Rank two affine submanifolds in \(\mathcal H(2,2)\) and \(\mathcal H(3,1)\) (Q321678)

The recent \textit{magic wand} theorem of \textit{A. Eskin} and \textit{M. Mirzakhani} [``Invariant and stationary measures for the \(\mathrm{SL}(2, \mathbb R)\) action on moduli space'', Preprint, \url{arXiv:1302.3320}] shows that all \(\mathrm{GL}(2, \mathbb{R})\)-orbit closures in \(\Omega_g\), the moduli space of holomorphic differentials on compact genus-\(g\) Riemann surfaces, are \textit{affine submanifolds} in period coordinates (local coordinates modeled on the homology relative w.r.t. the zeros of the differential). Understanding what submanifolds occur is an important open question with applications to billiards and low-dimensional dynamics. In [ibid. 19, No. 1, 413--438 (2015; Zbl 1318.32021)], \textit{A. Wright} introduced the notion of \textit{rank} of an affine manifold \(M\), half the dimension of the image of the tangent space \(TM\) projected to the absolute cohomology. Rank-\(1\) manifolds parametrize loci of branched cover of tori. Mirzakhani conjectured that if the rank of \(M\) is at least 2 and the \textit{field of definition} (the smallest real field in which the coefficients of the local systems of linear equations defining \(M\) lie) is \(\mathbb Q\), then \(M\) is either the entire stratum or a covering of a stratum of abelian or quadratic differentials. The paper under review completes the classification of rank-2 affine manifolds in \(\Omega_3\), which is stratified by the orders of the zeros of differentials, and consists of 3 strata \(H(4)\), \(H(2, 2)\), \(H(3, 1)\). \(H(3, 1)\) is connected, and contains no rank-2 manifolds. \(H(4)\) and \(H(2, 2)\) each contain two connected components. Earlier work of the authors and Wright classified the possibilities in \(H(4)\), and the current paper deals with case of \(H(2, 2)\). The techniques in this paper hold promise for analyzing the problem in higher genera.