J. Math. Anal. Appl.

274 (2002) 626–644

www.academicpress.com

Commutators of Hardy operators ✩

Long Shunchao ∗ and Wang Jian
Department of Mathematics, Xiangtan University, Hunan 411105, PR China
Received 11 December 2000
Submitted by L. Chen

Abstract
In this paper we establish Hardy’s integral inequalities for commutators generating
Hardy operators with one sided dyadic CMO functions.
 2002 Elsevier Science (USA). All rights reserved.

Keywords: Hardy’s integral inequality; Hardy operator; Commutator, CMO; One sided CMO;
Dyadic CMO

1. Introduction

The Hardy operators are defined by

x
1
Hf (x) = f (t) dt (x > 0),
x
0

and (dual form)

∞
∗
H f (x) = f (t) dt (x > 0).
x

✩
Supported by the Foundation of Province Educational Committee of Hunan (99B03).
* Corresponding author.
E-mail addresses: jxtu@xtu.edu.cn (L. Shunchao), jwang@hnedu.com (W. Jian).

0022-247X/02/$ – see front matter  2002 Elsevier Science (USA). All rights reserved.
PII: S 0 0 2 2 - 2 4 7 X ( 0 2 ) 0 0 3 2 1 - 9
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 627

Hardy [11] established the Hardy integral inequality

∞
∞
 p 
p
p
Hf (x) dx
p f (x) dx, f (x)  0, p > 1,
0 0

and generalized this result himself [12]:

∞
∞
 ∗
p  p
H f (x) dx
p p
xf (x) dx, f (x)  0, p > 1,
0 0

p
where = p/(p − 1).
The two inequalities above went by the name of Hardy’s integral inequality. In
view of the usefulness of these inequalities in analysis and their applications, they
have received considerable attention and a number of papers have appeared which
deal with its alternative proofs, various generalizations, and numerous variants
and applications. For the earlier development of this kind of inequality and many
important applications in analysis, see [13]. Among numerous papers dealing with
such inequalities, we choose to refer to the papers [1,2,4,7,8,11–14,16–25] and
some of the references cited therein.
In this paper, our main objective is establishing the Hardy’s integral inequality
for commutators generating Hardy operators and fractional order Hardy operators
with one sided dyadic CMO functions.
Let b(x) be a real measurable, locally integrable function, define the commu-
tators as follows:

x
1  
Hb f (x) = b(x) − b(t) f (t) dt (x > 0),
x
0

and

∞
 
Hb∗ f (x) = b(x) − b(t) f (t) dt (x > 0).
x

The space CMOp , 1
p < ∞, of functions of finite central mean oscillation

was introduced by Chen and Lau [5], Garcia-Cuerva [9], Lakey [15], and Garcia-
Cuerva and Herrero [10]. CMOp spaces bears a simple relationship with BMO,
the space of functions of bounded mean oscillation: g ∈ BMO precisely when g
and all of its translates belong to CMOp uniformly a.e. Many precise analogies
exist between CMOp and BMO from the point of view of real Hardy spaces,
for example, CMO1 is the dual of the Beurling–Hardy spaces HA1 , the duality is
then analogous to the H 1 ↔ BMO (see [5,9]), and CMOp is the dual of the Herz–
Hardy spaces HAp , 1
p < ∞ (see [10]). Here we find the boundedness of the
commutator generating Hardy operators with one sided dyadic CMOp functions
628 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

on Lp (R + ) is just the same as that of the commutators generating the Calderon–

Zygmund operators with BMO functions on Lp (R n ).
Let 1
p < ∞, we say b is a one sided dyadic CMOp function, if

2j 1/p
1
sup j |b(t) − b(0,2j ] |p dt = bCMOp < ∞,
+
j ∈Z 2
0

where

2j
1
b(0,2j ] = j b(t) dt,
2
0
p q
we then say that b ∈ CMO+ . Obviously, if 1
p < q < ∞, then CMO+ ⊂
p p
CMO+ , and bCMOp
bCMOq , and the classical BMO space ⊂ CMO+ for
+ +
p
all 1
p < ∞. Simply by CMOp , denote CMO+ below.
We obtain the following theorems:

