• 검색 결과가 없습니다.

Weighted Orlicz space integral inequalities for potential maximal operators

N/A
N/A
Protected

Academic year: 2022

Share "Weighted Orlicz space integral inequalities for potential maximal operators"

Copied!
12
0
0

로드 중.... (전체 텍스트 보기)

전체 글

(1)

WEIGHTED ORLICZ SPACE INTEGRAL INEQUALITIES FOR POTENTIAL MAXIMAL OPERATORS

YONG MAL KIM AND YOON JAE Yoo

ABSTRACT. We characterize a condition forMepto be of weak type (Ch,ep2) in terms of Orlicz norms.

1. Introduction

Given a function f in Rn, we define a function Mf in R~+l - {(x, s) : x ERn,s ~ O} by setting

Mf(x,s) =sup{I~I

h

If(y)1 dy: x E Qand sidelength(Q) ~ s}.

It is well known that this maximal operator M controls the Poisson integral defined by, for x E Rn and s ~0,

P(J)(x,s) = r f(y)P(x - y,s)dy,

JRn

where

ens

P(x,s) = (lxl2+s2)n+l/2

is the Poisson kernel. For a given positive measure v onRnX[0,00),un- der what condition onv can we assert thatM is bounded from £P(Rn) into weak-£P(Rn x [0,00),v)? Carleson([3]) showed that this was equiv- alent to the Carleson condition and later Feffermann-Stein([5]) found a sufficient condition, and Ruiz([13]), Ruiz-Torrea([14]) unified all these

Received January 15, 2000.

2000 Mathematics Subject Classification: 42B25.

Key words and phrases: maximal operator, Orlicz norm.

This research is supported by Kyungpook National University Fund, 2000.

(2)

614 Yong Mal Kim and Yoon Jae Yoo

results. Recently, Gallardo([6]) and Chen([4]) obtained characteriza- tions in terms of the Orlicz norm andin[8], we obtained a characteriza- tion for the fractional maximal operator. Inthis paper, we characterize a condition for Mep to be of weak type (~b ~2) having four weights in the Orlicz norm. In the next theorem, we shall assume that cp is essentially nondecreasing, Le., there exists a positive constant p for which

cp(t) ~ pcp(s), t ~ s and

lim cp(t) = o.

t-+oo t Our resultis as follows.

THEOREM 3.1. Let Mep! be the potential maximal operator on lR~+l. Mep! is defined by

(1) Z(Q) ~s,

where Z(Q) denotes the sidelength ofQ. Letu,v be weights on ]Rn, W

be a weight on ]R~+l and J.L be a nonnegative measure on ]R~+l. ~1 and ~2 areN-functions with complementary N-functions W1 and W2, respectively. Assume further that~20~11 is convex. Then weak type boundedness, Le.,

(2)

~21[ [ ~2~~~(x,S))dJ.L(X'S)] ~ ~11[1n ~1(CI!(x)lu(x))v(x)dx]

J{(x,s)EJR+ :M<pf(x,s»>..} JJR

holds if and only if

(3) [ ['Y(A,Q) CP(IQI)]

JQW1 CAu(y)v(y) IQI v(y) dy < ,(A,Q) <00

holds for each cubeQ, where (4)

'Y(A,Q) = ~1 o~21

[,4

~2(AW(X,S)) dJ.L(X,S)] , Q =Qx (O,Z(Q)].

(3)

When rp(IQJ) = 1, the Hardy-Littlewood maximal operator is ob- tained. The fractional maximal operator Ma(O < 0: < n) is given by rp(IQJ) = IQI~· Maximal operators connected to the Bessel poten-

~

tial operator are defined by rp(jQI) = fJQ'n 'I/J(s) ds, where 'l/J is the derivative ofrp.

2. Preliminaries

DEFINITION 2.1. Let <P : [0, (0) ---t IR. be a function satisfying (i) <p(s) > 0 for all s~ 0;

(ii) lims-+o<p(s)js= 0;

(iii) lims-+oo<p(s)js= 00.

Then <P is called an N-function. Each N-function has the integral representation: <p(s) = f; <p(t) dt, where <p(s) > 0 for s > 0,<p(0) = 0 and <p(s) ---t 00 as s ---t 00. Further, <p is right-continuous and nonde- creasing. <p is called the density function of<P. Define p : [0,(0) ---t IR.

bypet) =sup{s : <p(s) ~ t}. Thenpis called the generalized inverse of

<p. Finally, define

'11(t) =

it

pes) ds

and'11 is called the complementary N-function of<P. For further details, see [10].

DEFINITION2.2. An N-function<P is said to satisfy the .6.2-condition

. [0 )'f <1>(2s)

m ,00 1 sUPs>o <1>(s) < 00.

