Processing math: 2%

数学物理学报, 2020, 40(6): 1446-1460 doi:

论文

带变量核奇异积分算子的ρ-变差

龚珍兵,, 陈艳萍,, 陶文宇,

ρ-Variation for Singular Integral Operators with Variable Kernels

Gong Zhenbing,, Chen Yanping,, Tao Wenyu,

通讯作者: 陈艳萍, E-mail:yanpingch@126.com

收稿日期: 2020-03-22  

基金资助: 国家自然科学基金.  11871096
国家自然科学基金.  11471033

Received: 2020-03-22  

Fund supported: the NSFC.  11871096
the NSFC.  11471033

作者简介 About authors

龚珍兵,E-mail:gongzb@xs.ustb.edu.cn , E-mail:gongzb@xs.ustb.edu.cn

陶文宇,E-mail:wytao@xs.ustb.edu.cn , E-mail:wytao@xs.ustb.edu.cn

Abstract

In this paper, we will prove that the variation of singular integral operators with rough variable kernels are bounded on L2(Sn) if ΩL(Rn)×Lq(Sn1) for q>2(n1)/n, and n2. Moreover, we can also obtain the weighted variational inequalities for singular integral operators with smooth variable kernels. Finally, we extend the result to the Morrey spaces.

Keywords: Variational inequalities ; Variable kernel ; Singular integral ; Ap-weight ; Morrey space

PDF (377KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

龚珍兵, 陈艳萍, 陶文宇. 带变量核奇异积分算子的ρ-变差. 数学物理学报[J], 2020, 40(6): 1446-1460 doi:

Gong Zhenbing, Chen Yanping, Tao Wenyu. ρ-Variation for Singular Integral Operators with Variable Kernels. Acta Mathematica Scientia[J], 2020, 40(6): 1446-1460 doi:

1 引言

Sn1={xRn:|x|=1}Rn(n2)上的单位球面, dσ表示Sn1上的Lebesgue测度.设Rn×Sn1上的零阶齐次函数Ω(x,z)L(Rn)×Lq(Sn1)(q1)是指

Ω(x,λz)=Ω(x,z)对每个 x,zRn 且 λ>0,
(1.1)

(1.2)

其中 z' = z /|z|, z \in {{\Bbb R}} ^{n} \backslash\{0\} .

对于 \varepsilon>0 , 带变量核的截断奇异积分算子 T_{\Omega, \, \varepsilon} 被定义为

\begin{eqnarray} T_{\Omega, \varepsilon} f(x) = \int_{|x-y|>\varepsilon}\frac{\Omega(x, x-y)}{|x-y|^n}f(y){\rm d}y, \end{eqnarray}
(1.3)

其中 \Omega\in L^{\infty}({{\Bbb R}} ^{n}) \times L^{1}({\mathbb S}^{n-1}) 且满足消失条件

\begin{equation} \int_{{\mathbb S}^{n-1}} \Omega(x, y') {\rm d}\sigma(y^{\prime}) = 0, \end{equation}
(1.4)

则对于 f\in C^\infty_0({{\Bbb R}} ^n) , 带变量核奇异积分算子 T_\Omega 定义为

\begin{equation} T_\Omega f(x) = \lim\limits_{\varepsilon\rightarrow0^+}T_{\Omega, \, \varepsilon} f(x), \ a.e. \ x\in{{\Bbb R}} ^n. \end{equation}
(1.5)

关于变量核奇异积分算子的早期工作要追溯到1955年.当 \Omega 满足(1.4)式和 \Omega \in L^{\infty}({{\Bbb R}} ^{n}) \times L^{q}({\mathbb S}^{n-1})\, (q>2(n-1) / n) 时, Calderón和Zygmund (见文献[3])得到了 T_\Omega 是在 L^{2}({{\Bbb R}} ^{n}) 上有界的.同时, 他们也指出了 q>2(n-1) / n 是最优的条件.

众所周知, 奇异积分算子族的几乎处处收敛性问题往往归结为相关的极大奇异积分算子的有界性问题.在1980年, Aguilera和Harboure (见文献[1])得到了极大奇异积分算子 T_\Omega^{\ast} = \sup\limits_{\varepsilon>0}|T_{\Omega, \, \varepsilon} f(x)| L^{2}({{\Bbb R}} ^{n}) 有界性, 其中 \Omega \in L^{\infty}({{\Bbb R}} ^{n}) \times L^{q}({\mathbb S}^{n-1})(q>4(n-1) /(2 n-1)) 且满足(1.4)式.在 1985 年, Cowling和Mauceri (见文献[11])通过使用 \Omega 的球面调和展开的方法证明了当 q>2(n-1) / n 时, 上面结论仍然成立.利用旋转和混合范数估计的方法, Christ, Duoandikoetxea和Rubio de Francia (见文献[8])得到了相同的结果.回顾如下:

定理A[8, 11]  当 q>2(n-1) / n 时, 如果 \Omega \in L^{\infty}({{\Bbb R}} ^{n}) \times L^{q}({\mathbb S}^{n-1}) 且满足(1.4)式, 则 T_\Omega^{*} L^{2}({{\Bbb R}} ^{n}) 上的有界算子.

{\cal T}_\Omega 表示算子族 \{T_{\Omega, \varepsilon}\}_{\varepsilon>0} .众所周知, {\cal T}_\Omega 的变差不等式在学习带变量核奇异积分算子簇的几乎处处收敛性问题中也起到了至关重要的作用.受到定理A的启发, 一个有趣的问题是, {\cal T}_\Omega 的变差是否也能得到相同的结果.该文首先对这个问题做出了一个肯定的回答, 并建立了带粗糙核 {\cal T}_\Omega 的变差函数是 L^{2}({{\Bbb R}} ^{n}) 有界的.

在介绍该文的主要结果之前, 首先回顾一下变差算子的定义. 1976年, Lépingle[20]关于一般鞅序列得到了第一个变差不等式.十多年后, Pisier和Xu (见文献[27])给出了Lépingle结果的一个简单证明, 而Bourgain[2]利用Lépingle结果得到了遍历平均的变差不等式, 极大地改进了经典的Dunford-Schwartz极大遍历不等式的结果.从那时起, 大多数有关变差算子的工作在调和分析和遍历理论中出现(可参看文献[7, 9-10, 12, 16-17, 19, 22-25, 30]).

\rho\geq1 , 并且 {\mathfrak{a}} = \{a_{t}:t>0\} 是一个复数族, 复数族 {\mathfrak{a}} \rho -变差范数定义为

\|{\mathfrak{a}}\|_{V_{\rho}} = \sup\bigg(\sum\limits_{k\geq1}|a_{t_{k}}-a_{t_{k-1}}|^{\rho}\bigg)^{\frac{1}{\rho}},

这里的上确界取遍所有递增的正数序列 \{t_k:k\geq1\} .显然, 对于某个固定 t_0 , 有

\begin{equation} \|{\mathfrak{a}}\|_{L^{\infty}({{\Bbb R}} )}: = \sup\limits_{t \in {{\Bbb R}} }|a_{t}| \leq\|{\mathfrak{a}}\|_{V_{\rho}}+|a_{t_{0}}| \quad \rho \geq 1. \end{equation}
(1.6)

{\cal F} = \{F_{t}\}_{ t>0} 是定义在 {{\Bbb R}} ^{n} 上的Lebesgue可测函数族.通过复数族的 \rho -变差范数的定义, 可以定义函数族 {\cal F} 的强 \rho -变差函数 V_{\rho}({\cal F}) .固定 x\in{{\Bbb R}} ^{n} , 函数族 {\cal F} x 处的强 \rho -变差函数 V_{\rho}({\cal F}) 定义为

V_{\rho}({\cal F})(x) = \|\{F_{t}(x)\}_{t>0}\|_{V_{\rho}}\quad \rho\geq1.

类似地, 可以定义如下的强 \rho -变差算子

V_{\rho}({\cal A}f)(x) = \|\{{A}_{t}(f)(x)\}_{t>0}\|_{V_{\rho}} \quad \rho\geq1,

其中 {\cal A} = \{A_{t}\}_{t>0} L^{p}({{\Bbb R}} ^{n})\, (1\leq p\leq\infty) 上的算子族.

注意对于任意的 f\in L^{p}({{\Bbb R}} ^{n}) x\in{{\Bbb R}} ^{n} (见文献[7]或[12]), 有下面的逐点估计

\begin{equation} A^{\ast}f(x)\leq V_{\rho}({\cal A}f)(x)\quad \rho\geq1, \end{equation}
(1.7)

其中 A^{\ast} 是定义为 A^{\ast}f(x) = \sup\limits_{t>0}|{A}_{t}(f)(x)| 的极大算子.

