Curvature estimates for minimal hypersurfaces via generalized longitude functions (Q488120): Difference between revisions

It is said that an open subset \(V\) of a Riemannian manifold is a convex supporting set if every compact subset of \(V\) supports a strictly convex function. It is shown by \textit{J. Jost} et al. [J. Differ. Geom. 90, No. 1, 131--176 (2012; Zbl 1250.53061)] that the complement of a half-equator \(S^n\setminus \overline{S}_+^{n-1}\) is a maximal convex supporting set of \(S^n\). Combining the construction of convex functions and some analytic technologies, they showed the smoothness of a weakly harmonic map \(u: M \to S^n\) whose image is contained in a compact subset of \(S^n\setminus \overline{S}_+^{n-1}\). They also proved a Bernstein type theorem by using this regularity and the well-known Ruh-Vilms theorem in [\textit{E. A. Ruh} and \textit{J. Vilms}, Trans. Am. Math. Soc. 149, 569--573 (1970; Zbl 0199.56102)]. In this paper, the author considers a generalization of the Bernstein type theorem mentioned above by finding a generalized longitude function \(\theta\) whose level sets are totally geodesic on some specified convex supporting sets of spheres. Given a weakly harmonic map \(u\) into \(S^n\), the composition of \(\theta\) and \(u\) satisfies an elliptic equation of divergence type. With aid of Harnack's inequalities, the author establishes the image shrinking property and then regularity results. And applying these results to the Gauss image of minimal hypersurfaces in Euclidean space, the author obtains curvature estimates and a corresponding Bernstein type theorem. Let \(M\subset {\mathbb R}^{n+1}\) be an embedded complete minimal hypersurface with Euclidean volume growth. Assume that there is a point \(y_0 \in M\) such that the following Neumann-Poincaré inequality \[ \int_{B_R(y_0)}|v- {\bar v}_R| \leq CR^2 \int_{B_R(y_0)} |\nabla v|^2 \] for any \(v \in C^\infty(B_R(y_0))\) holds with a positive constant \(C\) not depending on \(R\). Under the hypotheses, the author proves that if the Gauss image of \(M\) is contained in \(S^n\setminus \overline{S}_+^{n-1}\), and \[ \sup_{B_R(y_0)} d(\cdot, S^{n-2})^{-1}\circ \gamma = \text{o}(\log \log R), \] then \(M\) has to be an affine hyperplane. Here \(\bar v\) is the average value of \(v\) on \(B_R(y_0)\) and \(\gamma \) is the Gauss map.