Symplectic coordinates on \(\mathrm{PSL}_3(\mathbb{R})\)-Hitchin components (Q2658465)

Let \(G\) be a real reductive Lie group and \(\Sigma_{n,g}\) an orientable, connected, compact surface of genus \(g\) with \(n\) boundary components. Let \(\Gamma_{n,g}=\pi_1(\Sigma_{n,g})\), and \(\mathrm{Hom}^*(\Gamma_{n,g},G)\) be the subspace of \(\mathrm{Hom}(\Gamma_{n,g},G)\) consisting of homomorphisms with closed conjugation orbits. The \(G\)-character variety of \(\Sigma_{n,g}\) is then the quotient space \(\mathfrak{X}_{n,g}(G)= \mathrm{Hom}^*(\Gamma_{n,g},G)/G\) by the conjugation action of \(G\) on \(\mathrm{Hom}(\Gamma_{n,g},G)\). The smooth locus of \(\mathfrak{X}_{n,g}(G)\) is a Poisson manifold that is symplectic if and only if \(n=0\), with symplectic leaves defined by fixing the conjugation classes of homomorphisms at boundary curves when \(n>0\). This follows by [\textit{W. Goldman}, Adv. Math. 54, 200--225 (1984; Zbl 0574.32032)] for \(n=0\) and [\textit{K. Guruprasad} et al., Duke Math. J. 89, No. 2, 377--412 (1997; Zbl 0885.58011)] for \(n>0\). The Poisson bracket is described in general in [\textit{W. Goldman}, Invent. Math. 85, 263--302 (1986; Zbl 0619.58021)] in the closed surface case and by the reviewer [Trans. Am. Math. Soc. 361, No. 5, 2397--2429 (2009; Zbl 1171.53052)] when \(n>0\). Let \(g\geq 2\). The subspace of conjugation classes of discrete faithful homomorphisms in \(\mathfrak{X}_{0,g}(\mathrm{PSL}(2,\mathbb{R}))\) is naturally homeomorphic to the space of complex structures on \(\Sigma_{0,g}\) (called \textit{Teichmüller space}), which in turn is equivalent (via uniformization) to the space of hyperbolic structures on \(\Sigma_{0,g}\) (called \textit{Fricke space}). We will denote it by \(\mathfrak{T}_g\). There is an irreducible representation \(\sigma_{m}:\mathrm{PSL}(2,\mathbb{R})\to \mathrm{PSL}(m,\mathbb{R})\) given by the \((m-1)\)-symmetric tensor. Define \(\mathrm{Hit}_{m}(\Sigma_{0,g})=\{\sigma_m\circ\rho\ |\ \rho\in \mathfrak{T}_g\}\); called the \textit{Hitchin component}. In [\textit{N. Hitchin}, Topology 31, No. 3, 449--473 (1992; Zbl 0769.32008)], it is shown that \(\mathrm{Hit}_{m}(\Sigma_{0,g})\cong \mathbb{R}^{(m^2-1)(2g-2)}\). In the case of \(m=3\), in [\textit{S. Choi} and \textit{W. Goldman}, Bull. Am. Math. Soc., New Ser. 34, No. 2, 161--171 (1997; Zbl 0866.57001)], Goldman and the first named author of this paper show \(\mathrm{Hit}_{3}(\Sigma_{0,g})\) corresponds to the space of convex real projective structures on \(\Sigma_{0,g}\), generalizing Fricke space. In [\textit{F. Labourie}, Invent. Math. 165, No. 1, 51--114 (2006; Zbl 1103.32007)], Labourie shows that \(\mathrm{Hit}_{m}(\Sigma_{0,g})\) consists of discrete and faithful homomorphisms for all \(m\) (generalizing the cases of \(m=2,3\)) and contributing to the advent of ``higher Teichmüller theory''. In [Am. J. Math. 107, 969--997 (1985; Zbl 0578.32039)], \textit{S. Wolpert} shows that Fenchel-Nielsen coordinates are Darboux for \(m=2\), and in [\textit{H. Kim}, J. Differ. Geom. 53, No. 2, 359-401 (1999; Zbl 1040.53093)] the third named author of this paper shows that Goldman's coordinates parametrizing \(\mathrm{Hit}_{3}(\Sigma_{0,g})\), kin to Fenchel-Nielsen coordinates, are likewise Darboux. The authors of the paper under review, which includes Kim, claim that the proof of this fact is incomplete. One of the goals of this paper, in the words of the authors, is to ``fill in the