\Omega(x, z') = \Omega(z') 且维数为 n = 1 的齐次核奇异积分算子 T_\Omega 是著名的Hilbert变换 H .在2000年, Campbell[9]等人首先考虑了截断Hilbert变换族 {\cal H}: = \{H_\varepsilon\}_{\varepsilon>0} 的强 \rho -变差算子 (\rho>2) L^p({{\Bbb R}} )\, (1<p<\infty) 有界的.在2002年, Campbell[10]等人也给出了 {\cal T}_\Omega 的强 \rho -变差算子 (\rho>2) L^p({{\Bbb R}} ^n) 有界的, 其中 \Omega(x, z') = \Omega(z') \in L\log^+ L({\Bbb S}^{n-1}) 且满足(1.4)式.

该文可以得到的第一个结果如下

定理1.1  令 {\cal T}_\Omega = \{T_{\Omega, \, \varepsilon}\}_{\varepsilon>0} 是由(1.3)式给出的带变量核的截断算子族.假设对于 q>2(n-1) / n , \Omega \in L^{\infty}({{\Bbb R}} ^{n}) \times L^{q}({\mathbb S}^{n-1}) 且满足(1.4)式.则存在一个常数 C > 0 使得

\begin{eqnarray} \|V_{\rho}({\cal T}_{\Omega}f)\|_{L^{2}({{\Bbb R}} ^{n})}\leq C\|f\|_{L^{2}({{\Bbb R}} ^{n})} \end{eqnarray}
(1.8)

对所有的 \rho>2 成立.

注1.1  通过类似于文献[3]的证明, 在定理1.1中的条件 q>2(n-1) / n 是最优的.从不等式(1.7)式, 不难发现, 定理1.1是定理A的改进和推广的结果.

关于(1.8)式的证明, 我们可以从文献[17]中获得一些想法.也就是说, 我们将分别证明长变差和短变差来得到所需要的估计(相关概念见第2节).但是, 因为需要处理的是粗糙变量核, 所以有一些细节是不同的.该文将使用核函数的球面调和展开和傅里叶变换来分别估计长变差和短变差算子.

值得注意的是在定理1.1的条件下, 核函数 \Omega 没有任何的光滑性.因此, 如果在 \Omega(x, z) 的第二个变量中加强其光滑性假设, 则可以得到带光滑变量核奇异积分算子的加权 \rho -变差不等式 (\rho>2) .

加权不等式理论在调和分析中有着广泛的应用(可参看文献[7, 14-15, 22-23, 31]).近年来, 加权变差不等式在遍历理论和调和分析中也得到了广泛的关注.最近Calderón-Zygmund算子的加权变差不等式已经在文献[22]和[23]中被证明.并且陈[7]等建立了卷积核粗糙算子的加权变差不等式.回顾 A_{p} 权的定义.令 w 是定义在 {{\Bbb R}} ^{n} 上的非负局部可积函数.对于 1<p<\infty , 如果存在常数 C>0 使得对每个方体 Q

\sup\limits_{Q}\bigg(\frac{1}{|Q|} \int_{Q} w(x) {\rm d}x\bigg)\bigg(\frac{1}{|Q|} \int_{Q} w(x)^{1-p'} {\rm d}x\bigg)^{p-1} \leq C,

其中 p' p 的共轭指数, 则称 w \in A_{p}

现在, 将介绍该文的第二个结果.

定理1.2  令 1<p<\infty , w\in A_p {\cal T}_\Omega = \{T_{\Omega, \, \varepsilon}\}_{\varepsilon>0} 是由(1.3)式给出的带变量核的截断算子族.假设 \Omega 满足(1.1)式, (1.4)式和

\begin{eqnarray} \max\limits_{|j|\le 2n}\| (\partial^j/\partial y^j)\Omega(x, y)\|_{L^\infty({\Bbb R}^n\times {\mathbb S}^{n-1})}<\infty, \end{eqnarray}
(1.9)

则对所有的 \rho>2 , 存在一个常数 C > 0 使得

\begin{eqnarray} \|V_{\rho}({\cal T}_{\Omega}f)\|_{L^{p}(w)}\leq C\|f\|_{L^{p}(w)}. \end{eqnarray}
(1.10)

1<p<\infty , \gamma \in(0, n) , Morrey空间 L^{p, \gamma}({{\Bbb R}} ^{n}) 定义为

L^{p, \gamma}({{\Bbb R}} ^{n}) = \{f \in L_{loc}^{p}({{\Bbb R}} ^{n}):\|f\|_{L^{p, \gamma}}<\infty\},

其中

\|f\|_{L^{p, \gamma}} = \bigg(\sup\limits_{x \in {{\Bbb R}} ^{n}, r>0} \frac{1}{r^{\gamma}} \int_{Q(x, r)}|f(y)|^{p} {\rm d}y\bigg)^{1 / p},

Q(x, r) 代表半径为 r , 圆心为 x 的任意方体.当 \gamma = 0 时, L^{p, \gamma}({{\Bbb R}} ^{n}) 与Lebesgue空间 L^{p}({{\Bbb R}} ^{n}) 一致.

Morrey空间 L^{p, \gamma}({{\Bbb R}} ^{n}) 是Morrey在1938年研究二阶椭圆型偏微分方程解的局部性质时所引进的(见文献[21]).对于Morrey空间的性质和应用, 读者可以参看文献[26, 28-29].

容易发现定理1.2的结果可以推广到Morrey空间上.即, 该文可以得到带光滑变量核奇异积分算子的强 \rho -变差函数 (\rho>2) 在Morrey空间 L^{p, \gamma}({{\Bbb R}} ^{n}) 上的有界性.

定理1.3  令 1<p<\infty , \gamma\in (0, n) , 且 {\cal T}_\Omega = \{T_{\Omega, \, \varepsilon}\}_{\varepsilon>0} 是由(1.3)式给出的带变量核的截断算子族.假设 \Omega 满足(1.1), (1.4)和(1.9)式, 则对所有的 \rho>2 , 存在一个常数 C >0 使得

\begin{equation} \|V_{\rho}({\cal T}_{\Omega}f)\|_{L^{p, \gamma}({{\Bbb R}} ^{n})}\leq C\|f\|_{L^{p, \gamma}({{\Bbb R}} ^{n})}. \end{equation}
(1.11)

我们可以将上述的变差结果应用到 \lambda -跳跃函数的估计.给定一个定义在 {{\Bbb R}} ^{n} 上的Lebesgue可测函数 {\cal F} = \{F_{t}\}_{ t>0} , 令 \lambda> 0 x \in {{\Bbb R}} ^{n} , 在 x 点的 \lambda -跳跃函数 N_{\lambda}({\cal F}) 被定义为

N_{\lambda}({\cal F})(x) = \sup\{J\in{\mathbb N} : \exists \; 0<s_{1} < t_{1}\leq s_{2} < t_{2} \leq \cdot\cdot\cdot \leq s_{J}< t_{J}\; \mbox{s.t.}\; |F_{t_{k}}(x)-F_{s_{k}}(x)| > \lambda\}.

在文献[17]中, Jones等人得到当 \lambda>0 \rho \geq 1 时, 有

\begin{equation} \lambda(N_{\lambda}({\cal F})(x))^{1 / \rho} \leq C V_{\rho}({\cal F})(x). \end{equation}
(1.12)

因此, 通过(1.12)式, 定理1.1, 定理1.2和定理1.3, 我们立即得到下面的跳跃估计.

推论1.1  与定理1.1的条件一样.令 \rho>2 , 存在一个常数 C > 0 使得

\sup\limits_{\lambda>0}\|\lambda(N_{\lambda}({\cal T}_{\Omega}f))^{\frac{1}{\rho}}\|_{L^{2}({{\Bbb R}} ^{n})}\leq C\|f\|_{L^{2}({{\Bbb R}} ^{n})}.

推论1.2  与定理1.2的条件一样.令 1<p<\infty , w\in A_p \rho>2 , 下面 \lambda -跳跃不等式成立

\sup\limits_{\lambda>0}\|\lambda(N_{\lambda}({\cal T}_{\Omega}f))^{\frac{1}{\rho}}\|_{L^{p}(w)}\leq C\|f\|_{L^{p}(w)}.

推论1.3  令 1<p<\infty \gamma\in (0, n) .与定理1.3的条件一样, 对于 \rho>2

\sup\limits_{\lambda>0}\|\lambda(N_{\lambda}({\cal T}_{\Omega}f))^{\frac{1}{\rho}}\|_{L^{p, \gamma}({{\Bbb R}} ^{n})}\leq C\|f\|_{L^{p, \gamma}({{\Bbb R}} ^{n})}.

