Boundedness of Iterated Spherical Average on α-Modulation Spaces∗

2021-11-03 12:07:42QiangHUANGXiaomeiWU
Chinese Annals of Mathematics,Series B 2021年5期

Qiang HUANG Xiaomei WU

Abstract For the iteration of spherical average (A1)N and the Laplace operator Δ, we consider the boundedness of the operator Δ(A1)N on the α-modulation spaces Ms,αp,q. The authors obtain some sufficient and necessary conditions to ensure the boundedness on the α-modulation spaces. The main theorems significantly improve some known results.

Keywords Spherical average, α-Modulation spaces, Bessel functions

1 Introduction

Let Sn−1be the unit sphere in the Euclidean space Rn, n≥2.We define the average operator of functionsfon the unit sphere as

where dσ(y′) is the normalized surface Lebesgue measure.

This operator has a profound background in harmonic analysis, dating back to early 1970’s(see [21–22]). About 120 years ago, Pearson (see [18]) first used it to study random walks in high dimensional spaces. AnN-steps uniform walk in Rnstarts at the origin and consists ofNindependent steps of length 1, each of which is taken into a uniformly random direction. The probability density functionof such a random walk is the Fourier inverse of (A1)N(see [4]), where (A1)Ndenotes theNiteration ofA1.

The operatorA1also plays a significant role in the approximation theory (see [1]). In order to obtain some equivalent forms of the K-functional inLp(Rn) spaces, Belinsky, Dai and Ditzian[1]studied the iterates Δ(A1)Nfor positive integersN,where Δ is the Laplacian. They obtained the following result.

Theorem A(see [1])Let1 ≤p≤∞, n≥2andThe inequality

holds for all f∈Lp(Rn).

Naturally,we may ask that what is the smallest positive integerNto guarantee the inequality

Fan and Zhao in [7] answered this question by using the well known estimates of wave operators(see [16, 19]). Recently, Fan, Lou and Wang [6] obtained the sufficient and necessary conditions for the smallest positive integerN, which completely solved the above question. We state their theorem as follows.

Theorem B(see [6])Let n/=3,5, and N be positive integers. The inequality

holds if and only if

Let n=3,5, and N be positive integers. The inequality

holds if and only if

Later, in [13], Huang extended the boundedness of Δ(A1)Non the modulation spaceson full ranges of 1 ≤p1,p2,q1,q2≤∞ands1,s2∈R. They obtained the sufficiency and necessity for the boundedness of Δ(A1)NfromMoreover, he found the smallest iterate stepNwhich ensures that Δ(A1)Nis bounded on modulation spacesfor all (p,q,s) ∈[1,+∞)×[1,+∞)×R iswhich is smaller than that inL1(Rn) spaces(see Theorem B). Precisely, the related result is stated as follows.

Theorem C(see [13])Let σ=2 −N and1 ≤pi,qi≤∞, si∈Rfor i=1,2. When q1≤q2, the iterated spherical averageΔ(A1)Nis bounded fromand only if

When q1>q2,the iterated spherical averageΔ(A1)Nis bounded fromif and only if

The modulation spacewas introduced by Feichtinger[8]in order to measure smoothness of a function or distribution in a way different fromLpspaces. Nowadays, spacesare recognized as a useful tool for studying functional analysis, pseudo-differential operators and certain Cauchy problems of nonlinear partial differential equations (see [2, 5, 11, 15, 17, 20, 23,24]). In addition, Gr¨obner in his unpublished thesis (see [9]) extended modulation space toαmodulation spaceby using theα-decomposition on the frequency space. Their definitions will be represented in Section 2. Here we first point out one of their significant properties: The parameterα∈[0,1) determines a segmentation of the frequency spaces. Whenα=0,is equivalent to the classical modulation spaceWhenα→1,is considered equivalent to the classical Besov space. Obviously, it is proposed to be an intermediate function space between Besov space and modulation space. Hence, it is very important to study some analysis and PDE’s problems inα-modulation space. Among numerous research papers,the reader may refer to [3, 10, 12, 14, 26] and the references therein.

