Abstract
In this paper, we study the quantitative weighted bounds for the q-variational singular integral operators with rough kernels, a stronger nonlinearity than the maximal truncations. The main result is for the truncated singular integrals itself
it is the best known quantitative result for this class of operators. In the course of establishing the above estimate, we obtain several quantitative weighted bounds which are of independent interest.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction and Statement of Main Results
1.1 Background and Main Result
Let \({\mathcal {F}}=\{F_t:t\in {\mathcal {I}}\subset \mathbb {R}\}\) be a family of Lebesgue measurable functions defined on \(\mathbb {R}^n\). For x in \(\mathbb {R}^n\), the value of the q-variation function \(V_q({\mathcal {F}})\) of the family \({\mathcal {F}}\) at x is defined by
where the supremum runs over all increasing subsequences \(\{t_k:k\ge 0\}\subset {\mathcal {I}}.\) Specially, when \(q=\infty ,\)
where the supremum runs over all increasing subsequences \(\{t_k:k\ge 0\}\subset {\mathcal {I}}.\) It is trivial that \( \sup _{t\in \mathbb R}|F_t(x)|\le V_\infty (\mathcal F)(x)+|F_{t_*}(x)|\) for any \(t_*\in {\mathcal {I}}.\)
Suppose \(\mathcal {A}=\{{A}_t\}_{t\in {\mathcal {I}}}\) is a family of operators on \(L^p(\mathbb R^n)\, (1\le p\le \infty )\), the associated strong q-variation operator \(V_q{\mathcal {A}}\) is defined as
There are two elementary but important observations that motivate the development of variational inequalities in ergodic theory and harmonic analysis. The first one is that from the fact that \(V_q({\mathcal {F}})(x)<\infty \) with finite q implies the convergence of \(F_t(x)\) as \(t\rightarrow t_0\) whenever \(t_0\) is an adherent point for \({\mathcal {I}}\), it is easy to observe that \(A_t(f)\) converges a.e. for \(f\in L^p\) whenever \(V_q{\mathcal {A}}\) for finite q is of weak type (p, p) with \(p<\infty \). The second one is that q-variation function dominates pointwisely the maximal function: for any \(q\ge 1\),
where \(A^*\) is the maximal operator defined by \(A^*(f)(x):=\sup _{t\in {\mathcal {I}}}|{A}_t(f)(x)|\) and \(t_0\in {\mathcal {I}}\) is any fixed number.
Because of the first observation, Bourgain [1] proved in ergodic theory the first variational estimate which had originated from the regularity of Brownian motion in martingale theory [28, 38] since the pointwise convergence may not hold for any non-trivial dense subclass of functions in some ergodic models and thus the maximal Banach principle does not work, see [22, 23] for further results in this direction. Because of the second observation, many maximal inequalities in harmonic analysis, such as maximal singular integrals, maximal operators of Radon type, the Carleson–Hunt theorem, Cauchy type integrals, dimension free Hardy–Littlewood maximal estimates etc, have been strengthened to variational inequalities [2, 4, 5, 24, 33,34,35,36,37, 43]. There are also many other publications on vector-valued and weighted norm estimates coming to enrich the literature on this subject (cf. e.g., [12, 16, 17, 24, 27]).
In this paper, we continue to study q-variational estimates but focus on the quantitative weighted bounds for singular integrals with rough kernels.
For \(\varepsilon >0,\) suppose that \(T_{\Omega ,\varepsilon }\) is the truncated homogeneous singular integral operator defined by
where \(y'={y/|y|}\), \(\Omega \in L\log ^+\!\!L(\mathbb S^{n-1})\) is homogeneous of degree zero and satisfies the cancellation condition
Denote the family \(\{T_{\Omega , \varepsilon }\}_{\varepsilon >0}\) of operator by \({\mathcal {T}}_\Omega \). For \(1< p<\infty \) and \(f\in L^p(\mathbb R^n)\), the Calderón–Zygmund singular integral operator \(T_\Omega \) with homogeneous kernel can be defined by
Considering the influence of (1.2) for any \(f\in L^p(\mathbb R^n)\) and \(x\in \mathbb R^n\) yields the inequality expression
where \(T_\Omega ^*\) which is given by \(T_\Omega ^*f(x)=\sup _{\varepsilon >0}|T_{\Omega , \varepsilon } f(x)|\) together with the maximal Banach principle.
In 2002, Campbell et al. [5] first proved the variational inequalities for \( T_\Omega \) extending their previous results for the Hilbert transform [4], which gives extra information on the speed of convergence in (1.5). See also [10, 24] for more results.
Theorem A. ([5]) Let \(\Omega \in L\log ^+\!\!L(\mathbb S^{n-1})\) satisfying (1.4) and \({\mathcal {T}}_\Omega \) be as defined above. Then the variational inequality
holds for \(1<p<\infty \) and \(q>2\).
The first weighted norm q-variational inequality, to the authors’ best knowledge, is due to Ma et al. [31]. Later on, their results were extended to higher dimensions [26, 32]. The weighted version of Theorem A was also considered in [7]. On the other hand, motivated by the solution of the \(A_2\) conjecture, Hytönen et al. [20] established the sharp weighted inequality for the variation of the smooth truncations of Calderón–Zygmund operator with \(\omega \)-Dini type regularity; in [9], Franca Silva and Zorin-Kranich extended the sharp weighted estimates for smooth truncations to the case of sharp truncations. The latter result was generalized to matrix weight by Duong et al. [14].
However, the (sharp) quantitative weighted version of Theorem A, that is for operators without regularity, has not been considered up to now except the special case of \(q=\infty \) due to Di Plinio et al. [11] and Lerner [30]. One main goal of the present paper is to address this question, and to provide the quantitative weighted variational inequality for \({\mathcal {T}}_\Omega \).
The arguments in [25, 39] show that strong q-variational estimates with respect to the martingale sequence in \(E_{ N}f(x)=\frac{1}{|Q|}\int _Q f\) (here Q is the unique dyadic interval of sidelength \(2^N\) containing x) fail whenever \(q < 2\) and the strong q-variational estimates of \({\mathcal {T}}_\Omega \) depends on \(\{E_{ N}\}\). Thus, throughout the paper we will only consider the case \(q > 2\) for strong q-variations.
Theorem 1.1
Let \(\Omega \in L^\infty ({\mathbb S}^{n-1})\) satisfying (1.4) and \({\mathcal {T}}_\Omega \) be defined as above. Then the variational inequality
holds for \(q>2\), \(1<p<\infty \) and \(w\in A_p\), where the implicit constant \(c_{p,q,n}\) is independent of f and w, and \((w)_{A_p}:=\max \{[w]_{A_\infty },\,[w^{1-p'}]_{A_\infty }\}\), \(\{w\}_{A_p}:=[w]_{A_p}^{1\over p}\max \{[w]_{A_\infty }^{1\over p'},\,[w^{1-p'}]_{A_\infty }^{1\over p}\}\), and the constant \([w]_{A_p}\), \(1<p\le \infty \), is given in (2.1) and (2.2).
Remark 1.2
Our quantitative estimate in (1.7) for \(q=\infty \) matches the best known results in [11, 30] for \(T_\Omega ^*\).
1.2 Approach and Main Methods
We now state the methods and techniques for proving our main result. Some estimates are of independent interest, and hence we will single them out as theorems or propositions. As usual, we shall prove Theorem 1.1 by verifying separately the corresponding inequalities for the long and short variations as follows (see e.g., [24, Lemma 1.3]).
Theorem 1.3
Let \(q>2.\) Let \(\Omega \in L^\infty ({\mathbb S}^{n-1})\) satisfying (1.4). Then for \(1<p<\infty \) and \(w\in A_p,\)
Theorem 1.4
Let \(q\ge 2.\) Let \(\Omega \in L^\infty ({\mathbb S}^{n-1})\) satisfying (1.4). Then for \(1<p<\infty \) and \(w\in A_p,\)
where
Remark 1.5
It is worthy to point out the above two results are of independent interest due to the presence of different quantitative behaviors. However for \(q<2\), We don’t know whether or not the \(L^p\) boundedness of the short variations \(\mathcal {S}_q({\mathcal {T}}_\Omega f)\) holds.
The proof of Theorem 1.3 is given in Sect. 6 and the proof of Theorem 1.4 in Sect. 7, while the important tools for the proof of these two theorems are established in Sects. 3, 4 and 5.
