Shengrong WANG Jingshi XU
Abstract In this paper, the authors obtain the boundedness of vector valued bilinear Calder´on-Zygmund operators on products of weighted Herz-Morrey spaces with variable exponents.
Keywords Bilinear Calder´on-Zygmund operator, Vector valued inequality, Muckenhoupt weight, Variable exponent, Herz-Morrey space
As a generalization of Calder´on-Zygmund singular integral operators, the theory of multilinear Calder´on-Zygmund singular integral operators began with the work of Coifman and Meyer [3] in 1975. After that, the multilinear Calder´on-Zygmund singular integral operators have attracted more and more attention from many authors, see [2, 9–12, 15, 24, 31–32]. In recent years, Hu and Meng [13] established the boundedness of multilinear Calder´on-Zygmund operators on products of Hardy spacesHp. Lu and Zhu [21] obtained the boundedness of multilinear Calder´on-Zygmund operators on Herz-Morrey spacewith variable exponents. Shen and the second author of this paper [27] obtained a vector valued inequality of multilinear Calder´on-Zygmund operators on products of Herz-Morrey spaceswith variable exponents. Wang and Liu [28] studied the boundedness of multilinear singular integral operators on the product of generalized Morrey spaces (Lp(w),Lq)α. Wang and the second author of this paper obtained Multilinear Calder´on-Zygmund operators and their commutators with BMO functions in variable exponent Morrey spaces in [30]. Hu and the second author of this paper obtained the boundedness of Multilinear Calder´on-Zygmund operators and their commutators with BMO functions in Herz-Morrey spaces with variable exponentsin [14]. In [29], we obtained a weighted norm inequality of bilinear Calder´on-Zygmund operators in weighted Herz-Morrey spaces with variable exponents and weight in the variable Muckenhoupt class.
In this paper, as a continuation of [29], we will consider the boundedness of vector valued bilinear Calder´on-Zygmund operators on products of weighted Herz-Morrey spaces with variable exponents. The plan of the paper is as follows. In Section 2, we collect some notations and state the main result. The proof of the main result will be given in Section 3.
In this section,we firstly recall some definitions and notations,then we state our result. Let Ω be a positive measurable subset of Rn,p(·) be a measurable function on Ω taking values in[1,∞),then the Lebesgue space with variable exopnentLp(·)(Ω) is defined by

The Lebesgue spaceLp(·)(Ω) becomes a Banach function space equipped with the norm

The spaceis defined byfor all compact subsetsK⊂Rn},where and what follows,χSdenotes the characteristic function of a measurable setS⊂Rn.Letp(·):Rn→(0,∞).We denoteThe set P(Rn)consists of all measurable functionsp(·) satisfyingp−>1 andp+<∞; P0(Rn) consists of all measurable functionsp(·) satisfyingp−>0 andp+<∞.Lp(·)can be similarly defined as above forp(·)∈P0(Rn).p′(·) is the conjugate exponent ofp(·),which means
Letp(·) ∈P(Rn) andwbe a nonnegative measurable function on Rn. Then the weighted variable exponent Lebesgue spaceLp(·)(w)is the set of all complex-valued measurable functionsfsuch thatfw∈Lp(·). The spaceLp(·)(w) is a Banach space equipped with the norm

Letf∈Then the standard Hardy-Littlewood maximal function offis defined by

where the supremum is taken over all balls containingxin Rn. In general,the Hardy-Littlewood maximal operator is not bounded on weighted variable Lebesgue spaces. But ifp(·) ∈P(Rn)and satisfies the following global log-H¨older continuous andw∈Ap(·), thenMis bounded onLp(·)(w).
Definition 2.1Let α(·)be a real-valued measurable function onRn.
(i)The function α(·)is locally log-H¨older continuous if there exists a constant C1such that

(ii)The function α(·)is log-H¨older continuous at the origin if there exists a constant C2such that

Denote bythe set of all log-H¨older continuous functions at the origin.
(iii)The function α(·)is log-H¨older continuous at the infinity if there exists α∞∈Rand a constant C3such that

Denote bythe set of all log-H¨older continuous functions at infinity.
(iv)The function α(·)is global log-H¨older continuous if α(·)are both locally log-H¨older continuous and log-H¨older continuous at infinity. Denote byPlog(Rn)the set of all global log-H¨older continuous functions.
Definition 2.2Fix p∈(1,∞). A positive measurable function w is said to be in the Muckenhoupt class Ap, if there exists a positive constant C for all balls B inRnsuch that

