Singular integral and maximal operators associated to hypersurfaces: \(L^p\) theory (Q2574450)

Given a real valued function \(\Psi\in C^{\infty}(\mathbb R^N)\) with \(\Psi(0)=0\), the authors consider the operators, associated to the hypersurface \(y_{N+1}=\Psi(y)\), defined by \[ M_{\Psi}f(x, x_{N+1})=\sup_{0<h<1}{1\over h^N} \int_{| y| \leq h} | f(x-y, x_{N+1}-\Psi(y)) | dy, \] and \[ H_{\Psi}f(x, x_{N+1})=p.v. \int_{| y| \leq 1} f(x-y, x_{N+1}-\Psi(y)) K(y) dy, \] with \(K\) a Calderón-Zygmund kernel, homogeneous of degree \(-N\) and with mean value zero over the unit sphere. It is well known that if a principal curvature of the hypersurface defined by \(\Psi\) does not vanish to infinite order at the origin, then both operators are bounded on \(L^p\) with \(1<p<\infty\) and a basic question treated in this paper is to study them when the curvature condition fails. The authors are mainly interested in the case when \(\Psi\) is a convex function on \(\mathbb R^N\), and, in particular, if \(\Psi(y)=\phi(| y| )\) defines a radial hypersurface it is known that without the convexity condition on \(\phi\), both operators can be unbounded on \(L^p\) for non trivial ranges of \(p\). However, \(H_\Psi\) is bounded on \(L^2\) for every measurable \(\phi\) as long as \(N\geq 2\) (these results can be found in some earlier works of the authors). The present paper continues exploring to what extend the above facts for radial hypersurfaces are true for more general ones, and the main purpose of this paper is to study the case when \(\Psi(x)=\phi(G(x))\) where \(G\) is a general convex function of finite type at the origin such that \(G(0)=\nabla G(0)=0\). In this setting, they consider \[ E_l=\{v\in{\mathbb {R^N}}; \;G(sv)=O(s^{l+1}) \text{ for \;small } s>0\} \] and they prove that under the above condition on \(G\), if \(l_0\) is the smallest positive integer such that \(E_l\neq \mathbb {R^N}\), the codimension of \(E_{l_0}\) is at least 2 and the function \(\gamma(s):=\phi(s^{l_0})\) is convex, then both operators are bounded in \(L^p\) for every \(1<p<\infty\). Moreover, if the codimension of \(E_{l_0}\) is 1, there are no positive results, in general, and they construct some counterexamples.