Professional Documents
Culture Documents
Combiantorias
24 al 28 de junio de 2017
iii
iv
Contents
1 Historical Background 1
2 Abstract Polytopes 7
2.1 Partially ordered sets . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Abstract polytopes . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Homomorphisms and anti-homomorphisms . . . . . . . . . . . 19
4 Symmetries of polytopes 27
4.1 Regular polytopes . . . . . . . . . . . . . . . . . . . . . . . . . 32
Bibliography 35
v
vi CONTENTS
Chapter 1
Historical Background
The greeks were the first to study the symmetries of polyhedra. Euclid, in
his Elements showed that there are only five regular solids (in Figure 1.1).
In this context, a polyhedron is meant to be a convex polyhedron and it is
regular if all its polygons are regular and equal, and all its vertices have the
same degree.
1
2 CHAPTER 1. HISTORICAL BACKGROUND
Figure 1.2: A regular convex pentagon and a pentagram. Both have 5 ver-
tices, and the pentagram has 5 “false” vertices, that is, points that are the
intersection of two edges, but that are not vertices.
It was until 1619 that Kepler found the other two regular polyhedra: the
great dodecahedron and the great icosahedron (Figure 1.3). To do so, he
allowed “false” vertices and edges, that is, the intersection of the (convex)
faces of the polyhedra at points that need not be vertices nor edges, just as
the pentagram allows intersection of edges at points that are not vertices of
the polygon. The great dodecahedron has regular pentagons as faces and
at each vertex five of them meet in the form of a pentagram. Similarly, the
great icosahedron has equilateral triangles as faces and five of them around
each vertex, arranged in form of a pentagram. In this way, the vertex-figure
of these two polyhedra are pentagrams (see Figure 1.2).
A couple of years later Cauchy showed that these are the only four regular
“star” polyhedra. We note that the convex hull of the great dodecahedron,
great icosahedron and small stellated dodecahedron is the icosahedron, while
the convex hull of the great stellated dodecahedron is the dodecahedron.
In the early 1920’s, Petrie realized that regular polyhedra can be con-
structed organizing the faces going “up and down”. He found two infinite
such polyhedra: the first one (shown in Figure 1.5) has square faces, six
around each vertex, its dual has regular hexagon as faces, four around each
vertex; Coxeter found a third one (having regular hexagonal faces with six
around each vertex) and showed there were no more. These three (infinite)
polyhedra are known as the Petrie-Coxeter polyhedra.
Grünbaum found 47 regular polyhedra in the seventies, including all the
ones described above. Looking at the Petrie-Coxeter polyhedra, one can
observe that the faces are arranged, around each vertex, in a skew manner.
In the Euclidean three space there are 4 different types: convex, star, skew
and helicoidal. While convex and star are always finite, skew polygons can
be either finite or infinite and helicoidal ones are always infinite. (Figure 1.6
shows a convex, a star and a skew 8-gon.) Hence a regular polyhedron can
have, in principle, faces of any of these types and the arrangement of edges
around each vertex can be of any of the finite types.
In the early 80’s, Dress found a new regular polyhedron and showed that
there are 48 regular polyhedra in Euclidean three space, 18 of which are
finite. We refer the interested reader to [4, 5] for details of the classification
of regular polyhedra in Euclidean 3-space.
The Platonic Solids can be thought as tessellations of the sphere. By
using polygons as faces (in a combinatorial sense), one can obtain “polyhe-
4 CHAPTER 1. HISTORICAL BACKGROUND
dra” (or maps) in different surfaces, that need not be realizable in Euclidean
space. More precisely, a map is a 2-cell embedding of a connected graph
(or multigraph) on a surface. The study of regular maps was initiated by
Brahana [1] and developed by Coxeter [2].