Theorem 1. Let p > 1, b ∈ CMOmax(p,p ) , then

∞
∞
 p
|Hb f (x)| dx
C bCMOmax(p,p )
p
|f (x)|p dx, (1)
0 0

where
  p/p 
  
C = 22p/p +p 1 + (i − k)p 2(k−i)/2 2(k−i)p/(2p ) ,
i−k
0 i−k
0

and

∞
∞
 p
|Hb∗ f (x)|p dx
C bCMOmax(p,p ) |xf (x)|p dx, (2)
0 0

where
  p/p 
  
C = 22p/p +p 1 + (k − i)p 2−(k−i)p /(2p) 2−(k−i)/2.
k−i
0 k−i
0

Let 0 < α < 1, define the fractional order Hardy operator as follows:

x
1
H f (x) =
α
f (t) dt (x > 0),
x 1−α
0
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 629

and the commutator as follows:

x
1  
Hbα f (x) = b(x) − b(t) f (t) dt (x > 0).
x 1−α
0

Theorem 2. Let 0 < α < 1, 1 < p0 < p < ∞, 1/p = 1/p0 − α, b ∈


CMOmax(p,p0 ) , then

∞ 1/p 
∞ 1/p0
|Hb f (x)| dx
α p

Cb max(p,p ) |f (x)| 0 dx
p
, (3)
0 CMO
0 0

where
  1/p
   
C = 3 × 22/p +1 (i − k)p 2(k−i)p /(2p0 )
i−k
0
  1/p
(k−i)p/(2p0 )
× 2 ,
i−k
0

and

∞ 1/p
|Hb∗ f (x)|p dx
0

∞ 1/p0

Cb max(p,p ) |x 1−α
f (x)| p0
dx , (4)
CMO 0
0

where
  1/p
  
C = 3 × 22/p +1 (k − i)p 2−(k−i)p /(2p)
k−i
0
  1/p
−(k−i)/2
× 2 .
k−i
0

The commutators Tb generating Calderon–Zygmund operators T with BMO

function b have been studied by many authors, for example [3,6,26]; in [3,26],
authors have used the sharp function. In this paper, we consider the commutators
generating Hardy operators and fractional order Hardy operators with one sided
dyadic CMO functions, a straightforward proof is given that shows that, for Hardy
operators H , the commutators (Hb ) theorem holds for more general functions b.
630 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

2. Proofs of theorems

Firstly, let us prove two lemmas.

Lemma 1. Let l  0, s > 1, p  r > 1, then

i
∞ 
 2
k+1 1/r p


(i) (i − k)l 2(k−i)/s |f (t)|r dt

i=−∞ k=−∞
2k
  p/p
 

(i − k)lp 2(k−i)p /(2s)
i−k
0

∞ p/r

× 2 (k−i)p/(2s)
|f (t)| dt
r

i−k
0 0
and
∞
∞ 
 2
k+1 1/r p


(ii) (k − i)l 2−(k−i)/s |f (t)|r dt

i=−∞ k=i
2k
  p/p
lp  −(k−i)p  /(2s)

(k − i) 2
k−i
0

∞ p/r

−(k−i)p/(2s)
× 2 |f (t)| dt
r
.
k−i
0 0

Proof. (i) We have

i
∞ 
 2
k+1 1/r p


(i − k)l 2(k−i)/s |f (t)|r dt

i=−∞ k=−∞
2k

∞
  2
k+1 p/r 
 
i

2 (k−i)p/(2s)
|f (t)| dt
r

i=−∞ k=−∞
2k
 p/p

i
lp  (k−i)p  /(2s)
× (i − k) 2 (by Hölder’s inequality)
k=−∞
  p/p
 
= (i − k)lp 2(k−i)p /(2s)
i−k
0
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 631