本文的提纲如下.在第2节中, 该文建立了定理1.1的证明.第3节讨论了定理1.2的证明.在第4节中, 该文建立了定理1.3的证明.在该文中, 字母 C 表示一个正常数, 它是独立的基本变量, 在每次出现时可能不一样.

2 定理1.1的证明

在证明定理1.1之前, 将引入一些概念和引理.假设 m\in{\mathbb N} .定义 {\cal H}_{m} {\mathbb S}^{n-1} 上的 m 阶的球面调和函数空间并且是一个有限维的向量空间, {\rm dim} {\cal H}_{m} = D_{m}. \{Y_{m, j}\}_{j = 1}^{D_{m}} {\cal H}_{m} 上的标准正交基且(见文献[5])

\bigg(\sum\limits_{j = 1}^{D_{m}}|Y_{m, j}(\xi')|^{2}\bigg)^{\frac{1}{2}}\sim m^{(n-2)/2}\quad\hbox{对任意}\; \xi\neq0.
(2.1)

引理2.1[6]  对 0<\beta<1 , \ell\in{\mathbb Z} , m \in {\mathbb N} .

v_{\ell, m, j}(x) = \frac{Y_{m, j}(x')}{|x|^{n}}\chi_{\{2^{\ell}< |x|\leq2^{\ell+1}\}}(x),

则对 m\geq1 , 有

\begin{equation} |\widehat{v_{\ell, m, j}}(\xi)|\leq C m^{-\varrho-1+\beta / 2}\min\{|2^{\ell} \xi|, |2^{\ell} \xi|^{-\beta / 2}\}|Y_{m, j}(\xi^{\prime})|, \end{equation}
(2.2)

其中 \varrho = (n-2)/2 , \xi' = \xi/|\xi| .

下面证明定理1.1.为了证明定理1.1, 需要使用下面逐点不等式(见文献[17, 引理1.3])

\begin{eqnarray} V_{\rho}({\cal T}_\Omega f)(x)\leq C\Big(S_{2}({\cal T}_\Omega f)(x)+V_{\rho}(\{T_{\Omega, 2^{k}} f\}_{k\in{\mathbb Z}})(x)\Big), \end{eqnarray}
(2.3)

其中

S_{2}({\cal T}_\Omega f)(x) = \bigg(\sum\limits_{k\in{\mathbb Z}}[V_{2, k}({\cal T}_\Omega f)(x)]^{2}\bigg)^{1/2},

V_{2, k}({\cal T}_\Omega f)(x) = \bigg(\sup\limits_{ t_1<\cdots<t_J\atop [t_l, t_{l+1}]\subset[2^k, 2^{k+1}]} \sum\limits_{l = 1}^{J-1}|T_{\Omega, t_{l+1}}f(x)-T_{\Omega, t_l}f(x)|^2\bigg)^{1/2}.

因此, 只需证明

\begin{equation} \Big\|V_{\rho}(\{T_{\Omega, 2^{k}} f\}_{k\in{\mathbb Z}})\Big\|_{L^{2}({{\Bbb R}} ^{n})}\leq C\|f\|_{L^{2}({{\Bbb R}} ^{n})} \end{equation}
(2.4)

\begin{equation} \|S_{2}({\cal T}_\Omega f)\|_{L^{2}({{\Bbb R}} ^{n})}\leq C\|f\|_{L^{2}({{\Bbb R}} ^{n})}. \end{equation}
(2.5)

下面将分别估计(2.4)式和(2.5)式.

(2.4)式的证明.正如文献[5], 可以通过一个极限讨论将不等式(1.8)减少到 f \in C_{0}^{\infty}({{\Bbb R}} ^{n}) 的情况, 并且

\Omega(x, z') = \sum\limits_{m \geq 0} \sum\limits_{j = 1}^{D_{m}} a_{m, j}(x) Y_{m, j}(z')

是一个有限核.注意 \Omega(x, z') 满足消失性条件 (1.4) 式, 因此对任意的 1\leq j\leq D_{m} a_{0, j} \equiv 0 .定义

a_{m}(x) = \bigg(\sum\limits_{j = 1}^{D_{m}}|a_{m, j}(x)|^{2}\bigg)^{1 / 2}

b_{m, j}(x) = \frac{a_{m, j}(x)}{a_{m}(x)},

\begin{equation} \sum\limits_{j = 1}^{D_{m}} b_{m, j}^{2}(x) = 1. \end{equation}
(2.6)

T_{m, j}f(x) = {\rm p.v.}\int_{{{\Bbb R}} ^{n}}\frac{Y_{m, j}(y')}{|y|^{n}}f(x-y){\rm d}y,

T_{m, j, \varepsilon}f(x) = \int_{|y|>\varepsilon}\frac{Y_{m, j}(y')}{|y|^{n}}f(x-y){\rm d}y.

由强 \rho -变差算子的定义可以得出

V_{\rho}(\{T_{\Omega, 2^{k}} f\}_{k\in{\mathbb Z}})(x) = \sup\bigg(\sum\limits_{l\geq1}\bigg|T_{\Omega, 2^{k_{l}}} f(x)-T_{\Omega, 2^{k_{l-1}}} f(x)\bigg|^{\rho}\bigg)^{\frac{1}{\rho}},

其中上确界取遍所有的递增序列 \{2^{k_{l}}:l\geq1\} .则通过Minkowski不等式, 两次Hölder不等式和(2.6)式, 可得到对 0<\theta<1 , 有

\begin{eqnarray} &&\Big(V_{\rho}(\{T_{\Omega, 2^{k}} f\}_{k\in{\mathbb Z}})(x)\Big)^{2}\\ & = &\sup\bigg(\sum\limits_{l\geq1}\bigg|T_{\Omega, 2^{k_{l}}} f(x)-T_{\Omega, 2^{k_{l-1}}} f(x)\bigg|^{\rho}\bigg)^{\frac{2}{\rho}}\\ &\leq & C\bigg\{\sum^{\infty}_{m = 1}a^{2}_{m}(x)m^{-\theta}\bigg\}\sup\bigg(\sum\limits_{l\geq1}\bigg( \sum^{\infty}_{m = 1}m^{\theta}\sum\limits_{j = 1}^{D_{m}} \bigg| T_{m, j, 2^{k_{l}}} f(x)-T_{m, j, 2^{k_{l-1}}} f(x)\bigg|^{2}\bigg)^{\frac{\rho}{2}}\bigg)^{\frac{2}{\rho}}\\ &\leq & C\bigg\{\sum^{\infty}_{m = 1}a^{2}_{m}(x)m^{-\theta}\bigg\}\sum^{\infty}_{m = 1}m^{\theta}\sum\limits_{j = 1}^{D_{m}} \sup\bigg(\sum\limits_{l\geq1} \bigg| T_{m, j, 2^{k_{l}}} f(x)-T_{m, j, 2^{k_{l-1}}} f(x)\bigg|^{\rho}\bigg)^{\frac{2}{\rho}} \\ & = &C\bigg\{\sum^{\infty}_{m = 1}a^{2}_{m}(x)m^{-\theta}\bigg\} \bigg\{\sum^{\infty}_{m = 1}m^{\theta}\sum\limits_{j = 1}^{D_{m}} V^{2}_{\rho}(\{T_{m, j, 2^{k}}f\}_{k\in{\mathbb Z}})(x)\bigg\}. \end{eqnarray}
(2.7)

由文献[5], 对 q>2(n-1)/n , 如果令 0<\theta<1 并充分的接近1时, 有

\begin{eqnarray} \bigg(\sum^{\infty}_{m = 1}a^{2}_{m}(x)m^{-\theta}\bigg)^{1/2}&\leq & C\bigg(\int_{{\mathbb S}^{n-1}}|\Omega(x, z')|^{q}\, {\rm d}\sigma(z')\bigg)^{1/q}\\ &\leq &C\|\Omega\|_{L^{\infty}({{\Bbb R}} ^{n})\times L^{q}({\mathbb S}^{n-1})}. \end{eqnarray}
(2.8)

通过(2.7)式和(2.8)式, 有

\begin{eqnarray*} &&\Big\|V_{\rho}(\{T_{\Omega, 2^{k}} f\}_{k\in{\mathbb Z}})\Big\|^{2}_{L^{2}({{\Bbb R}} ^{n})}\\ &\leq& C\|\Omega\|^{2}_{L^{\infty}({{\Bbb R}} ^{n})\times L^{q}({\mathbb S}^{n-1})}\sum^{\infty}_{m = 1}m^{\theta}\bigg\|\bigg(\sum\limits_{j = 1}^{D_{m}} V^{2}_{\rho}(\{T_{m, j, 2^{k}}f\}_{k\in{\mathbb Z}})\bigg)^{\frac{1}{2}}\bigg\|^{2}_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray*}