1.2.1 Methods for Proving Theorem 1.3
As a standard way to study variational inequalities for rough singular integrals (see e.g., [7, 10] originating from [13]), the first step is to exploit the Cotlar type decomposition of \(T_{\Omega ,2^k}\). Let \(\phi \in C^{\infty }_{0}(\mathbb {R}^n)\) be a non-negative radial function, supported in \(\{|x|\le 1/4\}\) with \(\int \phi (x)dx=1\), and denote \(\phi _{k}(x):= 2^{-kn}\phi (2^{-k}x)\). Let \(K_\Omega (x):={\Omega (x/|x|)}{|x|^{-n}}\), and \(\delta _0\) be the Dirac measure at 0. Then for a reasonable function f, we decompose \(T_{\Omega ,2^k}f\) as
which divides \(\mathcal {T}_\Omega =\{T_{\Omega , 2^k}\}_{k\in \mathbb Z}\) into three families:
Thus, it suffices to verify the weighted \(L^p\) estimate for the variation of \(\mathcal {T}_\Omega ^i(f)\), \(i=1,2,3\).
The second usual step to deal with variational estimates for rough singular integrals is to exploit the almost orthogonality principle based on Littlewood–Paley decomposition and interpolation. However, this step is not obvious at all in the present setting for quantitative weighted estimates. Indeed, the standard Littlewood–Paley decomposition seems to be insufficient. We will exploit the one associated to a sequence of natural numbers \({\mathcal {N}}=\{N(j)\}_{j\ge 0}\) used in [21]; moreover, it is quite involved to gain the sharp kernel estimates involving the smoothing version of \(T_\Omega \) which are necessary to obtain the quantitative weighted estimates. Now let us comment on the proof term by term in details.
Comments on the estimate of \(\mathcal {T}_\Omega ^1(f)\). Let us first formulate below the desired estimate of this term which is of independent interest since this variational estimate strengthens the result of Lerner [30]. Denote by \(\Phi (g)=\{\phi _j*g\}_{j\in \mathbb Z}.\)
Theorem 1.6
Let \(q>2\). Let \(\phi \) be as used in (1.10) and \( T_\Omega \) be given as in (1.3) with \(\Omega \in L^\infty (\mathbb S^{n-1})\) satisfying (1.4). Then for \(1<p<\infty \) and \(w\in A_p,\)
The proof of Theorem 1.6 is given in Section 5.
To obtain (1.11), we apply the Littlewood–Paley decomposition of \(T_\Omega \) in [21] to get \( T_{\Omega }=\sum _{j=0}^{\infty } {T}_{j}^{{\mathcal {N}}}, \) where each \( {T}_{j}^{\mathcal {N}}\) is a \(\omega _j\)-Dini Calderón–Zygmund operator of convolution type. This yields
Then to control the summation of the right-hand side of the above inequality, it suffices to show for each term the unweighed estimate with rapid decay (with respect to j) and the weighted norm \(L^p\) estimate for all \(1<p<\infty \) and then to apply the Stein–Weiss interpolation—Lemma 2.2. The rapid decay estimate will follow from the \(L^2\) estimate of \(T^{\mathcal {N}}_j\); the involved part lies in the estimates in terms of the weighted constant and the Dini norm. The latter is not only necessary for further application but also is of independent interest, we formulate it in the following theorem.
Theorem 1.7
Let \(q>2\). Let T be a \(\omega \)-Dini Calderón–Zygmund operator of convolution type with the kernel K satisfying the following cancellation condition: for any \(0<\varepsilon ,\,R<\infty ,\)
Then for any \(1<p<\infty \) and \(w\in A_p,\)
where \(\omega \) satisfies the Dini condition (2.3) and the estimate is sharp in the sense that the inequality does not hold if we replace the exponent of the \(A_{p}\) constant by a smaller one.
The proof of Theorem 1.7 is given in Section 5.
Firstly, it would be impossible to get the weighted bound in (1.13) via iteration by showing the weighted estimate of \(V_q\Phi \) since the one of T is already linear (see. e.g., [21, 29]).
We provide the proof of (1.13) based on the Cotlar type decomposition
where \(K_k=K\chi _{|\cdot |\le 2^{k}}\) and \(K^k=K\chi _{|\cdot |\ge 2^{k}}.\) Now, the sharp weighted bound of \(V_q(\{K^k*f\}_{j\in \mathbb Z})\) has been obtained in [9] (see also [6, 14]) with bounds \(\Vert {\omega }\Vert _{Dini}+C_K+\Vert T\Vert _{L^2\rightarrow L^2}\). Regarding the second term, at a first glance, it should be easier to handle than the second term in (1.10) since the kernel K enjoys some regularity; but it turns out that they are in the same level of complexity as they will follow from Proposition 1.8 below. Finally, the third term in (1.14) is hard to deal with which will follow from Proposition 1.9.
We first control the variation norm by a stronger norm. That is,
Then all the desired estimates for the left quantity involving \(V_q\) will follow from that involving \(L_q\).
We establish the following result.
Proposition 1.8
Let K be a kernel satisfying the mean value zero property (1.12) and the size condition (2.5). Then the following assertions hold.
- (1):
-
Weak type (1, 1) boundedness:
$$\begin{aligned} \big \Vert \big (\sum _{k\in \mathbb Z}|{\phi }_k*K_k*f|^2\big )^{1/2}\big \Vert _{L^{1,\infty }}\lesssim C_K\Vert f\Vert _{L^1}. \end{aligned}$$(1.15)Weak type (1, 1) boundedness of the grand maximal truncated operator (see (2.7)):
$$\begin{aligned} \Vert {\mathcal {M}}_{(\sum _{k\in \mathbb Z}|{\phi }_k*K_k*f|^2)^{1/2}}\Vert _{L^{1,\infty }}\lesssim C_K\Vert f\Vert _{L^1}. \end{aligned}$$(1.16) - (2):
-
Sharp weighted norm inequalities: for all \(1<p<\infty ,\) and \(w\in A_p,\)
$$\begin{aligned} \bigg \Vert \bigg (\sum _{k\in \mathbb Z}|\phi _k*K_k*f|^2\bigg )^{1/2}\bigg \Vert _{L^{p}(w)}&\lesssim C_K \{w\}_{A_p}\Vert f\Vert _{L^{p}(w)}. \end{aligned}$$(1.17)
The proof of Proposition 1.8 is given in Section 3.
Surely, by the fact that \(\ell ^2\subset \ell ^q\) for \(q\ge 2\), the desired estimate for the second term in the decomposition (1.14) can be deduced from the Proposition 1.8. Moreover, it is worth highlighting that in Proposition 1.8, we do not need any regularity assumption on the kernel K.
Again, we start with the comments on the estimate of the third term in the decomposition (1.14). We prove the following result.
Proposition 1.9
Let \(q>2.\) Let K be a kernel satisfying the mean value zero property (1.12) and the size condition (2.5) and the Dini condition (2.6). Let \( C(K,\omega ,q)= C_K+\Vert \omega \Vert _{Dini}+\Vert \omega \Vert _{Dini}^{1/q}\Vert \omega ^{1/q'}\Vert _{Dini}+\Vert \omega \Vert _{Dini}^{1/q'}\Vert \omega ^{1/2}\Vert _{Dini}^{2/q},\) where \(\omega \) satisfies the Dini condition (2.3). Then the following assertions hold.
- (1):
-
Weak type (1, 1) boundedness:
$$\begin{aligned} \big \Vert \big (\sum _{k\in \mathbb Z}|(\phi _{k}-\delta _0)*K^k*f|^q\big )^{1/q}\big \Vert _{L^{1,\infty }}&\lesssim C(K,\omega ,q)\Vert f\Vert _{L^{1}}. \end{aligned}$$(1.18)Weak type (1, 1) boundedness of the grand maximal truncated operator:
$$\begin{aligned} \Vert \mathcal {M}_{(\sum _{k\in \mathbb Z}|(\phi _{k}-\delta _0)*K^k*f|^q)^{1/q}}\Vert _{L^{1,\infty }}&\lesssim C(K,\omega ,q)\Vert f\Vert _{L^{1}}. \end{aligned}$$(1.19) - (2):
-
Sharp weighted norm inequalities: for all \(1<p<\infty ,\) and \(w\in A_p,\)
$$\begin{aligned} \big \Vert \big (\sum _{k\in \mathbb Z}|(\phi _k-\delta _0)*K^k*f|^q\big )^{1/q}\big \Vert _{L^p(w)}&\lesssim C(K,\omega ,q)\{w\}_{A_p}\Vert f\Vert _{L^p(w)}. \end{aligned}$$(1.20)
This proposition will be shown in Sect. 4.
