Documente Academic
Documente Profesional
Documente Cultură
Structures
Vincent J. Matsko
Revised 2011
Chapter 0
Trigonometry
(cos θ, sin θ)
Figure 0.1
sin θ cos θ 1 1
tan θ := , cot θ := , sec θ := , csc θ := . (0.1)
cos θ sin θ cos θ sin θ
1
2 CHAPTER 0. TRIGONOMETRY
Note that tan θ and sec θ are not defined when cos θ = 0 (i.e., at odd multi-
ples of π2 ), while cot θ and csc θ are not defined when sin θ = 0 (i.e., at even
multiples of π2 ).
Since the circle in Figure 0.1 has radius 1, it is evident that the Pythagorean
Theorem implies that
cos2 θ + sin2 θ = 1. (0.2)
If cos θ 6= 0 (or sin θ 6= 0), we may divide (0.2) by cos2 θ (or by sin2 θ) to
obtain the additional Pythagorean identities:
(cos θ, sin θ)
θ
−θ
(cos(−θ), sin(−θ))
Figure 0.2
c
b
p θ q
a
Figure 0.3
Here, we have
a b b c c a
cos θ = , sin θ = , tan θ = , sec θ = , csc θ = , cot θ = . (0.6)
c c a a b b
π
It is evident that the measure of ∠prq is 2 − θ, so that (0.6) implies
π π π
cos − θ = sin θ, sin − θ = cos θ, tan − θ = cot θ,
2
π 2π π2
sec − θ = csc θ, csc − θ = sec θ, cot − θ = tan θ. (0.7)
2 2 2
Although (0.7) was derived in the case 0 < θ < π2 , these relationships are
valid for all values of θ.
Virtually all other useful relationships may be derived from the identity
First, note that [Oc] = cos β, so that [Ob] = cos α cos β. Moreover,
[dc] = sin β.
4 CHAPTER 0. TRIGONOMETRY
f c
β
α
e O b a
Figure 0.4
sin(α ± β)
tan(α ± β) = ,
cos(α ± β)
substitute from (0.11) and (0.12), and then divide both numerator and de-
nominator by cos α cos β to obtain
tan α ± tan β
tan(α ± β) = . (0.13)
1 ∓ tan α tan β
This formula is valid whenever tan α, tan β, and tan(α ± β) are all defined.
Of frequent use are double-angle and half-angle formulas. Writing (0.8)
with β := α gives
cos 2α = cos2 α − sin2 α,
which in combination with (0.2) may be written in three forms:
while writing (0.13) with β := α and taking the upper signs results in
2 tan α
tan 2α = (0.18)
1 − tan2 α
wherever tan α and tan 2α are defined.
To find an expression for tan θ/2, use (0.16) with (0.2) to obtain
r r r
θ 1 − cos θ 1 − cos θ (1 − cos θ)2 1 − cos θ
tan = ± · =± 2 =± ,
2 1 + cos θ 1 − cos θ sin θ sin θ
or alternatively
r r s
θ 1 − cos θ 1 + cos θ sin2 θ sin θ
tan = ± · =± 2
=± .
2 1 + cos θ 1 + cos θ (1 + cos θ) 1 + cos θ
It happens that the choice of “+” is always correct in these cases, so that
θ 1 − cos θ sin θ
tan = = . (0.19)
2 sin θ 1 + cos θ
These trigonometric relationships are those most commonly used. Trigonom-
etry is also used in conjunction with triangles, so a few important results
are derived here.
Consider the triangle shown in Figure 0.5:
r
a h c
p θ q
s b
Figure 0.5
h = a sin(π − θ) = sin θ.
c
h a
π−θ θ q
s p b
Figure 0.6
Throughout this chapter, the reader is assumed to be familiar with the usual
angles in the plane, basic geometric constructions, and basic trigonometry.
A brief review of these ideas is included in Chapter 0 and the appendices.
1.1 Angles
In this Chapter, all angles will be measured in degrees. Other units of
measurement may be employed later on; the reader will be given sufficient
notice when this occurs.
Table 1.1 lists some commonly used values of trigonometric functions.
