See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228687755

On maximal function and fractional integral, associated with

the Bessel differential operator

Article in Mathematical Inequalities & Applications · April 2003

DOI: 10.7153/mia-06-30

CITATIONS READS

79 667

1 author:

Vagif Guliyev
Institute of Applied Mathematics of Baku State University
306 PUBLICATIONS 4,233 CITATIONS

SEE PROFILE

All content following this page was uploaded by Vagif Guliyev on 27 March 2014.

The user has requested enhancement of the downloaded file.

 M athematical
I nequalities
& A pplications
Volume 6, Number 2 (2003), 317–330

ON MAXIMAL FUNCTION AND FRACTIONAL INTEGRAL,

ASSOCIATED WITH THE BESSEL DIFFERENTIAL OPERATOR

VAGIF S. GULIEV

(communicated by V. Burenkov)

Abstract. In this paper we consider the generalized shift operator, generated by Bessel differen-
tial operator B , by means of which maximal functions ( B -maximal functions) and fractional
integrals ( B -fractional integrals) are investigated. The Lp (B) -boundedness result for the B -
maximal function and (Lp (B) Lq (B)) -boundedness result for the B -fractional integral are
obtained.

1. Introduction

Suppose that Rn is the n -dimensional Euclidean space, x = (x 1 : : : x n ) , ξ =

(ξ1 : : : ξn) are vectors in Rn , x ξ = x 1ξ1 + : : : + x nξn , jx j = (x x )1=2 , Rn+ = fx =
(x 1 : : : x n); x1 > 0 : : : x n > 0g , E+ (x r) = fy 2 Rn+ : jx ; y j < rg:
The Bessel differential operator B = (B1 : : : Bn ) is defined by

+ xγi
2
Bi = @
2
@xi @
@

xi
i=1 ::: n
i

γ =( R)
γ1 : : : γn γ1 > 0 : : : γn > 0 jγ j γ1 = + ::: + γn xγ = x γ11 : : : x γnn
( ) =
jE+ 0 r jγ
E+ (0
r)
x γ
dx Cr n+jγ j
: =
6 6
For 1 p n
1 let L p R+ B
n
(γ
 L p R+ x dx ) ( ) be the space of functions
n
measurable on R+ with the norm
Z != 1 p

f
k kL p (Rn
B)
+
= j f (x )j x dx
p γ
Rn+

f
k kL
1 Rn+
B
( )  k kL f 1 Rn+ = essx2Rsup
(
n
) f (x ) j j:

+
Denote by T y the generalized shift operator ( B -shift operator) defined by
Y  n  Z π Z Y π n
T f (x ) = π ; 2 Γ γi + Γ;1 (γ ) sinγi ;1 αi 
y n 1
i :::
2
q i=1
q
0 0 i=1

 f x 21 ; 2x 1 y 1 cos α1 + y 21 ::: x 2n ; 2x n y n cos αn + y 2n dα1 : : : dαn :

Mathematics subject classification (2000): 47F05, 46E35.

Key words and phrases: Bessel differential operator, B -maximal function, B -fractional integral.

317

c D l , Zagreb
Paper No. 06-30/9.00$
318 VAGIF S. GULIEV

Let f be in L p (Rn+ B) 1 6p6 1: Then for all x 2 Rn+ the function T x f

belongs to L p (Rn+ B) and
k T x f kLp (Rn+
B) 6 f
k kL p (Rn
B) :
+ (1)
We note that T y is closely connected with the Bessel differential operator B =
(B1 : : : Bn) (see  1] for details the one-dimensional case). The B -shift T y gener-
ates the corresponding B -convolution
Z
(f  g)B (x ) = T y f (x )g(y )y γ dy :
Rn+

We note the following properties of the B -convolution:

( f  g)B = (g f )B 

k (f  g)B kLp (Rn+
B) 6 f L R+
B

k k ( n
1
g +
B)
) k kL p (Rn :

The Bessel (Hankel) transformations can be defined by

^
Z Y n
ϕ λ ( ) = (FB ϕ ) (λ ) = ϕ x () j γi ;1 (x i λ i )x γ dx
Rn+ i=1
2

and its inverse transformations can be given by

_
ϕ x ( )=
;F; ϕ  (x) = C Z
1
^
ϕ λ ( )
Yn
j γi ;1 (x i λ i )λ γ dλ
B n
γ
Rn + i=1
2

where jν (t ) = t ;ν Jν (t ) Jν being the Bessel function of the first kind.

