Delaunay partitions (Q1925108)

The Teichmüller space \(\mathcal{T}( p,n) \) can be viewed as a space of metrics, up to scale factor and isotopy, on a closed oriented surface \(M_{p,n}=M_{p}\setminus S_{n}\) of genus \(p\) with \(n\) punctures, \( S_{n}\) being the set of punctures. The author considers flat metrics having cone singularities at the punctures with a fixed set of cone angles \(\theta >0\) satisfying the cone angle condition \(\sum_{S_{n}}\theta \left( s\right) =\left( 4p-4+2n\right) \pi \). A cone metric determines a unique cell decomposition of \(M_{p,n}\), called its Delaunay partition, with vertex set \( S_{n}\). Every \(2\)-cell \(F\) of this partition is isometric to a planar polygon inscribed in a circle. For a fixed \(\theta \), each \(\tau \in \mathcal{T}( p,n) \) determines an isotopy class of Delaunay partitions \(\mathcal{E}( \theta ,\tau) \in \Lambda( p,n) \) where \(\Lambda( p,n) \) is the set of all isotopy classes of all cell decompositions of \(M_{p,n}\). Let \(\widehat{C}( \mathcal{E},\theta) =\left\{ \tau |\mathcal{E}( \theta ,\tau) =\mathcal{E}\right\} \subset \mathcal{T}( p,n) \) and let \( \widehat{\mathcal{C}}( \theta) =\left\{ \widehat{C}( \mathcal{E},\theta) |\mathcal{E\in }\Lambda( p,n) \right\} \) be the resulting partition of \(\mathcal{T}( p,n) \). In Theorem 1.1, the pair \(( \mathcal{T}( p,n) ,\widehat{ \mathcal{C}}( \theta)) \) is described in elementary terms: the author constructs a homeomorphism \(h:\mathcal{R}( \theta) \rightarrow \mathcal{T}( p,n) \) such that \(\widehat{ \mathcal{C}}( \theta) =h( \mathcal{D}( \theta)) \). Here \(\mathcal{R}( \theta) \) is a certain quotient space of a locally compact space \(\mathcal{R}_{0}( \theta) =\bigcup_{\mathcal{E\in }\Lambda ( p,n) }\left\{ \mathcal{E}\right\} \times P( \mathcal{E},\theta) \), and \(\mathcal{D}( \theta) \) is its cell decomposition. The space \(P( \mathcal{E},\theta) \) is defined combinatorially as a subset of a certain relatively compact convex polyhedron in \(\mathbb{R} ^{E}\), \(E\) being the set of \(1\)-cells of \(\mathcal{E}\). In the sequel, the author uses his results from [\textit{W. A. Veech}, Am. J. Math. 115, No.~3, 589-689 (1993; Zbl 0803.30037)] to identify the set \(\mathcal{S}( p,n) \) of all ''admissible'' affine complex structures on \(M_{p,n}\) up to isotopy, with the product \(l^{+}( p,n) \times \mathcal{T}( p,n) \times G^{+}( p,n) \), where \(l^{+}( p,n) \) is the set of all \(\theta \) satisfying the cone angle condition, and \(G^{+}( p,n) \) is the positive character group of \(\pi _{1}( M_{p,n}) \). He then proves that each \(( \theta ,\tau ,v) \in \mathcal{S}( p,n) \) admits a unique minimal Delaunay partition. This is again given by \(\mathcal{E\in }\Lambda ( p,n) \) with each open \(2\)-cell affine complex equivalent to an inscribed polygon. If \(\theta \in l^{+}( p,n) \) and \(v\in G^{+}( p,n) \) are fixed, there arises a partition \(\widehat{\mathcal{C}}( \theta ,v) \) of \(\mathcal{T}( p,n) \) just as in the cone metric case, and the author proves in Theorem 1.2 a certain generalization of Theorem 1.1. A\ consequence of these theorems is that sets \(\widehat{ \mathcal{C}}( \theta) \) and \(\widehat{\mathcal{C}}( \theta ,v) \) are convex polyhedra, which was established previously in [\textit{I. Rivin}, Ann. Math., II. Ser. 139, No.~3, 553-580 (1994; Zbl 0823.52009)].