Thin Riemannian manifolds with boundary (Q1392393): Difference between revisions

In this article, the authors prove that there is a dimension independent constant \(c\) \((\geq 0.075)\) such that, if \(M\) is a complete connected Riemannian manifold with sectional curvature \(K\) in the interior, then: (1) if \(M\) has boundary \(B\) with normal curvature \(k\) and the inradius of \(M\) satisfies \(\text{inr}^{2}\max \{\sup \left| K\right| ,{\sup k}^{2}\}<c\), then \(M\) is diffeomorphic to the product of a manifold without boundary and an interval, or \(M\) can be doubly covered by such a product. (A statement similar to this one was given by Gromov in his address to the International Congress in 1978, but with a dimension dependent constant [see \textit{M. Gromov}, Proc. Int. Congr. Math., Helsinki, 1978, Vol. 1, 415-420 (1980; Zbl 0427.53018)]. (2) if \(M\) has no boundary, \(\left| K\right| \leq 1\), \(H_{1}(M;\mathbb{Z}/2\mathbb{Z})=0\), and \(H\) is a smooth connected hypersurface properly imbedded in \(M\) with normal curvature at most \(1\), then the closure of each component of \(M-H\) has inradius at least \(c\). If \(M\) is a complete connected Riemannian manifold with boundary \(B\), \(K\) the sectional curvature of the interior, and \(k\) the normal curvature of \(B\), the inradius of \(M\) is defined by: \(\text{inr}= \sup \{d(p,B):p\in M\}\). A manifold \(M\) with boundary is thin if \(\text{inr}^{2}\max \{\sup \left| K\right| ,\sup k^{2}\}\) is small. The theorem is proved by showing that if \(c\) is sufficiently small, then every cut point of \(B\) has degree 2. (\(p\) is a cut point of \(B\) if there is a geodesic segment \(\sigma \) realizing \(d(p,B)\), but no extension of \(\sigma \) realizes distance to \(B\); the degree of \(p\) is the number of such minimizing segments, the endpoints on \(B\) of these segments are the footpoints of \(p\)). The authors prove that there is an increasing sequence of positive constants \(c(k)\), \(k=2,3,\dots\), such that if the curvature-normalized inradius of \(M\) is less than \(c(k)\), then the cut points of \(B\) have degree at most \(k\). This is done showing that in the boundary of the lift, the footpoints of \(p\) have injectivity radius uniformly bounded below, and for \(c\) sufficiently small, the corresponding normal coordinate neighbourhoods subtend large disjoint cones with vertex \(p\). The methods used are those of comparison geometry and use previous work by the authors [see \textit{S. B. Alexander, I. D. Berg} and \textit{R. L. Bishop}, Trans. Am. Math. Soc. 339, 703-716 (1993; Zbl 0793.53039); \textit{S. B. Alexander} and \textit{R. L. Bishop}, Differ. Geom. Appl. 6, 67-86 (1996; Zbl 0859.53053)]. The comparison arguments to a model manifold with boundary and the relative comparison lemmas presented in Section 2 have independent interest for the study of manifolds with boundary.