With formulae such as sin(180◦ − α) = sin α, cos(180◦ − α) = − cos α, etc.,
we may easily obtain other values as necessary. For simplicity, we use the
abbreviation
√
τ = (1 + 5)/2. (1.1)
x x+y y 1
ρ= = =1+ =1+ .
y x x ρ
9
10 CHAPTER 1. ANGLES AND CONSTRUCTIONS
cos 0◦ = 1, sin 0◦ = 0,
q √ √ √
cos 18◦ = 1
2
5+ 5
2 = 1
2 τ + 2, sin 18◦ = 5−1
4 = 1
2τ = 12 (τ − 1),
√
cos 30◦ = 3
2 , sin 30◦ = 12 ,
√ q √ √
cos 36◦ = 5+1
4 = τ
2, sin 36◦ = 1
2
5− 5
2 = 1
2 3 − τ,
cos 45◦ = √1 ,
2
sin 45◦ = √1 ,
2
√
cos 54◦ = 1
2 3 − τ, sin 54◦ = τ2 ,
√
cos 60◦ = 21 , sin 60◦ = 3
2 ,
√
cos 72◦ = 21 (τ − 1), sin 72◦ = 1
2 τ + 2,
Table 1.1
1.2. REGULAR POLYGONS 11
q s
r
Figure 1.1
s2 = 12 + 12 − 2 · 1 · 1 · cos 72◦ .
12 CHAPTER 1. ANGLES AND CONSTRUCTIONS
b d
a O
Figure 1.2
36◦ 1
Figure 1.3
√
What is 2 − τ ? It happens (see §1.5 for a thorough discussion) that
2 − τ = τ −2 , so that √
−1 1 5−1
s=τ = = .
τ 2
We remark here that any expression involving √ “τ ” may be converted
√ into
an expression involving only rationals and “ 5” by using τ = ( 5 + 1)/2
and rationalizing
√ the denominator as necessary. Conversely, expressions
involving “ √ 5” may be written as expressions involving “τ ” by using the
relationship 5 = 2τ − 1. √
Now recall that in Figure 1.2, we have [Od] = ( 5 − 1)/2. Thus, in
creating our regular pentagon, we have created our decagon as well! This
might not have been √ so obvious if we failed to convert s = τ −1 into an
expression involving “ 5.”
—A hexagon variation.
One such polygon is a hexagon, each of whose angles has measure 120◦ ,
but whose sides alternate in the ratios τ and 1. Such a hexagon is called a
τ : 1-hexagon. As can be seen in Figure 1.4, there are two approaches to
14 CHAPTER 1. ANGLES AND CONSTRUCTIONS
Figure 1.4
2. Perpendicularly bisect bc at d.
3. With this same radius [bd], draw an arc centered at d from b counter-
clockwise to c.
−→
4. With center b and radius [bd], draw an arc counterclockwise from ab .
h
g
a
k j b
Figure 1.5
With data from above, we solve for [ch] to see that [ch] = √2 s, and hence
5
from Step 7 that [cj] = √2 s.
5
We now apply the cosine law for triangles to ∆cbj, yielding
1
[jb]2 − [jb]s + s2 = 0.
5
16 CHAPTER 1. ANGLES AND CONSTRUCTIONS
a c e
d
b
f
(a) (b)
Figure 1.6
1.3. POLYGON VARIATIONS 17
(a) (b)
Figure 1.7
Polygons described in this section; that is, equiangular polygons whose
sides alternate in length, are called quasi-regular polygons.
18 CHAPTER 1. ANGLES AND CONSTRUCTIONS
a c g b
d0 d e
Figure 1.8
1.5 Powers of τ
Recall that when discussing a construction of the regular decagon earlier,
it was helpful to know that 2 − τ = τ −2 . It is now time to look at such
calculations in more detail.
We first examine powers of τ . Since τ 2 = τ + 1, we know that
τ 3 = τ 2 + τ = (τ + 1) + τ = 2τ + 1.
Likewise,
τ 4 = τ 3 + τ 2 = (2τ + 1) + (τ + 1) = 3τ + 2.
20 CHAPTER 1. ANGLES AND CONSTRUCTIONS
Of course, we may keep going and calculate τ 5 , τ 6 , etc. On the other hand,
dividing the relationship τ 2 = τ + 1 by τ results in τ = 1 + 1/τ , so that
1
= τ −1 = τ − 1.
