
p1
du

_

_
x
u
p
du
_
1/p
_

_
x
_
1
u
u
_
0
|f (t )| dt
_
p
du
_
1/p

=
1
(p 1)
1/p
x
11/p
f
p1
C(p)
.
We consider the factorization f =g h, where
g(x) =
_
|f (x)|k(x)
_
1/p
sgnf (x) and h(x) =|f (x)|
1/p

k(x)
1/p
.
Then
g
p
p
=

_
0
|f (x)|

_
x
u
p
_
u
_
0
|f (t )| dt
_
p1
dudx
=

_
0
u
p
_
u
_
0
|f (t )| dt
_
p1
u
_
0
|f (x)| dx du =f
p
C(p)
and, by the HlderRogers inequality,
_
x
_
0
|h(t )|
p

dt
_
p
=
_
x
_
0
|f (t )|
1/p

|f (t )|
1/p
k(t )
p

/p
dt
_
p

_
x
_
0
|f (t )| dt
_
p1
_
x
_
0
|f (t )|k(t )
p

dt
_
.
337
Hence, by the above and using the fact that k is decreasing, it yields that

_
x
_
s
1
x
_
0
|h(t )|
p

dt
_
p
ds

_
x
s
p
__
x
_
0
|f (t )| dt
_
p1
x
_
0
|f (t )|k(t )
p

dt
_
ds
=k(x)
x
_
0
|f (t )|k(t )
p

dt

x
_
0
|f (t )|k(t )
1p

dt =
x
_
0
|h(t )|
p

dt
or, equivalently,

_
x
s
p
ds
_
x
_
0
|h(t )|
p

dt
_
p1
1,
which means that
_
x
_
0
|h(t )|
p

dt
_
p1
(p 1)x
p1
and, hence,
sup
x>0
1
x
x
_
0
|h(t )|
p

dt (p 1)
1/(p1)
or h
G(p

)
(p 1)
1/p
. We have proved that
Ces
p
L
p
G
_
p

_
and
inf{g
L
ph
G(p

)
: f =g h} (p 1)
1/p
f
C(p)
.
Imbedding . Let f =g h with g L
p
and h G(p

). Then
x
_
0
|h(t )|
p

dt h
p

G(p

)
x
_
0
dt
338
and then, for any positive decreasing function w on (0, ), we have by [32, property
18
0
, p. 72] that
x
_
0
|h(t )|
p

w(t ) dt h
p

G(p

)
x
_
0
w(t ) dt.
By the HlderRogers inequality we nd that
_
x
_
0
|f (t )| dt
_
p
=
_
x
_
0
|g(t )|w(t )
1/p

|h(t )|w(t )
1/p

dt
_
p

x
_
0
|g(t )|
p
w(t )
1p
dt
_
x
_
0
|h(t )|
p

w(t ) dt
_
p1

x
_
0
|g(t )|
p
w(t )
1p
dt h
p
G(p

)
_
x
_
0
w(t ) dt
_
p1
and, thus,

_
0
_
1
x
x
_
0
|f (t )| dt
_
p
dx

_
0
x
p
_
x
_
0
|g(t )|
p
w(t )
1p
dt
__
x
_
0
w(t ) dt
_
p1
dxh
p
G(p

)
.
Taking in the last estimate w(t ) =t
1/p
we obtain that
f
p
C(p)

_
0
x
p
_
x
_
0
|g(t )|
p
t
11/p
dt
_
_
x
11/p
1 1/p
_
p1
dxh
p
G(p

)
=
_
p

_
p1

_
0
_
x
_
0
|g(t )|
p
t
11/p
dt
_
x
1/p2
dxh
p
G(p

)
=
_
p

_
p1

_
0
_

_
t
x
1/p2
dx
_
|g(t )|
p
t
11/p
dt h
p
G(p

)
=
_
p

_
p

_
0
|g(t )|
p
dt h
p
G(p

)
=
_
p

_
p
g
p
p
h
p
G(p

)
or
f
C(p)
p

g
p
h
G(p

)
,
339
that is, L
p
G(p

) Ces
p
and
f
C(p)
p

inf{g
p
h
G(p

)
: f =gh}.
Putting these facts together we have that Ces
p
(p1)
1/p
L
p
G(p

)
p

Ces
p
and
the proof of Proposition 1 is complete. P
Proposition 2. If 1 p <, then
D(p) G(p) =L
p
and
f
L
p =inf{g
D(p)
h
G(p)
: f =gh, g D(p), h G(p)}.
Moreover, G(1)

=D(1) with equality of the norms.

Proof. It sufces to prove the statement for p =1 because the general result follows
by p-convexication. Suppose that f =gh with g D(1), h G(1). Then
f
L
1 =

_
0
|g(t )h(t )| dt

_
0
g(t )|h(t )| dt.
Moreover, from the denition of the norm in G(1) it follows that
t
_
0
|h(s)| ds h
G(1)
t =h
G(1)
t
_
0

[0,)
(s) ds, t >0.
Therefore, since g decreases it follows by [32, property 18
0
, p. 72], we nd that
f
L
1 h
G(1)

_
0
g(t ) dt =h
G(1)
g
D(1)
.
Hence, D(1) G(1) L
1
and
f
L
1 inf{g
D(1)
h
G(1)
: f =gh, g D(1), h G(1)}.
This also means that G(1) D(1)

and h
D(1)
h
G(1)
. We show that we have
in fact even equality. If f D(1)

, then
1
x
x
_
0
|f (t )| dt =
1
x
1
_
0

[0,x]
(t )|f (t )| dt

1
x

[0,x]

D(1)
f
D(1)
=f
D(1)
,
340
for all x > 0, i.e., f G(1) and so D(1)

G(1) with f
G(1)
f
D(1)
. Of
course, G(1)

= D(1)

= D(1) since the norm of D(1) has the Fatou property.

Finally, if f L
1
, then, by the Lozanovski factorization theorem ([40, Theorem
6, p. 429]; cf. also [43, p. 185]), we can nd g D(1) and h D(1)

=G(1) such
that f =g h and
g
D(1)
h
G(1)
=f
L
1 .
This ends the proof of Proposition 2. P
Remark 1. In particular, Proposition 2 shows that (Ces

[0, ))

=G(1)

=D(1).
Thus, for the Cesro function space on [0, ) we get the result analogous to the
LuxemburgZaanen theorem (cf. [42]): (Ces

[0, 1])

=

L
1
[0, 1], where f

L
1
=

f
L
1
[0,1]
with

f (x) =ess sup
t [x,1]
|f (t )|.
Remark 2. For a positive weight function w and 1 p < let us dene the
weighted spaces D(w, p) and G(w, p) by the norms f
D(w,p)
= (
_

0

f (x)
p

w(x) dx)
1/p
, where

f (x) = ess sup
t [x,)
|f (t )|, and f
G(w,p)
= sup
x>0
(
1
W(x)

_
x
0
|f (t )|
p
dt )
1/p
, W(x) =
_
x
0
w(t ) dt , respectively. Proposition 2 is valid for weight-
ed spaces: If 1 p <, then D(w, p)G(w, p) =L
p
and f
L
p =inf{g
D(w,p)

h
G(w,p)
: f =gh, g D(w, p), h G(w, p)}.
Proposition 3. Let 1 < p < . If g (Ces
p
)

, then g(x) = ess sup

t [x,)
|g(t )|
(Ces
p
)

and
g
C(p)
8g
C(p)
.
Proof. Let f Ces
p
, f 0. Then
_
x
0
f (t ) dt 0 if x 0
+
. Consider two cases:
(a) If
_

0
f (s) ds = , then we select a two-sided sequence {a
k
}
kZ
such that
0 a
k
<a
k+1
, a
k
when k and
a
k
_
a
k1
f (s) ds =2
k
, k Z. (6)
(b) If A =
_

0
f (s) ds < , we nd a one-sided sequence {a
k
}
k0
such that
0 a
k
<a
k+1
, a
0
= and
a
k
_
a
k1
f (s) ds =2
k1
A, k 0. (7)
By J let us denote either Z or {k Z: k 0} depending on which of the cases
(a) or (b) we have, and let
341
P =
_
k J: there is a set A
k
[a
k1
, a
k
) such that m(A
k
) >0
and |g(s)|
1
2
g(a
k1
) for all s A
k
_
.
Note that P =. In fact, let k J be arbitrary and let i be the rst time such that
i k and
m
_
s (a
i1
, a
i
]: |g(s)|
1
2
g(a
k1
)
_
>0.
Since g(a
i1
) = g(a
k1
), then i P.
Let P ={k
i
}
m
i=l
, where k
i
< k
j
(i < j) and l may be . Moreover, it is easily
seen that either m = and k
i
when i (in the case (a)) or k
m
= 0 and
t
k
m
= (in the case (b)).
Dene the function

f (t ) =
m

i=l
_

i
f (s) ds
1
m(A
k
i
)

A
k
i
(t ),
where
i
=(a
k
i
1
, a
k
i
], and estimate its norm in Ces
p
.
Let a =lim
i
a
k
i
if l = and a =a
k
l
if l is nite. If a > 0, then

f (t ) =0
for all t [0, a). Therefore
x
_
0

f (t ) dt =0 (0 <x a). (8)

Suppose x > a. Then either (1
o
) t
i
for some i or (2
o
) there is i < m such that
t (a
k
i
, a
k
i+1
1
]. In the rst case, by (6) or (7) it yields that
x
_
0

f (t ) dt =
i1

j=l
_

j
f (s) ds
1
m(A
k
j
)
m(A
k
j
)
+
m(A
k
i
(a
k
i
1
, t ])
m(A
k
i
)
_

i
f (s) ds

a
k
i
_
0
f (s) ds 2
x
_
0
f (s) ds.
Analogously, in the second case we have that
x
_
0

f (t ) dt
a
k
i
_
0
f (s) ds
x
_
0
f (s) ds.
342
The last inequalities and equality (8) show that

f
C(p)
2f
C(p)
. (9)
Moreover, for any i running from l to m we nd that
_

f (t )|g(t )| dt =
_
A
k
i

f (t )|g(t )| dt
1
2
g(a
k
i
1
)
_
A
k
i

f (t ) dt
=
1
2
g(a
k
i
1
)
_

i
f (t ) dt. (10)
Since g decreases, then (10) implies, in particular, that
_

f (t )|g(t )| dt
1
2
_

i
f (t ) g(t ) dt. (11)
Note that, by denition of the set P, it yields that g(t ) g(a
k
i
1
) a.e. on the interval
(a
k
i1
, a
k
i
1
] if i >l and on the interval (0, a
k
l
1
] if l is nite. Moreover, taking into
account (6) or (7) once again, we have that
a
k
i
1
_
a
k
i1
f (s) ds
_

i
f (s) ds if i >l
and
a
k
l
1
_
0
f (s) ds
_

l
f (s) ds if l is nite.
Therefore, by (10), it follows that
_

f (t )|g(t )| dt
1
2
g(a
k
i
1
)
_

i
f (t ) dt
1
2
g(a
k
i
1
)
a
k
i
1
_
a
k
i1
f (t ) dt

1
2
a
k
i
1
_
a
k
i1
g(t )f (t ) dt,
where a
l1
=0 if l is nite.
Since f =0 a.e. on the interval (0, a], when l = and a =lim
i
a
k
i
> 0,
then, by summing the last inequalities and inequality (11) over all i, we get that
2