显然, 如果可以证明对某个 0<\beta<1-\theta 使得

\begin{eqnarray} \bigg\|\bigg(\sum\limits_{j = 1}^{D_{m}} V^{2}_{\rho}(\{T_{m, j, 2^{k}}f\}_{k\in{\mathbb Z}})\bigg)^{\frac{1}{2}}\bigg\|_{L^{2}({{\Bbb R}} ^{n})}\leq Cm^{-1+\beta/2}\|f\|_{L^{2}({{\Bbb R}} ^{n})}, \end{eqnarray}
(2.9)

则(2.4)式将被证明.明显地, 对于 k\in{\mathbb Z} , 有

\begin{eqnarray*} T_{m, j, 2^k}f(x)& = & \int_{|y|>2^k}\frac{Y_{m, j}(y')}{|y|^{n}}f(x-y){\rm d}y\\ & = &\sum\limits_{s\geq k}\int_{2^{s}< |y|\leq2^{s+1}}\frac{Y_{m, j}(y')}{|y|^{n}}f(x-y){\rm d}y\nonumber\\ & = &\sum\limits_{s\geq k}v_{s, m, j}\ast f(x), \end{eqnarray*}

其中 v_{s, m, j}(x) = Y_{m, j}(x')|x|^{-n}\chi_{\{2^{s}< |x|\leq2^{s+1}\}}(x) .

\phi(x) 是径向的Schwartz函数使其在 |\xi|\leq2 {\hat{\phi}}(\xi) = 1 , 在 |\xi|>4 {\hat{\phi}}(\xi) = 0 , 则有如下相似于文献[17]的分解

\begin{eqnarray*} T_{m, j, 2^k}f(x)& = &\phi_{k}\ast T_{m, j}f(x)+\sum\limits_{s\geq 0}(\delta_{0}-\phi_{k})\ast v_{k+s, m, j}\ast f(x)-\phi_{k}\ast\sum\limits_{s<0}v_{k+s, m, j}\ast f(x)\nonumber\\ & = &:T_{k}^{1} f(x)+T_{k}^{2} f(x)-T_{k}^{3} f(x), \nonumber \end{eqnarray*}

其中 \phi_{k}(x) = 2^{-kn}\phi(2^{-k}x) \widehat{\phi_{k}}(\xi) = \hat{\phi}(2^{k}\xi) , \delta_{0} 是在 0 处的Dirac测度.令 {\cal T}_{m, j}^{i}f 表示族 \{T_{k}^{i} f\}_{k\in{\mathbb Z}} 其中 i = 1, 2, 3. 为了证明(2.9)式, 充分的去证明一下不等式

\begin{eqnarray} \bigg\|\bigg(\sum\limits_{j = 1}^{D_{m}} V_{\rho}^{2}({\cal T}_{m, j} ^{i}f)\bigg)^{\frac{1}{2}}\bigg\|_{L^{2}({{\Bbb R}} ^{n})}\leq Cm^{-1+\beta/2}\|f\|_{L^{2}({{\Bbb R}} ^{n})}, \; \; i = 1, 2, 3. \end{eqnarray}
(2.10)

(2.10)式在 i = 1 处的估计.由文献[12, 引理2.7]和文献[3], 可得

\begin{eqnarray*} \bigg\|\bigg(\sum\limits_{j = 1}^{D_{m}} V_{\rho}^{2}({\cal T}_{m, j} ^{1}f)\bigg)^{\frac{1}{2}}\bigg\|^{2}_{L^{2}({{\Bbb R}} ^{n})} &\leq & C\sum\limits_{j = 1}^{D_{m}}\big\| V_{\rho}({\cal T}_{m, j} ^{1}f)\big\|^{2}_{L^{2}({{\Bbb R}} ^{n})}\\ &\leq &C\sum\limits_{j = 1}^{D_{m}}\| T_{m, j}f\big\|^{2}_{L^{2}({{\Bbb R}} ^{n})}\\ &\leq& C m^{-2}\|f\|^{2}_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray*}

(2.10)式在 i = 2 处的估计.由Minkowski不等式得到

\begin{eqnarray*} \bigg(\sum\limits_{j = 2}^{D_{m}} V_{\rho}^{2}({\cal T}_{m, j} ^{2}f)(x)\bigg)^{1/2} & = & \bigg(\sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}\bigg|\sum\limits_{s\geq0}[(\delta_{0}-\phi_{k})\ast v_{k+s, m, j}]\ast f(x)\bigg|^{2}\bigg)^{1/2}\\ &\leq &C\sum\limits_{s\geq0} \bigg(\sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}|(\delta_{0}-\phi_{k})\ast v_{k+s, m, j}\ast f(x)|^{2}\bigg)^{1/2}\\ & = &:C\sum\limits_{s\geq0}G_{s}f(x). _{} \end{eqnarray*}

\psi(\xi) = \psi(|\xi|)\in C_{c}^{\infty}({{\Bbb R}} ^{n}) 使得 0\leq\psi\leq1 , \rm{supp}\psi\subset\{\xi:1/2\leq|\xi|\leq2\} 和在 |\xi|\neq1 \sum\limits_{l\in{\mathbb Z}}\psi^{2}(2^{-l}\xi) = 1 . \widehat{\Delta_{l}f}(\xi) = \psi(2^{-l}\xi)\hat{f}(\xi) 定义乘子 \Delta_{l} .则有

\begin{eqnarray*} G_{s}f(x)& = & \bigg(\sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}|(\delta_{0}-\phi_{k})\ast v_{k+s, m, j}\ast \sum\limits_{l\in{\mathbb Z}}\Delta^{2}_{l-k}f(x)|^{2}\bigg)^{1/2} \\ &\leq &\sum\limits_{l\in{\mathbb Z}}\bigg( \sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}|(\delta_{0}-\phi_{k})\ast v_{k+s, m, j}\ast \Delta^{2}_{l-k}f(x)|^{2}\bigg)^{1/2}\\ & = & :\sum\limits_{l\in{\mathbb Z}}G_{s}^{l}f(x). \end{eqnarray*}

l\in{\mathbb Z} s\geq0 , 将去建立 \|G_{s}^{l}f\|_{L^{2}({{\Bbb R}} ^{n})} 的估计.因为

\hbox{supp}((1-\widehat{\phi_{k}})\widehat{v_{k+s, m, j}})\subset \{\xi:|2^{k}\xi|>1\},

通过引理2.1, 对 0<\beta<1 \varrho = \frac{n-2}{2} , 有

\begin{eqnarray} |(1-\widehat{\phi_{k}})(\xi)||\widehat{v_{k+s, m, j}}(\xi)|\leq C m^{-1-\varrho+\beta/2}2^{-\beta s/2}\min\{|2^{k}\xi|^{-\beta/2}, |2^{k}\xi|\}|Y_{m, j}(\xi')|. \end{eqnarray}
(2.11)

因为 \rm{supp}\psi\subset\{\xi:1/2\leq|\xi|\leq2\} , 那是很容易看出

\begin{equation} |\psi(2^{k-l}\xi)||(1-\widehat{\phi_{k}})(\xi)||\widehat{v_{k+s, m, j}}(\xi)| \leq C m^{-1-\varrho+\beta/2}2^{-\beta s/2}\min\{2^{-l\beta /2}, 2^{l}\}|Y_{m, j}(\xi')|. \end{equation}
(2.12)

应用Plancherel定理, Littlewood-Paley理论和 \sum\limits_{j = 1}^{D_{m}}|Y_{m, j}(\xi')|^{2}\sim m^{2\varrho} , 可以得到对某个 0<\nu<1 , 有

\begin{eqnarray*} \|G_{s}^{l}f\|^{2}_{L^{2}({{\Bbb R}} ^{n})}& = &\int_{{{\Bbb R}} ^{n}} \sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}\psi^{2}(2^{k-l}\xi)|(1-\widehat{\phi_{k}})(\xi)|^{2}|\widehat{v_{k+s, m, j}}(\xi)|^{2}|\widehat{\Delta_{l-k}f}(\xi)|^{2}{\rm d}\xi\\ &\leq & Cm^{-2-2\varrho+\beta}2^{-\beta s}2^{-\beta\nu|l|}\int_{{{\Bbb R}} ^{n}}\sum\limits_{j = 1}^{D_{m}}|Y_{m, j}(\xi')|^{2} \sum\limits_{k\in{\mathbb Z}}|\widehat{\Delta_{l-k}f}(\xi)|^{2}{\rm d}\xi\\ &\leq & Cm^{-2+\beta}2^{-\beta s}2^{-\beta\nu|l|} \|f\|^{2}_{L^{2}({{\Bbb R}} ^{n})}, \end{eqnarray*}

