Abstract
Let \(\Omega \) be a homogeneous function of degree zero, have vanishing moment of order one on the unit sphere \(\mathbb {S}^{d-1}\)(\(d\ge 2\)). In this paper, our object of investigation is the following rough non-standard singular integral operator
where A is a function defined on \({\mathbb {R}}^d\) with derivatives of order one in \({\textrm{BMO}}({\mathbb {R}}^d)\). We show that \(T_{\Omega ,\,A}\) enjoys the endpoint \(L\log L\) type estimate and is \(L^p\) bounded if \(\Omega \in L(\log L)^{2}({\mathbb {S}}^{d-1})\). These results essentially improve the previous known results given by Hofmann (Stud Math 109:105–131, 1994) for the \(L^p\) boundedness of \(T_{\Omega ,\,A}\) under the condition \(\Omega \in L^{q}({\mathbb {S}}^{d-1})\) \((q>1)\), Hu and Yang (Bull Lond Math Soc 35:759–769, 2003) for the endpoint weak \(L\log L\) type estimates when \(\Omega \in \textrm{Lip}_{\alpha }({\mathbb {S}}^{d-1})\) for some \(\alpha \in (0,\,1]\).
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
This paper will be devoted to study the boundedness of certain non-standard Calderón-Zygmund operators with rough kernels. To begin with, let \(d\ge 2\), \({\mathbb {R}}^d\) be the d-dimensional Euclidean space and \({\mathbb {S}}^{d-1}\) be the unit sphere in \({\mathbb {R}}^d\). Let \(\Omega \) be a function of homogeneous of degree zero, \(\Omega \in L^1({{\mathbb {S}}^{d-1}})\) and satisfy the vanishing condition
Define the non-standard rough Calderón-Zygmund operator by
where A is a function on \({\mathbb {R}}^d\) such that \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\), that is, \(\partial _n A\in \textrm{BMO}({\mathbb {R}}^d)\) for all n with \(1\le n\le d\). This class of singular integrals is of interest in Harmonic analysis. It was well-known that \(T_{\Omega ,\,A}\) is closely related to the study of Calderón commutators [1, 2]. Even for smooth kernel \(\Omega \), since \({L^\infty ({\mathbb {R}}^d)}\varsubsetneq \textrm{BMO}({\mathbb {R}}^d)\), the kernel of the operator \(T_{\Omega ,\,A}\) may fail to satisfy the classical standard kernel conditions. This is the main reason why one calls them nonstandard singular integral operators.
Recall that if \(\nabla A\in L^\infty ({\mathbb {R}}^d)\), then the \(L^p({\mathbb {R}}^d)\) boundedness of \(T_{\Omega ,\,A}\) follows by using the methods of rotation in the nice work of Caldéron [2], Bainshansky and Coifman [1]. Since the method of rotations doesn’t work in the case of \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\), Cohen [7] and Hu [24] obtained the \(L^p({\mathbb {R}}^d)\) boundedness of \(T_{\Omega ,\,A}\) with smooth kernels by means of a good-\(\lambda \) inequality. More precisely, if \(\Omega \in \textrm{Lip}_\alpha ({\mathbb {S}}^{d-1})\) (\(0<\alpha \le 1\)), then Cohen [7] proved that \(T_{\Omega ,A}\) is a bounded operator on \(L^p({\mathbb {R}}^d)\) for \(1<p<\infty \). Later on, the result of Cohen [7] was improved by Hofmann [19]. It was shown that \(\Omega \in \cup _{q>1}L^q({\mathbb {S}}^{d-1})\) is a sufficient condition for the \(L^p({\mathbb {R}}^d)\) boundedness of \(T_{\Omega , A}\). If \(\Omega \in L^{\infty }({\mathbb {S}}^{d-1})\), Hofmann [19] demonstrated that \(T_{\Omega ,A}\) is bounded on \(L^p({\mathbb {R}}^d,\,w)\) for all \(p\in (1,\,\infty )\) and \(w\in A_p({\mathbb {R}}^d)\), where and in what follows, \(A_p({\mathbb {R}}^d)\) denotes the weight function class of Muckenhoupt, see [12, Chap. 9] for properties of \(A_p({\mathbb {R}}^d)\).
It is quite natural to ask if one can establish weak type inequalities for \(T_{\Omega ,\,A}\) or not. Hu and Yang [23] considered the operator
where a is a function on \({\mathbb {R}}\) such that \(a'\in \textrm{BMO}({\mathbb {R}})\). Hu and Yang showed that, \(T_{a}\) may fail to be of weak type \((1,\,1)\), which differs in this aspect from the property of the classical singular integral operators, see Remark 3 in [23, p. 762]. As a replacement of weak (1, 1) boundedness, it was shown in [23] that, when \(\Omega \in \textrm{Lip}_{\alpha }(S^{d-1})\) with \(\alpha \in (0,\,1]\), \(T_{\Omega ,A}\) still enjoys the endpoint \(L\log L\) type estimates. This, tells us that, when \(\Omega \) satisfies suitable regularity condition, the endpoint estimates of \(T_{\Omega , A}\) parallels to that of the commutator of Calderón-Zygmund operators with symbol in \(\textrm{BMO}({\mathbb {R}}^d)\). For the endpoint estimates of the commutator of Calderón-Zygmund operators, see [22, 29] and the references therein.
Now, we recall some known results of classical singular integrals and make a comparative analysis. It was first shown by Calderón and Zygmund [3] that the singular integrals \(T_{\Omega }\) defined by
is bounded on \(L^p({\mathbb {R}}^d)\) \((1<p<\infty )\) either when \(\Omega \) is an odd function and \(\Omega \in L^1({\mathbb {S}}^{d-1})\), or \(\Omega \) is an even function with \(\int _{{\mathbb {S}}^{d-1}}\Omega \,d\sigma =0\) and \(\Omega \in L\log L({\mathbb {S}}^{d-1})\). Later on, the condition \(\Omega \in L\log L({\mathbb {S}}^{d-1})\) was improved to \(\Omega \in H^1({\mathbb {S}}^{d-1})\) by Connett [8], Ricci and Weiss [30], independently. Since then, great achievements have been made in this field. Among them are the celebrated works of the weak type (1, 1) bounds given by Christ [5], Christ and Rubio de Francia [6], Hofmann [17], Seeger [31], and Tao [33]. It was shown that \(\Omega \in L\log L({\mathbb {S}}^{d-1})\) is sufficient condition for the weak type (1, 1) estimate of \(T_{\Omega }\). Recently, this result was generalized by Ding and Lai [9] for the operator \(T_{\Omega }^*\) defined by
where the kernel \(\Omega \in L\log L({\mathbb {S}}^{d-1})\) and K needs to satisfy some size and regularity conditions. For other related contributions, we refer the readers to references [10, 11, 15, 22, 25,26,27,28, 32, 34, 35] and the references therein.
Consider now the \(L^p({\mathbb {R}}^d)\) boundedness and endpoint estimates for the operator \(T_{\Omega ,\,A}\) when \(\Omega \) satisfies only size condition, things become more subtle. Hu [21] considered the \(L^2({\mathbb {R}}^d)\) boundedness of \(T_{\Omega , A}\) when \(\Omega \in GS_{\beta }({\mathbb {S}}^{d-1})\), which means,
The main result in [21] can be summarized as follows:
Theorem A
Let \(\Omega \) be homogeneous of degree zero which satisfies the vanishing condition (1.1), A be a function on \({\mathbb {R}}^d\) such that \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\). Suppose that \(\Omega \in GS_{\beta }(S^{d-1})\) for some \(\beta >3\), then \(T_{\Omega ,\,A}\) is bounded on \(L^2({\mathbb {R}}^d)\).
This size condition was introduced by Grafakos and Stefanov [14], to study the \(L^p({\mathbb {R}}^d)\) boundedness of the homogeneous singular integral operator. As it was pointed out in [14], there exist integrable functions on \({\mathbb {S}}^{d-1}\) which are not in \(H^1({\mathbb {S}}^{d-1})\) but satisfy (1.3) for all \(\beta \in (1,\,\infty )\). Thus, \(GS_{\beta }({\mathbb {S}}^{d-1})\) is also a minimum size condition for functions on \({\mathbb {S}}^{d-1}\). It is easy to verify that
For the \(L^p({\mathbb {R}}^{d})\) (\(1<p<\infty \)) boundedness of \(T_{\Omega , A}\), the best known condition \(\Omega \in \cup _{q>1}L^q({\mathbb {S}}^{d-1})\) is given in [19]. There is no any endpoint estimate for \(T_{\Omega , A}\) when \(\Omega \) only satisfies some size condition, even if \(\Omega \in L^{\infty }({\mathbb {S}}^{d-1}) \). Note that the following inclusion relationship holds
Therefore, it is quite natural to ask the following question:
Question: What is the minimal condition such that \(T_{\Omega ,\,A}\) is bounded on \(L^p({\mathbb {R}}^d)\) for all \(p\in (1,\,\infty )\)? Does the endpoint estimate of \(L\log L\) type still holds true when \(\Omega \) only satisfies size condition?
