PERIODIC TILINGS AND TILINGS BY REGULAR POLYGONS

by

DARRAH PERRY CRAVEY

A thesis submitted in partia1 fu1fi11ment of the

requirements for the degree of

Doctor of Philosophy

(Mathematics)

at the

UNIVERSITY OF WISCONSIN - MADISON

1984
© Copyright by Darrah Perry Chavey 1984

All Rights Reserved

 PERIODIC TILINGS AND TILINGS BY REGULAR POLYGONS

Darrah Perry Chavey

Under the supervision of Professor Donald W. Crowe

Abstract: We assume a tiling has, under its symmetry group, v orbits

of vertices; e orbits of edges; and t orbits of tiles. Inequalities

are established relating these parameters, both for arbitrary tilings

and for tilings by regular polygons, and we show that some of these

inequalities are sharp. In the case of tilings by regular pOlygons,

we classify those tilings with v ~ 3, e ~ 3, or t~ 2, and show that

the number of tilings with some fixed number of orbits of vertices

[or edges; or tiles] is finite. The edge figures which can occur in

a tiling by regular polygons are Classified, as are tilings which

contain at most three different types of these edge figures.

Progress is made towards classifying those tilings by regular

polygons which contain at most two different types of vertex figures.

with respect to tilings by regular polygons which contain only two

types of tiles (two congruence classes of polygons), the number of

possible orbits of each polygon is determined. Tilings by regular

polygons in which any two congruent tiles are equivalent under the

symmetries of the tiling are classified, as are tilings which satisfy

a similar condition on the edges.

 ii

"We're all in it - we're all tiled, here."

Olga •.
The Grand Duke, by Gilbert and Sullivan.

"He's got 'em on the list - he's got 'em on the list;
And they'll none of them be missed - they'll none of them
be missed."
Chorus of Men,
The Mikado, by Gilbert and Sullivan.

Dedicated to the two women I love

Peggy and Eunice Chavey.
 iii

Acknowledgements

Now that it's almost over, it seems amazing to me that my

friends and my thesis committee (which are not exclusive) have

managed to put up with me for the last month or so. They are among

the many people I wish to thank for helping to make this thesis

possibl.e.

None of this work woul.d have been possible without the excellent

survey of the subject by GrUnbaum and Shephard, and I wish to thank

them for making their advance copy avail.abl.eto us. Professors

Donald Crowe and Michael Bl.eicher deserve thanks for their efforts in

creating and sustaining a seminar covering this work, and it was from

this seminar that most of these results developed.

Much of the work in this thesis owes a great deal, in ways that

are difficult to pin down, to conversations with Don Crowe and Mike

Bleicher; but some of the work can be more directly attributed to my

colleagues. Mary Leland discovered one class of tilirlgs used in the

proof of theorem 2.3 (as mentioned there), and this class helps to

extend the known range of realizable parameters in tilings. The nice

proof of fact 1 in section 1.3 is a drastic improvement of my origi-

nal, and this proof was pointed out by John Rosenberg. Elsa Gunter

vol.unteered to draw most of the til.ings in figures 5.2 - 5.5 on a

Carnegie-Mellon laser printer, and these figures (one of prettier

aspects of the thesis) would have been impossible without her help.
 32

graph induced by E, the dual graph induced by E does not in general

include all edges of the dual graph which join vertices of this

induced subgraph.

Lemma 2.1: Assume T is a (v , e, t)-tiling. Then:

(1) ~t'hElre
is a representative set of v vertices whose induced

grclph is connected;

(2) ThE~re is a representative set of e edges whose induced

grClph is connected;

(3) Thl~re is a representative set of e edges whose induced dual

graph is connected; and

(4) Th~!re is a representative set of t tiles whose induced dual

graph is connected.

Pf: We first prove (1). Let G be a maximal set of inequiva~ent

vertices 'whose induced graph is connected. Let V be the set of

vertices 'equivalent to some vertex in G and ~et V' contain all

other vertices. If Viis non-empty then by connectivity there is

an edge [v, Vi] with v in V; Vi in V'. Let S be a symmetry of the

tiling taking v to a vertex in G. Then S(v') is not equiva~ent to

any vertex of G, and G U { S(v') } induces a connected graph,

contradicting the maxlmallty of G.

The proofs of (2) and (4) are quite similar, but (3) is less

intuitlve and so we a~so prove it. Let G be a maximal set of

 33

inequiva1ent edges whose induced dua1 graph is connected; let E be

the set of edges equivalent to an edge in G, and let E' be the

complement of E. Assume E' is non-empty. Let D be the set of

tiles incident with E and D' be the set of tiles incident with E'.

D and D' cannot be disjoint (else, look at an edge where D and D'

meet, whic,h exists by connectedness of the dual graph) so let r be

a tile in both D and 0', i.e. T is incident with both E in E, and

E' in E'. Pick a sYmmetry S so that SeE) is in G, then as above

G U { S{E') } induces a connected dual graph, contradicting the

maximality of G. II

Although we will have no need of it, it is worth mentioning that

this 1emma can be strengthened to state that the representative sets

all induce "simply connected" graphs or dua1 graphs. Here simply

connected has the obvious definitions: a dual graph is simply

connected if the union of the tiles dual to its vertices forms a

simply connected set, and a graph is simply connected if its vertices

and edges do not separate any pair of vertices which are in the

tiling but not in the graph.

The proof of lemma 2.1 uses quite heavily the SYmmetries of the

tiling, and one would not expect results like this to hold for, say,

a k-gonal tiling as opposed to a k-isogonal tiling. Some results in

this direction do hold, as shown by the following lemma.

 34

Lemma 2.2: Let T be a tiling without singular points. Then:

(1) If T is 2-gonal, there is a set of 2 incongruent vertices

whose induced graph is connected. Similar conclusions do not

necessarily hold if T is 3-gonali

(2) If T is 2-hedral, there is a set of 2 incongruent tiles whose

induc:ed dual subgraph is connected. Similar conclusions do

not necessarily hold if T is 3-hedrali

(3) If T is 2-toxal, there are 2 sets of pairs of edges whose

Lnduced graph and induced dual graph (respectively) are

conne ct.ed , Similar conclusions do not necessarily hold if T

is 5··toxal.
l
Pf: The proofs for the 2-gonal, 2-hedral, and 2-toxal assertions all

follow imnediately from connectedness of the graph and dual graph,

and from arguments similar to those of lemma 2.1. Counter-

examples for the 3-gonal, 3-hedral, and 5-toxal cases are given in

figure 2.1 (a, b, and c respectively). II

The mos1t:surprising fact contained in this lemma is that it

remains open whether conclusions like those of lemma 2.1 hold in the

case of 3- or 4-toxal tilings. Counter-examples in the case of

3-hedral and 3-gonal tilings can be constructed as edge-to-edge

tilings by rElgular polygons, but no such nice counter-example can be

constructed for the 3-toxal case (see theorem 4.8). In fact, for
 35

l 1

Figure 2.1 (a) Figure 2.1 (b)

Figure 2.1 (c) Figure 2 •.

1 (d)

Figure 2.1: Examples of (respectively) 3-gona1, 3-hedra1, S-toxa1,

and 6-toxal tilings which have no representative sets of elements
of the relevant class which induce connected graphs or dual graphs
(as appropriate). Examples of representative sets are marked in
each tiling. The bold edge in (a) is referred to in section 3.1.
 36

edge-to-edge tilings by regular polygons. the smallest value of k

with a k-toxal counter-example that we know of is the 6-toxal example

of figure 2.1 (d). This situation is an example of a phenomenon that

recurs in chapter 4. namely that imposing local regularity on the

edge figures is a stronger condition than imposing local regularity

on the tiles or the vertex figures. This is especially true for

tilings by regular polygons. As another example. although it is

fairly easy to find monohedral. non~isohedral tilings and monogonal.

non-isogonal tilings. we do not know of an example of a monotoxal

tiling that 1s not isotoxal.

We are now ready to establish the main result of this section, a

partial class:ification of the realizable triples (v , e. t).

Theorem 2.3: Let T be a (v. e. t)-tiling. Then

(1) v < e+ 1 with equality only if the graph of the til.ing is

bipartite;

(2) t < e+1 with equality only if the dual graph of the tiling is

bipartite; hence only if every vertex of the tiling has even

valence;

(3) (v , E~) is realizable if it satisfies condition 1;

(4) (v. t) is realizable if t > v.

(5) (e. t) is realizable if it satisfies condition 2 and

t > e - (and t .?- 1);

3
(6) No upper bound can be set on the value of e or t in terms of

v.
 37

Pf: We note that (6) is implicit in (3) and (4).

(1) By lemma 2.1, there is a representative set E of e edges which

induces a connected subgraph G. Every vertex is incident with

some edge e, and e is equivalent to an edge in E, hence each

vertex is equivalent to a vertex of G. A connected graph with e

edges has at most e+1 vertices, with equality only if the graph

is a tree, so v < e+1. If v = e+1, then G is a tree and no two

vertices of G are equivalent, i.e. every vertex of T is

equivalent to a unique vertex of G. Since G is bipartite, we may

2-color 1~he vertices of G so that no edge joins two vertices of

the same color. We now extend this coloring to T by coloring

vertex orbits monochromatically. Since every edge of T is

equivalent to an edge in G, this is a 2-coloring of the vertices

of T, hence the graph of T is bipartite.

(2) The definition of the dual graph implies that two dual vertices

have only one edge joining them, even- if the tiles they represent

intersect in more than one edge. With this in mind, the proof of

the first part of (2) duplicates that of (1), using the dual

graph induced by E. It remains to note that if a vertex v of the

tiling had odd valence, then the tiles incident to v (which are

distinct) form an odd cycle in the dual graph, contradicting the

biparti tieproperty of the dual graph.

(3) and (4) If each edge in the regular tiling by squares is replaced

by e pairs of arcs, with each pair placed symmetrically about the

omitted edge, we get a (1, e, e+l)-tiling. Figure 2.2 shows an

l

38

example with e=2. This shows that (3) and (4) hold if v=1 and,

in fact, that these parameters can be realized by tilings that

are extremal with respect to inequality (2).

If we take the regular tiling by squares and replace each

2k-th left-to-right diagonal of squares (k ~ 2) by circles which

pass through the same vertices, we get a (k+1, k, k+1)-tiling.

Figure 2.3 (without the dashed edges) shows an example with k=3.

(In this figure, representative sets of edges and vertices are

marked, and the tiles to the lower right of the circled vertices

form a representative set of tiles.) This class of tilings not

only shows that inequalities (1) and (2) are best possible, but

also shows that for any e both bounds can be achieved

simultaneously (this example works for e > 1; for an example of a

(2, 1, 2)-tiling see Grunbaum & Shephard [1978b], figure 4, #14).

Since this class of tilings is used for further constructions,

we list the symmetries of these tilings. In addition to the

obvious translations, there are:

a) reflections in the line of the circles;

b) reflections in parallel lines midway between the circles;

c) 2-fold rotation centers at the center of each tile on the

lines of type a or b; and

d) reflections in lines perpendicular to those of type a

passing through the center of each circle.

These symmetries are all preserved by our modifications belOW.

Dissecting the circles in these tilings by m pairs of arcs

 39

Figure 2.3: A (4,3,4)-tiling without

Figure 2.2: A (1,2,3)-tiling the dashed edges, or a (4,4,5)-
representative of a class tiling with these edges. This
of (1, e, e+l)-tilings. represents a class of (v, e, e+l)-
tilings; 1 < v < e+l.

Figure 2.4: A (4,4,4)-tiling Figure 2.5~ A (2k, 2k+l, k+l)-tiling

representative of a class (with k=2) used to construct a
of (k+l, k+m, k+l)-tilings class of (2k, 2k+m+l, k+m+l)-
with 0 < m < k+l. tilings; m < k+3.
 40

placed symmetrically around the reflection lines of type a gives

a class of (k+1, k+m, k+m+1)-tilings. Figure 2.3 (including the

dashed edges) is an example for m=1. When k=1 and m > 0, this

construction still yields examples whose parameters are

(k+1, k+m, k+m+1). These examples prove (3) and (4) when v > 1;

again with examples that are extremal with respect to inequality

(2).

(5) Starting with the class of (k+1, k, k+1)-tilings of (3), we take

any m of the tile orbits and bisect the tiles along the

ref~ection lines of type d, to get (k+1, k+m, k+1)-tilings where

m < k+1. Figure 2.4 is an example with k=3 and m=1. Bisecting

one or both of the orbits of tiles which lie on the reflection

lines of type a and b increases the value of e while leaving t

unchanged. This creates (k+x, k+m, k+l)-tillngs for any m with

o~ m ~ k+3, where x is 1, 2, or 3. Thus e = k+m < 2k+3 = 2t+1,

showing that the pair of parameters (e, t) is realizable if

e - < t < e+1. This construction works for k > 1; hence

2
for e > 1. For the case where e=1, the tilings #1 and #10 in

l figure 4.1 give examples with t=1 and t=2 respectively.

TO complete (5), we build a similar class of tilings from

the regular hexagonal tiling.1 If the edges between every 2k-th

1 This class of examples was pointed out to us by Mary Diane

Palmer Leland, A. T. & T. Bell Laboratories.

 41

vertical column of hexagons and the adjacent hexagons are

replaced by circular arcs, the result is a class of

(1k, 2k+1, k+1)-tilings. Figure 2.5 is an example with k=2. As

above, dissecting any of the tile orbits along the reflection

lines leaves t unchanged, but increases e by m, where

o < m ~ k+3. This gives tilings where:

2t - 1 = 2k+1 < e < 3k+4 = 3t+1;

e - e +
hence: < t < , establishing (5). II
3 2

Although the statement of theorem 2.3, and its proof, assume

that v, e, and t are finite (i.e. that the tiling has no singular

points), it is worth remarking that the conclusions hold even if we

allow them to be infinite. If either t or e is finite, then by

theorem 1.3 there are no singular points, so the tiling is a

(v, e, t)-tiling. It is conceivable that v is finite and that there

are singular points, in which case t and e must be infinite. The

existence of such a tiling would imply that no bounds are possible on

the values of t or e as a function of v, but theorem 2.3 shows that

this is true even without the existence of singular points!

All the inequalities of theorem 2.3 (6) involve only two of the

parameters v, e and t. It seems likely that a more complete

classification of the realizable triples would inClude relations

between all three parameters. Of special interest would be bounds on

the Euler characteristic of a periodic tiling T, Eu(T), which we

define by Eu(T) = v - e + t. We do not know any lower bounds for

 4,2

EU(T), but we have found no classes of tilings for which Eu(T) < O.

Sharp upper bounds are established by Corollary 2.4.

Corollary 2.4:

Eu(T) ~ v+1, and equality is possible for any value of v;

Eu(T) ~ t+1, and equality is possible for any value of t;

Eu(T) ~ e+2, and equality is possible for any value of e.

Pf: The inequalities follOW by replacing v and/or t in the

definition of Eu(T) by the upper bounds of e+1 from theorem 2.3.

The upper bounds are all achieved by the (k+1, k, k+1)-tilings

constructed in the proof of theorem 2.3, parts (3) and (4). II

 43

Chapter 3: Ti1ings by Regu1ar POlygons.

l The remainder of this thesis considers only edge-to-edge tilings

by regular polygons. Much more can be said about this special class

of tilings. Such tilings cannot contain singu1ar points, so the

results of section '.3 apply. The current chapter considers general

questions involving such tilings, while the next two chapters deal

with more specific classification theorems. Section 3.1 classifies

the elements of a tiling that can appear in an edge-to-edge tiling by

regular polygons. Section 3.2 shows that the number of k-isogona1,

k-isotoxa1, or k-isohedra1 ti1ings for a fixed k is finite. Finally,

section 3.3 is devoted to the improvements that are possib1e in the

bounds of theorem 2.3 for this special class of tilings.

Section 3.': Classification of the Elements of a Ti1ing.

In much of the notation developed in this section, the number n

is used to refer to a regular n-gon. In an edge-to-edge tiling by

regular polygons a vertex figure is completely determined by knowing

what polygons are incident to the vertex, and in what order. For

these tilings the type of a vertex is n,.n2 •••• if the vertex is

incident, in cyclic order, to an n,-gon, an n2-gon, etc.. Thus the

regular tilings by squares and hexagons have vertices of types

 44

4.4.4.4 and 6.6.6 respectively. For brevity, we write these symbols

as 44 and 63, with similar abbreviations in other cases. In order to

obtain a unique symbol for each vertex type, we shall choose that

symbOl which is lexicographically first among all possible

expressions. We do not view a reversal of the cyclic order as a

distinct type. Consequently, two vertices have congruent vertex

figures if and only if they have the same type.

It is readily established (e.g. Grunbaum and Shephard [1977a])

that there are only 15 vertex types which can occur in an edge-to-

edge tiling by regular polygons (6 other vertex figures can be built

from regular pOlygons, but cannot be extended to tilings of the

plane). These 15 vertex types are listed in table 1 and pictured in

figure 3.1. One consequence of this classification of vertex types

is that the only tiles which can occur in these tilings are n-gons

where n is 3, 4, 6, 8 or 12. It is well known that the only such

tiling which contains an octagon is the unique isogonal tiling in

which every vertex has type 4.82 (see, for example, Grunbaum and

Shephard [1977a] or theorem 4.3).

In much of this thesis we will have occasion to refer to tilings

which contain only a few vertex types. Such tilings are indicated by

a list of the vertex types which occur, enclosed in parenthesis.

Thus, the 2-gonal tiling of figure 3.4, which contains only the

vertex types 34.6 and 3.6.3.6, is a (34.6; 3.6.3.6) tiling.

Occasionally, we will abuse this notation to refer to, say, a 2-gonal

tiling which is 3-isogonal. In this case the tiling of figure 3.4

 J J

Vertex
Type

3
4
6
Isogonal
Tiling

Yes
Valence

6
® CJjffiz:K>(tJ 3.3.3.4.4 3.3.4.3.4 3.3.6.6
3 .6 Yes 5
3.3.3.3.3.3 3.3.3.3.6

e=>
33.42 Yes 5
2
3 .4.3.4
2
3 .4.12
2 2
3 .6
Yes
No

No
5
4 <X){D 3.4.4.6 3.4.6.4
4 3.6.3.6
3.4.3.12 No 4 4.8.8
2
3.4 .6 No 4
3.4.6.4 Yes 4
3.6.3.6
3.12
44
4.6.12
4.82
2
Y.es
Yes
Yes
Yes
Yes
4
3
4
3
3
3.3.4.12 3.4.3.12
00 6.6.6

3
6 Yes 3

Table 1: The vertex

can occur in edge-to-edge
types which
tilings 4.6.12 3.12.12
EE
4.4.4.4
by regular polygons, with their
valence and whether an isogonal Figure 3.1: The 15 vertex figures which occur
tiling of that type exists. in edge-to-edge tl1ings by regular polygons.
",.In
 46

might also be called a (34.6; 34.6; 3.6.3.6) tiling to emphasize the

fact that the set of vertices of type 34.6 splits into two orbits of

vertices.

As with vertex figures, an edge figure is completely determined

by knowing what pOlygons meet the edge, and in what order.

Consequently, in many of the arguments used from here on, we abuse

the definitions of edge and vertex figures by including all of the

edges of these polygons within the figure (as in figure 3.1). To

describe an edge figure, we use the idea of an edge type. This is,

essentially, a listing in cyclic order of the polygons which meet the

edge, with the list starting and ending at one of the polygons inci-

dent to the edge. Specifically, the edge type of an edge £ starts

with one of the two polygons incident to £; lists the polygons inci-

dent with one of its endpoints in the (unique) cyclic order around

that vertex which ends at the other polygon incident to Ei and then

lists the polygons around the other endpoint, ending at the original

polygon. For convenience we separate the tiles incident with each

vertex by a "/" and list the tiles incident with the edge twice, once

with each endpoint. As an example, the bold edge marked in figure

2.1a is of type 36 / 32.42.3. Again, we use exponent notation for

brevity. There are in general 4 distinct symbols we could use for a

given edge figure. As with the vertex type, we can choose that sym-

bol which lexicographically precedes the others, and two edge figures

will then be congruent if and only if they have the same edge type.

We will not always use this convention, .however, since we will some-
 47

times wish to list one endpoint before the other to emphasize the

first vertex.

There are times when we need, or have, less information about an

edge than its edge type. Sometimes, all that is necessary is to know

what polygons are incident with the edge. Thus an edge is said to

have simple edge type n1.n2 if the edge is incident with both an

n,-gon and an n2-gon. As an example, the bold edge in figure 2.1a

has simple type 3.3, or 32• When there is danger of confusion as to

whether we are referring to a "simple edge type" or a standard "edge

type"1 we will refer to the latter as a full edge type.

As a step towards solving the classification problems considered

later, it is useful to know what vertex and edge types can occur in a

tiling by regular polygons. As mentioned before, the possible vertex

types are listed in table 1. The possible edge types have not been
'\

previously compiled, and so this is achieved by the following

theorem. In the classification of edge figures it is useful to

distinguish between two types of edge figures. A monogonal edge

figure (or edge type) is one where both endpoints have the same

vertex type. If the endpoints have different types, the edge figure

or edge type is non-monogonal.

