APP下载

Some Weighted Norm Inequalities on Manifolds∗

2018-11-19ShiliangZHAO

Shiliang ZHAO

Abstract Let M be a complete non-compact Riemannian manifold satisfying the volume doubling property and the Gaussian upper bounds.Denote by∆the Laplace-Beltrami operator and by∇the Riemannian gradient.In this paper,the author proves the weighted reverse inequality,for some range of p determined by M and w.Moreover,a weak type estimate is proved when p=1.Some weighted vector-valued inequalities are also established.

Keywords Weighted norm inequality,Poincaré inequality,Riesz transform.

1 Introduction

LetMbe a complete Riemannian manifold,∆be the Laplace-Beltrami operator,and∇be the Riemannian gradient.Denote bythe Riesz transform onM.To start with,we recall some facts about Riesz transform on the Euclidean space Rn.Laplace-Beltrami operator on Rnis defined by∆=−The Riesz transform can be written as=(R1,R2,···,Rn),whereRj=∂xjfor 1 ≤j≤n.By the classical Calderón-Zygmund theory,there exists constantCp>0,such that for allf∈,

where 10,such that for allf∈,

(see for example[10,Subsection 2.1]).

As for manifolds,Strichartz[22]proved that(1.1)holds for 1

On the other hand,the weighted norm inequalities for singular integral operators have a long history and are of great interest in harmonic analysis(see[20,Chapter 5]for more details).In the series[4–7],Auscher and Martell established some criteria to prove the weighted norm inequalities.In particular,they considered the Riesz transform,the reverse inequalities,and some quadratic operators of Littlewood-Paley-Stein type associated with elliptic operators in[5].In[6],they showed the weighted estimates of Riesz transform on manifolds(see[9]for the corresponding results on graphs).In[8–9],these criteria were used to prove theLpboundedness of Riesz transform associated with Schrödinger operators.

In this paper,letMbe a complete non-compact Riemannian manifold withdits geodesic distance,andµbe its Riemannian measure.Moreover,assume that it has the volume doubling property,i.e.,there exists a constantC>0 such that

for allx∈Mandr>0,whereB(x,r)is the geodesic ball centered atxwith radiusr,andV(B)is the volume ofBwith respect to the Riemannian measureµ.For anyλ>0,denoteλB=B(x,λr).The volume doubling property implies that there existC,ν≥ 1 such that,for every ballBandλ>1,

Let∆be the Laplace-Beltrami operator onM.Denote bypt(x,y)the heat kernel of the semigroup e−t∆,wheret>0 andx,y∈M.It is said to have Gaussian upper bounds if there exist some constantsC,c>0 such that for allt>0,x,y∈M,

Note also thatMis stochastic complete whenMis geodesic complete and has volume doubling property(see[16,p.303]).That is

Now we recall the definition of Muckenhoupt weights and reverse Hölder classes in[3].

Definition 1.1We say that a nonnegative locally integrable function w belongs to the Muckenhoupt Ap-weight classes Ap(µ)on M for1

and for p=1if

We say that a nonnegative locally integrable function w belongs to the reverse Hölder classesRHs(µ)with exponent s>1if there exists a constant C such that,for every ball B,

The endpoint s=∞is defined as follows.For every ball B,

For properties of MuckenhouptAp-weightsAp(µ)and reverse Hölder classes RHs(µ),we refer the reader to[3,20].To proceed,we need the following definitions in[3].

Definition 1.2Given w∈Ap(µ),define

Then for1≤p0

where q′=is the conjugate exponent to q.Set

In[6],the weighted estimates of Riesz transform on manifolds were proved.

Theorem ALet M be a complete non-compact Riemannian manifold satisfying the volume doubling property and Gaussian upper bounds.Let w∈A∞(µ).

(i)For p∈Ww(1,q+),the Riesz transform is of strong-type(p,p)with respect to wdµ,that is,

for all f bounded with compact support.

(ii)If w∈A1(µ)∩ RH(q+)′(µ),then the Riesz transform is of weak-type(1,1)with respect to wdµ,that is,

for all f bounded with compact support.

We will prove the weighted estimates for the reverse inequality.Before we state the main result of this paper,recall thatMsupports ap-Poincaré inequality for 1 ≤p<∞ if there existsC>0 such that,for every ballBand every locally Lipschitz functionf,

whereris the radius ofBandfB=(V(B))−1RBfdµ.We define

Note that the unweighted estimates were proved in[1,Theorem 0.7].

The main result of this paper is as follows.

Theorem 1.1Let M be a complete non-compact Riemannian manifold satisfying the volume doubling property and Poincaré inequality with1 ≤r−<2.Let w∈A∞(µ).

(i)For p∈Ww(r−,∞),there exists Cp,w>0such that

for all f smooth with compact support.

