Hyperbolic surfaces with sublinearly many systoles that fill (Q2214030)

The main result of this paper is succinctly described in the title: there exists a sequence of hyperbolic surfaces \(X_n\), with genus going to infinity, such that the set of systoles (closed geodesics of shortest length) on \(X_n\) fills \(X_n\) (i.e., the complement is contractible), and such that the number of systoles on \(X_n\) grows more slowly than any linear function of the genus. This work fits into an interesting circle of ideas related to understanding spines of moduli space \(\mathcal{M}_g\). A spine of a space is a subset of smaller dimension onto which the whole space deformation retracts. It is known that the virtual cohomological dimension of \(\mathcal{M}_g\) is \(4g-5\), which gives a lower bound on the dimension of any spine. No spine of this dimension is known. A candidate natural geometric spine was explored by Thurston in an unpublished preprint. He considers the locus of surfaces whose systoles fill the surface. His argument that this is a spine, which involves constructing a flow that increases lengths of the systoles on surfaces away from this locus, is considered incomplete by experts in the field. The work under review can be thought of as an effort towards computing the dimension of Thurston's locus. One hopes that the existence of surfaces with a small set \(S\) of filling systoles should imply that the dimension of the Thurston locus is large (since near such a surface, the locus is cut out by equations stipulating that the lengths of the curves in \(S\) remain equal). However, without further information about the derivatives of the various length functions, this idea cannot be made into a proof. Nevertheless, by applying the main result and then deforming the surface, the author is able to give a proof that the Thurston locus has dimension at least \(4g-5\), and he conjectures that for any \(\varepsilon>0\), there is a genus \(g\) for which the locus has dimension at least \((6- \varepsilon)g\). The proof of the main result is achieved via a nice construction that starts by taking right-angled regular hyperbolic polygons and gluing them together according to certain highly symmetric graphs, which the author calls flag-transitive maps. This yields surfaces with boundary, called ``blocks'', for which the systoles are exactly the boundary curves. These blocks are then glued together along their boundaries via a different class of graphs, with the key property being high girth. Because of the symmetries of the construction, the systoles of the resulting closed surfaces can be identified without the need for intricate hyperbolic geometry arguments. By choosing the parameters in the construction appropriately, one can achieve a small collection of filling systoles.