GENERALIZED CARLESON INEQUALITY ON SPACES OF HOMOGENEOUS TYPE
YOON JAE Yoo*
1. Introduction
The purpose of this paper is to generalize the Carleson inequality, which is known to play important roles in harmonic analysis. The result given here is a generalization of Coifmann, Meyer, Stein [CMS].
Asimilar result is shown by Deng[D].
2. Preliminaries
Let (X, d,p.) be a space of homogeneous type, i.e.,X is a topological space, p. isa positive Borel measure onX, and disapseudo-metric on Xj more precisely, we assume that there are constant A and K so that for all x, y,z E X and all 8>0,
i) d(x, y) ~ 0;d(x, y) = 0 ifand only if x= Yj ii) d(x,y) = d(y,x)j
iii) d(x,z):5 K[d(x,y) +d(y,z)], iv).V(B(x,28) :5 AI'(B(x,8)),
where B(x,8) = {y EX: d(x, y) < 8}. We also assume that the collection {B(x, 6)} forms a basis for the topology ofX. Property iv) is referred to the doubling property of p.. For the details, see [CW].
Let X+ = X X(0,00), which is an analogue of a generalized upper- half space over X. Let r(x)'; ((y,s) E X+ :x E B(y,s)}. This is an analogue of nontangential or conical region. For any set E eX, the tent over E is the set
T(E) = {(x,r) EX+ :B(x,r) c E}.
Received August 12, 1994.
Key words: Carleson measure, Carleson inequality.
*This researchissupported bythe research grant of KOSEF 1992.
It is. then easY tov:erify that T(E) = IUxiEreX)]c. For a measurable functionI defined onX+, fot xEX,p;::: 1 and a ER,let
1
[ 1
i .
dP.(Y)dt]PtTp,,,,(f)(x) = sup (B(t5)) I/(Y, t)IP tOl '
xEB(6)p. T(B(6»·
where B(8) is a ball with radius 8> 0. Let M(f) denote the Hardy- Littlewood's maximal function
M(f)(x) = sup (;(8» f I/(y)ldp.(y).
xEB(6)p. JB(6)
For a positive Borel measureAonX+ = X x [0,00),if there exists a constant C» 0 such that A(T(B(8»)) :::; Cp.(B(8)) fdr any <5>0, then Ais called a Carleson measure. For a measurable function I on X+, we define a p-areafunctio~ Ap,OIas follows:
1
Ap,01(f)( x) = [ ( I/(Y, t)IPd~~~dt] P, if 1:::; p<00, a E R
11'(x)
and
Aoo,OI(f) = sup I/(y,t)J, if p=00.
(y,t)eI'(x)
For a nonnegative measurable function I on X+, if there exists a
constant C such that .
1 (Idp. < C inf I(x) p.(B(8» 1B(6) - xEB(6)
for all 8 > 0, then I is said to belong to the class Al [M].
3. Generalized Carleson inequality We begin with lemmas.
LEMMA 1 (COVERING LEMMA OF VITALI-WIENER TYPE). Let X be a space of homogeneous type. Let E c X be a bounded set i.e., E is contained in some ball. Let r(x) be a positive number fot each x E E. Then there is a sequence of balls B(Xi, r(Xi», Xi E E, such tbat the balls B(xi,4Kr(xi» coverE, wbere K is the constant in tbe triangle inequality. Furthermore, every x EE is containedin some ball B(xj,4Kr(xi» satisfying r(x) :5 2r(xi)'
(See Theorem 1.2 on p.69 of [S])
PROPOSITION 1. Letp~ 1. Tben Tp,a is of weaktype (p,p).
Proof. Fix A>O. Set
{lA = {x EX:Tp,a(f)(x) >A}.
Assume first that !lA is bounded. Apply lemma 1 to obtain a sequence of disjoint ball B(xi,ri) so that !lA C UB(Xi,4Kri) and
1 f If(y,t)lpdJl(y)dt > AP.
Jl(B(Xi, ri» IT(B(xi,r;}) ta
Then from the doubling property of Jl it follows that Jl(!lA) :5 2:Jl(B(Xi, 4Kri»
:5 C2:Jl(B(Xi,ra)
:5 ~ f If(y,t)IPdJl(~)dt
A IT(B(z;,r;)) t
:5 ~
L+
If(y, t)IPdJl;~)dt.If!lA is not bounded, fix a E X and R > O. Then!lAnB(a, R) is a bounded set, so we can apply the above argument to obtain
Jl(!lAnB(a,R»:5 ~ f If(y,t)IPdJl(y)dt.
/\p Jx+ ta
Letting R --+ 00 we obtain the same estimate. This completes the proof.
LEMMA 2. Let (Xl,JL1) and (X2,JL2) be measure spaces and T be a sublinear~pei'ator that transfoI'i:ds measurable· fupctionsIDeX1,JL1)
into measurable functionsin(X2,JL2). HT is of weak type(p,q), then
[L\T!(y)\rdJL2]·1r :5CJL2(K)::-tllfllp
holds forall r,O < r < q, and for any subset K of finite measure.