The main purpose of this paper is to show that \(\Omega \in L(\log L)^2({\mathbb {S}}^{d-1}) \) is a sufficient condition for the \(L^p({\mathbb {R}}^d)\) boundedness and weak type \(L\log L\) estimate for \(T_{\Omega , A}\). Our first result can be stated as follows.
Theorem 1.1
Let \(\Omega \) be homogeneous of degree zero, satisfy the vanishing moment (1.1), and A be a function on \({\mathbb {R}}^d\) such that \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\). Suppose that \(\Omega \in L(\log L)^{2}({\mathbb {S}}^{d-1})\). Then \(T_{\Omega ,\,A}\) is bounded on \(L^2({\mathbb {R}}^d)\).
Let \({\widetilde{T}}_{\Omega ,\,A}\) be the dual operator of \(T_{\Omega ,\,A}\), defined as
Theorem 1.2
Let \(\Omega \) be homogeneous of degree zero, satisfy the vanishing condition (1.1), and A be a function on \({\mathbb {R}}^d\) such that \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\). Suppose that \(\Omega \in L(\log L)^{2}({\mathbb {S}}^{d-1})\). Then for any \(\lambda >0\) and \(\mathrm{\Phi (t)=t\log ({e}+t)}\), the following inequalities hold
As far as we know, there is no previous study concerning the weak type endpoint estimates for \({\widetilde{T}}_{\Omega , A}\), even if \(\Omega \in \textrm{Lip}_{\alpha }({\mathbb {S}}^{d-1})\) for \(\alpha \in (0,\,1]\). We consider this operator mainly to deduce the following precise \(L^p({\mathbb {R}}^d)\) bounds of \(T_{\Omega ,\,A}\).
Theorem 1.3
Let \(\Omega \) be homogeneous of degree zero, satisfy the vanishing condition (1.1), and A be a function on \({\mathbb {R}}^d\) such that \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\). Suppose that \(\Omega \in L(\log L)^{2}({\mathbb {S}}^{d-1})\). Then
Remark 1.4
Theorem 1.1, along with Theorem 1.3, shows that \(\Omega \in L(\log L)^{2}({\mathbb {S}}^{d-1})\) is a sufficient condition such that \(T_{\Omega ,\,A}\) is bounded on \(L^p({\mathbb {R}}^d)\) for all \(p\in (1,\,\infty )\). This improves essentially the result obtained in [19, Theorem 1.1], in which, it was shown that if \(\Omega \in \cup _{q>1}L^q({\mathbb {S}}^{d-1})\), then \(T_{\Omega ,\,A}\) is bounded on \(L^p({\mathbb {R}}^d)\) for all \(p\in (1,\,\infty )\).
Remark 1.5
As it was pointed out, for \(\beta \in [1,\,\infty )\), \(L(\log L)^{\beta }({\mathbb {S}}^{d-1})\subset GS_{\beta }(S^{d-1})\). However, it is unknown whether \(L(\log L)^{\beta }({\mathbb {S}}^{d-1})\subset GS_{\beta '}(S^{d-1})\) when \(\beta '>\beta \). We conjecture that there is no inclusion relationship between \(L(\log L)^{\beta }({\mathbb {S}}^{d-1}) \) and \( GS_{\beta '}(S^{d-1})\) when \(\beta '>\beta \), and believe Theorem A and Theorem 1.3 do not imply each other in the case \(p=2\).
We believe that the condition \(\Omega \in L(\log L)^{2}({\mathbb {S}}^{d-1})\) is the weakest condition for these weak type results to hold, in the following sense.
Conjecture 1.6
\(\Omega \in L(\log L)^{2}({\mathbb {S}}^{d-1})\) is the minimal condition for the weak \(L\log L\) type estimate of \(T_{\Omega ,\,A}\), and weak (1, 1) estimate of \({\widetilde{T}}_{\Omega ,A}\), in the sense that the power 2 can’t be replaced by any real number smaller than 2.
The article is organized as follows. Section 2 will be devoted to demonstrate the \(L^2\) boundedness of \(T_{\Omega ,A}.\) In Sect. 3, we will prove Theorem 1.2 and Theorem 1.3. The proof of Theorem 1.2 is not short and will be divided into several cases and steps. Smoothness trunction method will play an important role and will be used several times.
Let’s explain a little bit about the proofs of the main results. In Sect. 2, we will introduce a convolution operator \(Q_s\) with the property that
This makes it possible to commutate with the paraproducts appeared in the proof and thus obtains more freedom in dealing with the estimates of the \(L^2\) boundedness. Moreover, the method of dyadic analysis has been applied in the delicate decomposition of \(L^2\) norm of \(T_{\Omega ,\,A}\). At some key points, we will use some properties of Carleson measure.
The key ingredient in our proof of Theorem 1.2 is to estimate the bad part in the Calderón-Zygmund decomposition of f. In the work of [31], Seeger showed that if \(\Omega \in L\log L({\mathbb {S}}^{d-1})\), then \(T_{\Omega }\) is bounded from \(L^1({\mathbb {R}}^d)\) to \(L^{1,\,\infty }({\mathbb {R}}^d)\). Ding and Lai [9] proved that if \(\Omega \in L\log L({\mathbb {S}}^{d-1})\) and for some \(\delta \in (0,\,1]\), the function K satisfies
and \({T}_{\Omega }^*\) is bounded on \(L^2({\mathbb {R}}^d)\), then \({T}_{\Omega }^*\) is bounded from \(L^1({\mathbb {R}}^d)\) to \(L^{1,\,\infty }({\mathbb {R}}^d)\). However, when A has derivatives of order one in \(\textrm{BMO}({\mathbb {R}}^d)\), the function \([{A(x)-A(y)-\nabla A(y)(x-y)}]{|x-y|^{-d-1}}\) does not satisfy the conditions (1.8)–(1.10). Let f be a bounded function with compact support, \(b=\sum _{L}b_L\) be the bad part in the Calderón-Zygmund decomposition of f. In order to overcome this essential difficulty, we write
where \(A_{L}(y)=A(y)-\sum _{n=1}^d\langle \partial _n A\rangle _Ly_n.\) \(\phi _s(x)=\phi (2^{-s}x)\). Here, \(\langle \partial _n A\rangle _L\) denotes the mean value of \(\partial _n A\) on the cube L, \(\phi \) is a smooth radial nonnegative function on \({\mathbb {R}}^d\) such that \(\textrm{supp}\, \phi \subset \{x:\frac{1}{4}\le |x|\le 1\}\) and \(\sum _s\phi _s(x)=1\) for all \(x\in {\mathbb {R}}^d\backslash \{0\}\). Then, our key observation is that, for each \(s\in {\mathbb {Z}}\) and L with side length \(\ell (L)=2^{s-j}\), the kernel \({|x-y|^{-d-1}}\phi _s(x-y)\big (A_L(x)-A_L(y)\big )\chi _{L}(y)\) instead satisfies (1.9) and (1.10).
In what follows, C always denotes a positive constant which is independent of the main parameters involved but whose value may differ from line to line. We use the symbol \(A\lesssim B\) to denote that there exists a positive constant C such that \(A\le CB\). Specially, we use \(A\lesssim _{n,p} B\) to denote that there exists a positive constant C depending only on \(n,\,p\) such that \(A\le CB\). Constant with subscript such as \(c_1\), does not change in different occurrences. For any set \(E\subset {\mathbb {R}}^d\), \(\chi _E\) denotes its characteristic function. For a cube \(Q\subset {\mathbb {R}}^d\), \(\ell (Q)\) denotes the side length of Q, and for \(\lambda \in (0,\,\infty )\), we use \(\lambda Q\) to denote the cube with the same center as Q and whose side length is \(\lambda \) times that of Q. For a suitable function f, \({\widehat{f}} \) denotes the Fourier transform of f. For \(p\in [1,\,\infty ]\), \(p'\) denotes the dual exponent of p, namely, \(1/p'=1-1/p\).
