Thick-thin decomposition for quadratic differentials (Q2459320): Difference between revisions

A quadratic differential \(q\) on a Riemann surface \(S\) defines a singular Euclidean metric on \(S\), while by the uniformization theorem, the surface \(S\) is also equipped with a hyperbolic metric \(\sigma\). The problem of comparing the two metrics has been studied extensively in recent years. In the paper under review, the author investigates the same problem on a thick part of \(S\), that is, the part on which the diameter and the injectivity radius are both bounded above and below by constants depending only on the topology of \(S\). Let \(Y\subset S\) be a thick piece. Consider the \(q\)-lengths \(l_{q}(\alpha)\) for all nonperipheral simple closed curves \(\alpha\) on \(Y\). If \(\alpha_Y\) is one of the simple closed curves on \(Y\) such that \(l_{q}(\alpha)\) takes the smallest value, then \(l_{q}(\alpha_Y)\) is called the size of \(Y\) and is denoted by \(\lambda(Y)\) or simply by \(\lambda\) if \(Y\) is fixed. In what follows the notation \(x\prec y\) means that there exists a constant \(c\) depending on the topology of \(S\) such that \(cx\geq y\); and the notation \(x\asymp y\) stands for \(y/C\leq x\leq Cy\) for a constant \(C\). Let \(l_{\sigma}(\alpha)\) denote the \(\sigma\)-length of \(\alpha\). The author proves the following theorem: For each thick piece \(Y\subset S\), let \(\lambda\) denote the size of \(Y\). Then for every nonperipheral simple closed curve \(\alpha\subset Y\), \[ l_q(\alpha)\asymp \lambda\;l_{\sigma}(\alpha).\tag{TH} \] To prove the theorem, the author first recalls a result of \textit{Y. Minsky} [J. Differ. Geom. 35, No. 1, 151--217 (1992; Zbl 0763.53042)] which claims that \[ i(\alpha, \mu)\asymp l_{\sigma}(\alpha) \] for each nonperipheral simple closed curve \(\alpha\subset Y\), where \(i(\alpha,\mu)\) denotes the geometric intersection number of \(\alpha\) with the short marking \(\mu\) that is defined as follows. Take the shortest \(\sigma\)-shortest pants decomposition \(\mathcal{P}\) of \(Y\), and then for each curve in \(\mathcal{P}\) one takes a transverse curve with the shortest \(\sigma\)-length. The set of these transverse curves is denoted by \(\mathcal{Q}\). Let \(\mu=\mathcal{P} \cup \mathcal{Q}\). It is easily seen that \(\mu\) fills \(Y\) in the sense that all the components of \(Y-\mu\) are either disks or annuli sharing the boundary components with \(Y\). Then he uses the Minsky theorem to prove the main theorem. More precisely, the fact that \[ \lambda\;l_{\sigma}(\alpha) \prec l_q(\alpha)\tag{LLK} \] follows from Lemma 3 (which states that the \(q\)-lengths of curves in a \(\sigma\)-short marking \(\mu\) of \(Y\) are comparable to \(\lambda\)), Lemma 5 (which says \[ l_q(\alpha)\;l_q(\beta)\succ \lambda^2\;i(\alpha,\beta) \] for any nonperipheral simple closed curves \(\alpha\) and \(\beta\)), and the Minsky theorem. To prove that \[ \lambda\;l_{\sigma}(\alpha)\succ l_q(\alpha),\tag{LK} \] the author gives the estimations for the area and the diameter of \(Y\) with respect to the \(q\) metric (Theorem 4). That is, \[ \text{diam}_q(Y) \asymp \lambda \;\;\text{and}\;\;\text{area}_q(Y) \asymp \lambda^2.\tag{LLP} \] From (\textbf{LLP}) the author is able to retrace the intersection pattern of \(\alpha\) with \(\mu\) and produce a curve in \(Y\) in the homotopy class of \(\alpha\) whose length is bounded by a multiple of \(\lambda\;\sum_{\beta\in \mu}i(\alpha,\beta)\). It follows that \(\lambda\;l_{\sigma}(\alpha) \succ l_q(\alpha)\). By combining (\textbf{LLK}) and (\textbf{LK}), Theorem \textbf{TH} holds.