A T(P) theorem for Sobolev spaces on domains (Q2345406)

In this paper, the authors study necessary and sufficient conditions on certain singular integral operators to be bounded on the Sobolev space of a Lipschitz domain of \({\mathbb R}^d\). Let \(n\in\mathbb{N}\), \(p\in(1,\,\infty)\) and \(U\subset \mathbb{R}^d\) be any open set. The Sobolev space \(W^{n,\,p}(U)\) consists of all the functions \(f\) which have weak derivatives up to order \(n\) and \(D^{\alpha}f\in L^p(U)\) for all \(|\alpha|\leq n\). If \(D^{\alpha}f\in L^p_{\text{loc}}(U)\) for all \(|\alpha|\leq n\), \(f\) is said to belong to \(W^{n,\,p}_{\text{loc}}(U)\). A measurable function \(K\in W^{n,\,1}_{\text{loc}}({\mathbb R}^d\setminus \{0\})\) is said to be a smooth convolution Calderón-Zygmund kernel of order \(n\) if there exists a positive constant \(C_K\), depending on \(K\), such that, for any \(j\in\{0,\ldots,n\}\) and \(x\neq 0\), \[ |\nabla^j K(x)|\leq \frac{C_K}{|x|^{d+j}}, \] where \(|\nabla^j K|:=\sum_{|\alpha|=j}|D^\alpha K|\). The smooth convolution Calderón-Zygmund operator \(T\) of order \(n\) is defined by setting, for any \(f\in L^1_{\text{loc}}({\mathbb R}^d)\) and \(x\in{\mathbb R}^n\setminus \mathrm{\,supp\,}f\), \[ T(f)(x):=\int_{{\mathbb R}^d}K(x-y)f(y)\,dy. \] For a domain \(\Omega\subset\mathbb{R}^d\), the truncated operator \(T_\Omega\) is defined by \(T_\Omega(f):=\chi_\Omega T(\chi_\Omega f)\) for any \(f\in L^1_{\text{loc}}({\mathbb R}^d)\). In this paper, the authors first prove that, if \(\Omega\subset\mathbb{R}^d\) is a Lipschitz domain, \(T\) is a smooth convolution Calderón-Zygmund operator of order \(n\in\mathbb{N}\) and \(p\in(d,\,\infty)\), then the following two statements are equivalent: {\parindent=6mm \begin{itemize} \item[(i)] The truncated operator \(T_\Omega\) is bounded on \(W^{n,\,p}(\Omega)\). \item [(ii)] For every polynomial \(P\) of degree at most \(n-1\), \(T_\Omega(P)\in W^{n,\,p}(\Omega)\). \end{itemize}} For \(p\in(1,\,d]\), the authors find a sufficient condition in terms of Carleson measures and prove that for \(n=1\) this condition is also necessary. A positive and finite Borel measure \(\mu\) is said to be a \(p\)-Carleson measure if there exists a positive constant \(C\) such that, for any \(a\in\Omega\) which is close enough to the boundary \(\partial\Omega\), \[ \int_{\widetilde{\text{Sh}}(a)}[\text{dist}(x,\,\partial\Omega)]^{(d-p)(1-p')} \left[\mu(\text{Sh}(x)\cap\text{Sh}(a))\right]^{p'}\frac{dx} {[\text{dist}(x,\,\partial\Omega)]^d}\leq C\mu(\text{Sh}(a)), \] where \(\text{Sh}(x)\), \(\widetilde{\text{Sh}}(a)\) and \(\text{Sh}(a)\) are, respectively, the shadows of \(x\) and \(a\), which can be understood as Carleson boxes of \(\Omega\). To be precise, the authors show that, if \(\Omega\subset\mathbb{R}^d\) is a Lipschitz domain, \(T\) is a smooth convolution Calderón-Zygmund operator of order \(n\in\mathbb{N}\), \(p\in(1,\,d]\) and the measure \(|\nabla^n T_\Omega (P)(x)|^p\,dx\) is a \(p\)-Carleson measure for every polynomial \(P\) of degree at most \(n-1\), then \(T_\Omega\) is bounded on \(W^{n,\,p}(\Omega)\).