2 Proof of Theorem 1.1
This section will be devoted to prove Theorem 1.1, the \(L^2({\mathbb {R}}^{d})\) boundedness of \(T_{\Omega ,\,A}\) when \(\Omega \in L(\log L)^{2}({\mathbb {S}}^{d-1})\). We will employ some ideas from [19], together with many more refined estimates. We begin with some notions and lemmas. Let \(\psi \in C^{\infty }_0({\mathbb {R}}^d)\) be a radial function with integral zero, \(\textrm{supp}\,\psi \subset B(0,\,1)\), \(\psi _s(x)=s^{-d}\psi (s^{-1}x)\) and assume that
Consider the convolution operator \(Q_sf(x)=\psi _s*f(x). \) It enjoys the property that
Moreover, by the classical Littlewood-Paley theory, it follows that
Let \(\phi \) be a smooth radial nonnegative function on \({\mathbb {R}}^d\) with \(\textrm{supp}\, \phi \subset \{x:\frac{1}{4}\le |x|\le 1\}\), \(\sum _s\phi _s(x)=1\) with \(\phi _j(x)=2^{-jd}\phi (2^{-j}x)\) for all \(x\in {\mathbb {R}}^d\backslash \{0\}\). For each fixed \(j\in {\mathbb {Z}}\), define
where
The following lemmas are needed in our analysis.
Lemma 2.1
([19]) Let \(\Omega \) be homogeneous of degree zero, satisfies the vanishing condition (1.1) and \(\Omega \in L^1 ({\mathbb {S}}^{d-1})\). Let A be a function on \({\mathbb {R}}^d\) such that \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\). Then for any \(k_1,\,k_2\in {\mathbb {Z}}\) with \(k_1<k_2\), the following inequality holds
Lemma 2.2
([19]) Let \(\Omega \) be homogeneous of degree zero, integrable on \({\mathbb {S}}^{d-1}\) and satisfy the vanishing moment (1.1). Let A be a function on \({\mathbb {R}}^d\) such that \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\). Then there exists a constant \(\epsilon \in (0,\,1)\), such that for \(s\in (0,\,\infty )\) and \(j\in {\mathbb {Z}}\) with \(s2^{-j}\le 1\),
Lemma 2.3
([19]) Let \(\Omega \) be homogeneous of degree zero and \(\Omega \in L^{\infty }({\mathbb {S}}^{d-1})\). Let A be a function on \({\mathbb {R}}^d\) such that \(\nabla A\in \textrm{BMO}({\mathbb {R}}^d)\). Then there exists a constant \(\varepsilon \in (0,\,1)\), such that for \(s\in (0,\,\infty )\) and \(j\in {\mathbb {Z}}\) with \(2^{-j}s\le 1\),
Lemma 2.4
([20]) Let \(\Omega \) be homogeneous of degree zero, have mean value zero on \({\mathbb {S}}^{d-1}\) and \(\Omega \in L(\log L)^2({S}^{d-1})\). Then for \(b\in \textrm{BMO}({\mathbb {R}}^d)\), \([b,\,T_{\Omega }]\), the commutator of \(T_{\Omega }\) with symbol b, defined by
is bounded on \(L^p({\mathbb {R}}^d)\) for all \(p\in (1,\,\infty )\).
Lemma 2.5
([19]) Let \(\Omega \) be homogeneous of degree zero, and integrable on \({\mathbb {S}}^{d-1}\) and satisfy the vanishing moment (1.1), A be a function in \({\mathbb {R}}^d\) with derivatives of order one in \(\textrm{BMO}({\mathbb {R}}^d)\). Then for any \(r\in (0,\,\infty )\), functions \({\widetilde{\eta }}_1,\,{\widetilde{\eta }}_2\in C^{\infty }_0({\mathbb {R}}^d)\) whose supported on balls of radius r,
The following lemma plays an important role in our analysis.
Lemma 2.6
([4]) Let A be a function on \({\mathbb {R}}^d\) with derivatives of order one in \(L^q({\mathbb {R}}^d)\) for some \(q\in (d,\,\infty ]\). Then
where \(I_{(x,|x-y|)}\) is a cube which is centered at x with length \(2|x-y|.\)
We need a lemma from the book of Grafakos.
Lemma 2.7
([12, p. 140]) Let \(\Phi \) be a function on \({\mathbb {R}}^d\) satisfying for some \(0<C,\delta <\infty ,\) \(|\Phi (x)|\le C {(1+|x)|)^{-d-\delta }}\). For \(t>0\), set \(\Phi _t(x)=t^{-d}\Phi (t^{-1}x)\). Then a measure \(\mu \) on \({\mathbb {R}}_+^{d+1}\) is a Carleson if and only if for every p with \(1<p<\infty \) there is a constant \(C_{p,d,\mu }\) such that for all \(f\in L^p({\mathbb {R}}^d)\) we have
Proof of Theorem 1.1
Invoking (2.1), to prove that \(T_{\Omega ,\,A}\) is bounded on \(L^2({\mathbb {R}}^d)\), it suffices to show the following inequalities hold for \(f,\,g\in C^{\infty }_0({\mathbb {R}}^d)\),
First, we will prove (2.4). To this aim, the kernel \(\Omega \) will be decomposed into disjoint forms. Let
Set
For \(i\in {\mathbb {N}}\cup \{0\}\), let \(T_{\Omega ,\,A;\,j}^i\) be the same as in (2.3) for \(T_{\Omega ,\,A;\,j}\) with \(\Omega \) replaced by \(\Omega _i\). Then
Let \(\alpha \in \big (\frac{d+1}{d+2},\,1\big )\) be a constant. Fix \(j\in {\mathbb {Z}}\), we decompose the set \(\{(s,t):\, 0<t<\infty ,\,0<s\le t\}\) into three regions:
In the following three subsections, we will discuss the contribution of each \(E_{j,s,t}\) on the right ride of (2.6) to inequality (2.4). \(\square \)
2.1 Contribution of \({E_1(j,s,t)}\)
Let \(\varepsilon \) be the same constant appeared in Lemma 2.3 and denote \(N=2(\lfloor \varepsilon ^{-1}\rfloor +1)\). For each fixed \(i\in {\mathbb {N}}\), we introduce the notion \( E_{1,1}^i\) and \(E_{1,2}^i\) as follows
Then, one gets obviously that \({E_1(j,s,t)}=E_{1,1}^i(j,s,\,t)\cup E_{1,2}^i(j,s,\,t):=E_{1,1}^i\cup E_{1,2}^i.\) Therefore
We first consider term I. Let \(\{I_l\}_{l}\) be a sequence of cubes having disjoint interiors and side lengths \(2^j\), such that
For each fixed l, let \(\zeta _l\in C^{\infty }_0({\mathbb {R}}^d)\) such that \(\textrm{supp}\,\zeta _l\subset 48dI_l\), \(0\le \zeta _l\le 1\) and \(\zeta _l(x)\equiv 1\) when \(x\in 32dI_l\). Let \(x_l\) be a point on the boundary of \(50dI_l\) and
with \(A_{I_l}^*(y)={\widetilde{A}}_{I_l}(y)-{\widetilde{A}}_{I_l}(x_l). \) Note that for \(x\in 30 dI_l\) and \(y\in {\mathbb {R}}^d\) with \({|x-y|\le 2^{j}}\), we have
An application of Lemma 2.6 then implies that \(\Vert A_{I_l}\Vert _{L^{\infty }({\mathbb {R}}^d)}\lesssim 2^j. \)
For each fixed \(j\in {\mathbb {Z}}\), consider the operators \(W^i_{\Omega ,\,j}\) and \(U^i_{\Omega ,\,m;j}\) defined by
and
The method of rotation of Caldeón-Zygmund states that for \(p\in (1,\,\infty )\), they enjoy the following properties:
see [12, pp. 272–274]. For each fixed l, let \(h_{s,l}(x)=Q_sg(x)\chi _{I_l}(x)\) and \(I_l^*=60dI_l\). For \(x\in \textrm{supp}h_{s,l}\), we have
Hence, to show the estimate for \(\textrm{I}\), we need to consider the following three terms.
and
For \( \textrm{R}^1_{i},\) note that
Then, the well-known Littlewood-Paley theory for g-function leads to that
For \(x\in 48dI_l,\) since \(\sup \{\phi _j\}\subset [2^{j-2},2^j]\) and note that \(\phi _j(x-y)Q_t^4f(y)=\chi _{I_l^*}(y)\phi _j(x-y){Q_t^4f(y),}\) then, \(W_{\Omega ,j}^i(Q_t^4f)=W_{\Omega ,j}^i(\chi _{I_l^*}Q_t^4f)\). It then follows from Hölder’s inequality, Cauchy-Schwarz inequality and the boundedness of \(W_{\Omega ,j}^i\) that
where in the last inequality we have used the fact that the cubes \(\{60 dI_l\}_l\) have bounded overlaps.