Theorem 3.1: The only edge types which occur in edge-to-edge tilings

by regular polygons are the 41 monogonal edge types listed in

table 2 and the 57 non-monogonal edge types listed in table 3.

 48

Pf: Under the symmetries of the vertex figure (see figure 3.1)

certain edges incident to a vertex are equivalent. Using the

inequivalent edges it is straight-forward to write down all

combinations of vertex figures which could meet along, say, an

edge of simple type 32• Such a list includes separate entries for

inequivalent edges of the same simple type in a vertex figure. In

listing the edge figures for edges of simple types 32, 42, or 62

it is also necessary to include two entries if there are two

inequivalent orientations for matching a pair of vertex figures

along that edge; i.e. a reflection of one vertex. This list will

include some duplicates in the case of monogonal edge types, but

when these are deleted, the resulting list contains 101 edge

types. Three of these, 33.6.3 I 3.12.4.3, 3.4.12.3 I 3.62.3, and

3.12.3.4 I 42.6.3, cannot occur because such an edge type forces

the existence of a vertex incident to a triangle, hexagon, and

dodecagon and yet table 1 indicates that no such vertex exists.

To show that the other edge types do occur in tilings, we

must demonstrate tilings which contain them. The 3rd column in

table 2 and the 4th column in table 3 give the minimum values of v

for which there is a v-isogonal tiling containing that edge type.

The 1- and 2-isogonal tilings are pictured in figures 4.1 and 4.2

respectively, .and an entry in the tables of the form 2-17

indicates that the edge type exists in the 17th tiling shown in

the collection of 2-isogonal tilings. The 3-isogonal, vertex-

homogeneous tilings are shown in figure 5.1 and an entry in the

 49

tables of the form 3-19 refers to these tilings. Similarly. an

entry of the form 4-19 refers to the 4-isogonal. vertex-

homogeneous tilings of figure 5.2. An entry of the form 4'-2

refers to a 4-isogonal tiling which is not vertex-homogeneous.

The three non-homogeneous. 4-isogonal tilings referred to by the

table are shown in figure 3.2. The only edge types remaining are

the three in table 3 where v = 6, all of which occur in the tiling

of figure 3.3 (b) (marked as bold edges). II

The statement of theorem 3.1 does not claim that the "minimum v"

columns of tables 2 and 3 are correct. but this is easily verified.

Since the 1-, 2-, and 3-isogonal tilings have all been classified (or

are classified in Chapter 5), a straight-forward, (tedious) search of

these tilings combined with the 4-isogonal tilings of figure 3.2

verifies all of the values in these columns except for those where
•
v = 6. In these three cases it is easily seen that any of the edges

forces the configuration of figure 3.3(a). The vertices marked 1 to

4 must be in different orbits (1 and 2 have the same vertex types,

but vertex 1 is adjacent to a vertex of type 3.42.6, and vertex 2

cannot be). Vertex 5 could only be equivalent to one of the others

if it had type 32.4.12. in which case vertex x would be inCident to a

triangle, hexagon, and dodecagon. which is impossible. Thus vertices

1 to 5 are inequivalent and either the value v = 6 is correct or else

every vertex of a tiling containing this figure is equivalent to one

of vertices 1 to 5. In this case vertex z must be of type 4, and x

 50

Figure 3.2 - 1

Figure 3.2 - 2

Figure 3.2 - 3

Figure 3.2: The three 4-isogonal, non vertex-homogeneous tilings used

to show the existence of certain edge types in theorem 3.1.
 51

Figure 3.3 (a)

Figure 3.3 (b)

Figure 3.3: Part (a) shows the configuration forced by the existence
of anyone of the edge figures marked as bold edges. This cannot
be embedded in any v-isogonal tiling for v < 5. Part (b) shows a
6-isogonal tiling which contains these edge figures.
 52

cannot be (4 is adjacent to a vertex of type 1 along the edge of

simple type 3.4, and x isn't). Vertex x must then be of type 3,

hence is adjacent to a vertex of type 4; i.e. vertex y is of type 4.

Since every vertex of type 4 is adjacent to another such vertex (i.e.

4 and z), w is also of type 4; but then w is not adjacent (along the

hexagon) to a vertex of type 3, hence is not equivalent to vertex 4

and must be in a sixth orbit.

Tables 2 and 3 contain some additional information about the 98

possible full edge types. Within a given edge figure we can identify

various edges of several different simple edge types. For example,

in figure 2.1a, the edge figure corresponding to the bold edge

contains edges with simple types 32, 3.4, and 42. The numbers of

different simple edge types which are contained in the edge figures

become important in section 4.3, and these are listed in table 3.

 53

Vertex Minimum
Type Edge Figure v
============= =======-============== =======
6 6 6
3 3 I 3 1-1

3.6.3.6 3.6.3.6 I 6.3.6.3 1-8

44 44 I 44 1-2
3 3 3
6 6 / 6 1-3

4 4 4
3 .6 3 .6 / 6.3 1-2
3 3
" 3 .6.3 / 3 .6.3 1-2
3 2 2
" 3 .6.3 I 3 .6.3 1-4
3 3
" 3 .6.3 / 3.6.3 3-4
2 2 2 2
•• 3 .6.3 /3 .6.3 2-2

------------- ----------------------
2 2
-------
3 2 3 .42.3
3 .4 / 3 .42.3 1-5
2 2 2
" 3 .42.3 I 3.4 .3 . 2-11

•• 3 2 3
3 .4 / 42.3 1-5
3 3
" 4.3 .4 / 4.3 .4 1-5

------------- ---------------------- -------

2
3 .4.3.4 3.4.3.4.3 / 3.4.3.4.3 1-6
2 2
•• 3 .4.3.4 I 4.3.4.3 2-5

n
2 2
3 .4.3.4 / 4.3 .4.3 1-6
2 2
II
3.4.3 .4 I 4.3 .4.3 2-5
------------- ---------------------- -------
2
3 .4.12 3.4.12.3 I 3.4.12.3 3-27

n
3.4.12.3 / 3.12.4.3 3-28
2 2
" 3 .12.4 / 4.12.3 2-6

!~~!~_!!
The 41 monogonal edge types which occur in edge-to-edge
tilings by regular polygons.
 54

Vertex Minimum
Type Edge Figure v
============= ====================== =======
2 2 2
3 .4.12 3 .4.12 / 12.4.3 2-6

2 2
" 4.3 .12 / 12.3 .4 2-6

2 2 2 2 2 2
3 .6 3 .6 /6 .3 2-7
2 2
" 3.6 .3 / 3.6 .3 2-15
2 2
" 6.3 .6 / 6.3 .6 2-7

------------- ---------------------- _ .._----

3.4.3.12 3.4.3.12 / 12.3.4.3 3-33

II
3.12.3.4 / 4.3.12.3 2-115

------------- ---------------------- ---_._--

2 2 2
3.4 .6 3.4 .6 / 6.4 .3 2-17

II
2 2
3.6.4 / 4 .6.3 2-11:3

" 4.3.6.4 / 4.3.6.4 2-19

II
4.3.6.4 / 4.6.3.4 2-17
2 2
" 4 .3.6 / 6.3.4 2-17

------------- ---------------------- -------

3.4.6.4 3.4.6.4 / 4.6.4.3 1-7

" 4.3.4.6 / 6.4.3.4 1-7

1-9

" 12.3.12 / 12.3.12 1-9

------------- ---------------------- -------

4.6.12 4.6.12 / 12.6.4 1-10

" 4.12.6 / 6.12.4 1-10

" 6.4.12 / 12.4.6 1-10

------------- ---------------------- -------
1-11

II
8.4.8 / 8.4.8 1-11
 55

Simple
Vertex Vertex Min. Edge Thm. Thm.
1 v 4.1 4.8
======== ===~=========!~~2=~~~~2============ =~~2~ ===== =====
4
3 .6 36 I 33.6•3 2-1 2 I 8
6 2 2
" " 3 I 3 .6.3 2-1 2 I 7

" 33.42 36 I 32.42.3 2-2 3 I *

•• 2 6
3 .4.3.4 3 I 3.4.3.4.3 2-5 2 I 6
6
" 3 I 3.4.12.3 2-6 4 I

6 2
" 3 I 3.6 .3 2-7 3 I 3

33.6•3 I 32.42.3 3-20 4 **

II
3.6.3 I 3.42.32 4
" 3 3-21 *
2
" " 3 .6.3
2
I 32.42.3 41-1 4 **
2 3
•• 3 .4.3.4 3 .6.3 I 3.4.3.4.3 41-2 3 * *
2 2
" " 3 .6.3 I 3.4.3.4.3 3-20 3 ** *
2 3
" 3 .4.12 3 .6.3 I 3.4.12.3 6 5 *
2 2
" " 3 .6.3 I 3.4.12.3 6 5 **
3 2
" 3 .6.3 I 3.6 .3 4-18 3 * *
2 2 2
" " 3 .6.3 I 3.6 .3 2-8 3 I
*
4 2 2
" " 3 .6 I 6 .3 2-8 3 I A

2 4 2
3.4 .6 3 .6 I 6.4 .3 3-21 5 *
II
3-4 2 II 5

It It
4 .3.6 I 6.4.3.4 3-26 4 II

Table 3: The 57 non-monogonal edge types which occur in edge-to-edge

tilings by regular polygons.
 56

Simple
Vertex Vertex Min. Edge Thm. Thm.
1 2 v 4.1 4.8
======== ======== ====== =~~~~ =====
----=~~~~-!~~~~==---= =====
33.42 2 2
3 .4.3.4 3 .42.3 I 3.4.3.4.3 2-9 3 I B

33.42 2
II
" I 4.3 .4.3 2-9 3 I *
33.42 2
II
" I 4.3.4.3 2-10 3 I *
2
" 3 .4.12 32.42.3 I 3.4.12.3 3-24 5 *
II
" 32.42.3 I 3.12.4.3 4-20 5 *
33.42 2
" " I 4.12.3 3-8 5 **
2 2 2 2
•• 3 .6 3 .42.3 I 3.6 .3 3-25 5 *
33.42 I 4.3.12.3
l " 3.4.3.12
2 2 2
4-20 4 *
" 3.4 .6
3
3 .4 I 4 .6.3 3-21 5 * ---
3
" " 4.3 .4 I 4.3.6.4 4-2 5 **
3 2
" 3.4.6.4 3 .4 J 4.6.4.3 2-11 4 I

44 3
" 4.3 .4 I 44 2-12 3 I C
--------- -------- ---------------------- ------ ------
2 2
3 .4.3.4 3 .4.12 3.4.3.4.3 I 3.4.12.3 4'-3 4 *
2 2
" " 3 .4.3.4 I 4.12.3 3-14 4 *
2 2
II
" 3.4.3 .4 I 4.12.3 4-15 4 **
2 2 2
" 3 .6 3.4.3.4.3 I 3.6 .3 4-26 4 *
2
" 3.4.3.12 3 .4.3.4 I 4.3.12.3 4'-3 3 * *
2
" " 3.4.3 .4 I 4.3.12.3 4-16 3 ** '*
2 2 2
" 3.4 .6 3 .4.3.4 I 4 .6.3 4-26 5 *
2 2
" " 3.4.3 .4 I 4 .6.3 3-15 5 **
2
" 3.4.6.4 3 .4.3.4 I 4.6.4.3 2-14 3 I *
2
" " 3.4.3 .4 I 4.6.4.3 3-16 3 III *
2
" " 4 .3.6 I 6.4.3.4 3-26 4 II
 57

Simple
Vertex Vertex Min. Edge Thm. Thm.
1 2 v 4.1 4.8
======== ======== ====== =~~g~~===== =====
=====~~~~=~~~~~======
2 2
3 .4.12 3.4.3.12 3 .12.4 I 4.3.12.3 3-27 4 **
2
" " 3 .4.12 I 12.3.4.3 3-27 4 **
2 2 2
" 3.4 .6 3 .12.4 I 4 .6.3 6 7 *
2
" 3.4.6.4 3 .12.4 I 4.6.4.3 3-24 5 **
2 2
" 3.12 3 .4.12 I 122.3 3-28 5 **
2
" 4.6.12 4.3 .12 I 12.6.4 3-18 6 **
-------- -------- ---------------------- ------ ------
2 2 2 2 2 2
3 .6 3.4 .6 3 .6 I 6.4 .3 3-25 6 *
2 2
" 3.6.3.6 3 .6 I 6.3.6.3' 2-15 3 I 4
3 2 3
" 6 6.3 .6 I 6 3-19 3 ** 2
-------- -------- ---------------------- ------ ------
3.4.3.12 3.4.6.4 3.12.3.4 I 4.6.4.3 3-33 3 ** *
2
" 3.12 3.4.3.12 I 122.3 2-16 3 I
-------- -------- ---------------------- ------ ------
2 2
3.4 .6 3.4.6.4 3.6.4 I 4.6.4.3 2-17 4 I

2
" " 4 .3.6 I 6.4.3.4 3-26 4 II

2
" 3.6.3.6 3.4 .6 I 6.3.6.3 2-18 4 I

44 I 44 **
" 4.3.6.4 3-34 4
2
" 4.6.12 4 .3.6 I 6.12.4 3-39 6 **
-------- -------- ---------------------- ------ ------
3.4.6.4 4.6.12 4.3.4.6 I 6.12.4 2-20 4 I
 58

Section 3.2: Finiteness of Tilings with k Orbits of an Element.

This section shows that there are only a finite number of

tilings by regular polygons which have a fixed number of orbits of

some class of elements. In fact, for the proof of the main result,

the only facts needed are that there are a finite number of tiles,

vertex figures, or edge figures possible in such cl tiling and that

any element can be adjacent to only a finite number' of other elements

of the same class. For simplicity, the theorem Ls stated and proved

only for edge-to-edge tilings by regular pOlygons, and then a more

general formulation (which can be established by the same proof) is

stated as a corollary. This result answers two questions posed by

Griinbaum and Shephard [1983] (see sec·tion 1.2).

To prove this result, we need some preparation. Assume T is a

v-isogonal tiling. Given a representative set C of v vertices that

induce a connected graph, we define the fundamental region R that

they generate to be the set C together with all edges incident with

at least one vertex of C. We call the original set C the center of

R, and the vertices in R which are not in C are denoted by C'. An

example of a fundamental region is the set of edges marked in figure

3.4 with bOld edges. A tiling can have incongruent fundamental

regions, and the centers of two such regions are marked by the

circled vertices in figure 3.4. Also, the same fundamental region

can generate more than one tiling -- for example, the two (33.42; 44)

tilings of figure 4.2 have the same fundamental regions. However,

the idea of the proof of theorem 3.3 is to show that there are at
 59
.. , ' , ,
.. ..,
,

»
... -6. . ...
c
,
y , ,
, ,
, ,
<
',CJ'
.. ..
,"
"

,..
, > ., "

, >
.
,
,
,'I

.,.
..
.. -1, ... ,
'c

, ..
,
.•.
,
~
....
.!,
,"
, ' ..• . ..
,,
.' '"
I
•
'

Figure 3.4:
4 4
A 2-gonal, 3-isogonal (3 .6; 3 .6; 3.6.3.6) tiling.
Two different representative sets of vertices are marked with
open circles, and the fundamental region R generated by one set
is marked with bold edges. R includes all of the solid edges,
and is the union of R and the fundamental regions generated by
the five centers marked by inscribed triangles.

I I I I
Figure 3.5: An example of two different tilings which have the
same fundamental regions and labeled closures with respect to
vertices, edges, and tiles.
 60

most a finite number of fundamental regions possible, and that each

fundamental region can be extended to only a finite number of

v-isogonal tilings. The second half of this proof uses the idea of a

closure of a fundamental region. Let v be any vertex in ct. By the

isogonality condition, v is contained in the center of at least one

image of R under some symmetry of T, and this image is denoted by Rv'

In figure 3.4 the centers of each such Rv are indicated by the five

inscribed triangles. In a more general tiling, Rv may not be

uniquely determined. A closure of R, denoted R, is the union of R

and one Rv for every v in ct. In figure 3.4, R consists of all the

solid edges. Again, R is not always uniquely dete!rmined. Finally, a

labeled closure of R is a closure of R in which every vertex v' in C'

is labeled by a pair (v, S) where v is the vertex in C to which v' is

equivalent, and S is a symmetry of T that takes v to v' and R to Rv,1.

For further examples of the ideas involved in the proof of lemma 3.2,

we suggest that the reader consider Rand R in the two (33.42; 44)

tilings of figure 4.2.

Of course, knowing S tells us which vertex of C is equivalent to

v', and knowing S for each vertex in C' also describes the struc-
ture of R, so there is a good deal of redundant information here.
This definition, however, seems to make the proof follow more
easily. Exactly how much information about Rand R is needed to
give the result of lemma 3.2 is unclear, and this is discussed
further in chapter 6.
 61

Lemma 3.2: No two different tilings without singular points have the

same fundamental region R and labeled closure R.

Pf: We claim that a knowledge of Rand R, with R labeled, uniquely

determines the tiling. Rather than comparing two different

tilings, we imagine trying to construct a tiling T from Rand R,

and show that this construction is unique. The idea of the

construction is straight-forward: at any stage in extending T,

there will be a vertex v' near the "border" of the construction

with a label (v, S). v is a vertex in R which is equivalent to v'

under the symmetry S, and S forces the existence of a copy Rv' of

R which covers v'. Adding the tiles of Rv' to our construction of

T will, in general, force the existence of a larger portion of T.

Note that while the vertex v' may be contained in several

different copies of R, the symmetry S specifies exactly one of

these copies. Thus, while there may not be a unique extension of

T which embeds v' in a copy of R, nevertheless the extension we

will choose is forced, and hence the construction of T will be

unique.

We need to show that this kind of extension is always

possible. and that it eventually forces all of the tiling. To

make the idea of the "border" of the construction more precise, we

let the distance from a vertex v' to C, the center of R, be the

minimum distance from v' to some vertex of C.

Every image of R under a symmetry a of T is embedded in the

 62

same way as R in an image of R, and the image of a label (v,S)

will be (v, o(S». For every vertex Vi at distance 1 from the

center of R, Rv' is contained in R (R being the first step in our

construction of T from R). and the sYmmetry which labels Vi

extends Rv' uniquely

.•....
to an RVI. This gives a larger portion of

the tiling which, for every vertex 00 at distance 2 from the

center, contains 00, a fundamental region Roo' and a label for w.

Iterating this process n times gives a portion of the tiling which

includes the vertex figures and a copy of Roo for each vertex 00 at

distance n+1 from the center. Since the graph of the tiling is

connected, every vertex is at finite distance from the center,

hence every vertex figure in the tiling is eventually forced. II

Lemmata similar to 3.2 hold for the edges and tiles. If we define a

fundamental edge region [fundamental tile region] to be the union of

a representative set of edges [tiles] that induce a connected graph

[dual graph) together with all adjacent edges [tiles), and define the

labeled closures in the natural way, then arguments identical to

those used above show:

Lemma 3.2': No two different tilings without singular points have the

same fundamental edge region and labeled closure.

Lemma 3.2": No two different tllings without singular points have the

same fundamenta1 ti1e region and 1abe1ed c1osure.

 63

Throughout this chapter, we have been assuming that our ti1ings

contain no singu1ar points, and these 1ast three 1emmata are a11

false if we allow such points. Figure 3.5 shows two tilings which

have the same fundamental regions and labeled closures with respect

to vertices, edges, and tiles.

We are now ready to move on to the main result of this section.

Theorem 3.3: Let S be the set of all edge-to-edge ti1ings by regu1ar

polygons and let k be a fixed integer. Then the number of

k-isogonal tilings in S is finite; the number of k-·isotoxal

tilings in S is finite; and the number of k-isohedral tilings in S

is finite.

Pf: The proofs of the three parts of this theorem are essentially the

same, so we give only the proof for the k-isogonal tilings, using

1emma 3.2. For this theorem (tillngs by regular polygons), as

opposed to corollary 3.4 (more general ti1ings). the finiteness of

the k-isotoxal and k-isohedral tilings also follows from the

finiteness of the k-isogonal tilings via theorem 3.6.

Fix k. We claim there are a finite number of possible

fundamental regions with center of size k. This fol10ws since

there are a finite number of connected graphs G on k vertices;

each vertex in G can be occupied by at most 1 of 15 vertex

figures; and each edge of G can correspond to at most 1 of 6 edges

in the vertex figure of either endpoint.

 64

Fix a fundamental region R with center C and let CI be the

set of vertices in R but not in C. For each vertex v' in C' there

are k choices for which vertex v in C is equiva~ent to Vi. Vi is

adjacent to C, and there are at most valence(v') (~6) choices

for what edge in the vertex figure of Vi joins it to C, and at

most 1 symmetry of T which fixes that edge and its endpoints.

Thus there are at most 12 symmetries of T which take v to v'.

Hence, for a fixed v, this gives at most 12 choices for Rv" and

for the ~abel attached to Vi. Since CI is finite, there are thus

a finite number of choices for the labeled closure of R, and each

of these choices extends to at most one tiling. II

Corollary 3.4: Let k be a fixed number. Then:

(1) Given a finite set of vertex figures, each of finite valence,

then the number of k-isogonal ti~ings that contain on~y

these vertex figures is finite;

(2) Given a finite set of edge figures, with each incident vertex

having finite va~ence, then the number of k-isotoxa~ ti~ings

that contain only these edge figures is finite.