We sayw∈A1, ifMw(x) ≤Cw(x) for a.e.x. If 1 ≤p < q <∞, thenAp⊂Aq. We denoteThe MuckenhouptApclass with constant exponentp∈(1,∞)was firstly proposed by Muckenhoupt in [22]. In[5], Cruz-Uribe,Fiorenza and Neugebauer introduced the variable MuckenhouptAp(·). For more details, see [4–5, 8, 16–17].
Definition 2.3Let p(·) ∈P(Rn). A positive measurable function w is said to be in Ap(·),if there exists a positive constant C for all balls B inRnsuch that

Remark 2.1It is easy to see that ifp(·)∈P(Rn) andw∈Ap(·), thenw−1∈Ap′(·).
Lemma 2.1(see [5, Theorem 1.5])If p(·) ∈Plog(Rn)∩P(Rn)and w∈Ap(·), then there is a positive constant C such that for each f∈Lp(·)(w),

To give the definitions of the weighted Herz space and Herz-Morrey space with variable exponents,we use the following notations. For eachk∈Z,we defineBk:={x∈Rn:|x|≤2k},We also need the notation of the variable mixed sequence spaceℓq(·)(Lp(·)), which was firstly defined by Almeida and H¨ast¨o in[1]. Letwbe a nonnegative measurable function. Given a sequence of functions {fj}j∈Z,we define the modular


The norm is

Definition 2.4Let p(·)∈P(Rn), q(·)∈P0(Rn). Let α(·)be a bounded real-valued measurable function onRn. The homogeneous weighted Herz spaceand non-homogeneousweighted Herz spaceare defined respectively by

and

where

and

For any quantitiesAandB, if there exists a constantC >0 such thatA≤CB, we writeAB. IfABandBA, we writeA≈B.
The following lemma is a corollary of [20, Theorem 3].
Lemma 2.2Let α(·) ∈L∞(Rn), p(·), q(·) ∈P0(Rn)and w be a weight. If α(·)and q(·)are log-H¨older continuous at infinity, then

Additionally, if α(·)and q(·)are log-H¨older continuous at the origin, then

Definition 2.5Let p(·), q(·) ∈P0(Rn), λ∈[0,∞). Let α(·)be a bounded real-valued measurable function onRn. The homogeneous weighted Herz-Morrey spaceand non-homogeneous weighted Herz-Morrey spaceare defined respectively by

and

where

and

Proposition 2.1Let p(·), q(·)∈P0(Rn), w be a weight, λ∈[0,∞)and α(·)∈L∞(Rn).
(i)If α(·), q(·)∈then for any f∈

where and hereafter, q0:=q(0).
(ii)If α(·), q(·)∈then

ProofObviously,

WhenL≤0,from Lemma 2.2 we know that

WhenL>0,from Lemma 2.2 again we also obtain

Thus we obtain (i). Similarly, we can obtain (ii).
The following Lemma 2.3 has been proved by Izuki and Noi in [18–19].
Lemma 2.3If p(·) ∈Plog(Rn) ∩P(Rn)and w∈Ap(·), then there exist constants δ1,δ2∈(0,1)and C >0such that for all balls B inRnand all measurable subsets S⊂B,

We denote by S(Rn) the space of all Schwartz functions on Rnand S′(Rn) the space of all tempered distributions on Rn. Denote byL∞C(Rn) the space of compactly supported bounded functions, and supp(f) the support set of functionf. LetTbe a bilinear operator, which is originally defined on the 2-fold of Schwartz function space S(Rn), and its value belongs to S′(Rn):

Tis called a bilinear Calder´on-Zygmund operator, if it extends to a bounded bilinear operator fromLp1×Lp2toLpfor somep1, p2∈(1,∞) andpsuch thatand forf1,f2∈L∞C(Rn), x /∈supp(f1)∩supp(f2),

where the kernelKis a function in R3nofffrom the diagonalx=y1=y2and there exist positive constantsε, Asuch that