∞
 2
k+1 p/r
 
× 2 (k−i)p/(2s)
|f (t)| dt
r

k=−∞ i−k
0
2k
  p/p
 
= (i − k)lp 2(k−i)p /(2s)
i−k
0

∞
 2
k+1 p/r
 
× 2(k−i)p/(2s) |f (t)| dt
r

i−k
0 k=−∞
2k
  p/p
 

(i − k)lp 2(k−i)p /(2s)
i−k
0

∞ p/r

× 2 (k−i)p/(2s)
|f (t)| dt r
(since p  r).
i−k
0 0

(ii) We have
∞
∞ 
 2
k+1 1/r p

l −(k−i)/s
(k − i) 2 |f (t)| dt
r


i=−∞ k=i
2k

∞
 ∞
 2
k+1 p/r 
 

2−(k−i)p/(2s) |f (t)|r dt
i=−∞ k=i
2k
 ∞
p/p
  
× (k − i)lp 2−(k−i)p /(2s) (by Hölder’s inequality)
k=i
  p/p
 
= (k − i)lp 2−(k−i)p /(2s)
k−i
0

∞
 2
k+1 p/r
 
−(k−i)p/(2s)
× 2 |f (t)| dt
r

k=−∞ k−i
0
2k
  p/p 
 
= (k − i)lp 2−(k−i)p /(2s) 2−(k−i)p/(2s)
k−i
0 k−i
0

∞

2k+1 p/r

× |f (t)| dt
r

k=−∞
2k
632 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

  p/p 
lp  −(k−i)p  /(2s)

(k − i) 2 2−(k−i)p/(2s)
k−i
0 k−i
0

∞ p/r
× |f (t)|r dt (since p  r). ✷
0

Lemma 2.
|b(t) − b(0,2i+1 ] |
|b(t) − b(0,2k+1] | + 2|i − k|bCMO1 .

Proof. Since

2j
1
|b(0,2j ] − b(0,2j+1 ] |
j |b(t) − b(0,2j+1] | dt
2
0
2
j+1
1

2× |b(t) − b(0,2j+1 ] | dt
2j +1
0

2bCMO1 ,
if k < i, then


i
|b(t) − b(0,2i+1 ] |
|b(t) − b(0,2k+1] | + |b(0,2j ] − b(0,2j+1 ] |
j =k+1

|b(t) − b(0,2k+1] | + 2(i − k)bCMO1 ,
if k > i, then


k
|b(t) − b(0,2i+1 ] |
|b(t) − b(0,2k+1] | + |b(0,2j ] − b(0,2j+1 ] |
j =i+1

|b(t) − b(0,2k+1] | + 2(k − i)bCMO1 . ✷

Proof of (1). We have

∞
|Hb f (x)|p dx
0
2i+1
x
∞

p
 1  

= b(x) − b(t) f (t) dt dx
x
i=−∞
2i 0
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 633

2i+1
∞
2
k+1 p
 1 i
 
b(x) − b(t) f (t) dt dx

i
2
i=−∞ k=−∞
2i 2k

∞
2
k+1
2i+1 p
p/p 
 1 
i
 

2 i
b(x) − b(0,2i+1] f (t) dt dx
2
i=−∞ k=−∞
2i 2k

∞
2
k+1
2i+1 p
p/p 
 1 
i
 
+2 i
b(t) − b(0,2i+1 ] f (t) dt dx
2
i=−∞ k=−∞
2i 2k
= I + II.

For I , we have

∞

2i+1  2
k+1 p
 1 p 
i
p/p  b(x) − b dx
I =2 (0,2i+1 ] |f (t)| dt
2ip
i=−∞ k=−∞
2i 2k

∞

2i+1(1/p)p
 2i+1 1
p/p  b(x) − b i+1 p dx

2 (0,2 ]
2ip 2i+1
i=−∞ 0
  2
k+1 1/p  2
k+1 1/p p

i
× |f (t)|p dt dt
k=−∞
2k 2k
(by Hölder’s inequality)
∞
  2
k+1 1/p p
 
i
p/p  +1 k/p 

2 (bCMOp )p 2i(1−p) 2 |f (t)|p dt
i=−∞ k=−∞
2k

∞
  2
k+1 1/p p
 
i
p/p  +1 (k−i)/p 

2 (bCMOp ) p
2 |f (t)| dt
p

i=−∞ k=−∞
2k
  p/p 


2p/p +1 2(k−i)/2 2(k−i)(p−1)/2(bCMOp )p
i−k
0 i−k
0

∞
× |f (t)|p dt (by Lemma 1(i): l = 0, s = p , r = p).
0
634 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