_
0

f (t )|g(t )| dt 2
m

i=l
_

f (t )|g(t )| dt
1
2

_
0
g(t )f (t ) dt,
343
whence,

_
0
g(t )f (t ) dt 4

_
0

f (t )|g(t )| dt.
Combining the last inequality with (9), we obtain that
g
C(p)
=sup
_

_
0
g(t )f (t ) dt : f
C(p)
1
_
4sup
_

_
0

f (t )|g(t )| dt : f
C(p)
1
_
4sup
_

_
0

f (t )|g(t )| dt :

f
C(p)
2
_
=8g
C(p)

and the proof is complete. P

Proof of Theorem 2. Firstly, we show that D(p

)
1
(L
p
G(p

))

. In fact, let f
D(p

) and g L
p
G(p

), then g =h k with h L
p
and k G(p

). By the Hlder
Rogers inequality and the imbedding D(p

) G(p

)
1
L
p

proved in Proposition 2
we obtain that
fg
L
1 =f hk
L
1 h
L
pf k
L
p
h
L
pk
G(p

)
f
D(p

)
,
from which it follows that D(p

) (L
p
G(p

))

and f
(L
p
G(p

))
f
D(p

)
.
Since, by Proposition 1 we have equality Ces
p
=L
p
G(p

), it follows that
D
_
p

_ p

(Ces
p
)

.
To prove the converse, take f (Ces
p
)

. Since

f |f | and D(p

) is a Banach
lattice, then by Proposition 3, we may (and will) assume that f is a non-negative
decreasing function on (0, ), i.e., f =

f . Then, by the Hardy inequality, we nd
that
f
D(p

)
=f
L
p
=sup
_

_
0
|f (x)g(x)| dx: g
L
p 1
_
p

sup
_

_
0
|f (x)g(x)| dx: g
C(p)
1
_
=p

f
(Ces
p
)
.
Therefore, f D(p

) and (Ces
p
)

8p

D(p

). P
344
4. THE DUAL SPACES OF THE CESRO FUNCTION SPACES Ces
p
[0, 1]
We describe the dual and Kthe dual of Ces
p
(I) for 1 < p < in the case I =
[0, 1]. Surprisingly this will have a different description than in the case I =[0, ).
For p = the space Ces

[0, 1] introduced by Korenblyum, Kren and Levin [31]

we denote by K and its separable part by K
0
.
As we already mentioned the Kthe dual space K

was found by Luxemburg and

Zaanen [42]: K

=

L
1
with equality of norms, where
f

L
1
=

f
L
1 , with

f (x) =ess sup
t [x,1]
|f (t )|.
Earlier the dual space of K
0
was found by Tandori [56]: (K
0
)

=

L
1
with equality
of norms.
We will nd the Kthe dual space (Ces
p
[0, 1])

for 1 < p < . Consider, for

1 < p < , a Banach ideal space U(p) on I = [0, 1] which norm is given by the
formula
f
U(p)
=
_
_
_
_
1
1 x
1/(p1)

f (x)
_
_
_
_
L
p
=
_
1
_
0
_

f (x)
1 x
1/(p1)
_
p
dx
_
1/p
, (12)
where

f (x) =ess sup
t [x,1]
|f (t )|.
Remark 3. Since min(1, p1)
1x
1x
1/(p1)
max(1, p1) for all x (0, 1), then
the norm (12) in U(p) is equivalent to the norm
f
0
U(p)
=
_
1
_
0
_

f (x)
1 x
_
p
dx
_
1/p
.
Theorem 3. If 1 <p <, then
(Ces
p
)

=(Ces
p
)

=U
_
p

_
, p

=
p
p 1
, (13)
with equivalent norms.
Before the proof of this theorem we prove some auxiliary results of independent
interest. First, for 1 < p < we dene the Banach ideal space V(p) on I =[0, 1]
generated by the functional
f
V(p)
= sup
0<x1
_
(1 x
1/(p1)
)
p1
x
x
_
0
|f (t )|
p
dt
_
1/p
. (14)
Proposition 4. If 1 <p <, then
Ces
p
L
p
V
_
p

_
, p

=
p
p 1
, (15)
345
that is, if f Ces
p
, then f =gh with g L
p
, h V(p

) and
inf
_
g
p
h
V(p

)
: f =g h, g L
p
, h V
_
p

__
(p 1)
1/p
f
C(p)
. (16)
Proof. The proof is analogous to the proof of Proposition 1 (for the case I =
[0, )) but we put details to see how the weight w(x) =(1x
p1
)
1/(p1)
appeared
in the denition of the space V(p

). For f Ces
p
, f =0, dene
k(x) =
1
_
x
u
p
_
u
_
0
|f (t )| dt
_
p1
du, x [0, 1].
Then k(x) >0, k is decreasing and, by the HlderRogers inequality, we nd that
k(x) =
1
_
x
u
1
_
1
u
u
_
0
|f (t )| dt
_
p1
du

_
1
_
x
u
p
du
_
1/p
_
1
_
x
_
1
u
u
_
0
|f (t )| dt
_
p
du
_
1/p

1
(p 1)
1/p
_
1 x
p1
x
p1
_
1/p
f
p1
C(p)
.
Let
g(x) =
_
|f (x)|k(x)
_
1/p
sgnf (x) and h(x) =|f (x)|
1/p

k(x)
1/p
, 0 <x <1.
Then f =g h and
g
p
p
=
1
_
0
|f (x)|
1
_
x
u
p
_
u
_
0
|f (t )| dt
_
p1
dudx
=
1
_
0
u
p
_
u
_
0
|f (t )| dt
_
p1
u
_
0
|f (x)| dx du =f
p
C(p)
,
and, by the HlderRogers inequality,
_
x
_
0
|h(t )|
p

dt
_
p
=
_
x
_
0
|f (t )|
1/p

|f (t )|
1/p
k(t )
p

/p
dt
_
p

_
x
_
0
|f (t )| dt
_
p1
_
x
_
0
|f (t )|k(t )
p

dt
_
.
346
Hence, by the above and using the fact that k is decreasing, we obtain that
1
_
x
_
s
1
x
_
0
|h(t )|
p

dt
_
p
ds

1
_
x
s
p
__
x
_
0
|f (t )| dt
_
p1
x
_
0
|f (t )|k(t )
p

dt
_
ds

1
_
x
s
p
__
s
_
0
|f (t )| dt
_
p1
x
_
0
|f (t )|k(t )
p

dt
_
ds
=k(x)
x
_
0
|f (t )|k(t )
p

dt

x
_
0
|f (t )|k(t )
1p

dt =
x
_
0
|h(t )|
p

dt
or, equivalently,
1
_
x
s
p
ds
_
x
_
0
|h(t )|
p

dt
_
p1
1,
which means that
_
x
_
0
|h(t )|
p

dt
_
p1
(p 1)
x
p1
1 x
p1
and, thus,
sup
x>0
(1 x
p1
)
1/(p1)
x
x
_
0
|h(t )|
p

dt (p 1)
1/(p1)
or h
V(p

)
(p 1)
1/p
. Summing up we have proved that Ces
p
L
p
V(p

) and
inf{g
L
ph
V(p

)
: f =g h} (p 1)
1/p
f
C(p)
.
P
Remark 4. In the above imbedding we cannot take instead of the space V(p

),
where the weight w(x) = (1 x
p1
)
1/(p1)
appeared, the corresponding space
without this weight, that is, the p

-convexication K
(p

)
of K. This space is too
small since if the imbedding Ces
p
[0, 1] L
p
K
(p

)
would be valid, then since
L
p
K
(p

)
L
p
L
p

L
1
[0, 1] we will have a contradiction because Ces
p
[0, 1]
is not embedded into L
1
[0, 1] (cf. Theorem 1(d)) and the problem is near 1,
therefore this weight w is really needed in the imbedding (15).
347
Proposition 5. If 1 <p <, then
(a) U(p) V(p) L
p
with
f
L
p inf{g
U(p)
h
V(p)
: f =g h, g U(p), h V(p)}.
(b) U(p) (V(p) L
p

and f
(V(p)L
p

f
U(p)
for all f U(p).
Proof. (a) Let f =g h, g U(p), h V(p). Since |g| g it follows that
f
p
L
p

1
_
0
g(t )
p
|h(t )|
p
dt. (17)
On the other hand, by the denition of the norm in V(p) and using the equality
d
dx
_
x
(1 x
1/(p1)
)
p1
_
=
1
(1 x
1/(p1)
)
p
,
we obtain that
x
_
0
|h(t )|
p
dt h
p
V(p)
x
(1 x
1/(p1)
)
p1
=h
p
V(p)
x
_
0
1
(1 t
1/(p1)
)
p
dt
for all x (0, 1]. Since g
p
decreases, then, by [32, property 18
0
, p. 72], the last
inequality implies that
1
_
0
g(t )
p
|h(t )|
p
dt h
p
V(p)
1
_
0
_
g(t )
1 t
1/(p1)
_
p
dt.
Therefore, by (17), f L
p
and
f
L
p g
U(p)
h
V(p)
,
and the proof of (a) is complete.
(b) For any f U(p) and g V(p) L
p

we have g =h k with h V(p), k L

p

and, by the HlderRogers inequality and Proposition 5(a), we obtain that

1
_
0
|fg| dx =
1
_
0
|f hk| dx
_
1
_
0
|f h|
p
dx
_
1/p
_
1
_
0
|k|
p

dx
_
1/p

f
U(p)
h
V(p)
k
L
p
=h
V(p)
k
L
p
f
U(p)
or f (V(p) L
p

and f
(V(p)L
p

f
U(p)
. The proof of (b) is complete. P
348
Proposition 6. Let 1 p < . If g (Ces
p
)

, then g(x) = ess sup

t [x,1]
|g(t )|
(Ces
p
)

and
g
C(p)
8g
C(p)
.
Proof. When p = 1, then the assertion is obvious since Ces
1
= L
1
(ln1/t ) and
(Ces
1
)

=L

(ln
1
1/t ). Let p >1 and f Ces
p
, f 0. Consider two cases:
(a) If
_
1
0
f (s) ds = , then we select a two-sided sequence {a
k
}
kZ
such that 0
a
k
<a
k+1
, a
k
1 when k and
a
k
_
a
k1
f (s) ds =2
k
, k Z. (18)
(b) If A =
_
1
0
f (s) ds < , then we can nd an one-sided sequence {a
k
}
k0
such
that 0 a
k
<a
k+1
, a
0
=1 and
a
k
_
a
k1
f (s) ds =2
k1
A, k 0. (19)
The remaining part of the proof is completely analogous to the proof of Proposi-
tion 3 so we omit the details. P
Proof of Theorem 3. Imbedding . If f U(p

), then, by Proposition 5(b) and

Proposition 4, we obtain that
U
_
p

_
V
_
p

_
L
p
_

(Ces
p
)

and f
C(p)
(p 1)
1/p
f
U(p

)
.
Imbedding . Let f (Ces
p
)

. Since

f |f | and U(p

) is a Banach lattice,
then by Proposition 6 we may (and we will) assume that f is a non-negative
decreasing function on (0, 1], i.e., f =

f . Dene the weight
w(x) =
[0,1/2]
(x) +(1 x)
[1/2,1]
(x), 0 <x 1.
Since 1 x w(x) 2(1 x) for x (0, 1], then according to Remark 3 it is
enough to prove that for some constant A
p
>0 we have that
_
_
_
_
f
w
_
_
_
_
L
p