missing links and make the proof of H. Kim more complete and clear.'' More precisely, let \(\mathcal{P}=\{P_1,\dots,P_{2g-2}\}\) be a pants decomposition of \(\Sigma_{0,g}\), which gives a collection of disjoint, non-isotopic, oriented, essential, simple closed curves \(\alpha_1,\dots,\alpha_{3g-3}\). In [\textit{W. Goldman}, J. Differ. Geom. 31, No. 3, 791--845 (1990; Zbl 0711.53033)] it is shown that \(\mathrm{Hit}_{3}(\Sigma_{0,g})\) can be parametrized by three sets of coordinates: (1) \textit{internal} parameters \(\{s_1,t_1,\dots,s_{2g-2},t_{2g-2}\}\) with each pair \((s_i,t_i)\) corresponding to coordinates on the \(2\) dimensional moduli space of convex real projective structures on \(P_i\) with fixed holonomy on the boundary curves, (2) \textit{length} parameters \(\{\ell_1,m_1,\dots,\ell_{3g-3},m_{3g-3}\}\) where each pair \((\ell_i,m_i)\) corresponds to horizontal and vertical translation length (Hilbert metric) along the curve \(\alpha_i\), and (3) \textit{twist-bulge} parameters \(\{u_1,v_1,\dots,u_{3g-3},v_{3g-3}\}\) where each pair \((u_i,v_i)\) corresponds to ``twisting'' or ``bulging'' along the curve \(\alpha_i\). We note that the reviewer, in the case of a pair-of-pants, wrote out the correspondence between the more general trace coordinates on the character variety and the internal parameters \((s,t)\) in [\textit{S. Lawton}, Thesis (Ph.D.). University of Maryland, College Park (2006)]. In the above terms, one of the main theorems in this interesting and well-written paper takes half the internal parameters (the \(s_i\)'s) and all of the length parameters (making \(8g-8\) parameters) and shows that there exist additional \(8g-8\) parameters such that this new collection of \(16g-16\) parameters is a global Darboux coordinate system for \(\mathrm{Hit}_{3}(\Sigma_{0,g})\). The short version of the proof is this: They show that the \(8g-8\) parameters \[\{s_1,\dots,s_{2g-2},\ell_1,m_1,\dots,\ell_{3g-3},m_{3g-3}\}\] define a Lagrangian fiber bundle \(\mathrm{Hit}_{3}(\Sigma_{0,g})\to \mathbb{R}^{8g-8}\) satisfying some general conditions: (a) the base has trivial second cohomology, (b) the fibres are simply connected Lagrangian submanifolds, and (c) the Hamiltonian vector fields of the coordinate functions defining the fiber bundle are everywhere linearly independent and complete. The result then follows from a general symplectic geometry theorem (attributed to Duistermaat) that says that under such conditions the coordinates functions defining the fiber bundle map can be completed to a global Darboux coordinate system. The other main theorem in the paper under review concerns \(\Sigma_{n,g}\) when \(n\geq 0\) and is true for general \(G=\mathrm{PSL}(m,\mathbb{R})\). In that case there is an analogue of \(\mathrm{Hit}_{m}(\Sigma_{0,g})\) defined by considering homomorphisms with fixed boundary classes (of loxodromic type). Denote this space by \(\mathrm{Hit}^{\mathcal{B}}_{m}(\Sigma_{n,g}).\) The theorem says that if one decomposes \(\Sigma_{n,g}\) into a finite collection of hyperbolic subsurfaces (by removing an appropriate set of simple closed curves), fixes the conjugation classes of homomorphisms in \(\mathrm{Hit}^{\mathcal{B}}_{m}(\Sigma_{n,g})\) at the removed curves, and takes the quotient of this subspace by a natural Hamiltonian action, then the resulting quotient space is symplectomorphic to the cartesian product of the Hitchin spaces of the hyperbolic subsurfaces.