τ
Proceeding as above, we have 1 = τ −1 + τ −2 , so that
τ −2 = 1 − τ −1 = 1 − (τ − 1) = 2 − τ.
τ1 = τ, τ −1 = τ − 1,
τ 2 = τ + 1, τ −2 = 2 − τ ,
τ 3 = 2τ + 1, τ −3 = 2τ − 3,
τ 4 = 3τ + 2, τ −4 = 5 − 3τ ,
τ 5 = 5τ + 3, τ −5 = 5τ − 8,
τ 6 = 8τ + 5, τ −6 = 13 − 8τ .
Table 1.2
The reader familiar with the famous Fibonacci sequence will no doubt
find this table intriguing. For those unfamiliar, the Fibonacci sequence is
defined as follows: put F0 = 0, F1 = 1, and calculate the remaining members
of the sequence by the recurrence relation
Fn+2 = Fn+1 + Fn
Of course these are precisely the numbers occurring in Table 1.2. In fact,
the whole of this table may be summarized by the following rules, valid for
n ≥ 1:
τ n = Fn τ + Fn−1 , τ −n = (−1)n+1 (Fn τ − Fn+1 ).
1.6 Exercises
1. A regular 15-sided polygon is called a pentadecagon, or 15-gon.
c
g
b e d
a O
Figure 1.9
(a) Show that [Oe] = cos 72◦ and [Og] = sin 72◦ .
Show that ∠Oef has measure 60◦ , and conclude that [Of ] =
(b) √
3 cos 72◦ .
(c) Show that the sides of a regular 15-gon inscribed in a circle of
radius 1 have length 2 sin 12◦ .
(d) Since 12◦ = 72◦ − 60◦ , we may write
2 sin 12◦ = 2 sin (72◦ − 60◦ ) = 2 (sin 72◦ cos 60◦ − cos 72◦ sin 60◦ ) .
Using the results of (a) and (b), show that [Og] = 2 sin 72◦ cos 60◦
and [Of ] = 2 cos 72◦ sin 60◦ . Finally, show that [f g] = 2 sin 12◦ .
(e) (Due to Sara Fessler) Alternatively, consider ∆ef g. Knowing
∠ef g, ∠f eg, and [eg], conclude that [f g] = 2 sin 12◦ .
e
m n
d
k g
b c
j h
Figure 1.10
4. Recall from §1.1 that the ratio of the length of the longer sides to that
of the shorter side of an isosceles triangle with apex angle 36◦ (such
as ∆abc in Figure 1.6(b)) is τ . Argue geometrically that the ratio of
the length of the longer side to that of the shorter sides of an isosceles
triangle with apex angle 108◦ (such as ∆bcd in Figure 1.6(b)) is also
24 CHAPTER 1. ANGLES AND CONSTRUCTIONS
τ . Use these facts to show that the ratio of the lengths of the sides of
the decagram in Figure 1.6(b) is
√
[ef ] 5
= .
[ae] 2
√
5. Consider the following construction of a 2 : 1-octagon. Begin by
creating a, b, and c as in the construction of the regular pentagon (see
Figure 1.2). Draw a line through c and b, intersecting
√ the circle again
at f . Show that af and f e are two sides of a 2 : 1-octagon; that is,
[f e] √
= 2.
[af ]
g h j
a b c d e
f
k m n
Figure 1.11
27
28 CHAPTER 2. THE PLATONIC SOLIDS
V − E + F = 2. (2.1)
Take, for example, a cube, which has eight vertices (or corners), twelve
edges, and six square faces. Then we see that
V − E + F = 8 − 12 + 6 = 2.
The stage is now set for a discussion of the Platonic solids. A Platonic
solid is a polyhedron with the following properties:
(P1 ) It is convex.
(P2 ) Its faces are all the same regular polygon.
(P3 ) The same number of polygons meet at each of its vertices.
Note that since a Platonic solid is convex, the polygons referred to in
(P2 ) must also be convex.
Which polyhedra satisfy properties (P1 )–(P3 )? We provide two different
approaches to answering this question.