For II, by Lemma 2, we have

2i+1

2
k+1 p
∞
 
i
 
2p/p  1 b(t) − b k+1 f (t) dt dx
II
2 (0,2 ]
2ip
i=−∞ k=−∞
2i 2k

∞


i 2i+1 2
k+1 p

 1
+ 22p/p 2(i − k)b 1 |f (t)| dt dx
2 ip CMO
i=−∞ k=−∞
2i 2k
= II 1 + II 2 .
For II 1 , by Hölder’s inequality, we have

2i+1

 2
k+1 1/p
∞
 
i
p 
2p/p  1
II 1
2 b(t) − b(0,2k+1] dt
2ip
i=−∞ k=−∞
2i 2k
 2
k+1 1/p p


× |f (t)|p dt dx

2k
i  2
k+1 1/p p
2i  (k+1)/p
∞

2p/p 

2 (bCMOp )p 2 |f (t)| p
dt
2ip
i=−∞ k=−∞
2k
i
∞ 
 2
k+1 1/p p

 

23p/p (bCMOp )p 2(k−i)/p |f (t)|p dt

i=−∞ k=−∞
2k
  p/p 


23p/p 2(k−i)/2 2(k−i)(p−1)/2(bCMOp )p
i−k
0 i−k
0

∞
× |f (t)|p dt (by Lemma 1(i): l = 0, s = p , r = p).
0

For II 2 , by Hölder’s inequality, as II 1 , we have

  p
II 2
22p/p +p bCMO1
∞ 
i  2
k+1 1/p p

(k−i)/p 
× (i − k)2 |f (t)| dt
p


i=−∞ k=−∞
2k
(using Hölder’s inequality)
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 635

  p/p
2p/p +p p  (k−i)/2

2 (i − k) 2
i−k
0


∞

× 2(k−i)p/(2p ) (bCMO1 )p |f (t)|p dt
i−k
0 0

(by Lemma 1(i): l = 1, s = p , r = p).
Combining the estimate above, we obtain (1). ✷

Proof of (2). We have

∞
∗
H f (x) p dx
b
0
2i+1
∞
∞

p
  

= b(x) − b(t) f (t) dt dx

i=−∞ x
2i
∞ 2
k+1
∞
 2i+1 p
  

b(x) − b(t) f (t) dt dx

i=−∞ k=i
2i 2k
∞ 2
k+1
∞
 2i+1 p
   

2p/p b(x) − b i+1 f (t) dt dx
(0,2 ]

i=−∞ k=i
2i 2k
∞ 2
k+1
∞
 2i+1 p
p/p 
  
+2
b(t) − b(0,2i+1 ] f (t) dt dx

i=−∞ k=i
2i 2k
= J + JJ.
For J , we have
∞

2i+1  ∞ 2

k+1 p
 p 
p/p  b(x) − b dx
J =2 (0,2i+1 ] |f (t)| dt
i=−∞ k=i
2i 2k

∞


2i+1 (1/p)p
 1 p
p/p  b(x) − b dx

2 2 i+1
(0,2i+1 ]
2i+1
i=−∞ 0
 ∞
 2
k+1 1/p  2
k+1 1/p p

× |f (t)|p dt dt
k=i
2k 2k
636 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

(by Hölder’s inequality)

