Tangled Platonic Polyhedra
Total Page:16
File Type:pdf, Size:1020Kb
Tangled Platonic Polyhedra Toen Castle, V. Robins and S. T. Hyde Last typeset on February 7, 2010 Abstract A method to find all embeddings of a given graph on the genus-one two-dimensional torus is used to construct a variety of entangled Platonic polyhedra graphs. These graph embeddings contain knotted and/or linked subgraphs of arbitrary complexity, the simplest of which include the ‘trefoil’ knot and the ‘Hopf’ and ‘Whitehead’ links. Analyses of knotting and linking, point group symmetries and an energy function are presented for embeddings of the topological tetrahedron, cube and octahedron. 1 Contents 1 Introduction 3 1.1 The Big Picture (surfaces in space) . 7 2 Method - generalising Platonic Polyhedra 10 2.1 Universal cover . 11 2.2 2-cell embeddings . 12 2.3 Tilings of the torus dissolving to polyhedron nets . 15 2.4 Tangling . 17 2.5 Knots and links . 18 2.6 Energy . 22 2.7 Symmetry . 22 3 Tetrahedron 25 3.1 2-cell embeddings of the tetrahedron . 26 3.2 The tetrahedron in the universal covers: <8, 4> and <9, 3> .... 30 3.3 Comparison of tetrahedral isotopes . 31 4 Cube 34 4.1 2-cell embeddings of the cube . 34 4.2 The cube in its universal cover . 35 4.3 Symmetries of tangled cubes . 39 4.4 High symmetry cube isotopes . 40 4.5 Honeycomb symmetries . 42 4.6 Cube isotope knots, links and energies . 43 5 Octahedron 49 5.1 2-cell embeddings of the octahedron . 49 5.2 The toroidal octahedron in its universal cover . 49 5.3 High symmetry toroidal octahedra isotopes . 52 5.4 Properties of toroidal octahedra . 52 6 Further constructions 60 6.1 Non- 2-cell torus graph embeddings . 60 6.2 Duals . 61 6.3 Chirality . 64 7 Conclusion 64 2 1 Introduction Polyhedra have long been of interest to mathematicians and philosophers, the fas- cination aided by the spatial, geometric, and combinatorial patterns commensurate with their simple structures. One notable early discussion of regular polyhedra was by Plato in his Timaeus [1]. He identified the ‘classical elements’ of fire, air, earth, water and quintessence or aether with the five regular polyhedra - tetrahedra, oc- trahedra, cubes, icosahedra and dodecahedra - as universal building blocks in his early ‘theory of everything’. Just as the understanding of the constitutive nature of elements has since grown, so too has the conception of polyhedra expanded. Polyhedra were initially envisaged as solid bodies with planar polygonal faces, such as used intrinsically by the ancient Egyptians in the design for their pyramids. The polyhedra now known as Platonic polyhedra due to the interest recorded in them by Plato, are defined by being regular, as each face is a congruent regular polygon, with the same number of faces meeting around each vertex. Later the concept had developed so that in the fifteenth century, Leonardo da Vinci famously illustrated Fra Luca Bartolomeo de Pacioli’s De divina proportione with an image of a rhombicuboctahedron that emphasised only the edges and vertices; omitting the faces and solid body (see Fig. 1). Keppler broadened the concept further by considering stellated polyhedra [2], before the idea was extended further still to infinite polyhedra [3] and myriad other still broader generalisations. Each gen- eralisation allows interesting new structures to be created as the principles of the Platonic polyhedra are relaxed or extended. Within the mathematical literature, the definitions of polyhedra have expanded to mirror this growing variety of structures. We follow the conventions of tiling theory, as described by Grunbaum¨ [5], in identifying the face of a polyhedron with the cyclically ordered set of vertices and connecting edges that surrounds it. Using this definition allows us to maintain the intuitive link with regular polyhedra while creating topologically interesting finite structures, embedded in oriented manifolds of genus greater than zero. This definition is suitably broad to allow a freedom of structure that is not constrained to convexity, planar faces, or topologically spheri- cal polyhedra. In turn, this freedom allows the presence of knots and links, as well as other entanglement modes [6] within the polyhedra. Interestingly, topologically toroidal polyhedra can be created within a definition more restrictive than ours, such as the dual pair of the Szilassi and Csasz´ ar´ polyhedra which have planar faces and straight edges [7, 8]. However, these requirements of straightness and linearity need to be dropped to allow a fuller variety of structures. Our method generates topologically distinct ways in which graphs (sets of vertices and edges with all edges joining pairs of vertices) can be embedded in space. Some of these ways are topologically trivial, whereas up to ambient isotopy 3 Figure 1: A ‘skeletal’ ployhedron is composed of just the edges and corners, and omits the faces and solid body. This illustration of a rhombicuboctahedron allows viewing of both the front and back elements[4]. 