(3) Given a finite set of ti~es, each with a finite number of

edges, then the number of k-isohedral tilings that contain

only these tiles is finite.

Remark: In (3), we emphasize that the "edges" of the tiles must be

edges in the tiling. This is equivalent to saying that for each tile
 65

we know what points along its boundary will be vertices of the

tiling. If we were to try to omit this restriction we would find,

for example. that there are infinitely many isohedral tilings by

congruent squares, i.e. the standard (44) tiling by squares with rows

slid along each other by a constant a.

 66

Section 3.3: Bounds on the Number of Orbits of an E~ement.

This section considers improvements in the bounds of theorem 2.3

under the added assumption that the tiling is edge-to-edge and all

tiles are regu~ar polygons. Specifica~~y, bounds are estab~ished on

the number of orbits of one c~ass of elements in the tiling as a

fUnction of the number of orbits of another c~ass of e~ements. To

take advantage of the fact that equality can ho~d in theorem 2.3 (1),

on~y if the graph of the ti~ing is bipartite, we first classify these

graphs.

Lemma 3.5: Let T be an edge-to-edge ti~ing by regu~ar po~ygons whose

graph is bipartite. Then T is one of the four isogonal tilings

(44); (63); (4.82); or (4.6.12); hence v = 1 < e.

Pf: A ti~ing whose graph is bipartite contains no triang~es, and the

four vertex types mentioned in the lemma are the only such vertex

types in table 1. No non-monogonal edge exists which uses only

these four vertices (see table 3), hence the tiling must be

monogonal. It is well known (see section 4.1), that this implies

the tiling is isogonal. II

Grunbaum and Shephard [1977a] ask if for (v, e, t)-ti~ings by regular

polygons there are functions t(v) and vet) such that for a fixed t,

v ~ vet) and for a fixed v, t~ t(v). We answer these questions, and

the four re~ated questions invo~ving the parameter e, in theorem 3.6.

 67

Theorem 3.6: Let T be an edge-to-edge (v, e, t)-tiling by regular

polygons. Then:

(1) v < e and t~ e + 1;

(2) e < 5v and t < 4v + 1;

(3) e < 6t - 2 and v < 6t - 2.

Pf: (1) From theorem 2.3, t~ e+1; and v ~ e except possibly if the

graph of T is bipartite. Lemma 3.5 shows that v < e for these

tilings also.

(2) By lemma 2.1 there is a representative set V of v vertices whose

induced subgraph is connected. Every edge and tile in T is

equivalent to an edge or tile incident with a vertex in V, so e

and t are bounded by the number of edges and tiles incident to V.

From table 1, valence(v) < 6 for each vertex v in V. Also, if

every vertex of V has valence 6, then every vertex is of type 36

(see table 1) and the tiling is the isogonal (36) tiling (see

section 4.1) which satisfies (2). Thus, w.l.o.g., there is some

v in V with valence(v) ~ 5. Hence the number of edges and the

number of tiles incident with V is at most 6v-1, even counting

multiplicities for tiles or edges incident with more than one

vertex of V. Since the graph induced by V is connected, it

contains a spanning tree with v-1 edges, each of which has been

counted twice, and each of these edges is incident with two tiles

that have been counted twice. Thus we have over counted the edges

by at least v-1 and the tiles by at least 2(v-1), proving (2).

 68

(3) Let D be a representative set of t tiles whose induced dual graph

is connected. Since D may contain 12-gons, imitating the argu-

ment of (2) does not give strong results.

Let tn be the number of n-gons in D. As remarked earlier,

the only tiling with an octagon is the isogonal (4.82) tiling,

which satisfies (3). We may thus assume tn - 0 except for n = 3,

4, 6 or 12. Since no vertex is incident with three 12-gons, at

least six edges of each 12-gon are incident with an m-gon where

m ~ 6, and are thus equivalent to an edge of such an m-gon in D.

Thus, the number of inequivalent edges incident to a tile in D is

at most:

3t3 + 4t4 + 6t6 + 6t'2 < 6t

so e < 6t. If e ~ 6t-l, then t3 and t4 = O. Lemma 3.5 shows

this happens only if T is the isogonal (63) tiling (where e = 1),

hence e < 6t - 2.

By table 1, every vertex of T is incident with an m-gon

where m~ 6, so to count the inequivalent vertices, we need only

count the vertices incident with such an m-gon in D. As with the

edges, we find that v ~ 6t, with v > 6t-1 only if t3 and t4 - 0

(in which case v = 1), hence v < 6t - 2. II

One immediate corollary to this theorem is the proof of a

conjectured enumeration of the 2-isotoxal tilings by Grunbaum and

Shephard [1977a] (see section 1.2). Theorem 4.8 will prove a

stronger version of corollary 3.7.

---, 69

Corollary 3.7: The only isotoxal edge-to-edge tilings by regular

polygons are the four isogonal tilings: (36); (44); (63); and

(3.6.3.6). The only 2-isotoxal edge-to-edge tilings by regular

pOlygons are the four isogonal tilings: (3.122); (4.82);

(3.4.6.4); and (32.4.3.4).

Pf: By theorem 3.6, isotoxal tilings are isogonal, and 2-isotoxal

tilings are either isogonal or 2-isogonal. These tilings have

been classified (Kepler [1619] and Krotenheerdt [1969], see

figures 4.1 and 4.2), and inspection of these tilings yields the

desired result. II

We have no reason to believe that the bounds of theorem 3.6 are

sharp. except for (1) with small values of e. The isogonal (3.6.3.6)

tiling is a (1, 1, 2)-tiling which achieves equality in both parts of

(1). There are also tilings where v .••e for e ~ 8; and one tiling

where t •••e+ 1 = 3. Ineql1alities (2) and (3) can probably be

improved, and possibly (1) can be improved for larger values of e.

Theorem 3.8, however, gives some limits on possible improvements. To

establish these limits. we need to construct various classes of

tilings. One way to do this is to combine "strips" of tiles to cover

the plane. This technique will be used here and again in chapter 4,

so it is helpful to be precise as to what we mean by this. A strip

is a set of polygons which lie between two parallel lines and cover

the area between those lines; the lines are assumed to be close

enough that polygons meet both lines. There are three kinds of
 70

strips: strips of squares; strips of triangles; and strips of

hexagons. These strips are all demonstrated in figure 3.8. A "strip

of hexagons" also includes triangles (although only the hexagons meet

both lines bounding the strip) and the hexagons all meet at vertices

of type 3.6.3.6.

Theorem 3.8: For each of the following equations there are

infinitely many edge-to-edge (v, e, t)-tilings by regular polygons

which satisfy that equality:

(1) v e - 1; t e - 1;

(2) e = 3v - 2; t = 2v;

(3 ) e = 3t - 9; v 2t - 1.

Pf: Figure 3.6 is an example of a strip of squares separated by k

strips of triangles (here, k=4). This family of tilings has

parameters (v, e, t) = ( Lk;2J ' k+2, k+1 ) where LxJ denotes the

greatest integer function. This class of tilings satisfies

t = e-1, for any e > 3 and also satisfies t - 2v when k is odd.

Figure 3.7 is an example of a column of triangles separated by k

columns of hexagons (in figure 3.7, k=4). This family has

parameters (v, e, t) = ( k+1, k+2, Lk;3 J ) . These tilings

satisfy v e-1, for any e ~ 3, and also satisfy v = 2t-1 when k

" is even. Figure 3.8 may be modified by using 2k strips of squares

(in the figure, k=2), to give a family of ( 2k+4, 3k+6, k+5 )-

 71

Figure 3.6: A (3.6.5)-tiling which demonstrates a class of

( Lk+ 2-'
L Z.J' k+2, k+l)-tilings. Representative sets of edges and

vertices are marked.

Fig. 3.7: A (5,6,3)-tiling which Fig. 3.8: A (8,lZ,7)-tiling which

demonstrates a class of demonstrates a class of
( 2k+4, 3k+6, k+5)-tilings.
( k+l, k+2, ~~3J )-tHings.
Representative sets of vertices
Representative sets of vertices and edges are marked.
and edges are marked.
 72

tilings. which satisfies e = 3t-9. If we remove the strips of

triangles (and translate the squares over to meet with the

hexagons), and use 2k+1 strips of squares, the result is a class

of ( k+2, 3k+4, 2k+4 )-tilings, which satisfies e = 3v-2. II

With til.ings by arbitrary tiles (theorem 2.3). we showed the

existence of a wide. continuous range of realizable pairs of

parameters. We can also show the existence of continuous ranges of

realizable parameters in the case of tilings by regular pOlygons.

Unlike theorem 2.3. however. these ranges cover most, but not all,

of the values within the extremes of theorem 3.8.

Theorem 3.9: For edge-to-edge tilings by regular polygons:

I
(1) Any pair of parameters (v, e) is realizable if

v + 1 < e < 2v

(2) Any pair of parameters (v, t) is realizable if

Lt;1J .s v < 2t-1 ;

(3 ) Any pair of parameters (e, t) is realizable if

t + < e < 2t and t > 2

or t + < e
-< 3t - 19.

Pf: (1) If we broaden the class of tilings referred to in figure 3.7

to allow j columns of triangles and k columns of hexagons, the

resulting tilings have parameters (v, e, t)

(l~J ;+211J
+ k , j + k + 1 ,
l~J ) .
 73

(This class of examples is used again for theorem 4.7, and two

examples with j > are shown in figure 4.9). When j = 1,

e = v+1; and when k = 1 and j is odd, e = 2v. It is readily

verified that any intermediate values of e and v can be obtained.

In this class of examples, we must have v > 2; but corollary

3.7 gives four tilings with v = 1 and e = 2.

(2) We look again at the class of examples of (1). When k = 1,

v = t; and when j = 1 and k is even, v 2t-1. The range is

complete between these extremes, so this gives tilings with all

values (v, t) satisfying t ~ v ~ 2t-1.

We now generalize the class of ex~mples of figure 3.6 so as

to alternate j strips of squares with k strips of triangles. The

resulting tilings have parameters (v, e, t) =

l~J +L~J +1, j+k+1
lJ iI
.i:!:...L2
+ k i.

When k = 1 and j is even, v = t; and when j - 1, v = It;] .

The range between these extremes is again complete.

In these classes of examples, t~ 2, but the three isohedral

tilings (tilings #1, 2, and 3 in figure 4.1) satisfy t = 1 and

v = 1.

(3) The second class of examples from (2) (strips of squares and

triangles) gives tilings with any pair (e, t) satisfying

t+1 < e < 2t (with t * 1). To conClude the argument, we

need only consider the range 2t+1 < e < 3t-19, which is

empty unless t > 20. Hence we need not concern ourselves with
 74

small values of t.

We t:irst generalize the example of figure 3.8 to allow j

strips of hexagons and 2k strips of squares, with the strips of

hexagons arranged according to the pattern of the isogonal

(3.6.3.6) tiling. These tilings have (e, t) =

(
j + l~J +
3k + 4 , T
j + l"+1J + k + 3 ) .
This range extends from e = t+3 (when k = 1) to e = 3t-9 (when

j = 1). Unfortunately, the range is not continuotLs since

e - t = 2k+ 1, which is odd. If, however, we replalce the single

strips of triangles by three strips of triangles, we have a class

of tilings where (e, t) =

( j + Lj;1J + 3k + 11 , j + li;1J + k + 9 ).

These tilings cover the range of values from e = t+4 to

e = 3t-20. In this case, e-t 2k+2 is even, so this fills in

the gaps of the previous class. (We remark that neither class of

examples gives a continuous range of values even if we allow an

odd number of strips of squares). II

 75

Chapter 4: Local Regularity in Tilings by Regular Polygons.

Chapter 1 defined three kinds of local regularity conditions

that one might impose on a tiling: i.e. asking that the tiling be

k-gonal. k-toxal. or k-hedral for some sma~~ va~ue of k. In genera~

these conditions do not force very much structure on the tiling. even

in the special case of tilings by regular polygons. Still. some

interesting results can be established. For example. the partial

analysis we shall do of the 2-gonal tilings is quite helpful in

classifying the 2- and 3-isogona~ ti~ings. and the results on

e-isotoxal ti~ings are somewhat surprising. As with chapter 3, we

consider only edge-to-edge tilings by regular pOlygons.

Section 4.1: v-gonal Tilings.

It is known' that the only monogonal tilings by regular polygons

are the 11 isogonal tilings shown in figure 4.1. The complete

determination of the 2-gonal tilings seems like a difficult problem,

although not impossib~e. As a special case, the determination of all

2-gonal (33.42; 32.4.3.4) tilings is unknown. This is a particularly

1 See, for example, Grunbaum and Shephard [1977a).

 76

,
/'
\I
'\
/ ,/ \I \
'\ '\
\

V
I'v
I
Ie
I I
/

'- <r.
,/
'\
'\1\/ '\ '\/

4 2
4: (3 .6) 6: (3 .4.3.4)

2
7: (3.4.6.4) 8: (3.6.3.6) 9: (3.12 )

2
10: (4.6.12) 11: (4.8 )

Figure 4.1: The eleven isogonal edge-to-edge tilings by

regular polygons.
 77

interesting special case, since here each vertex has the same number

of incident polygons of each type (in the terminology of GrUnbaum and

Shephard, each vertex is of the same "species"). The other 2-gonal

tilings with this property, specifically the (32.62; 3.6.3.6) and

(3.42.6; 3.4.6.4) tilings, have been Classified by Sommerville [1905)

(see also lemmata 4.4 and 4.5). A knowledge of the 2-isogonal

tilings is helpful in discussing the 2-gonal tilings. These tilings

have been Classified by Krotenheerdt [1969], and are shown in figure

4.2.

Although we cannot currently classify the 2-gonal tilings, we

can determine which pairs of vertices occur in 2-gonal tilings. By

lemma 2.2, there must be an edge connecting two vertices of different

types, and we can determine all such non-monogonal edge figures which

occur in 2-gonal tilings.

Theorem 4.1: The only non-monogonal edge types which occur in

2-gonal edge-to-edge tilings by regular pOlygons are:

Class I: 36 I 33.6.3; 36 I 32.6.32; 36 / 32.42.3;

36 I 3.4.3.4.3; 36 I 3.4.12.3; 36 I 3.62.3;

r+.
34.6 I 62.32; 32.6.32 I 3.62.3;

33.42 I 4.32.4.3; 33.42 I 4.344.32;

33.42 I 4.6.4.3; 32.42.3 I 3.4.3.4.3;

3.42.6 I 6.3.6.3; 3.6.42 / 4.6.4.3;

 78

32.4.3.4 I 4.6.4.3; 32.62 I 6.3.6.3;

3.4.3.12 I 122.3; 4.3.4.6 I 6.12.4;

Class II:· 34.6 I 6.3.6.3;

Class III: 3.4.32.4 I 4.6.4.3

Furthermore, each edge type of c~ass I occurs in a 2-isogona~

ti1ing; each edge type of c1ass II occurs in a 3-isogona1 ti1ing,

but in no 2-isogonal tiling; and the edge type of class III occurs

in a 4-isogona1 tiling but in no v-isogona1 ti1ing for v < 4.

Coro11ary 4.2: The on1y types of 2-gona1 ti1ings are:

C~ass I: (36; 34.6); (36; 33.42); (36; 32.4.3.4);

(36; 32.4.12); (36; 32.62); (34.6; 32.62);

(33.42; 32.4.3.4); (33.42; 3.4.6.4); (33.42; 44);

(32.4.3.4; 3.4.6.4); (32.62; 3.6.3.6);

(3.4.3.12; 3.122); (3.42.6; 3.4.6.4);

(3.42.6; 3.6.3.6); (3.4.6.4; 4.6.12);

Class II: (34.6; 3.6.3.6).

Furthermore, every vertex pair of c1ass I occurs in a 2-isogona~

tiling whi~e the vertex pair of class II occurs in a 3-isogonal

tiling but in no 2-isogonal tiling.

Pf: The corollary follows immediately from the theorem.

 79

~ x
,/
.r-:
1\.1'
.r.
'\ ) ~
~
"-/
,/\ f\.
\/ V\/

9: (33.42;32.4.3.4)
1

10: (33.42;32.4.3.4)
2

Figure 4.2: The twenty 2-isogona1 edge-to-edge ti1ings by

regular polygons.
 80

I I I I I I I
/ '\ 'V I I I I I I I I
/ / '\/
I
I J
I I I
I I I
, '\ J <./\ ~
I I I I I I I IJ

13: (33.44;44) 14: (32.4.3.4;3.4.6.4)

2

16: (3.4.3.12;3.122) 17: (3.42.6;3.4.6.4)

18: (3.42.6;3.6.3.6) 19: (3.42.6;3.6.3.6) 20: (3.4.6.4;4.6.12)

1 2

Figure 4.2 (continued)

 81

Pf of thm: In table 3. under the column labeled "thm 4.1." those edge

figures £ which are marked by a * all contain a vertex which

cannot be completed to have the same type as either endpoint of E.

Those edge figures E marked ** all share the property that any
attempt to extend the edge figure to a larger portion of the

tiling using only the vertex types of the endpoints of E. forces

several vertex types until a contradiction is reached. The only

interesting example of this form is 3.4.32.4 I 4.3.12.3 (the solid

edge:; in figure 4.3). Vertices 1 and 2, in that order, are forced

to be of type 32.4.3.4. Now vertex 3 cannot contain a 12-gon,

else vertex 4 is of type 3.122, so vertex 3 has type 32.4.3.4.

There are 2 locations for the additional triangle at vertex 3. In

one case. vertex 5 has 3 triangles together and cannot have either

allowed type. In the second case. vertex 4 has 2 triangles

together, giving the same contradiction.

Omitting the edge figures marked * or ** leaves the 22 edge

figures of the theorem. Which edge figures do or do not arise in

2-isogonal tilings can. be read off from the classification of

2-isogonal tilings (see figure 4.2). (Alternatively, one could

use these 22 edges as a starting point for that classification.)

For examples of 3-isogonal tilings containing the edge figures of

class II see figure 5.6, #10 and #19; for a 4-isogonal tiling

containing the edge of class III, see figure 4.4. It remains to

show that there is no 3-isogonal tiling containing the edge figure

of class III (we cannot use the classification of 3-isogonal

I .
 J ) J J J

\-- - -
.••..••.. I ••••
, , I
, , 1/ ,
I ,
\
\
, \
,

Figure 4.3: The 2-gonal configuration forced

2 .
by an edge of type 3.4.3 .4 / 4.3.12.3.

Figure 4.5: The 2-gooal conf iguration forced Figure 4.4: A 4-isogonal tiling containing
2 2
by an edge of type 3.4.3 .4 / 4.6.4.3. edges of type 3.4.3 .4 / 4.6.4.3. en
t\)
 83

tilings for this step since that classification will depend on

this thE~orem). Starting with this edge figure (corresponding to

the bold edge in figure 4.5), and filling in some vertex figures

that arE~ forced by the assumption that it is 2-gonal, we get the

portion of the tiling of figure 4.5. If we think of pairs of

adjacen1t triangles as "diamonds" and count the number of hexagons

and diamonds incident to each of the marked vertices, we see that

these 4 marked vertices are inequivalent. II

As remarked above, we cannot currently classify all the 2-gonal

tilings; but this can be done with many of the pairs listed in

corollary 4.2. Since this palrtial classification of the 2-gonal

dissected into 6 triangles, ,'orming a vertex of type 36 and changing

the types of the vertices inc'ident to the dissected hexagon.

Similarly, a 12-gon may be dlssected into a hexagon surrounded by six

adjacent squares and six nei2~hbOring triangles. In a given tiling

I
this dissection of a 12-gon can be done in two essentially different

ways, differing by a rotatiorl of 300• Figure 4.8 shows three 12-gons

dissected (by the dashed edgis) and the dissection of the '2-gon on

the left is a 300 rotation oj! the dissection of the two 12-gons on

the right. The opposite of Jhe process of dissection is the process

of fusion. We will speak of fusing a vertex of type 36 into a

 84

hexagon, or fusing a hexagon surrounded by vertices of type 3.4.6.4

into a 12-90n. The fusing and dissecting processes will help us

change certain 2-gonal tilings into other, known tilings.

One wcly of constructing many til.ings is by til.ing the pl.ane with

strips of polygons, as defined in section 3.3. That this is the onl.y

way of conf;tructing some of the 2-gonal tilings is demonstrated in

l lemma 4.4. The three kinds of strips (strips of squares, triangles,

and hexagons) which are defined in section 3.3 are demonstrated in

figure 3.8.

The fact that a tiling is 2-gonal does not, in general, imply

that it is 2-isogonal. Lemmaj 4.3 shows that this does happen in

certain cases, e.g. for some of the possible pairs of vertex figures.

In one case, a 2-gonal tiling can exist with either of 2 different

non-monogonal. edge types present, but if onl.y one of these appears,

the resulting tiling is forcet to be 2-isogonal.