∞
 ∞
 2
k+1 1/p p
 
p/p  +1 k/p 

2 (bCMOp )p 2i 2 2−kp |tf (t)|p dt
i=−∞ k=i
2k

∞
 ∞
 2
k+1 1/p p
 
p/p  +1 −(k−i)/p
=2 (bCMOp )p 2 |tf (t)| dt
p

i=−∞ k=i
2k
  p/p 
 

2p/p +1 2−(k−i)p /(2p) 2−(k−i)/2 (bCMOp )p
k−i
0 k−i
0

∞
× |tf (t)|p dt (by Lemma 1(ii): l = 0, s = p, r = p).
0
For JJ, by Lemma 2, we have
2i+1 ∞ 2
k+1
∞

p
   

JJ
2 2p/p
b(t) − b(0,2k+1] f (t) dt dx

i=−∞ k=i
2i 2k
2i+1 ∞
∞
2
k+1 p
 
2p/p 
+2 2(k − i)bCMO1 |f (t)| dt dx

i=−∞ k=i
2i 2k
= JJ 1 + JJ 2 .
For JJ 1 , by Hölder’s inequality, we have
∞
2i+1 ∞  2
k+1 1/p
  p 
2p/p  b(t) − b k+1 dt
JJ 1
2 (0,2 ]

i=−∞ k=i
2i 2k
 2
k+1 1/p p


× |f (t)|p dt dx

2k
2p/p 

2 (bCMOp )p
∞  2
k+1 1/p p
∞
 

× 2i 2(k+1)/p 2−kp |tf (t)|p dt

i=−∞ k=i
2k
∞
∞ 
 2
k+1 1/p p


= 23p/p (bCMOp )p 2−(k−i)/p |tf (t)|p dt

i=−∞ k=i
2k
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 637

  p/p 
3p/p  −(k−i)p  /(2p)

2 2 2−(k−i)/2 (bCMOp )p
k−i
0 k−i
0

∞
× |tf (t)|p dt (by Lemma 1(ii): l = 0, s = p, r = p).
0

For JJ 2 , by Hölder’s inequality, we have


JJ 2
22p/p +p (bCMO1 )p
∞  2
k+1 1/p  2
k+1 1/p p
∞
 
i
× 2 (k − i) |f (t)| dt
p
dt

i=−∞ k=i
2k 2k
(by Hölder’s inequality)
2p/p  +p

2 (bCMO1 )p
∞
∞ 
 2
k+1 1/p p

−(k−i)/p
× (k − i)2 |tf (t)| dt
p


i=−∞ k=i
2k
  p/p
  

22p/p +p (k − i)p 2−(k−i)p /(2p)
k−i
0


∞
−(k−i)/2
× 2 (bCMO1 ) p
|tf (t)|p dt
k−i
0 0
(by Lemma 1(ii): l = 1, s = p, r = p).
Combining the estimate above, we obtain (2). ✷

Proof of (3). We have

∞
α
H f (x) p dx
b
0
2i+1
∞

p
 1
x  

= 1−α b(x) − b(t) f (t) dt dx
x
i=−∞
2i 0

∞
2
k+1
2i+1 p
 1 
i
 

i(1−α) b(x) − b(t) f (t) dt dx
2
i=−∞ k=−∞
2i 2k
638 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

2i+1
∞
2
k+1 p
 1 
i
 
p/p  b(x) − b i+1 f (t) dt dx

2 i(1−α) (0,2 ]
2
i=−∞ k=−∞
2i 2k

2i+1
∞
2
k+1 p
 1 
i
 
p/p  b(t) − b i+1 f (t) dt dx
+2 i(1−α) (0,2 ]
2
i=−∞ k=−∞
2i 2k

= K + KK.

For K, we have

∞

2i+1  i 2
k+1 p
 1 
p/p 
K =2 |b(x) − b(0,2i+1] | dx
p
|f (t)| dt
2ip(1−α)
i=−∞ k=−∞
2i 2k