The same reasoning applies to \(\textrm{R}^2_{i}\) with small and straightforward modifications yields that
Now we are in a position to consider each term \(\textrm{R}_{i,m}^3\). For \(x\in 32dI_l\), it is easy to check
Therefore \(\textrm{R}_{i,m}^3\) can be controlled by the sum of the following terms:
Observe that \(|[\partial _mA,\,Q_t]h(x)|\lesssim M_{\partial _mA}h(x),\) where \(M_{\partial _m A}\) is the commutator of the Hardy-Littlewood maximal operator defined by
Hölder’s inequality, along with the \(L^2({\mathbb {R}}^d)\) boundedness of \(M_{\partial _m A}\) and \(U_{\Omega ,m,j}^i\), it yields that
Exactly the same reasoning applies to \( \textrm{R}_{i,m}^{3,2}\), we obtain
As for \(\textrm{R}_{i,m}^{3,3},\) observing that for fixed \(l\in {\mathbb {Z}}\), \(s,\,t\le 2^j\), one gets
Henceforth we have
Let \(x\in 48dI_l\), \(q\in (1,\,2)\) and \(s\in (2^{j-1}, 2^j)\). A straightforward computation involving Hölder’s inequality and the John-Nirenberg inequality gives us that
where \(I(x,\,s)\) is the cube center at x and having side length s.
This inequality, together with the boundedness of \(U^i_{\Omega ,m,j}\) and maximal function \(M_qh\), implies that
On the other hand, by the \(L^2\) boundedness of convolution operators and the Littlewood-Paley theory for g-function again, we have that
Therefore
Combining the estimates for \(\textrm{R}_{i}^1\), \(\textrm{R}_i^2\) and \(\textrm{R}_{i,m}^{3,n}\) (with \(1\le m\le d, \,n=1,\,2,\,3\)) in all yields that
since
It remains to discuss the contribution of terms II and III. For \(i\in {\mathbb {N}}\cup \{0\}\), by Lemma 2.3, one gets
Note that
Thus
which further implies that
Similarly, we have that
Therefore, these inequalities, together with the fact that \(E_{1,1}^0=\emptyset \) may lead to
Inequality (2.11), together with the estimate (2.9) for \(\textrm{I}\), gives that
2.2 Contribution of \({E_2(j,s,t)}\)
Let \(\alpha \in (\frac{d+1}{d+2},\,1)\), \(i\in {\mathbb {N}}\cup \{0\}\), and write
Firstly, we consider the term \( \textrm{IV}\). When \(i=0\), the integral \(\int ^{2^j}_{2^{j-Ni}}\int ^{(2^js^{-\alpha })^{1/{(1-\alpha )}}}_{2^j}\frac{dt}{t}\frac{ds}{s}\) vanishes, we only need to consider the case \(i\in {\mathbb {N}}\). Since \(s>2^{j-Ni}\), then \((2^j s^{-\alpha })^{\frac{1}{1-\alpha }}\le 2^j 2^{iN\frac{\alpha }{1-\alpha }}\). Therefore
where \(h_{s,l}(x)=Q_sg(x)\chi _{I_l}(x)\), and \(\{I_l\}_{l}\) be the cubes in (2.7).
Observe that when \(x\in 4dI_l\), \( T_{\Omega ,A,j}^i(Q_t^4f)(x)Q_s^3h_{s,l}(x)= T_{\Omega ,A,j}^i(\zeta _lQ_t^4f)(x)Q_s^3h_{s,l}(x), \) we rewrite
Similar to the estimate for \(\textrm{R}_i^1\) and \(\textrm{R}_i^2\), we know that
and
On the other hand, for each fixed \(1\le m\le d\), the same reasoning as what we have done for \({\textrm{R}^{3,1}_{i,m}}\) and \({\textrm{R}^{3,2}_{i,m}}\) yields that
and
Note that if \(x\in 4dI_l(x)\), then \(U_{\Omega ,\,m,j}^i(Q_tQ_t(\partial _m\widetilde{A_{I_l}}Q_t^2f))(x)=U_{\Omega ,\,m,j}^i(\zeta _lQ_tQ_t(\partial _m\widetilde{A_{I_l}}Q_t^2f))(x). \) Since the kernel of \(Q_t\) is radial and it enjoys the property that
Hence, we have
A trivial argument then yields that
Now we write
A standard argument involving Hölder’s inequality leads to that
Similarly, one can verify that
and
On the other hand, the fact (see [20, Lemma 4 and Lemma 3])
implies that
Applying Hölder’s inequality and inequality (2.8) in the case \(s\in (2^{j-1}, 2^j)\), we obtain
Collecting the estimates from (2.14) to (2.23) in all, we deduce that
To show the estimate for \(\mathrm V\), note that for each fixed j, it holds that
It then follows that
and
Thus, by Lemma 2.3, we obtain
Combining estimates (2.24)–(2.25) yields
Therefore, by (2.13), (2.24) and (2.26), it holds that
which gives the contribution of \(E_2(j,s,t)\).
To finish the proof of (2.4), it remains to show the contribution of the term \(E_3^i(j,s,t)\).
2.3 Contribution of \({E_3(j,s,t)}\)
Our aim is to prove
where
Since the sum of i and the sum of j are independent and the sum of j depends only on the functions \(\phi _j\) and \({\chi _{E_{3}}(j,s,t)}, \) one may put \(\phi _j \cdot {\chi _{E_{3}}(j,s,t)}\) together in the place of \(\phi _j\) in (2.28), and temporary moves the summation over j before \(\phi _j \cdot {\chi _{E_{3}}(j,s,t)}\), which indicates that it is possible to move the summation over i inside the integral again before \(\Omega _i\) to obtain \(\Omega \). After that, one may move the sum of j outside the integral. Therefore, to prove (2.27), it suffices to show that
To this purpose, we set
Let \(H^{(1)}\) be the integral operator corresponding to kernel \(h^{(1)}\). It then follows that
Applying Lemma 2.5 and reasoning as the same argument as in [18, p. 1282] give us that
where \(\gamma =(d+2)\alpha -d-1\). This in turn indicates that \(|H^{(1)}Q_tf(x)|\lesssim \big (\frac{s}{t}\big )^{\gamma }M(Q_tf)(x). \) Therefore
It remains to show the corresponding estimate for the second term on the rightside of (2.30).
Let \(F_x^j(s,t)=(Q_sT_{\Omega ,A;\,j}1)(x)Q_t^4f(x)Q_s^3g(x)\). Then
Therefore, it is sufficient to consider the contributions of each terms in Eq. (2.32) to the second term in (2.30).
Consider the first term in (2.32). Let \(P_s=\int ^{\infty }_sQ_t^4\frac{dt}{t}.\) Han and Sawyer [16] observed that the kernel \(\Phi \) of the convolution operator \(P_s\) is a radial bounded function with bound \(cs^{-d}\), supported on a ball of radius Cs and has integral zero. Therefore, it is easy to see that \(\Phi \) is a Schwartz function. Since \( {P_sg=\Phi _s * g}\), it then follows from the Littlewood-Paley theory that
On the other hand, whenever \(\Omega \in L^1({\mathbb {R}}^d)\), it was shown in [19, p.121, Lemma 4.1] that \(T_{\Omega ,A}1\equiv b\in \textrm{BMO}({\mathbb {R}}^d)\). Therefore, by [19, p.114, (3.1)], \(\int ^{\infty }_0Q_s^3(Q_sbP_s)\frac{ds}{s}\) defines an operator which is bounded on \(L^2({\mathbb {R}}^d)\). However, we can’t use this boundedness directly in our case, since once using Hölder’s inequality, we have to put the absolute value inside the integral and the \(L^2({\mathbb {R}}^d)\) boundedness may fail in this case. To overcome this obstacle, we apply the property of Carleson measure.
Note that \(|Q_s T_{\Omega ,\,A}1(x)|^2\frac{dxds}{s}\) is a Carleson measure since \(T_{\Omega ,\,A}1\in \textrm{BMO}({\mathbb {R}}^d)\). By Hölder’s inequality, Lemma 2.7, it yields that
On the other hand, by Lemma 2.2, one gets
Denote by \(D^1_{j,s,t}=\{(j,s,t): s\le t\le 2^j\}\), \(D^2_{j,s,t}=\{(j,s,\,t):\,s\le t,s^{\alpha }t^{1-\alpha }\le 2^j\le t\}\). It then follows that
where in the last inequality, we used the property (2.2).
Combining (2.32)–(2.34), we have
which, together with (2.31), leads to (2.27). This finishes the proof of \(E_3(j,s,t)\), and also completes the proof of inequality (2.4).
2.4 Proof of (2.5)
To finish the proof of Theorem 1.1, it remains to show the estimate (2.5). Observe that
where \({\widetilde{\Omega }}(x)=\Omega (-x)\) and \({\widetilde{T}}_{{\widetilde{\Omega }},\,A}\) is the operator defined by (1.5), with \(\Omega \) replaced by \({\widetilde{\Omega }}\). Let \(T_{{\widetilde{\Omega }},\,m}\) be the operator defined by
It then follows that
Inequality (2.4) tells us that
For each fixed m with \(1\le m\le d\), by duality, involving Lemma 2.4 and Hölder’s inequality may lead to
This estimate, together with (2.35), leads to (2.5) and then completes the proof of Theorem 1.1.