We point out that at the moment we are unable to prove the above result for \(q=2\) since our argument in showing the weak type (1, 1) estimate of the localized maximal operator (1.19) depends heavily on the same estimate of q-variation operator associated to the Hardy–Littlewood averages.
In this process, together with the weak type (1, 1) estimates of variational Calderón–Zygmund operators [14], we also obtain the following result as a byproduct.
Corollary 1.10
Let \(q>2\) and T be as in Theorem 1.7. Then
where \(\omega \) satisfies the Dini condition (2.3).
Comments on the estimate of \(\mathcal {T}_\Omega ^2(f)\). About \(\mathcal {T}_\Omega ^2(f),\) we start with the comments on the estimate of the second term in the decomposition (1.14) since the latter involving regular kernel K seems to be easier to handle. In particular, Proposition 1.8 applies well to \(\mathcal {T}_\Omega ^2(f)\), and thus the almost orthogonality technique can be avoided, improving the corresponding argument in [7].
Comments on the estimate of \(\mathcal {T}_\Omega ^3(f)\). For the rough kernel \(K_\Omega \) in \(\mathcal {T}_\Omega ^3(f)\), we have to decompose it by using the Littlewood–Paley decomposition from [21], repeat the argument of Proposition 1.9, require a slightly more involved proof and then exploit the almost orthogonality principle.
1.2.2 Methods for proving Theorem 1.4
Following from the sharp weighted boundedness of the Hardy–Littlewood maximal operator (see (7.1)),
Thus by interpolation, it suffices to prove that
To verify this, we write
where \(T_{k,t}f(x)=v_{k,t}*f(x)\) for \(t\in [1,2]\), and \(v_{0,t}(x)={\Omega (x/|x|)}{|x|^{-n}}\chi _{_{t\le |x|\le 2}}(x)\) and \(v_{k,t}(x)={2^{-kn}}\nu _{0,t}(2^{-k}x)\) for \(k\in \mathbb {Z}\). Next, by using the Littlewood–Paley decomposition as in [21], we further have \(T_{k,t}=\sum _{j=0}^\infty T^\mathcal {N}_{k, t, j}\), whose definition will be given in Section 7. Therefore, by the Minkowski inequality, we get
where
Hence, to prove (1.21), it again suffices to verify for \(\mathcal {S}_{2,j}^\mathcal {N}( f)\) the unweighted norm with rapid decay in j and the weighted norm estimate and then to apply the Stein–Weiss interpolation—Lemma 2.2.
Notation. From now on, \(p'=p/(p-1)\) represents the conjugate number of \(p\in [1,\infty )\); \(X\lesssim Y\) stands for \(X\le C Y\) for a constant \(C>0\) which is independent of the essential variables living on \( X\ \& \ Y\); and \(X\approx Y\) denotes \(X\lesssim Y\lesssim X\).
2 Preliminaries
2.1 Muckenhoupt Weights
Let w be a non-negative locally integrable function defined on \(\mathbb R^n\). For \(1<p<\infty \), we say that \(w\in A_p\) if there exists a constant \(C>0\) such that
We will adopt the following definition for the \(A_\infty \) constant for a weight w introduced by Fujii [15] and later by Wilson [42]:
Here \(w(Q):=\int _Qw(x)\,dx,\) and the supremum above is taken over all cubes with edges parallel to the coordinate axes. When the supremum is finite, we will say that w belongs to the \(A_\infty \) class. \(A_\infty :=\bigcup _{p\ge 1} A_p\). Recall in Theorem 1.1, \((w)_{A_p}:=\max \{[w]_{A_\infty },\,[w^{1-p'}]_{A_\infty }\}.\) Using the facts that \([w^{1-p'}]_{A_p'}^{1/p'}=[w]_{A_p}^{1/p}\) and \([w]_{A_\infty }\le C[w]_{A_p}\)(see [19]), one easily checks that \((w)_{A_p}\le [w]_{A_p}^{\max (1,1/(p-1))}\) (see [21]).
2.2 \(\omega \)-Dini Calderón–Zygmund Operators of Convolution Type
A modulus of continuity \(\omega \) if it is increasing and subadditive in the sense that
\(\omega \) satisfies the classical Dini condition
For any \(c>0\) the integral can be equivalently (up to a c-dependent multiplicative constant) replaced by the sum over \(2^{-j/c}\) with \(j\in \mathbb N.\)
Let T be a bounded linear operator on \(L^2({\mathbb R}^n)\) represented as
We say that T is an \(\omega \)-Dini Calderón–Zygmund operators of convolution type if the kernel K satisfies the following size and smoothness conditions:
for any \(x\in {\mathbb R}^n\backslash \{0\}\),
for any \(x,\,y\in {\mathbb R}^n\) with \(2|y|\le |x|\),
2.3 Criterion for Sharp Weighted Norm Estimate
The grand maximal truncated operator \(\mathcal {M}_U\), associated to a sub-linear operator U, is defined by
where the supremum is taken over all cubes \(Q\subset {\mathbb R}^n\) containing x.
Lemma 2.1
([29]) Assume that both U and \(\mathcal {M}_U\) be of weak type (1, 1), then for every compactly supported \(f\in L^p({\mathbb R}^n)\,(1<p<\infty )\) and \(w\in A_p,\)
where \(K=\Vert U\Vert _{L^1\rightarrow L^{1,\infty }}+\Vert \mathcal {M}_U\Vert _{L^1\rightarrow L^{1,\infty }}.\)
2.4 The Stein–Weiss Interpolation Theorem with Change of Measures
The following interpolation result with change of measures due to Stein and Weiss plays an important role in dealing with rough singular integrals. Again, we will use frequently this tool.
Lemma 2.2
([40]) Assume that \(1\le p_0,p_1\le \infty \), that \(w_0\) and \(w_1\) be positive weights, and T be a sublinear operator satisfying
with quasi-norms \(M_0\) and \(M_1\), respectively. Then
with quasi-norm \(M\le M_0^\lambda M_1^{1-\lambda }\), where
3 Proof of Proposition 1.8
We first give a useful Lemma which plays an important role in the proof of Proposition 1.8.