For f 2 L p (Rn+ B) p = 1 or 2 we have FB ( f  g)B = FB f FB g:
We shall define function spaces, generated by the Bessel differential operator
B = (B1 : : : Bn ):
DEFINITION 1.  2 ] Let 1 6 p < 1 0 6 λ 6 n + jγ j  t ]1 = minf1 t g: We
denote by L p
λ (Rn+ B) B -Morrey spaces and by L p
λ (Rn+ B) modified B -Morrey e
spaces which are the sets of functions f locally integrable on Rn+ with finite norms
Z != 1 p

f
k k
Lp λ (Rn+
B) = sup t ;λ T j f (x )j y dy
y p γ
Rn+ (0
1) E+ (0
t )

Z != 1 p

f
k k
e
Lp λ (Rn+
B)
= sup  t ];λ1 T j f (x )j y dy
y p γ
:
Rn+ (0
1) E+ (0
t )

DEFINITION 2. Let us now introduce, as in  2 ], the B - BMO space BMO(Rn+ B)

as the set of functions locally integrable on Rn+ with finite norm
Z
f 
B
k k = sup E (0 r) ;γ 1
j + j j T y f (x ) ; fE+ (0
r) (x )jy γ dy < 1
x
r E+ (0
r)
 ON MAXIMAL FUNCTION AND FRACTIONAL INTEGRAL 319

where Z
fE+ (0
r) (x ) = jE+ (0 r)j;
γ
1
T y f (x )y γ dy :
E+ (0
r)

Note that
e
L p
0 (Rn+ B) = L p
0 (Rn+ B) = L p (Rn+ B)
L p
n+jγ j (Rn+ B) = L1 (Rn+ B)

e
L p
λ (Rn+ B)  L p (Rn+ B) and k k f Lp (Rn+
B) 6 f
k k
e
Lp λ (Rn+
B)
:

LEMMA 1. Let 1 6 p < 1 0 6 λ 6n+ jγ j: Then for α p =n+ γ j j ; λ

L p
λ (Rn+ B)  L1
n+jγ j;α (Rn+ B) and 6C f L

k k f L1 n+ γ j j; α (R+
B)
n k k
p λ (R+
B)
n :

Proof. Let f 2 L p
λ (Rn+ B) 1 6 p < 1 0 6 λ 6 n + jγ j 1= p + 1= p0 =1

and α p = n + jγ j ; λ . By applying Hölder’s inequality we have

Z Z != Z 1 p !=0 1 p

T j f (x )jy dy
y γ
6 (T f (x ) )
y
j j
p γ
y dy y dyγ
E+ (0
t ) E+ (0
t ) E+ (0
t )

Z !=
jγ j)=p0
1 p

6 Ct (n+
T j f (x )j y dy
y p γ
:
E+ (0
t )

Moreover
Z Z !=1 p

t ;(n+jγ j)
α
T j f (x )jy dy 6 Ct ;(n+jγ j)=p
y γ α
T j f (x )j y dy
y p γ
E+ (0
t ) E+ (0
t )

Z != 1 p

6 C t;λ T j f (x )j y dy
y p γ
E+ (0
t )

6C k k f Lp λ (Rn+
B) :

Therefore f 2 L1
n+jγ j;α (Rn+ B) and

k k f
j j;
L1 n+ γ α (R+
B)
n 6C k k f Lp λ (Rn+
B) :

2. L p (B) -boundedness of B -maximal function

We now consider the B -maximal operator (see  2 ],  3 ])

Z
MB f (x ) = sup jE+ (0 r)j;
γ
1
T y j f (x )jy γ dy :
r>0 E+ (0
r)
320 VAGIF S. GULIEV

THEOREM 1. 1. If f 2 L1 (Rn+ B) , then for every α > 0

Z
x : MB f (x ) > α g jγ 6α f (x )jx γ dx
C
jf j
Rn+

where C > 0 is independent of f .

