Deforming convex real projective structures (Q684220)

A marked complete convex projective structure on a surface \(S\) can be viewed as a diffeomorphism from \(S\) to a quotient \(\Gamma \backslash \Omega\) where \(\Omega\) is a convex domain in the real projective plane \(\mathbb{RP}^2\) and \(\Gamma \cong \pi_1(S)\) is a discrete subgroup in \(\mathrm{PGL}(3, \mathbb R)\) which acts freely on \(\Omega\). Examples are hyperbolic structures, where \(\Gamma\) lies inside \(\mathrm{PO}(2, 1)\) and hence preserves the negative cone, but there are many more: in fact if \(g\) is the genus of \(S\), the Teichmüller space \(\mathrm{Hyp}(S)\) (which parametrises marked hyperbolic structures) is of dimension \(6g - 6\) while the space \(\mathcal C(S)\) of convex projective structures has dimension \(16g - 16\). The paper under review is a contribution to the global study of \(\mathcal C(S)\). A well-known fact in Teichmüller geometry is that there is a family of natural actions of \(\mathbb R^{3g - 3}\) on \(\mathrm{Hyp}(S)\) (equivalently, \(3g-3\) commuting flows), one for each pants decomposition of \(S\), given by translating along the twist parameters for the decomposition. The orbits of this action are an important feature of the large-scale geometry of \(\mathrm{Hyp}(S)\) (see for example [\textit{A. Eskin} et al., Duke Math. J. 166, No. 8, 1517--1572 (2017; Zbl 1373.32012)]). Such commuting twist flows also exist for convex projective structures but they account only for \(6g-6\) dimensions, while half the dimension equals \(8g - 8\). In this paper the authors define for each pants decomposition two additional flows (``bulging flows'') commuting with each other and with the twist flows, thus achieving this goal. The definition is technical but we can attempt to give an idea of how it works: each pair of pants admits an ideal triangulation by two ideal triangles. The gluing of their sides (known as ``shear coordinates'' in hyperbolic geometry) can be viewed as the remaining parameters: in Teichmüller space they account for the boundary lengths of the pants, and a generalisation to convex projective geometry is due to F. Bonahon and G. Dreyer. Not all flows along these parameters commute with the twist flows (in Teichmüller space none does), but the authors show that it possible to change additional parameters while commuting with these flows which account for \(2g - 2\) additional commuting flows in total. Most of the paper is dedicated to describing these parameters, following Fock-Goncharov and Bonahon-Dreyer. The holonomy of convex projective structures on \(S\), as a subset of the character variety of \(\pi_1(S)\) with values in \(\mathrm{PGL}_3(\mathbb R)\), form the so-called ``Hitchin component'' which is also defined for higher-dimensional representation spaces. The authors announce further work of theirs with Z. Sun where they prove analogous results in this generality.