(a) (b)
(c)
Figure 2.2
antiprism) is that exactly three triangles (and one pentagon) meet at each
vertex. As a result, appending a pentagonal pyramid to both the top and
bottom of this antiprism yields a polyhedron with precisely five triangles
meeting at each of its twelve vertices (see Figure 2.4). The pentagonal
antiprism contributes ten equilateral triangles, and each of the pentagonal
pyramids contributes five, for a total of twenty triangles – hence the name
“icosahedron.”
(a)
(b)
(c)
Figure 2.3
Our search for Platonic solids with triangular faces ends here, for one is
easily convinced that the angles of six equilateral triangles comprise 360◦ ,
and hence any vertex with six equilateral triangles would be “flat.” This,
of course, does not result in a convex polyhedron, but rather a tiling of the
plane by equilateral triangles.
So now we have enumerated all possible Platonic solids with equilateral
triangles as faces. What about the next regular polygon, the square? Three
squares at a vertex results in our old friend the cube (sometimes called a
hexahedron), and at four squares we are already flat.
2.2. A GEOMETRIC ENUMERATION 31
(a) (b)
Figure 2.4
Figure 2.5
Perhaps the best way to imagine such a solid is to begin with an arrange-
ment of six pentagons as shown in Figure 2.5(a). Five of these pentagons
may be folded up to yield a bowl-like shape; as it happens, two such bowls
fit exactly together. The result, as it requires precisely twelve pentagons, is
called a dodecahedron (or sometimes a pentagonal dodecahedron).
32 CHAPTER 2. THE PLATONIC SOLIDS
Now each angle of a regular pentagon had measure 108◦ . Hence four
such angles have measure 432◦ > 360◦ , and hence it is impossible to fashion
a vertex of a convex polyhedron with four (or more) pentagons at a vertex.
So on to regular hexagons. With three hexagons at each vertex we are
already flat, yielding a hexagonal tiling of the plane. As a result, there are
no Platonic solids with regular hexagonal faces.
Our search stops here. Since three hexagons result in a flat vertex, three
regular polygons with more than six sides, if they met at a vertex, would
comprise more than 360◦ . So as with the case of four pentagons, no convex
polyhedron may be formed.
(P10 ), yielding
2E 2E
−E+ = 2.
q p
A little algebra yields
1 1 1 1
+ = + , (2.2)
p q 2 E
which must be valid for any Platonic solid.
Now p and q are integers 3 or greater, and E is a positive integer. So if
p ≥ 4 and q ≥ 4, we would have 1/p + 1/q ≤ 1/2, making (2.2) impossible.
As a consequence, we must have p = 3 or q = 3 (or possibly both).
Assume for the moment that p = 3. Then (2.2) becomes
1 1 1
= + . (2.3)
q 6 E
Since E is a positive integer, this means that 1/q > 1/6, or equivalently,
q < 6. Since at least three polygons must meet at the vertex of a polyhedron,
the only possibilities are q = 3, q = 4, or q = 5. With each choice of q, E
may be determined from (2.3). V may then be determined from (P30 ), and
F from (P20 ).
Since (2.2) is symmetric in the occurrence of “p” and “q” (each playing
precisely the same role), the reader is invited to make an analogous argument
with the assumption that q = 3. When this is done, Table 2.1 is obtained,
which includes all possibilities for p and q as described in the previous few
paragraphs.
As a final note, the polyhedron with regular faces of p sides, where q are
assembled at each vertex, is sometimes denoted by {p, q}. This notation for
referring to a polyhedron is called a Schläfli symbol.
3 3 6 4 4 Tetrahedron {3, 3}
3 4 12 6 8 Octahedron {3, 4}
4 3 12 8 6 Cube {4, 3}
3 5 30 12 20 Icosahedron {3, 5}
5 3 30 20 12 Dodecahedron {5, 3}
Table 2.1
34 CHAPTER 2. THE PLATONIC SOLIDS
Notice that there are exactly five possibilities, each corresponding to one
of the Platonic solids described in the previous section. As expected, an
algebraic approach yields the same set of Platonic solids as an incremental
geometric approach.