2. If f 2 L p (Rn+ B) 1 < p 6 1 , then MB f 2 L p (Rn+ B) and
k MB f kLp (Rn+
B) 6 Cp f
k kL p (Rn
B)
+

where Cp > 0 is independent of f .

Proof. We need to introduce a maximal function defined on a space of homo-
geneous type. By this we mean a topological space X equipped with a continuous
pseudo-metric ρ and a positive measure µ satisfying the doubling condition
( (
µ E x 2r )) 6 Cµ (E(x r)) (2)
where C is independent of x and r > : 0 Here E(x r) = f y 2 X : ρ (x y ) < rg: Let
(X ρ µ ) be a space of homogeneous type. Define
Z
Mµ f (x ) = sup µ (E(x r));1 j f (y )jdµ (y ):
r>0 E(x
r)

It is well known that the maximal operator Mµ is of weak type (1 1) and is

bounded on L p (X dµ ) for 1 < p < 1 (see  4 ]). We shall use this result in the case
in which X = Rn+ ρ (x y ) = jx ; y j dµ (x ) = x γ dx : It is clear that this measure
satisfies the doubling condition (2).
Also
Z
( )=
µE x r j E+ (x r)jγ = y γ dy
fy2Rn+ : jx;yj<rg
Y n Z Y Z
n x i +r

6 y γi dy i = y γi dy i
fyi >0; jxi ;yi j<rg maxf0
xi ;rg

Y(r
i=1 i=1

n 1+γi
r xi Y
n
6C = Crn jγ j maxf1 (x i =r)γi g:
>
+

i=1 rx γi i r 6 xi i=1

We shall show that

MB f (x ) 6 CMµ f (x ):
By the definition of the B -shift operator it follows that T y χ E+ (0
r) (x ) is supported
in E+ (x r) .
Next we estimate T y χ E+ (0
r) (x ) .
Z Z
T y χ E+ (0
r) (x ) 6 C sinγi ;1 αi dα1 : : : dαn

f(α1
:::
αn )2(0
π )n :
Pn
(x 2i ;2xi yi cos αi +y2i )<r2 g
i=1
 ON MAXIMAL FUNCTION AND FRACTIONAL INTEGRAL 321

Y n Z
6C sinγi ;1 αi dαi
i=1
fαi 2(0
π ):x2i ;2xi yi cos αi +y2i <r2 g
Y n Z
=C sinγi ;2 αi d cos αi
n
i=1
x 2 +y 2 r2 ; o
2
α i (0
π ): i 2x ii y i <cos α i

Y n Z1
;1  γi
;1
=C t i2 2
dt i
n ; o
;

i=1
;1
xi +2xyii yi r
2 2 2
max

Y n  rγi =2 (r ; jx i ; y i j)γi =2
 Y n
6C min 1
(x iy i)γi =2 6C minf1 (r=x i );γi g:
i=1 i=1
Consequently there exists C > 0 such that for all x 2 Rn+ r > 0 and y 2

E+ (x r)
Yn
T y χ E+ (0
r) (x ) 6 C minf1 (r=x i )γi g:
i=1
Thus
Xn
MB f (x ) 6 MB
k f (x )
k=0
Xn Z
= sup jE+ (0 r )jγ
;1 j f (y )jT y χ E+ (0
r) (x )y γ dy
k=0 r>xij j=1 k E+ (x
r)
6
r xij j=k+1 n
6
ij =i p j= p 6
where
Z
MB
0 f (x ) = sup jE+ (0 r )jγ
;1 j f (y )jT y χ E+ (0
r) (x )y γ dy
6 E+ (x
r)
r x j
j=1
n
Z
MB
n f (x ) = sup jE+ (0 r )jγ
;1 j f (y )jT y χ E+ (0
r) (x )y γ dy :
r>x j
j=1
n E+ (x
r)