REMARK 1. If ijJ is the density function of <P, then <P satisfies the .6.2-condition if and only if there exists a constant 0: > 1 such that s<p(s) < o:<P(s), for anys> o.

REMARK 2. If'11 is the complementary N-function of <P, then st ~

<p(s)+'11(t) for all s, t ~ o. Futher, = holds if and only if<p(8-) ~ t ~

<p(s) or else p(t-) ~ s~ pet).

(4)

616 Yong Mal Kim and Yoon Jae Yoo

DEFINITION 2.3. Let(X, M,f-L)is a (I-finite measure space and~be an N-function. Then the Orlicz space£,ip(df-L) and£,4.(df-L) are defined by

£,ip(df-L) = {f:

Ix

~(lfl)d/-L <oo}

and

£,4.(df-L) = {f : fg E £'1(df-L) for all 9 E £,\f!} , where Wis the complementary N-functionof~.

Keeping these definitions and notions, the following properties about the Orlicz space will be used in the proof of the Theorem 3.1.

PROPOSITION 2.4.

(i) The Orlicz space £,4.(df-L) is a Banach space with the Orlicz norm

IIfllip=sup { /Ifgldf-L :9 ES\f! } ,

where S\f! = {g E£,\f! :I W(/gl) d/-L ~ I}, or with the Luxem- burg norm

IIfll(ip) =inf{A> 0: / ~('{I)df-L ~ I}.

(ii) (Holder's inequality) Hf E £,4.(df-L) and9 E £,; (df-L), then

(5) IIfgllip ~ 211fll(ip)lIgll(\f!)'

(iii) (Young's inequality)

(6) ab~ ~(a)+W(b) for all a, b>O.

LEMMA 2.5. Let~ be an N-function with complementary function W. Let x and y >O. Then

(7) ~(x) ~ x<jJ(x) ~ ~(2x),

(8) and (9)

~(x) +~(y) ~ ~(x+y)

~ [W~x)] ~ W(x).

(5)

LEMMA 2.6. Suppose that f is an integrable function in ~n. For each A>0, let EA = {(x, s) E ]R~+1 : Mcpf(x, s) > A}. Then, ifEA is not empty, wehave

(10)

(11)

whereQj is the family of nonoverlapping maximal dyadic cubes satis- fying

~ < <p(IQjl) r f(y) dy·< ~

4np IQj! }Qj - 2n

for each integerj. Furthermore, wehave that

Proof. Following [7](p.160), we let CA = {Pj} be the family of the dyadic maximal nonoverlapping cubes satisfying the condition

A< <p(!Pjl) r f(y) dy.

!Pjl }Pj

To show that there is such a familyCA' observe that

<p(jQI) r f(y) dy ---. 0

IQI }Q

as Qt ]Rn, since f is integrable and since limt-+oo<p(t) / t = O. If, for some dyadic cube Q,

then Q is contained in dyadic cubes satisfying this condition, which are maximal with respect to the inclusion. Thus, there is a family of maximal nonoverlapping dyadic cubes{Pj } yield

(12)

(6)

618 Yong Mal Kim and Yoon Jae Yoo

where Pj denotes the only dyadic cube containing Pj . From this dis- cussion, it is clear that

{x E:!Rn : M;f(x) >>'} = UPj •

j

Let (x, s) E EA; by definition, there is a cube R containing x with l(R) 2: s such that

>. < cp~~I)

L

f(y) dy.

Let k be the unique integer such that 2(k+l)n < IRI ~ 2- kn . There are some dyadic cubes with side length 2-k, and at most 2n of them, {Ji :i = 1, ... ,2n} meeting the interior ofR. It is easy to see that, for one of these cubes, say Jl,

~ < cp(IRI) ( f(y) dy.

2n IRI JRnh

Now, since IRI ~ IJ11< 2nIR!, cp(IRI) ~ PCP(IJil) and

4:IJ11<

2:

IRI < cp(IRI) ( f(y) dy ~PCP(IJ11) ( f(y) dy.

JRnJ1 JJ1

Hence,

~ < cp(!J11) ( f(y) dy.

4np IJ11 lh

By letting Ct/4np = {Qj}, we see that J1 C Qk, for some k, and x ERe Jp c Q~.

Onthe other hand, s ~ l(R) ~ l(QJ)' From this, we conclude that

EA c UQ;'

j

Finally, it follows from (12) that

for eachj, concluding the proof of the lemma. o

(7)

3. Main Results

THEOREM 3.1. Let Mepj be the potential maximal operator on R~+l. Mepj isdefined by

<p(IQI) r

Mepj(x,s) = ~~g IQI lQ Ij(y)1 dy, where l(Q) 2: s.