则有

\begin{eqnarray*} \bigg\|\bigg(\sum\limits_{j = 2}^{D_{m}} V_{\rho}^{2}({\cal T}_{m, j} ^{2}f)\bigg)^{1/2}\bigg\|_{L^{2}({{\Bbb R}} ^{n})}&\leq & Cm^{-1+\beta/2}\sum\limits_{s\geq0}\sum\limits_{l\in{\mathbb Z}}2^{-\beta s/2}2^{-\beta\nu|l|/2} \|f\|_{L^{2}({{\Bbb R}} ^{n})}\\ &\leq & Cm^{-1+\beta/2}\|f\|_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray*}

(2.10)式在 i = 3 处的估计.类似地, 有下面逐点分解

\begin{eqnarray*} \bigg(\sum\limits_{j = 2}^{D_{m}} V_{\rho}^{2}({\cal T}_{m, j} ^{3}f)(x)\bigg)^{1/2} &\leq & C\sum\limits_{s<0}\bigg(\sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}} |\phi_{k}\ast v_{k+s, m, j}\ast f(x)|^{2}\bigg)^{1/2}\\ & = &: C\sum\limits_{s<0}H_{s}f(x). \end{eqnarray*}

由Minkowski不等式和 \sum\limits_{l\in{\mathbb Z}} \Delta^{2}_{l}f = f , 可以得到

\begin{eqnarray*} H_{s}f(x)& = & \bigg( \sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}\bigg|\phi_{k}\ast v_{k+s, m, j}\ast \sum\limits_{l\in{\mathbb Z}}\Delta^{2}_{l-k}f(x)\bigg|^{2}\bigg)^{1/2}\nonumber\\ &\leq & \sum\limits_{l\in{\mathbb Z}}\bigg( \sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}|\phi_{k}\ast v_{k+s, m, j}\ast \Delta^{2}_{l-k}f(x)|^{2}\bigg)^{1/2}\\ & = &: \sum\limits_{l\in{\mathbb Z}}H_{s}^{l}f(x). \end{eqnarray*}

因此, 只需要考虑当 l\in{\mathbb Z} s<0 时, H_{s}^{l}f(x) L^{2}({{\Bbb R}} ^{n}) 有界性.因为 \hbox{supp}(\widehat{\phi_{k}}\widehat{v_{k+s, m, j}})\subset \{\xi:|2^{k}\xi|<1\} , 通过引理2.1, 可以得到对 0<\beta<1 \varrho = \frac{n-2}{2} ,

\begin{eqnarray*} |\widehat{\phi_{k}}(\xi)||\widehat{v_{k+s, m, j}}(\xi)|\leq C m^{-1-\varrho+\beta/2}2^{ s}\min\{|2^{k}\xi|^{-\beta/2}, |2^{k }\xi|\}|Y_{m, j}(\xi')|. \end{eqnarray*}

因为 \rm{supp}\psi\subset\{\xi:1/2\leq|\xi|\leq2\} , 有

\begin{eqnarray} |\psi(2^{k-l}\xi)||\widehat{\phi_{k}}(\xi)||\widehat{v_{k+s, m, j}}(\xi)|\leq C m^{-1-\varrho+\beta/2}2^{s}\min\{2^{-\beta l/2}, 2^{l}\}|Y_{m, j}(\xi')|. \end{eqnarray}
(2.13)

应用上面的估计, Littlewood-Paley理论和 \sum\limits_{j = 1}^{D_{m}}|Y_{m, j}(\xi')|^{2}\sim m^{2\varrho} , 可以得到对某个 0<\vartheta<1 , 有

\begin{eqnarray*} \|H_{s}^{l}f\|^{2}_{L^{2}({{\Bbb R}} ^{n})}& = &\int_{{{\Bbb R}} ^{n}} \sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}|\psi(2^{k-l}\xi)|^{2}|\widehat{\phi_{k}}(\xi)|^{2}|\widehat{v_{k+s, m, j}}(\xi)|^{2} |\widehat{\Delta_{l-k}f}(\xi)|^{2}{\rm d}\xi\\ &\leq& Cm^{-2-2\varrho+\beta}2^{ 2s}2^{-\vartheta|l| }\int_{{{\Bbb R}} ^{n}}\sum\limits_{j = 1}^{D_{m}}|Y_{m, j}(\xi')|^{2} \sum\limits_{k\in{\mathbb Z}}|\widehat{\Delta_{l-k}f}(\xi)|^{2}{\rm d}\xi\\ &\leq & Cm^{-2+\beta}2^{2 s}2^{-\vartheta|l|} \|f\|^{2}_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray*}

因此, 得到

\begin{eqnarray*} \bigg\|\bigg(\sum\limits_{j = 2}^{D_{m}} V_{\rho}^{2}({\cal T}_{m, j} ^{3}f)\bigg)^{1/2}\bigg\|_{L^{2}({{\Bbb R}} ^{n})} &\leq & Cm^{-1+\beta/2}\sum\limits_{s<0}\sum\limits_{l\in{\mathbb Z}}2^{ s}2^{-\vartheta|l|/2} \|f\|_{L^{2}({{\Bbb R}} ^{n})}\\ &\leq & Cm^{-1+\beta/2}\|f\|_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray*}

(2.5)式的证明.令 t\in[1, 2) , 首先定义 v_{0, t, m, j}

\begin{eqnarray*} v_{0, t, m, j}(x) = \frac{Y_{m, j}(x')}{|x|^{n}}\chi_{\{t\leq|x|\leq2\}}(x) \end{eqnarray*}

v_{k, t, m, j}(x) = 2^{-k n}v_{0, t, m, j}(2^{-k}x) 其中 k\in{\mathbb Z} .很容易得到

V_{2, k}({\cal T}_\Omega f)(x) = \bigg\|\bigg\{\sum\limits_{m \geq 1} a_{m}(x) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(x)v_{k, t, m, j}\ast f(x)\bigg\}_{t\in[1, 2]}\bigg\|_{V_{2}}.

由Minkowski不等式和 \sum\limits_{l\in {\mathbb Z}}\Delta_l^2f = f , 有

\begin{eqnarray*} S_{2}({\cal T}_\Omega f)(x)& = & \bigg(\sum\limits_{k\in{\mathbb Z}}[V_{2, k}({\cal T}_\Omega f)(x)]^{2}\bigg)^{1/2}\\ & = &\bigg(\sum\limits_{k\in{\mathbb Z}}\bigg\|\bigg\{\sum\limits_{m \geq 1} a_{m}(x) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(x)v_{k, t, m, j}\ast f(x)\bigg\}_{t\in[1, 2]}\bigg\|_{V_{2}}^{2}\bigg)^{\frac{1}{2}}\\ &\leq& \sum\limits_{l\in{\mathbb Z}}\bigg(\sum\limits_{k\in{\mathbb Z}}\bigg\|\bigg\{\sum\limits_{m \geq 1} a_{m}(x) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(x)v_{k, t, m, j}\ast \Delta^{2}_{l-k} f(x)\bigg\}_{t\in[1, 2]}\bigg\|_{V_{2}}^{2}\bigg)^{\frac{1}{2}}\\ & = & :\sum\limits_{l\in{\mathbb Z}}S_{2, l}({\cal T}_\Omega f)(x), \end{eqnarray*}

则仅仅需要去估计当 l\in{\mathbb Z} 时, S_{2, l}({\cal T}_\Omega f) L^{2}({{\Bbb R}} ^{n}) 有界性.因为

\|{\mathfrak{a}}_{t}\|_{V_{2}}\leq C\|{\mathfrak{a}}_{t}\|_{L^{2}(X)}^{1/2}\|\frac{{\rm d}t}{t}{\mathfrak{a}}_{t}\|_{L^{2}(X)}^{1/2},

其中 X = L^{2}([1, 2], \frac{{\rm d}t}{t}) (见文献[17]), 则

\begin{eqnarray*} [S_{2, \, l}({\cal T}_\Omega f)(x)]^{2}&\leq& C\sum\limits_{k\in{\mathbb Z}}\bigg\|\sum\limits_{m \geq 1} a_{m}(x) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(x)v_{k, t, m, j}\ast \Delta^{2}_{l-k} f(x)\bigg\|_{L^{2}_{t}([1, 2])}\\ &&\times\bigg\|\frac{\rm d}{{\rm d}t}\bigg(\sum\limits_{m \geq 1} a_{m}(x) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(x)v_{k, t, m, j}\ast \Delta^{2}_{l-k} f(x)\bigg)\bigg\|_{L^{2}_{t}([1, 2])}. \end{eqnarray*}