3 Proof of Theorems 1.2 and 1.3
This section is devoted to prove Theorem 1.2, the weak type endpoint estimates for \(T_{\Omega ,\,A}\) and \({\widetilde{T}}_{\Omega ,A}\). To this end, we first introduce the definition of standard dyadic grid. Recall that the standard dyadic grid in \({\mathbb {R}}^d\), denoted by \({\mathcal {D}}\), consists of all cubes of the form
For each fixed \(j\in {\mathbb {Z}}\), set \({\mathcal {D}}_j=\{Q\in {\mathcal {D}}:\, \ell (Q)=2^j\}\).
3.1 Proof of (1.6) in Theorem 1.2
The key ingredient of our proof lies in the step of dealing with the bad part of the Calderón-Zygmund decomposition of f. By homogeneity, it suffices to prove (1.6) for the case \(\lambda =1\). Applying the Calderón-Zygmund decomposition to \(|f|\log (\textrm{e}+|f|)\) at level 1, we can obtain a collection of non-overlapping closed dyadic cubes \({\mathcal {S}}=\{{\mathbb {L}}\}\), such that
-
(i)
\(\Vert f\Vert _{L^{\infty }({\mathbb {R}}^d\backslash \cup _{{\mathbb {L}}\in {\mathcal {S}}}{\mathbb {L}})}\lesssim 1;\)
-
(ii)
\(\int _{{\mathbb {L}}}|f(x)|\log (\textrm{e}+|f(x)|)dx \lesssim |{\mathbb {L}}|; \)
-
(iii)
\(\sum _{{\mathbb {L}}\in {\mathcal {S}}}|{\mathbb {L}}|\lesssim \int _{{\mathbb {R}}^d}|f(x)|\log (\textrm{e}+|f(x)|)dx\).
Let g be the good part and b be the bad part of the decomposition of f, which are defined by
It is easy to see that \(\Vert g\Vert _{L^{\infty }({\mathbb {R}}^d)}\lesssim 1\), and for \(E=\cup _{{\mathbb {L}}\in {\mathcal {S}}}100d{\mathbb {L}}\), it holds that
The \(L^2({\mathbb {R}}^d)\) boundedness of \(T_{\Omega ,\,A}\) then yields that
Therefore, it is sufficient to show that
To prove (3.2), let \(\phi \) be a smooth radial nonnegative function on \({\mathbb {R}}^d\) with \(\textrm{supp}\, \phi \subset \{x:\frac{1}{4}\le |x|\le 1\}\) and \(\sum _s\phi _s(x)=1\) with \(\phi _s(x)=\phi (2^{-s}x)\) for all \(x\in {\mathbb {R}}^d\backslash \{0\}\). Set \({\mathcal {S}}_{j}=\{{\mathbb {L}}\in {\mathcal {S}}:\, \ell ({\mathbb {L}})=2^j\}\). Then, we have
where
Let \(A_{{\mathbb {L}}}(y)=A(y)-\sum _{n=1}^d\langle \partial _n A\rangle _{\mathbb {L}}y_n.\) A trivial computation leads to the fact that
Now write \(T_{\Omega ,A}b\) as
where
Fixed \(1\le n\le d\), since the kernel \( {\Omega (x)x_n}{|x|^{-1}}\) is still in \(L\log L {({\mathbb {S}}^{d-1})}\), homogenous of degree zero and satisfies the vanishing condition on the unit sphere, by the weak endpoint estimate of the operators \(T_{\Omega }^n\) (see [31] or [9]), it follows that
where in the last inequality, we have used the fact that \(\Vert b_{\mathbb {L}}\Vert _{L\log L,\,{\mathbb {L}}}\lesssim 1\) for each cube \({\mathbb {L}}\in {\mathcal {S}}\).
Therefore, to prove inequality (1.6), by (3.1), (3.2) and (3.4), it is sufficient to show that
In order to prove inequality (3.5), we first give some estimate for \(\sum _s\sum _{{\mathbb {L}}\in {\mathcal {S}}_{s-j}}T_{\Omega ,\,A_{\mathbb {L}};s,j}b_{{\mathbb {L}}}\). For this purpose, we need to introduce some notations.
For \({\mathbb {L}}\in {\mathcal {S}}_{s-j}\), s, \(j\,\in {\mathbb {Z}}\) with \(j\ge \log _2(100d/2)=:j_0\). Let \(L_{j,1}=2^{j+2}d{\mathbb {L}}\), \(L_{j,2}=2^{j+4}d{\mathbb {L}}\), \(L_{j,3}=2^{j+6}d{\mathbb {L}}\), and \(y^j_{{\mathbb {L}}}\) be a point on the boundary of \(L_{j,3}\). Set
where \(\varphi _{{\mathbb {L}}}\in C^{\infty }_c({\mathbb {R}}^d)\), \(\textrm{supp}\,\varphi _{{\mathbb {L}}}\subset L_{j,1}\), \(\varphi _{{\mathbb {L}}}\equiv 1\) on \(3\cdot 2^{j}d{\mathbb {L}}\), and \(\Vert \nabla \varphi _{{{\mathbb {L}}}}\Vert _{L^{\infty }({\mathbb {R}}^d)}\lesssim 2^{-s}. \) Let \(y_0\) be the center point of \({\mathbb {L}}\). Observe that for \(x\in {\mathbb {R}}^d\backslash E\), \(j\le j_0\), \(y\in {\mathbb {L}}\), we have \(|x-y|\ge |x-y_0|-|y-y_0|>2^s\). The support condition of \(\phi \) then implies that \(T_{\Omega ,A_{{\mathbb {L}}};s,j}b_{\mathbb {L}}(x)=0\) if \(j\le j_0\). For \(y\in {\mathbb {L}}\in {\mathcal {S}}_{s-j}\), \(s,\,j\in {\mathbb {Z}}\) with \(j> j_0\), we have \(\varphi _{{\mathbb {L}}}(y)= 1\). By the support condition of \(\phi \), it follows that \(|x-y_0|\le |x-y|+|y-y_0|\le 1.5d2^s \). Hence \(x\in 3\cdot 2^{j}d{\mathbb {L}}\) and \(\varphi _{{\mathbb {L}}}(x)= 1\). Collecting these facts in all, it follows that
The kernel \(\Omega \) will be decomposed into disjoint forms as in Section 2 as follows:
where \(E_0=\{\theta \in {\mathbb {S}}^{d-1}:\, |\Omega (\theta )|\le 1\} \ \hbox {and } E_k=\{\theta \in {\mathbb {S}}^{d-1}:\,2^{k-1}<|\Omega (\theta )|\le 2^k\}\) for \( k\in {\mathbb {N}}.\)
Let the operator \( T^i_{\Omega ,A_{{\mathbb {L}}};\,s,j}b_{\mathbb {L}}\) be defined in the same form as \( T_{\Omega ,A_{{\mathbb {L}}};\,s,j}b_{\mathbb {L}}\), with \(\Omega \) replaced by \(\Omega _i\). Then we can divide the summation of \(T_{\Omega ,A_{{\mathbb {L}}};\,s}b_{\mathbb {L}}\) into two terms as follows
where N is some constant which will be chosen later. If we can verify that
and
the inequality (1.6) then follows directly. The proofs of these two estimate will be given in the next two subsections respectively.
3.2 Proof of Inequality (3.6)
We first claim that if \({\mathbb {L}}\in {\mathcal {S}}_{s-j}\), then
This claim is a consequence of the following lemma, which will also be used several times later.