and

whenever |x−x′|≤max{|x−y1|,|x−y2|}, and the two analogous difference estimates with respect to the variablesy1andy2hold.
Our main result is as follows.
Theorem 2.1Assume that T is a bilinear Calder´on-Zygmund operator, p1(·)and p2(·) ∈Plog(Rn)∩P(Rn)satisfyingand p(·)∈P0such that there exists s≤p−such thatand M is bounded onwhere w=w1w2and wi∈Api(·), i=1,2.Suppose that α(·) ∈α(0)=α1(0)+α2(0), α∞=α1∞+α2∞,δi1, δi2∈(0,1)are the constants in Lemma2.3for exponents pi(·)and weights wi, i= 1,2.Let ri∈(1,∞)for i=1,2andIf λi−nδi1<αi∞,αi(0) for all To prove Theorem 2.1, we need a series of lemmas. Forδ >0, we denoteThen the sharp maximal function is defined by where the supremum is taken over all the cubesQcontaining the pointx, and where as usualfQdenotes the average offonQ. We denote Lemma 3.1(see[25,Theorem 2.1])Let0 for every function f such that the left hand side is finite. Lemma 3.2(see [23, Theorem 2.1])Let T be a bilinear Calder´on-Zygmund operator and let0<δ <. Then, there exists a constant C >0such that for any functions Lemma 3.3(see [15, Theorem 2.3])Let p(·), p1(·), p2(·) ∈P0(Rn)such thatfor x∈Rn. Then there exists a constant Cp,p1independent of functions f and g such that holds for every f∈Lp1(·)(Rn)and g∈Lp2(·)(Rn). If p∈P(Rn),w is a weight with w=w1×w2,then Lemma 3.4(see [26, Proposition 1.2])Let0< p≤∞, δ >0. Then there is a positive constant C such that for non-negative sequenceHere, when p=∞,(3.1)stands for Lemma 3.5(see [7, Proposition 3.20])The following are equivalent: (a)For every p∈(0,∞)and every w∈A∞, (b)There exists p0>0such that for every p,0 LetLp(w) be the weighted Lebesgue space with respect to the measurew(x)dx. Lemma 3.6(see[6,Theorem 2.1])Assume that for some p0∈(0,∞)and every w0∈A∞,letFbe a family of pairs of non-negative functions such that Then for all0 Furthermore, for every p, q∈(0,∞), w0∈A∞, and sequence{(fj,gj)}j⊂F, Lemma 3.7(see [8, Theorem 2])Assume that for some p0∈(0,∞), every w0∈A1, and letFbe a family of pairs of non-negative functions, Let p(·) ∈P0(Rn). If there exists p0≤p−such thatand M is bounded onThen for all(f,g)∈Fwith f∈Lp(·)(w), Lemma 3.8Assume that for some p0, every w0∈A∞, and letFbe a family of pairs of non-negative functions such that(3.2)holds. Let p(·) ∈P0(Rn). If there exists s≤p−such that ws∈and M is bounded onThen for every q∈(1,∞)and sequence{(fj,gj)}j∈N⊂F, ProofFixq∈(1,∞), we define the new family Fqto consist of the pair (Fq,Gq), where Since (3.2) holds we have (3.4). Thus, by (3.4) and (3.5), we obtain By Lemmas 3.5 and 3.7, we have This completes the proof. Lemma 3.9(see [8, Corollary 3.2])Let p(·) ∈P(Rn)and w be a weight. If the maximal operator M is bounded both on Lp(·)(w)and Lp′(·)(w−1), q∈(1,∞), then Lemma 3.10Let T be a bilinear Calder´on-Zygmund operator and p(·)∈P0such that there exists s≤p−such that ws∈and M is bounded onSuppose that w=w1w2and wi∈Api(·), i=1,2. If pi(·)∈Plog(Rn)∩P(Rn)for i=1,2satisfyingfor x∈Rn, i= 1,2, then there exists a constant C independent of the compactly supported bounded functions fj1, fj2∈Lp0(Rn), j∈Nsuch that where qi∈(1,∞)for i=1,2and ProofSincefj1, fj2are bounded functions with compact support,T(fj1,fj2) ∈Lp(Rn) for every 0 Therefore, by Lemmas 3.6 and 3.8, we have This completes the proof. Proof of Theorem 2.1Since the set of all bounded compactly supported functions is dense in weighted variable Lebesgue spaces (see [8, Lemma 3.1]), we only consider bounded compact supported functions. Letfv1andfv2be bounded functions with compact support forv∈N and write By Proposition 2.1, we have where Since to estimateGis essentially similar to estimateE, it is suffice to obtain thatEandHare bounded in Herz-Morrey space with variable exponents. It is easy to see that where We shall use the following estimates. Ifl≤k−1, then by H¨older’s inequality, Lemma 2.3 and Definition 2.3, we have Ifl=k, then Ifl≥k+1, then By the interchange off1andf2, we