=
_
1
_
0
_
f (x)
w(x)
_
p

dx
_
1/p
A
p
f
C(p)
(20)
since
f
U(p

)
=
_
_
_
_
1
1 x
1/(p

1)
f (x)
_
_
_
_
L
p

max
_
1, p

1
_
_
_
_
_
1
1 x
f (x)
_
_
_
_
L
p

2max
_
1, p

1
_
f/w
L
p
.
349
We now prove that if h L
p
, h 0, then h/w Ces
p
and
h/w
C(p)

_
p

+2p
_
h
L
p. (21)
To prove this we rst show that the operator S
w
dened by
S
w
h(x) =
x
_
0
h(t )
w(t )
dt (0 <x 1)
is bounded in L
p
[0, 1] for 1 p <. In fact, for 0 <x 1/2 we have that
S
w
h(x) =
x
_
0
h(t ) dt =
1
_
1x
h(1 t ) dt
1
_
1x
h(1 t )
t
dt
and for 1/2 x 1
S
w
h(x) =
1/2
_
0
h(t ) dt +
x
_
1/2
h(t )
1 t
dt
=
1
_
1/2
h(1 t ) dt +
1/2
_
1x
h(1 t )
t
dt
1
_
1x
h(1 t )
t
dt.
Thus,
S
w
h(x) H

h)(1 x) for 0 <x <1,

where

h(t ) = h(1 t ) and H

is the associated Hardy operator, i.e., H

h(x) =
_
1
x
h(t )
t
dt . It is well known that H

is bounded in L
p
[0, 1] for 1 p < (cf. [32],
pp. 138139) and, thus,
S
w
h
L
p
_
_
H

h)
_
_
L
p

_
_
H

_
_

h
L
p =
_
_
H

_
_
h
L
p .
Since
1
x
S
w
h(x) =
1
x
x
_
0
h(t )
w(t )
dt
1
x
x
_
0
h(t ) dt
[0,
1
2
]
(x) +2S
w
h(x)
[
1
2
,1]
(x)
it follows that
h/w
C(p)
=
_
_
_
_
1
x
S
w
h(x)
_
_
_
_
L
p
Hh
L
p +2S
w
h
L
p
p

h
L
p +2ph
L
p =
_
p

+2p
_
h
L
p
350
and the estimate (21) is proved. Moreover, by using this fact we obtain that
_
_
_
_
f
w
_
_
_
_
L
p

=sup
_
1
_
0
f (t )
w(t )
h(t ) dt : h 0, h
L
p 1
_
sup
_
1
_
0
f (t )
w(t )
h(t ) dt : h 0,
_
_
_
_
h
w
_
_
_
_
C(p)
p

+2p
_

_
p

+2p
_
f
C(p)

and also the estimate (20) is proved, which shows that (Ces
p
)

U(p

) and for
every f (Ces
p
)

f
U(p

)
16max
_
1, p

1
__
p

+2p
_
f
C(p)
,
and the proof is complete. P
Remark 5. Let 1 < p < . The L
p
spaces have the property that the restriction
of L
p
[0, ) to [0, 1] gives the space L
p
[0, 1]. The situation is different for Cesro
function spaces. In fact, if f Ces
p
[0, ) and suppf [0, 1], then
f
p
Ces
p
[0,)
=

_
0
_
1
x
x
_
0
|f (t )| dt
_
p
dx
=
1
_
0
_
1
x
x
_
0
|f (t )| dt
_
p
dx +

_
1
_
1
x
1
_
0
|f (t )| dt
_
p
dx
=f
p
Ces
p
[0,1)
+
1
p 1
f
p
L
1
[0,1]
,
which means that
Ces
p
[0, )|
[0,1]
=Ces
p
[0, 1] L
1
[0, 1].
Therefore,
(Ces
p
[0, 1] L
1
[0, 1])

=
_
Ces
p
[0, )
_

|
[0,1]
=D
_
p

_
|
[0,1]
or
U
_
p

_
+L

[0, 1] =D
_
p

_
|
[0,1]
.
The last equality can be easily veried. For example, for f D(p

)|
[0,1]
we can take
as a decomposition f =g +h, g U(p

), h L

[0, 1] the functions

g(x) =(1 x)f (x) and h(x) =xf (x), x [0, 1].
Then f =g +h and g(x) =ess sup
t [x,1]
(1 t )|f (t )| (1 x)

f (x), which shows
351
that g U(p

) since f D(p

). Moreover,
h

=ess sup
x[0,1]
x|f (x)| ess sup
x[0,1]
x

f (x)

f
L
1

f
L
p
=f
D(p

)
,
so that h L

[0, 1].
5. ON p-CONCAVITY, TYPE AND COTYPE OF CESRO SEQUENCE AND FUNCTION
SPACES
A Banach lattice X is said to be p-convex (1 p < ) with constant K 1,
respectively q-concave (1 q <) with constant L 1 if
_
_
_
_
_
_
n

k=1
|x
k
|
p
_
1/p
_
_
_
_
_
K
_
n

k=1
x
k

p
_
1/p
,
respectively
_
n

k=1
x
k

q
_
1/q
L
_
_
_
_
_
_
n

k=1
|x
k
|
q
_
1/q
_
_
_
_
_
,
for every choice of vectors x
1
, x
2
, . . . , x
n
in X.
Of course, every Banach lattice is 1-convex with constant 1. In particular, ces
p
and Ces
p
(I) are 1-convex with constant 1. The spaces L
p
(I) are p-convex and
p-concave with constant 1.
If the above estimates hold for pairwise disjoint elements {x
k
}
n
k=1
in X, that is,
_
_
_
_
_
n

k=1
x
k
_
_
_
_
_
K
_
n

k=1
x
k

p
_
1/p
,
respectively
_
n

k=1
x
k

q
_
1/q
L
_
_
_
_
_
n

k=1
x
k
_
_
_
_
_
,
then we say that X satises an upper p-estimate with constant K and a lower q-
estimate with constant L, respectively. It is obvious that a p-convex (q-concave)
Banach lattice satises upper p-estimate (lower q-estimate).
Let r
n
: [0, 1] R, n N, be the Rademacher functions, that is, r
n
(t ) =
sign(sin2
n
t ). A Banach space X has type 1 p 2 if there is a constant K > 0
such that, for any choice of nitely many vectors x
1
, . . . , x
n
from X,
1
_
0
_
_
_
_
_
n

k=1
r
k
(t )x
k
_
_
_
_
_
dt K
_
n

k=1
x
k

p
_
1/p
.
352
A Banach space X has cotype q 2 if there is a constant K > 0 such that, for
any choice of nitely many vectors x
1
, . . . , x
n
from X,
_
n

k=1
x
k

q
_
1/q
K
1
_
0
_
_
_
_
_
n

k=1
r
k
(t )x
k
_
_
_
_
_
dt.
In order to complete this denition for q =the left-hand side should be replaced
by max
1kn
x
k
.
We say that the space X has trivial type or trivial cotype, if it does not have any
type bigger than one or any nite cotype, respectively.
More information and connections among the above notions may be found in
[17] and [38].
Theorem 4. If 1 <p <, then Ces
p
(I) are p-concave with constant 1, that is,
_
n

k=1
f
k

p
C(p)
_
1/p

_
_
_
_
_
_
n

k=1
|f
k
|
p
_
1/p
_
_
_
_
_
C(p)
, (22)
for all f
1
, f
2
, . . . , f
n
Ces
p
(I).
Proof. Inequality (22) taken to the power p means that
n

k=1
_
I
_
1
x
x
_
0
|f
k
(t )| dt
_
p
dx
_
I
_
1
x
x
_
0
_
n

k=1
|f
k
(t )|
p
_
1/p
dt
_
p
dx.
If we show that
n

k=1
_
1
x
x
_
0
|f
k
(t )| dt
_
p

_
1
x
x
_
0
_
n

k=1
|f
k
(t )|
p
_
1/p
dt
_
p
for every x I , then we are done. The last estimate can also be written as
_
n

k=1
_
x
_
0
|f
k
(t )| dt
_
p
_
1/p

x
_
0
_
n

k=1
|f
k
(t )|
p
_
1/p
dt,
which is the p-concavity of L
1
[0, x] for every x I .
It is clear that L
1
(J), J = J
x
= [0, x] is 1-convex with constant 1 and it is well
known that then L
1
(J) is p-concave with constant 1 (cf. [38, Proposition 1.d.5] or
[44, Theorem 4.3]). We can also prove this fact directly as in [44, Theorem 4.3]: by
the HlderRogers inequality for t J it yields that
n

k=1
|f
k
(t )||a
k
|
_
n

k=1
|f
k
(t )|
p
_
1/p
{a
k
}
p

353
and, by integrating over J,
_
J
n

k=1
|f
k
(t )||a
k
| dt {a
k
}
p

_
I
_
n

k=1
|f
k
(t )|
p
_
1/p
dt
={a
k
}
p

_
_
_
_
_
_
n

k=1
|f
k
|
p
_
1/p
_
_
_
_
_
L
1
(J)
.
Taking the supremum over all {a
k
} such that {a
k
}
p
1 we obtain, by the Landau
theorem,
sup
_
_
J
n

k=1
|f
k
(t )||a
k
| dt : {a
k
}
p
1
_
=sup
_
n

k=1
|a
k
|
_
J
|f
k
(t )| dt : {a
k
}
p
1
_
=
_
_
_
_
__
J
|f
k
(t )| dt
__
_
_
_
p
=
_
n

k=1
__
J
|f
k
(t )| dt
_
p
_
1/p
=
_
n

k=1
f
k

p
L
1
(J)
_
1/p
.
Thus,
_
n

k=1
f
k

p
L
1
(J)
_
1/p

_
_
_
_
_
_
n

k=1
|f
k
|
p
_
1/p
_
_
_
_
_
L
1
(J)
,
and putting these facts together we obtain the estimate (22). P
Theorem 5. If 1 < p < , then the space Ces
p
(I) has trivial type and cotype
max(p, 2). The space Ces

(I) has trivial type and trivial cotype.