Motivated by the above works, in this paper, we consider the boundedness of Δ(A1)Nonα-modulation spaces and give some sufficient and necessary conditions on the boundedness of Δ(A1)NfromWe state our main results as follows.

Theorem 1.1Let σ=2 −and1 ≤pi,qi≤∞, si∈Rfor i=1,2, α∈[0,1).Whenor

the iterated spherical averageΔ(A1)Nis bounded from

Theorem 1.2Let1 ≤pi,qi≤∞, si∈Rfor i= 1,2, α∈[0,1). If the iterated spherical averageΔ(A1)Nis bounded fromthen the following conditions must hold

or

Remark 1.1Whenα= 0, we can see that the above results are sharp and coincide with that in modulation spaces which was obtained in[13]. But for the caseα∈(0,1),our results are not sharp for some technical problems. Essentially, these difficulties are due to two properties of Bessel function. The first one is thatVδ(r) andshare the same upper bound asr→∞. The other fact is that the first term of asymptotic expansion of the Bessel function iswhich exists root in every interval [2kπ,2(k+1)π] fork∈Z.

The proof of Theorem 1.1 is somewhat routine with the help of Bernstein’s multiplier theorem. However the proof of Theorem 1.2 is quite involved. Based on the structure ofand the asymptotic form of the Fourier transform of Δ(A1)N,we construct a sequence of functions{fkj,λ} to achieve the necessary conditions.

This paper is organized as follows. In Section 2, we will introduce some preliminary knowledge which includes some properties ofα-modulation spaces and some useful lemmas. The proofs of main results will be presented in Section 3.

Throughout this paper,we use the inequalityABto mean that there is a positive numberCindependent of all main variables such thatA≤CB, and use the notationA≃Bto meanABandBA.

2 Preliminaries and Lemmas

In this section, we give the definition and discuss some basic properties ofα-modulation spaces. Also, we will state some estimates and lemmas which will be used in our proofs.

Definition 2.1(α-Modulation Space)Let ρ be a nonnegative smooth radial bump function supported in B(0,2), satisfying ρ(ξ) = 1for|ξ|<1and ρ(ξ) = 0for|ξ| ≥2. For any k=(k1,k2,···,kn)∈Zn, we set

and

We define the ball

It is easy to check thatsatisfy

and

This type of decomposition on frequency space is called α-decomposition which is a generalization of the uniform decomposition and the dyadic decomposition. Corresponding to the above sequence{}k∈Zn, we can construct an operator sequence{}k∈Znby

whereFandF−1denote the standard Fourier transform, and inverse Fourier transform respectively. For α∈[0,1),0

where

andSee [12] for details.

We now list some basic properties aboutα-modulation spaces.

Proposition 2.1(Almost Orthogonality)(see [12])For any k∈Zn, we define

Then the cardinality ofΛαkis uniformly finite for all k∈Zn.

Proposition 2.2(Isomorphism) (see [12])Let0

Proposition 2.3(Embedding) (see [12])Suppose that0

or

we have

The Fourier multiplierm(D)is a linear operator whose action on a test functionfis formally defined by

The functionm(ξ)is called the symbol or multiplier ofm(D). Up to a constant multiple,m(D)is a convolution operator with the kernel. In the sense of distribution, it is defined as

By the Young inequality, if ‖m∨‖∈L1andf∈Lp, then we have

for any 1 ≤p≤∞. We will use the following Bernstein multiplier theorem to estimate‖m∨‖L1.

Lemma 2.1(Bernstein’s multiplier Theorem) (see [25])Assume that0< p≤2and∂γm(ξ)∈L2for all multi-indices γ withWe have

By checking the Fourier transform (see [22]), we have that

where

andJδ(r) is the Bessel function of orderδwhich is defined as

Here we state some basic properties about the Bessel function which will be used in our proofs.