Lemma 3.1
For any fixed \(k\in \mathbb Z,\) let \(\mu _k\) be a function such that supp \( \mu _k\subset \{x:|x| \lesssim 2^k\}.\) If there exist constants \(\gamma _1,\,\gamma _2,\,\gamma _3\) such that \(\mu _k\) satisfy
and
and for \(0<2|y|\le |x|,\)
Then for \(1<p<\infty ,\) and \(w\in A_p,\)
Proof
To prove (3.4), by Lemma 2.1, it suffices to verify that
and
By virtue of (3.1)–(3.3), we know \(\{\mu _k*f\}\) is a vector Calderón–Zygmund operator. By using a Calderón–Zygmund decomposition and a trivial computation, it is easy to verify that (3.5) holds. Next, we verify (3.6). Take x and \(\xi \) in any fixed cube Q. Let \( B_x=B(x,2\sqrt{n}\ell (Q)),\) then \(3Q\subset B_x.\) By using the triangle inequality, we get
For the term I, since x and \(\xi \) in Q, then \(|x-\xi |\le \sqrt{n}\ell (Q),\) and \(|x-y|\ge 2\sqrt{n}\ell (Q) \), then we get \(2|x-\xi |\le |x-y|\), by (3.3) we have
For the term \(I\!I\), since \(\xi \) in Q and \(y\in B_x{\setminus } 3Q\), then \(|y-\xi |\ge 3\ell (Q)\) and \(|y-\xi |^{-n}\lesssim \ell (Q)^{-n}\) and \( |x-y|\le 2\sqrt{n}\ell (Q) \), then by the size condition (3.2), we get
For the last term, recalling that supp \(\mu _k\subset \{x:|x|\lesssim 2^{k}\}\) and that \(B_{x}=\{y:|y-x|\le 2\sqrt{n} \ell (Q)\}\), we have that \(\mu _k*f\chi _{\mathbb {R}^{n}{\setminus } B_{x}}=\int _{2\sqrt{n} \ell (Q)<|x-y|\lesssim 2^{k}}\mu _k(x-y)f(y)dy.\) Thus,
By using the size estimate (3.2), we have
which gives
Combining the estimates of \(I,\,I\!I\) and \(I\!I\!I,\) we get
Then the weak type (1, 1) of the Hardy–Littlewood maximal function M and the estimate in (3.5) imply
which shows that (3.6) holds. The proof of Lemma 3.1 is complete. \(\square \)
Now we return to the proof of Proposition 1.8. We point out that to estimate \(\{\phi _{k}*K_{k}\},\) we do not need any regularity condition on K, we only assume K satisfies the size condition (2.5) and the cancellation condition (1.12). We first verify the size estimate of \(\phi _k*K_k(x)\) for any fixed \(k\in \mathbb Z.\) Since supp \({\phi }_k\subset \{x: |x|\le 2^k/4\},\) we have supp \(\phi _k*K_k\subset \{x:|x|\le 2^{k+1}\}.\) Then from (2.5) and (1.12) we have for some \(\theta \in (0,1)\)
Now, we estimate \(\nabla \phi _{k}*K_k(x)\). Recall that \(\phi \in C_0^\infty ({\mathbb R}^n)\) with \(\int \phi =1\) and supp \(\phi \subset \{x: |x|\le 1/4\}.\) It is easy to verify that \(\int \nabla \phi =0\) and supp \((\nabla \phi )\subset \{x: |x|\le 1/4\}\) with \(\nabla \phi \in C_0^\infty ({\mathbb R}^n).\) Since supp \((\nabla \phi )_k\subset \{x:|x|\le 2^k/4\}\) and supp \(K_k\subset \{x:|x|\le 2^k\}\) then we get supp \((\nabla \phi )_k*K_k\subset \{x:|x|\le 2^{k+1}\}.\) Then repeating the argument of (3.7), we get
Thus for \(|x|\ge 2|y|,\) and for some \(\theta \in (0,1),\)
Since \(|x|\ge 2|y|,\) then \(x/2\le |x-\theta y|\le 3x/2\), we get \(|\nabla {\phi }_k*K_k(x )|\le \frac{C_K}{2^{k{(n+1)}}}\chi _{|x|\le 2^{k+1}}. \) Thus for \(|x|\ge 2|y|,\) we get
On the other hand, it is easy to verify that
and
for some \(\eta \in (0,1)\). In addition, since \({{\phi }}_k*K_k*1=0,\) by applying the T1 Theorem for square functions (see [8]) we have
Then applying Lemma 3.1 to \(\{\mu _k\}=\{\phi _k*K_k\},\) we get
and then by Lemma 2.1, for \(1<p<\infty \) and \(w\in A_p,\)
This verifies (1.15)–(1.17). The proof of Proposition 1.8 is complete. \(\square \)
4 Proof of Proposition 1.9
Step 1. First, we would like to prove that
We first use Fourier transform and Plancherel theorem to deal with the \(L^2\)-norm of \((\sum _{k\in \mathbb Z}|({{\phi }}_k-\delta _0)*K^k*f|^2)^{1/2}.\) For \(|2^k\xi |<1,\) write
For the term \(I_{1,k}\), by the size estimate (2.5) and cancellation (1.12), we have
uniformly in k.
We now consider \(I_{2,k}(\xi )\). Let \(z=\frac{\xi }{2|\xi |^2}\) so that \(e^{2\pi i z \cdot \xi }=-1\) and \(2|z|=|\xi |^{-1}\). By changing variables \(x=x'-z\), we rewrite \(I_{2,k}(\xi )\) as
Taking the average of the above inequality and the original definition of \(I_{2,k}\) gives
Now we split
For the first term, by noting that \(2|z|=|\xi |^{-1}\) and using (2.6) we get
For the second term, we note that \(\chi _{|\xi |^{-1}<|x|}-\chi _{|\xi |^{-1}<|x-z|}\) is nonzero if and only if \(|\xi |^{-1}-|z|\le |x-z|\le |\xi |^{-1}+|z|\). Thus by (2.5)
Combining the above estimates, we get \( |I_{2,k}(\xi )|\lesssim C_{K}+\Vert \omega \Vert _{Dini}, \) which, together with the estimate of \(|I_{1,k}(\xi )|\), gives
For \(|2^k\xi |>1,\) let \(z=\frac{\xi }{2|\xi |^2}\) so that \(e^{2\pi i z \cdot \xi }=-1\) and \(2|z|=|\xi |^{-1}\). Similarly to \(I_{2,k}(\xi )\), we get
For the term \(J_{1,k}\), by noting that \(2|z|=|\xi |^{-1}\) and using (2.6) we get
For the term \(J_{2,k}\), we note that \(\chi _{2^k<|x|}-\chi _{2^k<|x-z|}\) is nonzero if and only if \(2^k-|z|\le |x-z|\le 2^k+|z|\). Thus by (2.5)
Combining the estimates of \(J_{1,k}(\xi )\) and \(J_{2,k}(\xi ),\) we get
Since \({\widehat{\phi }}\in \mathcal {S}({\mathbb R}^n)\) and \({\widehat{\phi }}(0)=1,\) then \(|1-\widehat{\phi _k}(\xi )|\lesssim \min (|2^k\xi |, 1).\) By the Plancherel Theorem, (4.2) and (4.3), we get
On the other hand, we claim that for any fixed \(k\in \mathbb Z,\)
which will be proved in Step 2. Then we can get
Interpolating between (4.4) and (4.6), we get for \(q\ge 2\)
We get (4.1).
Step 2. We now estimate \(|(\delta _0-\phi _k)*K^k(x)|\) for any fixed \(k\in \mathbb Z.\)
We first consider \(A_{k,1}\). By noting that supp \( {\phi }_k\subset \{x: |x|\le 2^k/4\}\) and that \(|z|>2^{k},\) we have \(|x|\ge \frac{3|z|}{4}> 3\cdot 2^k/4.\) Then by the regularity condition of K in (2.6),
For the term \(A_{k,2}\), we first note that \(\chi _{|z|> 2^{k}}-\chi _{|x|> 2^{k}}\) is nonzero if and only if \(\frac{3}{4}\cdot 2^k\le |x|\le \frac{5}{4}\cdot 2^k\) since \(|x-z|\le 2^k/4\). Thus,
Combining the two cases, we get that for any fixed \(k\in \mathbb Z,\)
which verifies (4.5) and implies that
We now estimate \(\Vert \{({{\phi }}_k-\delta _0)*K^k(x-y)-({{\phi }}_k-\delta _0)*K^k(x)\}_{k}\Vert _{\ell ^q}\) for \(0<|y|\le |x|/2.\) To begin with, we claim that for \(|y|\le \frac{2^k}{2},\)
Assuming (4.10) for the moment, we write
For the term \(I\!I_1\), by noting that \(|x|\ge 2|y|>0\) and by using (4.8) we get
For the term \(I\!I_2\), by using (4.10),
Combining the estimates of \(I\!I_1\) and \(I\!I_2\), we get
Now we return to give the proof of (4.10). We write
For the term \(I\!I\!I_1\), recall that \(|y|\le \frac{2^k}{2}\). By noting the facts that \(0<|y|\le \frac{2^k}{2}\) and \(|x-y|\ge 2^{k}\) imply \(|x|>\frac{2^k}{2}\) and that \(\chi _{|x-y|>2^{k}}-\chi _{|x|>2^{k}}\) is nonzero if and only if \(2^{k}-|y|\le |x|\le 2^{k}+|y|\), we have
Similarly, for the term \(I\!I\!I_2\), we get
Since \(|x-z|\le 2^k/4\) and \(|z|>\frac{2^k}{2}\), we get \( 3|z|/2\ge |x|\ge |z|/2>2^k/4 \). Thus, there exists some \(\eta \in (0,1)\),
Combining the estimates of \(I\!I\!I_1\) and \(I\!I\!I_2\), we get (4.10).
Step 3. By Lemma 2.1, to prove (1.20), we need to verify that
and
The two above inequalities just are (1.18) and (1.19).