THEOREM 1. Let f ELI( dradt
),0:5 € < 1. Then TI,a(f)E E Al for anya E R.
Proof. Fix B(60 ) andXo EB(60 ). Put
A2 = {B(6) :Xo E B(6),6 > 60 }.
Then we have
(1)
where
and
1
l
dJL(y)dtTI,a(f)(xO):5 sup (B(6)) If(y, t)l ta
B(6)€A1 J.L T(B(6»
1 1" If( t)ldJL(y)dt
+ sup y, a
B(6)€A2 1-'(B(6)) T(B(6» t
:=A{x(}.)+B(xo},
1
l'
dp(y)dtA(xo)= sup' (B(6)) \f(y, t)1 ta ' B(6)€A1 JL . T(B(6»
To estimate B(xo), suppose B(o) E A2 • Since B(o)nB(oo) :F 4> and
0> 00, there exists a constant C ~ 1 such that B(Co) :) B(oo). Hence
1 I dp.(y)dt
p.(B(o» 1T(B(6» If(y,t)1 tQ
< p.(B(Co» 1 I I f ( )Idp.(y )dt - p.(B(o» p.(B(o» 1T(B(6» y,t tQ
~ C inf T1cif)(x) :rEB(C6) ,
~ C inf T1,a(f)(x), :rEB(60)
and so
(2) B(xo) ~ C inf T1,a(f)(x).
:rEB(60)
As forA(xo),note thatB(Coo) :) B(o)for some constantC. Indeed, since B(oo)nB(o) :F 4> and 0:500, B(o) c B(Coo)for some constant C by the property of the pseudo-metric. Thus T(B(o» c T(B(Coo»,
and so ifwe put h(y, t) =!(y, t)XT(B( Cc5o»(y,t), then we have (3) A(xo)= sup 1 I I!(y,t)1dp.(y)dt
B(6)eA2 p.(B(o» 1T(B(6» tQ
< sup 1 I If (y t)1dp.(y)dt - B(6)EA2 p.(B(o» 1T(B(6» 1 , t a
~ T1,Q(fd(x).
Since 0 ~ f <1,
(4) T1,Q(f)(xo)f ~ A(XO)f +B(XO)f.
Hence it follows from (1),(2),(3), and (4) that p.(;(0» j T1,a(f)(xO)fdp.(xo)
(B(60»
(5) ~ (B~O» I A(XO)fdp.(xo) +C inf T1,a(f)(xr
p. 0 1B(60) :reB(60)
~ (B~O» I T1,aUd(xrdp.(x) +C inf T1,Q(f)(X)f
p. 0 1B(60) xeB(60)
By lemma 2, the first term. of the right-hand side of (5) is less than
[
1 f dP.(Y)dt]E
C p.(B(So» 1B(06
0) If(y,t)1 tOt and so (5) is reduced as
which proves theorem.
LEMMA 3(WHITNEY DECOMPOSITION). LetX bea space ofhomo- geneous type. Let 0 c X be an open set. Then there are positive constants M,Cl,C2,Ca, andasequence {B(xi,6i)} of balls such,that
(i) UiB{XhSi) = 0
(H) B(Xi'C16i ) c 0 andB(Xi, C26i)nQC=I- cP,
(iii) the balls B(Xi,CaDi) are pairwise disjoint;
(iv) no pointin0 liesin more than M of the ball B(xi,6i).
(See [eMS].)
THEOREM .2. Let i + ~ = 1, 1 ~ p ~ 00. Suppose there exist constants A > O,B > 0 such that A6n ~ p.(B(x,6» ~ B6n for all S> o. Then there exists a constant C such that
{ dp.(y)dt (
1x+ f(x, t)v(x, t) tOt ~ C1xAp,n_I+Ot(f)(x)Tq,Ot(v)(x)dp.(x)
for all !,v.
Proof Assume first that 1~P<00. For each integer k, set
and 1
. nk={x EX: M(Xnk)(X) >'2}.