Lemma 4.3: A 2-gona1 til.ing of type (36; 32.4.12), (36; 32.62), or

(33.42; 3.4.6.4) or one of type (32.4.3.4; 3.4.6.4) with no edges

of type 3.4.32.4 I 4.6.4.3 is 2-isogonal.. Each of these

2-is09ona1 tHings is uniTe.

Pf: The uniqueness assertion is proved by Krotenheerdt [1969], but is

also demonstrated by the flOllowing arguments.

1: (36; 32.4.12). If a verteK of type 36 is adjacent to a vertex of

type 32.4.12. then it is rradilY verified that it is adjacent to

only such vertices. Thus no two vertices of type 36 are adjacent

 85

and we may simultaneously fuse all vertices of type 36 into

hexagons. No two of these hexagons can be adjacent, else a vertex

where they meet is forced, in the original tiling, to be of type

36 and ad jacent; to =r=

I
vertices of type 36; contradiction. Thus

the new tiling has only vertices of type 4.6.12 (incident to these

fused haxaqona ) and, ~OSSiblY, vertices of type 32.4.12. But by

corollary 4.2, there are no 2-gonal (32.4.12; 4.6.12) tilings,

hence the tiling is the monogonal. isogonal (4.6.12) tiling.

Dissecting these hexagons gives us the unique 2-gonal, 2-isogonal

(36; 32.4.12) tiling of flgure 4.2, #6.

2: (36; 32.62). A similar argument here shows that every vertex of

type 36 is adjacent only to vertices of type 32.62, hence we may

fuse all of the vertices elf type 36• The new tiling will have

only vertices of type 63 and , possibly, 32.62• Again, corollary

4.2 shows that there can be no vertices of the latter type, so

this tiling is the isogonal (63) tiling. It is easy to see that

there is only one way to dissect a set of the hexagons in the (63)

tiling to get a 2-gonal (36; 32.62) tiling, and the result is the

2-isogonal tiling of figure 4.2, #7.

3: (33.42; 3.4.6.4). Every h1eXagOn is incident only with vertices of

type 3.4.6.4, and we WOUldj like to fuse all of these hexagons into

12-gons. This is POSSiblej only if no two such 12-gons would

overlap -- i.e. only if there is no edge incident with 2 different

hexagons. If there is suc an edge, then the configuration of

figure 4.6 is forced, the ltarred vertices must be incident with

 J J J J J J

,,
\ I

\ ••• \'
.,...... I

- - - ~ .," " I
r ,- - - -
, '[ , I
I ,
l ,
"'\ ...t- - --
'" \
, .•. " \
Figure 4.6: The configuration forced by an edge ;( \
I \ \
meeting 2 hexagons in a (33.42; 3.4.6.4) tiling. I \

,,,
I \
~
...
-, ,- .,. .•." '
I

- "!.. .,
,
,, ... ....
.•.
••.
••.••
.•.
I
.•.\
I
I
I

,...
\~ .•.
I"'"
,
I.
1
,

T
I

.- - --
.••••

I
I

\\ II '... _• •
I

t
\ I I .••.•• •
.. ',----
---- I •••

- - .... ....
I
~
.,.,;',
.•. I
"...
I"'... I
I \
\
,
, , .-
,I ,
'

Figure 4.7: A forbidden configuration in a Figure 4.8: A forbidden configuration in a

(33.42~ 3.4.6.4) tiling with fused disks. (32.4.3.4; 3.4.6.4) tiling. OJ
0\
 87

other hexaqons (which are surrounded by vertices of type 3.4.6.4)

and add:Ltional hexagons continue being forced. forcing the mono-

gonal, :lsogonal (3.4.6.4) tiling, contradicting our assumptions.

l Hence WE! may simultaneously fuse all the hexagons, and neighboring

tiles. iLnto 12-gons. Where 2 such 12-gons meet. there will be

vertices of type 3.122• Other vertices incident to these 12-gons

will be of type 32.4.12. Also. there may still be vertices of

l As~mme there is a vertex of type 33.42 in this new tiling,

Le. thcLt there is a pair of adjacent squares. If this pair is

embedded in an infinite strip of squares. then this strip is

bounded by strips of triangles, which must be bounded by strips of

squares, etc •• and the tiling is the isogonal (33.42) tiling

l (figure 4.1. 1t5). contradiLtion. Hence. this pair of squares is

embedded in a strip of squ.ares which terminates at a 12-gon. and

the configuration of figur/1e4.7 is forced. Consider the dissec-

tion of the 12-gon in the original tiling. Edge cannot have

l
€

been incident to a triangl~ (else vertex 1 could not have been

l completed).

dissection
hence it was ircident to another square.

of ·the 12-gon forces vertex 2 to have type 33.42;

But this

forces the dashed edges; a'nd thus forces vertex 3 to be incident

to four triangles. contraJiction.

We have shown that thle original tiling with hexagons fused

into 12-gons has no vertices of type 33.42. and hence has only

vertices of type. 3.122 a~ 32.4.12. Since (corollary 4.2) there

 88

are no 2-gonal tilingsof type (32.4.12; 3.122), this tiling must

be isog<mal. There is no isogonal (32.4.12) tiling, so this

tiling :Ls the isogonal (3.122) tiling. It is straight-forward to

see that there is a unique way to dissect these 12-gons to get a

2-gonal (33.42; 3.4.6.4) tiling, and that is the 2-isogonal tiling

of figure 4.2, #11.

4: (32.4.3,,4; 3.4.6.4) with no edges of type 3.4.32.4 I 4.6.4.3. As

wi th the previous case, ea.ch hexagon is surrounded by vertices of

type 3.4.6.4. If there we1re an edge which met two different

hexagons, then the confi ration of figure 4.6 would again be

forced, and the starred vertices could not have type 32.4.3.4

(else there would be an edge of type 3.4.32.4 I 4.6.4.3) so as in

the previous case, the is gonal (3.4.6.4) tiling would be forced,

contradiction. Thus, as tefore, we may fuse all of the hexagons,

with their neighboring tiles, into 12-gons. The new tiling could,
l conceivably, have vertices of types 32.4.3.4, 3.4.3.12, or 3.122,

but no other types are po sible. (The absence of edges of type

3.4.32.4 / 4.6.4.3 pravenns the existence of vertices of type

32.4.12.) This tiling mus/t contain vertices other than those of

type 32.4.3.4 (since the criginal tiling did); and since there is
l
no monogonal tiling of tYfe (3.4.3.12) and no 2-gonal tiling of

type (32.4.3.4; 3.4.3.12), there must be a vertex of type 3.122•

Assume this new tiling is not the monogonal (3.122) tiling.

In this case, there must te an edge E of type 3.4.3.12 / 122.3,

givlng us the bold edges af figure 4.8. If both of the vertices

 89

Ai are of type 3.4.3.12, then both vertices Bi are of type

32.4.3.4, and vertex C is incident with 3 consecutive triangles,

contradlction. Thus we may assume A, is of type 3.122, forcing

the othE~r solid edges in the figure. We now consider the edges of

the ori~Jinal tiling. By symmetry, we may assume the 12-gon T is

dissectEld as shown by the dashed edges, and this forces the

l dissectlons of the other two 12-gons (the dissection of T' is

forced by the absence of edges of type 3~4.32.4 I 4.6.4.3).

Vertex J~2 is forced, in the original tiling, to have type 32.4.3.4

(as shown), and B2 is then also forced to have this type. This

again gives 3 consecutive triangles at vertex C, contradiction.

We have shown that t~e tiling T which results from fusing the

hexagons into 12-gons is, in fact, the isogonal (3.122) tiling.

To recover the original tfling from T, we note that some 12-gon

must be dissected so that the triangles within it are incident to

the edges of simple type 122 in T (otherwise, the original tiling

would have edges of simpl~ type 42, contradiction). But then the

non-existence of edges of type 3.4.32.4 I 4.6.4.3 in the original

tiling forces all of the 12-gons to be dissected this way, giving

the 2-isogonal (32.4.3.4; 3.4.6.4) tiling of figure 4.2, #14. II

As mentioned above, several of the 2-gonal tilings occur only as

tilings by strips of pOlygon~, although there may be many possibil-

ities for how to arrange the ',striPS. These 2-gonal tilings are

Classified in the next lemma.

 90

Lemma 4.4: (1) The on1y 2-gona1 (36; 33.42) ti1ings are ti1ings by

strips (If squares and strips of triangles such that no two strips

of squares are adjacent and some pair of strips of triangles is

adjacent.

(2} The only 2-gonal (33.42; 44) tilings are tilings by

strips elf squares and strips of triang1es such that no two strips

of triangles are adjacent and some pair of strips of squares is

adjacent.

(3) (Sommerville, [1905) The only 2-gonal (32.62; 3.6.3.6)

tilings are tilings by strips of hexagons in which at least one

pair of adjacent strips meet along edges of simple type 62•

(4) The only 2-gonal (3.42.6; 3.6.3.6) tilings are tilings by

strips of squares and strips of hexagons such that no two strips

of squares are adjacent and no two strips of hexagons meet along

edges of simple type 62•

Pf: 1 & 2: Consider the edges of simple type 3.4 incident to a vertex

of type 33.42. The restrirtions on the vertex types force the

endpoints of these edges to have type 33.42. Repeating this

argument gives a sequence of edges of simple type 3.4 whose union

is a full line in the plane. On one side of this line is a strip

of squares, and side is a strip of triang1es. The

rest of (1) and (2) follows easily from this.

3: This result is stated by Sommervil1e [1905) and by Grunbaum and

Shephard ([1977a] and [1983]) without proof, so we give our own.

 91

Edgel3 of simple type 32 and 62 can only be incident with vertices

of type 32.62• As with the previous case, since the two edges of

these s:lmple types in the vertex figure 32.62 are collinear, these

two edg4~s are embedded in a full line which is incident only with

vertices of this type. This line has strips of hexagons above and

below i1~, and they force the rest of the plane to be tiled by such

strips.

4: The edgE~s of simple type 3.4 and 4.6 in the vertex figure 3.42.6

are collinear and can be incident only with vertices of this type,

so again the tiling is forced to contain a line incident only with

such vertices. This line is surrounded by a strip of squares and

a strip of hexagons. The Irest of the statement then follows. II

Some other 2-gonal tilings cainbe classified as specific modifica-

tions of other, simpler tilings. This kind of classification turns

out to be less informative than the classifications of the previous 2

lemma ta, because it is not aI/wayS easy to see how these modifications

affect the parameters v , e, Bind t , For the following lemma we need

vertices of type 3.4.6.4,

-,-
to formalize the idea of a d~Sk, which is a hexagon, surrounded by

tog'ether with the neighboring tiles. A

disk is thus equivalent to a 12-gon which has been dissected. In

lemma 4.5, we will modify certain tilings by fusing these disks into

12-gons or by rotating the disks.

 92

Lemma 4.5: (1) The only· 2-gonal (36; 34.6) tHings are those which

can be cbt.eLned from the isogonal (36) tiling by fusing some set

of vertices V into hexagons·where V has the property that the

distanc(~ between any two vertices in V is at least 3.

(2) The only 2-gonal (3.4.6.4; 4.6.12) tHings are those

which can be constructed from the isogonal (3.4.6.4) tiling by

fusing some set of non-overlapping disks.

(3) (Sommerville [1905]) The only 2-gona1 (3.42.6; 3.4.6.4)

tilings are those which can be constructed from the isogonal

(3.4.6.4) tiling by rotating a set of non-overlapping disks by

300• For a given 2-gonal tiling, the choice of disks to rotate in

order to get the isogonal (3.4.6.4) tiling is uniquely determined.

Pf: 1: Dissecting the hexagons of such a 2-gonal tiling into

triangles gives the isogon,al (36) tiling. To recover the 2-gonal

tiling fromthe triangula, tiling. it is necessary to fuse certain

vertices into hexagons in such a way that no two hexagons meet at

I
a vertex. The. condition that no two hexagons meet is easily seen

to be equivalent to the distance condition on the vertices being

fused.

2: Let T be a 2-gonal. (3.4.6fi 4.6.12) tHing. Each 12-gon in T is

incident only to vertices of type 4.6.12 and hence is surrounded,

alternately, by squares and hexagons. Each of these 12-gons can

be dissected so that the new squares are adjacent· to hexagons in

T. In this way all verti \es of type 4.6.12 are changed into
 93

verticels of type 3.4.6.4. The resulting tiling is thus the

monogonal, isogonal (3.4.6.4) tiling.

3: Levy [11391] and [1894J asks about this classification. The

solution is given by both Sommerville [1905] and by Grunbaum and

l Shephard ([1977a] and [1983) without proof, so we give our own.

Let T be a 2-gonal (3.42.6; 3.4.6.4) tiling. Every vertex of type

3.42.6 :InT must be adjacent, along the edge of simple type 3.4,

to a vertex of tyPe 3.4.6.4 (otherwise, we introduce a vertex of

l type 3.6.3.6), and this edge has type 3.6.42 / 4.6.4.3. Let e be

any edge of this type with endpoints v, of type 3.4.6.4, and v'.

of type 3.42.6. This edge figure forces the hexagon incident with

v to be contained in a disk, while none of the other hexagons

meeting this disk can be contained in disks. As a result, the set

S of all disks defined by such edges contains no overlapping

disks. Rotating anyone juch disk changes the vertices on its

I
boundary from type 3.42.6 (to type 3.4.6.4. Since the disks in S

are non-overlapping, and 1very vertex of type 3 ..2.6 is incident

:::9::s:,:n~:hr:::::::si::t:fo::: ::~::n3:.:::~t:::o:::
converts the tiling into the isogonal (3.4.6.4) tiling. The

uniqueness is established by noting that the set S is uniquely

determined; leaving any dil,Skin S unrotated will leave a vertex of

type 3.42.6 in the tiling; and rotating any disk not in Swill

create a new vertex of this type.

 94

Section..!:1l. t-hedral Tilings.

It is well-known (see section 1.2) that the only monohedral

edge-to··edge tilings by regular polygons are the three isohedral

tilings (figure 4.1, #1, 2, and 3). The full determination of

2-hedral tilings seems even more difficult than the classification

of 2-gonal tilings, and may be intractable. This relative difficulty

can be seen by considering our attempts to classify the 2-gonal

tilings in section 4.1, and noticing that the problem of classifying

the tilings with triangles and hexagons includes classifying the

made in the direction of thiS cJ.assification, and theorem 4.6

mentions these. Sommerville

I [1905J proves most of this theorem, and

does a partial classification of which combinations of vertices may

occur in 2-hedral tilings.

Theorem 4.6: (Sommerville, [[1905J) A 2-hedral edge-to-edge tiling by

I
regular polygons is either the isogonal (4.82) tiling, the

isogonal (3.122) tiling, al tiling by triangles and squares, or a

tiling by triangles and he agons. A tiling by triangles and

hexagons is equivalent to the isogonal (36) tiling with some set

of vertices V fused into ~exagons, where V has the property that

the distance between any two vertices in V is at least 2.

I
Pf: As mentioned earlier, a ~iling by regular polygons which contains

an octagon must be the isogonal (4.82) tiling. This can be

verified by noting that i such a tiling was not isogonal it would

 95

contain a non-monogonal edge incident to a vertex of type 4.82•

yet table 3 shows that no such edge exists.

Assume the tiling contains a 12-gon. The only vertex figure

which corrte Lns a 12-gon but is 2-hedral is the vertex 3.122•

There il3no non-monogonal edge figure which contains the vertex

3.122 and is 2-hedral, hence the tiling is monogonal and thus

isogonal.

If the tiling contains no 8-gons or 12-gons, then it contains

only triangles, squares, and hexagons. There are no 2-hedral

vertex figures which contain only squares and hexagons.

To prove the last sentence, note that if the tiling is a

tiling by triangles and hexagons, then dissecting the hexagons

into triangles gives the l.sogonal (36) tiling. To recover a

2-hedral tiling from the t~riangular tiling. it suffices to fuse

any set of non-adjacent vrtices into hexagons, II

To emphasize the lack o ' information we have on these two

remaining classes of 2-hedra , tilings, we mention some problems which

are very special cases of the:ir classification. First, in the case

of tilings by squares and triangles, if we further ask that every

.,
vertex be incident with at le!ast one triangle and one square, then

this is equiva~ent to the pr tblem mentioned earlier of the classifi-

cation of the 2-gonal (33.42; 32.4.3.4) tilings, which is still open.

With respect to tilingsby t'iangles and hexagons, the classification

above is not particularly us :ful for investigating many of the

 96

questions that come up about such tilings. For example, GrUnbaum and

Shephard [1983] list as an open problem the determination of the

pairs (t, h) such that there is a tiling by triangles and hexagons

with t orbits of triangles and h orbits of hexagons. No particular

insight into this problem seems to arise from theorem 4.6. Using a

slightly different view of the construction of some of the tilings by

triangles and hexagons, however, gives the following solution.

Theorem 4.7: There is an edge-to-edge tiling by triangles and

hexagons with t orbits of triangles and h orbits of hexagons if

and only if the pair (t, h) satisfies one of the fOllowing

I
properties:

(1) ( t,; h) ••• (1, 0);

(2) (t, h) (0, 1);

(3) t > 0 and h > O.

Pf: If either t I
or h is 0, then the tiling is monohedral hence, as

remarked earlier; isohedr 1. We may thus assume t and h are both

positive. lie remind the reader of the "co1.umns" of triangles and

hexagons used in the construction of the class of examp1.es of

figure 3.7. Figure 3.7 is a tiling with 1 column of triangles

alternating with 4 cOlumnS of hexagons. Figure 4.9 shows two

I
tilings, one which a1.tern~tes 2 co1.umns of triang1.es and 4 co1.umns

of hexagons, and the othe with 6 columns of triangles and 2

columns of hexagons. In grneral, if we use 2t (or 2t-1) columns

of triangles, there will e t orbits of triangles (regardless of

 97

Figure 4.9 (a)

I '\ /
I
\
/ "
'\
,/
/
\
/ f\ /
\ ,/ \ ,/
/ '\ / '\
\
I
, ,/ \ / J
/ \
J \ /
:
'\
J \
/ I
!r J
/ / '\ I
\
/ \ \ J
\ / / /
\ ./ \
\ I \

" v\/ v v \/ v

Figure 4.9 ( b)

Figure 4.9 : Examples of tHings with varying numbers of orbits

of tr~angles and hexagons. Part (a) has 1 orbit of
triangles abd 2 orbits of hexagons while part (b)
has 3 orbit1s of triangles and 1 orbit of hexagons.
 98

how many co~umns of hexagons are used); and if we use 2h (or 2h-1)

co~umns of hexagons, there will be h orbits of hexagons. This

class of examples thus gives tilings having any pair of parameters

(t, h) with both t and h > O. II

This last resu~t is worth emphasizing, and we have shown several

sample ti~ings, because Grunbaum and Shephard [1983] imply that this

result is false. As one of their exercises (see section 1.2), they

ask the reader to find a pair (t, h) for which no tiling exists with

those numbers of orbits of triang~es and hexagons. Presumably, they

did not have in mind one of the trivial pairs, such as (5, 0).

It is worth remarking th,at the equivalent question for tilings

by triangles and squares has ,a similar answer. This result was known

at least to Grunbaum and Shephard [1983], who leave it as an exercise

for the reader (Ex. 2.3.6). lince it seems appropriate for this

section, we mention the solution. A tiling with 2s (or 2s-1) strips

I
of squares alternating with t strips of triangles gives a tiling with

s orbits of squares and t orb"ts of triangles (s and t > 0). Figure

3.5 shows an example with s •.• and t = 4. Thus, there is a tiling

by squares and triangles with s orbits of squares and t orbits of

triangles if and on~y if both t and s are> 0, or if (t, s) = (0, 1>

or (t , s ) •.•(1, 0).
 99

Section 4.3: k-toxal Tilings.

As remarked earlier, it is an established fact that a monohedral

tiling is isohedral and that a monogonal tiling is isogonal. It is

not surpri~iing, then, that monotoxal tilings are also all isotoxal.

We suspect this has not been noted earlier only because the concept

of e-toxal seems not to have been previously studied. What is a

little surprising, in view of the difficulties involved in the clas-

sification of 2-hedral or 2-gonal tilings, is the fact that 2-toxal

and 3-toxal tilings are also 2-isotoxal and 3-isotoxal, respectively,

and can be fully classified. This is established in theorem 4.8, and

the concurrent classification of the 2- and 3-isotoxal tilings in

this theorem answers two questions of GrUnbaum and Shephard [1983].

The 1-, 2-, and 3-isogonal tilings referred to in theorem 4.8 are

pictured in figures 4.1, 4.2, and 5.1 respectively.

Theorem 4.8: If T is an e-toxal edge-to-edge tiling by regular

pOlygons with e ~ 3 then T is also e-isotoxal. Furthermore:

I) The only isotoxal tillingS are the four isogonal tilings:

(36); (44);1 (63); and (3 •6 • 3 •6 ) •

II) The only 2-isotoxal tilings are the four isogonal tilings:

(3.4.6.4); and

III) The only 3-isotoxal tilings are the three isogonal tilings:

and (4.6.12);

the five 2-isogonal ilings:

 100

and

and the two 3-isogonal tilings:

and

Pf: Inspection of the 11 isogonal tilings of figure 4.1 shows that

the first sentence of this theorem hOlds for these tilings and

that the enumerations of I, II, and III are complete with respect

to the isogonal tilings. Thus we may assume that the tiling

contains at least one of the non-monogonal edge figures of table

I
3. Furthermore, it is easily checked that the tilings of It III,

I
and III are. in fact. 1-, 2- and 3-isotoxal; so we need only check
I

that there are no other tilings.