(ii)If1-Poincaré inequality holds,then for every w∈A1(µ)there exists C1,w>0such that

for all f smooth with compact support.

To prove Theorem 1.1,we need to estimate the operatorSdefined by setting for anyh:M×(0,∞)→R,

Moreover,for anyp∈ [1,∞],let

where|h|H(x)=for anyx∈M.Denote by(µ)the function space consisting of bounded functions with compact support with respect toµ,and set

Then we have the following theorem.

Theorem 1.2Let M be a complete non-compact Riemannian manifold satisfying the volumedoubling property and Gaussian upper bounds.For any h∈L∞c,H(µ),we have

(i)If10such that

(ii)If p=1and w∈A1(µ),there exists C1,w>0such that

Now we introduce some notations.Given any ballB,denoteC1(B)=4B,Cj(B)=2j+1B2jBforj≥ 2.Denote byχEthe indicator function of subsetEofM,and setfj=fχCj(B)for any given functionf.The constantsC,candc′may change from line to line.

2 Proof of Theorem 1.2

To begin with,we recall some properties of the operatorSin the unweighted cases.Note that the Littlewood-Paley-Stein square operator forf∈Lp(µ)is defined by

Since e−t∆is a symmetric diffusion semigroup in our settings,Gis bounded onLp(µ)for 1

In fact,for anyϕ∈(M),we have

where 1

Before proving Theorem 1.2,we need to prove some lemmas.Denote by(µ)the function space consisting of locally integral functions with respect toµ.

Lemma 2.1Given any fixed l≥1,there exist C,c>0,such that the following holds forevery ball B=B(x0,r),h∈(µ)and j≥ 2:

ProofWe first deal with|e−lr2∆hj(x)|.According to the Gaussian upper bounds of the heat kernel,one gets

where the third and fifth inequalities follow from(1.2).Here and in what follows,we denoteV(B(x,r))byV(x,r).Note that∆e−t∆f= −f.By the Gaussian upper bounds of the time derivative of the heat kernel(see for example[15–16,19]),

Remark 2.1Given anyl≥1 fixed,there existC,c>0,such that the following holds for every ballB=B(x0,r),h∈(µ)supported inBandj≥ 2:

Observing thatV(x0,r)≤Cthe remark follows according to the proof of the lemma.

Lemma 2.2There exist C,c>0,such that the followings hold for every f∈L1(µ)and a>0:

ProofSinceBy(2.2),(i)follows.

Note that

According to(2.2),there existsC>0 such thatfort>0 andf∈L1(µ).As a result,one gets

In order to prove(1.6),we need the following vector-valued extension of Theorem 3.3 in[6].

Lemma 2.3Let1≤p00be a family of operators acting from(µ)(µ).Assume the following conditions:

(a)There exists q∈Ww(p0,q0),such that T is bounded from(w)to Lq,∞(w).

(b)For all j≥ 1,there exists a constant αj,such that for any ball B with r(B)as its radiusand for any f∈(µ)supported in B,

(c)There exists β>(sw)′,i.e.,w∈ RHβ′(µ),with the following property:For all j≥ 2,there exists a constant αj,such that for any ball B with r(B)as its radius and for any f∈≥ 2,

If w∈then there exists a constant C>0such that for all f∈

ProofThe proof of this lemma is similar to that of[7,Theorem 3.3].Since we have the vector-valued Calderón-Zygmund decomposition,the methods in proving[7,Theorem 3.3]apply well.

Proof of(1.5)in Theorem 1.2Fix 1

We will prove(II) first.Note thatSAr=ArS.Then expand Arand lethj=(Sf)χCj(B).Forj≥2,apply Lemma 2.1.Sinceh1is supported inC1(B),it is a direct result of the Gaussian upper bounds of the heat kernel and(1.2).

To prove(I),let

We will treatf1andfj(j≥2)separately.

According to(2.1)withp=p0,one obtains

Since the variablexoff1is supported inC1(B),it follows that

For 1≤l≤n,one has

Apply Remark 2.1 wheni≥2 and Gaussian upper bounds wheni=1.Then one gets

where we have used(1.2).Thus we have proved(I)forf1.

To treatfjforj≥2,we use the Minkowski integral inequality and get

For I1,we expand(I−e−r2∆)n,and Lemma 2.1 gives

Since

we have

For I2,from Lemma 2.2,we have that for anyx∈B=B(x0,r)andt>0,

We have used the fact thatin the second line.Substituting the above estimate into I2and using Hölder inequality,one gets

Combining the above estimate and the estimate ofI1,we obtain that(1.5)holds providedn>.

Proof of(1.6)in Theorem 1.2Takep0=1,q0= ∞,Ar=I−(I−)n.We will check the conditions(a),(b),(c)and(d)required by Lemma 2.3.