通过Cauchy-Schwarz不等式, 有

\begin{eqnarray} &&\|S_{2, \, l}({\cal T}_\Omega f)\|^{2}_{L^{2}({{\Bbb R}} ^{n})}\\ &\leq& C \bigg\|\bigg(\sum\limits_{k\in{\mathbb Z}}\Big\|\sum\limits_{m \geq 1} a_{m}(\cdot) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(\cdot)v_{k, t, m, j}\ast \Delta^{2}_{l-k} f\Big\|^{2}_{L^{2}_{t}([1, 2])}\bigg)^{\frac{1}{2}}\bigg\|_{L^{2}({{\Bbb R}} ^{n})}\\ && \times \bigg\|\bigg(\sum\limits_{k\in{\mathbb Z}}\Big\|\frac{\rm d}{{\rm d}t}\bigg(\sum\limits_{m \geq 1} a_{m}(\cdot) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(\cdot)v_{k, t, m, j}\ast \Delta^{2}_{l-k} f\bigg)\Big\|^{2}_{L^{2}_{t}([1, 2])}\bigg)^{\frac{1}{2}}\bigg\|_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray}
(2.14)

现在估计(2.14)式右边的第一项.利用两次Hölder不等式, (2.6)式和(2.8)式, 可以得到

\begin{eqnarray*} & &\bigg\|\bigg(\sum\limits_{k\in{\mathbb Z}}\Big\|\sum\limits_{m \geq 1} a_{m}(\cdot) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(\cdot)v_{k, t, m, j}\ast \Delta^{2}_{l-k} f\Big\|^{2}_{L^{2}_{t}([1, 2])}\bigg)^{\frac{1}{2}}\bigg\|^{2}_{L^{2}({{\Bbb R}} ^{n})}\\ &\leq & C\|\Omega\|^{2}_{L^{\infty}({{\Bbb R}} ^{n})\times L^{q}({\mathbb S}^{n-1})}\bigg\|\bigg(\sum\limits_{k\in{\mathbb Z}}\sum^{\infty}_{m = 1}m^{\theta} \sum\limits_{j = 1}^{D_{m}}\Big\|v_{k, t, m, j}\ast \Delta^{2}_{l-k} f\Big\|^{2}_{L^{2}_{t}([1, 2])}\bigg)^{\frac{1}{2}}\bigg\|^{2}_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray*}

通过引理2.1, 得到对 0<\beta<1 \varrho = \frac{n-2}{2} , 有

\begin{eqnarray*} |\widehat{v_{k, t, m, j}}(\xi)|\leq m^{-\varrho-1+\beta / 2}\min\{|2^{k} \xi|, |2^{k} \xi|^{-\beta / 2}\}|Y_{m, j}(\xi^{\prime})|, \end{eqnarray*}

t\in[1, 2) 上一致的.因此, 通过Plancherel定理, Littlewood-Paley理论, \sum\limits_{j = 1}^{D_{m}}|Y_{m, j}(\xi')|^{2}\sim m^{2\varrho}, 并且令 0<\beta<1-\theta , 可以得到对某个 0<\mu<1 , 有

\begin{eqnarray} && \bigg\|\bigg(\sum\limits_{k\in{\mathbb Z}}\sum^{\infty}_{m = 1}m^{\theta} \sum\limits_{j = 1}^{D_{m}}\|v_{k, t, m, j}\ast \Delta^{2}_{l-k} f\|^{2}_{L^{2}_{t}([1, 2])}\bigg)^{\frac{1}{2}}\bigg\|^{2}_{L^{2}({{\Bbb R}} ^{n})}\\ & = & \sum^{\infty}_{m = 1}m^{\theta} \sum\limits_{j = 1}^{D_{m}}\sum\limits_{k\in{\mathbb Z}}\int_{1}^{2}\int_{{{\Bbb R}} ^{n}}|\psi(2^{k-l}\xi)|^{2}|\widehat{v_{k, t, m, j}}(\xi)|^{2}|\widehat{\Delta_{l-k} f}(\xi)|^{2}{\rm d}\xi \frac{{\rm d}t}{t}\\ &\leq & C\sum^{\infty}_{m = 1}m^{-2\varrho-2+\beta+\theta}2^{-\mu|l|}\int_{{{\Bbb R}} ^{n}}\sum\limits_{k\in{\mathbb Z}}|\widehat{\Delta_{l-k} f}(\xi)|^{2}\sum\limits_{j = 1}^{D_{m}}|Y_{m, j}(\xi')|^{2}{\rm d}\xi\\ &\leq& C2^{-\mu|l|}\|f\|^{2}_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray}
(2.15)

接下来去估计(2.14)式右边的第二项.通过球坐标变换, 和一个简单的计算可以得出

\begin{eqnarray*} &&\frac{\rm d}{{\rm d}t}\bigg(\sum\limits_{m \geq 1} a_{m}(x) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(x)v_{k, t, m, j}\ast h(x)\bigg)\\ & = & \frac{\rm d}{{\rm d}t}\bigg[\int_{2^{k}t<|y|\leq 2^{k+1}}\frac{\Omega(x, y')}{|y|^{n}}h(x-y){\rm d}y\bigg]\\ & = & \frac{\rm d}{{\rm d}t}\bigg[\int_{{\mathbb S}^{n-1}}\Omega(x, y')\int_{2^{k}t}^{ 2^{k+1}}\frac{1}{r}h(x-r y'){\rm d}r {\rm d}\sigma(y')\bigg]\\ & = & -\frac{1}{t}\int_{{\mathbb S}^{n-1}}\Omega(x, y')h(x-2^{k}t y'){\rm d}\sigma(y'). \end{eqnarray*}

注意 t\in[1, 2) , 通过Minkowski不等式, 有

\begin{eqnarray*} & &\bigg(\int_{{{\Bbb R}} ^{n}}\bigg|\frac{\rm d}{{\rm d } t}\bigg(\sum\limits_{m \geq 1} a_{m}(x) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(x) v_{k, t, m, j}\ast h(x)\bigg)\bigg|^{2} {\rm d}x\bigg)^{1 / 2}\\ & = &\bigg(\int_{{{\Bbb R}} ^{n}}\bigg|\frac{1}{t} \int_{{\mathbb S}^{n-1}} \Omega(x, y^{\prime}) h(x-2^{k} t y^{\prime}) {\rm d}\sigma(y^{\prime})\bigg|^{2} {\rm d}x\bigg)^{1 / 2}\\ &\leq& C \int_{{\mathbb S}^{n-1}} \sup\limits_{x \in {{\Bbb R}} ^{n}}|\Omega(x, y^{\prime})|\bigg(\int_{{{\Bbb R}} ^{n}}|h(x-2^{k} t y^{\prime})|^{2} {\rm d} x\bigg)^{1 / 2} {\rm d}\sigma(y^{\prime}) \\ &\leq& C \int_{{\mathbb S}^{n-1}} \sup\limits_{x \in {{\Bbb R}} ^{n}}|\Omega(x, y^{\prime})| {\rm d}\sigma(y^{\prime})\|h\|_{L^{2}({{\Bbb R}} ^{n})} \\ &\leq& C\|\Omega\|_{L^{\infty}({{\Bbb R}} ^{n}) \times L^{q}({\mathbb S}^{n-1})}\|h\|_{L^{2}({{\Bbb R}} ^{n})}, \end{eqnarray*}

则通过Littlewood-Paley理论知道

\begin{eqnarray} & & \bigg\|\bigg(\sum\limits_{k\in{\mathbb Z}}\Big\|\frac{\rm d}{{\rm d}t}\bigg(\sum\limits_{m \geq 1} a_{m}(\cdot) \sum\limits_{j = 1}^{D_{m}} b_{m, j}(\cdot)v_{k, t, m, j}\ast \Delta^{2}_{l-k} f\bigg)\Big\|^{2}_{L^{2}_{t}([1, 2])}\bigg)^{\frac{1}{2}}\bigg\|_{L^{2}({{\Bbb R}} ^{n})}^2\\ &\leq& C\|\Omega\|^{2}_{L^{\infty}({{\Bbb R}} ^{n})\times L^{q}({\mathbb S}^{n-1})}\int_1^2\int_{{{\Bbb R}} ^{n}}\sum\limits_{k\in{\mathbb Z}}|\Delta^{2}_{l-k}f(x)|^{2}{\rm d}x\frac{{\rm d}t}{t}\\ &\leq & C\|\Omega\|^{2}_{L^{\infty}({{\Bbb R}} ^{n})\times L^{q}({\mathbb S}^{n-1})}\|f\|^{2}_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray}
(2.16)