Assume e=1. This case is established by proving the stronger

statement that any tiling which contains only one simple edge type

is isotoxal. This statemJnt hOlds for the isogonal tilings, and

there are no non-monogonaj edge figures which contain only one

simple edge type, so it hTdS in general.

In the case where e >11. we frequently use the fact that the

number of simple edge typelS cannot exceed the number of full edge

types; and that if these numbers are equal, then every simple edge

type corresponds to a unique full edge type. Specifically:

I
Fact 1: An e-toxal tiling Icontains at most e simple edge types,

and if it contains e s mple edge types, then any two edges with
 101

the same simple type must have the same full type.

Assume e=2. From table 3, the only non-monogonal edge types

that contain at most two simple edge types are:

In the first two cases, if the edge figure is extended by applying

fact 1 to the edges of simple type 32, a contradiction is quickly

reached. In the third case (see figure 4.10), since € and €I must

have the same full type (fact 1), the dashed edges are forced and

edge x has type 32.4.3.4 / 4.3.4.32• It is not possible, however,

to extend the drawing so that edge y has the same full type as

In the final

1 to the edges of

simple type 3.6 soon yields a situation in which two edges of

I

simple type 32 have differEmt full types.

Assume e=3. The non-J1lonogonaledge figures which contain 4

or more simple edge types are indicated by a dashed line in the

column of table 3 labeled "Thm. 4.8". This leaves 23 non-

monogonal edge figures WhiJh might occur in a 3-toxal tiling.

Those edge figures marked Jith a ,,*" in this column contain 3

simple edge types and can be ruled out by extending the edge
I
figure by applying fact 1 to the edges which have the same simple

type as the original edge. In some cases, the resulting portion

of the tiling cannot be filled in at certain vertices, while in

 102

Figure 4.10 Figure 4.11

, , ....
,,
\
" I
\ I
\,. "
',," ' .•••.
1
I'~ ~ ,. "\ ,
I', ,'" \
I 1\ /\ \
I / \ /' '"
" I \ I \ ,,.
',I \/ \",
---- ----
Figure 4.12 Figure 4.13

---,.,
,,-- --'
I
I \
\ \
I I \
I 1 ,
, 1\

1
I
" \
\
\
,
, ,
I
I ,
\
,
1

Figure 4.14 Figure 4.15

Figures 4.10 - 4.15: Edge figures and portions of tilings forced by

them in the proof of theDrem 4.8. In each case, the edge being
studied is marked as a bJld edge.
 103

other cases two edges of the same simple type but with different

full types are forced. The edge figures labeled A, B, and C in

table 3 also cannot occur in a 3-toxal tiling, but the arguments

are slightly less trivial. These edges all contain 3 simple edge

types, and we can apply fact 1.

A: 34.6 I 62.32• This edge figure contains one edge of simple

type 32 which is incident to a vertex of type 32.62 and another

edge of the same simple type which cannot be incident to such a

vertex.

B: 32.42.3 I 3.4.3.4.3. This edge figure contains one edge of

simple type 3.4 which is incident to a vertex of type 33.42 and

another edge of the same simple type which cannot be incident

to such a vertex.

C: 4.33.4 I 44. This edge figure, and fact 1 applied to the edges

of simple type 42, forces the edges of simple type 3.4 either

to have type 33.42 I 4.32.4.3 or to have type 33.42 I 4.3.4.32;

neither of which is poS~ible in a 3-toxal tiling.

This eliminates all blut 8 of the non-monogonal edge tyPes,

and these 8 all occur in 3-toxal, 3-isotoxal tilings. To classify

the 3-toxal tilings containing these edges we will frequently use:

Fact 2: Given any three full edge types which have different

simple edge types, there is at most one 3-toxal tiling

containing all of those full edge types.

Pf: Start with anyone of rhe 3 edge figures and extend the tiling
 104

successively by filling in the vertex figure of any vertices

whose types are forced. At any step in extending the tiling,

there will be a vertex v whose vertex type is not yet

determined, but which is adjacent to a vertex of known type.

The ledge connecting them is then of known simple type, hence

its full type is forced, so the vertex type and orientation of

V is also forced. The extension of the tiling thus continues

to be forced at each stage, giving at most one such tiling.

Of the 8 remaining full edge types, 4 of them (labeled 1 to 4

in table 3) contain 3 simple edge types, and we consider these

four first.

1) 3.4.3.12 / 122.3. This edge figure contains edges of each of

the 3 simple types wheJe the full edge type is determined by

the figure. Fact 2 forces the 2-isogonal (3.4.3.12; 3.122)

tiling of figure 4.2. 4~16.

2) 6.32.6 / 63• See figure 4.11. Fact 1 applied to the edges of

simple type 62 forces 1:he edge figure of the edge a to extend

J
as shown. Fact 1 apPlled to the edges of simple type 3.6 then

forces vertices u, u', and u" to have the same vertex type. If

u and u' have type 3.6t306. then edges band c have the same

simpl.e type, but diffe2~ent ful.l.types. Thus, by considering

the possible full. edge types remaining in our enume.ration, we

This forces v r
see that edge b must bE!monogonal., so u ' and u" have type

have type 36 (or else it i. incident

 105

with an edge of type 32.6.32 ! 3.62.3, which was ruled out

earlier). The three full edge types are now determined. and

fact 2 forces the 3-isogonal (36; 32.62; 63) tiling of figure

5.1, #19.

Romark: Since any non-isogonal tiling with a vertex of type

63 contains a non-monogonal edge meeting such a vertex, and since

the edge figure just considered is the only non-monogonal edge

meeting such a vertex, we may assume in the following that there

are no vertices of type 63• Thus, any edge of simple type 62 must

3) 36 ! 3.62.3. This edge figure (and fact 1 applied to the edges

of simple type 32) forces the full edge type of the edges with

simple type 32 and 3.6. Hence. by the remark. all 3 edge types

are fully determined and the 2-isogonal (36; 32.62) tiling of

figure 4.2, #7, is forded.

4) 32.62 ! 6.3.6.3. This edge figure (and fact 1 applied to the

edges of simple type 3.6) forces the full edge type of the

edges with simple type 32 and 3.6. Hence, by the remark, all 3

edge types are fully determined and the 2-isogonal

(32.62; 3.6.3.6) tiling of figure 4.2, #15, is forced.

The four remaining edge figures each contain only 2 simple edge

types, and the tilings lie:ted in the theorem that contain them all

have 2 edges of the same JimPle type but of different full types.

TO deal with this situation, we need a strengthening of fact 2:

 106

Fact 3: Assume we have 3 different full edge types with the

following property: Each edge in the vertex figure of an

endpoint of one of these 3 edges can be extended in only one

way so as to have one of those three full types. Then in this

case there is at most one 3-toxal tiling containing those full

edge types.

Pf: The proof given for fact 2 also proves this, and fact 2 is a

special case of fact 3.

The wording of fact 3 is important: although it implies that the

full type of each edge of a vertex figure 1s forced, that property

alone is not sufficient tO force a unique tiling. For example, if

I
we know that the central edge of simple type 32 in the vertex

figure 34.6 is supposed to be extended so as to have full type

I
32.6.32 / 33.6.3, there aze still two different ways to carry out

this extension.

5) 34.6 / 6.3.6.3. The figure is shown by the solid edges in

figure 4.12. Edges band c cannot have the same full type, so

a, b, and c represent tre 3 full edge types. Any edge of

simple type 3.6 must have the same full type as a, so a and a'

have the same type and the dashed edges are forced. Now c'

cannot have the same type as b, so c and c' have the same type.
I

Vertex v cannot be of type 34.6 and still have c and c' be of

the same full type, so V must be of type 36• It is then

I

straight-forward to ver1lfY that fact 3 al'plies to these three

n
 107

edge types. and the 3-isogonal (36; 34.6; 3.6.3.6)2 tiling of.

figure 5.1. #5. is forced.

6) 36 / 3.4.3.4.3. The solid edges of figure 4.13 are part of the

edge figure for an edge a of this type. Experimentation shows

that the only way to surround the square by vertices of type

32.4.3.4 without introducing 3 different edges of simple type

3.4 is the configuration indicated by the dashed edges. There

are now two edges of simple type 3.4, and these are distin-

guishable in the vertex figure of 32.4.3.4 as follows: the

vertex figure contains a pair of adjacent triangles, and the

edges of simple type 3.4 incident with this pair are of

different full type than those that are not. In this case,

knowing the full edge type uniquely determines how the edge can

be extended. The other edges in the vertex figures of the

I
endpoints are easily seen to satisfy fact 3, hence the

2~isogonal (36; 32.4.3.4) tiling of figure 4.2, #5, is forced.

7) 36 / 32.6.32• This edge figure is shown as the solid edges in

figure 4.14. Edges a and b cannot have the same full type, so

a, b, and c represent the 3 full edge types. It is impossible

to fill out the vertex figure of v so that edges d and b have

the same type, so d must have the same type as a, the dashed

edges are forced, and the 3 edges a, b, and c satisfy fact 3.

Consequently, the 3-isdgonal (36; 34.6; 3.6.3.6)2 tiling of

figure 5.', #5, is for ed. (Note: This tiling is the same as

the one forced in case 5 above, which explains why we have 8

 108

cases in the proof. but only 7 non-monogonal tilings in the

theorem.)

8) 36 I 33.6.3. Referring to the edge figure shown as the solid

edges in figure 4.15, we note that edges a and b cannot have

the same full type, so the edges a, b, and c represent the 3

full edge types. As with the last case. d and b cannot have

the same full type, hence d has the same type as a and the

dashed edges are forced. Edges a, b, and c now satisfy fact 3,

so the 2-isogonal (36; 34.6)2 tiling of figure 4.2, *2, is

forced. II

To demonstrate that th,aorem 4.8 is the strongest theorem of this

form, note that the tiling '~f figure 4.16 is a 4-toxal tiling which

is not 4-isotoxal; and in fact is not e-isotoxal for any finite e.

\/\/\/\/\ \/ /\/\/\/

Figure 4.16: An example of a 4-toxal tiling by regular polygons

which is not 4-isotoxal. Representatives of the four edge types
are marked as bold edges.
 109

Chapter 5: Global Regularity in Tilings by Regular Polygons.

Having looked at the consequences of assuming local regularity

in a tiling, we now consider the various kinds of global regularity

conditions that can be imposed. Again, all tilings are assumed to be

edge-to-edge tilings by regular polygons. The 3 sections consider,

respectively, k-isogonal, k-isohedral, and k-isotoxal tilings. The

three sections also consider (respectively) the questions of vertex-,

tile-, and edge-homogeneous tilings.

In most cases, the enumeration of a class of tilings requires a

tedious case-by-case analysis. While we cannot hope to completely

avoid such tedium, a principal goal is to simplify the analyses

when.ever possible by utilizing previous classifications. Ini tially,

the main classification at hand is Krotenheerdt's classification of

the vertex-homogeneous tilings, which includes the 2-isogonal

tilings. While this is usefJ.l in some situations, especially in

section 5.3, other results r1quire the classification of the

3-isogonal tHings, e.g. section 5.2. Consequently, we begin our

investigations with these ti~ings. We warn the reader that most of

the tedium in tOhis chapter odcurs in this classification.

 110

Section 5.1: Vertex Regularity.

The 2-isogonal tilings, all of which are vertex-homogeneous,

have been classified by Krotenheerdt [1969), and these are shown in

figure 4.2. Krotenheerdt also determined the other vertex-

homogeneous tilings (Krotenheerdt [1970a and 1970b). Since such

case-by-case classifications are easily subject to error, we have

done an independent enumeration of these tilings, and our work

verifies his results.1 The papers of Krotenheerdt do not give

drawings of the vertex-homogeneous tilings. Since these tilings will

be of interest to us, and some of the results require inspection of

these tilings, theY are Pictired in figures 5.1 to 5.5. These

figures show, respectively, the 3-, 4-, 5-, 6-, and 7-isogonal

vertex-homogeneous tilings. There are no v-isogonal, vertex-

homogeneous tilings for v > 7.

Since Krotenheerdt has Classified the 3-gonal, 3-isogonal

tilings, to determine all of the 3-isogonal tilings it is only

necessary to classify those Jhich are 2-gonal. The work of section

4.1 is fundamental to this classification, which is achieved in the

fOllowing theorem.

1 There are, however, tw10 typographical errors in his papers:

In [1970a), tiling #14 has tYipe (36; 33.42; 32.4.3.4; 32.4.12), not
(36; 32.4.12; 32.4.3.12; 32J.12) -- the third vertex in his list
does not exist; and in [1970b), Hilfssatz III, tiling #6 (corre-
sponding to our figure 5.1, 25) has type (c; f; h), not (b; f; h).
 111

6 4 2 2 6 4 2 2
1: (3 ; 3 .6; 3 .6 \ 2: (3 ; 3 .6; 3 .6 )2
'i

6 4 2 2 6 4
3: (3 ; 3 .6; 3 .6 )3 4: (3 ; 3 .6; 3.6.3.6)1

Figure 5.1: The 39 3-isogona1., vertex-homogeneous edge-to-edge

tilings by regular polygons. Repres ent at ave sets

of vertices are marked by solid dots.

 112

5:

I
9: (36, , 33 ,2,
• '"+ , 3.4.6.4)l 10:

F:gure 5.1 (cont.)

I .

I .
 113

12:

13:

6 2
15: (3 ; 3 .4.3.4;

Eigure 5.1 (cont.)

l
 114

6 2 18: (36; l.4.12; 4.6.12)

17: (3 ; 3 .4.3.4; 3.4.6.4)2

l
21:

5.1 (cent.)
 115

2 2
26: (3 .4.3.4; 3.4 .6; 3.4.6.4)

2 2 2 2
27: (3 .4.12; 3.4.3.12; 3.12 ) 28: (3 .4.12; 3.4.6.4; 3.12)
,
F1 gure 5.1 (cont.)
j
 '16

2 2 2
29: (3 .6 ; 3.4 .6; 3.6.3.6)1

2 2 3.6.3.6;63) 2
32: (3.6;

33: (3.4.3.12; 3.4.6.4;

2
.12 ) 34: (3.42 .6;
.
3.4.6.4; 44)

Ftgure 5.1 (cont.)

 117

35: (3.42 .6; 3.6.3.6;44) 1 36: ( 3.42 .6; 3.6.3.6;44) 2

38: (3.4 2 .6; 3.6.3.6;44) 4

I .

2
39: (3.4 .6; 3.6.3.6; 4.6.12)

Fi~e 5.1 (concluded)

n
 118

6 4 322
1: (3 ; 3 .6; 3 .4 ; 3.4 .6)1

Figure 5.2: The 33 4-isogonal, vertex-homogeneous edge-to-

edge tilings by regular polygons. Representative
sets of vertices are marked by solid dots.

l
 119

/ \ 1\ / \

> X X X X
1 / /\/

X
X
\ ,/
/ 1 \ 1\

X X
x X
\ \ /
1.1' .1\ \ 1 \

X X x X 'X

X X x x X
/'\ '\ '\
,/ ,/
/\
\/'
/>; /
'\/\
'\ \
" '\
\
/ " \/\ '\/

Jour. 5.2 (cont.)

 120

9: (36,. 3 4 .6; 3. 42 .6; 3 ...6 3 6) 2 2

10: 3 .4.3.4;
2
3 .4.12)

F gure 5.2 (conto)

I
 '21

2 12: (36,. 3.;

3 42 3 2 .4.3.4.
11: (36,_33.42,_3 . 4 .3. 4 ;
3.4.6.4)1 3.4.6.4)2

\ /' ))\ '\ /\)\1 r-; '\1\ '\

I" ' '\ "\./ , "

-, "\"
\
.
I I
x: '1 X X X xX X X X
I
,/\

XXXXXX'XXX ';x
I I I I
\

I I
<~XXXXXXXX
IT I
1\7'\ ,/ '"
\
I I I I I I I I I I I I I I I

13 :

Figure 5.2 (cont.)

 122

I
I

6 2 2
15: (3 ; 3 .4.3.4; 3 .4.12; 16: (36; 32.4.3.4; 3.4.3.12;
3.122) 3.122)

17: (36; 32.4.3.4; 3.42.6; 18:

3.4.6.4)

~igure 5.2 (cont.)

 123

19: 20:

21 : 22: (33.42; 32.4.12;

3.4.3.12; 44)

Fi re 5.2 (cont.)
· r-"

124

3 2 2 2 2 63)
24: (3.4; 3 .6 ; 3.4 .6; 1

25 : (33.42;
2 2
3 .6 ; 3.42.6; ()Z 26:
2 2
(32.4.3.4; 3 .6 ; 3.42.6;
3.4.6.4)

Filgure 5.2 (cont.)

 125

27:

29: (32 .4.12; 3.4.3.12; 3.4.6.4; 4.6.12)

Figure 5.2 (cont.)

I
 126

2 2
30: (3 .6 ; 3.42.6; 31:
2
(3.6;
2 2
3.4 .6;
3. 6. 3 • 6; 44) 1 3.6.3.6; 44)2

x X X X X X X X X X x X X X X X X X X X
xY.XXXXY.Y.Y.X X X X X y X X Xx X
I I I

I I I
I I I
X X X :x :x )< Y.X X :x X X X X :x :x X
X X X X -::. X Y.X X X X X 'x X X :x :x X
I
I
I
x'XXXXXX7'XX X X X X X X X X XX
X:x7'XXXXXXX X
I
X X X X X X X X-X:
11111 I I I I I I iiTl I I I I I I I I I I I 11111
I ! I I I I I I I I 1I
,Till I I I I I II I I"" I I I II I I fl I I I I I I I I I I I I IITI!'

2 2 2 2
32: (3 .6 ; 3.4 .6; 33: (}2.62; 3.4 .6;
3.6.3.6; 44)3 3.6.3.6; 44) 4

Figu e 5.2 (concluded)

 127

I I I I I I

I I I I

~ ~ ~~~ ~~~ ~~1<

I I I I I I I I I

6 1+ 3 2 32.62,_
1: (3 ; 3 .6; 3 .4 ;
2
3.4 .6)

Figure,5.3: The fifteen

i

5-isogonal vertex-

homogeneous edge-to--edge

tilings by regular

polygons.

6 1+ 3 2 2
3: (3 ; 3 .6; 3 .4 ; 3.4 .6; 4.6.12)
 128

6 4 2 2 6 4 2
4: (3 ; 3 .6; 3 .• 6 ; 5: (3 ; 3 .6; 3.4 .6; 3.4.6.4:
3
3.6.3.6; 6 ) 3.6.3.6)

6 4 2 6 1+ 2 44)
6: (3 ; 3 .6; 3.4 .6; 7: (3 ; 3 .6; 3.4 .6; 3.6.3.6; 2
3.6.3.6; 44)1

Filgure 5.3 (cont.)

 129

6 4 2
8: (3 ; 3 .6; 3.4 .6; 9: (3 6 ; 3 4 .6; 3.4 2 .6; 3.6.3.6; 44)
4
3.6.3.6; 44)3

6 3 2 2 2 6 3 2
10: (3 ; 3 .42; 3 .6 ; 3.4 .6; 11: (3 ; 3 .42; 3.4 .6;
3.6.3.6) 3.6.3.6; 44)

Figure 5.3 (cont.)

 130

6" 4 3"
12: (3 ; 3 .4.3.4: 3.4 .6; 13: (3 .6: 3 .4 ; 3 .4.3.4;

l 3.4.6.4; 3.6.3.6) 3.4'.6; 44)

l
l 14:
2
(3 .4.3.4;
2
3 .4.12;
2
15: (3 .4.3.4;
2 2 2
3 .6 ; 3.4 .6;
2 3
3.4.6.4; 3.12 ; 4.6.12) 3.4.6.4; 6 )

(conc~uded)
 131

6 It 3 2 2 6 It 322
1: (3; 3 .6; 3.4; 3 .4.3.4; 2: (3 ; 3 .6; 3 .4 ; 3 .4.12;
2 2 2 2
3 .6 ; 3.4 .6) 3.4 .6; 3.6.3.6)

Figure 5.4: The ten 6-1sogonal vertex-~omogeneous

edge-to-edge tilings by regular polygons.

 132

3'.
3 2 2
4. 32 62.,
(36,. 3 .4 ,. 3 4 3 ., .. . 2
3 .4.3.4; 3.4
2
.6;
2 1+
3.4 .6; 3.4.6.4) 3.4.6.4; 4 )

I I I I I I I I I I I I I I I I I I
< } ): ):
X x x X X >
< I> ')(

I I I
')(
X
I
:x x
I
X
I
>
I I I I I I
r-:r, f\ 7',7'\ /\ '\ '\ ,\ '\1\ '\ '\ '\
./ 1 \ 7 ,I ././
I I I I I T I I I I I I I
I I I I T I TTT I I I I I I
X X X X X X X X X

l X X Xx X X X X X
I I I I I IT TT I I I I I I
I I I I IT I I I I I I I I I
1\/\11""/\/ '\1\1-/\1,'\1\1\1 . '\
<r.». ,/\/\/ '\1\1\1\1 \1\1\/\1\ '\ \
I I I II I I I I I I I I
I I I I I I T-1 IIf I I I I I I i

6 3 2 2 2 2
5: (3 ; 3 .4 ; 3 .6 ; 3.4 .6;
4
3.6.3.6;/4 )

F1rre 5.4 (cont.)