Schläfli, at the end of the XIX century, discovered regular convex poly-
topes and honeycombs in higher dimensions. In the classical theory, the
facets and vertex-figures of higher dimensional polytopes are spherical. In
1978, Grünbaum [6] proposed the study of a more general class of polytopes,
in which facets and vertex-figures might not be spherical. Using Coxeter and
Tits ideas about how polytopes can be seen as combinatorial objects and the
inclusion relation between its elements, Danzer and Schulte [3, 7, 8] extended
this further and set out the basic theory of combinatorial objects which are
now known as abstract polytopes.
6 CHAPTER 1. HISTORICAL BACKGROUND
Chapter 2
Abstract Polytopes
When dealing with abstract polytopes we always try to keep in mind the
geometry behind them. Given a “classical” polytope P, we consider it as
the set containing all its vertices, edges, 2-faces, . . . , facets, together with ≤
relation between them, given by the incidence. Such relation is easily seen to
be a partial order on all the “faces”. Abstract polytopes are defined as posets
with certain properties that resemble classical polytopes. For example, since
a vertex cannot be in a polygon of the polytope without being in an edge,
and two vertices cannot be incident to each other, then the maximal chains of
the order have one element of each kind. A diagram of how we would see the
tetrahedron in this way is shown in Figure 2.1. When replacing each vertex,
7
8 CHAPTER 2. ABSTRACT POLYTOPES
edge and face of this diagram by a vertex, we obtain the Hasse diagram of
the face-lattice of the tetrahedron.
Before going into the definition of abstract polytopes and start studying
them, we shall briefly review some basic facts about posets.
F = F0 , F1 , F2 , . . . , Fk = G,
Figure 2.2: Minimal and maximal faces of a tetrahedron, and a flag of it.
Figure 2.2 shows a way in which we can see the minimal and maximal
face of the tetrahedron and has one flag of it in the partial order. As we
shall see later, the minimal and maximal face of a polytope are required for
technical reasons. Since a flag is a maximal chain of the order, together with
the minimal and maximal faces, it contains a vertex, an edge, a polygon, etc.,
in such a way that the vertex is incident to the edge, the edge is incident to
10 CHAPTER 2. ABSTRACT POLYTOPES
We also require that P be strongly connected, that is, every open interval
of P having elements of at least two different ranks is connected (as a poset).
Finally, we require a homogeneity property (often called the diamond condi-
tion), namely, whenever F ≤ G, with rank(F ) = j − 1 and rank(G) = j + 1,
there are exactly two faces H of rank j such that F ≤ H ≤ G. Figure 2.4
represents the diamond condition on a tetrahedron (for j = 0, 1, 2) in a geo-
metric way. Figure 2.5 shows the same thing in the Hasse diagram.
3. strong connectivity,
Figure 2.5: The diamond condition on the Hasse diagram of the tetrahedron
shape (see Figure 2.6), and we can think of it as an edge with its two end-
vertices.
If P is a 2-polytope, it is easy to see that the number of vertices and edges
of P are exactly the same. Furthermore, every vertex is incident to exactly
two edges and every edge is incident to exactly two vertices (see Figure 2.7).
For this reason, a 2-polytope is called a polygon, or if it is finite and has p
vertices (and hence also p edges), a p-gon. Note that every two geometric
p-gons are combinatorially equivalent. For example, a convex pentagon and
a pentagram (in Figure 1.2) are different representations of the same abstract
polytope, the 5-gon (in Figure 2.7).
Given two faces F and G of a polytope P such that F ≤ G, the section
G/F of P is the closed interval {H|F ≤ H ≤ G}. Often, and motivated by
12 CHAPTER 2. ABSTRACT POLYTOPES
F2
Edges
Vertices
F-1
the phenomena in convex polytopes, we identify the section G/F−1 with the
face G. If F0 is a vertex, then the section Fn /F0 is called the vertex-figure of
F0 , and more generally Fn /F is the co-face at F .