Observe that Ok ~.nk+landnk :) fik+1 for each k. Further U~_oo
T(nk) contains the support off. By Lemma 3, there exist collections
of disjoint balls B(c5k,j) and B(tS'k) such that fh C UjB(ctSk,j) and Ok C UjB(cc5k,j). Also we can choose a constant cso that T(Ok) C UT(B( cc5k,j)) and thatT(flk)C UjT(B(cc5'k,j))' PutIk,j =T(B(cc5k,j)) and Jk,j = T(B(cc5'k)) for simplicity. Then it follows from Holder's inequality that
li+
f(y, t)v(y, t)dp.~~)dtI< f: f f(y, t)v(y, t)dJ.l~~)dt
k=-ooIT(O;l\T(O;+I)
(1) < ,t;oo
t,
L.,;lu.J.+". !(y,t)v(y,t)dl';~)dt00 00 1
:S k~OO; [lk,j\u.Jk+1
,. If(y, t)IPdJl~~)dt] p
[jJk,j\u.Jk+1 ,. Iv(y,t)19dJ.l~:)dt] q.1
Put
and
IT we show
and
then "we have
I"( fey,t)v(y, t) dp.(:)dtI·
lx+ t··
(4)
Since Tq,a( v)E AI, it follows from proposition 1 that
and so
(5) L~Tq,a(v)dp. $ 4
L
M(Xnk)(x?Tq,a(v)(x)dp.(x)$ C
L
Xnk(X)2Tq,a(v)(x)dp.(x)= C ( Tq,cr(v)(x)dp(x).
lnk "
Therefore by (5), (4) can be rewritten as
ILJ(y, t).(y,t)dP~~)dtIS c,=f;~2'fa: T".(')dp
(6) $ C f 2k
1
Tq,a(v)dp.k=-oo n~
=C (·lxAp,n-I+a(f)Tq,~(v )dp(x).
So ifwe show (2) and (3), then (6) holds for 1~p <00.
To see (2), observe that (7)
f Ap,n-l+a(f)(x)Pdl-t(X)
1B(c6k,j \u, B(c6.+1,,)
= f dp(x) f If(y,t)IPdl-t~~:dt
lB(c6:,j \u,B(c6Hl,,) lr(x) t
=
1
If(y, t)/PXr(X)(y, t)dp(x)n~~Y)dt[B(c6k,j)\u,B(c6.+1,'»)xX+ t
~ f If(y,t)lpdp~~~dt f xr(x)(y,t)dp(x).
1J.,j \u, J.+l" t 1B(c6;,j )\u, B(c6Hl,,) For any fixed (y, t) E Jk,j\ UII Jk+l,8' we have
(8) B(y,t)n(Ok+l)C :I <1>.
In fact, suppose B(y,t) C 0;+1 == UjB(c6k+l,/)' Then (y,t) E T(B(c6kH,j))' for some j, that is, (y, t) E Jk+l,j, which leads to a contradiction. Hence (8) holds. So we can choose a point P E B(y,t) so that M(XOH 1)(X) ~ !' which implies
(9) P(Bt t))y, lB(y,t)f XO.+ 1(x)dl-t(x) ~ -21,
and so by (9), we have
(10) 1 f Xr(x)(y,t)dl-t(x)
I-t(B(y,t))1B(c6Z,j)\u,B(c6.+1,,)
~ 1 f XB(y,t)(x)dl-t(x)
I-t(B(y, t)) 1B(c6j"j)\u,B(c6.+1,,)
~ 1 f XB(y,t)(x)dl-t(x)
I-t(B(y, t))1B(c6i.,j)\u,B(c6.+1 , , )
= (Bt t)) f [1- XB(y,t)nU,B(c6.+1,,)(x)]dl-t(x) I-t y, 1B(y,t)
= (Bt t)) f [1 - XB(y,t)nOHl (x)]dl-t(x) I-t y, 1B(y,t)
>-1 - 2'
In the second inequality of (10), we used the fact B(y, t) C B(cbk,j).
Thus
( Xr(x)(y,t)dp(x) ~ Cp(B(y,t»
JB(coi.,j)\u.B(cbk+l,.)
>Ctn
- ,
which implies (2). To estimate (3), note that
This shows (3).
Finally, assumep = 00. Then
I( fey, t)v(y, t) dp(:)dtI
Jx+ . t
00
1
d()dt~ L If(y, t)v(y, t)! Pt:
k=-oo T(Ok)\T(Ok+t>
f f 2k+l
1
v(y, t)dJL~:)dtk=-oo j=1 Ik,j
~ C f 2kf
1
T1,a(v)(x)dp(x)k=-oo j=1 B(COk,j)
:5 C
L
Aoo(f)(x)T1,a(v)(x)dp(x).This completes the proof.
References
[CMS] Coifman, R. R., Meyer, Y., and Stein, E., Some new junction spaces and their applications to harmonic analysis,J.of Func. Anal.62 (1985), 304-335.
[CW] Coifman, R. R. and Weiss G., Extension of Hardy spaces and their use in analysis, Bull. Amer. Math. Soc. 83 (1977), 569-645.
[D] Deng, D. E., Ona generalized Carleson inequality, Studia Math. 78 (1984), 245-251.
[M] Muckenhaupt,8., Weighted norm inequalities for the Hardy maximal func- tion, Trans. Amer. Math. Soc. 165 (1972), 207-226.
[8] Sueiro,J., On maximal junctions and Poisson-Szego integrals, Trans. Amer.
Math. Soc. 298 (1986), 653-669.
Department of Mathematical Education Teachers College
Kyungpook National University Taegu 702-701, Korea