Proof. Let 1 < p < . The space Ces
p
(I) contains a copy of L
1
(I) (cf. [4],
Lemma 1 for I =[0, 1] and Theorem 2 for I =[0, )) which implies that Ces
p
(I)
has trivial type.
On the other hand, since, by Theorem 4 the space Ces
p
(I) is p-concave, then
by a well-known theorem (cf. Lindenstrauss and Tzari [38, p. 100]) it has cotype
max(p, 2). The fact that this space has no smaller cotype follows, for example, from
Theorem6 showing that Ces
p
(I) contains an isomorphic copy of l
p
and the fact that
the space l
p
has cotype max(p, 2) and this value is the best possible (cf. [38, p. 73]
or [44, pp. 9194]).
For p = the space Ces

(I) has no absolutely continuous norm and, by the

Lozanovski theorem (see [40, Theorem 5, p. 65]; cf. also [28, Theorem 4 in X.4]
354
and [59, Theorem 4.1]), it contains an isomorphic copy of l

, therefore it has trivial

type and trivial cotype. The proof is complete. P
Remark 6. Similarly as in Theorem 4 we can prove that the Cesro sequence
spaces ces
p
are p-concave with constant 1 since l
1
is p-concave with constant 1.
Moreover, similarly as in Theorem 5 we can obtain that the Cesro sequence spaces
ces
p
have trivial type and cotype max(p, 2) for 1 < p < . Also ces

has trivial
type and trivial cotype.
6. COPIES OF l
p
SPACES IN THE CESRO FUNCTION SPACES Ces
p
The Cesro function space Ces
p
(I) contains a copy of L
1
(I) and as we will see in
the next theorem also complemented copies of l
p
.
Theorem 6. If 1 < p < , then Ces
p
(I) contains an order isomorphic and
complemented copy of l
p
.
Proof. Let I = [0, 1]. We shall construct a sequence {f
n
}

n=1
Ces
p
[0, 1] with
disjoint supports which spans an isomorphic copy of l
p
in Ces
p
[0, 1] and the closed
linear span [f
n
]
Ces
p
is complemented in Ces
p
[0, 1]. Let us denote
f
n
=
[2
n1
,2
n
]
and F
n
(t ) =
1
t
t
_
0
f
n
(s) ds, n =1, 2, . . . .
Since
F
n
(t ) =

0, if 0 <t 2
n1
,
1
1
2
n+1
t
, if 2
n1
t 2
n
,
1
2
n+1
t
, if t 2
n
,
it follows that
f
n

p
C(p)
=F
n

p
L
p
=
2
n
_
2
n1
_
1
1
2
n+1
t
_
p
dt +2
(n+1)p
2
n(p1)
1
p 1
.
Note that the rst term in the above sum is not bigger than 2
pn1
and the second
one satises the inequalities
1 2
p+1
p 1
2
pn
2
(n+1)p
2
n(p1)
1
p 1

2
pn
p 1
.
Therefore,
f
n

C(p)
f
n

L
p 2
n/p
(23)
355
with constants which depend only on p. If

f
n
=
f
n
f
n

C(p)
, n =1, 2, . . . ,
then
1 =

f
n

C(p)

f
n

L
p , n N.
Let us denote
x(t ) =

n=1

n

f
n
,
n
R.
Since

f
n
are disjoint functions we may assume that
n
0. By Theorem 1(c) (the
Hardy inequality) and the above equivalence
x
C(p)

p
p 1
x
L
p =
p
p 1
_

n=1

p
n

f
n

p
L
p
_
1/p
C
p
_

n=1

p
n
_
1/p
.
On the other hand, by Theorem 4, for any n N,
_
_
_
_
_
n

k=1

k

f
k
_
_
_
_
_
C(p)

_
n

k=1

k

f
k

p
C(p)
_
1/p
=
_
n

k=1

p
k
_
1/p
and passing to the limit as n we arrive at the inequality
x
C(p)

k=1

p
k
_
1/p
x
L
p,
which together with estimation from above gives us that
[

f
n
]
Ces
p
[

f
n
]
L
p l
p
. (24)
Next, we prove that [

f
n
]
Ces
p
is complemented in Ces
p
for 1 < p < . Let x
Ces
p
, x 0 and suppx [2
n1
, 2
n
], n N. Then
1
t
t
_
0
x(s) ds =
1
t
t
_
2
n1
x(s) ds
[2
n1
,2
n
]
(t ) +
1
t
x
L
1
[2
n
,1]
(t )
and
x
p
C(p)
=
2
n
_
2
n1
_
1
t
t
_
2
n1
x(s) ds
_
p
dt +x
p
L
1
1
_
2
n
t
p
dt.
356
The rst term in the last sum is not bigger than
x
p
L
1
2
n
_
2
n1
t
p
dt =
2
p1
1
p 1
2
n(p1)
x
L
1 ,
and the second one is equal to
x
p
L
1
2
n(p1)
1
p 1
.
Therefore,
x
C(p)
x
L
1 2
n(11/p)
, n =1, 2, . . . , (25)
with constants which depend only on p. We consider the orthogonal projector
T x(t ) :=

k=1
2
k+1
2
k
_
2
k1
x(s) ds
[2
k1
,2
k
]
(t ) (26)
and prove that it is bounded in Ces
p
.
For arbitrary x Ces
p
, x 0 we set x
k
=x
[2
k1
,2
k
]
(k =1, 2, . . .). Since
T x
k
=x
k

L
1 2
k+1

[2
k1
,2
k
]
,
then (23) and (25) imply that
T x
k

C(p)
=x
k

L
1 2
k+1
f
k

C(p)
Bx
k

L
1 2
k+1
2
k/p
Cx
k

C(p)
.
Therefore, by (24) and Theorem 4, we have that
T x
C(p)
C

k=1
T x
k

p
C(p)
_
1/p
C

C
_

k=1
x
k

p
C(p)
_
1/p
C

C
_
_
_
_
_

k=1
x
k
_
_
_
_
_
C(p)
=C

Cx
C(p)
,
and the proof of the boundedness of T in Ces
p
is complete. Since the image of T
coincides with [x
n
]
Ces
p
, then Theorem 6 is proved. P
The above theorem shows that the Cesro function spaces Ces
p
[0, 1] behave
near zero similar to the l
p
spaces. The authors proved in [4] that in the middle
Cesro function spaces Ces
p
(I) contain an asymptotically isometric copy of l
1
,
that is, there exist a sequence {
n
} (0, 1),
n
0 as n and a sequence of
functions {f
n
} Ces
p
[0, 1] such that, for arbitrary {
n
} l
1
, we have that

n=1
(1
n
)|
n
|
_
_
_
_
_

n=1

n
f
n
_
_
_
_
_
C(p)

n=1
|
n
|. (27)
357
Consequently, these spaces are not reexive and do not have the xed point property.
This is a big difference with the Cesro sequence spaces ces
p
, which for 1 <p <
are reexive and have the xed point property.
Let us recall that a Banach space X has the DunfordPettis property if x
n
0
weakly in X and f
n
0 weakly in the dual space X

imply f
n
(x
n
) 0. The
classical examples of Banach spaces with the DunfordPettis property are the
AL-spaces and AM-spaces. It is clear that if X

has the DunfordPettis property,

then X has itself this property (cf. [2, pp. 334336]). Of course, the Cesro sequence
spaces ces
p
, 1 < p < , as reexive spaces do not have the DunfordPettis
property.
Corollary 1. If 1 <p <, then Ces
p
(I) do not have the DunfordPettis property.
Proof. By Theorem 6, Ces
p
(I) contains a complemented copy of l
p
and l
p
do
not have the DunfordPettis property. On the other hand, it is easy to show that, if a
Banach space has the DunfordPettis property, then its complemented subspace has
also the DunfordPettis property (cf. Wnuk [58, Lemma 1(i)] or [23, Proposition
11.37]). Thus, Ces
p
(I) do not have the DunfordPettis property. P
As it was mentioned before the Cesro sequence spaces ces
p
are not isomorphic
to the l
q
space for any 1 q . An analogous theorem is true for Cesro function
spaces.
Theorem 7. If 1 < p , then Ces
p
(I) are not isomorphic to any L
q
(I) space
for any 1 q .
Proof. If 1 <q <, then Ces
p
(I) has trivial type but L
q
(I) has type min(q, 2) >1
and therefore they cannot be isomorphic. The space Ces
p
(I) for 1 < p < is not
isomorphic to L
1
(I) since L
1
(I) has the DunfordPettis property but Ces
p
(I), as
we have seen in Corollary 1, do not have the DunfordPettis property. Also Ces
p
(I)
for 1 <p <is not isomorphic to L

(I) since the rst space is separable and the

second one is non-separable.
It only remains to show that Ces

(I) is not isomorphic to L

(I). Since, by
Peczy nski theorem L

(I) is isomorphic to

(cf. Albiac and Kalton [1, Theorem

4.3.10]), therefore it is enough to show that Ces

(I) is not isomorphic to

.
We show this for K = Ces

[0, 1] since for the case of Ces

(0, ) the proof

is similar. For xed a (0, 1) dene a continuous projection P : K K by Pf =
f
[a,1]
. Then
1
_
a
|Pf (t )| dt
1
_
0
|Pf (t )| dt Pf
K
= sup
0<x1
1
x
x
_
0
|f (t )
[a,1]
(t )| dt
= sup
ax1
1
x
x
_
a
|f (t )
[a,1]
(t )| dt
1
a
1
_
a
|Pf (t )| dt.
358
Hence, P(K) is isomorphic to L
1
[a, 1], i.e., K contains a complemented copy of a
separable space while no separable subspace of

is complemented in

because

is prime, that is, every innite dimensional complemented subspace of

is
isomorphic to

(see Lindenstrauss and Tzafriri [37, Theorem 2.a.7], or Albiac

and Kalton [1, Theorem 5.6.5]). Therefore, K and

are not isomorphic. P

7. ON THE WEAK BANACHSAKS PROPERTY OF THE CESRO FUNCTION SPACES
Let us recall that a Banach space X is said to have the weak BanachSaks property
if every weakly null sequence in X, say (x
n
), contains a subsequence (x
n
k
) whose
rst arithmetical means converge strongly to zero, that is,
lim
m
1
m
_
_
_
_
_
m

k=1
x
n
k
_
_
_
_
_
=0.
It is known that uniformly convex spaces, c
0
, l
1
and L
1
have the weak Banach
Saks property, whereas C[0, 1] and l

do not have. We should mention that the

result on L
1
space, proved by Szlenk [55] in 1965, was a very important break-
through in studying of the weak BanachSaks property.
In 1982, Rakov [48, Theorem 1] proved that a Banach space with non-trivial
type (or equivalently B-convex) has the weak BanachSaks property (cf. also
[57, Theorem 1]). Recently Dodds, Semenov and Sukochev [19] investigated the
weak BanachSaks property of rearrangement invariant spaces and Astashkin and
Sukochev [6] have got a complete description of Marcinkiewicz spaces with the
latter property.
The spaces Ces
p
[0, 1] for 1 p < are neither B-convex (they have trivial
type) nor rearrangement invariant. Nevertheless, we will prove that Ces
p
for all
1 p <have the weak BanachSaks property.
Theorem 8. If 1 p <, then the Cesro function space Ces
p
[0, 1] has the weak
BanachSaks property.
We begin with some auxiliary notation and results.
If I =[a, b] and J =[c, d] are two closed intervals, then we write I <J if b c.
Let {I
n
}

n=1
be a sequence of closed intervals I
n
= [a
n
, b
n
] [0, 1]. Then I
n
0
means that I
1
> I
2
> and b
n
0
+
. Analogously, I
n
1 means that I
1
< I
2
<
and a
n
1

. Moreover, in what follows suppf ={t : f (t ) =0}.

