Skip to topic | Skip to bottom
Home
WildSurfaces
WildSurfaces.ThurstonSurfacer1.1 - 17 Feb 2006 - 17:38 - JoshuaBowmantopic end

Start of topic | Skip to actions
In his notes ``On the geometry and dynamics of diffeomorphisms of surfaces'', Thurston describes a method for constructing surfaces (generally half-translation surfaces) for whom a large subgroup of the Veech group, isomorphic to the free group on two generators, is known.

(See also: BouillabaisseSurfaces.)

Let $\Gamma$ and $\Delta$ be transverse (not necessarily maximal) multicurves on a compact orientable surface $S$, and assume that they fill the surface, i.e., that each component of their complement $S - (\Gamma \cup \Delta)$ is a simply connected polygon. Because $\Gamma$ and $\Delta$ are transverse to each other, the edges of the cylinders will also be transverse to each other, and will cut each other into rectangles. The corners of these rectangles meet inside the polygons that form the complement of $\Gamma$ and $\Delta$, and so the number of sides of the polygon determines what kind of singularity it will contain. We want to find appropriate side lengths of these rectangles.

Suppose $\Gamma = \{ \gamma_1, \dots, \gamma_m\}$ and $\Delta = \{ \delta_1,  \dots, \delta_n \}$, and let $M = [m_{ij}]$ be the $m \times n$ matrix such that $m_{ij}$ is the number of crossings of $\gamma_i$ and $\delta_j$. Thus $m_{ij}$ counts the number of rectangles comprising $S$ that contain a crossing of $\gamma_i$ and $\delta_j$. Now consider the matrix $M^\top M$; it is symmetric, which means all of its eigenvalues are real. We let $\lambda$ denote the largest eigenvalue of $M^\top M$. Choose an eigenvector $v$ of $M^\top M$ corresponding to $\lambda$ and having positive entries, and set $w = Mv$. The calculation

\[ MM^\top w = MM^\top Mv = M(\lambda v) = \lambda Mv = \lambda w \]
shows that $w$ is an eigenvector of $MM^\top$, also with eigenvalue $\lambda$. Now let each rectangle that contains an intersection of $\gamma_i$ and $\delta_j$ have width $v_j$ and height $w_i$. (We now assume that $\Gamma$ is horizontal and $\Delta$ is vertical with respect to the flat structure.) The following proposition shows that these coordinates are the correct ones.

Proposition. With respect to this flat structure, the Dehn twists around $\Gamma$ and $\Delta$ are affine automorphisms, and their derivatives have the forms

\[ D_\Gamma =  \left( \begin{array}{ c c } 1 & \lambda \\ 0 & 1 \end{array} \right) \ \ \mbox{and}\ \ D_\Delta =  \left( \begin{array}{ c c } 1 & 0 \\ -1 & 1 \end{array} \right). \]

These two Dehn twists are parabolic transformations. The group they generate contains other parabolic elements, which are conjugates of the generators; all other elements are hyperbolic (pseudo-Anosov).

Variants of this construction are possible. For example, each curve can be assigned a positive integral weight, so that the corresponding parabolic transformation consists of different multiples of a Dehn twist for each curve. One can also choose $w = aMv$ for some positive real number $a$; then $v = bM^\top w$, where $ab = 1/\lambda$. But this just amounts to applying $\mathrm{diag}(e^{-t},e^t)$, for some real $t$, to the surface described above.

-- JoshuaBowman - 17 Feb 2006
to top


You are here: WildSurfaces > ThurstonSurface

to top

Copyleft 2005-2009. Tout est libre ici! Everything is free here!
This project is not supported by any grant, organisation,...
Ideas, requests, problems regarding TWiki? Send feedback.