Without loss of generality we assume that ij  j j=1 ::: n: Then

Z
MB
k f (x ) = sup j E+ (0 r)jγ;1 j f (y )jT y χ E+ (0
r) (x )y γ dy :
r>xj j=1 k E+ (x
r)
6
r xj j=k+1 n

In the case k = 0 by taking into account that µ E + (x r) 6 Crn+jγ j , jE+ (0 r)jγ =

rn+jγ j and T y χ E+ (0
r) 6 1 , we have
Z
MB
0 f (x ) 6 f (y )jy γ dy 6 CMµ f (x )
1
µ E+ (x r)
sup j :
r6x
j=1
n
j E+ (x
r)
322 VAGIF S. GULIEV

In the case 1 6 k 6 n , by taking into account

Yn Y k
( ) 6 Crn
µ E+ x r +jγ j maxf1 (x i =r)γi g = Crn+jγ j (x i r)γ
=
i

i=1 i=1
Y n Y k
T y χ E+ (0
r) 6C minf1 (r=x i )γi g = (r x i )γ
=
i

i=1 i=1

we have

Yk Y k
MB
k f (x ) 6 C sup j E+ (0 r)j; 1 n+jγ j
γ r (x i r)γ
=
i
(r xi )γ
=
i

r>xj j=1 k i=1 i=1
6
r xj j=k+1 n
Z
f (y )jy γ dy 6 CMµ f (x )
1
µ E+ (x r)
 j :
E+ (x
r)

Finally we get
MB f (x ) 6 CMµ f (x ):
This completes the proof.

COROLLARY 1. If f 2 Lp (Rn+ B) 1 6 p 6 1 then

Z
lim jE+ (0 r)jγ;1 T y f (x )y γ dy = f (x )
r !0 E+ (0
r)

for a. e. x 2 Rn+ :
REMARK 1. In the one-dimensional case Theorem 1 was proved earlier by another
method by K. Stempak  5 ].

3. Sobolev – type theorem for the B -fractional integral

Consider the B -Riesz potentials

Z
IBα f (x ) = T y jx jα ;n;jγ j f (y )y γ dy 0<α < n + jγ j
Rn +

and the modified of B -Riesz potentials

Z
Ie f (x ) =
α
B T y jx jα ;n;jγ j ; jy jα ;n;jγ j χ E+ (0
1) (y ) f (y )y γ dy
Rn+

where E+  (0 r) = Rn nE+ (0 r) r > 0:

+
n+jγ j
The examples considered below show that if p > α then the B -potentials IBα
are not defined for all functions f 2 L p (Rn+ B):
 ON MAXIMAL FUNCTION AND FRACTIONAL INTEGRAL 323

EXAMPLE 1. Let x 2 Rn+ 0 < α < n + jγ j f (x ) = jx j α 

ln jx j χ E+ (0
2)
1
(x ) . For
n+jγ j
p= α
f 2 L p (R+ B) and
n
IBα f (x ) = +1:

EXAMPLE 2. Let x 2 Rn+ 0 < α < n + jγ j f (x ) = x ;α χ E+ (0
2) (x ) . For

j j
n+jγ j
p> α
f 2 L p (Rn+ B) and IBα f (x ) = +1:
For these B -Riesz potentials the following analogue of Hardy–Littlewood–Sobolev
theorem is valid.
THEOREM 2. Let 0 < α < n + jγ j , 1 6 p < n+αjγ j and 1p ; 1q = n+αjγ j .
a) If f 2 L p (Rn+ B) then the integral IBα f is absolutely convergent for almost
every x 2 Rn+ :
b) If 1 < p < n+αjγ j f 2 L p (Rn+ B) then IBαn f 2 Lq (Rn+ B) and
k IBα f kLq (Rn
B)
+
6 Cp f
k kL (Rn
B)
p + (3)
where Cp > 0 is independent of f .
c) If f 2 L1 (Rn+ B) 1q = 1 ; n+αjγ j then
C  q
jf x 2 Rn+ : IBα f (x ) > β gjγ 6 f
k kL1 (Rn
B)
+ β > 0 (4)
β
where C > 0 is independent of f .
n+jγ j
Proof. a) Let f 2 L p (Rn+ B) 1 6 p < α : We denote
f1 (x ) = f (x )χE+ (0
1) (x ) f2 (x ) = f (x ) ; f1 (x ):
Then
IBα f (x ) = IBα f1 (x ) + IBα f2 (x ) = J1 (x ) + J2 (x ):
Let us estimate J1 (x ) .
Z
J1 (x )j 6
j j j y α ;n;jγ j T y j f (x )jy γ dy
Z E+ (0
1)