 133

4 3 2 2 2
7: (3 .6; 3 .42; 3 .4.3.4; 3 .4.3.4; 3.4 .6;
2 2 2
3 .6 ; 3.4 .6; 4.6.12) 3.4.6.4; 3.6.3.6)

32222
9: (3.4; 3 .4.3.4; 3.6 i
4
·3.4 .6; 4 ; 63)
2 4
3.4.6.4; 3.6.3.6; 4 );

Figu) e 5.4 (concluded)

 134

6 1+ 3 2 2
1: (3; 3 .6; 3.4; 3 .4.3.4;
2 2
3 .4.12; 3.4 .6; 4.6.12)

Figure 5.5: The seven

7-isogonal vertex-

homogeneous edge-to-edge

tilings by regu1ar

polygons.

6,. 1+ 3 2 2
3: (3 3 .6; 3 .4 ; 3 .4.3.4;
2
3.4 .6: 3.4.6.4; 3.6.3.6)2
135
 136

'\/\/ ,/\/\/\ ' \/ \ / \/

x \ '\/
7\
\ '\. 'V \/v
/ .1\ '\ '*-
'\/ \1 '\/ '\ .1\/
J\ '\ / ,/\ \/ =:r:
,A/\ \I ,/
x '\7\/
J
A ~
'\I J '\ \1 '\1
J\ ,1\ .A
'\/ \/
\ I 7\ 7'\ ~
./\1\ '\/ 1\/
l\J\7I l\ / .1\ ..1 .r.
>(
/\/\
> .1\ 1\
7\ /\ I' /\
'\/ \/\1\/\1\
\ '\-/
I' 7'7\.7 /\
\7\.7\1' j\
X-
, "r; '\ /\ 1\ / 1\/\ /\ '\ ,

,kf:Y'~ "" k [>k

./I'S/ /
,;>'(,., t-
/kl/
r-, /r--,.v <, V~
;; -.....1/
kl)k Dk
r-,
~vf' ./

K[)
.i-r-. I./f....v
V r-, V ./r--... ./
'C
t--..
""r<l)k
> [)k
~r; V
f'k?)K
1/1'-1/
'X
;1<..
"::.<
Vl'-I/
",. I ",;1<..
ykl>
,I'...
I'-V

Figure 5.6: The 22 3-isogonal, 2-gonal edge-to-edge

tilings by regular polygons. These are the

3-isogonal tilings which are not vertex-
homogeneous. Representative sets of vertices
are marked by solid dots.
 137

8:

4 4
10: (3 .6; 3 .6; 3.6.3.6)1

Figlre 5.6 (cont.)

 138

4 4
11: (3 .6; 3 .6; 3.6.3.6)2 12:

3 2. 2 2
13 : (3 .4 ; 3 .4.3.4; 3 .4.3.4) 14:

..•. ~ ;;l-
I--
K /'
<, I)l-
I- f(
I-- f(
/
I) -
I- k:'
/'
<, -
')
/
<, )-
I-
K /
)-
I--
f( /'
<, )-
I- f(
I--
K
,/
<,
>-
I-
<
/'

>-
-
:..- <"" 1
,/
<, >
15: 16:

Figure .6 (cont.)
 139

, 1/
)-
:: <;
1/
0 r-
~ l./
t/ r-
/
1/
t/ r-
~ l./
./
I' D I-
~
1/

l./
-
I)
)-
:: <,
/
>-
~ L>
<, )-
::: L>
<, )r-
-'

17: 18:

2 2 2
19: (3.4 .6; 3.4.6.4; 3.4.6.4) 20: (3.4 .6; 3.4 .6; 3.6.3.6)

2 2
21: (3.4 .6; 3.6.3.6; 3.6.3.6)1 22: 0.4 .6; 3.6.3.6;,3.6.3.6)2

Figure 5.6 (concluded)

 140

Theorem 5.1: The only 3-isogonal edge-to-edge tilings by regular

pOlygons are the 39 3-gonal tilings of figure 5.1 and the 22

2-gonal tilings of figure 5.6.

Pf: We need only consider the 2-gonal tilings. The possible vertex

types for this tiling are restricted by corollary 4.2. Three of

these 16 possibilities are eliminated by lemma 4.3, which says

that certain 2-gonal tllings must be 2-isogonal. Furthermore,

any (32.4.3.4; 3.4.6.4) tiling either has an edge of type

3.4.32.4 I 4.6.4.3. in which case it is at least 4-isogonal

(theorem 4.1), or else has no such edge and is then 2-isogonal

(lemma 4.3), hence this case is also eliminated. It is straight-

forward to verify that the only tilings by strips such as those

referred to in lemma 4.4 which give 3-isogonal tilings are the

ones included in figure 5.6, specifically numbers 5-8, 14-18. and

20-22. We thus need only consider 2-gonal tilings of the

following 8 types:

(36; 32.4.3.4); (3.42.6; 3.4.6.4); (3.4.6.4; 4.6.12);

(3.4.3.12; 3.122); (34.6; 3.6.3.6); (34.6; 32.62);

One technique for classifying these tilings is to modify them to

get a v-isogonal tiling with v < 3; using the classifications of

these tilings (figures 4.1 and 4.2); then re-constructing the

original tilings from them. This technique is used in the first 3

I
cases. In these cases it ilsnecessary to verify that the sYmmetry
 141

group of the new ti1ing is no smaller than the symmetry group of

the original tiling (to guarantee that no single orbit of vertices

splits into two orbits). Because the sets of hexagons or disks

involved in our modifications are uniquely determined by the

geometry of the original tilings, these sets are "recognized" by

the symmetries of the original tiling, hence all such symmetries

are in the symmetry group of the modifiedtiilng also.

Case 1) (36; 32.4.3.4). If a vertex of type 36 is adjacent

to a vertex of type 32.4.3.4, then it is readily verified that it

is adjacent to only such vertices. Consequently, no two vertices

of type 36 are adjacent, and we may fuse all of these vertices

into hexagons surrounded by vertices of tyPe 3.4.6.4. Fusing

these vertices does not change the symmetry group of the tiling.

Since at least one orbit of vertices of type 36 has been removed,

this new tiling is either the monogonal, isogonal (3.4.6.4) tiling

(figure 4.1, #7) or the 2-isogonal (32.4.3.4; 3.4.6.4) tiling

(figure 4.2, #14). Dissecting the hexagons in the former tiling

gives the 2-isogonal (36; 32.4.3.4) tiling (figure 4.2, #5) while

dissecting the hexagons in the later ti1ing gives the 3-isogona1

(36; 32.4.3.4; 32.4.3.4) tiling, figure 5.6, #9.

Case 2) (3.42.6: 3.4.6.4). Denote the original tiling by T.

Lemma 4.5 says there is a set of disks in T which can be rotated

to eliminate all vertices rf type 3.42.6. Fusing these disks also

eliminates these vertices, and forms a 2-gonal (3.4.6.4, 4.6.12)

 142
l
tiling TI. Since the symmetry group of T permutes these rotated

disks, no symmetries are lost from this fusing operation. Since

at least one orbit of vertices of type 3.4.6.4 inside the fused

disks has been removed, T' has at least one fewer orbit of

vertices than T and is either 1- or 2-isogonal. This implies that

T' is one of the following: the isogonal (3.4.6.4) tiling; the

isogonal (4.6.12) tiling; or the (unique) 2-isogonal

(3.4.6.4; 4.6.12) tiling. Since T contained vertices of type

3.42.6, T' must contain 12-gons, hence is one of the later two.

Since the 12-gons in T' were fused from the hexagons in T that are

surrounded by vertices of type 3.42.6, to recover T from these

possibilities for T', each 12-gon must be dissected in this

manner. The former case gives rise to the (3.42.6; 3.4.6.4)

2-isogonal tiling (figure 4.2, #17) while the later case gives

rise to the 3-isogonal (3.42.6; 3.4.6.4; 3.4.6.4) tiling, figure

5.6, #19.

Case 3) (3.4.6.4; 4.6.12). By lemma 4.5, any such tiling T

arises from the isogonal (3.4.6.4) tiling by fusing some set of

disks into 12-gons, so T is equivalent to figure 5.7a with certain

additional disks fused.

Claim 1: Every hexagon in T is adjacent to at least two 12-gon5.

Pf: Otherwise, let H be a hexagon in T adjacent to only one

1 12-gon T. Then:
 143

Figure 5.7 (a)

Figure 5.7 (b)

Figure 5.7: TheorE>m5.1, case 3.

 144
l
a) Let £ be the edge incident with both Hand T. then T

mus·t have as a symmetry a refl.ection perpendicul.ar to £ which

fixes £ (el.se no symmetry woul.d take H onto itsel.f, and its 6

vertices would fal.l in 6 orbits); and

b) All hexagons in T are equivalent (else the vertices of H

would fall in three orbits, and the vertices incident with

hexagons not equival.ent to H woul.d fal.l.in other orbits), and

hence al.l hexagons are adjacent with exactly one 12-gon.

NOw, referring to figure S.7a, let H1 be a hexagon adjacent

to only one 12-gon. HO must be a hexagon in T (i.e. it is not

fused into a disk), and hence is adjacent to some 12-gon which

must be fused from one of the hexagons H3' H4' or HS. H3 is

rul.ed out, el.se H2 would be adjacent to two 12-gons,

contradicting (b)i HS is eliminated since the reflection which

fixes £ would force H6 to be adjacent to two 12-gonsi and H4 is

eliminated since the reflection which fixes £' would force H1

I

to be adjacent to two 12-gons. This establishes the claim.

Claim 2: If a hexagon H is adjacent to exactly two 12-gons, then

they must be incident to opposite edges of H.

Pf: Otherwis~, we have the configuration of figure S.7b, with

some disks fused, but with the hexagons HOt H1t and H2 all

hexagons in T (i.e., not fused). Since vertices 1, 2, and 3

are then not equivalent, they form a representative set and

each hexagon in T must be equivalent to H2. But the only way

 145

to fuse disks adjacent to HO so it is equivalent to H2 places a

12-gon on the edge of H1 opposite T.

Claim 3: No hexagon is adjacent to three 12-gons.

Pf: If HO (figure 5.7b) was fused into a disk so that H2 was

adjacent to three 12-gons, then by claim 2 hexagons situated

like H1 would also have to be adjacent to three 12-gons. This

continues outward, and forces the isogonal (4.6.12) tiling.

We are ~eft with the conc~usion that every hexagon is

adjacent with exact~y two 12-gons, and that they are incident to

opposite edges of the hexagon. This easily forces the tiling to

be the 2-isogonal (3.4.6.4; 4.6.12) tiling. There are thus no

3-isogonal (3.4.6.4; 4.6.12) tilings.

The last 5 cases use three general arguments several times.

These arguments are not as precise as the "facts" used in the

proof of theorem 4.8. In the current analysis it seems more

difficult to write down specific conditions which imply a unique

extension to a 3-isogonal tiling. Neverthe~ess, the first two

arguments below frequently imply a unique extension, but the

reader will need to verify this when the arguments are invoked.

The three arguments, which will be referred to by name, are:

SYmmetry argument: Once enough SYmmetries of the tiling are

known, a portion of the tiling may force the structure of the

I
entire tiling. To show 'that an isometry is a sYmmetry of the
 146

ti~ing, note that for any pairs (v,. e,) and (v2' €2)' with vi

an endpoint of Ei' there are only two isometries which take one

pair to the other, and these two differ by a reflection in E2"

Hence if we know that the pairs (v., e.) must be equiva~ent,

a 1

and can eliminate one isometry as a possible symmetry, then the

other isometry must be a symmetry of the tiling.

Neighborhood argument: We define the neighborhood of a vertex v

to be the union of the vertex figure of v and the vertex

figures of all vertices adjacent to v. Frequently (but not

always), if we know the neighborhoods of representatives of

each vertex orbit, the tiling is forced. For the tiling to be

forced, it is necessary (at each stage in extending the tiling)

to be able to recognize which orbits some incompleted vertices

-I
belong to, and that the orientations of the neighborhoods of

these vertices.are recognizable. In the cases where we use

this argument, it is readily verified that this is the case.

Representative Sets argument: T~e proof of lemma 2.' not only

shows that representative sets of verticJs exist which induce

connected graphs, but a~so shows that any set of inequivalent

vertices which induce a connected graph can be extended t.o such

a representative set. Starting with a non-monogonal edge E,

whose endpoints are 2 such inequiva~ent vertices, this argument

shows that a representative of the third orbit can be found

I

which is adjacent to aniendpoint of e.

 147

Case 4) (3.4.3.12; 3.122). Every square must be contained

in a configuration 1ike the s01id edges of figure 5.8. No

triang1e is adjacent to on1y one 12-gon since then it wou1d be

adjacent to two squares, which wou1d force the existence of a

vertex incident to two squares. If every triangle is adjacent to

two 12-gons and 1 square, then the neighborhood argument forces

the tiling to be the 2-isogonal (3.4.3.12; 3.122) tJLling of figure

4.2, #16. Thus we may assume there is a triangle adjacent to

three 12-gons. Now (by connectivity of the dua1 graph) there must

be such a triang1e "near" a triangle which meets only two 12-gons;

i.e. we may assume that triangle A in figure 5.B meets three

12-gons. In this case, vertices 1, 2, and 3 are all inequivalent;

hence form a representative set. This means that all vertices

labeled 3 are equivalent, and there must be a 3-fold rotation

center which fixes A. This forces the dashed edges. If edge £

has simple type 3.12, then vertex x must be equivalent to vertex

1, but has the wrong neighborhood. Similarly, if edge E has

simple type 122, then vertex x must be equivalent to vertex 3, but

this also is impossible, since x would not be adjacent to any

vertex of type 2. We conclude there are no 3-isogonal

(3.4.3.12; 3.122) tilings. (We remark in passing that a more

careful analysis also shows that there are nO 4-isogonal tilings

of this sort, and a unique 5-isogonal tiling.)

 148

, - ... --
, " ...
,,
,,
,
,,'---- ..
I' ' ••...
,
, ,,
/ I
,
---::,": ,
I. ,
,
, ••••• 1 I
, " r
.....•. " \
... ':-'---~ .. .•......•.
\ .•._ .•..•.6',,'

,
,,
,,
I

,,
I

Figure 5.8: Theorem 5.1, case 4.

,-- -_ .•. \ ,

'/ "
,
,',
'.z "
I
/
I

~-
,,--+--=.--"'--1
,
,I. ~_"",,,~_..:Y
\ I ,

" \ I
\ I " I
,/ \/
\ I

,
' .,.1
/
I

Figure 5.9 (a) Figure 5.9 (b)

Figure 5.9: Theorem 5.1. case 5.

 149

Case 5) (34.6; 3.6.3.6). By theorem 4.1, there is an edge

of type 34.6 / 6.3.6.3, and the 2-gonal assumption forces this

edge figure, marked as a bold edge in figure 5.9a, to extend to

the solid edges. The vertex x has type 34.6 and we may assume (by

symmetry) that this extends to the dashed edges. Vertices 1, 2,

and 3 now form a representative set (2 and 3 are inequivalent

because the only symmetry which could take 2 to 3 would be a

reflection perpendicular to the edge joining them - which is ruled

out because of its action on the dashed hexagon). The vertices x

and yare both incident to an edge of type 32.6.32 / 32.6.32 (the

edge joining them), and vertex 2 is not, so x and yare in orbit

3. The symmetry which takes vertex 3 to y can only be a transla-

tion, forcing the solid edges of figure 5.9b. Since both vertices

of type 34.6 are adjacent, along the hexagon, to another vertex of

that type, it is not possible for exactly one of the vertices w

and z to have type 34.6. If both have type 3.6.3.6, the dashed

edges are forced and neither w nor z are adjacent to vertices of

type 2 or 3, contradiction. Thus both wand z have type 34.6.

One of these must be in orbit 2 and one in orbit 3.

Regardless of which choice is made, the symmetry which takes 3 to

the appropriate vertex (w or z) is forced, and the symmetry

argument shows that the tiling is forced. This gives the two

tilings (34.6; 34.6; 3.6.3.6)1,2 of figure 5.6, #10 and #11.

 150

Case 6) (34.6; 32.62). By theorem 4.1 there must be an edge

of type 32.6.32 / 3.62.3 or else one of type 34.6 / 62.32• The

latter edge forces the existence of one of the former type, so we

assume the existence of such an edge. The fact that any triangle

adjacent to a hexagon at a vertex is adjacent to another triangle

at that vert~x (this is true for both vertex types), forces the

edge figure 32.6.32 / 3.62.3 to extend as shown in figure 5.10a.

By the representative set argument, the 3 vertex orbits are

represented by 1,2 and one of { x, y, z }.

Claim 1: Vertices 2 and x are equivalent.

Pf; If not, there is a unique orbit of vertices of type 34.6;

vertices 1 and ware equivalent; and the (unique) edges of type

32.6.32 / 3.62.3 incident to vertices 1 and w force vertices 2

and x to be equivalent.

Claim 2: Vertex z has type 34.6.

Pf: Assume z has type 32.62• If vertices 21 and z are inequiva-

lent, then there is a unique orbit of vertices of type 34.6, so

vertices and yare equivalent, and the argument of claim 1

forces 2 and z to be equivalent. Thus we may assume vertices 2

and z are equivalent, which (via the same argument) forces

and y to be equivalent, and the tiling is 2-isogonal.

Using claim 2, and the non-existence of edges of type

33.6.3 / 3.62.3, the solid edges of figure S.10b are forced.

Vertices 1, 2, and 3 are n w inequivalent (1 and 3 have different

 151

Figure 5.10 (a)

I
I
,
\
\

Figure 5.1·0 (b)

Figure 5.10: Theorem 5.1, case 6.

 152

neighborhoods), hence form a representative set. The vertices

labeled z can only be equivalent to vertex 3, so the dashed edges

are forced. Note that the vertex types 1 and 3 have the property

that if we travel out along the central edge of simple type 32,

continuing in a straight line, a vertex of type 32.62 is reached

after traversing 1 or 2 edges. The same path from the vertex u,

however, traverses at least 4 edges before reaching such a vertex.

We conclude that u is not equivalent to either 1 or 3, and hence

there is no 3-isogonal (34.6; 32.62) tiling.

Case 7) (33.42; 32.4.3.4~. There are three non-monogonal

edge types which are possible in such a tiling, and we will refer

to these by number: I) 33.42 / 4.3.4.32; II) 33.42 / 4.32.4.3;

III) 32.42.3 / 3.4.3.4.3.

Claim 1: There is an edge of type I.

Pf: Assume not. The existence of a type III edge easily forces

the existence of a type II edge, so we ~y assume the tiling

has such an edge. Using the non-existence of type I edges, the

-edge figure of type II extends to the solid edges of 5.11a (the

vertex figures a, b, all the 1's, and 2 are forced in that

order). We claim that all the vertices labeled 1 are equiva-

lent. This is certainly true if there is a unique orbit of

vertices of that type. Otherwise, there is a unique orbit of

vertices of type 32.4.3.4; all the vertices x and b are equiva-

lent to 2; each of these 4 vertices is incident to a unique

 153

1- - - -i-- .•..
---.:
I
I
,
I

I
I

,I, __ :; ....__~~;.-.=«
1

1 " I ,
1
I .r '
'f---~""~~~ I \

/-- - ~'~---iIio-
•••••
'_
, "
"
1
!,,'
,
~.~---=~----~~ ,
I

'~
- - ~:'---j"'--K
\
,
\ \ '
',. ~ \~ \L __ ~

-, I I
...'1- I,"

Figure 5.11 (a) Figure 5.11 (b)

Figure 5.11 (c)

I
1'----"
\ r

I
, ...
" \ I
1-- - -1- - --, .•....
''

1 1
I I ,',
I, ,
~---,. - - --.... I
" ••••. '/1

I
I

Figure 5.11 (d) Figure 5.11 (e)

Figure 5.11 Theorem 5.1, case 7.

 154

edge of type II; and the other endpoints of these equivalent

edges are the four vertices labeled'. The equivalence of the

vertices 1 then forces the dashed edges and the representative

set argument (starting with' and b) says that all vertices of

type 33.42 are equivalent. The neighborhood argument on the

vertices of type 33.42 alone is then sufficient to force the

2-isogonal (33.42; 32.4.3.4)1 tiling (figure 4.2, #9).

Claim 2: There is a unique 3-isogonal (33.42; 32.4.3.4) tiling

with only one orbit of vertices of type 32.4.3.4.

Pf: The edge figure I forces the solid edges of figure 5.11b.