Two flags are said to be adjacent if they differ by exactly one face. Fig-
ure 2.8 has a geometric view of a flag and its 0-adjacent, in two different
ways, and Figure 2.9 shows the Hasse diagram of the tetrahedron empha-
sizing a flag and its 1-adjacent flag. The diamond condition tells us that
Figure 2.9: A flag on the Hasse diagram of the tetrahedron and its 1-adjacent
one.
1. (Φi )i = Φ.
1. Flag connected.
2. Strong connected.
a) Prove that v ≤ e.
Exercise 12 For each of the two preceding constructions, find the number
of flags and number of k-faces of the output.
Exercise 13 Consider all the geometric structures that are polytopes of Fig-
ure 2.10. For each of them, find its Petrial and decide if it is a polytope or
not. Do the same with the Platonic Solids.
2.3. HOMOMORPHISMS AND ANTI-HOMOMORPHISMS 19
Figure 2.12: A Petrie polygon of the cube and the Petrial (or Petrie-dual) of
the cube.
It is not difficult to see that the dual of a cube is an octahedron and the
dual of a dodecahedron is an icosahedron (Figure 2.13 implicitly defines the
anti-isomorphims between a dodecahedron and an icosahedron).
Exercise 15 Show that a convex pentagon and a pentagram (see Figure 1.2)
are isomorphic.
Exercise 18 If we take a cube and identify opposite vertices, edges and faces,
we obtain a hemi-cube. The hemi-cube can be represented in the projective
plane as in Figure 2.14.
a) P yr(P) ∼
= P yr(Q) if and only if P ∼
= Q.
b) (P ri(P)) ∼
= P ri(Q) if and only if P ∼
= Q.
c) (P yr(P))∗ ∼
= P yr(P) if and only if P ∼
= P ∗.
Chapter 3
Flag graph and equivelar
polytopes
Let P be an n-polytope. We define the flag graph GP of P as follows. The
vertices of GP are the flags of P, and we put an edge between two of them
whenever the corresponding flags are adjacent. Hence GP is n-valent (i.e.
every vertex of GP has exactly n incident edges; to reduce confusion we avoid
the alternative terminology ‘n-regular’). Furthermore, we can colour the
edges of GP with n different colours as determined by the adjacencies of the
flags of P. That is, an edge of GP has colour i, if the corresponding flags of
P are i-adjacent. In this way GP is an n-edge coloured graph. Hence, every
vertex of has exactly one edge of each colour (see Figure 3.1). Note that the
edges of a given colour form a perfect matching of the graph.
Figure 3.1: The flag graph of a cube. Green, blue and yellow segments
represent colours 0, 1 and 2, respectively.
23
24 CHAPTER 3. FLAG GRAPH AND EQUIVELAR POLYTOPES
Let i ∈ {0, . . . , n − 1} and consider GPî as the graph GP taken away all the
edges of colour i. Let Φ ∈ F(P), and consider Φ as a vertex of GP . Then,
the connected component of GPî that contains Φ, GPî (Φ), corresponds of all
the flags of P that can be written as Φi0 ,i1 ,...,ik for some integer k such that
ij ∈ {0, . . . , n − 1} \ {i}. In other words, the flags in GPî (Φ) are precisely all
the flags of P that have (Φ)i as its i-face. Hence, we can think of GPî (Φ) as
the i-face of P contained in Φ.
Consider now the (i, j) 2-factor of GP , that is, the subgraph of GP that
has all the vertices of GP but only the edges of colours i and j, and denote
it by GPi,j . If |i − j| > 1, the fact that for any flag Φ, Φi,j = Φj,i implies
that the connected components of GPi,j are always 4-cycles. However, if |i −
j| = 1, then the connected components of GPi,j are either (even) cycles with
alternating colours or infinite alternating paths of the graph. Moreover,
fixing i and j, the connected components of GPi,j need not be isomorphic,
for example a cuboctahedron (see Figure 3.2) has two types of connected
components of colours 0 and 1 (hexagons and octagons). Let Φ ∈ F(P)
and consider the flags that are in the same connected component of GPi,i+1 as
Φ. Each such flag can be written as Φi0 ,i1 ,...,ik for some integer k such that
ij ∈ {i, i + 1}. That is, the vertices of the connected component of Φ in GPi,i+1
are Φ, Φi , Φi,i+1 , Φi,i+1,i , . . . together with Φi+1 , Φi+1,i , . . . Note that all these
flags contain the faces (Φ)j , whenever j 6= i, i + 1. Hence, the connected
component of Φ in GPi,i+1 can be thought of as the section (Φ)i+2 /(Φ)i−1 of
P.