see that the estimates ofE2,E3andE6are similar to those ofE4, E7andE8, respectively. Thus we are only necessary to estimateE1, E2, E3, E5, E6andE9. To estimateE1, sincel, j≤k−2,we deduce that fori=1,2, Therefore, forx∈Dk, we have Thus, for anyx∈Dkandl, j≤k−2, we have Therefore, by H¨older’s inequality and Minkowski’s inequality, we obtain where Sincenδi2−αi(0)>0, by (3.6) and Lemma 3.4 we obtain where we write 2−|k−l|(nδi2−αi(0))=2−|k−l|εiforεi=nδi2−αi(0)>0.Thus, we obtain To estimateE2, sincel≤k−2 andk−1 ≤j≤k+1, then we have Therefore, forx∈Dk, we have Thus, for anyx∈Dk,l≤k−2, k−1 ≤j≤k+1, we have Therefore, by H¨older’s inequality and Minkowski’s inequality, we obtain It is obvious that Now we estimateE2,2. Taking (3.6)–(3.8) together, we have where we use 2−nδ22<1 and 2(j−k)n(1−δ12)<2(j−k)n<22n,j∈{k−1,k,k+1} for (3.6) and(3.8) respectively. Thus, we obtain To estimateE3, sincel≤k−2 andj≥k+2, we have Therefore, fir anyx∈Dk, l≤k−2, j≥k+2, we get Thus, by H¨older’s inequality and Minkowski’s inequality, we have It is obvious that Sincenδ21+α2(0)>0, by (3.8), we obtain First, we considerI1. By Lemma 3.4, we have where we write 2−|k−j|(nδ21+α2(0))=2−|k−j|η2forη2=nδ21+α2(0)>0. Next, we considerI2. Sincenδ21+α2(0)−λ2>0, we obtain Then, we considerI3. Sincenδ21+α2(0)−λ2>0, we obtain Thus, we have To estimateE5, using H¨older’s inequality and Lemma 3.10, we have To estimateE6, sincek−1 ≤l≤k+1 andj≥k+2, we obtain Thus, for anyx∈Dk, k−1 ≤l≤k+1 andj≥k+2, we obtain Therefore, by H¨older’s inequality and Minkowski’s inequality, we obtain By the interchange off1andf2, we see that the estimate ofE6,1is similar to that ofE2,2andE6,2=E3,2. To estimateE9, sincel, j≥k+2, we get Therefore, for anyx∈Dk, l, j≥k+2, we have Thus, by H¨older’s inequality and Minkowski’s inequality, we have Obviously, the estimates ofE9,iare similar to those ofE3,2fori=1,2,respectively. Taking all estimates forEitogether,i=1,2,···,9, we obtain Finally, we estimateH.By the interchange off1andf2, we see that the estimates ofH2,H3andH6are similar to those ofH4, H7andH8, respectively. Thus we are only necessary to estimateH1, H2, H3, H5, H6andH9. To go on, we need further preparation. Ifl <0, by Proposition 2.1, we have To estimateH1, sincel, j≤k−2,andλ=λ1+λ2,by (3.9) and H¨older’s inequality, we have where By (3.6), we obtain Ifqi∞≥1, sincenδi2−αi∞>0 andnδi2−αi(0)>0, by the Minkowski’s inequality and(3.14), we obtain Ifqi∞<1, sincenδi2−αi∞>0 andnδi2−αi(0)>0, by (3.14), we have We considerI5. Sincenδi2−αi∞>0, by Lemma 3.4, we have where we write 2−|k−l|(nδi2−αi∞)2−|k−l|ηiforηi=nδi2−αi∞. Thus, we get To estimateH2, sincel≤k−2, k−1 ≤j≤k+1,andλ=λ1+λ2,by(3.10) and H¨older’s inequality, we have It is obvious that Now we estimateH2,2. Combining (3.6)–(3.8), we have where we use 2−nδ22<1 and 2(j−k)n(1−δ21)<2(j−k)nfor (3.6) and (3.8), respectively. Thus,we obtain To estimateH3, sincel≤k−2, j≥k+2,andλ=λ1+λ2,by (3.11) and H¨older’s inequality, we have It is obvious that Sincenδ21+α2∞>0, by (3.8), we obtain ForI6, by Lemma 3.4, we obtain where we write 2−|k−j|(nδ21+α2∞)=2−|k−j|ϑ2forϑ2=nδ21+α2∞>0. ForI7, sincenδ21+α2∞−λ2>0, we have Thus, we get To estimateH5, using H¨older’s inequality and Lemma 3.10, we have To estimateH6, sincek−1 ≤l≤k+1, j≥k+2,andλ=λ1+λ2,by(3.12) and H¨older’s inequality, we have By the interchange off1andf2, we see that that ofH6,1is similar to the estimate ofH2,2andH6,2=H3,2. To estimateH9, sincel,j≥k+2,andλ=λ1+λ2,by(3.13)and H¨older’s inequality, we have Obviously, the estimates ofH9,iare similar to those ofH3,2fori=1,2,respectively. Taking all estimates forHitogether,i=1,2,···,9, we obtain This completes the proof. AcknowledgementThe authors would like to thank the referee for his (or her) careful reading.
3 Proof of Theorem 2.1


































































































Chinese Annals of Mathematics,Series B
2021年5期