Fixw∈A1(µ).(a)follows from(1.5)as the operatorSis bounded from(w)toLq(w)for 1

By expanding Ar,it is sufficient to show(b)for e−lr2∆with 1 ≤l≤n.Fixing anyl,we have forj≥2,

Forj=1,it follows by the Gaussian upper bounds of the heat kernel and(1.2).

To see(c),we use the same method as in proving(1.5)and identifyαj=2−j(n−ν2)forj≥ 2.In fact,takeβ= ∞.Then forj≥ 2,

For I1,using Remark 2.1,we obtain

For I2,since the following inequality holds with everyt>0:

we have

As a result,we can conclude that(d)holds ifn>νw+,whereνwis the doubling constant of the measurewdµ.

Thus we have finished the proof of Theorem 1.2.

3 Proof of Theorem 1.1

In order to prove(1.4),we need a weighted version of the Calderón-Zygmund decomposition.

Lemma 3.1Let M be a complete non-compact Riemannian manifold satisfying the volume doubling condition,and it supports1-Poincaré inequality.Let w∈A1(µ).For any function f such that|∇f|∈L1(w)and λ>0,one can find a collection of balls(Bi)i,functions(bi)iand g,such that

where C and N depend on the constants in Poincaré inequality,νand w.

ProofAccording to[3,Proposition 9.1],the results hold when we have the following weighted Poincaré inequality:

Proof of(1.3)in Theorem 1.1We will use Theorem 3.7 in[3]again.Sincep∈Ww(r−,∞),there existr−

Now we deal with(I′).GivenB=B(x0,r),leth=f−f4B,hj=hχCj(B)forj≥ 3,h2=h(χ8B−φ1),h1=hφ1and

wheredis the geodesic distance.Note that|∇φ1|is bounded with compact support.

According to[2],one has the following equalities:

where we have used the stochastic complete property ofMin the first equality and

We have the following estimates ofgr(t)(see[2]):

and

For anyc,r>0,j≥2,

We first estimate I1and one has

whereCdepends onc,n.On the other hand,

whereCdepends onc,n.Then,combining the estimates of I1and I2,we get that there existsC>0 depending onc,n,such that for everyr>0,

With the help of Poincaré inequality,we obtain

In the forth inequality,we have used the fact

Thus we have

The sum of the right-hand side converges whenn>.

Forj=1,we have

Note that

andφ1is bounded with compact support.By Poincaré inequality,we have

Forj=2,notice thath2is supported in 8B/2Band we have a similar estimate forh2in the spirit of Lemma 2.1.Using the similar method as in the proof forj≥3,the result forj≥3 also holds forj=2 and this leads to(I′).Then we have proved(1.3).

Proof of(1.4)in Theorem 1.1We follow the method in[1].Forfsmooth with compact support,apply Lemma 3.1.Sincew∈A1(dµ),we havew∈Aq(dµ)for anyq>1.By(1.3)and the property ofg,we have the following:

Letri=2kif 2k≤r(Bi)<2k+1,and set

Then it is enough to estimate

For I1,we get

We have

By Remark 2.1 and the properties ofbi,we obtain

As a result,we have

where we have used the definition ofA1(µ)in the third inequality.

Hence,

It remains to handle I2.Define

Then we have

where

By Cauchy-Schwarz inequality,it gives

Thus we have

SinceSis a linear operator and(µ)is dense in(w),(1.6)holds for allh∈(w).Then by Theorem 1.2,we get

which,together with the estimate ofI1,implies(1.4).Thus we have proved Theorem 1.1.

4 Appendix

In this section,we give[3,Theorem 3.7]which plays an important role in this paper.

Given two vector spaces of measurable functionsAandB,we say that an operatorTacts fromAintoB,ifTis a map defined onAand valued inB.An operatorTacting fromAtoBis sublinear,if

for allf,g∈Aandλ∈ R or C.LetMbe the Hardy-Littlewood maximal operator.[3,Theorem 3.7]is stated as follows.

Theorem 4.1(see[3,p.234,Theorem 3.7])Let1≤p00be a family of operators acting fromDintoE.Assume that

and

for all f∈D,for all ball B,where r(B)denotes its radius and all x∈B.Let p0

for all f∈D.Furthermore,for all p0

for all fj∈D.

As emphasized in[3],(4.1)–(4.2)are unweighted assumptions by which one can obtain the weighted inequalities as well as the vector valued estimates.

AcknowledgementsThis work is done during the author’s visit to Macquarie University.The author thanks Professor Duong Xuan Thinh and Professor Li Hong-quan for their generous help.The author also thanks the referees for their careful reading of the manuscript and for their constructive and detailed comments.


登录APP查看全文