To verify (4.12), we apply the Calderón-Zygmund decomposition to f at height \(\alpha \) to obtain that there is a disjoint family of dyadic cubes \(\{Q\}\) with total measure \(\sum _Q|Q| \lesssim {\alpha }^{-1}\Vert f\Vert _{L^1},\) and that \(f=g+h,\) with \(\Vert g\Vert _{L^\infty }\lesssim \alpha ,\) and \(\Vert g\Vert _{L^1}\lesssim \Vert f\Vert _{L^1}\), \(h=\sum _{Q}h_Q,\) where supp\((h_Q)\subseteq Q,\) \(\int _{\mathbb R^n} h_Q(x)\,dx=0\) and \(\sum \Vert h_Q\Vert _{L^1}\lesssim \Vert f\Vert _{L^1}.\)
By Chebychev’s inequality and (4.1), we get
Then it suffices to estimate \((\sum _{k\in \mathbb Z}|({{\phi }}_k-\delta _0)*K^k*h(x)|^q)^{1/q}\) away from \(\bigcup {\widetilde{Q}}\) with \({\widetilde{Q}}=2Q.\) Let \(y_Q\) denote the center of Q. Since \( \int _{\mathbb R^n} h_Q(x)\,dx=0\), then we can write
Now by Chebychev’s inequality and (4.11),
If \(|y-y_Q|\le 2^k/2,\) then \(|x-y_Q|\ge 2^k-|y-y_Q|\ge 2^k/2,\) thereby \(|x-y_Q|^{-1}\le 2\cdot 2^{-k},\)
This finishes the proof of (4.12).
Now, we verify (4.13). Let Q be a cube, and take \(x,\,\xi \in Q.\) Let \(B(x)=B(x,2\sqrt{n}\ell (Q)),\) then \(3Q\subset B_x.\) By the triangle inequality, we get
For the term I, by using the triangular inequality,
By using the size estimate (4.8) and the fact that \(\frac{1}{2}|x-y|\le |\xi -y|\le \frac{3}{2}|x-y|\) (since \(2|x-\xi |\le |x-y|\)), we get
For the term \(I_2\), by (4.10), we get
To estimate \({\widetilde{I}}_2\), it suffices to consider the form
where \(|s_{k+1}-s_{k}| \le 2|x-\xi |,\) and \(\frac{2^{k}}{2}\le s_k\le \frac{3}{2}\cdot 2^{k}.\) Using the hypothesis that \(|x-\xi |<\ell (Q)\) and the kernel estimate we can bound the above by a dimensional constant times
The above \(\ell ^{q}\) norm can be written as
where \(V_{q}\mathcal {A} (|f|)(x)=V_q\{A_{t}(|f|)(x)\}_{t>0}, \) and
Here \(B_{t}\) denotes the open ball in \(\mathbb {R}^{n}\) of center at the origin and radius t. Also note that
Thus we get
Thus combining the estimates of \(I_1\) and \(I_2,\) we get
For the term II, by using (4.9) and the fact that \(|x-y|\simeq |\xi -y|\) (since \(2|x-\xi |\le |x-y|\)), we get
For the term III, since supp \((\phi _{k}-\delta _{0})*K^{k}\subset \{x:|x|\ge \frac{3}{4}\cdot 2^k\},\) we obtain that
By using the size estimate (4.8) we get
which implies
Thus, we obtain that
Combining the estimates of \(I,\,II\) and III, we get
Then by the weak type (1, 1) of M (see [41]) and \({V}_{q}\mathcal {A}(|f|)\)(see [23]), and (4.12), we get
which gives the proof of (4.13). The proof of Proposition 1.9 is complete.\(\square \)
5 Proofs of Theorems 1.6 and 1.7
Proof of Theorem 1.7
Recalling that for any fixed \(k\in \mathbb Z,\) we denote by
Then for any fixed \(k\in \mathbb Z,\) write
where \(\delta _0\) is the Dirac measure at 0. Thus by the triangle inequality, we get
Since the following results have been established in [9, Theorem 1.5], [14, Prosition 2.1] and [6, Theorem 1.3],
and for \(1<p<\infty ,\) \(w\in A_p,\)
Then combining Propositions 1.8 and 1.9, we finish the proof of Theorem 1.7. \(\square \)
Proof of Theorem 1.6
Recall the definition of the operator \(T_{\Omega }\) given in the introduction. It can be written as
We consider the following partition of unity. Let \(\varphi \in C_{c}^{\infty }(\mathbb {R}^{n})\) be such that supp \(\varphi \subset \{x:|x| \le \frac{1}{100}\}\) and \(\int \varphi d x=1,\) and so that \({\widehat{\varphi }} \in \mathcal {S}(\mathbb {R}^{n}).\) Let us also define \(\psi \) by \({\widehat{\psi }}(\xi )={\widehat{\varphi }}(\xi )-{\widehat{\varphi }}(2 \xi ).\) Then, with this choice of \(\psi ,\) it follows that \(\int \psi d x=0\). We write \(\varphi _{j}(x)=\frac{1}{2^{j n}} \varphi (\frac{x}{2^{j}}),\) and \(\psi _{j}(x)=\frac{1}{2^{j n}} \psi (\frac{x}{2^{j}})\). We now define the partial sum operators \(S_{j}\) by \(S_{j}(f)=f*\varphi _{j}\). Their differences are given by
Since \(S_{j}f \rightarrow f\) as \(j \rightarrow -\infty \), for any sequence of integers \(\mathcal {N}=\{N(j)\}_{j=0}^{\infty }\), with \(0 = N(0)< N(1)<\cdot \cdot \cdot < N(j)\rightarrow \infty \), we have the identity
In this way,
where
and, for \(j\ge 1\),
Therefore,
To prove Theorem 1.6, we claim that the following inequalities hold for \(1<p<\infty \) and \(w\in A_p\):
and
Assume the above two claims at the moment. Set \(\varepsilon :=\frac{1}{2}c_n/(w)_{A_p},\) \(c_n\) is small enough (see [21, Corollary 3.18]). By (5.10), we have, for this choice of \(\varepsilon \),
Now we are in position to apply the interpolation theorem with change of measures by Stein and Weiss—Lemma 2.2. We apply it to \(T=V_q\{\phi _{l}*{T}_{j}^{\mathcal {N}}\}_{l\in \mathbb Z}\) with \(p_0=p_1=p,~~w_0=w^0=1\) and \(w_1=w^{1+\varepsilon }\), so that by \(\lambda =\varepsilon /(1+\varepsilon )\), (5.9) and (5.11), one has for some \(\theta ,\,\gamma >0 \) such that
Thus
We now choose \(N(j)=2^j\) for \(j\ge 1\). Then, using \(e^{-x}\le 2x^{-2}\), we have
by summing the two geometric series in the last step. This implies
and hence the proof of Theorem 1.6 is complete under the assumptions that (5.9) and (5.10) hold.
Now we turn to proving the claims (5.9) and (5.10). The following inequality can be found in [21, Lemma 3.7],
for some \(0< \alpha < 1\) independent of \(T_{\Omega }\) and j. Then by (see [10, 24])
we get
The operator \({T}_{j}^{\mathcal {N}} \) is a \(\omega \)-Calderón–Zygmund operator with the kernel \(K_j^\mathcal {N}\) (see [21, Lemma 3.10]) satisfying
and for \(2|y|\le |x|,\)
where \(\omega _{j}(t)\le \Vert \Omega \Vert _{L^ \infty } \min (1,2^{N(j)}t), \) and \(\Vert \omega _j\Vert _{Dini}\lesssim \Vert \Omega \Vert _{L^{\infty }}(1+N(j)). \) For \(r>1,\) the Dini norm of \(\omega _{j}^{1/r}\) is estimated as
Applying Theorem 1.7 to \(T={T}_{j}^{\mathcal {N}}\), we get for \(1<p<\infty \) and \(w\in A_p\)
which gives the proof of (5.10). Taking \(w=1,\) we get for \(1<p<\infty ,\)
Interpolating between (5.14) and (5.18), we get
which establishes the proof of (5.9). The proof of Theorem 1.6 is complete. \(\square \)
6 Proof of Theorem 1.3
Recalling that for any fixed \(k\in \mathbb Z,\)
and \(K_\Omega (x)=\frac{\Omega (x')}{|x|^n}\). Following [13], for any fixed \(k\in \mathbb Z,\) we write
which splits \(\mathcal {T}_\Omega =\{T_{\Omega , 2^k}\}_{k\in \mathbb Z}\) into three families:
Thus, it suffices to estimate the weighted \(L^p\) norm of \(\mathcal {T}_\Omega ^i(f)\), \(i=1,2,3\).