∞


2i+1 (1/p)p
 2i+1 1 p
p/p  b(x) − b dx

2 (0,2i+1 ]
2ip(1−α) 2i+1
i=−∞ 0

  2
k+1 1/p0  2
k+1 1/p p


i 0
× |f (t)|p0 dt dt
k=−∞
2k 2k

(by Hölder’s inequality)

p/p  +1

2 (bCMOp )p

∞
  2
k+1 1/p0 p
 
i
k/p0
× 2i(1−p(1−α)) 2 |f (t)| dt
p0

i=−∞ k=−∞
2k

∞
  2
k+1 1/p0 p
 
i
p/p  +1 (k−i)/p0

2 (bCMOp )p 2 |f (t)| dt
p0

i=−∞ k=−∞
2k

(notice that 1/p = 1/p0 − α)

  p/p 
p/p  +1 (k−i)p  /(2p0 ) 

2 2 2(k−i)p/(2p0) (bCMOp )p
i−k
0 i−k
0

∞ p/p0
× |f (t)|p0 dt (by Lemma 1(i): l = 0, s = p0 , r = p0 ).
0
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 639

For KK, by Lemma 2, we have

2i+1

2
k+1 p
∞
 
i
 
2p/p  1 b(t) − b k+1 f (t) dt dx
KK
2 (0,2 ]
2ip(1−α)
i=−∞ k=−∞
2i 2k
∞

 1
+ 22p/p
2ip(1−α)
i=−∞
2i+1

2
k+1 p

i

× 2(i − k)bCMO1 |f (t)| dt dx

k=−∞
2i 2k
= KK 1 + KK 2 .
For KK 1 , we have
2i+1

 2
k+1 1/p
∞
 
i
p  0
2p/p  1
KK 1
2 b(t) − b(0,2k+1] dt
0
2ip(1−α)
i=−∞ k=−∞
2i 2k
 2
k+1 1/p0 p


× |f (t)| dt
p0
dx

2k
(by Hölder’s inequality)
∞

 2i

22p/p (b p  ) p
ip(1−α)
CMO 0 2
i=−∞
i  2
k+1 1/p0 p


× 2(k+1)/p0 |f (t)|p0 dt

k=−∞
2k
2p/p  +p/p0

2 (b p 
p
)
CMO 0
i
∞ 
 2
k+1 1/p0 p

(k−i)/p0
× 2 |f (t)| dt
p0


i=−∞ k=−∞
2k
(notice that 1/p = 1/p0 − α)
  p/p
   

22p/p +p/p0 2(k−i)p /(2p0 )
i−k
0

∞ p/p0

(k−i)p/(2p0 )
× 2 (b p ) p
|f (t)| p0
dt
CMO 0
i−k
0 0
640 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

(by Lemma 1(i): l = 0, s = p0 , r = p0 ).

For KK 2 , by Hölder’s inequality, we have
∞

 1
KK 2
22p/p +p (bCMO1 )p 2i
2ip(1−α)
i=−∞
i  2
k+1 1/p0  2
k+1 1/p p
 0

× (i − k) |f (t)| dt
p0
dt

k=−∞
2k 2k
(by Hölder’s inequality)
2p/p  +p

2 (bCMO1 )p
i
∞ 
 2
k+1 1/p0 p

(k−i)/p0
× (i − k)2 |f (t)| dt
p0


i=−∞ k=−∞
2k
(notice that 1/p = 1/p0 − α)
  p/p
2p/p  +p p  (k−i)p  /(2p0 )

2 (i − k) 2
i−k
0

∞ p/p0

(k−i)p/(2p0 )
× 2 (bCMO1 ) p
|f (t)| p0
dt
i−k
0 0

(by Lemma 1(i): l = 1, s = p0 , r = p0 ).
Combining the estimate above, we obtain (3). ✷

Proof of (4). By the proof of (2), we have

∞
∗
H f (x) p dx
b
0
2i+1 ∞ 2
k+1
∞

p
   
p/p  b(x) − b i+1 f (t) dt dx

2 (0,2 ]

i=−∞ k=i
2i 2k
2i+1 ∞ 2
k+1
∞

p
   
p/p  b(t) − b i+1 f (t) dt dx
+2 (0,2 ]

i=−∞ k=i
2i 2k
= L + LL.
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 641