Let U and v be weights on Rn, w be a weight on R~+l and J.L be a nonnegative measure on R~+l. epl and ep2 are N -functions with complements q,1 and q,2, respectively. Assume further thatep2 0 epl 1 is convex. Then weak type boundedness, i.e.

holdsifandonly if

r ['Y(A'Q) <P(IQI)] -

JQ q,1 CAU(X)V(x). IQI v(x) dx ::; 'Y(A, Q) < 00

holds for each cubeQ, where

Proof. For the necessity, we follow the idea of [2]. Since \111;10) is increasing in c and has full range R+, for given A> 0, we can choose c such that

r ( c ) v(y) IQI

JEq,l u(y)v(y) -c-dy =2CA<p(IQI) , EcQ

1 ( c ) v(x)

j(x) = Cq,1 u(x)v(x) -c-XE(x).

(8)

620 Yong Mal Kirn and Yoon Jae Yoo

If(y,s) E Q, then yE Q ands::; l(Q). So

cp(IQI) (

M<p!(y, s) = =~g IQI J

Q1!(y)1dy 2: cp~~I)

k

1!(y)1dy

cp(IQD {I ( g ) v(y)

= IQI JQ Cqrl u(y)v(y) -g-XE(y) dy

= 2A> A.

Thus if(y,s) E Q, then (y,s) E EA ={(x,r) E lR~+l : M<p!(x,r) >

A}. Hence

,(A,Q) = <PI 0 <P21

[k

<P2 (AW(X, s)) dp,(x,S)]

::; <PI 0 <P21[ { <P2(AW(X, s)) dp,(x, s)]

J{M,.,f>A}

::; L

<PI [U(y~v(y)qrl (u(y~v(y))] v(y)dy

::; L

qr1 (u(y~v(y)) v(y) dy

IQI

= 2CAcp(IQI)g.

The third inequality follows from (2). Put

.6.

-1

- E'l/J (1 4CAU(Y)V(Y)/(A,Q) CP(IQI))IQI _1_u(y) dy.

Then

1 (

g) 1

.6.< 'l/J1 - d y

- E 2u(y)v(y) u(y)

1 (

g) v(y)

::; 2 Eqr1 u(y)v(y) - g -dy

IQI

= 4CAcp(IQlf

(9)

Also

Thus

~ > 4CAIQI r'1J [( ,(A,Q) )<P(IQI)] d

- <p(IQlh(A,Q) lE 1 4CAU(Y)V(Y) IQI v(y) y.

r [( ,(A,Q) )<P(IQI)] -

lE'1J1 4CAU(Y)V(y) IQI v(y) dy ::; ,(A, Q) < 00.

As E -+ Q, (3) holds.

For the sufficiency, we follow the idea of [4] and [11]. From Lemma 2.6, for each A> 0, let E>.. = {(x, s) E JR~+l : M<pf(x, s) > A}. Then, ifE>.. is not empty, we have

where Qj is the family of nonoverlapping maximal dyadic cubes satis- fying

~ < <p(lQjl) r f(y)d < ~

4np IQj! lQj y - 2n

for each integer j. Then 4~P < <p~~~j) IQj f(y) dy and {QJ} is a cov-

ering ofE>... Hence it follows from (6) that

2,(A,QJ)::;

hj

4np!{(X)I<P~~11)2'(A,QJ)dX

r n n ,(A, QJ) <p(IQJI)

= lQj23 4 Cpjf(x)lu(x) CAU(X)V(x) IQJI v(x) dx

::; hj

<PI(23n4nCplf(x)lu(x))v(x) dx

r (,(A, QJ) <P(IQJI))

+lQj '1J1 CAU(X)V(x) IQJI v(x) dx

::; hj

<PI(23n4npClf(x)lu(x))v(x) dx +,(A, QJ).

(10)

622

So

YongMal Kim andYoonJaeYoo

and thus

k~(<P2(AW(X,s)))dp,(x,s) ~<P20 <PII[£j <PI (23n4nCplf(x)lu{x) )v(x)dx].

Summing overj gives

By using that <b20 <b11 is convex, this last sum is bounded by

<b20 <b11[~

k

j

<b1(23n4nCplf(x)lu(x) )v(x) dX]

::; <b2 0 <b11[C~Ln<b1(23n4ncplf(x)lu(x) )v(X) dx

1

::; <b2O<b11[Ln <b1 (Cnlf(x)\u(x))V(X)dX]. D COROLLARY 3.2. From (3), put w = U = 1,<b1 = <b2. Then (3) gives

(13)

[IL(Q)CP(,Q')]4>fcp(\QI) f 1/1(_C_)dy ] ::;Cc, for each c> o.

IQI ~ CIQI JQ v(y) .

Proof We follow the proof of Corollary 3.2 of [8]. o

COROLLARY 3.3. Hip10 <b21has thel:!,.' conditionand (3) holds for w = U = 1, then there exist constants C', C" > 0 such that, for any

c>0,

-1( (-))cp(IQD [C' f (C)] "

(14) <b10 <b2 JL Q IQI 4>1 1QIJ

Q1/11 v(y) dy ::; C c.