Part 1 Let us first consider \(\mathcal {T}_\Omega ^1(f)\). By Theorem 1.6, we get that for \(1<p<\infty \) and \(w\in A_p,\)
Part 2 For the term \(\mathcal {T}_\Omega ^2(f)\), by the Minkowski inequality, we get for \(1<p<\infty \) and \(w\in A_p,\)
It is easy to verify that \(K_\Omega \) satisfies the cancellation condition (1.12) and the size estimate \(|K_\Omega (x)|\le \frac{\Vert \Omega \Vert _{L^\infty }}{|x|^n}.\) Then applying Proposition 1.8 to \(K=K_\Omega \), we get
by \(\ell ^2\subset \ell ^q\) for \(q>2.\)
Part 3 We now estimate \(\mathcal {T}_\Omega ^3(f).\) We claim that for \(1<p<\infty \) and \(w\in A_p,\)
which, together with the estimates of \(\mathcal {T}_\Omega ^1(f)\), \(\mathcal {T}_\Omega ^2(f)\) and \(\mathcal {T}_\Omega ^3(f),\) gives
This finishes the proof of Theorem 1.3.
We now provide the proof of (6.1). Recall that \(\mathcal {T}_\Omega ^3(f):=\{(\delta _0-\phi _{k})*T_{\Omega ,2^k}f\}_{k\in \mathbb Z}\). For \(k\in \mathbb Z\), we define \(v_{k}\) as \(v_{k}(x)=\frac{\Omega (x')}{|x|^n}\chi _{2^k<|x|\le 2^{k+1}}(x).\) Then we can write \(T_{\Omega , 2^k}f(x)=\sum _{s\ge 0}v_{k+s}*f(x).\) Recall that \(S_{j}(f)=f*\varphi _{j}\) and their differences are given by \( S_{j}(f)-S_{j+1}(f)=f*\psi _{j}, \) where \(\varphi _{j}\) and \(\psi _{j}\) are defined as in Section 5. Denote by \(T_{k}f(x)=v_k*f(x).\) Similarly, for any sequence of integers \(\mathcal {N}=\{N(j)\}_{j=0}^{\infty }\), with \(0 = N(0)< N(1)<\cdot \cdot \cdot < N(j)\rightarrow \infty \), we also have the identity.
where \(T_{k+s,0}^\mathcal {N}=T_{k+s}S_{k+s}\) and for \(j\ge 1\),
Then
So
We first estimate the \(L^2-\)norm of \((\sum _{k\in \mathbb Z}|(\delta _0-\phi _k)*\sum _{s\ge 0} T^\mathcal {N}_{k+s,j}f|^q)^{1/q}\).
Lemma 6.1
We have for \(j\ge 0\), there is a positive \(0<\tau <1\) such that
Proof
By the Plancherel Theorem and (6.3), we get for \(j\ge 1,\)
Since \({\widehat{\phi }}\in \mathcal {S}({\mathbb R}^n)\) and \({\widehat{\phi }}(0)=1,\) \(|1-{\widehat{\phi }}(\xi )|\lesssim \min (1, |\xi |).\) Also by \({\widehat{\psi }}\in \mathcal {S}({\mathbb R^n})\) and \({\widehat{\psi }}(0)=0,\) then \(|{\widehat{\psi }}(\xi )|\lesssim \min (1, |\xi |).\) Therefore, \(|1-\widehat{\phi _k}(\xi )|\lesssim \min (1,|2^k\xi |^\alpha )\), \(|{\widehat{\psi }}(2^{k+s-i}\xi )|\lesssim \min (|2^{k+s-i}\xi |^\gamma ,1),\,|{\widehat{v}}_{k+s}(\xi )|\lesssim \Vert \Omega \Vert _{L^\infty }|2^{k+s}\xi |^{-\beta }\) (see [13]) for some \(0<\beta ,\,\gamma ,\,\alpha <1\), \(\alpha +\gamma -\beta >0,\) \(\gamma <\beta \) and \(\alpha <\beta ,\)
Therefore, by summing the geometric series and the Plancherel theorem, we get for some \(\tau \in (0,1),\)
For \(j=0\), \(|1-{\widehat{\phi }}(2^{k}\xi )|\lesssim \min (|2^{k}\xi |,1),\,|{\widehat{\varphi }}(2^{k}\xi )|\lesssim 1,\,|{\widehat{v}}_{k+s}(\xi )|\lesssim \Vert \Omega \Vert _{L^\infty }|2^{k+s}\xi |^{-\beta }\) for some \(0<\beta <1\) (see [13]). Therefore, by the Plancherel theorem, we get
The proof of Lemma 6.1 is complete by \(\ell ^2\subset \ell ^q\) for \(q>2\). \(\square \)
Denote \(T_{k,j}^\mathcal {N}:=\sum _{s\ge 0} T^\mathcal {N}_{k+s,j}\), and let \(K_{k,j}^\mathcal {N}\) be the kernel of \(T_{k,j}^\mathcal {N}\) given by
for \(j\ge 1\) and for \(j=0,\)
In the following, we will verify that \(K_{k,j}^\mathcal {N}\) is a \(\omega \)-Dini Calderón–Zygmund kernel satisfying (2.5) and (2.6).
Lemma 6.2
For \(j\ge 0\) and \(k\in \mathbb Z\). Then we have the size estimate
and the regularity estimate
where \(\omega _j(t)\le \Vert \Omega _j\Vert _{L^{\infty }}\min (1, 2^{N(j)}t).\)
Proof
In order to get the required estimates for the kernel \(K_{k,j}^\mathcal {N}\), we first study the kernel of each \(H_{k, s, N(j)}\) which is defined by
First, we estimate \(|H_{k, s, N(j)}(x)|.\) Since supp \(\varphi \subset \{x: |x|\le \frac{1}{100}\}\), a simple computation gives that
From the triangular inequality, and \(N(j-1)<N(j)\), we obtain that the kernel \(K_{k,j}^\mathcal {N}:=\sum _{s\ge 0}(H_{k,s,N(j)}-H_{k,s,N(j-1)})\) satisfies
On the other hand, we compute the gradient. Again by the support of \(\varphi \), we have
Since \(|x-y|\le \frac{2^{k+s}}{100}, 2^{k+s}< |y|\le 2^{k+s+1}\), then \(|x|\simeq |y|\) and \(\frac{3}{4}\cdot 2^{k+s}\le |x|\le \frac{5}{4}\cdot 2^{k+s}.\) Thus, we get
Therefore, if \(|h|<\frac{|x|}{2},\) we get \(|x-\theta h|\simeq |x|\) and
If \(|h|<\frac{|x|}{2},\) combining (6.11) and (6.13), we get for \(j\ge 0\) and \(k\in \mathbb Z,\)
where \(\omega _j(t)\le \Vert \Omega \Vert _{L^{\infty }}\min (1, 2^{N(j)}t).\) The proof of Lemma 6.2 is complete. \(\square \)
We now continue the proof of (6.1). By noting that supp \(K_{k,j}^\mathcal {N}\subset \{x: |x|\ge \frac{3}{4}\cdot 2^k\}\) and that \(K_{k,j}^\mathcal {N}\) satisfies mean value zero, (6.8) and (6.9), repeating the argument of Proposition 1.9 only with \(K^k\) replaced by \(K_{k,j}^\mathcal {N},\) we can also get for \(1<p<\infty \) and \(w\in A_p\)
Taking \(w=1\) in (6.14) we get for \(1<p<\infty ,\)
Interpolating between (6.5) and (6.15), we get
Similar to the proof of Theorem 1.6, based on (6.14) and (6.16), applying the interpolation theorem with change of measures (Lemma 2.2), we get
which gives (6.1). The proof of Theorem 1.3 is complete.\(\square \)
7 Proof of Theorem 1.4
Recall that we can write
where
Observe that
Then by the sharp weighted boundedness of the Hardy–Littlewood maximal operator M (see Hytönen–Pérez [19, Corollary 1.10], the original version was due to Buckley [3]),
we get
Now we claim that
In fact, interpolating between (7.1) and (7.2), we get for \(q\ge 2,\)
Now we turn to verify (7.2). Recall that \(S_{j}(f)=f*\varphi _{j}\) and their differences are given by
where \(\varphi _{j}\) and \(\psi _{j}\) are defined as in Section 5. Similarly, for any sequence of integers \(\mathcal {N}=\{N(j)\}_{j=0}^{\infty }\), with \(0 = N(0)< N(1)<\cdot \cdot \cdot < N(j)\rightarrow \infty \), we also have the identity
where \(T_{k,t,0}:=T_{k,t,0}^\mathcal {N}:=T_{k,t}S_k\) and for \(j\ge 1\),
Therefore, by the Minkowski inequality, we get
Then for \(1<p<\infty \) and \(w\in A_p,\)
Part 1. We first give the \(L^2\)-norm of \(\mathcal {S}_{2,j}^\mathcal {N}( f)(x)\).