Since vertices x and yare equivalent, and z cannot have type

33.42, the vertex " adjacent to y must have type 33.42 and the

dashed edges on the right are forced. By the symmetry

argument, vertex x is equivalent to z by,a translation and to

vertex y by a glide reflection. Under t~ese symmetries, the

vertices labeled 1 (or '1), and those labeled 2 (or 2'), are

equivalent. If vertex a had type 32.4.3.4 then the symmetry

which takes x to a forces vertices 1 and 2' to be equivalent

and, by the representative set argument on y and 2', the tiling

would be 2-isogonal. Thus vertices a and a' have type 33.42;

the dashed edges on the left are forced; a must be equivalent

to vertex " (look at the edges of simple type 32); and this

equivalence is realizable only by a rotation through the center

of the edge E. The symmetry argument noV forces the tiling

I

1
(33.42; 33.42; 32.4.3.41 of figure 5.6, #12.
 155

We may thus assume there is one orbit of vertices of type 33.42.

C~aim 3: No square is adjacent to two other squares.

Pf: If not then we must have 3 squares in a row. This forces

figure 5.11c, and every vertex of type 33.42 is equivalent to x

(i.e. is "across from" two other vertices of the same type),

which forces these squares to be contained in a strip of

squares. This then forces the isogona1 (33.42) ti1ing (figure

4.1, #5).

An edge figure of type I is forced (as in claim 2) to extend

to the solid edges of figure 5.11b. Claim 3 forces this to extend

to the solid edges of figure 5.11d. The edge £' cannot be of type

III, and if it was of type II, then vertex 4 would be of type

33.42; but wou~d be incident to 2 edges of type III (un~ike vertex

1) which contradicts the assumption of a unique orbit of vertices

of type 33.42. Thus E' has type I, forcing the dashed edges of

5.11d. Since vertices x, y, and z are equivalent, the sYmmetry

argument shows there is a translation from x to z and a 2-fold

rotation in the center of the edge E. This forces the tiling to

contain the configuration of figure S.lle, and forces the vertices

labeled i or i' (i = 1, 2, 3) to be equivalent.

The vertices marked 1 and 2 in 5.11e represent 2 distinct

orbits. Starting with these vertices, the representative set

argument imp~ies that the third orbit contains either a vertex

labeled 3 (which are all e uivalent) or else the vertex x. If x

 156

has type 33.42, then it is equivalent to 2, which also forces 1

and 3 to be equivalent; hence forces the tiling to be 2-isogonal.

Thus vertex x has type 32.4.3.4, as shown by some of the dashed

edges. Vertices 1 and 3 are not equivalent (1ook at the incident

edges of simple type 32), so vertices 1, 2, and 3 represent the 3

orbits. The remaining dashed edges are now forced since every

vertex of type 32.4.3.4 must be adjacent to a vertex of type

33.42, and vertex y is the on1y candidate adjacent to x. Since

vertices 2 and yare equivalent, another translation is forced,

and the symmetry argument forces this ti1ing to be the

Case 8) (36; 34.6). Each hexagon is surrounded by vertices

of type 34.6. For any vertex v of type 36, there are 6 edges in

the tiling which connect two of the vertices adjacent to V (these

would be the edges of the hexagon formed if we fused v). We say

that these edges are ~ v, and the types (and order) of the
edges near a vertex is unchanged by a symmetry of the tiling.

Looking at the structure of these edges will frequently enable us

that two vertices of type 36 are inequivalent.

I
to determine

The only symmetry groups which fix a hexagon but do not split

its vertices into 3 or more orbits are groups which contain a

3-fold rotation center and the dihedral group of order 4.

Claim 1: The only 3-isogona1 (36; 34.6) tiling where the

symmetries which fix a hexagon form the dihedral group of

 157

order 4 is the tiling (36; 34.6; 34.6) of figure 5.6, #4.

Pf: The solid edges, and lines of reflection, shown in figure

5.12a are forced. The vertices x cannot have type 34.6 (else

the rE~flection R2 is violated), and the edge E cannot have

simple type 3.6 (else the reflection R1 forces vertex y to

have type 32.62), so the dashed edges are also forced.

Since the vertices of one hexagon split into two orbits.

all hexagons are equivalent, and all vertices of type 36 are

equivalent. Thus, if vertex y has type 36, then (since x and y

are equivalent) every vertex of type 36 has one reflection in a

line collinear with some of its edges (like y) and One

reflection which bisects the angle between two edges (like x).

Considering the choices with either x or y shows that these

reflections are perpendicular, hence the solid edges and lines

I
of reflection in 5.12b are forced. Since y is incident to a

hexagon along the line R" x must be inc~dent to one along R3'

and the dashed edges are forced. The sYmmetry argument with

these reflections forces the tiling (36; 34.6; 34.6) of figure

5.6, #4~

Thus, we may assume that vertex y (in figure 5.12a) has

type 34.6, which forces the solid edges of 5.12c. The fact

that vertex 1 is near two edges of simple type 3.6, which are
I
separated by one edge, combined with the equivalence of

vertices 1 and 2, forces some of the dashed edges; the

,
reflection R2 then forc\s the rest of these edges. Equivalence
 158

'\"-- -"'"
I, '\
I \ ' \
I ,I ,

" \
-L__
,
", "

_
, /
,
JI
I
I

Figure 5.12 (b)

Figure 5.12 (a)

\
",
. \ ,\,
""
"
'(-.--~:-" " , ,--1; '.,
\ I \' " \ I \' \ ,
't/_.,J,I. __ Jl. ~ __ ~_ •. ~

Figure 5.12 (c) Figure 5.12 (d)

Figure 5.12: Theorem 5.1, case 8.

 159

Figure 5.12 (e) Figure 5.12 (f)

Figure 5.12 (g) Figle 5.12 (h)

Figure 5.12 (cant.): Theorem 15.1, case 8.

I
.J ) ) ] J

,, ,
,
.' A \
\
, ,~
I

,, '.
,
; - -x, - y .' '( , - ---).
\ ,\ ~ a . 'x' " ,"
\

":---*---'~'
I '. " \ I

')t'---:i---'t
\' ,
,,
\ / .\ I \. /:', " '. "
\. \ I, I.. I •• ,

''t.- '~-'-'\-----*-.-,*- .. " \

\ :
/ \
\ ,
" '~
•• ,
/ "
•• '
I..
\ I
I
B )
" '.
'f - - -- -'It - - - *' - --"'It
'"
- •. y
\'
\ ."', ,", ,." I ,,- - --'
\\,;,,/.~_':i.. __'~/ 'II....,"

Figure 5.12 (j) Figure 5.12 (k)

Figure 5.12 (i)

Figure 5.12 (cont.): Theorem 5.1, case 8.

0\
o
 161
l
of hexagons forces the reflection R3' The hexagon H must be

fixed by some reflection passing through its vertices, and this

must be the reflection R4, since any other choice for a

reflection axis creates a 6-fold rotation center in one of the

other hexagons. The symmetry argument, with these reflections,

shows that the 2-isogonal (36; 34.6)2 tiling of figure 4.2, #2,

is forced.

Claim 2: There are edges of type 33.6.3 I 36•

Pf: The only other possible non-monogonal edge is 32.6.32 I 36, as

shown by the bold edge in figure 5.12d. If vertex x is of type

36, then edge e: has the desired type. Otherwise, x is of type

34.6 and edge £1 has the desired type.

Since we may now assume the existence of at least a 3-fold

rotation center, the solid edges of figure r.12e are forced from

claim 2. Consider a set S of three vertices situated as shown.

If vertex 1 has type 34.6, then we may assume (by symmetry) that
I

the dashed edges exist. Now vertex 2 cannot have type 34.6, else

vertex x has type 3.6.3.6, so at least one vertex of S has type

36• If only one vertex in 5 has type 36, there are essentially

two choices for that vertex, either vertex 1 or 2. If vertex 2 is

the only vertex in S of type 36, then figure 5.12f is forced

(using the 3-fold rotation); the vertices 1, 2, and 3 form a

representative set; there is a unique orbit of hexagons; and the

equivalence of the hexagons H1, H2, and H3 'forces vertex 4 to be

 162

near only edges of simple type 32, so it represents a 4th orbit.

If the only vertex of S with type 36 is the center One (vertex

in 5.12e), the solid edges of figure 5.12g are forced. By the

represen·tative set argument (starting with and 2), whichever

vertex represents the 3rd orbit has the same neighborhood as the

other vertex of the same type, i.e. the neighborhood of a vertex

is completely determined by its type. Consequently, the

...., neighborhood argument forces this tiling to be the 2-isogonal

tiling (36; 34.6)2 of figure 4.2, #2.

If exactly 2 vertices of S have type 36• this must include

the center vertex (if the other two vertices have type 36, they

force the center one to have that type also) and figure 5.12h is

forced (up to symmetry). Vertices 1, 2, 3, and 4 are now all

different (vertex 2 cannot have 2 near edges of simple type 3.6),

contradiction.

Thus. we may assume that every vertex in S has type 36,. and the

solid edges in figure 5.12i are forced. We I now distinguish 3

cases: 1) Edges a have simple type 3.6; 2) Edges b have simple type

3.6; 3) All the outside solid edges of figure 5.12i have simple

type 32• By symmetry and the 3-fold rotation. these are the only

possible cases.

Case 8.1) Edge a has simple type 3.6.

Equivalence of the hexagons forces the dashed edges in figure

5.12i, with vertices 1, 2'land 3 as a representative set.

Vertices x are then equivalent to vertex 1 (they have at least one

 163

near edge of simple type 3.6, and vertex 2 does not), hence

l vertices y have type 34.6. The neighborhood argument now forces

the tiling (36; 36; 34.6)1 of figure 5.6, #1.

Case B.2) Edge b has simple type 3.6.

Equivalence of the hexagons forces the solid edges in figure

5.12j. If edges c have simple type 32, then vertices 1, 2, 3, and

4 are all distinct, hence we may assume the dashed hexagons exist.

Since each hexagon must be surrounded by vertices of type 36, the

same arguments continue to apply as the tiling is extended,

forcing the 2-isogonal tiling (36; 34.6)1 of figure 4.2, #1.

Case 8.3) Edges a and b have simple type 32•

The solid edges of figure 5.12k are forced. We refer to the

vertices at distance n from the central hexagon as the n-th ring

of vertices. The vertices 1, 2, and 3 are a representative sete

Vertices l' and 2' must be equivalent to and 2, respectively.

If the hexagon does not contain a 6-fold rotation center, then 1

and l' can be equivalent only by a reflection which fixes 2; and 2

and 2' can be equivalent only by a reflection which fixes 1.

These two reflections then generate a 6-fold rotation center.

hence we may assume such a rotation exists. If vertex y (and its

6 images under the rotation) had type 34.6, then equivalence of

the hexagons forces the (36; 36; 34.6)2 tiling of figure 5.6, #2.

Thus we may assume the dashed triangles exist. If every vertex of

a set such as S had type 36, then vertex x would be adjacent only

to vertices of type 36, and would be in a fourth orbit.

 164

Consequently, we must have an edge of simple type 3.6 joining two

3rd ring vertices, i.e. a hexagon such as A or B (but not both).

In the former case, since all hexagons are equivalent, vertex z

will be adjacent only to vertices of type 36, contradiction.

Thus, the tiling must contain hexagon B. The 6-fold rotation

center in the central hexagon and in B now force the tiling

(36; 36; 34.6)3 of figure 5.6, #3. /1

 165

Section 5.2: Tile Regularity.

The fact that the only isohedral edge-to-edge tilings by regular

polygons are the tilings #1, 2, and 3 in figure 4.1 has been known

since antiql.llity
(see section 1.2). It is thus a little surprising

that the 2-i.sohedral tilings have not been previOUSly classified.

Grunbaum and Shephard [1983] conjectured that there are only 11

such tilings.2 In fact, there are 13 2-isohedral tilings.

A direct classification of the 2-isohedral tilings is not much

simpler than the more inclusive enumeration of tile-homogeneous

tilings, so we find these tilings first and get the 2-isohedral

tilings as a corollary. We remind the reader that a tiling is

tile-homogeneous if any two congruent tiles are equivalent under the

symmetries of the tiling.

Theorem 5.2: A tile-homogeneous edge-to-edge tiling by regular

polygons is v-isogonal for v ~ 3, or else is the 4-isogonal tiling

of figure 5.13.

Pf: Case 1) Every vertex of T is incident with a triangle.

Fix a triangle T. By homogeneity, every triangle is equivalent to

2 Actually, they conjecture that there are 14 2-isohedral

tilings, which they claim are the ones they show. They then show
only 11 such tilings. It appears that they are including the
isohedral tilings in their lis:t of 2-isohedral tillngs, even though
this disagrees with their deflnitions.
 166

Figure 5.13: The single tile-homogeneous tiling by regular polygons

with more than 3 orbits.of vertices. This tiling is 4-isogonal,
and representatives of the 4 vertex orbits are marked.

Figure 5.14 (a) Figure 5.14 (b)

Figure 5.14: Theorem 5.2, ~case 3.

---"""
,
\
I
I

'- .J
,
... l.,.r"- ." "...
"- - - ,.
- .I

Figure 5.15 (a) Figure 5.15 (b)

Figure 5.15: Theoremls.2, case 4.

 167

T, so every vertex of T is equivalent to one of the three vertices

of T, and T is at most 3-isogonal.

Thero are only four vertex types without triangles possible in

an edge-to-edge tiling by regular polygons, hence we need only

consider the possibility that T contains one of these four.

Case 2) T contains a vertex of type 4.82•

As mentioned in section 4.2, T must be the isogonal (4.82) tiling

of figure 4.1, #11.

In the other three cases we refer to the neighborhood of a

tile T, i.e. the set of neighbors of T. Given a Bet of tiles with

their neighborhoods, there is usually at mo~t one tile-homogeneous

tiling containing them. As with the neighborhood argument used in

proving theorem 5.1, it is necessary (for uniqueness) to check

that the orientation for placing the neighborhood of a tile is

uniquely determined. We may assume, in these 3 cases, that the

tiling is not isogonal, and hence contains J

I
non-monogonal edge

figure which meets the vertex type under consideration.

Case 3) T contains a vertex of type 44.

The only non-monogonal edge types which meet a vertex of type 44

are 4.3.6.4 / 44 and 4.33.4 / 44. If there is an edge of the

first type, then one square in the edge figure is adjacent to a

I

triangle while another is adjacent to a hexagon, so, by

homogeneity, each square i; adjacent to a triangle and a hexagon

 168

(and two squares). Hence, the edge figure extends to figure

5.14a. Homogeneity of the squares then forces the 3-isogonal

(3.42.6; 3.4.6.4; 44) tiling of figure 5.1, #34.

We may thus assume the tiling contains the edge 4.33.4 I 44

and the neighborhood of a square contains only squares and

triangles. This edge figure implies that the neighborhood of a

square contains at least 2 squares and at least 1 triangle.

Claim: Each square is adjacent to 3 squares and 1 triangle.

Pf: Otherwise, figure 5.14b is forced. Triangle 1 is adjacent to

two triangles, so triangles 2 and 3 must be also, forCing the

dashed edges. But now, vertex x cannot be completed.

The claim, together with the edge figure 4.33.4 I 44, easily

forces two strips of squares surrounded by strips of triangles.

The equivalence of the triangles then forces the 2-isogonal tiling

Case 4) T contains a vertex of type 4.6.12.

Claim: Each 12-gon is incident only to vertices of type 4.6.12.

Pf: If not, then some vertex of this type must be adjacent to a

vertex of type 32.4.12 as shown by the solid edges of figure

5.15a. The neighborhood of a triangle must then consist of a

triangle, a square, and a 12-gon, which forces the dashed

edges. Vertex x must now have type 3.4.6.4, and the triangle

incident to this vertex has the wrong neighborhood.

I
This claim forces the soli edges of figure 5.1Sb. Edge £ cannot
 169

have simple type 42, else one square would be adjacent to a

triangle (at vertex x) while the other square would be adjacent to

no triangles. If E has simple type 4.12, the isogonal (4.6.12)

tiling is forced by the neighborhood argument. Thus we may assume

£ has simple type 3.4, and the dashed edges are forced. The

homogeneity of the squares forces the hexagons to be adjacent to

two 12-gons, and the neighborhood argument forces the 2-isogonal

(3.4.6.4; 4.6.12) tiling of figure 4.2, #20.

Case 5) T contains a vertex of type 63•

The only possible non-monogonal edge figure is 6.32.6 / 63, shown

in figure 5.16a. Homogeneity of the triangles forces the exis-

tence of one of three symmetries taking the left triangle to the

right, and each of these forces the vertices x to be equivalent.

Vertex x must then have type 3.42.6, 3.6.3.6, 34.6, or 32.62• In

the first case, vertex y must have type 32.4.3.4, and some

triangles are adjacent to 2 squares, while dthers are adjacent to

only 1 square, contradiction. In the second and third case, the

solid edges of the figures in 5.16b are forced.

I
I
The dashed edges

are forced by the homogeneity of the triangles. In both cases,

homogeneity of the triangles also prevents the edges e from having

simple type 3.6, hence they must have simple type 62• The neigh-

borhoods of a triangle and a hexagon are then determined, and the

I

3-isogonal tilings (32.62; 13.6.3.6; 63)1 and (34.6; 32.62; 63) of

figure 5.1, #31 and #22, a 'e now forced.

 170

...•.
,r l

Figure 5.16 (a) Figure 5.16 (c)

(
""
•.'IIt,
.....,
~
-r>;
...' ,'
~'
-,
...
~
I I
I
.~
I

. ~ ..
c. .. 1 ,
I
. ' I
••• I

I
-: '~
, .... "
I .,
,~
..... , I
I

Figure 5.16 (b)

Figure 5.16: Theorem 5.2, case 5.

 171

The only possibility remaining is if the vertices x of figure

5.16a have type 32.62, which (combined with the homogeneity of the

triangles) forces the solid edges of figure 5.16c. For the 6

triangles to be equivalent, the vertex y must be the center of at

least a 3-fold rotation, which forces the dashed hexagons labeled

21. If edge a (and hence al) is of simple type 3.6, the homoge-

neity of the triangles forces hexagons 1 and 2 to have different

neighborhoods, contradiction. Hence all of the dashed hexagons

are forced. The edge b can have simple type 3.6 or 62• In either

case, the neighborhoods of both hexagons and triangles are fully

determined by this choice, and the tiling is forced. If b has

simple type 3.6, the 3-isogonal (36; 32.62; 63) tiling of figure

5.1, #19, is forced. If b has simple type 62, the 4-isogonal

(36; 32.62; 63; 63) tiling of figure 5.13 is forced. II

Corollary 5.3: The only 2-isohedral edge-to-edge tilings by regular

polygons are:

the following five isogonal tilings of figure 4.1:

(33.42); (32.4.3.4); (32.12); (3.6.3.6); (4.82);

the following four 2-isogonal tilings of figure 4.2:

(36; 32.62); (34.6; 32.62); (33.42; 44); (32.62; 3.6.3.6);

the following three 3-isogonal tilings of figure 5.1:

(36; 32.62; 63); (34.6; 32.62; 63); (32.62; 3.6.3.6; 63);

and the 4-isogonal tiling of figure 5.13:

(36; 32.62; 63; 63).

 172

Pf: Since a1l monohedra1 tilings are isohedra1, the 2-isohedral

tilings must a1so be 2-hedra1, hence are tile-homogeneous. The

resu1t then fo11ows by inspection of the tilings described in

theorem ~i.
2• //

Corollary 5.4: There are exact1y 22 ti1e-homogeneous edge-to-edge

tilings by regular pOlygons, the 13 2-isohedral tilings of

coro1lary 5.3 and:

the fo1lowing five isogonal tilings of figure 4.1:

1· 1 31 3 3
(36); (44); (6); (3.4.6.4) ; (4.6.12) ;

and the following four 2-isogonal ti1ings of figure 4.2:

4 3
(3.4.6.4; 4.6.12) (3.4.3.12; 3.122) ;
3 3
(3.42.6; 3.6.3.6)2; and (3.42.6; 3.4.6.4; 44) •

The superscript at the end of the tiling symbol is the number of

orbits of tiles in the tiling. Thus, there are 3 tile-homogeneous

tilings which are isohedral; 13 which are 2-isohedral; 5 which are

3-isohedral, and 1 which is 4-isohedral.

Pf: This follows by inspection of the tilings described in theorem

5.2. II
 173

Section 5.3: Edge Regularity.

The e-isotoxal tilings with e ~ 3 were classified in section 4.3

(for these values of e the e-toxal tilings and the e-isotoxal tilings

are equivalent). If we were interested only in the e-isotoxal

tilings, we could also get this classification as a corollary to the

classifications of the v-isogonal tilings ( v < 3 ), since by theorem

3.6, v ~ e.
In chapter 1 we defined edge-homogeneous tilings to be those in

which any two edges of the same full type were equivalent. The

classification of edge-homogeneous tilings seems to be a very

difficult task. For example, there are at most 5 tiles which could

arise in a tile-homogeneous tiling, and at most 15 vertex types in a

vertex-homogeneous tiling, but an edge-homogenous tiling might have

as many as 98 different edge types. Thus, instead of considering

this problem, we investigate those tilings which satisfy a stronger

homogeneity condition. A tiling is strongly edge-homogeneous if any

two edges of the same simple type are equivalent. The classification

of these tilings is easily accomplished.