Lemma 1 (Weakly null sequences in Ces
p
[0, 1], 1 < p < ). Let {x
n
}

n=1

Ces
p
. Then x
n
w
0 in Ces
p
if and only if
(a) there exists a constant M >0 such that x
n

C(p)
M for all n =1, 2, . . . ,
and
(b) for every set A [0, 1] such that A [h, 1 h] for some h (0,
1
2
) we have
_
A
x
n
(t ) dt 0 as n .
359
Proof. It is enough to check that the set of all functions of the form
a(t ) =
n

k=1
a
k

A
k
(t ), (28)
where n N, a
k
R and A
k
[0, 1] are pairwise disjoint sets such that A
k
[h, 1
h] for some h (0,
1
2
), is dense in the space U(p

) =(Ces
p
)

=(Ces
p
)

, p

=
p
p1
,
with the norm
y
U(p

)
=
_
1
_
0
_
y(t )
1 t
_
p

dt
_
1/p

, y(t ) =ess sup

s[t,1]
|y(s)|.
Let y U(p

) and > 0. Note that y(1

) = lim
t 1
y(t ) = 0. In fact, if y(t )
c > 0 (0 < t < 1), then since p

> 1 we have that y

p

U(p

)
c
_
1
0
1
(1t )
p

dt = .
Therefore, we may choose (0, 1) and h (0, ) so that
max
_

_
0
_
y(t )
1 t
_
p

dt,
1
_
1
_
y(t )
1 t
_
p

dt
_

(29)
and
y(1 h)
_
p

1p

1
_
1/p

. (30)
Since y U(p

), then y(t ) is nite for every t (0, 1) which implies that y(t ) is
a bounded measurable function on the interval [h, 1 h]. Therefore, there exists a
function a(t ) of the form (28) such that suppa [h, 1 h] and
(y a)
[h,1h]

L

_
p

1
h
1p

1
_
1/p

. (31)
By the triangle inequality we have that
y a
U(p

)
y
[0,h]

U(p

)
+(y a)
[h,1h]

U(p

)
+y
[1h,1]

U(p

)
, (32)
and let us estimate each of the three terms separately. At rst, since 0 <h <, then,
by (29),
y
[0,h]

U(p

_
0
_
y(t )
1 t
_
p

dt
p

. (33)
Next, (31) implies
(y a)
[h,1h]

U(p

1h
_
0
dt
(1 t )
p

_
p

1
h
1p

1
_
=
p

. (34)
360
Finally, (30) and (29) imply that
y
[1h,1]

U(p

1
_
0
_
y(1 h)
1 t
_
p

dt +
1
_
1
_
y(t )
1 t
_
p

dt 2
p

. (35)
Thus, by (32)(35), we have that y a
U(p

)
4
1/p

, and the proof is com-

plete. P
Corollary 2. Let {I
n
}

n=1
be a sequence of intervals from [0, 1] such that either
I
n
0 or I
n
1. Then, for every p (1, ), we have

I
n

I
n

C(p)
w
0 in Ces
p
[0, 1].
Following Kadec and Peczy nski [27] (see also [46] and [47]) we will use the
following notation: Let X be a Banach function lattice on [0, 1]. For every x X
and (0, 1] we set
(x, ) = sup
A[0,1],m(A)=
x
A

X
.
Moreover, if K X, then
(K, ) = sup
xK
(x, ), (K, 0
+
) = lim
0
+
(K, ).
Lemma 2. If a Banach function lattice X on [0, 1] satises a lower p-estimate
(1 p < ) with constant one, then for any disjointly supported x, y X and
>0, >0 we have that
(x +y, +)
p
(x, )
p
+(y, )
p
.
Proof. For any >0 choose the sets A and B from[0, 1] such that Asuppx, B
suppy, m(A) , m(B) , and
x
A

p
X
(x, )
p
, y
B

p
X
(y, )
p
.
Since m(A B) + and X satises a lower p-estimate with constant one it
follows that
(x +y, +) (x +y)
AB

X
=x
A
+y
B

_
x
A

p
X
+y
B

p
X
_
1/p

_
(x, )
p
+(y, )
p
2
_
1/p
,
and the proof of the lemma follows by letting 0
+
. P
Let X be a Banach function lattice on [0, 1] and a set K X. We say that
K consists of elements having equicontinuous norms in X if
lim
A[0,1],m(A)0
sup
xK
x
A

X
=0.
361
An important tool in the proof of Theorem 8 will be the following assertion:
Proposition 7 (Subsequence splitting property). Let 1 < p < , {x
n
}

n=1

Ces
p
[0, 1], x
n

C(p)
=1 and x
n
w
0 in Ces
p
[0, 1]. Then there exists a subsequence
{x

n
} {x
n
} such that
x

n
=y
n
+z
n
, n =1, 2, . . . ,
where {y
n
}

n=1
consists of elements having equicontinuous norms in Ces
p
and
suppz
n
I

n
I

n
with {I

n
, I

n
}

n=1
being a sequence of pairwise disjoint intervals
from [0, 1] such that I

n
0, I

n
1. Moreover, y
n
w
0, z
n
w
0 in Ces
p
.
Proof. We set
0
= ({x
n
}, 0
+
). If
0
= 0, then the sequence {x
n
} consists of
elements with equicontinuous norms in Ces
p
and we have nothing to prove.
Therefore, assume that
0
> 0. By the denition of
0
, there exists a sequence of
sets A
n
[0, 1], m(A
n
) =
n
0 and a subsequence of {x
n
} (which will be denoted
also by {x
n
}) such that for all n N
x
n

A
n

C(p)

1
n
. (36)
Let us denote
u
n
=x
n

A
n
and v
n
=x
n
u
n
. (37)
Since Ces
p
[0, 1] is p-concave with constant one, then, by Lemma 2, it yields that
(v
n
, )
p
(x
n
, +
n
)
p
(u
n
,
n
)
p
(x
n
, +
n
)
p

1
n
_
p
.
Hence, for 0 < 1/2 we have that
lim sup
n
(v
n
, )
p
({x
n
}, 2)
p

p
0
.
Since Ces
p
is a separable space the last inequality implies that

_
{v
n
}, 0
+
_
=0, (38)
that is, {v
n
} consists of elements with equicontinuous norms in Ces
p
.
According to Lemma 1, for every h (0,
1
2
),
x
n

[h,1h]
w
0 in Ces
p
. (39)
Therefore, since Ces
p
[0, 1]
|[h,1h]
= L
1
[h, 1 h] with equivalent norms (see [4,
Lemma 1]) we have that x
n

[h,1h]
w
0 in L
1
. Moreover, since ({v
n
}, 0
+
) = 0 it
follows that

L
1
_
{v
n

[h,1h]
}, 0
+
_
=0
362
(where
L
1 is calculated in the space L
1
) and
v
n

[h,1h]

L
1 Cv
n

[h,1h]

C(p)
C.
Thus, by the classical DunfordPettis criterion (see, for example, [22, Theorem
4.21.2] or [1, Theorem 5.2.9]), the sequence {v
n

[h,1h]
}

n=1
is a relatively weakly
compact subset of L
1
and, hence, simultaneously in Ces
p
. Therefore, there is a
subsequence {v
n
k
} {v
n
} such that v
n
k

[h,1h]
w
v, where v Ces
p
. By combining
the last mentioned facts with (39) and with the equality x
n
= u
n
+v
n
, we get that
u
n
k

[h,1h]
w
v in Ces
p
, and, hence, in L
1
. Taking into account the denition of
u
n
(see (36) and (37)) and using again the DunfordPettis criterion we conclude
that for every h (0, 1/2) there exists a subsequence {u
n
k
} {u
n
} (depending on h)
such that
u
n
k

[h,1h]

C(p)
0 as k .
Since Ces
p
is a separable space, then by a standard procedure, we may choose
a subsequence of {u
n
k
} (denote it again by {u
n
k
}) and pairwise disjoint intervals
{I

k
, I

k
}
kN
, I

k
0, I

k
1 such that
u
n
k

[0,1]\(I

k
I

k
)

C(p)
0 as k . (40)
Setting x

k
=y
k
+z
k
, with
y
k
=v
n
k
+u
n
k

[0,1]\(I

k
I

k
)
, z
k
=u
n
k

k
I

k
,
we see that, by (38) and (40), this representation satises all conditions. In
particular, according to Lemma 1, we have that z
k
w
0 and y
k
w
0. The proof
is complete. P
Now, we may proceed with the proof of Theorem 8.
Proof of Theorem 8. Since Ces
1
[0, 1] = L
1
(ln1/t ) (with equality of norms) and
L
1
(ln1/t ) is isometric to L
1
, then in the case p = 1 the result follows from the
Szlenk theorem [55]. Therefore, we will consider the case when 1 <p <. Taking
into account Proposition 7 it is enough to prove the following: if {x
n
} Ces
p
, x
n
w

0 and either
(a) {x
n
} consists of elements with equicontinuous norms
or
(b) suppx
n
I
n
, where I
n
1
or
(c) suppx
n
I
n
, where I
n
0,
then there is a subsequence {x

n
} {x
n
} such that
1
m
_
_
_
_
_
m

k=1
x

k
_
_
_
_
_
C(p)
0 as m. (41)
363
Case (a). We will use the following remark from Szlenk paper [55, Remark 1]:
a sequence {x
n
} X, x
n
w
0 in X (X is a Banach space) contains a subsequence
{x

n
} such that
1
m

m
k=1
x

X
0 as m if and only if it contains a subse-
quence {x
n
k
} such that
lim
m
sup
k
1
<<k
m
_
_
_
_
_
1
m
m

i=1
x
n
k
i
_
_
_
_
_
X
=0. (42)
Let {x
n
} Ces
p
, x
n
w
0 and >0. At rst, setting
A
n,m
={t [0, 1]: |x
n
(t )| m}, m, n =1, 2, . . . ,
we prove that
lim
m
sup
nN
m(A
n,m
) =0. (43)
We may assume that x
n

C(p)
=1(n =1, 2, . . .). Therefore,
1 =x
n

C(p)
x
n

C(1)
=x
n

L
1
(ln1/t )
=
1
_
0
|x
n
(t )| ln
1
t
dt

_
A
n,m
|x
n
(t )| ln
1
t
dt m
_
A
n,m
ln
1
t
dt,
i.e.,
_
A
n,m
ln
1
t
dt
1
m
for all n, m N. (44)
Assume that (43) does not hold, that is, there exists a >0 such that for every m N
there is n
m
N such that m(A
n
m
,m
) > . Clearly, we may assume that n
m
as
m. Since
m
_
A
n
m
,m

_
0, 1

2
__
>

2
,
then we have that, for any m N,
_
A
n
m
,m
ln
1
t
dt
_
A
n
m
,m
[0,1

2
]
ln
1
t
dt
ln
2
2
m
_
A
n
m
,m

_
0, 1

2
__
>

2
ln
2
2
.
The last inequality contradicts (44) and, therefore, (43) is proved.
364
We recall that {x
n
} consists of functions having equicontinuous norms. Hence, by
(43), for some m
0
and all n N,
x
n

A
n,m
0

C(p)
<

3
. (45)
Denote y
n
=x
n

A
n,m
0
(n =1, 2, . . .). Then |x
n
(t ) y
n
(t )| m
0
for t [0, 1], so that,
in particular, x
n
y
n
L
p
and x
n
y
n

L
p m
0
. Since L
p
is a reexive space
for 1 < p < and since L
p
has the BanachSaks property (cf. [7, Chapter XII,
Theorem 2]), we may choose an increasing sequence of natural numbers {n
k
}

k=1
such that x
n
k
y
n
k
w
v in L
p
, where v L
p
, and (see (42))
lim
m
sup
k
1
<<k
m
_
_
_
_
_
1
m
m

i=1
(x
n
k
i
y
n
k
i
) v
_
_
_
_
_
L
p
=0. (46)
Using the imbedding L
p
Ces
p
(see Theorem 1(c)) we obtain that x
n
k
y
n
k
w
v in
Ces
p
. Therefore, since x
n
k
w
0 in Ces
p
, we get that y
n
k
w
v in Ces
p
. Moreover,
from(45) it follows that y
n
k

C(p)
<

3
so that v
C(p)