For L, we have

∞

2i+1  ∞ 2
k+1 p
 
p/p 
L=2 |b(x) − b(0,2i+1] | dx
p
|f (t)| dt
i=−∞ k=i
2i 2k

∞


2i+1 (1/p)p
 1
p/p 

2 2i+1 |b(x) − b(0,2i+1 ] |p dx
2i+1
i=−∞ 0
 ∞
 2
k+1 1/p0  2
k+1 1/p p
 0
× |f (t)|p0 dt dt
k=i
2k 2k
(by Hölder’s inequality)
p/p  +1

2 (bCMOp )p
∞
 ∞  2
k+1 1/p0 p
 
k/p0 −kp0 (1−α) 1−α
× 2 i
2 2 |t p0
f (t)| dt
i=−∞ k=i
2k
p/p  +1

2 (bCMOp ) p

∞
 ∞  2
k+1 1/p0 p
 
× 2−(k−i)/p |t 1−α f (t)|p0 dt
i=−∞ k=i
2k
(notice that 1/p = 1/p0 − α)
  p/p 
p/p  +1 −(k−i)p  /(2p)

2 2 2−(k−i)/2(bCMOp )p
k−i
0 k−i
0

∞ p/p0
× |t 1−α f (t)|p0 dt
0
(by Lemma 1(ii): l = 0, s = p, r = p0 ).
For LL, by Lemma 2, we have
2i+1 ∞ 2
k+1
∞

p
   
2p/p 
LL
2 b(t) − b(0,2k+1] f (t) dt dx

i=−∞ k=i
2i 2k
2i+1 ∞

2
k+1 p
∞
 
2p/p 
+2 2(k − i)bCMO1 |f (t)| dt dx

i=−∞ k=i
2i 2k
= LL1 + LL2 .
642 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

For LL1 , by Hölder’s inequality, we have

2i+1 ∞  2
k+1
∞

1/p
  0
2p/p  
LL1
2 |b(t) − b(0,2k+1] |p0 dt

i=−∞ k=i
2i 2k
 2
k+1 1/p0 p


× |f (t)| dt
p0
dx

2k
(by Hölder’s inequality)


22p/p (b p )p
CMO 0
∞  2
k+1 1/p0 p
∞
 
 −kp0 (1−α) 1−α
× i
2 2 (k+1)/p0
2 |t p0
f (t)| dt

i=−∞ k=i
2k
2p/p  +p/p0

2 (b p )p
CMO 0
∞
∞ 
 2
k+1 1/p0 p

−(k−i)/p
× 2 |t 1−α p0
f (t)| dt

i=−∞ k=i
2k
(notice that 1/p = 1/p0 − α)
  p/p
  

22p/p +p/p0 2−(k−i)p /(2p)
k−i
0

∞ p/p0

−(k−i)/2
× 2 (b p0 ) p
|t 1−α
f (t)| p0
dt
CMO
k−i
0 0
(by Lemma 1(ii): l = 0, s = p, r = p0 ).
For LL2 , by Hölder’s inequality, as LL1 , we have

LL2
22p/p +p (bCMO1 )p
∞  2
k+1 1/p0  2
k+1 1/p p
∞
  0

× 2i (k − i) |f (t)|p0 dt dt

i=−∞ k=i
2k 2k
(by Hölder’s inequality)
2p/p  +p

2 (bCMO1 )p
∞
∞ 
 2
k+1 1/p0 p


× (k − i)2−(k−i)/p |t 1−α f (t)|p0 dt

i=−∞ k=i
2k
 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644 643

(notice that 1/p = 1/p0 − α)

  p/p
  

22p/p +p (k − i)p 2−(k−i)p /(2p)
k−i
0

∞ p/p0

−(k−i)/2
× 2 (bCMO1 ) p
|t 1−α
f (t)| p0
dt
k−i
0 0
(by Lemma 1(ii): l = 1, s = p, r = p0 ).
Combining the estimate above, we obtain (4). ✷

Acknowledgments

The authors are grateful to the unknown referee(s) for many helpful comments on this paper.

