On commutators of vector BMO functions and multilinear singular integrals with non-smooth kernels (Q986573)

To study multilinear singular integrals, the authors consider a class of integral operators \(\{A_t\}_{t>0}\), which are a sort of approximations of the identity: \(A_tf(x)=\int_{\mathbb R^n}a_t(x,y)f(y)dy\), and \(a_t(x,y)\) satisfy the condition \[ |a_t(x,y)|\leq h_t(x-y)=t^{-n/s}h(|x-y|/t^{1/s}), \tag \(*\) \] where \(s\) is a positive constant and \(h\) is a positive, bounded and decreasing function satisfying \(\lim_{r\to\infty}r^{(n+\eta)/s}h(r)=0\) for some \(\eta>0\). Let \(T:\mathcal S(\mathbb R^n)\times\cdots\times\mathcal S(\mathbb R^n)\mapsto \mathcal S'(\mathbb R^n)\) be an \(m\)-linear operator which has a kernel function \(K(x,y_1,\ldots,y_m)\in L_{\roman{loc}}^1((\mathbb R^n)^m\setminus \{x=y_1=\cdots=y_m\})\) satisfying \[ \big\langle T(f_1,\ldots,f_m),g\big\rangle= \int_{\mathbb R^n} \int_{(\mathbb R^n)^m} K(x,y_1,\dots,y_m) f_1(y_1)\cdots f_m(y_m)g(x)\,dy_1\cdots dy_m\,dx \] for \(f_1,\dots, f_m, g\in{\mathcal S}(\mathbb R^n)\) with \(\bigcap_{j=1}^m \operatorname {supp}f_j\cap \operatorname {supp}g=\emptyset\). Since \(T\) is linear in every entry, it has \(m\) formal transposes \(T^{*,j}\), whose kernels \(K^{*,j}\) satisfy \(K^{*,j}(x,y_1,\dots,y_j,\dots,y_m)=K(y_j,y_1,\dots,x,\dots,y_m)\). The authors consider the following two conditions on \(K\) and \(K^{*,k}\). \(({\mathbf H}_1)\) There exist operators \(\{A_t^{(j)}\}_{t>0}\) with kernels \(a_t^{(j)}(x,y)\) satisfying the above condition \((*)\) and there exist kernels \(K_t^{*,k,(j)}(x,y_1, \dots,y_m)\), a function \(\varphi\in C(\mathbb R)\) with \(\operatorname{supp}\varphi \subset [-1,1]\) and positive constants \(\varepsilon\), such that \[ \begin{multlined}\big\langle T^{*,k}(f_1,\dots,A_t^{(j)}f_j,\dots,f_m),g\big\rangle\\ = \int_{\mathbb R^n} \int_{(\mathbb R^n)^m} K_t^{*,k(j)}(x,y_1,\dots,y_m)f_1(y_1)\cdots f_m(y_m)g(x)\,dy_1\cdots dy_m\,dx \end{multlined} \] for \(f_1,\dots, f_m, g\in\mathcal S(\mathbb R^n)\) with \(\bigcap_{j=1}^m \operatorname{supp}f_j\cap \operatorname{supp}g=\emptyset\), and \[ \begin{multlined}\big|K^{*,k}(x,y_1,\dots,y_m)-K_t^{*,k,(j)}(x,y_1,\dots,y_m)\big|\\ \leq A(|x-y_1|+\cdots+|x-y_m|)^{-mn}\sum_{k=1,k\neq j}^{m}\varphi(|y_j-y_k|/t^{1/s}) +At^{\varepsilon/s}(|x-y_1|+\cdots+|x-y_m|)^{-mn-\varepsilon} \end{multlined} \] whenever \(t^{1/s}\leq|x-y_j|/2\). \(({\mathbf H}_2)\) There exist operators \(\{B_t\}_{t>0}\) with kernels \(b_t(x,y)\) satisfying the above condition \((*)\) and a function \(\varphi\in C(\mathbb R)\) with \(\operatorname{supp}\varphi\subset [-1,1]\) and positive constants \(\varepsilon, A\) such that for \(K_t^{(0)}(x,y_1,\dots,y_m) =\int_{\mathbb R^n}K_(z,y_1,\dots,y_m)b_t(x,z)\,dz\), it holds \[ \big|K_t^{(0)}(x,y_1,\dots,y_m)\big|\leq A(|x-y_1|+\cdots+|x-y_m|)^{-mn} \] whenever \(2t^{1/s}\leq \min_{1\leq j\leq m}|x-y_j|\), and \[ \begin{multlined} \big|K(x,y_1,\ldots,y_m)-K_t^{(0)}(x,y_1,\dots,y_m)\big|\\ \leq A(|x-y_1|+\cdots+|x-y_m|)^{-mn}\sum_{k=1,k\neq j}^{m}\varphi(|y_j-y_k|/t^{1/s})\\ +At^{\varepsilon/s}(|x-y_1|+\cdots+|x-y_m|)^{-mn-\varepsilon} \end{multlined} \] whenever \(2|x-x'|\leq t^{1/s}\) and \(2t^{1/s}\leq \min_{1\leq j\leq m}|x-y_j|\). The authors' main theorem is the following: Let \(T\) be an \(m\)-linear operator with generalized Calderón-Zygmund kernel \(K\) satisfying \(({\mathbf H}_1)\) and \(({\mathbf H}_2)\). Assume that for some \(1\leq q_1,\dots,q_m<\infty\) and some \(0<q<\infty\) with \(1/q=1/q_1+\cdots+1/q_m\), \(T\) maps \(L^{q_1}(\mathbb R^n)\times\cdots\times L^{q_m}(\mathbb R^n)\) to \(L^{q,\infty}(\mathbb R^n)\). Then, for \(\vec b=(b_1,\dots,b_m)\) with \(b_j\in{\text{BMO}}\) \((j=1,\dots,m)\), the commutator \([T,\vec b]\) of \(T\) and \(\vec b\) is bounded from \(\prod_{j=1}^m L^{p_j}(M\omega_j)\) to \(L^p(\nu_{\vec\omega})\) provided \(1<p_j<\infty\) and \(1/p=1/p_1+\cdots+ p_m\), where \(M\omega_j\) is the Hardy-Littlewood maximal function of weight \(\omega_j\), and \(\nu_{\vec \omega}=\prod_{j=1}^m \omega_j^{p/p_j}\). To show this, they give certain pointwise estimates on the Fefferman-Stein sharp maximal function for \(T\) and \([T,\vec b]\), and also weighted estimates involving maximal function associated with \(L(\log L)\). They show that Calderón's \(m\)-th commutator satisfies the assumptions in the above result, but it does not satisfy the standard Calderón-Zygmund kernel regularity conditions in Grafakos-Torres' sense.