A sharp estimate for multilinear Marcinkiewicz integral operator (Q819596): Difference between revisions

Let \(m\in \mathbb N\), \(\Omega\) be a homogeneous function of degree zero, satisfying \(\Omega\in\text{Lip}_\gamma(S^{n-1})\) (\(0<\gamma\leq 1\)), \(\int_{S^{n-1}}\Omega(x)\,d\sigma(x)=0\), and \(A\) be a fuction satisfying \(\partial^\beta A\in\text{BMO}(\mathbb R^n)\) for \(|\beta|=m\). Let \(\mu_\Omega^A\) be the multilinear Marcinkiewicz operator defined by \[ \mu_\Omega^Af(x)=\bigl[\int_0^\infty|F_t^A(f)(x)|t^{-3}dt\bigr]^{1/2}, \] where \[ F_t^A(f)(x)=\int_{|x-y|\leq t}\frac{\Omega(x-y)}{|x-y|^{n-1}} \frac{R_{m+1}(A;x,y)}{|x-y|^m}f(y)dy, \] and \[ R_{m+1}(A;x,y)=A(x)-\sum_{|\beta|\leq m}\frac{1}{\beta!} \partial^\beta A(y)(x-y)^\beta. \] The author claims: For any \(0<r<p<1\), there exists \(C>0\) such that for any \(f\in C_0^\infty(\mathbb R^n)\) and any \(x\in \mathbb R^n\), \[ ([(|\mu_\Omega^A(f)|^r)^\#](x))^{1/r}\leq C\sum_{|\alpha|=m}\|\partial^\alpha A\|_{\text{BMO}}\bigl((M(|\mu_\Omega(f)|^p))^{1/p}(x)+M^2f(x)\bigr), \] where \(g^\#\) is the sharp function of \(g\), \(Mg\) is the Hardy-Littlewood maximal function of \(g\), and \(\mu_{\Omega}f(x)=\bigl[\int_0^\infty|F_t(f)(x)|t^{-3}dt\bigr]^{1/2}\), and \(F_t(f)(x)=\int_{|x-y|\leq t}\frac{\Omega(x-y)}{|x-y|^{n-1}}f(y)dy\). As consequenes, he states weighted estimates for these operators. However, the reviewer could not follow his proofs. I mention two gaps in his proofs. Firstly, the conclusion of Lemma 4 does not follow from the cited papers. Secondly, in the proof of Theorem 1 one cannot deduce \[ I(x)=\mu_\Omega\bigl(F_t(R_m(\widetilde A;x,\cdot)|x-\cdot|^{-m}f(\cdot))\bigr)= \| F_t\bigl(R_m(\widetilde A;x,\cdot)|x-\cdot|^{-m}f(\cdot)\bigr)(x)\| <+\infty \] in general. In fact, take a \(\varphi\in C_0^\infty(\mathbb R)\) satisfying \(\varphi(t)=1\) (\(|t|\leq 1\)), \(\varphi(t)=0\) (\(|t|\geq 2)\). Let \(M=\max\{\|\varphi^{(k)}(t)\|_\infty;\;k=1,2,3\}\), and set \(\Omega(x)=x_1/|x|\), \(A(x)=x_1^3\varphi(x_1)\), and \(Q=[-5M,5M]^n\), \(\widetilde Q=[-50M\sqrt n,50M\sqrt n]^n\). Then, \(\partial^\alpha A(x)=0\) for \(|\alpha|=3\) and \(\alpha\neq(3,0,\dots,0)\), \[ |\widetilde Q|^{-1}|\int_{\widetilde Q}\partial_{x_1}^3A(x)\,dx|\leq 4\cdot 86/(100\sqrt n)<6, \] and hence \((\partial^\alpha A)_{\widetilde Q}=|\widetilde Q|^{-1}\int_{\widetilde Q}\partial^\alpha A(x)\,dx=0\) for \(|\alpha|=3\) and \(\alpha\neq(3,0,\dots,0)\), and \(|(3!)^{-1}(\partial_{x_1}^3A)_{\widetilde Q}|<1\). Thus, \(\widetilde A(x)=A(x)-\sum_{|\alpha|=3}\frac{1}{\alpha!} (\partial^\alpha A)_{\widetilde Q}x^\alpha=c_1x_1^3\) \((|x|<1)\) for some \(c_1>0\). Hence \(R_3(\widetilde A;x,y)=c_1(x_1-y_1)^3\) for \(|x|,|y|<1/2\). Therefore, if \(f(x)\in C_0^\infty(\mathbb R^n)\) satisfies \(f(x)=1\) for \(|x|\leq 1\), we have for \(0<t<1/2\) and \(|x|<1/2\), \[ \begin{multlined} F_t\bigl(R_3(\widetilde A;x,\cdot)|x-\cdot|^{-3}f(\cdot)\bigr)(x) =\int_{|x-y|\leq t}\frac{\Omega(x-y)}{|x-y|^{n-1}} R_3(\widetilde A;x,y)|x-y|^{-3}f(y)\,dy=\\ =\int_{|x-y|\leq t}(x_1-y_1)|x-y|^{-n}c_1(x_1-y_1)^3|x-y|^{-3}dy=\\ =c_1\int_{|y|\leq t}y_1^4|y|^{-n-3}dy =c_1t\int_{|y|\leq 1}y_1^4|y|^{-n-3}dy=c_2t.\end{multlined} \] Hence, for \(|x|<1/2\) we have \[ \begin{multlined} I(x)=\| F_t\bigl(R_3(\widetilde A;x,\cdot)|x-\cdot|^{-3}f(\cdot)\bigr)(x)\|\geq\\ \geq \Bigl(\int_0^{1/2} |F_t\bigl(R_3(\tilde A;x,\cdot)|x-\cdot|^{-3}f(\cdot)\bigr)(x)|^2 t^{-3}dt\Bigr)^{1/2}=+\infty.\end{multlined} \] This implies the term \(I\) on the page 181 is infinite.