= y α ;n;jγ j χ E+ (0
1) (y )T y j f (x )jy γ dy :

j j

Rn+

By Young’s inequality

kJ1 ( )kLp (Rn
B) 6C j j

α ;n;jγ j χ
E+ (0
1) f
k k
Lp (Rn+
B) :
+ L1 (Rn+
B)

Here
Z
j j
α ;n;jγ j χ
E+ (0
1)
L1 (Rn+
B)
= y α ;n;jγ j y γ dy
j j

Z
E+ (0
1)
1
=C α ;1
r dr = C1 < 1:
0
324 VAGIF S. GULIEV

Then for f 2 L p (Rn+ B) , 1 6 p 6 1

kJ1 ( )kLp (Rn
B) 6 C1 f

k kL p (Rn
B)
+
+

e.g. J1 (x ) is absolutely convergent almost every x 2 Rn+ .

By Hölder’s inequality we have
Z
j J2 (x )j 6 y α ;n;jγ j T y j f (x )jy γ dy
j j

Rn+ nE+ (0
1)

0 10
B@ Z
1

0C
p

6 k T x j f ( )jkLp (Rn+
B) y
j j
(α ;n;jγ j) p A :

n
Rn+ E+ (0
1)

By using inequality (1)

0 10
B@ Z
1

0C
p

j J2 (x )j 6 Ck f kLp (Rn+
B) y
j j
(α ;n;jγ j) p A :

n
Rn+ E+ (0
1)

n+jγ j
Hence for 1 6 p < α
for some C > 0

j J2 (x )j 6 Ck f kLp (Rn+
B) x 2 Rn+ :

n+jγ j
Thus for all functions f 2 L p (Rn+ B) 1 6 p < α B -Riesz potentials IBα f (x )
are absolutely convergent for almost every x 2 R n+ :
b) We have
0 1
Z
B + Z CA T f (x) y
IBα f (x ) = @ y
j j
α ;n;jγ j y γ dy = A(x t ) + C(x t ):
E+ (0
t ) Rn+ nE+ (0
t )

By taking sum with respect to all integer k < 0 we get

Z
j A(x t )j 6 j T y f (x )j jy jα ;n;jγ j y γ dy
E+ (0
t )

X
; 1 Z
= j T y f (x )j jy jα ;n;jγ j y γ dy
k=;1 k
6jyj<2k+1 t
Z
2 t

X
; ; 
1
; ;j j
6C 2t k α n γ
j T y f (x )j y γ dy
k=;1
2k t 6jyj<2k+1 t
6 Ct (MB f ) (x )
α
:
 ON MAXIMAL FUNCTION AND FRACTIONAL INTEGRAL 325

Therefore it follows that

j A(x t )j 6 Ct α MB f (x ) (5)
where C > 0 does not depend f x and t :
By Hölder’s inequality and the inequality (1) we have
0 1 0 10
B Z Z
1 1
p

C B 0 C
p

C (x t ) 6 @ T f (x ) y dy A @ y ; ;j j y dy A
y p γ (α n γ )p γ
j j j j  j j

+n +

Rn E (0 t ) +n +
Rn E
0
(0 t )

10
Z
1

6 T f +
B 0 C
p

k
y
@ k y ; ;j j y dy A
Lp (Rn B) j j
(α n γ )p γ

+n +
Rn E (0 t )