Lemma 7.1
There is a positive \(0<\tau <1\) such that
Proof
For \(t\in [1,2]\), we define \(v_{k,t}\) as \(v_{k,t}(x)=\frac{\Omega (x')}{|x|^n}\chi _{_{2^kt<|x|\le 2^{k+1}}}(x).\) Then \(T_{k,t}\) can be expressed as \(T_{k,t}f(x)=v_{k,t}*f(x)\), \(t\in [1,2]\). Recall that
Note that
where \(a'=\{\frac{d}{dt}a_{t}:t\in \mathbb {R}\}\) (see [24]). We have applied (7.9) to \(a=\{T_{k,t,j}^\mathcal {N}f(x)\}_{t\in [1,2]},\) then we have
Then by the above inequalities, we get
This, along with the Cauchy–Schwartz inequality, yields
By the Plancherel theorem and (7.6), we get for \(j\ge 1,\)
Since \(|{\widehat{\psi }}(2^{k-i}\xi )|\lesssim \min (|2^{k-i}\xi |,1),\,|{\widehat{v}}_{k,t}(\xi )|\lesssim \Vert \Omega \Vert _{L^\infty }|2^{k}\xi |^{-\beta }\) for some \(0<\beta <1\) (see [13]),
Then by summing the geometric series and the Plancherel theorem, we get
For \(j=0\), \(|{\widehat{v}}_{k,t}(\xi )|\lesssim \Vert \Omega \Vert _{L^\infty }\min (|2^k\xi |,\,|2^{k}\xi |^{-\beta })\) for some \(0<\beta <1\) and \(|{\widehat{\varphi }}(2^{k}\xi )|\lesssim 1\),
Then by the Plancherel theorem, we get
Combining the above estimates for \(j>0\) and \(j=0,\) we get
On the other hand, we recall the elementary fact: for any Schwartz function h,
Then by \(\int _{\mathbb S^{n-1}}\Omega (y')d\sigma (y')=0\) and \(t\in [1,2]\), we get
Therefore, by the Plancherel theorem, we get for \(j\ge 1\)
By (7.14) and \(|{\widehat{\psi }}(2^{k-i}\xi )|\lesssim \min (|2^{k-i}\xi |, \,|2^{k-i}\xi |^{-\beta })\) for some \(0<\beta <1,\)
Then by the Plancherel theorem, we get for \(j\ge 1\)
For \(j=0,\)
Then by (7.14), we have
By \(|{\widehat{\varphi }}(2^{k}\xi )|\lesssim \min (1, \,|2^{k}\xi |^{-\beta }),\) then
Then by the above inequalities, we get
Then by the Plancherel theorem and the above inequality, we get
Combining the case of \(j>0\) and \(j=0,\) we get
This along with (7.10) and (7.13), we get for \(0<\beta <1,\)
which proves Lemma 7.1. \(\square \)
Part 2. Next, we give the \(L^p(w)\)-norm of \(\mathcal {S}_{2,j}^N( f)(x).\) For \(j\ge 1,\) we denote by \(S_{k,j}:=S_{k-N(j-1)}-S_{k-N(j)}\). For \(j=0,\) we denote by \(S_{k,0}:=S_{k}.\) We have the following observation:
where the operator \({{T}}_{k,t_l,t_{l+1}}\) is given by
Denote by
the kernel of \({T}_{k,t_l,t_{l+1}}S_{k,j }\) for \(j\ge 1\) and by
the kernel of \({T}_{k,t_l,t_{l+1}}S_{k,0 }\) for \(j=0.\) Then
In the following, we give the kernel estimates.
Lemma 7.2
For every \(x \in \mathbb {R}^{n}\backslash \{0\}\), \(j\ge 0\) and \(k\in \mathbb Z,\)
If \(0<|y|\le \frac{|x|}{2},\)
where \(\omega _j(t)\le \Vert \Omega \Vert _{L^\infty }\min (1, 2^{N(j)}t).\)
Proof
Let \(x \in \mathbb {R}^{n}\backslash \{0\}\). Since supp \(\varphi \subset \{x:|x| \le \frac{1}{100}\},\) we get that
Then we get
On the other hand, we compute the gradient. Again by taking into account the support of \(\varphi \), we obtain that
Therefore, by the gradient estimate, for \(|y| \le \frac{1}{2}|x|\) we have
Then using (7.23) and (7.26), we get
where \(\omega _j(t)\le \Vert \Omega \Vert _{L^\infty }\min (1,2^{N(j)}t)\) (for \(j=0,\) the subtraction is not even needed). From the triangle inequality, and \(N(j-1)<N(j)\) it follows that the kernel \(v_{k,t_l,t_{l+1}}*(\varphi _{k-N(j)}-\varphi _{k-N(j-1)})\) satisfies the same estimates (7.24) and (7.27) which proves Lemma 7.2. \(\square \)
Lemma 7.3
Let \(\mathcal {S}_{2,j}^\mathcal {N}\) be defined as in (7.7). Then we get
Proof. We perform the Calderón-Zygmund decomposition of f at height \(\alpha \), thereby producing a disjoint family of dyadic cubes \(\{Q\}\) with total measure \(\sum _Q|Q| \lesssim {\alpha }^{-1}\Vert f\Vert _{L^1}\) and allowing us to write \(f=g+h\) as in the proof of Theorem 1.7. From Lemma 7.1,
It suffices to estimate \(\mathcal {S}_{2,j}^\mathcal {N}( h)\) away from \(\bigcup {\widetilde{Q}}\) where \({\widetilde{Q}}\) is a large fixed dilate of Q. Denote
Then from the definition of \(\mathcal {S}_{2,j}^\mathcal {N}\) in (7.20) we get \( \mathcal {S}_{2,j}^\mathcal {N}(h)(x)\le \Big (\sum _{k\in \mathbb Z}|\mathcal {S}_{2,j,k}^\mathcal {N} h(x)|^2\Big )^{1/2}. \) Since
and
then we get
By the above inequalities we get
For the term I, by Chebyshev’s inequality we have
Denote by \(y_Q\) the center of Q. For \(x\notin {\widetilde{Q}}\) and \(y\in Q,\) we get \(2|y-y_Q|\le |x-y_Q|,\) then by \(|y-y_Q|\le \ell (Q)\) and using the vanishing mean value of \(h_Q\) and (7.22),
Hence
where the last inequality follows from \(\Vert \omega _j\Vert _{Dini}\le \Vert \Omega \Vert _{L^\infty }(1+N(j)).\) On the other hand, for the term II, by (7.21), we have
Then by the weak type (1, 1) of M (see [41]),
Combining the estimates of (7.30) and (7.31), we get
The proof of Lemma 7.3 is complete.\(\square \)
Lemma 7.4
For \(j\ge 0,\) let \(S_{2,j}^\mathcal {N}\) be defined as in (7.7). Then we get
where \(\omega _j(t)\le \Vert \Omega \Vert _{L^\infty }\min (1, 2^{N(j)}t).\)
Proof
Let Q be a cube, and take \(x,\,\xi \in Q.\) Let \(B(x)=B(x,2\sqrt{n}\ell (Q)),\) then \(3Q\subset B_x.\) By the triangular inequality, we get
We begin by estimating term I.
