Singular integrals and potential theory (Q931046): Difference between revisions

For a function \(F(x,h)\) on the upper half space \({\mathbb R}_+^{n+1}\), define \[ S^2(F)(x)=\int_{|x-y|<1}|F(y,h)|^2h^{-n-1}\,dy\,dh \] and \(F^*(x)=\sup_{|x-y|<1}|F(x,h)|\). For a smooth function \(A\) on \({\mathbb R}_+^{n+1}\), define \(\nabla A=(\frac{\partial}{\partial x_1}, \ldots, \frac{\partial}{\partial x_n}, \frac{\partial}{\partial h})\), and call \(A\in D_p\) \((0<p<\infty)\) if \(A^*,\,S(h\nabla A)\in L^p({\mathbb R}^n)\). The author gives the duality in \(L^p({\mathbb R}_+^{n+1})\) and \(D_{p'}\) via the Newtonian potential, relating the Dirichlet problem in a Lipschitz domain. To prove it, he gave the following: Let \(Q(x)\) be an appropriately smooth function on \(\mathbb R^n\) satisfying \(Q(x)=O((1+|x|)^{-n-\varepsilon}\) and \[ \nabla Q(x)=O(r^{-n+\varepsilon_1}(1+r)^{-\varepsilon_2-\varepsilon_1} \] for some \(\varepsilon>1/2, \varepsilon_1, \varepsilon_2>0\). Define \[ Q_{h,F}f(x)=\int h^{-n}Q((x-y)/h)F([a_1]_{x,y}, \ldots, [a_k]_{x,y})f(y)\,dy; \] \(f\in C_0^\infty\), where \(a_1,\ldots, a_k\in C_0^\infty\), \([a]_{x,y}=\int_{0}^{1}a(x+t(y-x))\,dt\), and \(F\in \mathbb C^N\) for sufficiently large \(N\). Let further \(\int Q\,dx=0\) and \(1<p\leq +\infty\). Then there exists \(C>0\) depending only on \(n\), \(p\), \(\|a_j\|_\infty\), \(\|F\|_{\mathbb C^N}\) and constants appearing in the estimate of \(Q\), such that \[ \left|\int\!\!\!\int Q_{h,F}f(y)A(y,h)\frac{dydh}{h}\right|\leq C\|f\|_p\|A\|_{D_{p'}}; \quad A\in C_0^\infty({\mathbb R}_+^{n+1}). \] As a corollary, he gave the \(L^p\) boundedness of the following singular integrals with rough kernels: Let \(\Omega\) be a homogeneous function of degree zero with \(\Omega\in L^r(S^{n-1})\) for some \(r>1\) and \(\int_{S^{n-1}}\Omega(y')d\sigma(y')=0\), \(F\), \(a_j\) be as above. Then the singular integrals \[ T_{\varepsilon.A}\int_{\varepsilon<|x-y|<A}\Omega((x-y)/|x-y|)|x-y|^{-n} F([a]_{x,y})f(y)\,dy \] are \(L^p\)-bounded \((1<p<\infty)\) uniformly in \(0<\varepsilon, A<\infty\). Some probabilistic applications also are given.