Heegaard splittings with large subsurface distances (Q611918)

\textit{M.~Scharlemann} and \textit{M.~Tomova} [Geom. Topol. 10, 593--617 (2006; Zbl 1128.57022)] showed that for a closed oriented \(3\)-manifold \(M\) and a Heegaard surface \(\Sigma\) of \(M\), if the Hempel distance \(d(\Sigma)\) of \(\Sigma\) satisfies \(d(\Sigma)>2g(\Sigma)\) where \(g(\Sigma)\) is the genus of \(\Sigma\), then any other Heegaard surface \(\Lambda\) of \(M\) satisfying \(d(\Sigma)>2g(\Lambda)\) is a stabilization of \(\Sigma\). \textit{K.~Hartshorn} [Pac. J. Math. 204, No.~1, 61--75 (2002; Zbl 1065.57021)] showed a similar result for the case that \(\Lambda\) is incompressible. The authors generalize the result as follows: For a closed oriented \(3\)-manifold \(M\), a Heegaard surface \(\Sigma\) of \(M\) with genus \(g(\Sigma)\geq 2\), and a compact essential subsurface \(F\) of \(\Sigma\) (i.e.\ no component of \(\partial F\) bounds a disk on \(\Sigma\) except the case \(F=\Sigma\setminus \{\text{open disk}\}\)), let \(d_F(\Sigma)\) be the \(F\)-distance (subsurface distance or relative Hempel distance) of \(\Sigma\) which is defined from the curve complex of \(\Sigma\) and the arc and curve complex of \((\Sigma, F)\). Then if another Heegaard surface \(\Lambda\) of \(M\) satisfies \[ d_F(\Sigma)>2g(\Lambda)+c(F), \] where \(c(F)=2\) if \(F\) is an annulus (i.e. \(2\)-holed sphere), a \(4\)-holed sphere, a \(1\)-holed torus or a \(2\)-holed torus, and \(c(F)=0\) otherwise, then \(\Lambda\) contains \(F\) as a subsurface up to isotopy (i.e. we may assume \(F\subset \Lambda\)). The main tool of the present paper is the Rubinstein-Scharlemann graphic [\textit{H.~Rubinstein} and \textit{M.~Scharlemann}, Topology 35, No.~4, 1005--1026 (1996; Zbl 0858.57020)]. In particular, the first author [\textit{J.~Johnson}, J. Topol. 1, No.~3, 671--692 (2008; Zbl 1148.57029)] introduced the notion that \(\Sigma\) is mostly above (mostly below, respectively) \(\Lambda\). The authors generalize it into the notion mostly above (mostly below, respectively) with respect to \(F\). Let \(f\) and \(h\) be Morse functions of \(M\) to \([-1, 1]\) with respect to \(\Sigma\) and \(\Lambda\). The Rubinstein-Scharlemann graphic is the image of the discriminant set \(\Delta\) of \(f\times h\) on \([-1, 1]\times [-1, 1]\), where slight modifications are allowed by [\textit{T.~Kobayashi} and \textit{O.~Saeki}, Pac. J. Math. 195, No.~1, 101--156 (2000; Zbl 1019.57010)]. Then \((-1, 1)\times (-1, 1)\) is divided into three connected subspaces: \[ \begin{aligned} R^F_a = \{(t, s)\;|\;\Sigma_t\text{ is mostly above }\Lambda_s\text{ with respect to }F\} \text{(red region)},\\ R^F_b = \{(t, s)\;|\;\Sigma_t\text{ is mostly below }\Lambda_s\text{ with respect to }F\} \text{(blue region)},\\ (-1, 1)\times (-1, 1)\setminus (R^F_a\cup R^F_b) \text{(white region)}, \end{aligned} \] where both \(R^F_a\) and \(R^F_b\) are disjoint open sets in \((-1, 1)\times (-1, 1)\) (Lemma 3.3) and where both \(\partial R^F_a\) and \(\partial R^F_b\) consist of a finite union of differentiable arcs. The authors introduce the notion that \(h\) \(F\)-spans (\(F\)-splits, respectively) \(f\). The map \(h\) \(F\)-spans \(f\) if there exists \(s\in (-1, 1)\) such that \((-1, 1)\times \{s\}\) intersects with both \(R^F_a\) and \(R^F_b\). The map \(h\) \(F\)-splits \(f\) if there exists \(s\in (-1, 1)\) such that \((-1, 1)\times \{s\}\) is disjoint from the closures of both \(R^F_a\) and \(R^F_b\). The notion ``\(h\) weakly \(F\)-splits \(f\)'' is the complementary one of ``\(h\) \(F\)-spans \(f\)''. Except for the special cases for \(F\), ``weakly \(F\)-splits'' implies ``\(F\)-splits'', and in the exceptional cases, the situation can be suitably controlled (Lemma 3.7), which corresponds to the term \(c(F)\) in the main theorem. The maps \(f\) and \(h\) in the notions above can be replaced with \(\Sigma\) and \(\Lambda\) respectively (i.e.\ to use \(f\) and \(h\) has technical purposes). If \(\Lambda\) \(F\)-spans \(\Sigma\), then \(\Lambda\) contains \(F\) as a subsurface up to isotopy (Proposition 4.1). Finally the authors show that if \(\Lambda\) weakly \(F\)-splits \(\Sigma\), then \[ d_F(\Sigma)\leq 2g(\Lambda)+c(F) \] (Proposition 6.1). By Lemma 3.7, Proposition 4.1 and Proposition 6.1, the proof is completed. All lemmas in Section 5 and 6 are for the proof of Proposition 6.1. Reviewer's remark: {\parindent6.5mm \begin{itemize}\item[(1)] If \(\Lambda=\Sigma\), then it is clear that \(\Sigma\) \(F\)-spans \(\Sigma\). Thus the assumption ``\(d(\Sigma)>2g(\Sigma)\)'' in the theorem of Scharlemann and Tomova is appropriate. \item[(2)] The theorem of Scharlemann and Tomova corresponds to the case that \(F=\Sigma\setminus \{\text{open disk}\}\). Then \(d_F(\Sigma)=d(\Sigma)\) and \(c(F)=0\). \item[(3)] I suspect that the authors would like to know about the converses of Proposition 4.1 and Proposition 6.1. They would lead to the characterization of stabilizations by the Rubinstein-Scharlemann graphic. \end{itemize}}