0 10
Z
1

6 f +
B 0 C
p

@
k kL (Rn
p y ; ;j j y dy A 6 Ct ; j j = f
B) j j
(α n γ )p γ (n+ γ ) q
k kL p (Rn
B) :
+
n
Rn+ E+ (0
t )

Consequently
jC(x t )j 6 Ct ;(n+jγ j)=q k f kLp (Rn+
B) : (6)
Thus, from (5) and (6), we have

jIBα f (x )j 6 C t α MB f (x ) + t ;(n+jγ j)=q k f kLp (Rn
B) :

h +
i= p (n+jγ j)
Minimum of the right-hand side is attained at t = (MB f (x ));1 f
k kL p (Rn
B)
+
,
so
1; p=q
IBα f (x )j 6 C (MB f (x ))
p=q
j k k f Lp (Rn+
B) :

Z Hence, by the Theorem 1, we;have Z

IBα f (y )j y γ dy 6C (MB f (y ))p y γ dy 6 C
q q p q
j k k f Lp (Rn+
B) k k f Lp (Rn+
B)
E+ (0
t )
Rn+ nE+ (0
t )

and (3) follows.

c) Let f 2 L1 (Rn+ B): Note that
jf x : jIBα f (x )j > 2β gjγ 6 jf x : jA(x t )j > β gjγ + jf x : jC(x t )j > β gjγ :

By taking into account the inequality (6) and applying Theorem 1 we have
Z Z
β jfx 2 R+ : jA(x t )j > β gjγ
n
=β x dxγ
6β x γ dx

2 + j
j>
fx Rn : A(x t ) βg fx 2Rn+ : Ctα MB f (x)>β g

= β  x R : M f (x ) 2
n
+ B >
β
Ct α

Z α
γ

6β C t 1
β
j f (x )jx γ dx = C1tα k k f L1 (Rn+
B)
Rn+
326 VAGIF S. GULIEV

and also
Z
j C(x t )j 6 jT y f (x )j jy jα ;n;jγ j y γ dy
Rn+ nE+ (0
t )
Z
6 tα ;n;jγ j j T y f (x )j y γ dy

+j j
Z +n + Rn E (0
t )

jj
n+ γ
= t; f (x )jx γ dx = t;
n γ
q j q f
k k
L1 (Rn+
B) :

Rn+

n+jγ j
Thus, if t ; q k f kL1 (Rn
B) = β , then C(x t ) 6 β and, consequently,
j j
+
jf x 2 Rn : C(x t )
+ β γ =0
j j > gj :

Finally
jj
n+ γ
jf x 2 Rn+ : jIBα f (x )j > 2β g jγ 6 Cβ1 tα k k f L1 (Rn+
B) = C1tα + q

= C1tn + jγ j = C1 β ;q k k f
q
L1 (Rn+
B) = Cβ f
k k
q
L1 (Rn+ ) :

Therefore the mapping f ! R αB f is of weak type (1 q) .

Theorem has been proved.
n+jγ j
THEOREM 3. Let 0 < α < n + jγ j , 1 < p < α
then the condition
1
p
;
1
q
= n +α γ j j

is necessary for inequality (3) to be valid.

n+jγ j
Proof. Let 1 < p < α
f 2 L p (Rn+ B) and assume that the inequality
k IBα f kLq (Rn
B)
+
6 Cp f
k k
Lp (Rn+
B)

holds.
Define ft (x ) := f (t a x ) , then
Z ! p
1

k ft kLp (Rn
B)

+
= j f (tx )j p x γ dx
Rn+

Z ! 1
p
n+ γjj
= j f (y )j p
t ;n;jγ j y γ dy = t; p f
k k
Lp (Rn+
B)
Rn+

and
Z
α
kIB ft k
Lq (Rn+
B) = ft (y )T y jx jα ;n;jγ j y γ dy
Rn+
Lq (Rn+
B)
 ON MAXIMAL FUNCTION AND FRACTIONAL INTEGRAL 327
Z
= T y jx jα ;n;jγ j f (ty )y γ dy
Rn+
Lq (Rn+
B)