For \(2^k\le \ell (Q),\) since \(|x-\xi |\le 2|x-y|,\) we can get \(|x-y|\simeq |\xi -y|.\) Therefore by (7.21)
For \(2^k>\ell (Q)\), by (7.22) we get
Combining the estimates of \(\sum _{2^k\le \ell (Q)}\) and \(\sum _{2^k> \ell (Q)}\), we get
For the term II, by using (7.21) and noting that \(|x-y|\simeq |\xi -y|\) (since \(3|x-\xi |\le |x-y|\)), we have
We now turn to the term III. Note that
where for \(j\ge 1,\)
and for \(j=0,\)
Since supp \(v_{k,t_l,t_{l+1}}*\varphi _{k}\subset \{x: |x|\le 2^{k+1}\}\) and supp \(v_{k,t_l,t_{l+1}}*\varphi _{k-N(j)}\subset \{x: |x|\le 2^{k+1}\},\) from (7.21) we get
Combining the estimates of \(I,\,II\) and III, we get
which leads to
The proof of Lemma 7.4 is complete. \(\square \)
Then by Lemma 7.3, Lemma 7.4 and the weak type (1, 1) of M (see [41]) we get
since \(\Vert \omega _j^{2}\Vert _{Dini}^{1/2}\lesssim \Vert \Omega \Vert _{L^{\infty }}(1+N(j))^{1/2}.\)
Since \(\mathcal {S}_{2,j}^\mathcal {N}\) satisfies (7.28) and (7.33), therefore by applying Lemma 2.1 to \(U=\mathcal {S}_{2,j}^\mathcal {N}\), we get that for \(1<p<\infty \) and \(w\in A_p\),
Taking \(w=1\) in the above inequality gives
Interpolating between (7.8) and (7.35) shows that there exists some \(\theta \in (0,1)\)
Combining (7.34) and (7.36), and by using the interpolation theorem with change of measures-Lemma 2.2, we obtain that
which gives the proof of (7.2). The proof of Theorem 1.4 is complete. \(\square \)
References
Bourgain, J.: Pointwise ergodic theorems for arithmetic sets. Inst. Hautes Études SCI Publ. Math. 69, 5–41 (1989)
Bourgain, J., Mirek, M., Stein, E.M., Wróbel, B.: On dimension-free variational inequalities for averaging operators in \(\mathbb{R} ^d\). Geom. Funct. Anal. 28, 58–99 (2018)
Buckley, S.M.: Estimates for operator norms on weighted spaces and reverse Jensen inequalities. Trans. Am. Math. Soc. 340, 253–272 (1993)
Campbell, J., Jones, R., Reinhold, K., Wierdl, M.: Oscillation and variation for the Hilbert transform. Duke Math. J. 105, 59–83 (2000)
Campbell, J., Jones, R., Reinhold, K., Wierdl, M.: Oscillation and variation for singular integrals in higher dimensions. Trans. Am. Math. Soc. 355, 2115–2137 (2002)
Cheng, W., Ma, T.: Quantitative and sharp weighted estimates for q-variations of singular operators. Acta Math. Sin. (Chin. Ser.) 62, 279–286 (2019)
Chen, Y., Ding, Y., Hong, G., Liu, H.: Weighted jump and variational inequalities for rough operators. J. Funct. Anal. 274, 2446–2475 (2018)
Christ, M., Journé, J.-L.: Polynomial growth estimates for multilinear singular integral operators. Acta Math. 159, 51–80 (1987)
Clara de Franca Silva, F., Zorin-Kranich, P.: Sparse domination of sharp variational truncations. http://arxiv.org/abs/1604.05506
Ding, Y., Hong, G., Liu, H.: Jump and variation inequalities for rough operators. J. Fourier Anal. Appl. 23, 679–711 (2017)
Di Plinio, F., Hytönen, T., Li, K.: Sparse bounds for maximal rough singular integrals via the Fourier transform. Ann. Inst. Fourier 70, 1871–1902 (2020)
Do, Y., Muscalu, C., Thiele, C.: Variational estimates for paraproducts. Rev. Mat. Iberoam. 28, 857–878 (2012)
Duoandikoetxea, J., Rubio de Francia, J.: Maximal and singular integral operators via Fourier transform estimates. Invent. Math. 84, 541–561 (1986)
Duong, X.T., Li, J., Yang, D.: Variation of Calderón-Zygumund operators with matrix weight. Commun. Contemp. Math. 23, 2050062 (2021)
Fujii, N.: Weighted bounded mean oscillation and singular integrals. Math. Jpn. 22, 529–534 (1977/1978)
Gillespie, T., Torrea, J.: Dimension free estimates for the oscillation of Riesz transforms. Isr. J. Math. 141, 125–144 (2004)
Hong, G.: The behaviour of square functions from ergodic theory in \(L^\infty \). Proc. Am. Math. Soc. 143, 4797–4802 (2015)
Hytöen, T., Lacey, M.T., Pérez, C.: Sharp weighted bounds for the \(q\)-variation of singular integrals. Bull. Lond. Math. Soc. 45, 529–540 (2013)
Hytönen, T., Pérez, C.: Sharp weighted bounds involving \(A_\infty \). Anal. PDE 6, 777–818 (2013)
Hytönen, T., Lacey, M., Parissis, I.: A variation norm Carleson theorem for vector-valued Walsh-Fourier series. Rev. Mat. Iberoam. 30, 979–1014 (2014)
Hytönen, T., Roncal, L., Tapiola, O.: Quantitative weighted estimates for rough homogeneous singular integrals. Isr. J. Math. 218(1), 133–164 (2017)
Jones, R., Kaufman, R., Rosenblatt, J., Wierdl, M.: Oscillation in ergodic theory. Ergod. Theory Dyn. Syst. 18, 889–935 (1998)
Jones, R., Rosenblatt, J., Wierdl, M.: Oscillation in ergodic theory: higher dimensional results. Isr. J. Math. 135, 1–27 (2003)
Jones, R., Seeger, A., Wright, J.: Strong variational and jump inequalities in harmonic analysis. Trans. Am. Math. Soc. 360, 6711–6742 (2008)
Jones, R., Wang, G.: Variation inequalities for the Fejér and Poisson kernels. Trans. Am. Math. Soc 356, 4493–4518 (2004)
Krause, B., Zorin-Kranich, P.: Weighted and vector-valued variational estimates for ergodic averages. Ergod. Theory Dyn. Syst. 38, 244–256 (2018)
Le Merdy, C., Xu, Q.: Strong q-variation inequalities for analytic semigroups. Ann. Inst. Fourier (Grenoble) 62, 2069–2097 (2012)
Lépingle, D.: La variation d’ordre \(p\) des semi-martingales. Z. Wahrsch. Verw. Gebiete. 36, 295–316 (1976)
Lerner, A.K.: On pointwise estimates involving sparse operators. N. Y. J. Math. 22, 341–349 (2017)
Lerner, A.K.: A note on weighted bounds for rough singular integrals. C. R. Acad. Sci. Paris Ser. I(356), 77–80 (2018)
Ma, T., Torrea, J., Xu, Q.: Weighted variation inequalities for differential operators and singular integrals. J. Funct. Anal. 268, 376–416 (2015)
Ma, T., Torrea, J., Xu, Q.: Weighted variation inequalities for differential operators and singular integrals in higher dimensions. Sci. China Math. 60, 1419–1442 (2017)
Mas, A., Tolsa, X.: Variation for the Riesz transform and uniform rectifiability. J. Eur. Math. Soc. 16, 2267–321 (2014)
Mirek, M., Trojan, B.: Discrete maximal functions in higher dimensions and applications to ergodic theory. Am. J. Math. 138, 1495–1532 (2016)
Mirek, M., Stein, E.M., Trojan, B.: \(\ell _p(Z^d)\)-estimates for discrete operators of Radon type: variational estimates. Invent. Math. 209, 665–748 (2017)
Mirek, M., Trojan, B., Zorin-Kranich, P.: Variational estimates for averages and truncated singular integrals along the prime numbers. Trans. Am. Math. Soc. 69, 5403–5423 (2017)
Oberlin, R., Seeger, A., Tao, T., Thiele, C., Wright, J.: A variation norm Carleson theorem. J. Eur. Math. Soc. 14, 421–464 (2012)
Pisier, G., Xu, Q.: The strong \(p\)-variation of martingales and orthogonal series. Probab. Theory Relat. Fields 77, 497–514 (1988)
Qian, J.: The \(p\)-variation of partial sum processes and the empirical process. Ann. Prob. 77, 1370–1383 (1998)
Stein, E.M., Weiss, G.: Interpolation of operators with change of measures. Trans. Am. Math. Soc. 87, 159–172 (1958)
Stein, E.M.: Harmonic Analysis, Real Variable Methods, Orthogonality, and Oscillatory Integrals. Princeton Univ. Press, Princeton (1993)
Wilson, J.M.: Weighted inequalities for the dyadic square function without dyadic \(A_1\). Duke Math. J. 55, 19–50 (1987)
Zorin-Kranich, P.: Variation estimates for averages along primes and polynomials. J. Funct. Anal. 268, 210–238 (2015)
Acknowledgements
The authors would like to express their gratitude to the referees for giving several valuable suggestions, which have greatly improved the exposition of the paper.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Elijah Liflyand.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The project was in part supported by: Yanping Chen’s (the corresponding author)’s NSF of China (# 11871096); Guixiang Hong’s NSF of China (# 12071355); Ji Li’s ARC DP 170101060.
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
Chen, Y., Hong, G. & Li, J. Quantitative Weighted Bounds for the q-Variation of Singular Integrals with Rough Kernels. J Fourier Anal Appl 29, 31 (2023). https://doi.org/10.1007/s00041-023-10006-6
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00041-023-10006-6