Lemma 3.1
Let A be a function in \({\mathbb {R}}^d\) with derivatives of order one in \(\textrm{BMO}({\mathbb {R}}^d)\). Let \(s,\,j\in {\mathbb {Z}}\) and \( {\mathbb {L}}\in {\mathcal {S}}_{s-j}\) with \(j>j_0\) and let \(R_{s,{\mathbb {L}};j}(x,y)\) be the function on \({\mathbb {R}}^d\times {\mathbb {R}}^d\) defined by
Then, \(R_{s,{\mathbb {L}};j}\) enjoys the properties that
-
(i)
For any \(x,\, y\in {\mathbb {R}}^d\),
$$\begin{aligned} \big |R_{s,{\mathbb {L}};j}(x,y)\big |\lesssim \frac{j}{|x-y|^{d}}\chi _{\{2^{s-2}\le |x-y|\le 2^{s+2}\}}(x,\,y); \end{aligned}$$ -
(ii)
For any \(x,\,x'\in {\mathbb {R}}^d\) and \(y\in {\mathbb {L}}\) with \(|x-y|>2|x-x'|\),
$$\begin{aligned} \big |R_{s,{\mathbb {L}};j}(x,y)-R_{s,{\mathbb {L}};j}(x',y)\big |\lesssim \frac{|x-x'|}{|x-y|^{d+1}}\Big (j+\Big |\log \big ({2^{s-j}}{|x-x'|^{-1}}\big )\Big |\Big ); \end{aligned}$$ -
(iii)
For any \(x,\,y'\in {\mathbb {R}}^d\) and \(y\in {\mathbb {L}}\) with \(|x-y|>2|y-y'|\),
$$\begin{aligned} \big |R_{s,{\mathbb {L}};j}(x,y)-R_{s,{\mathbb {L}};j}(x,\,y')\big |\lesssim \frac{|y-y'|}{|x-y|^{d+1}}\Big (j+\Big |\log \big ({2^{s-j}}{|y-y'|^{-1}}\big )\Big |\Big ). \end{aligned}$$
Proof
We first prove (i). It is obvious that \(\textrm{supp}\,R_{s,{\mathbb {L}};j} \subset {L_{j,2}\times L_{j,2}}\). Fixed \(x\in L_{j,1}\), we know that \(2^{s-j}<|x-y_{\mathbb {L}}^j|\) and
Note that if \(x\in 4{\mathbb {L}}\), then \({I_{(x, 2^{s-j})}}\subset 8{\mathbb {L}}\) and it holds that
If \(x\in L_{j,1}\backslash 4{\mathbb {L}}\), then the center of \({\mathbb {L}}\) and the center of \({I_{(x, 2^{s-j})}}\) are at a distance of \(a2^{s-j}\) with \(a>1\). Hence, the results in [13, Proposition 3.1.5, p. 158 and 3.1.5–3.1.6, p. 166.] gives that
since \(2^s<|x-y_{\mathbb {L}}^j|<2^{s+5+d^2}\).
Therefore, for \(x\in {\mathbb {L}}_{j,1}\), it holds that
Lemma 2.6, together with John-Nirenberg inequality then gives that
which finishes the proof of (i).
Now we give the proof of (ii). For any \(x,\,x'\in {\mathbb {R}}^d\) and \(y\in {\mathbb {L}}\) with \(|x-y|>2|x-x'|\), it is easy to see that
-
(1)
if \(x\notin L_{j,1}\) and \(x'\notin L_{j,1}\), then \(R_{s,{\mathbb {L}};j}(x,y)=R_{s,{\mathbb {L}};j}(x',y) =0\);
-
(2)
if \(x\notin L_{j,1}\), then \(x'\notin 3\cdot 2^{j}d{\mathbb {L}}\), hence \(R_{s,{\mathbb {L}};j}(x,y)=R_{s,{\mathbb {L}};j}(x',y) =0\);
-
(3)
if \(x'\notin L_{j,1}\), then \(x\notin 3\cdot 2^{j}d{\mathbb {L}}\), hence \(R_{s,{\mathbb {L}};j}(x,y)=R_{s,{\mathbb {L}};j}(x',y) =0\).
If \(z\in I_{(x,|x-x'|)}\), another application of Lemma 2.6 and John-Nirenberg inequality indicates
and the similar method as what was used in the proof of (3.8) further implies that
By Lemma 2.6, (3.10) and (3.11), we have
Similarly, we obtain
In a similar way, we have
Therefore,
This completes the proof of (ii) in Lemma 3.1. (iii) can be proved in the same way as (ii). \(\square \)
Let us turn back to the contribution of \(D_1.\) It follows from the method of rotation of Calderón-Zygmund that
This verifies (3.6).
3.3 Proof of the Inequality (3.7)
The estimate of \(\mathrm{D_2}\) is long and complicated. We split the proof into three steps.
Step 1. A reduction for the estimate of \(D_2\).
Let \(l_{\tau }(j)=\tau j+3\), where \(0<\tau <1\) will be chosen later. Let \(\omega \) be a nonnegative, radial \(C^{\infty }_c({\mathbb {R}}^d)\) function which is supported in \(\{x\in {\mathbb {R}}^d:|x|\le 1\}\) and has integral 1. Set \(\omega _{t}(x)=2^{-td}\omega (2^{-t}x)\). For \(s\in {\mathbb {N}}\) and a cube \({\mathbb {L}}\), we define \(R^j_{s,{\mathbb {L}}}\) as
It is obvious that \(\textrm{supp}R_{s,{\mathbb {L}}}^{j}(x,y)\subset \{(x,\,y):\, 2^{s-3}\le |x-y|\le 2^{s+3}\} \). Moreover, if \(y\in {\mathbb {L}}\) with \({{\mathbb {L}}\in {\mathcal {S}}_{s-j}} \), then (i) of Lemma 3.1 implies that
We define the operator \(T^{i,j}_{\Omega ,{\mathbb {L}};\,s}\) by
and let \(\textrm{D}_{2}^*\) be the operator as follows
The following lemma indicates the intrinsically close relationship in each subtract terms between \(\textrm{D}_{2}\) and \(\textrm{D}_{2}^*\). Thus, the corresponding proof is transferred to verify it for each term of \(\textrm{D}_{2}^*\).
Lemma 3.2
Let \(\Omega \) be homogeneous of degree zero, A be a function on \({\mathbb {R}}^d\) with derivatives of order one in \(\textrm{BMO}({\mathbb {R}}^d)\). For \(j>j_0\) and \(i\ge 0\), it holds that
Proof
For each \(y\in {\mathbb {L}}\) and \(z\in \textrm{supp}\,\omega _{s-l_{\tau }( j)}\), notice that \(R_{s,{\mathbb {L}};j}(x,\,y)-R_{s,{\mathbb {L}};j{\tiny }}(x-z,\,y)=0\) if \(x\in L_{j,1}\backslash {3\cdot 2^{j}d{\mathbb {L}}}\). In fact, since \(|z|\le 2^{s-\tau j-3}\), then we have \(2^{s+1}<|x-y|<3\cdot 2^s\) and \(2^s<|x-y-z|<2^{s+2}\).
By Lemma 3.1, we have
Observing that the function \(\Theta (t)=t\log (\textrm{e}+\frac{1}{t})\) is bounded at \(t\in (0,\,1]\), and then for \(0<t\le r\),
we deduce that
Therefore
This leads to the desired conclusion of Lemma 3.2. \(\square \)
With Lemma 3.2 in hand, we only need to estimate \(\textrm{D}_{2}^*\). This is the content of the second step.
Step 2. Estimate for each term of \(\textrm{D}_{2}^*\).
Define \(P_tf(x)=\omega _t*f(x)\). Now we split
where \(\kappa \in (0,1)\) will be chosen later. In the following, we will estimate this two terms one by one. We have the following norm inequality for \(P_{s-j\kappa }T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s}\).
Lemma 3.3
Let \(\Omega \) be homogeneous of degree zero, A be a function in \({\mathbb {R}}^d\) with derivatives of order one in \({\textrm{BMO}}({\mathbb {R}}^d)\), \(b_{\mathbb {L}}\) satisfies the vanishing moment with \(\ell ({\mathbb {L}})=2^{s-j}\). For each \(j\in {\mathbb {N}}\) with \(j>j_0\), we have
Before proving Lemma 3.3, we need the following lemma for \(R_{s,{\mathbb {L}}}^{j}\).
Lemma 3.4
Let \(R^j_{s,{\mathbb {L}}}\) be defined as (3.14), \(\theta \in {\mathbb {S}}^{d-1}\), \(y,\,y'\in {\mathbb {L}}\) with \(\ell ({\mathbb {L}})=2^{s-j}\). Then
Proof
By the triangle inequality, the mean value theorem and the support condition of \(\phi \), we get
If \(r\notin [2^{s-4}, 2^{s+4}]\), by the support of \(R^j_{s,{\mathbb {L}}}\), it gives that \(|R^j_{s,{\mathbb {L}}}(y'+r\theta ,\,y)-R^j_{s,{\mathbb {L}}}(y'+r\theta ,\,y')|=0\).
For \(y,\,y'\in {\mathbb {L}}\), (3.13) gives us that
For \(\textrm{I}\), since \(|z-y'|\ge 2^{s-2}\), \(y,\,y'\in {\mathbb {L}}\), then (3.13) gives us that
Consider now the other two terms. If \(y,\,y'\in {\mathbb {L}}\) and \(|z-y'|\le 2^{s}\), (i) of Lemma 3.1 gives us that
On the other hand, for \(j>j_0\), when \(y,\,y'\in {\mathbb {L}}\) and \(|z-y'|\ge 2^{s-2}\), it holds that
Therefore,
where the second inequality follows from the fact that
Similarly, we have
Estimates for \(\textrm{I}\), \(\textrm{II}\) and \(\textrm{III}\) above lead to that
Similar to (3.16), we also have
Notice that
Combining (3.16) with (3.17), it gives that
This finishes the proof of Lemma 3.4. \(\square \)
With Lemma 3.4, we are ready to prove Lemma 3.3 now.