3
. At last, by Theorem1(c)
and (46), for large enough m N,
sup
k
1
<<k
m
_
_
_
_
_
1
m
m

i=1
(x
n
k
i
y
n
k
i
) v
_
_
_
_
_
C(p)
p

sup
k
1
<<k
m
_
_
_
_
_
1
m
m

i=1
(x
n
k
i
y
n
k
i
) v
_
_
_
_
_
L
p

3
.
The last inequalities give us that
sup
k
1
<<k
m
_
_
_
_
_
1
m
m

i=1
x
n
k
i
_
_
_
_
_
C(p)
sup
k
1
<<k
m
__
_
_
_
_
1
m
m

i=1
(x
n
k
i
y
n
k
i
) v
_
_
_
_
_
C(p)
+
1
m
_
_
_
_
_
m

i=1
y
n
k
i
_
_
_
_
_
C(p)
+v
C(p)
_
,
if m is large enough. Thus, {x
n
k
} satises condition (42) and in this case everything
is proved.
Case (b). Let {x
n
} Ces
p
, x
n
w
0, suppx
n
I
n
= [a
n
, b
n
](n = 1, 2, . . .), I
1
<
I
2
< and a
n
1

. We may suppose that x

n
0, x
n

C(p)
=1 and a
1
1/2.
Since p >1, it is enough to show that {x
n
} contains a subsequence (for simplicity,
it will be denoted also by {x
n
}) such that
_
_
_
_
_
n

k=1
x
k
_
_
_
_
_
C(p)
Cn
1/p
, (47)
365
where C > 0 is independent of n N. We will choose x
n
inductively. Suppose
that m 2 and x
1
, x
2
, . . . , x
m1
are already chosen. Then a
1
< b
1
a
m1
<
b
m1
<1 are xed and, since a
n
1

, we may take a
m
so that
1 a
m
(1 b
m1
) 2
p
. (48)
Then for x
m
we take the function corresponding to the interval I
m
=[a
m
, b
m
] (that
is, suppx
m
I
m
). Lets check that inequality (47) holds. For all n N and t (0, 1]
we have that
1
t
t
_
0

k=1
x
k
(s)

ds =
1
t
n

m=1
_
m1

i=1
b
i
_
a
i
x
i
(s) ds +
t
_
a
m
x
m
(s) ds
_

[a
m
,b
m
]
(t )
+
1
t
n

m=1
m

i=1
b
i
_
a
i
x
i
(s) ds
[b
m
,a
m+1
]
(t )
=S
1
(t ) +S
2
(t ) +S
3
(t ),
where a
n+1
=1 and
S
1
(t ) =
1
t
n

m=2
m1

i=1
b
i
_
a
i
x
i
(s) ds
[a
m
,b
m
]
(t ),
S
2
(t ) =
1
t
n

m=1
m

i=1
b
i
_
a
i
x
i
(s) ds
[b
m
,a
m+1
]
(t ),
S
3
(t ) =
1
t
n

m=1
t
_
a
m
x
m
(s) ds
[a
m
,b
m
]
(t ).
Since, by Theorem 3, (Ces
p
[0, 1])

= (Ces
p
[0, 1])

= U(p

) it follows that, for all

i N,
b
i
_
a
i
x
i
(s) ds Ax
i

C(p)

[a
i
,b
i
]

U(p

)
=A
_ b
i
_
0
dt
(1 t )
p

_
1/p

=A(p 1)
1/p

_
1
(1 b
i
)
p

1
1
_
1/p

B
(1 b
i
)
1/p
,
where B >0 depends only on p. Moreover, by (48), for every i =1, 2, . . . , m1,
_
1 a
m
1 b
i
_
1/p

m1

j=i
_
1 a
j+1
1 b
j
_
1/p
2
im
.
366
Therefore,
S
1

p
p
=
n

m=2
_
m1

i=1
b
i
_
a
i
x
i
(s) ds
_
p
b
m
_
a
m
dt
t
p
B
p
n

m=2
_
m1

i=1
(1 b
i
)
1/p
_
p
b
p1
m
a
p1
m
(p 1)a
p1
m
b
p1
m
C
p
1
n

m=2
_
m1

i=1
(1 b
i
)
1/p
_
p
(1 a
m
)
C
p
1
n

m=2
_
m1

i=1
2
im
_
p
C
p
1
n,
so that S
1

p
C
1
n
1/p
, where C
1
>0 depends only on p. Similarly,
S
2

p
p
B
p
n

m=1
_
m

i=1
(1 b
i
)
1/p
_
p
a
p1
m+1
b
p1
m
(p 1)a
p1
m+1
b
p1
m
C
p
2
n

m=1
_
m

i=1
_
1 b
m
1 b
i
_
1/p
_
p
C
p
2
n

m=1
_
1 +
m1

i=1
2
im
_
p
(2C
2
)
p
n,
which implies that S
2

p
2C
2
n
1/p
, where C
2
>0 depends only on p. Finally, it is
easy to see that
S
3

_
n

m=1
x
m

p
C(p)
_
1/p
=n
1/p
.
Thus, combining the estimates of S
1
, S
2
and S
3
we get
_
_
_
_
_
n

k=1
x
k
_
_
_
_
_
C(p)

k=1
S
k

p
(1 +C
1
+2C
2
) n
1/p
,
where C :=1 +C
1
+2C
2
is independent of n N, that is, inequality (47) is proved.
Case (c). Let {x
n
} Ces
p
, x
n
w
0, suppx
n
I
n
= [a
n
, b
n
] (n = 1, 2, . . .), I
1
>
I
2
> and b
n
0
+
. Again we may assume that x
n
0, x
n

C(p)
= 1 and b
1

1/2. As in the case (b) it is enough to prove inequality (47) for some subsequence
of {x
n
} (it will be denoted also by {x
n
}), which will be chosen inductively.
Suppose that m 2 and the functions x
1
, x
2
, . . . , x
m1
are chosen. Then b
1
>
a
1
b
m1
>a
m1
>0 are xed and, since b
n
0
+
, we may take b
m
so that
b
m
2
p

a
m1
. (49)
367
Let us show that the corresponding subsequence {x
n
} satises inequality (47). For
any n N and t (0, 1] we have that
1
t
t
_
0

k=1
x
k
(s)

ds
=
1
t
n

j=1
_
n

i=nj+2
b
i
_
a
i
x
i
(s) ds +
t
_
a
nj+1
x
nj+1
(s) ds
_

[a
nj+1
,b
nj+1
]
(t )
+
1
t
n

j=1
n

i=nj+1
b
i
_
a
i
x
i
(s) ds
[b
nj+1
,a
nj
]
(t )
=T
1
(t ) +T
2
(t ) +T
3
(t ),
where a
0
=1 and
T
1
(t ) =
1
t
n

j=2
n

i=nj+2
b
i
_
a
i
x
i
(s)] ds
[a
nj+1
,b
nj+1
]
(t ),
T
2
(t ) =
1
t
n

j=1
n

i=nj+1
b
i
_
a
i
x
i
(s) ds
[b
nj+1
,a
nj
]
(t ),
T
3
(t ) =
1
t
n

j=1
t
_
a
nj+1
x
nj+1
(s) ds
[a
nj+1
,b
nj+1
]
(t ).
Using again the duality result, as in the proof of (b), we nd that
b
i
_
a
i
x
i
(s) ds Ax
i

C(p)

[a
i
,b
i
]

U(p

)
=A(p 1)
1/p

_
1 (1 b
i
)
p

1
(1 b
i
)
p

1
_
1/p

b
1/p

i
(i =1, 2, . . .),
where B

>0 depends only on p. Since, by (49), for any k <i

_
b
i
a
k
_
1/p

m=k
_
b
m
a
m1
_
1/p

2
ki1
,
368
then
T
1

p
p
=
n

j=2
_
n

i=nj+2
b
i
_
a
i
x
i
(s) ds
_
p
b
nj+1
_
a
nj+1
dt
t
p

_
B

_
p
n

j=2
_
n

i=nj+2
b
1/p

i
_
p
b
p1
nj+1
a
p1
nj+1
(p 1)a
p1
nj+1
b
p1
nj+1
B
p
1
n

j=2
_
n

i=nj+2
b
1/p

i
_
p
a
1p
nj+1
B
p
1
n

j=2
_
n

i=nj+2
2
nij+1
_
p
B
1
n,
so that T
1

p
B
1
n
1/p
, where B
1
>0 depends only on p. Similarly,
T
2

p
p

_
B

_
p
n

j=1
_
n

i=nj+1
b
1/p

i
_
p
a
p1
nj
b
p1
nj+1
(p 1)a
p1
nj
b
p1
nj+1
B
p
2
n

j=1
_
n

i=nj+1
_
b
i
b
nj+1
_
1/p

_
p
B
p
2
n

j=1
_
1 +
n

i=nj+2
2
nij+1
_
p
(2B
2
)
p
n.
Hence, T
2

p
2B
2
n
1/p
, where B
2
> 0 depends only on p. Moreover, it is clear
that
T
3

_
n

j=1
x
j

p
C(p)
_
1/p
=n
1/p
.
Thus, combining the estimates of T
1
, T
2
and T
3
we get that
_
_
_
_
_
n

k=1
x
k
_
_
_
_
_
C(p)

k=1
T
k

p
(1 +B
1
+2B
2
) n
1/p
,
where B := 1 + B
1
+ 2B
2
is independent of n N. Since all cases (a)(c) are
examined, the theorem is proved. P
8. THE CESRO FUNCTION SPACES Ces
p
[0, ) AND Ces
p
[0, 1] ARE ISOMORPHIC FOR
1 <p
The main result in this Section is a construction of an isomorphism between the
Cesro function spaces Ces
p
[0, ) and Ces
p
[0, 1] for 1 <p .
369
Theorem 9. If 1 < p , then the Cesro function spaces Ces
p
[0, 1] and
Ces
p
[0, ) are isomorphic.
Proof. The proof will go in two parts. Let 1 < p <. Sy, Zhang and Lee proved
in [54] that the norm in Ces
p
[0, ) is equivalent to the functional
f
0
=
_

n=1
_
1
n
n

k=1
s
k
(f )
_
p
+

m=1
_
m

k=m
t
k
(f )
_
p
m
2
_
1/p
, (50)
where
s
k
(f ) =
k+1
_
k
|f (s)| ds and t
k
(f ) =
1
k
_
1
k+1
|f (s)| ds, k =1, 2, . . . .
Lets prove the analogous assertion for the space Ces
p
[0, 1]. At rst, if
1
m+1
x
1
m
, m=1, 2, . . . , then
m+1
2
1/(m+1)
_
0
|f |
1
x
x
_
0
|f | (m+1)
x
_
0
|f | 2m
1/m
_
0
|f |.
Therefore,
2
p

m=1
_
(m+1)
1/(m+1)
_
0
|f |
_
p_
1
m

1
m+1
_

m=1
1/m
_
1/(m+1)
_
1
x
x
_
0
|f |
_
p
dx
=
1
_
0
_
1
x
x
_
0
|f |
_
p
dx
and
1/2
_
0
_
1
x
x
_
0
|f |
_
p
dx =

m=2
1/m
_
1/(m+1)
_
1
x
x
_
0
|f |
_
p
dx
2
p

m=2
_
m
1/m
_
0
|f |
_
p_
1
m

1
m+1
_
.
370
The rst of these inequalities implies that
2
p

m=2
_
m
1/m
_
0
|f |
_
p_
1
m

1
m+1
_
=2
p

m=1
_
(m+1)
1/(m+1)
_
0
|f |
_
p_
1
m+1

1
m+2
_

1
_
0
_
1
x
x
_
0
|f |
_
p
dx =f
p
C(p)
,
and the second one yields that
2
p
1/2
_
0
_
1
x
x
_
0
|f |
_
p
dx