由(2.15)式和(2.16)式可以推出

\begin{eqnarray} \|S_{2, \, l}({\cal T}_\Omega f)\|_{L^{2}({{\Bbb R}} ^{n})}\leq C2^{-\mu|l|/4} \|\Omega\|_{L^{\infty}({{\Bbb R}} ^{n})\times L^{q}({\mathbb S}^{n-1})}\|f\|_{L^{2}({{\Bbb R}} ^{n})}, \end{eqnarray}
(2.17)

\begin{eqnarray*} \|S_{2}({\cal T}_\Omega f)\|_{L^{2}({{\Bbb R}} ^{n})}&\leq & C\sum\limits_{l\in{\mathbb Z}}2^{-\mu|l|/4} \|\Omega\|_{L^{\infty}({{\Bbb R}} ^{n})\times L^{q}({\mathbb S}^{n-1})}\|f\|_{L^{2}({{\Bbb R}} ^{n})}\\ &\leq & C\|\Omega\|_{L^{\infty}({{\Bbb R}} ^{n})\times L^{q}({\mathbb S}^{n-1})}\|f\|_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray*}

因此, 完成了定理1.1的证明.

3 定理1.2的证明

为证明定理1.2, 首先引入一些 \{Y_{m, j}(x')\}_{j = 1}^{D_{m}} 的性质.如参考文献[5]中所示, 有

\begin{eqnarray} D_{m}\leq Cm^{n-2}, \quad m\geq1. \end{eqnarray}
(3.1)

对任意的多重指标 \alpha (见文献[4])

\begin{equation} \sup\limits_{x' \in {\mathbb S}^{n-1}}|(\frac{\partial}{\partial x'})^{\alpha} Y_{m, j}(x')| \leqslant C m^{|\alpha|+(n-2) / 2}, \quad m = 1, 2, \cdots.\nonumber \end{equation}
(3.2)

如果 \varphi \in C^{\infty}({\mathbb S}^{n-1}) 那么 \sum\limits_{m} \sum\limits_{j} a_{m, j} Y_{m, j}(x')\varphi(x') 是关于 \{Y_{m, j}(x')\}_{m, j} 的Fourier级数展开, 其中

a_{m, j} = \int_{{\mathbb S}^{n-1}} \varphi(y') Y_{m j}(y') {\rm d}\sigma(y').

由文献[4]知

\begin{eqnarray} a_{m, j} = (-1)^{l} m^{-l}(m+n-2)^{-l} \int_{{\mathbb S}^{n-1}} L^{l}(\varphi) Y_{m, j}(x') {\rm d}\sigma(x'), \quad m \geq 1, \end{eqnarray}
(3.3)

其中 L(\varphi) = |x|^{2} \Delta \varphi(x) 和任意的正整数 l , L^{l+1}(\varphi) = L(L^{l}(\varphi)) .特别地, \varphi 的球调和展开一致收敛于 \varphi .

现在去证明定理1.2.对 x, y\in {{\Bbb R}} ^{n} , 由(1.4)式知

\begin{eqnarray*} \Omega(x, z') = \sum\limits_{m \geq 0} \sum\limits_{j = 1}^{D_{m}} a_{m, j}(x) Y_{m, j}(z'), \end{eqnarray*}

其中

a_{m, j}(x) = \int_{{\mathbb S}^{n-1}}\Omega(x, z')Y_{m, j}(z'){\rm d} z'.

且对每个 m\in \Bbb N j = 1, \dots, D_{m} a_{0, j}\equiv 0 .根据(1.9)式和(3.3)式, 可以得到

\begin{eqnarray} \|a_{m, j}\|_{L^{\infty}} \leq C m^{-2 n}. \end{eqnarray}
(3.4)

由强 \rho -变差算子的定义, Minkowski不等式和(3.4)式可以得出

\begin{eqnarray} V_{\rho}({\cal T}_{\Omega}f)(x) & = &\sup\bigg(\sum\limits_{l\geq1}\bigg|T_{\Omega, \varepsilon_{l}} f(x)-T_{\Omega, \varepsilon_{l-1}} f(x)\bigg|^{\rho}\bigg)^{\frac{1}{\rho}}\\ & = &\sup\bigg(\sum\limits_{l\geq1}\bigg|\int_{|x-y|>\varepsilon_{l}}\frac{\Omega(x, x-y)}{|x-y|^{n}}f(y){\rm d}y-\int_{|x-y|>\varepsilon_{l-1}}\frac{\Omega(x, x-y)}{|x-y|^{n}}f(y){\rm d}y\bigg|^{\rho}\bigg)^{\frac{1}{\rho}}\\ &\leq& C\sum\limits_{m \geq 1} \sum\limits_{j = 1}^{D_{m}}|a_{m, j}(x)|\sup\bigg(\sum\limits_{l\geq1}\bigg|\int_{|x-y|>\varepsilon_{l}}\frac{Y_{m, j}(x-y)}{|x-y|^{n}}f(y){\rm d}y\\ &&-\int_{|x-y|>\varepsilon_{l-1}}\frac{Y_{m, j}(x-y)}{|x-y|^{n}}f(y){\rm d}y\bigg|^{\rho}\bigg)^{\frac{1}{\rho}}\\ & = &C\sum\limits_{m \geq 1} \sum\limits_{j = 1}^{D_{m}}|a_{m, j}(x)|\sup\bigg(\sum\limits_{l\geq1}\bigg|T_{m, j, \varepsilon_{l}} f(x)-T_{m, j, \varepsilon_{l-1}} f(x)\bigg|^{\rho}\bigg)^{\frac{1}{\rho}}\\ & = &:C\sum\limits_{m \geq 1} \sum\limits_{j = 1}^{D_{m}}m^{-2 n}V_{\rho}({\cal T}_{m, j}f)(x), \end{eqnarray}
(3.5)

其中上确界取遍所有的递增序列 \{\varepsilon_{l}:l\geq1\} 并且 {\cal T}_{m, j}: = \{T_{m, j, \, \varepsilon}\}_{\varepsilon>0} .

接下来, 只需去证明 V_{\rho}({\cal T}_{m, j}f) 的加权有界性.可以发现

(1) 对于 x\neq0 , 有

\begin{eqnarray*} \frac{|Y_{m, j}(x)|}{|x|^{n}}\leq Cm^{n/2}\frac{1}{|x|^{n}}; \end{eqnarray*}

(2) 对于 |y|>2|x|>0 , 有

\begin{eqnarray*} \bigg|\frac{|Y_{m, j}(x-y)|}{|x-y|^{n}}-\frac{|Y_{m, j}(-y)|}{|-y|^{n}}\bigg|\leq Cm^{n/2}\frac{|x|}{|y|^{n+1}}, \end{eqnarray*}

这个基本不等式在参考文献[13]已被证明.

(3) 用证明定理1.1相同的方法可以得到

\begin{eqnarray*} \|V_{\rho}({\cal T}_{m, j}f)\|_{L^{2}({{\Bbb R}} ^{n})}\leq Cm^{n/2}\|f\|_{L^{2}({{\Bbb R}} ^{n})}. \end{eqnarray*}

由上面三个估计和文献[23, 定理1.1], 易证当 1<p<\infty w\in A_{p} 时, 有

\begin{eqnarray*} \|V_{\rho}({\cal T}_{m, j}f)\|_{L^{p}(w)}\leq Cm^{n/2}\|f\|_{L^{p}(w)}. \end{eqnarray*}

因此, 通过上面估计, 以及(3.1)和(3.5)式, 对于 1<p<\infty w\in A_{p} , 可得

\begin{eqnarray*} \|V_{\rho}({\cal T}_{\Omega}f)\|_{L^{p}(w)}&\leq &C\sum\limits_{m \geq 1} \sum\limits_{j = 1}^{D_{m}}m^{-2 n}\|V_{\rho}({\cal T}_{m, j}f)\|_{L^{p}(w)}\\ &\leq& C\sum\limits_{m \geq 1} m^{n-2}m^{-2 n}m^{n/2}\|f\|_{L^{p}(w)}\leq C\|f\|_{L^{p}(w)}. \end{eqnarray*}

因此, 完成了定理1.2的证明.

