Counterexamples to the \(L^p\)-Calderón-Zygmund estimate on open manifolds (Q2291465)

Given a number \(p\in [1,\infty]\), a Riemannian manifold \((\mathcal{M},g)\) is said to obey the \(L^p\)-Calderón-Zygmund property \(\mathrm{CZ}(p)\) if there exist constants \(C_1\) and \(C_2\) such that for all smooth and compactly supported functions \(f\) on \(\mathcal{M}\) one has \[ \|\nabla_g\nabla_g f\|_{L^p}\leq C_1\|\Delta_gf\|_{L^p}+C_2\|f\|_{L^p}, \] where the 2-tensor field \(\nabla_g\nabla_g f\) denotes the Hessian of \(f\) and its trace under the metric \(g\) is the Laplace-Beltrami operator \(\Delta_gf\). In the simplest case of the Euclidean space \(\mathbb{R}^m\), \(\mathrm{CZ}(p)\) was proved by Calderón and Zygmund in their seminal paper [\textit{A. P. Calderón} and \textit{A. Zygmund}, Acta Math. 88, 85--139 (1952; Zbl 0047.10201)] for all \(p\in (1,\infty)\) with constants \(C_1\) and \(C_2\) depending only on \(m\) and \(p\), while \(\mathrm{CZ}(p)\) fails to hold in the limit cases \(p=1\) and \(p=\infty\) (see [\textit{D. Ornstein}, Arch. Ration. Mech. Anal. 11, 40--49 (1962; Zbl 0106.29602)] and [\textit{C. T. McMullen}, Geom. Funct. Anal. 8, 304--314 (1998; Zbl 0941.37030)]). In the case of a Riemannian manifold \((\mathcal{M},g)\), \(\mathrm{CZ}(p)\) has been proved under suitable geometric assumptions on the manifold such as boundedness of the Ricci or the sectional curvatures, boundedness of the injectivity radius away from zero, or validity of the doubling property for the Riemannian volume measure. Recently, a two-dimensional complete manifold \((\mathcal{M},g)\) has been constructed in [\textit{B. Güneysu} and \textit{S. Pigola}, Adv. Math. 281, 353--393 (2015; Zbl 1320.53043)] for which \(\mathrm{CZ}(2)\) is false. In the very interesting paper under review, the author proves that for any integer \(m\geq2\) there exists an \(m\)-dimensional Riemannian manifold \(\mathcal{M}\) on which the Calderón-Zygmund property \(\mathrm{CZ}(p)\) fails for all \(p\in[1,\infty]\). The construction of the manifold relies on special warped product metrics in \(\mathbb{R}^m\), the properties of the Green function of the Laplace-Beltrami operator on \(\mathcal{M}\) and some crucial arguments from [Güneysu and Pigola, loc. cit.].