Lemma 2.2(see [22])For any δ >we have

Lemma 2.3(see [22])Let r >1and δ >For any positive integer L and r∈[1,∞),we have

where ajand bjare constants for all j, and E(r)is a C∞function satisfying

for any k=0,1,2,···.

3 Proofs of Main Results

Now we show the proof of Theorem 1.1. By the definition ofα-modulation spaces, we need to estimate ‖□αkΔ(A1)Nf‖Lp2(Rn). First, we obtain the following lemma.

Lemma 3.1Let1 ≤p2≤∞and σ=Then

ProofFor anyk∈Zn, □αkΔ(A1)Nfis a Fourier multipliermαk(D)(f)=Ωαk(x)∗f, where

By the almost orthogonality ofα-decomposition(Proposition 2.1),there exists an integerk0(n)which depends only onn,such thatηαl(ξ)ηαk(ξ)=0 when |l−k|≥k0(n). By the definition ofα-decomposition, we have

whereIis the identity operator. Then, Young’s inequality and Minkowski’s inequality yield

Thus, we only need to estimate

for everyk∈Zn.By Proposition 2.1, it suffices to estimate

for 〈l〉≃〈k〉.

When |k|<100, by (2.3) in Lemma 2.2 and suppηαl(ξ), we have that |Ωαl(x)| 1 for|x|<100.

On the other hand,when|k|<100 and|x|≥100,without loss of generality,we may assumeBy the derivative formula ofVδ(r)(see (2.2) in Lemma 2.2) and taking integration by part onξ1variable in (3.1), we obtain that

for |x|≥100. This estimate implies that ‖Ωαl(x)‖L11 when |k|<100, since 〈l〉≃〈k〉.

Next, we consider the case |k|≥100. ChoosingL=1 in Lemma 2.3, we have the following asymptotic form ofVδ(r)

for |r|>1.

By the definition of {ηαk} and noticing thatα∈[0,1), we have that

forξ∈suppηαk. Therefore, when |k|>100 and 〈l〉≃〈k〉, we have

forξ∈suppηαl(ξ). Now, by the chain rule and the derivative formula ofVδ(r)(see (2.2) in Lemma 2.2) , we obtain

By the asymptotic form ofVδ(r), we obtain that

forξ∈suppηαl(ξ).

Thus,Vδ(|ξ|)Nandshare the same upper bound which isfor anyδ >−andξ∈suppηαl(ξ). By the fact

and (2.1), using Bernstein’s multiplier Theorem (Lemma 2.1), we can obtain that

forl∈Λk.

Combining all the above estimates of ‖Ωαl(x)‖L1, we have finished the proof of Lemma 3.1.

Next, we turn to prove Theorem 1.1. By the definition ofα-modulation spaces and Lemma 3.1, we have that

By the embedding properties ofα-modulation spaces (Proposition 2.3), we can easily obtain that

When

or

Thus, we have completed the proof of Theorem 1.1.

Now, we turn to prove Theorem 1.2. For this purpose, we need to establish the following two lemmas (Lemma 3.2 and Lemma 3.3). The idea of Lemma 3.3 is derived from [13], and Lemma 3.2 has been proved in [13]. For the sake of completeness, we will show all details in the following text.

Lemma 3.2For j∈N+, define

and

When j is big enough, we have

where C(n)is a positive constant depending only on n.

ProofThe proofs forn=2 andn>2 share the same idea. We only prove the casen=2 explicitly and leave the proof of another case to the reader.

By symmetry, we only need to consider the case {(x,y) ∈R2:x,y≥0}.Forj∈Z+, we define

and

Moreover,forr,a>0, 0 ≤y≤r, we define an auxiliary function

Taking derivative, we know that

andfr,a(y) is a monotone increasing function.

Then, for any (x0,y0)∈Ix>y, we have

Therefore,

whenjis big enough.