4 定理1.3的证明

这个定理证明的想法来源于参考文献[13].对于固定的 t \in {{\Bbb R}} ^{n} r>0 , 令 B = B(t, r) .对于 f \in L^{p, \gamma}({{\Bbb R}} ^{n}) , 有

f(y) = f(y) \chi_{2 B}(y)+\sum\limits_{k = 1}^{\infty} f(y) \chi_{2^{k+1} B \backslash 2^{k} B}(y) = :\sum\limits_{k = 0}^{\infty} f_{k}(y).

k = 0 , 由(1.10)式和 w(x) = 1 , 可得

\begin{eqnarray*} \int_{B}|V_{\rho}({\cal T}_{\Omega}f_{0})(x)|^{p} {\rm d}x & \leq&\|V_{\rho}({\cal T}_{\Omega}f_{0})\|_{L^{p}({{\Bbb R}} ^{n})}^{p} \leq \|f_{0}\|_{L^{p}({{\Bbb R}} ^{n})}^{p} \\ & = &C \int_{2B}|f(y)|^{p} {\rm d}x \leq C(2 r)^{\gamma}\|f\|_{L^{p, \gamma}({{\Bbb R}} ^{n})}^{p}. \end{eqnarray*}

k>0 , 注意当 x \in 2^{k+1} B \backslash 2^{k} B 时, M(\chi_{B})(x) \sim 2^{-k n} . \eta \in(\gamma / n, 1) , 知道 (M \chi_{B})^{\eta} \in A_{1} \subset A_{p} 其中 A_{1} 的常数仅仅取决于 \gamma n .对于 k>0 , 由(1.10)式, 令 1<p<\infty \gamma\in (0, n) , 有

\begin{eqnarray*} \int_{B}|V_{\rho}({\cal T}_{\Omega}f_{k})(x)|^{p} {\rm d}x & = &\int_{{{\Bbb R}} ^{n}} |V_{\rho}({\cal T}_{\Omega}f_{k})(x)|^{p}(\chi_{B}(x))^{\eta} {\rm d}x \\ & \leq & \int_{{{\Bbb R}} ^{n}}|V_{\rho}({\cal T}_{\Omega}f_{k})(x)|^{p}(M \chi_{B}(x))^{\eta} {\rm d}x \\ & \leq& C \int_{{{\Bbb R}} ^{n}}|f_{k}(x)|^{p}(M \chi_{B}(x))^{\eta} {\rm d}x \\ & \leq& C 2^{-k n \eta} \int_{2^{k+1} B}|f_{k}(x)|^{p} {\rm d}x \\ & \leq& C 2^{-k n(\eta-\gamma / n)} 2^{-k \gamma} \int_{2^{k+1} B}|f_{k}(x)|^{p} {\rm d}x \\ & \leq& C 2^{-kn(\eta-\gamma / n)} r^{\gamma}\|f\|_{L^{p, \gamma}({{\Bbb R}} ^{n})}^{p}. \end{eqnarray*}

应用上面估计和Minkowski不等式, 可以得到(1.11)式.因此, 完成了定理1.3的证明.

参考文献

Aguilera N , Harboure E .

Some inequalities for maximal operators

Indiana Univ Math J, 1980, 29 (4): 559- 576

[本文引用: 1]

Bourgain J .

Pointwise ergodic theorems for arithmetic sets

Inst Hautes Études SCI Publ Math, 1989, 69 (10): 5- 41

[本文引用: 1]

Calderón A P , Zygmund A .

On a problem of Mihlin

Trans Amer Math Soc, 1955, 78: 209- 224

[本文引用: 3]

Calderón A P , Zygmund A .

Singular integral operators and differential equations

Amer J Math, 1957, 79: 901- 921

[本文引用: 2]

Calderón A P , Zygmund A .

On singular integrals with variable kernels

Applicable Anal, 1977/78, 7 (3): 221- 238

[本文引用: 4]

Chen Y , Ding Y , Li R .

L^{2} boundedness for commutators of rough singular integral with variable kernel

Rev Mat Iberoam, 2008, 24 (2): 531- 547

[本文引用: 1]

Chen Y , Ding Y , Hong G , Liu H .

Weighted jump and variational inequalities for rough operators

J Funct Anal, 2018, 274 (8): 2446- 2475

[本文引用: 4]

Christ M , Duoandikoetxea J , Rubio de Francia J L .

Maximal operators related to the Radon transform and the Calderón-Zygmund method of rotations

Duke Math J, 1986, 53 (1): 189- 209

[本文引用: 2]

Campbell J T , Jones R L , Reinhold K , Wierdl M .

Oscillation and variation for the Hilbert transform

Duke Math J, 2000, 105 (1): 59- 83

[本文引用: 2]

Campbell J T , Jones R L , Reinhold K , Wierdl M .

Oscillation and variation for singular integrals in higher dimensions

Trans Amer Math Soc, 2003, 355 (5): 2115- 2137

[本文引用: 2]

Cowling M , Mauceri G .

Inequalities for some maximal functions I

Trans Amer Math Soc, 1985, 287 (2): 431- 455

[本文引用: 2]

Ding Y , Hong G , Liu H .

Jump and variational inequalities for rough operators

J Fourier Anal Appl, 2017, 23 (3): 679- 711

[本文引用: 3]

Di Fazio G , Ragusa M A .

Interior estimates in Morrey spaces for strong solutions to nondivergence form equations with discontinuous coefficients

J Funct Anal, 1993, 112 (2): 241- 256

[本文引用: 2]

Gong R , Yan L .

Weighted L^{p} estimates for the area integral associated to self-adjoint operators

Manuscripta Math, 2014, 144 (1/2): 25- 49

[本文引用: 1]

Gong R , Yan L .

Littlewood-Paley and spectral multipliers on weighted L^{p} spaces

J Geom Anal, 2014, 24 (2): 873- 900

[本文引用: 1]

Hong G , Ma T .

Vector-valued q-variation for differential operators and semigroups I

Math Z, 2017, 286 (1/2): 89- 120

[本文引用: 1]

Jones R L , Seeger A , Wright J .

Strong variational and jump inequalities in harmonic analysis

Trans Amer Math Soc, 2008, 360 (12): 6711- 6742

[本文引用: 6]

Kurtz D .

Littlewood-Paley and multipliers theorems on weighted L^{p} spaces

Trans Amer Math Soc, 1980, 259 (1): 235- 254

Krause B , Zorin-Kranich P .

Weighted and vector-valued variational estimates for ergodic averages

Ergodic Theory Dynam Systems, 2018, 38 (1): 244- 256

[本文引用: 1]

Lépingle D .

La variation d'ordre p des semi-martingales

Z Wahrsch Verw Gebiete, 1976, 36 (4): 295- 316

[本文引用: 1]

Morrey C .

On the solutions of quasi-linear elliptic partial differential equations

Trans Amer Math Soc, 1938, 43 (1): 126- 166

[本文引用: 1]

Ma T , Torrea J L , Xu Q .

Weighted variation inequalities for differential operators and singular integrals

J Funct Anal, 2015, 268 (2): 376- 416

[本文引用: 3]

Ma T , Torrea J L , Xu Q .

Weighted variation inequalities for differential operators and singular integrals in higher dimensions

Sci China Math, 2017, 60 (8): 1419- 1442

[本文引用: 3]

Mas A , Tolsa X .

Variation for the Riesz transform and uniform rectifiability

J Eur Math Soc, 2014, 16 (11): 2267- 2321

Mirek M , Trojan B , Zorin-Kranich P .

Variational estimates for averages and truncated singular integrals along the prime numbers

Trans Amer Math Soc, 2017, 369 (8): 5403- 5423

[本文引用: 1]

Palagachev D , Softova L .

Singular integral operators, Morrey spaces and fine regularity of solutions to PDE's

Potential Anal, 2004, 20 (3): 237- 263

[本文引用: 1]

Pisier G , Xu Q .

The strong p-variation of martingales and orthogonal series

Prob Theory, 1988, 77 (4): 497- 514

[本文引用: 1]

Shen Z .

Boundary value problems in Morrey spaces for elliptic systems on Lipschitz domains

Amer J Math, 2003, 125 (5): 1079- 1115

[本文引用: 1]

Taylor M .

Analysis on Morrey spaces and applications to Navier-Stokes and other evolution equations

Comm Partial Differential Equations, 1992, 17 (9/10): 1407- 1456

[本文引用: 1]

Wu H , Yang D , Zhang J .

Oscillation and variation for the Riesz transform associated with Bessel operators

Proc Roy Soc Edinburgh Sect A, 2019, 149 (1): 169- 190

[本文引用: 1]

Wen Y , Wu H , Zhang J .

Weighted variation inequalities for singular integrals and commutators

J Math Anal Appl, 2020, 485 (2): 123825

[本文引用: 1]

/