m=2
_
m
1/m
_
0
|f |
_
p_
1
m

1
m+1
_
2
p
f
p
C(p)
. (51)
Denote

n
=
1
2
_
2 n
1p
_
, n =1, 2, . . . .
It is easy to check that
1
2
=
1

n
<
n+1
2
n
, n = 1, 2, . . . , and
n
1 as
n . Thus, if
n
x
n+1
, then
1
2
n

n
_
0
|f |
1

n+1

n
_
0
|f |
1
x
x
_
0
|f |

n+1
_
0
|f |
2

n+1

n+1
_
0
|f |,
which implies that
2
p

n=1
_
1

n
_
0
|f |
_
p
(
n+1

n
) (52)

n=1

n+1
_

n
_
1
x
x
_
0
|f |
_
p
dx
=
1
_
1/2
_
1
x
x
_
0
|f |
_
p
dx
371
and
1
_
1/2
_
1
x
x
_
0
|f |
_
p
dx =

n=1

n+1
_

n
_
1
x
x
_
0
|f |
_
p
dx (53)
2
p

n=1
_
1

n+1

n+1
_
0
|f |
_
p
(
n+1

n
).
Moreover, since

n+1

n
=
1
2
_
1
n
p1

1
(n +1)
p1
_
=
p 1
2
n+1
_
n
1
t
p
dt
and
1
n
p

_
n+1
n
1
t
p
dt
1
(n+1)
p

1
(2n)
p
=2
p 1
n
p
, it follows that
p 1
2
p+1
n
p

n+1

p 1
2n
p
, n =1, 2, . . . , (54)
and we conclude that

n+1

p 1
2n
p
4
p
p 1
2
p+1
(n +1)
p
4
p
(
n+2

n+1
), n =1, 2, . . . .
Therefore,

n=1
_
1

n
_
0
|f |
_
p
(
n+1

n
)
=
_
1

1
_
0
|f |
_
p
(
2

1
) +

n=1
_
1

n+1

n+1
_
0
|f |
_
p
(
n+2

n+1
)
4
p

n=1
_
1

n+1

n+1
_
0
|f |
_
p
(
n+1

n
).
By combining the last inequality with (52) we obtain that

n=1
_
1

n+1

n+1
_
0
|f |
_
p
(
n+1

n
) 8
p
1
_
1/2
_
1
x
x
_
0
|f |
_
p
dx. (55)
372
From (51), (53) and (55) it follows that
f
p
C(p)
2
p

m=2
_
m
1/m
_
0
|f |
_
p_
1
m

1
m+1
_
+2
p

n=1
_
1

n+1

n+1
_
0
|f |
_
p
(
n+1

n
)
and

m=2
_
m
1/m
_
0
|f |
_
p_
1
m

1
m+1
_
+

n=1
_
1

n+1

n+1
_
0
|f |
_
p
(
n+1

n
)
2
p
1
_
0
_
1
x
x
_
0
|f |
_
p
dx +8
p
1
_
1/2
_
1
x
x
_
0
|f |
_
p
dx
2
p
_
4
p
+1
_
f
p
C(p)
.
Thus, taking into account (54), we obtain that
f
C(p)

n=1
_
1
n
n+1

n+1
_
0
|f (t )| dt
_
p
+

m=2
_
m
1/m
_
0
|f (t )| dt
_
p
m
2
_
1/p
. (56)
Note that
1/m
_
0
|f (t )| dt =

k=m
t
k
(f ),
where t
k
(f ) =
1/k
_
1/(k+1)
|f (t )| dt,
and

n+1
_
0
|f (t )| dt =
1/2
_
0
|f (t )| dt +
n

k=1
b
k
(f ),
where b
k
(f ) =

k+1
_

k
|f (t )| dt.
373
Since
n

1
2
(n =1, 2, . . .) it follows that the rst sum in (56) does not exceed
2
p

n=1
_ 1/2
_
0
|f (t )| dt +
n

k=1
b
k
(f )
_
p
n
p
2
2p
_

n=1
_
1
n
n

k=1
b
k
(f )
_
p
+
_ 1/2
_
0
|f (t )| dt
_
p

n=1
n
p
_
.
Because p > 1 and the second sum on the right-hand side of (56) contains
(
_
1/2
0
|f (t )| dt )
p
, then
f
C(p)

n=1
_
1
n
n

k=1
b
k
(f )
_
p
+

m=2
_
m

k=m
t
k
(f )
_
p
m
2
_
1/p
. (57)
Denote by k
n
and l
m
one-to-one afne mappings such that
k
n
: [n, n +1] [
n
,
n+1
],
(n, m=1, 2, . . .)
l
m
:
_
1
m+1
,
1
m
_

_
1
m+2
,
1
m+1
_
and dene the linear operator T for f Ces
p
[0, 1] by
Tf (x) =

n=1
(
n+1

n
)f
_
k
n
(x)
_

[n,n+1]
(x)
+

m=1
f
_
l
m
(x)
_

[
1
m+1
,
1
m
]
(x).
Since
n+1
_
n

f
_
k
n
(x)
_

dx =
1

n+1

n+1
_

n
|f (t )| dt
and
1/m
_
1/(m+1)

f
_
l
m
(x)
_

dx =
m+2
m
1/(m+1)
_
1/(m+2)
|f (t )| dt
for n, m=1, 2, . . . , then the equivalences (50) and (57) show that
Tf
C
p
[0,)
Tf
0
=
_

n=1
_
1
n
n

k=1
s
k
(Tf )
_
p
+

m=1
_
m

k=m
t
k
(Tf )
_
p
m
2
_
1/p
374

n=1
_
1
n
n

k=1
b
k
(f )
_
p
+

m=2
_
m

k=m
t
k
(f )
_
p
m
2
_
1/p
f
C
p
[0,1]
.
Therefore, T : Ces
p
[0, 1] Ces
p
[0, ) is an isomorphism and the proof for 1 <
p < is complete.
If p = the construction of isomorphism will be different and the proof
is even working for the p-convexications, that is, if 1 p < , then the
spaces Ces
(p)

[0, 1] and Ces

(p)

[0, ) are isomorphic. In particular, Ces

[0, 1] and
Ces

[0, ) are isomorphic.

It is easy to check that
f
C()
(p)
[0,)
sup
kZ
_
2
k+1
_
{2
k1
<t 2
k
}
|f (t )|
p
dt
_
1/p
(58)
and
f
C()
(p)
[0,1]
sup
k=0,1,2,...
_
2
k+1
_
{2
k1
<t 2
k
}
|f (t )|
p
dt
_
1/p
. (59)
Moreover, for every k Z,
2
k+1
2
k
_
2
k1
|f (t )|
p
dt =
1
_
0

f
_
2
k1
(t +1)
_

p
dt. (60)
Dene the linear transforms
T
1
: Ces

[0, ) l

k=
L
p
[0, 1]
_
, T
1
f =
_
f
_
2
k1
(t +1)
_
kZ
_
and
T
2
: Ces

[0, 1] l

k=0
L
p
[0, 1]
_
, T
2
f =
_
f
_
2
k1
(t +1)
_

k=0
_
.
Formulas (58)(60) show that T
1
and T
2
are isomorphisms. It is obvious that the
spaces l

k=
L
p
[0, 1]) and l

k=0
L
p
[0, 1]) are isomorphic. There-
fore, the spaces Ces
(p)

[0, ) and Ces

(p)

[0, 1] are isomorphic. P

Problem 1. Is the Cesro function space Ces

(I) isomorphic with the Cesro

sequence space ces

?
375
9. A DESCRIPTION OF ISOMORPHIC COPIES OF l
q
IN Ces
p
[0, 1]
In Theorem 6 we proved that Ces
p
[0, 1] contains an isomorphic copy of l
p
. Now
we try to investigate when this is true for the spaces l
q
.
Theorem 10.
(a) If 1 p 2, then the space l
q
is embedded isomorphically into Ces
p
[0, 1] if
and only if q [1, 2].
(b) If 2 < p <, then the space l
q
is embedded isomorphically into Ces
p
[0, 1] if
and only if either q [1, 2] or q =p.
Proof. Firstly, Ces
p
[0, 1] contains a copy of L
1
[0, 1] (cf. [4, Lemma 1]) and in turn
l
q
is embedded into L
1
[0, 1] if 1 q 2 (cf. [1, Theorem 6.4.18]). Moreover, by
Theorem 6, l
p
is embedded into Ces
p
[0, 1] for every p [1, ) so we have to prove
only the necessity.
In the case when 1 p 2 necessity is obvious as a consequence of the fact that
Ces
p
[0, 1] has cotype 2.
If p >2 noting that Ces
p
[0, 1] Ces
1
[0, 1] =L
1
(ln1/t ) we consider two cases:
(a) Assume that the norms of the spaces Ces
p
[0, 1] and L
1
(ln1/t ) are equivalent
on a subspace X Ces
p
[0, 1] which is isomorphic to l
q
. In other words, X is a
subspace of L
1
(ln1/t ). Since the last space has cotype 2, then q 2.
(b) The norms of the spaces Ces
p
[0, 1] and L
1
(ln1/t ) are not equivalent on X
l
q
. Then there is a sequence {x
n
} X such that x
n

C(p)
=1 and x
n

L
1
(ln1/t )
0.
In particular, x
n
0 weakly in L
1
(ln1/t ), i.e.,
1
_
0
x
n
(t )y(t ) dt 0 for every y L

_
ln
1
1/t
_
.
Denote F :=

0<<1
L

[0, ]. Obviously, it yields that F L

(ln
1
1/t ) and F is
dense in (Ces
p
[0, 1])

=U(p

) (see Theorem 3). Therefore, x

n

C(p)
=1 and x
n

0 weakly in Ces
p
[0, 1]. By a known result (cf. [37, Proposition 1.a.12]) there exists
a subsequence {x

n
} {x
n
} which is equivalent to a seminormalized block basis of
the canonical basis of l
q
and, consequently, is equivalent to the canonical basis of l
q
itself (see [1, Lemma 2.1.1 and Remark 2.1.2]). Moreover, x