On the other hand, for any (x0,y0)∈II, we have

By monotonicity offr,a(y),

and

Combining all above analysis, we have

Now,we consider the domain IIIx>y. By the same argument,for any(x0,y0)∈IIIx>y, we have

It is easy to see that

Thus, for any (x0,y0)∈IIIx>y, we have

whenjis big enough. Moreover, it is obvious

whenj≥3. So, for every (x0,y0)∈IIIx>y,

whenj≥3. By (3.2), we can also obtain

On the other hand, for Iy≥xand IIIy≥x, by the same method on the auxiliary function

we can obtain that

and

Combining all above estimates, we have

Lemma 3.3Let1 ≤p≤∞. These exists a constant ρ=ρ(n)>0which only depends on n and a subsequence{kj}⊆Znsuch that

where{gkj(x)}is a sequence of Schwartz functions withsuppρ}.

ProofLetδ(ξ) be a smooth function with suppδ(ξ) ⊂{ξ: |ξ| ≤2ρ} andδ(ξ) ≡1 forξ∈{ξ:|ξ|≤ρ}. We define

For anygkwith suppwe have

By the same method as in Lemma 3.1, it is easy to get

By the definition ofδk(ξ) and Bernstein’s multiplier Theorem (Lemma 2.1), we have

for allk∈Zn. Therefore, for ∀k∈Zn, we have

Thus, we only need to prove the inverse inequality. By Lemma 2.3, we have

for |r|>1. We consider

in every semiperiod∈[jπ,(j+1)π], j=1,2,···.

Choosingε0=sin(0.07), we have

for∈[jπ+0.07,jπ+π−0.07],which is equivalent to

By Lemma 3.2, for everyj∈N+, the set

is not empty. So, there exists a subsequence of integer {kj}, such thatkj∈Λ1,jand

Moreover,

which means that

for

and

For the remainderin the expansion ofit is obvious that, whenris large enough,

Let∊,ρ=We obtain that there exist some constants∊,ρ >0 and a subsequence{kj}⊆Z+such that

forξ∈{ξ:|ξ−whenjis large enough. Moreover the subsequence {kj}⊆Znsatisfies

when the positive integerjis large enough.

Using the chain rule and the derivative formula ofVδ(t),

By the asymptotic form ofVδ(r) and (3.4), we have

whenshare the same upper bound which is

Moreover, sinceρ=sin(0.07)<, for the definition ofα-decomposition {ηαk} (see Definition 2.1), we have that

with

Therefore, by the definition ofδk(ξ), Bernstein multiplier theorem (Lemma 2.1) and (3.4), we obtain that

Combining the above estimate with (3.3), Lemma 3.2 is proved.

Now, we continue to prove Theorem 1.2. We first consider the caseq1≤q2. Letf(x) be a nonzero Schwartz function with suppDefine

forλ∈(0,ρ], whereρand {kj} are defined in Lemma 3.3. By the definition offkj,λ(x), we have

and

Then, by Lemma 3.3, we have

On the other hand,

By the assumption that Δ(A1)Nis bounded fromwe have that

for all |kj| sufficiently large and 0<λ≤ρ. Fixkjand letλ→0. We have

Thus, the conditionp1≤p2must be held. Moreover, whenλis fixed andkjgoes to infinity,we have

Next, we consider the caseq1>q2. LetMbe a large positive number. Define

whereaj>0 are constants to be chosen later andfkj,ρ(x) are defined in (3.6) with allkjsatisfying

for someL∈N+.

By (3.7)–(3.8) and the almost orthogonality of {ηαk}, we have

and

By the assumption that Δ(A1)Nis bounded fromwe have

By choosingwe obtain

By the assumptionq1>q2, the above series must be convergent asM→+∞. By Lemma 3.2,we have

Therefore, it must yield

which is equivalent toTheorem 1.2 is proved.


登录APP查看全文