Systole and inradius of noncompact hyperbolic manifolds (Q2517169)

Let \(M\) be a complete hyperbolic manifold of dimension \(n\geq 2\). If \(M\) has finite volume then a natural problem is to study the relation between the many metric invariants of the hyperbolic structure on \(M\) and its volume. The present paper investigates in this regard the systole \(\mathrm{sys}(M)\) (smallest length of a closed geodesic loop on \(M\)) and the maximal radius \(\mathrm{R}(M)\) (maximal radius of an embedded ball in \(M\)). Obviously the maximal radius detemines a lower bound for the volume of the manifold; it is a consequence of Margulis' lemma that this is also true for the systole. These relations are not sharp (at least because the value of Margulis' constant is not known for \(n\geq 3\)). The present paper gives a sharp bound for the systole of a three--dimensional \textit{non-compact} manifold \(M\) of finite volume, namely \[ \cosh(\mathrm{sys}(M)/2) \leq \frac{1 + \sqrt{13}} 4 \cdot \frac{\mathrm{vol}(M)}{\mathrm{\mathrm{vol}}(\Delta)} \] (the term \(\mathrm{vol}(M)/\mathrm{\mathrm{vol}}(\Delta)\) on the right is the simplicial volume of \(M\): here \(\Delta\) is a regular ideal tetrahedron). Moreover it characterizes the equality case: it is attained if and only if the manifold is isometric to a specific manifold known as the Gieseking manifold, which is the result of gluing the faces of \(\Delta\) in the correct pattern, and is the three--manifold with cusps of smallest volume [\textit{C. C. Adams}, Proc. Am. Math. Soc. 100, 601--606 (1987; Zbl 0634.57008)]. In higher dimensions a similar (but likely not sharp) inequality is established, with the constant on the right tending to 0 as \(n\to+\infty\). For the maximal radius the paper establishes the inequality \[ \cosh \mathrm{R}(M) \geq \sqrt 5/2 \] for \(n\geq 3\), which is again sharp because of the Gieseking manifold (it is not known if the latter is the only manifold attaining equality in this case). The proof of this inequality is quite simple: the bound is obtained by computing the injectivity radius at a point of self-intersection for a horosphere. The proof of the systolic inequality and the characterisation of the equality case in dimension three are more involved and proceed roughly as follows. First the volume is bounded below in terms of the volume of a cuspidal end. It then remains to give an upper bound for the systole using a cusp and to compare it to this lower bound. This is achieved as follows: take a maximal cusp in \(M = \Gamma\backslash\mathbb H^3\), suppose that it lifts in \(\mathbb H^3\) to an horoball \(B_\infty\) around \(\infty\) with stabilizer \(\Gamma_\infty\) (so that the cusp is isometric to \(\Gamma_\infty\backslash B_\infty\)). Then the collection of \(\Gamma\)-translates of \(B_\infty\) contains a subfamily of horoballs tangent to \(\partial B_\infty\), which project on \(\partial B_\infty\) as a disk packing. The crucial observation (due to C. Adams, loc. cit.) is that there are at least two \(\Gamma_\infty\)-orbits of disks in this packing, and then one can use a \(\gamma\in\Gamma,\,\gamma\not\in\Gamma_\infty\) moving one of these orbits to another to estimate the systole (this is done differently according to whether \(\gamma\) is loxodromic, unipotent or orientation-reversing parabolic). The characterization of equality relies on delicate estimates in dimension three.