Moduli space of meromorphic differentials with marked horizontal separatrices (Q2663322)

A (compact) translation surface is a pair \((X, \omega)\), where \(X\) is a (compact) Riemann surface and \(\omega\) is a holomorphic 1-form on the surface. Locally integrating the form defines a flat metric on the surface, with conical singularities. If one only asks the form \(\omega\) to be \emph{meromorphic}, obtaining a non-compact translation surface with infinite area (if the poles of \(\omega\) are not all simple). The study of such objects is motivated by the fact that they appear naturally when dealing with compactifications of the moduli space of translation surfaces. More precisely, if a sequence \((X_n,\omega_n)\) converges to the boundary of the moduli space in the Deligne-Mumford compactification, then the thick components of \(X_n\), appropriately rescaled, converge to meromorphic differentials, see, e.g., [\textit{A. Eskin} et al., Publ. Math., Inst. Hautes Étud. Sci. 120, 207--333 (2014; Zbl 1305.32007)]. In this article, the author studies the topology of the moduli spaces of translation surfaces with poles equipped with an extra combinatorial data: the choice, for each singularity of (an equivalence class of) horizontal separatrix, denoted \(\mathcal{H}^{\text{hor}}\). Here, by horizontal separatrix we mean either an horizontal geodesic line ending (or beginning) at a conical singularity or an horizontal geodesic going to infinity in the flat metric if the singularity is a non simple pole. The main result of the paper is a complete characterization of the connected components of \(\mathcal{H}^{\text{hor}}\). Similarly to the case of compact translation surfaces, proven in [\textit{M. Kontsevich} and \textit{A. Zorich}, Invent. Math. 153, No. 3, 631--678 (2003; Zbl 1087.32010)], in the general case of genus greater than 1 and underlining surfaces not belonging to the hyperelliptic component, there are at most 2 connected components, classified but an invariant \(\operatorname{Sp}\) which is a generalization of the classical \(\operatorname{Arf}\). In the hyperelliptic case the extra symmetry of the underlining surface yields more components. Finally, the genus-0 case is the most complicated one, depending also on the combinatorics of the singularities. The main result is obtained by reducing the problem to the study of some cyclic group coming for the forgetful map from \(\mathcal{H}^{\text{hor}}\) to \(\mathcal{H}^{\text{sing}}\), which is the space of translation surfaces with poles in which singularities have given names. The latter space has the same connected components of the moduli space of translation with poles, which were classified by the author in [Comment. Math. Helv. 90, No. 2, 255--286 (2015; Zbl 1323.30060)]. Generalizing two local surgeries introduced in the article by Kontsevich and Zorich [loc. cit.], called \emph{breaking up a zero} and \emph{bubbling a handle}, so that they can be performed also on poles, one constructs some important elements in the cyclic groups coming from the covering. Using some topological analysis, one then proceeds to show that these elements always exist. If the above elements generate the whole cyclic group, then the corresponding moduli space is connected. If they generate a finite index subgroup, then the index gives the number of connected components. Depending on the genus and on whether or not we are in the hyperelliptic case, components are distinguished by some topological invariant, related to the classical \(\operatorname{Arf}\) invariant and to the parity of the spin structure.