C(p)
=1 and x

n
0
in the Lebesgue measure m. Next, since Ces
p
[0, 1] is separable for 1 p <, then
applying the KadecPeczy nski procedure we may nd a subsequence {x

n
} {x

n
}
and a sequence of disjoint sets A
n
[0, 1] such that x

n
x

A
n

C(p)
0. Using
a standard argument we can select a subsequence {x

n
k
} {x

n
}, which is equivalent
to the sequence of disjoint functions z
k
:=x

n
k

A
n
k
. Note that {x

n
k
} and {z
k
} as well
are equivalent to the canonical basis of l
q
. To show that either q [1, 2] or q = p
we consider separately two cases:
(1) rstly, assume that there is h (0,
1
2
) such that suppz
k
[h, 1 h] for all
k = 1, 2, . . . . Since Ces
p
[h, 1 h] L
1
[h, 1 h] (cf. [4, Lemma 1]), then l
q
will
be embedded into L
1
[h, 1 h] L
1
[0, 1], so that q [1, 2].
376
(2) otherwise, there is a subsequence {z

k
} {z
k
} such that suppz
k
I
k
for some
intervals I
k
satisfying either I
k
0 or I
k
1. Then, using the same arguments as
in the proof of Theorem 8, we may select a subsequence {z

k
} {z

k
} such that
_
_
_
_
_
m

k=1
z

k
_
_
_
_
_
C(p)
Cm
1/p
,
where the constant C >0 does not depend on m=1, 2, . . . . Since [z

k
] l
q
, then we
have q p. On the other hand, q p because Ces
p
[0, 1] has cotype p, thus q =p
and the proof is complete. P
Let us remind that L
p
[0, 1] contains an isomorphic copy of l
q
if and only if
q [p, 2] for the case 1 p 2 and in the case when p >2 this can be when either
q = p or q = 2. We can see then the difference between L
p
[0, 1] and Ces
p
[0, 1]
spaces. In particular, if 1 <p <, then Ces
p
[0, 1] contains an isomorphic copy of
l
1
but not L
p
[0, 1].
ACKNOWLEDGEMENT
We would like to thank referee for some useful corrections in this paper.
REFERENCES
[1] Albiac F., Kalton N.J. Topics in Banach Space Theory, Graduate Texts in Mathematics, vol. 233,
Springer-Verlag, New York, 2006.
[2] Aliprantis C.D., Burkinshaw O. Positive Operators, Academic Press, Orlando, 1985.
[3] Alspach D.E. A xed point free nonexpansive map, Proc. Amer. Math. Soc. 82 (1981) 423424.
[4] Astashkin S.V., Maligranda L. Cesro function spaces fail the xed point property, Proc. Amer.
Math. Soc. 136 (12) (2008) 42894294.
[5] Astashkin S.V., Maligranda L. Rademacher functions in Cesro type spaces, Studia Math., to
appear.
[6] Astashkin S.V., Sukochev F.A. BanachSaks property in Marcinkiewicz spaces, J. Math. Anal.
Appl. 336 (2) (2007) 12311258.
[7] Banach S. Thorie des Oprations Linaires, Monograe Matematyczne, vol. 1, PWN, Warsaw,
1932, Reprinted in: S. Banach, Oeuvres, Vol. II, PWN, Warszawa, 1979, 13219; English
transl.: Theory of Linear Operations, North-Holland Math. Library, vol. 38, North-Holland,
Amsterdam, 1987.
[8] Bennett G. Factorizing the Classical Inequalities, Mem. Amer. Math. Soc., vol. 120, American
Mathematical Society, Providence, RI, 1996.
[9] Bennett C., Sharpley R. Interpolation of Operators, Academic Press, Boston, MA, 1988.
[10] Chen S., Cui Y., Hudzik H., Sims B. Geometric properties related to xed point theory in some
Banach function lattices, in: Handbook on Metric Fixed Point Theory, Kluwer Acad. Publ.,
Dordrecht, 2001, pp. 339389.
[11] Cui Y., Hudzik H. Some geometric properties related to xed point theory in Cesro sequence
spaces, Collect. Math. 50 (1999) 277288.
[12] Cui Y., Hudzik H. On the BanachSaks and weak BanachSaks properties of some Banach
sequence spaces, Acta Sci. Math. (Szeged) 65 (1999) 179187.
[13] Cui Y., Hudzik H. Packing constant for Cesro sequence spaces, Nonlinear Anal. 47 (2001)
26952702.
377
[14] Cui Y., Hudzik H., Li Y. On the GarcaFalset coefcient in some Banach sequence spaces,
in: Function Spaces, Proc. of the 5th International Conf. in Pozna n (Aug. 28Sept. 3, 1998),
Marcel Dekker, New York, 2000, pp. 141148.
[15] Cui Y., Jie L., Puciennik R. Local uniform nonsquareness in Cesro sequence spaces,
Commentat. Math. Prace Mat. 37 (1997) 4758.
[16] Cui Y., Meng C., Puciennik R. BanachSaks property and property () in Cesro sequence
spaces, Southeast Asian Bull. Math. 24 (2000) 201210.
[17] Diestel J., Jarchow H., Tonge A. Absolutely Summing Operators, Cambridge Univ. Press,
Cambridge, 1995.
[18] Dilworth S., Girardi M., Hagler J. Dual Banach spaces which contain an isometric copy of L
1
,
Bull. Polish Acad. Sci. Math. 48 (1) (2000) 112.
[19] Dodds P.G., Semenov E.M., Sukochev F.A. The BanachSaks property in rearrangement invariant
spaces, Studia Math. 162 (3) (2004) 263294.
[20] Dowling P.N., Lennard C.J. Every nonreexive subspace of L
1
[0, 1] fails the xed point property,
Proc. Amer. Math. Soc. 125 (2) (1997) 443446.
[21] Dowling P.N., Lennard C.J., Turett B. Renormings of l
1
and c
0
and xed point properties, in:
Handbook of Metric Fixed Point Theory, Kluwer Acad. Publ., Dordrecht, 2001, pp. 269297.
[22] Edwards R.E. Functional Analysis. Theory and Applications, Holt Rinehart and Winston, New
York, 1965.
[23] Fabian M., Habala P., Hjek P., Montesinos Santaluca V., Pelant J., Zizler V. Functional Analysis
and Innite-Dimensional Geometry, CMS Books in Math, vol. 8, Springer-Verlag, New York,
2001.
[24] Hardy G.H., Littlewood J.E., Plya G. Inequalities, Cambridge Univ. Press, Cambridge, 1952.
[25] Hassard B.D., Hussein D.A. On Cesro function spaces, Tamkang J. Math. 4 (1973) 1925.
[26] Jagers A.A. A note on Cesro sequence spaces, Nieuw Arch. Wiskund. (3) 22 (1974) 113124.
[27] Kadec M.I., Peczy nski A. Bases, lacunary sequences and complemented subspaces in the spaces
L
p
, Studia Math. 21 (1962) 161176.
[28] Kantorovich L.V., Akilov G.P. Functional Analysis, Nauka, Moscow, 1977. English transl.:
Pergamon Press, Oxford, 1982.
[29] Kashin B.S., Saakyan A.A. Orthogonal Series, Nauka, Moscow, 1984. English transl.: American
Mathematical Society, Providence, RI, 1989.
[30] Kerman R., Milman M., Sinnamon G. On the BrudnyKrugljak duality theory of spaces formed
by the K-method of interpolation, Rev. Mat. Complut. 20 (2) (2007) 367389.
[31] Korenblyum B.I., Kren S.G., Levin B.Ya. On certain nonlinear questions of the theory of singular
integrals, Doklady Akad. Nauk SSSR (N.S.) 62 (1948) 1720 (in Russian).
[32] Krein S.G., Petunin Yu.I., Semenov E.M. Interpolation of Linear Operators, Nauka, Moscow,
1978. English transl.: American Mathematical Society, Providence, RI, 1982.
[33] Kufner A., Maligranda L., Persson L.-E. The Hardy Inequality. About Its History and Some
Related Results, Vydavatelski Servis Publishing House, Pilzen, 2007.
[34] Lee P.Y. Cesro sequence spaces, Math. Chronicle, New Zealand 13 (1984) 2945.
[35] Lee P.Y. Cesro sequence spaces, Manuscript, Singapore, 1999, pp. 117.
[36] Leibowitz G.M. A note on the Cesro sequence spaces, Tamkang J. Math. 2 (1971) 151157.
[37] Lindenstrauss J., Tzafriri L. Classical Banach Spaces, I. Sequence Spaces, Springer-Verlag,
Berlin, 1977.
[38] Lindenstrauss J., Tzafriri L. Classical Banach Spaces, II. Function Spaces, Springer-Verlag,
Berlin, 1979.
[39] Liu Y.Q., Wu B.E., Lee P.Y. Method of Sequence Spaces, Guangdong of Science and Technology
Press, 1996 (in Chinese).
[40] Lozanovski G.Ja. Isomorphic Banach lattices, Sibirsk. Mat. Zh. 10 (1969) 9398. English transl.:
Siberian Math. J. 10 (1) (1969) 6468.
[41] Lozanovski G.Ja. Certain Banach lattices, Sibirsk. Mat. Zh. 10 (1969) 584599. English transl.:
Siberian Math. J. 10 (3) (1969) 419431.
[42] Luxemburg W.A.J., Zaanen A.C. Some examples of normed Kthe spaces, Math. Ann. 162
(1965/1966) 337350.
378
[43] Maligranda L. Orlicz Spaces and Interpolation, Seminars in Math., vol. 5, Univ. of Campinas,
Campinas, SP, Brazil, 1989.
[44] Maligranda L. Type, cotype and convexity properties of quasi-Banach spaces, in: M. Kato,
L. Maligranda (Eds.), Banach and Function Spaces, Proc. of the Internat. Symp. on Banach
and Function Spaces (Oct. 24, 2003, Kitakyushu-Japan), Yokohama Publishers, 2004,
pp. 83120.
[45] Maligranda L., Petrot N., Suantai S. On the James constant and B-convexity of Cesro and
CesroOrlicz sequence spaces, J. Math. Anal. Appl. 326 (1) (2007) 312331.
[46] Novikov S.Ya., Semenov E.M., Tokarev E.V. The structure of subspaces of the space
p
(),
Dokl. Akad. Nauk SSSR 247 (1979) 552554. English transl.: Sov. Math. Dokl. 20 (1979)
760761.
[47] Novikov S.Ya., Semenov E.M., Tokarev E.V. On the structure of subspaces of the spaces
p
(),
Teor. Funkts., Funkts. Anal. Prilozh. 42 (1984) 9197. English transl.: Ser. 2, Amer. Math.
Soc. 136 (1987) 121127.
[48] Rakov S.A. The BanachSaks exponent of some Banach spaces of sequences, Mat.
Zametki 32 (5) (1982) 613625. English transl.: Math. Notes 32 (5, 6) (1982) 791797.
[49] Sawyer E. Boundedness of classical operators on classical Lorentz spaces, Studia Math. 96 (2)
(1990) 145158.
[50] Shiue J.S. Cesro sequence spaces, Tamkang J. Math. 1 (1970) 1925.
[51] Shiue J.S. A note on Cesro function space, Tamkang J. Math. 1 (1970) 9195.
[52] Sinnamon G. The level function in rearrangement invariant spaces, Publ. Mat. 45 (1) (2001)
175198.
[53] Sinnamon G. Transferring monotonicity in weighted norm inequalities, Collect. Math. 54 (2)
(2003) 181216.
[54] Sy P.W., Zhang W.Y., Lee P.Y. The dual of Cesro function spaces, Glas. Mat. Ser. III 22 (1)
(1987) 103112.
[55] Szlenk W. Sur les suites faiblement convergentes dans lespace L, Studia Math. 25 (1965) 337
341.
[56] Tandori K. ber einen speziellen Banachschen Raum, Publ. Math. Debrecen 3 (1954) 263268,
(1955).
[57] Tokarev E.V. The BanachSaks property in Banach lattices, Sibirsk. Mat. Zh. 24 (1) (1983)
187189 (in Russian).
[58] Wnuk W. l
(p
n
)
spaces with the DunfordPettis property, Comment. Math. Prace Mat. 30 (2)
(1991) 483489.
[59] Wnuk W. Banach Lattices with Order Continuous Norms, Polish Scientic Publishers PWN,
Warszawa, 1999.
(Received August 2009)
379