References

[1] Y. Bicheng, Z. Zhuohua, L. Debnath, Note on new generalizations of Hardy’s integral inequality,
J. Math. Anal. Appl. 217 (2) (1998) 321–327.
[2] Y. Bicheng, Z. Zhuohua, L. Debnath, Generalizations of Hardy integral inequality, Internat.
J. Math. Math. Sci. 22 (3) (1999) 535–542.
[3] S. Bloom, A commutator theorem and weighted BMO, Trans. Amer. Math. Soc. 292 (1985)
103–122.
[4] T.A.A. Broadbent, A proof of Hardy’s convergence theorem, J. London Math. Soc. 3 (1928)
242–243.
[5] Y.Z. Chen, K.S. Lau, Some new classes of Hardy spaces, J. Functional Anal. 84 (1989) 255–278.
[6] R.R. Coifman, R. Rochberg, G. Weiss, Factorization theorems for Hardy spaces in several
variables, Ann. Math. 103 (1976) 611–635.
[7] E. Copson, Some integral inequality, Proc. Roy. Soc. Edinburgh Sect. A 75 (1975–76) 157–164.
[8] E.B. Elliott, A simple exposition of some recently proved facts as to convergency, J. London
Math. Soc. 1 (1926) 93–96.
[9] J. Garcia-Cuerva, Hardy spaces and Beurling algebras, J. London Math. Soc. 39 (1989) 499–513.
[10] J. Garcia-Cuerva, M.J.L. Herrero, A theory of Hardy spaces associated to the Herz spaces, Proc.
London Math. Soc. 69 (3) (1994) 605–628.
[11] G.H. Hardy, Note on a theorem of Hilbert, Math. Z. 6 (1920) 314–317.
[12] G.H. Hardy, Note on some points in the integral calculus, Messenger Math. 57 (1928) 12–16.
[13] G.H. Hardy, J.E. Littlewood, G. Polya, Inequalities, Cambridge University Press, Cambridge,
UK, 1959.
[14] M. Izumi, S. Izumi, On some inequalities for Fourier series, J. Anal. Math. 21 (1968) 277–291.
[15] J.D. Lakey, Constructive decomposition of functions of finite central mean oscillation, Proc.
Amer. Math. Soc. 127 (8) (1999) 2375–2384.
[16] E. Landau, A note on a theorem concerning series of positive terms, J. London Math. Soc. 1
(1926) 38–39.
[17] N. Levinson, Generalizations of an inequality Hardy, Duke Math. J. 31 (1964) 389–394.
[18] E.R. Love, Generalizations of Hardy’s inequality, Proc. Roy. Soc. Edinburgh Sect. A 100 (1985)
237–262.
644 L. Shunchao, W. Jian / J. Math. Anal. Appl. 274 (2002) 626–644

[19] J.A. Oguntuase, C.O. Imoru, New generalizations of Hardy’s integral inequality, J. Math. Anal.
Appl. 241 (2000) 73–82.
[20] B.G. Pachpatte, On a new class of Hardy type inequality, Proc. Roy. Soc. Edinburgh Sect. A 105
(1987) 265–274.
[21] B.G. Pachpatte, On some integral inequality similar to Hardy’s integral inequality, J. Math. Anal.
Appl. 129 (1988) 596–606.
[22] B.G. Pachpatte, A note on certain inequality related to Hardy’s inequality, Indian J. Pure Appl.
Math. 23 (1992) 773–776.
[23] B.G. Pachpatte, On some generalizations of Hardy’s integral inequality, J. Math. Anal.
Appl. 234 (1) (1999) 15–30.
[24] B.G. Pachpatte, E.R. Love, On some new inequality related to Hardy’s integral inequality,
J. Math. Anal. Appl. 149 (1990) 17–25.
[25] J.E. Pecaric, E.R. Love, Still more generalizations of Hardy’s integral inequality, J. Austral.
Math. Soc. Ser. A 58 (1995) 1–11.
[26] C. Segovia, J.L. Torrea, Higher order commutators for vector-valued Calderon–Zygmund
operators, Trans. Amer. Math. Soc. 336 (1993) 537–556.