Z Z  !

q 1
q

=  T x Rn
; ;j j f (y )t
Rn
; ;j
ty j y dy  x dx
j j
α

n γ n γ γ γ

!
+ +
Z Z 
 R t j j; T tx ; ;j j f (y)t; ;j j y dy
1
q q

= n n+ γ
+
α y
j j
α n γ n γ γ γ
x dx
+ Rn
n+ γjj
= t;α ; q k IBα f kLq (Rn
B) :
+

By inequality (3)
n+ γ j j ; n+pj j
6 Cp
q tα
γ
kIBα f kLq (Rn ) + q f
k kL (Rn
B) :
p +
+

If 1
p >
1
q + n αjγ j+
then in the case t ! 0 we have kIBα f kLq (Rn
B)
+
=0 for all
f 2 Lp (R+ B) , which is impossible. Similarly, if < + n+jγ j then at t
n 1
p
1
q
α
! 1 we
obtain kIBα f kLq (Rn
B) = 0 for all f 2 L p (Rn+ B) , which is also impossible.
+
Therefore 1
p = 1q + n αjγ j . +

THEOREM 4. Let 0 < α < n + jγ j p = n+jγ j

α
f 2 L p (Rn+ B): Then IBα f e 2

BMO(Rn+ B) and
e
IBα f 6 Cp f
k kL p (Rn
B) :
+
BMO(Rn+
B)

Proof. Let f 2 L p (Rn+ B): Given t > 0 we denote

f1 (x ) = f (x )χE+ (0
2t ) (x ) f2 (x ) = f (x ) ; f1 (x )
where χ E+ (0
2t ) is the characteristic function of the set E+ (0 2t ): Then
e e e
IBα f (x ) = IBα f1 (x ) + IBα f2 (x ) = F1 (x ) + F2 (x )
where
Z
F 1 (x ) = T y jx jα ;n;jγ j ; jy jα ;n;jγ j χ E+ (0
1) (y ) f (y )y γ dy
Z
E+ (0
2t )

F 2 (x ) = T y jx jα ;n;jγ j ; jy jα ;n;jγ j χ E+ (0
1) (y ) f (y )y γ dy :

Rn+ nE+ (0
2t )

Note that the function f1 has compact support and thus

Z
a1 = ; j j y α ;n;jγ j f (y )y γ dy
E+ (0
2t )nE+ (0
minf1
2t g)

is finite.
328 VAGIF S. GULIEV

Note also that

Z
F1 (x ) ; a1 = T y jx jα ;n;jγ j f (y )y γ dy
E+ (0
2t )
Z
; y α ;n;jγ j f (y )y γ dy
j j

E+ (0
2t )nE+ (0
minf1
2t g)

Z
+ y α ;n;jγ j f (y )y γ dy
j j

E+ (0
2t )nE+ (0
minf1
2t g)

Z
= T y jx jα ;n;jγ j f1 (y )y γ dy = IBα f1(x ) :
Rn +
Therefore
Z
jF 1 (x ) ; a 1 j 6 y α ;n;jγ j jT y f1 (x )j y γn dy
j j
Rn+
Z
= y α ;n;jγ j jT y f (x )j y γ dy :
j j

fy2Rn+ : T y jxj<2t g
Furthermore, for jx j < t and T y jx j < 2t we have
j j y 6x+ j j j x ; y j 6 jx j + T y jx j < 3t :
Consequently
Z
j F1 (x ) ; a1 j 6 y α ;n;jγ j jT y f (x )j y γ dy
j j (7)
E+ (0
3t )

if x 2 E+ (0 t ):
By Theorem 1, (1) and (7) for α p = n + jγ j
Z
E+ (0 t )jγ;1
j T z F1 (x ) ; a1 j z γ dz
j
E+ (0
t )
Z
6 j E+ (0 t )jγ;1 T z jF1 (x ) ; a1 j z γ dz
Z
E+ (0
t )
Z !
6 jE+ (0 t )jγ;1 α
jy j
;n;jγ j T y T z j f (x )jy γ dy zγ dz
E+ (0
t ) E+ (0
3t )