Theorem 5.5: Strongly edge-homogeneous edge-to-edge tilings by

regular polygons are vertex-homogeneous.

Pf: Assume T is a tiling that is strongly edge-homogeneous, but not

vertex-homogeneous. Then T has two vertices v and Vi of the same

type which are not equivalent. Let n be the valence of v (and

Vi), and let E

i
, i = 1, 2, ••• n, be the edges incident to v, and
 174

let Ei' be the corresponding edges incident to v'. i and Ei'

E

have the same simple type, so for each i there is a symmetry S. of

1

T such that Si(€i') = Ei• Since Si(v') * v, but Siev') is an

endpoint of Ei' we conclude that Ei is the line segment

[v, Siev'»), hence every vertex adjacent to v has the same type as
v. The same argument shows that Jf v" is any vertex in T with the

same type as v, then it is adjaOe~t on1y to vertioes of this type.

By connectivity of the graph of T, all vertices in T have the same

type, i.e. T is monogonal, hence isogonal and thus vertex-

homogeneous. II

Cor011a~ 5.6, There are 22 strong11 edge-homogeneous edge-to-edge

tilings by regular polygons namely:

the ten isogonal tilings of figure 4.1 excluding (34.6);

the following eight 2-isogonal tilings of figure 4.2:

(36; 32.62); (33.42; 3.4.6.4);

I
e3.4.3.12; 3.122); (3.42.6; 3.6.3.6)1

and the following four 3-isogonal tilings of figure 5.1:

Pf: This follows by inspection of the vertex-homogeneous tilings of

figures 5.1 - 5.5. To simplify this inspection, we note that the

tiling can contain at most two types of vertices whose vertex

figures contain the same simple type. This follOWS since if E is

 175

an edge of, say, simp1e type 3.4, then every vertex whose figure

contains an edge of this simp1e type must be equiva1ent to one of

the endpoints of E. Four set~ in particu1ar are worth checking.

These set:s, with the simp1e type which they a11 contain, are

1isted be1ow:

32: { 36; 34.6; 33.42; 32.4.3.4; 32.4.12; 32.62 }

3.4: { 33.42; 32.4.3.4; 32.4.12; 3.4.3.12; 3.42.6; 3.4.6.4 }
3.6: { 34.6; 32.62; 3.42.6; 3.6.3.6 }
42: { 33.42; 3.42.6; 44 }
E1iminating those vertex-homogeneous ti1ings which contain 3

vertex types from one of these sets e1iminates a11 of the

v-isogona1 ti1ings for v > 4, a11 but 7 of the 4-isogona1 ti1ings,

and 11 of the 3-isogona1 ti1ings. II

 176

Chapter 6: Summary and Suggestions for Further Research.

Most theses seem to contain a certain amount of solid research,

and some filler. This thesis is no exception. The additional

material is intended to fill some of the corners and niches that

arise fairly naturally in studying these tilings, but these results

are definately of less overall importance than some of the other

results. My opinion is that there are three main, important results

in this thesis:

1) The bounds on v, e, and t as functions of each other; both for

general periodic tilings and for tilings by regular polygons;

2) The classification results in chapter 5; and

3) Theorem 4.8, partly due to its classification of e-isotoxal

tilings, but especially due to the unexpecJed strength of its

connection between local and global regularity conditions on the

I
edges.

In addition to these specific results, there are 3 techniques which I

have used (for the first time, as far as I knOj) which appear to be

of significant further interest:

I) The systematic use of the ideas of connected representative sets

of elements in establishing other results (e.g. #1 above);

II) The use of edge types, simple edge types, and (especially) non-

monogonal edge types as a principal step towards the various

classification results;

III) The algorithm implied in the finiteness arguments of section

 177

3.2, and the idea of a closure of a fundamental region, as a

possible computer tool for further classification problems.

The ideas of fusing and dissecting hexagons and 12-gons as a

teChnique for studying certain kinds of tilings was used by

Sommerville [1905] and is extended here, especially in the proof of

lemma 4.3. This technique is very useful, and it would seem that it

has potential for further research -- especially if generalized

appropriately. To demonstrate its strength, we note that lemma 4.5

(b) is easily established in a paragraph using this approach.

Without this approach, Levy [1891] investigated this problem and

could not discover the solution. A few years later, still unsuccess-

ful at the solution, he submitted this as a research question (Levy

[1894]).

These are the highlights of the thesis. The rest of this

chapter gives a more detailed survey of the thesis results and their

importance, together with a collection of various unanswered

questions which arise from this research and se'em to merit further

study. The survey is arranged in approximately the same order as the

thesis, and the open questions are numbered seJuentiallY.

Chapter 1:

The facts and lemmata contained here are straight-forward. The

counter-examples to the claims of Grunbaum and IShephard [1983] are at

least as interesting as the pDoofs that these facts hold under the

assumption that the tiling has no singUlar points. My primary

 178

interest, and that of other researchers in geometry, is in ti1ings

without singu1ar points. Whi1e lemma 1.1 [connectedness' of graphs

and dual graphs in the absence of singular points] is thus of

interest, lemma 1.2 [on connectedness of dual graphs] is included

only because its statement and proof are quite simple. From this

perspective a comparable result for the connectedness of graphs seems

of less interest, since the examples imply that necessary and suf-

ficient conditions cannot be particularly simple.

Lemma 1.1 [connectedness of graphs and dual graphs in the

absence of singular pOints] seems nearly self-evident, and yet is

more powerfu1 than one might suspect. In particular, it gives rise

to lemma 2.1 [existence of connected, representative sets]. While

1emma 2.1 itself may seem straight-forward, Krotenheerdt [1969] takes

two journal pages to prove the equivalent of part 1 for the very

special case of vertex-homogeneous tilings by regular polygons. Our

proof of part 1, using lemma 1.1, requires only 8 lines.

Theorem 1.3 [equiva1ence of periodicity with the finiteness of

v, e, or t] seems to be of independent interest, but I have found it

quite frustrating not to be able to decide if the assumption of no

vertices of infinite valence is necessary. This raises the first

question I would like to see settled:

1) Are there any tilings with singular points which have only a
finite number of vertex orbits?

It seems unlike1y that such ti.1ings exist.

 179

Chapter 2:

Lemma 2.1 [existence of connected, representative sets] is quite

important, as mentioned earlier in connection with lemma 1.1. This

idea is alse, fundamental to many of the classification results.

Nevertheless, it is not a surprising result. Lemma 2.2 [the

comparable situation for locally regular tilings] is one of those

corners I felt was worth mentioning -- primarily because of the

curious gap in part 3. Thus, even though it may not be of great

import to further research in the area, I would like to know:

2) If a tiling T is 3-toxal, does it necessarily have sets of 3

edges which induce connected graphs (dual graphs)? What if T is

4-toxal?

I suspect that the answer is yes for 3-toxal tilings. Another

question mentioned in chapter 2 also points out some of the

differences between e-toxality and the other local regularity

conditions, namely:

3) Does there exist a monotoxal non-isotoxal tiling?

Theorem 2.3 [bounds on v, e, and t as functions of each other]

is the first major result in the thesis, and this has already been

published in Chavey [1984]. As remarked in section 1.2, GrUnbaum and

Shephard [1983] ask if there are triples (v, e, t) that are not

realizable; and this theorem answers that question with a vengeance.

Nevertheless, many open questions remain on the road to a fuller

classification of the realizable triples. Specifically:

 180

4} Is it possible to bound the value of v or e in a periodic tiling

as a function of t?

5} Establish a lower bound for Eu(T}, where Eu(T} = v - e + t. In

particu~ar, is Eu(T) ~ O?

6}Show that wider ranges of parameters are realizabl.e than those

given in theorem 2.3, (4) and (5).

The examples of (k+1, k+m, k+m+1}-tilings with m > 0 (used in

the proof of theorem 2.3, parts 3 and 4) all used tilings in which

the intersection of some pair of til.es was disconnected. If we

restrict our attention to til.ings in which this does not happen

(which Grunbaum and Shephard [1983] cal.l.normal. til.ings), it may be

possible to improve some of these bounds. In particular:

7) Are there bounds on e or t as a function of v for the class of

normal tilings?

Chapter 3:

The classification of edge types (theorem 3.1) seems like the

sort of thing that should be in print somewhere, and this

classification becomes very useful for many of the later results.

Neverthel.ess, the main contribution of section 3.1 is probabl.y the

introduction of the ideas of edge types and simple edge types. These

ideas are extremely useful in much of the later work. As mentioned

before, the classification of the non-monogonal edge types is of

great use in later classifications. The definition of "edge figure"

in chapter 1 appears to be ne~, and this is possibl.y due to an uncer-

 181

tainty as to what the correct definition should be. There are two

reasonable choices: 1) the edge and all incident edges (our choice)

which is equivalent to the edge and both incident vertex figures; or

2) the edge and both incident tiles. In the case of tilings by regu-

lar pOlygonn, the first definition corresponds to the (full) edge

type while the second definiton corresponds to the simple edge type.
1 As we have seen. both of these ideas are quite fruitful in investiga-

tions of these tilings. The homogeneity condition for these two

definitions would correspond to what we have called edge-homogeneous

and strongly edge-homogeneous. In the case of tillngs by regular

pOlygons, the first-definition always gives more information than the

second. but this is not true for more general tilings (hence the term

"strongly edge-homogeneous" is probably inappropriate for more

general tilings). Although we will discuss chapter 4 later, it is

worth noting here that, with respect to the alternate definition of

an edge figure as the edge and the two tiles incident to it (and

'hence with an alternate definition of e-toxal tilings), the proof of

theorem 4.8 shows that all monotoxal tilings are isotoxal. This

proof also shows that the only 2-toxal tilings (under this alternate

definition) are the 2-isotoxal tilings and the following two classes:

A) Tilings by triangles and hexagons with no adjacent hexagons; and

B) The 2-gonal (36; 32.4.3.4) tilings (which are not yet classified).

Thus, although the general question of the classification of 2-gonal

tilings is listed later, this special case has independent interest:

 182

8) Classify the 2-gonal (36; 32.4.3.4) tilings.

Equiva1ent1y, by fusing the vertices of type 36 into hexagons:

8') C1assify the 2-gona1 (32.4.3.4~ 3.4.6.4) ti1ings.

The finiteness resu1ts of section 3.2 are of reasonab1e

interest, a1though not particu1ar1y surprising. As discussed in

section 1.2, these resu1ts are closely related to those of

Krotenheerdt [1969] (also, see the discussion of 1emma 1.1 above).

These results answer two questions of GrUnbaum and Shephard [1983]

(see section 1.2). One obvious further prob1em might be:

9) Calculate reasonable upper bounds on the number of v-isogonal

[e-isotoxali t-isohedral] tilings.

The basic technique of theorem 3.3 (with certain straight-forward

improvements) yie1ds the ridiCUlous upper bound of at most

(12v)4v+1.14v. 4; (4v)! v-isogonal tilings. This, for example,

shows that there are at most about 7.7 x 106 isogonal tilings (there

are actually 11). We can also show that there are at least l. 2v-1
v
v-isogonal tilings. Improvements in these bounds should not be

difficult. Nevertheless. because of the vast disparity between these

bounds, it would seem that improvements in either direction would

have to be fairly extreme to be of significant interest.

Lemma 3.2 [uniqueness of tilings with a given fundamental region

and labeled closure] raises some questions that are probably less

significant than many others listed here, but which I would still be

interested in:
 183

10) Can the word "labeled" be dropped from the statement of l"emma

3.2? If not, would it be sufficient just to label the vertices

in R according to their vertex orbits?

TO a J.arge ext.en t, lemma 3.2 (and the answer to # 10) telJ.s us how

much of a tilling must be drawn in order to guarantee that there is at

most one ext.ensf.on to a tiling of the plane. An obvious related

question is:

11) How much of a tiling must be drawn in order to guarantee that

there is at least one extension to a tiling of the plane?

Since there are many non-monogonal edge figures by regular polygons

which cannot be embedded in 2-isogonal tilings (see section 4.1), it

is generally necessary to draw more than just a fundamental region.

The obvious candidate for an answer to question 11 is the closure of

a fundamental. region.

The final. question that arises from lemma 3.2 is the possibility

of using the techniques of the proof to develop an al.gorithm for a

computer search of certain classes of tilings. There seems to be no

theoretical difficulty with the computer successively generating all

possibl.e centers; fundamental. regions; and l.abel.edcl.osures -- and

then drawing the resulting tiling. The researcher would probably

have to eliminate duplicates by hand, and would either need an answer

to problem #11 or else verify by hand that the computer's tilings can

be extended to tiJ.ings of the plane.

l Although
I
lemma 3.2 generaltes several questions, the major result

l
 184

in chapter 3 is probably the improvement of the bounds of theorem 2.3

for the case of ti~ings by regu~ar po~ygons. This resu~t, theorem

3.6, also answers two questions of Grunbaum and Shephard [1977a] and,

together with corollary 3.7 and theorem 3.8, has been published in

Chavey [1984]. Such bounds automatica~~y raise the question of

possible improvements, and we phrase this as two prob1ems to

l emphasize a distinction:

12) Improve the bounds of theorem 3.6 (2) and (3).

13) Determine whether or not the bounds of theorem 3.6 (1) can be

improved.

While it seems plausib~e that the bounds of (1) might a~ready be

sharp, it seems inconceivable that this is true for (2) or (3). In

the other direction, one could look for tilings which pull the

equalities of theorem 3.8 closer to the bounds of theorem 3.6. Thus:

14) Can the equalities of theorem 3.8 be improved so as to widen the

known range of realizable triples (v, e. t) for tilings by

regular polygons?

In connection with this problem. it might be worth investigating

tilings with 12-gons (none of the classes used in theorem 3.8 contain

12-gons).

Corollary 3.7 [classification of 2-isotoxal tilings] is

superseded by theorem 4.8; but it is still quite pleasant to be able

to solve an open problem of Grunbaum and Shephard [1977a] in two

sentences.
 185

Chapter 4:

Certainly the most interesting result in chapter 4 is theorem

4.8, which shows that e-toxal tilings are e-isotoxal if e ~ 3, and

also classifies these tilings. The partial classification of 2-gonal

tilings is very helpful in the later classification theorems, but

will probably not have a great deal of independent interest unless

this classification can be completed. Thus, the next problem is:

15) Complete the classification of the 2-gonal tilings.

Theorem 4.6 [rough classification of 2-hedral tilings] is not of deep

significance, and is included primarily for completeness. It does,

however, lead naturally into theorem 4.7 [how many orbits of

triangles and hexagons can exist in a 2-hedral tiling]. This theorem

is of reasonable interest, especially since it answers another

question of Grunbaum and Shephard [1983] and contradicts a claim of

theirs. Nevertheless, this would be a ,nicer result if it could be

combined with the solution to another problem of Grunbaum and

Shephard [1983] (Exercise #2.3.9):

16) Determine the triples (t, s, h) such that there is a 3-hedral

tiling with t orbits of triangles, s orbits of squares, and h

orbits of hexagons.

So far I have examples to show that h) is realizable so long

": s,
as: (a) s > 2; (b) t > 2h - 1; and (c) t == 2h+1 mod 3.
I
The fact that so little can be said about 2-hedral tilings seems

somewhat incongruous with the state of affairs for 2-gonal and

 186

2-toxa1 ti1ings. To a 1arge extent this can be b1amed on the fact

that, whi1e we may fee1 that vertices and tiles should be dual, this

duality is not fully reflected by the ideas of "tile" and "vertex

figure." Fc)r example, in a tiling by regular polygons, a vertex

l figure implicitly defines the tiles which are incident to that

vertex; but a tile does not (generally) define the types of its

incident vertices. Thus, to attempt to restore the duality, it seems

natural to define a tile figure as the union of a tile and all the

edges which meet that tile. For edge-to-edge tilings by regular

po1ygons, a ti1e figure would then fix the vertex types of the
l incident vertices. Many questions arise fairly naturally from here;

but the most interesting is probably:

17) For what values of t is it the case that any edge-to-edge tiling
l by regular polygons with only t tile figures is t-isohedral?

It is fairly easy to see that the answer to #17 includes t = 1 and

t = 2, and that these tilings are precisely the isohedral and

2-isohedral tilings.

Chapter 5:

All four of the classification results in this chapter

[3-1sogonal; 2-isohedral; tile-homogeneous; and strongly edge-

homogeneous] are of independent interest: and all but the last one

answer questions raised by Grunbaum and Shephard [1977a]. The most

important resu1 t is probably t.he c1assification of the 3-isogonal

tilings, and this result is u l.edin several places. This theorem has
l
187

a long, tedious proofi but it makes it possible to give the more

elegant proofs of the classifications of strongly edge-homogeneous

tilings and tile-homogeneous tilings. These latter two proofs make

effective use of the earlier classifications to eliminate most of the

work.

Of couz'se we can always try to extend these results to higher

values of v, e. and ti but most further results will probably require

the use of a computer search. The three most interesting questions

are:

18) Classify the 3-isohedral tilings.

19) Use a computer to classify the 4-isogonal tilings.

20) Use a computer to classify the edge-homogeneous tilings.

Since theorem 3.6 tells us that v ~ e. a solution to problem 17 will

also give the classification of 4-isotoxal tilings.

The word "homogeneous" has been used in this thesis, and in the

works of both Krotenheerdt and Grunbaum & Shephard, to imply tilings

where any two objects that look alike are, in fact, equivalent under

the symmetries. We have extended the use of this term in the present

work, but further extensions are also possible. For example, if

applied to the idea of "colored tilings". a homogeneous coloring

would be a coloring where any tile of one color could be carried, by

a (color preserving) symmetry, to any other tile of the same color.

 188

References

A. Badoureau

1881 Memoire sur les figures isosceles.

J. tcole Poly technique, 49(1881), 47-172.

M. Breen

1983 A Characterization Theorem for Tilings having Countably

Many Singular Points.

J. of Geometry, 21(1983), 131-137.

D. Chavey

1984 Periodic Tilings and Tilings by Regular Polygons I: Bounds

on the Number of Orbits of Vertices, Edges, and Tiles.

Mitteilungen aus dem Mathem. Seminar Giessen, 164(2),

1984, pp. 37-50.

J. P. Conlan

1976 Derived Tilings.

J. Combinatorial Theory (A) 20(1976), 34-40.

E. S. Fedorov

1891 Symmetry in the Plane. (Russian).

zapiski Rus. Mineralog. Obscestva, Sere 2, 28(1891),

I
345-390 + 2 plates.

n
 189

B. Grunbaum and G. C. Shephard

n
1977a Tilings by Regular Polygons.

" Math. Magazine, 50(1977), 227-247.

1977b ~~he Eighty-one Types of Isohedral Tilings in the Plane.

Math. Proc. Cambridge Philos. Soc. 82(1977), 177-196.

1978a Isohedral Tilings of the Plane by Polygons.

Comment Math. Helv., 53(1978), 542-571.

1978b Isotoxal Tilings.

Pacific J. Math., 76(1978), 407-430.

1978c The Ninety-one Types of Isogonal Tilings in the Plane.

Trans. Amer. Math. Soc., 242(1978), 335-353.

1979a Incidence Symbols and their Applications.

Relations between combinatorics and other parts of

mathematics. Proc. Sympos. Pure Math., XXXIV, Amer. Math.

soe ,, 1979.

1979b Erratum to "The Ninety-one ~pes of Isogonal Tilings in

the Plane".

Trans. Amer. Math. Soc. 249(1979), 446.

1980a Tilings with Regular Polygons. Models in the Plane from

the Days of Kepler to Today, with Recent'Results and

Unsolved Problems. (Italian)

Archimede 32(1980), 15-45.

1980b Errata corrige: "Tilings with Regular Polygons". (Italian)

Archimede 32(1980), 135.

1983 Tilings and patterJs, preliminary edition, 1983.

 190

T. Heath

1921 A History of Greek Mathematics, Vol. II, 1921.

1947 Euclid, The Elements, Vol. II, 1947.

J. Kepl.er

1619 Harmonice Mundi. Lincii, 1619.

German translation: Weltharmonik, M. Caspar, 1939.

O. Krotenheerdt

1969 Die homogenen Mosaike n-ter Ordnung in der euklidischen

1970a Ebene. I, II, III.

1970b Wiss. Z. Martin-Luther-Univ. Halle-Wittenberg, Math.-

Natur. Reihe 18(1969), 273-290, 19(2) (1970), 19-38 and

19(6) (1970),97-122.

1891 Sur les pavages a l'aide de polygones reguliers.

Bull. de la Societe Philomatique de Paris, (8) 3(1891),

46-50.

1894 Question 262.

I
Intermed. Math. 1(1894), p. 147 and 7(1900), p. 153.
 191

P. Niggli

/ 1926 Die regelmassige Punktverteilung langs einer Geraden in

einer Ebene. (Symmetrie yon Bordurmuster.)

z, Krist. 63(1926), 255-274.

D. M. Y. Sommerville

1905 Semi-regul.ar Networks of the Pl.ane in Absol.u1;eGeometry.

Trans. Roy. Soc. Edinburgh 41(1905), 725-747 and 12

plates.

A. Val.ette

1981 Tilings of the Plane by Topolgical Disks.

Geometriae Dedicata, 11(1981), 447-454.