Proof of Lemma 3.3
Write
Let \(z-y=r\theta \). By Fubini’s theorem, \( P_{s-j\kappa } T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s}b_{\mathbb {L}} (x)\) can be written as
Let \(y' \in {\mathbb {L}}\). By the vanishing moment of \(b_{\mathbb {L}}\), we have
where
and
Note that \(|y-y'|\lesssim 2^{s-j}\), when y, \(y'\in {\mathbb {L}}\). By (3.15) and the mean value formula, it follows that
By Lemma 3.4 and the Fubini’s theorem one can get
This finishes the proof of Lemma 3.3. \(\square \)
To estimate the term \((I-P_{s-j\kappa })T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s}\), we introduce a partition of unity on the unit surface \({\mathbb {S}}^{d-1}\). For \(j>j_0\), let \({\mathfrak {E}}^j=\{e_{\nu }^j\}\) be a collection of unit vectors on \({\mathbb {S}}^{d-1}\) such that
-
(a)
for \(\nu \not =\nu '\), \(|e_{\nu }^j-e_{\nu '}^j|>2^{-j\gamma -4}\);
-
(b)
for each \(\theta \in {\mathbb {S}}^{d-1}\), there exists an \(e^j_{\nu }\) satisfying that \(|e^j_{\nu }-\theta |\le 2^{-j\gamma -4},\) where \(\gamma \in (0,\,1)\) is a constant.
The set \({\mathfrak {E}}^j\) can be constructed as in [31]. Observe that \(\textrm{card}({\mathfrak {E}}^j)\lesssim 2^{j\gamma (d-1)}\).
Below, we will construct an associated partition of unity on the unit surface \({\mathbb {S}}^{d-1}\). Let \(\zeta \) be a smooth, nonnegative, radial function with \(\zeta (u)\equiv 1\) when \(|u|\le 1/2\) and \(\textrm{supp}\, \zeta \subset \{|x|\le 1\}\). Set
It is easy to verify that \(\Gamma ^j_{\nu }\) is homogeneous of degree zero, and for all j and \(\xi \in {\mathbb {S}}^{d-1}\), \(\sum _{\nu }\Gamma ^j_{\nu }(\xi )=1. \) Let \({\widetilde{\psi }}\in C^{\infty }_c({\mathbb {R}})\) such that \(0\le {\widetilde{\psi }}\le 1\), \(\textrm{supp}\, {\widetilde{\psi }}\subset [-4,\,4]\) and \({\widetilde{\psi }}(t)\equiv 1\) when \(t\in [-2,\,2]\). Define the multiplier operator \(G_{\nu }^j\) by
Denote the operator \(T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }\) by
It is obvious that \(T^{i,j}_{\Omega ,{\mathbb {L}};\,s}h(x)=\sum _{\nu }T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }h(x). \) For each fixed \(i,\,s,\,j,\,{\mathbb {L}}\) and \(\nu \), \((I-P_{s-j\kappa })T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }\) can be decomposed in the following way
Estimate for the term \(G_{\nu }^j(I-P_{s-j\kappa })T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }\).
For the term \(G_{\nu }^j(I-P_{s-j\kappa })T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }\), we have the following lemma.
Lemma 3.5
Let \(\Omega \) be homogeneous of degree zero, A be a function in \({\mathbb {R}}^d\) with derivatives of order one in \(\textrm{BMO}({\mathbb {R}}^d)\). For each \(j\in {\mathbb {N}}\) with \(j> j_0\), we have that,
Proof
The proof is similar to the proof of Lemma 2.3 in [9]. For the sake of self-contained, we present the proof here. Observe that
This, together with Plancherel’s theorem and Cauchy-Schwartz inequality, leads to that
Applying (3.15), we see that for each fixed s, \({\mathbb {L}}\), and \(x\in {\mathbb {R}}^d\),
where \( H_{s,\nu }^j(x)=2^{-sd}\chi _{{\mathcal {R}}_{s\nu }^j}(x)\), and \({\mathcal {R}}_{s\nu }^j=\{x\in {\mathbb {R}}^d:\,|\langle x,e_{\nu }^j \rangle |\le 2^{s+3},\,|x-\langle x,e_{\nu }^j\rangle e_{\nu }^j|\le 2^{s+3-j\gamma }\}. \) This means that \({\mathcal {R}}_{s\nu }^j\) is contained in a box having one long side of length \(\lesssim 2^s\) and \((d-1)\) short sides of length \(\lesssim 2^{s-j\gamma }\). Therefore, we have
Let \(\widetilde{{\mathcal {R}}}_{s\nu }^j={{\mathcal {R}}}_{s\nu }^j+{{\mathcal {R}}}_{s\nu }^j\). As in the proof of Lemma 2.3 in [9], for each fixed \({\mathbb {L}}\in {\mathcal {S}}_{s-j}\), \(x\in {\mathbb {L}}\), \(\nu \) and s, we obtain
where we have used the fact that \(\int _{{\mathbb {R}}^d} |b_I(y)|dy\lesssim |I|\) and the cubes \(I\in {\mathcal {S}}\) are pairwise disjoint.
This, in turn, implies further that
which finishes the proof of Lemma 3.5. \(\square \)
Estimate for the term \((I-G_{\nu }^j)\)\((I-P_{s-j\kappa })T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }.\)
We need to present a lemma for \((I-G_{\nu }^j)\)\((I-P_{s-j\kappa })T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }\).
Lemma 3.6
Let \(\Omega \) be homogeneous of degree zero, A be a function in \({\mathbb {R}}^d\) with derivatives of order one in \(\textrm{BMO}({\mathbb {R}}^d)\). For each \(j\in {\mathbb {N}}\) with \(j> j_0\), \(\ell ({\mathbb {L}})=2^{s-j}\), some \(s_0>0\), we have that
Next we give the estimate of \(\textrm{D}_{2}^*\) and postpone the proof of Lemma 3.6 later.
Let \(\varepsilon =\min \{ (1-\kappa ), (1-\tau ), s_0,\gamma \}\). With Lemma 3.3, Lemma 3.5 and Lemma 3.6, we have
The proof of Lemma 3.6 is similar to the proof of Lemma 2.4 in [9]. For the completeness of this paper, we give the proof for the remaining term \((1-G_{\nu }^j)(I-P_{s-j\kappa })T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }\) here. Let’s introduce the Littlewood-Paley decomposition first. Let \(\alpha \) be a radial \(C^{\infty }\) function such that \(\alpha (\xi )=1\) for \(|\xi | \le 1\), \(\alpha (\xi )=0\) for \(|\xi | \ge 2\) and \(0\le \alpha (\xi )\le 1\) for all \(\xi \in {\mathbb {R}}^d\). Define \(\beta _k(\xi )=\alpha (2^k\xi )-\alpha (2^{k+1}\xi )\). Choose \({\tilde{\beta }}\) be a radial \(C^{\infty }\) function such that \({\tilde{\beta }}(\xi )=1\) for \(1/2\le |\xi | \le 2\), \({\textrm{supp}} \,{\tilde{\beta }} \in [1/4,4] \) and \(0\le {\tilde{\beta }}\le 1\) for all \(\xi \in {\mathbb {R}}^d\). Set \( {\tilde{\beta }}_k(\xi )={\tilde{\beta }}(2^k\xi ) \), then it is easy to see \(\beta _k={\tilde{\beta }}_k{\beta }_k\). Define the convolution operators \(\Lambda _k\) and \({\tilde{\Lambda }} _k\) with Fourier multipliers \(\beta _k\) and \(\tilde{\beta _k}\), respectively.
It is easy to have \(\Lambda _k={\tilde{\Lambda }} _k \Lambda _k\).
Proof of Lemma 3.6
We first write \((I-{G_{\nu }^j)}T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }= \sum \limits _{k}(I-{G_{\nu }^j)} \Lambda _k T^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu }\). Then
We can write
where \(M_k\) is the kernel of the operator \( (I-{G_{\nu }^j)}\Lambda _kT^{i,j}_{\Omega ,{{\mathbb {L}}};\,s,\nu } \). Then
Applying the method of Lemma 4.2 in [9], there exists \(M>0\) such that
Hence, note that \( \Vert (I-P_{s-j\kappa } ){\tilde{\Lambda }} _k\Vert _{L^1({\mathbb {R}}^d)\rightarrow L^1({\mathbb {R}}^d)} \le \Vert {\mathcal {F}}^{-1}( {\tilde{\beta }}_k)-\omega _{s-j\kappa }*{\mathcal {F}}^{-1}( {\tilde{\beta }}_k) \Vert _{L^1({\mathbb {R}}^d)}\lesssim 1\), we have
On the other hand, we can write
By (3.18), it is easy to show that
Let \(W_{k,s,\,\kappa }^j\) be the kernel of \((I-P_{s-j\kappa } ){\tilde{\Lambda }} _k\), then by the mean value formula, we obtain
By the proof of [26, Lemma 3.2], it holds that \(\Vert (I-{G_{\nu }^j)}\Lambda _k\Vert {_{L^1({\mathbb {R}}^d)\rightarrow L^1({\mathbb {R}}^d)}}\lesssim 1 \). Hence
Let \(m=s-[j\varepsilon _0]\), with \(0<\varepsilon _0<1\). Since \(\textrm{card}({\mathfrak {E}}^j)\lesssim 2^{j\gamma (d-1)}\). Then (3.21) and (3.23) lead to
where \(s_1=\big (\tau -\varepsilon _0+\gamma (1+2M)\big )\) and \(s_2=-\kappa +\varepsilon _0\).
