A uniform Poincaré estimate for quadratic differentials on closed surfaces (Q2349158)

Let \((M, c)\) be a closed Riemann surface of genus \(\geq 2\), where \(c\) is a complex structure on the underlying surface \(M\) and \(\mathcal{Q}(M, c)\) the complex vector space of smooth quadratic differentials on \((M, c)\). Using the unique hyperbolic metric \(g=\rho^2dzd\bar{z}\) compatible with \((M, c)\), define an \(L^2\)-inner product on \(\mathcal{Q}(M, c)\) by \[ \langle \phi dz^2,\psi dz^2\rangle=4\int_M \phi\cdot \bar{\psi}\cdot \rho^{-2} \frac{i}{2}dz\wedge d\bar{z}. \] Let \(\mathcal{H}(M, c)\) be the closed subspace of \(\mathcal{Q}(M, c)\) of holomorphic quadratic differentials, and let \(P_g: \mathcal{Q}(M,c)\to \mathcal{H}(M, c)\) be the orthogonal projection. The main purpose of this paper is to establish the following uniform Poincaré estimate: for all \(\Psi\in \mathcal{Q}(M,c),\) \[ \|\Psi-P_g(\Psi)\|_{L^1(M,g)}\leq C\|\bar{\partial}\Psi\|_{L^1(M,g)}, \] where \(C\) is a constant depending only on the genus of \(M\). This inequality has applications to the Teichmüller harmonic map flow. The outline of the proof is as follows. If the inequality is not true, then there exist Riemann surfaces \((M,c_i)\) and quadratic differentials \(\Phi_i\in \mathcal{Q}(M,c_i)\) such that \(P_{g_i}(\Phi_i)=0\), \(\|\Phi_i\|_{L^1(M,g_i)}=1\) and \(\|\bar{\partial}\Phi_i\|_{L^1(M,g_i)}\to 0\) as \(i\to\infty\), where \(g_i\) is the hyperbolic metric on \((M,c_i)\). Then, by the results in [\textit{M. Rupflin} et al., Adv. Math. 244, 874--893 (2013; Zbl 1288.32018)], a subsequence of \((M,c_i)\) converges to a Riemann surface with nodes, and \(\Phi_i\) to a holomorphic quadratic differential, which must be identically zero on the limit surface. However this contradicts the fact that the \(L^1\)-norm of the limit quadratic differential is equal to \(1\), which the authors prove in the paper. The key part of the paper is the estimate in Lemma 2.3: for all \(\Psi\in\mathcal{Q}(M,g)\) with \(P_g(\Psi)=0\), \[ \|\Psi\|_{L^1(\delta\mathrm{-thin}(M,g))}\leq C\Big(\|\bar{\partial}\Psi\|_{L^1(M,g)}+\delta^{1/2}\|\Psi\|_{L^1(M,g)}\Big), \] where \(\delta\)-\(\mathrm{thin}(M,g)\) is the \(\delta\)-thin part of a hyperbolic surface \((M,g)\) and the constant \(C\) depends only on the genus of \(M\). This estimate ensures that the \(L^1\) norm of a quadratic differential which is orthogonal and very close to \(\mathcal{H}(M,c)\) cannot concentrate on a long collar about a short closed geodesic curve, and hence the norm will not vanish when the collar is degenerating. Lemma 2.3 follows from two other lemmas and they involve several ingenious calculations.