An n-polytope P, n ≥ 2, is said to be equivelar if, for each j = 1, . . . , n−1,
25
there exists an integer pj , such that, for each flag Φ ∈ F(P), the section
(Φ)j+1 /(Φ)j−2 is a pj -gon. In other words, P is equivelar if for each j ∈
{0, . . . , n − 2} the connected components of the (j, j + 1) 2-factor of GP are
all isomorphic (to a 2pj bi-coloured cycle). In this case, we say that P has
Schläfli type (or sometimes only type) {p1 , p2 , . . . pn−1 }.
All 2-polytopes are equivelar; furthermore, a p-gon has Schläfli type {p}.
We note that an infinite 2-polytope or aperiogon has Schläfli type {∞}.
Alternatively, the Schläfli type of an equivelar polytope can be defined
as follows. For rank 2, we say that the Schläfli type of a p-gon is {p}. For
higher rank, an n-polytope P is said to have Schläfli type {p1 , p2 , . . . , pn−1 }
if all its facets have Schläfli type {p1 , p2 , . . . , pn−2 } and all its vertex-figures
have Schläfli type {p2 , p3 , . . . , pn−1 }.
A polyhedron is equivelar of Schläfli type {p, q} if and only if all its facets
are p-gons and all its vertex-figures are q-gons. Hence, all the Platonic Solids
are equivelar: the tetrahedron has Schläfli type {3, 3}, the cube has type
{4, 3}, the octahedron has type {3, 4}, the dodecahedron has type {5, 3} and
the icosahedron has type {3, 5}.
An example of a 4-polytope is a hypercube or 4-cube. It has cubes as
facets, three of them around each edge (see Figure 3.3) implying that its
vertex-figures are tetrahedra. Hence, its Schläfli type is {4, 3, 3}.
b) Show that if P is finite, the number of flags of Q is six times the number
of flags of P.
Exercise 25 a) Show that every two 2-polytopes with the same Schläfli
type are isomorphic abstract polytopes.
b) Find an abstract polytope that is not isomorphic to the cube, but has
the same Schläfli type.
c) Show that for each n > 2, there exist two n-polytopes with the same
Schläfli type that are not isomorphic.
Equivelar polytopes with exactly 2p1 · · · pn−1 flags are called tight.
27
28 CHAPTER 4. SYMMETRIES OF POLYTOPES
(Φi )γ = (Φγ)i ,
φ : Γ(P) → O
γ 7→ Φγ
is a bijection. Therefore there exists a bijection between every two flag orbits
|F (P)|
of P. In particular, if P is finite, |Γ(P)| = |O| = |Orb(P)| . Hence, |Γ(P)| ≤
|F(P)|.
If the action of Γ(P) has k orbits on the flags, we shall say that P is a
k-orbit polytope. If P is a 1-orbit polytope, then the action of Γ(P) on F(P)
is transitive and therefore regular. In such case we say that P is a regular
polytope.
30 CHAPTER 4. SYMMETRIES OF POLYTOPES
Lemma 8 Let P be an n-polytope, and let us denote by Orb(P) the set of all
flag orbits of P under the action of Γ(P). Let O1 , O2 ∈ Orb(P) and Φ ∈ O1 .
If for some i ∈ {0, . . . n − 1}, Φi ∈ O2 , then for any Ψ ∈ O1 , Ψi ∈ O2 .