We can now choose \(0\ll \tau \ll \gamma \ll \varepsilon _0< \kappa < 1\) such that \(\max \{s_1,s_2\}<0\). Let \(s_0=-\max \{s_1,s_2\}\), then the proof of Lemma 3.6 is finished now. \(\square \)
With Lemma 3.2 and (3.20) in hand, we can deduce (3.7) by
3.4 Proof of (1.7) in Theorem 1.2
It suffices to prove (1.7) for \(\lambda =1\). For a bounded function f with compact support, we employ the Calderón-Zygmund decomposition to |f| at level 1 then obtain a collection of non-overlapping dyadic cubes \({\mathcal {S}}=\{Q\}\), such that
Let \(E=\cup _{Q\in {\mathcal {S}}}100dQ\). With the same notations as in the proof of (1.6), for \(x\in {\mathbb {R}}^d\backslash E\), we write
By estimate (3.5), the proof of (1.7) can be reduced to show that for each n with \(1\le n\le d\),
But this inequality has already been proved in [22, inequality (3.3)]. Then the proof of (1.7) is finished. \(\square \)
3.5 Proof of Theorem 1.3
The proof of Theorem 1.3 is now standard. We present the proof here mainly to make the constant of the norm inequality clearly. Consider the case \(p\in (1,2]\). Let
and
By (1.6), we have
\(L^2({\mathbb {R}}^d)\) boundedness of \(T_{\Omega ,\,A}\) implies that
Since \({p\in (1,2)}\), we have
When \(p\in (2,\infty )\), by (1.7), we know \({\widetilde{T}}_{\Omega ,A}f(x)\) is of weak type (1, 1). Combining the \(L^2({\mathbb {R}}^d)\) boundedness of \({\widetilde{T}}_{\Omega ,A}f(x)\) and the Marcinkiewicz interpolation theorem, we have
By duality, it holds that
This completes the proof of Theorem 1.3. \(\square \)
Data Availability
Our manuscript has no associated data.
References
Bainshansky, B.M., Coifman, R.: On Singular Integrals. Proceedings of Symposia in Pure Mathematics, vol. 10, pp. 1–17. American Mathematical Society, Providence (1967)
Calderón, A.P.: Commutators of singular integral operators. Proc. Natl Acad. Sci. U.S.A. 53, 1092–1099 (1965)
Calderón, A.P., Zygmund, A.: On singular integrals. Am. J. Math. 78, 289–309 (1956)
Calderón, C.P.: On commutators of singular integrals. Stud. Math. 53, 139–174 (1975)
Christ, M.: Weak type \((1,1)\) bounds for rough operators. Ann. Math. (2) 128(1), 19–42 (1988)
Christ, M., Rubio de Francia, J.L.: Weak type \((1,1)\) bounds for rough operators II. Invent. Math. 93(1), 225–237 (1988)
Cohen, J.: A sharp estimate for a multilinear singular integral on \({\mathbb{R} }^d\). Indiana Univ. Math. J. 30, 693–702 (1981)
Connett, W.C.: Singular integral near \(L^1\). Proc. Sympos. Pure Math. 35, 163–165 (1979)
Ding, Y., Lai, X.: Weak type \((1,\,1)\) bounded criterion for singular integral with rough kernel and its applications. Trans. Am. Math. Soc. 371, 1649–1675 (2019)
Duoandikoetxea, J.: Weighted norm inequalities for homogeneous singular integrals. Trans. Am. Math. Soc. 336(2), 869–880 (1993)
Duoandikoetxea, J., Rubio de Francia, J.L.: Maximal and singular integral operators via Fourier transform estimates. Invent. Math. 84(3), 541–561 (1986)
Grafakos, L.: Classical Fourier Analysis, GTM249, 2nd edn. Springer, New York (2008)
Grafakos, L.: Modern Fourier Analysis, GTM250, 3rd edn. Springer, New York (2014)
Grafakos, L., Stefanov, A.: \(L^p\) bounds for singular integrals and maximal singular integrals with rough kernels. Indiana Univ. Math. J. 47, 455–469 (1998)
Grafakos, L., Stefanov, A.: Convolution Calderón–Zygmund singular integral operators with roughkernels. In: Analysis of Divergence. Applied and Numerical Harmonic Analysis, pp. 119–143. Birkhäuser Boston, Boston (1999)
Han, Y., Sawyer, E.T.: Para-accretive functions, the weak boundedness properties and the \(Tb\) theorem. Rev. Mat. Iberoam. 6, 17–41 (1990)
Hofmann, S.: Weak \((1,1)\) boundedness of singular integrals with nonsmooth kernel. Proc. Am. Math. Soc. 103(1), 260–264 (1988)
Hofmann, S.: Weighted norm inequalities for commutators of rough singular integrals. Indiana Univ. Math. J. 39, 1275–1304 (1990)
Hofmann, S.: On certain non-standard Calderón-Zygmund operators. Studia Math. 109, 105–131 (1994)
Hu, G.: \(L^p\) boundedness for the commutator of a homogeneous singular integral operator. Stud. Math. 154, 13–27 (2003)
Hu, G.: \(L^2({\mathbb{R} }^n)\) boundedness for a class of multilinear singular integral operators. Acta Math. Sinica. 19, 397–404 (2003)
Hu, G., Tao, X.: An endpoint estimate for the commutators of singular integral operators with rough kernels. Potential Anal. 58, 241–262 (2023)
Hu, G., Yang, D.: Sharp function estimates and weighted norm inequalities for multilinear singular integral operators. Bull. Lond. Math. Soc. 35, 759–769 (2003)
Hu, Y.: An estimate for multilinear singular integrals on \({\mathbb{R}}^d\), Beijing Daxue Xuebao,(3) 19-26, (1985). (Chinese. English summary)
Hytönen, T., Pérez, C.: The \(L(\log L)^{\epsilon }\) endpoint estimate for maximal singular integral operators. J. Math. Anal. Appl. 428, 605–626 (2015)
Lai, X.: Vector-valued singular integral operators with rough kernels. Math. Nachr. 296, 2417–2439 (2023)
Lerner, A.K.: A simple proof of the \(A_2\) conjecture. Int. Math. Res. Not. 14, 3159–3170 (2013)
Lerner, A.K.: A weak type estimates for rough singular integrals. Rev. Mat. Iberoam. 35, 1583–1602 (2019)
Pérez, C.: Endpoint estimates for commutators of singular integral operators. J. Funct. Anal. 128, 163–185 (1995)
Ricci, F., Weiss, G.: A characterization of \(H^1({\mathbb{S} }^{d-1})\). Proc. Sympos. Pure Math. 35, 289–294 (1979)
Seeger, A.: Singular integral operators with rough convolution kernels. J. Am. Math. Soc. 9, 95–105 (1996)
Stein, E.M.: Harmonic Analysis, Real Variable Methods, Orthogonality, and Oscillatory Integrals. Princeton University Press, Princeton (1993)
Tao, T.: The weak-type (1,1) of \(L\log L\) homogeneous convolution operator. Indiana Univ. Math. J. 48(4), 1547–1584 (1999)
Vargas, A.M.: Weighted weak type \((1,1)\) bounds for rough operators. J. Lond. Math. Soc. (2) 54(2), 297–310 (1996)
Watson, D.K.: Weighted estimates for singular integrals via Fourier transform estimates. Duke Math. J. 60(2), 389–399 (1990)
Acknowledgements
The authors want to express their sincere thanks to the referee for his/her valuable remarks and suggestions, which made this paper more readable. The author would also like to thank Dr. Xudong Lai for helpful discussions.
Funding
The author X. Tao was supported by the NNSF of China (No. 12271483), Z. Wang and Q. Xue were partly supported by the National Key Research and Development Program of China (Grant No. 2020YFA0712900), NSFC (No. 12271041), NSF of Jiangsu Province of China (Grant No. BK20220324) and Natural Science Research of Jiangsu Higher Education Institutions of China (Grant No. 22KJB110016).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors state that there is no conflict of interest.
Additional information
Communicated by Elena Cordero.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Hu, G., Tao, X., Wang, Z. et al. On the Boundedness of Non-standard Rough Singular Integral Operators. J Fourier Anal Appl 30, 32 (2024). https://doi.org/10.1007/s00041-024-10086-y
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00041-024-10086-y