(11)

Proof. We follow the proof of corollary 3.3 of [8]. o

EXAMPLE 3.4. Let cp(IQI) = 1 and c= t. From Corollary3.2, (13) gives A~ of [4], Le.

But (13) is weaker than A;, because an N-function <P of A; in [4]

satisfies the ~2-condition.

EXAMPLE 3.5. From Corollary 3.3, put dp,(x, t) = u(x)dx0 d8(t), where 8 is the Dirac mass on [0,00), concentrated at O. Also set

<Pl(X) = x; and <P2(X) = xq

q

, where 1 < p ~ q < 00. Then (14) gives (u, v) E A(cp,p, q) of [12], Le.

cp(IQI) IQlt-i[I~Iku(x)dx]t L~Ikv(x)-P!-l dxr-i ::;A for all Q.

References

[1] S. Bloom and R. Kerman, Weighted£if> integral inequalities for operators of Hardy type, Studia Math. 110(1994),35-52.

[2] _ _ , Weighted Orlicz space integral inequalities for the Hardy-Littlewood maximal operator, Studia Math. 110 (1994), 149-167.

[3] L. Carleson, Interpolation by bounded analytic functions and the corona prob- lem, Ann. Math. 16 (1962), 547-559.

[4] Jie-Cheng Chen, Weighted and .cif>-boundedness of the Poisson integral opera- tor, Israel J. Math. 81 (1993), 193-202.

[5] C. Fefferman andE. M. Stein, Some maximal inequalities, Amer. J. Math. 93 (1971), 107-115.

[6] D. Gallardo, Weighted weak type integral inequality for the Hardy-Littlewood maximal operator, Israel J. Math. 61 (1989), 95-108.

[7] J. Garcia-Cuerva and J. L. Rubio De Francia, Weighted Norm Inequalities and Related Topics, Elsevier Science, North-Holland, 1985.

[8] Yong Mal Kim and Yoon Jae Yoo, Weighted Orlicz space integral inequalities for fractional maximal operators, Far East J. Math. Sci. 5 (1997), no. 2, 225- 236.

(12)

624 Yong Mal Kim and Yoon Jae Yoo

[9] B. Muckenhoupt and R. L. Wheeden, Weighted norm inequalities for fractional integrals, Trans. Amer. Math. Soc. 191 (1974), 261-274.

[10] J. Musielak, Orlicz Spaces and Modular Spaces, Lecture Notes in Math. 1034 (1983).

[11] C. P'erez, Two Weighted Inequalities for Potential and Fractional Type Maxi- mal Operators, Indiana University Mathematics Journal 43(1994), no. 2.

[12] Y. Rakotondratsimba, Weighted Norm Inequalities for Maximal Functions from the Muckenhoupt conditions, Publications Matem'atiques38 (1994), 25- 36.

[13] F. Ruiz, A unified to Carleson measures and Ap weights, Pacific J. Math. 117 (1985), 397-404.

[14] F. J. Ruiz and J. L. Torrea, A unified approach to Carleson measures and Ap weights Il, Pacific J. Math. 117(1985).

Department of Mathematics Education Kyungpook National University

Taegu 702-701, Korea

E-mail: [email protected]

참조

관련 문서

“With the MySQL Query Analyzer, we were able to identify and analyze problematic SQL code, and triple our database performance. More importantly, we were able to accomplish

Quasi-entropy was introduced by Petz in 1985, [22] as the quantum generalization of Csisz´ ar’s f -divergence in the setting of matrices or von Neumann algebras... We

A 27-year-old man with spondyloarthropathy, oblique coronal fat-saturated T2-weighted (A) and oblique coronal postcontrast fat-saturated T1-weighted (B) images show

Utility of T1-and T2-weighted high-resolution vessel wall imaging for the diagnosis and follow up of isolated posterior inferior cerebellar artery dissection with

[표 12] The true model is inverse-gaussian, out-of-control ARL1 and sd for the weighted modeling method and the random data driven

Min Seok Ko, Bo An Lee, Sung Yong Lee, Yeong Jun Jang, Yoon Jeong Hong, and Sin Kim(JNU) TR-PIV Performance Test for a Flow Field Measurement in a Single Rod Test Section Ju

IL Soon Hwang, Myung Hyun Kim, Han Gyu Joo, Kyung Woo Yi, Bong Yoo, Moo Hwan Kim, Seung Rok Oh, Yoon Jae Kim, Jong Gye Shin, Kwang Myung Lee, Jae Yong

We have already looked at this problem from one standpoint for which we found the value of the potential was doubled at the surface and the expression for the