Exercise 33 Show that the five platonic solids are regular polytopes.
Exercise 35 Let P be a 3-polytope and let Q be such that its 2-faces corre-
spond to the vertices, edges and faces of P, and its 1-skeleton is isomorphic
to the flag-graph of Q (that is, Q is the polytope from exercise 24).
a) What can you say about the automorphism group of Π(P) in terms of
the automorphism group of P?
c) If P is a k-orbit polytope, what can you say about the number of flag-
orbits of Pyr(P)?
c) If P is a k-orbit polytope, what can you say about the number of flag-
orbits of Pri(P)?
32 CHAPTER 4. SYMMETRIES OF POLYTOPES
Since the above lemma implies that the rank 2 sections on a given “level”
are isomorphic, the following corollary is a consequence of the lemma.
Moreover, since the polytope is flag connected, then the set {ρ0 , . . . ρn−1 }
generate the group Γ(P). To see this, we take γ ∈ Γ(P) to be any automor-
phism of P, and let Ψ := Φγ. Hence, there exist i0 , i1 , . . . ik ∈ {0, . . . n − 1}
such that
Φ = Φi0 ,i1 ,...ik .
By Proposition 7, Φi0 ,i1 ,...ik = (Φi0 )i1 ,...ik = (Φρi0 )i1 ,...ik = (Φi1 ,...ik )ρi0 , and by
doing this k more times, we get that Φi0 ,i1 ,...ik = Φρik ρik−1 . . . ρi0 . That is,
Φγ = Φρik ρik−1 . . . ρi0 . Since the action of Γ(P) is free on F(po), this implies
that γ = ρik ρik−1 . . . ρi0 and hence Γ(P) is generated by ρ0 , . . . ρn−1 .
Exercise 41 Suppose that P is an n-polytope such that for some base flag Φ
there exist automorphisms ρi such that Φρi = Φi for every i ∈ {0, . . . , n − 1}.
Show that P is a regular polytope.
Exercise 42 Let P be the cube in Figure 4.2, and let Φ be the base flag
shown in the figure.
b) Consider the Petrial Π(P) of P, and let Ψ be a flag having the same
vertex and edge than Φ. Describe the distinguished generators of Π(P)
with respect to Ψ.
34 CHAPTER 4. SYMMETRIES OF POLYTOPES
Figure 4.2: A cube with three reflections as its distinguished generators with
respect to some base flag.
We now turn our attention to the chains of P that are contained in the
base flag Φ. For any J ⊆ {0, . . . , n − 1}, we set ΦK := {(Φ)j | j ∈ K}, and
note that if i ∈
/ K, then ρi fixes ΦK . Hence, hρi | i ∈
/ Ki is a group contained
in the stabilizer (under the action of Γ(P)) of ΦJ¯. We denote such stabilizer
by ΓK . Further, we have the following lemma.
Lemma 12 Let P be a regular polytope and let ρ0 , . . . , ρn−1 be the dis-
tinguished generators of Γ(P) with respect to some base flag Φ. For any
K ⊆ {0, . . . , n − 1} we have that
ΓK = hρi | i ∈
/ Ki.
This lemma implies that for each i, 0 < i < n, we have that
Γ((Φ)i+1 /(Φ)i−2 ) ∼
= hρi−1 , ρi i.
4.1. REGULAR POLYTOPES 35
(ρi−1 ρi )pi = ε.
ΓK∪L = ΓK ∩ ΓL .
We shall not discuss the details here, but the converse of the above the-
orem is also true. That is, if Γ is a string C-group, then there is a regular
polytope whose automorphism group is Γ. In fact, there is a bijection between
regular polytopes and string C-groups given by an ordered set of generators.
36 CHAPTER 4. SYMMETRIES OF POLYTOPES
Bibliography
[1] H.R. Brahana, Regular maps and their groups, Amer. J. Math. 49
(1927), 268–284.
37