Z !=
6 Ct;n;jγ j tα t(n+jγ j)=p0
1 p

(MB (T f (x )))
z p γ
z dz
E+ (0
t )

6 Cp k
z
T f kLp (Rn
B)
+
6 Cp f
k kL p (Rn
B) :
+
(8)
Denote Z
a2 = y α ;n;jγ j f (y )y γ dy :
j j

E+ (0
maxf1
2t g)nE+ (0
2t )

 ON MAXIMAL FUNCTION AND FRACTIONAL INTEGRAL 329

Let us estimate jF2 (x ) ; a2 j : By applying Hölder’s inequality we have

Z  
jF2 (x ) ; a2 j 6 j 
f (y )j T y jx jα ;n;jγ j ; jy jα ;n;jγ j y γ dy 
Rn+ nE+ (0
2t )
Z
6Cx j j j f (y )jjy jα ;n;jγ j;1 y γ dy
Rn+ nE+ (0
2t )
0 1=0
B@ Z
1 p
0 y dy C
6Cx f
j jk kL p (Rn
B)
+
j j y (α ;n;jγ j;1) p γ
A
Rn+ nE+ (0
2t )
n+ γ jj
6 C x tα ;1;
j j p k k f
Lp (Rn+
B) 6 C x t;1 j j f
k kL p (Rn
B) :
+

Note that if jx j 6 t , jz j 6 2t , then T z jx j 6 jx j + jz j 6 3t : Thus for α p = n + jγ j

we obtain

T z F2 (x ) ; a2 j 6 T z jF2 (x ) ; a2 j 6 CT z jx jt ;1 k f kLp(Rn+
B)
j 6C f
k kL p (Rn
B) :
+
(9)
Denote Z
af = a1 + a2 = y α ;n;jγ j f (y )y γ dy :
j j

E+ (0
maxf1
2t g)

Finally, by (8) and (9) we have

Z  
sup jE+ (0 t )jγ;1
x
t E+ (0
t )
T Ie f (x)
y α
B ; 
a f y γ dy 6C f
k kL p (Rn
B) :
+

Finally
Z  
e
IBα f 6 2 sup E (0 t ) ;γ 1
j + j T Ie f (x)
y α
B ; 
a f y γ dy
BMO(Rn+
B) x
t E+ (0
t )
6 C f L R+
B
k k p( n ):

and the statement of the theorem follows.

COROLLARY 2. Let p = n+αjγ j f 2 L p (Rn+ B):

If integral IBα f exists everywhere, then IBα f 2 BMO(Rn+ B) and the inequality

k IBα f kBMO(Rn
B)

+
6 Cp f
k kL p (Rn
B)
+

is valid.
Acknowledgements. The author thanks prof. V. I. Burenkov for the discussions of
results.
 330 VAGIF S. GULIEV

REFERENCES

1] B. M. LEVITAN, Bessel function expansions in series and Fourier integrals. Uspekhi Mat. Nauk 42(2)
(1951), 102–143.
 2 ] V. S. GULIEV, Sobolev theorems for B -Riesz potentials. Dokl. RAN, 358(4) (1998), p. 450–451.
 3 ] V. S. GULIEV, Sobolev theorems for anisotropic Riesz–Bessel potentials on Morrey-Bessel spaces. Dokl.
RAN, 367(2) (1999), p. 155–156.
 4 ] R. R. COIFMAN, G. WEISS, Analyse harmonique non commutative sur certains expaces homogenes.
Lecture Notes in Math., v. 242, Springer-Verlag. Berlin, 1971.
5] K. STEMPAK, Almost everywhere summability of Laguerre series, Studia Math. 100(2) (1991), 129–147.

(Received November 16, 2001) V. S. Guliev

Department of Mathematical Analysis
Institute of Mathematics and Mechanics
Azerbaijan Academy Sciences
F. Agayev st. 10
Baku 370148
Azerbaijan
e-mail: guliev@azdata.net

View publication stats