4 (stretching and deformations without passing edges through each other or them- selves) others are inherently knotted or interlinked. While the method is applicable to any graph, in each case considered in this paper, the graph will be that of a tetra- hedron, cube, or octahedron; but the embeddings of the constituent vertices and edges will vary between examples. If these graphs are considered in a chemical context – as atoms and bonds, or metal complexes and ligands – the results shed light on various topologically distinct conformational isomers that may be attain- able by the compound. As these isomers differ by their ambient isotopy, we follow the nomenclature of [9] in calling them isotopes. In literature regarding structural DNA manipulation [10] they are denoted topoisomers. A C D D B E F A G H G H B F E C Figure 2: A topologically simple cube and a tangled version. Both are shown reticulating underlying surfaces. There are different ways to quantify the ‘tangledness’ of a graph embedded in space. These methods are mostly imported from or inspired by knot theory: cycles within the graph embedding form loops which may be knotted or linked, and so can be described by knot and link tables, minimal crossing numbers, or knot polynomials etc. [11]. Indeed there has been some recent interest in the knots and links present in embedded graphs [12], however this work focuses on the knots and links unavoidably created in the embedding of sufficiently complex graphs, rather than the means of introducing controlled complexity into the embeddings of otherwise simple structures. An alternative to analysing the knots and links contained within an isotope is a generalisation of the knot polynomials called the Yamada polynomial [13] which provides an invariant (up to factors) of the entire embedded graph. This method 5 has been used to analyse the entanglement of the ‘theta graph’ – the graph with two vertices connected by three edges[14], however its utility for providing useful information about embeddings of larger graphs is unclear. Instead we prefer to use the hierarchy of graph embedding complexity provided by the minimum genus of the oriented surface necessary for a graph embedding as introduced in [9]. As the genus increases, so too does the complexity of knots [15], links and other entanglement modes [6] which manifest in the graph embedding. This hierarchy fits naturally with our method for generating these structures, which uses the 2-cell embeddings (see Section 2.2) of graphs to reticulate an ori- ented surface with the vertices and edges of the graph. A ‘reticulation’ of a surface indicates that the vertices and edges of a graph form a network on the surface, with edges only meeting or crossing at designated vertices. The use of 2-cell embed- dings is a common technique within graph theory [16]. It is used in [17] and with a different emphasis in [18]. As we only consider small (comparatively combinator- ically simple) graphs, we can omit many of the group theoretic aspects discussed in these works. As we are considering isotopes, within the equivalence class of ambient iso- topy, it suffices to work with a representative in a canonical form. As this paper focuses on the genus-one surface, we consider only the ‘doughnut’ manner of torus embedding in space. This avoids issues arising from self-interesection of the sur- face, and facilitates the analysis using its universal cover: an abstraction which represents the surface, irrespective of its specific shape in space. The universal cover of the torus is the plane, with a two-periodic translational symmetry echoing the longitudinal and meridional lines of the surface structure of the torus. The particular geometric placement of the points within the universal cover is determined to be the average of the placements of the connected points (barycen- tric). This gives a unique and canonical representation for a given isotopy class of graph surface embeddings [19]. It is also well-formed, in terms of points and lines having distinct placements, for all graphs we consider here. In the representation in the universal cover, the graph becomes a translational unit cell of a periodic re- peating pattern. In an alternative formulation, the surface can be cut open to form a surface patch reticulated with one unit of the repeating pattern in the universal cover. The manner that the patches fit together depends upon the periodicity of the universal cover. To find the actual placement of vertices, edges, and implicit faces in three-dimensional space, the surface is embedded in space in a suitable way, and then dissolved, leaving behind the points and lines, which can form a topologically tangled structure. While the isotopes in this paper generalise Platonic polyhedra by virtue of shar- ing their graphs, some can share another feature of Platonic polyhedra: equally 6 sized faces.