2.4. EXERCISES 35
2.4 Exercises
1. Build the five Platonic solids using the nets provided at the end of
the chapter. (An arrangement of polygons which may be folded to
produce a polyhedron is called a net for that polyhedron.)
2. Find all possible nets for the cube. In other words, find all arrange-
ments of six contiguous squares in the plane which may be folded to
obtain a cube. (Two nets are considered the same if one may be ob-
tained from the other by a rotation and/or a reflection.)
4. Color the faces of an octahedron with four colors so that all four colors
are incident at each vertex. Then color the faces of an icosahedron with
five colors so that all five colors are incident at each vertex.
9. (a) Build four tetrahedra and one octahedron so that all of the poly-
hedra have the same edge length. Arrange them to form a larger
tetrahedron. Use this construction to find the ratio of the volume
of an octahedron to the volume of a tetrahedron with the same
edge length as the octahedron.
(b) Build eight tetrahedra and one octahedron so that all polyhedra
have the same edge length. On each face of the octahedron,
affix a tetrahedron. What Platonic solid do the exposed vertices
form? The resulting figure is called a stella octangula, and may
be thought of as two large tetrahedra intersecting in a common,
smaller octahedron.
36 CHAPTER 2. THE PLATONIC SOLIDS
10. A tetrahedron may be cut into two congruent parts by a plane parallel
to and midway between a pair of opposite edges of the tetrahedron
(see Figure 2.6). Build two of these pieces and arrange them to form
a tetrahedron.
Figure 2.6
11. A cube may be cut into two congruent parts by a plane perpendicularly
bisecting a long diagonal of the cube (see Figure 2.7). Build two of
these pieces and arrange them to form a cube.
Spherical Trigonometry
30◦
Figure 3.1
Now in the plane, we measure the six “parts” of a triangle by measuring
the lengths of the sides (relative to some standard unit), and the measures
of the angles between adjacent sides. The situation is somewhat different
for spherical triangles, where all “parts” are, in fact, angles.
37
38 CHAPTER 3. SPHERICAL TRIGONOMETRY
surface of the sphere occupied by the spherical triangle. Thus, our triangle
1 ◦ ◦ 1
occupies 720 ◦ (210 − 180 ) = 24 of the surface of the sphere, a fact which
1
0< (Σ − 180◦ ) < 1,
720◦
from which it readily follows that 180◦ < Σ < 900◦ . In other words, not
only must the vertex angles of a spherical triangle sum to more than 180◦ ,
they must also sum to less than 900◦ .
For our purposes, we shall always assume that the vertex angles of a
spherical triangle are less than 180◦ in measure. It is also possible to specify
that the three arcs described above determine a large triangle enclosing the
other 23 ◦ ◦ ◦
24 of the sphere whose vertex angles are 330 , 270 , and 270 . The
necessity of always making such distinctions is avoided with this simplifying
assumption.
3.2. FUNDAMENTAL RELATIONSHIPS 39
c
B r
a
A C
q
b p
p0
r0
q0
Figure 3.2
q r
q0 p0
r0
r0
b a
c
Figure 3.3
sec2 a + sec2 b − 2 sec a sec b cos c = tan2 a + tan2 b − 2 tan a tan b cos C.
2 sec a sec b cos c = (sec2 a − tan2 a) + (sec2 b − tan2 b) + 2 tan a tan b cos C.
3.3. EDGE ANGLES OF PLATONIC SOLIDS 41
Using the fact that sec2 a − tan2 a = sec2 b − tan2 b = 1 and then dividing by
2 gives
sec a sec b cos c = 1 + tan a tan b cos C.
Of course, the same derivation could be used to find formulas for cos a or
cos b; we would find that
There are three other relationships among the various angles of the spher-
ical triangle ∆abc which will also be important to us, but whose derivation
is not so straightforward. Thus, we include the results but omit their deriva-
tion here.
We now wish to apply our results from spherical trigonometry to the deriva-
tion of angular properties of the Platonic solids. Our first task is to calculate
the “edge angle” of the regular icosahedron; that is, the angle subtended by
an edge when its endpoints are connected to the center of the polyhedron
(see Figure 3.4). To do this, we first imagine the icosahedron to be inscribed
in a sphere. Three vertices of a triangular face will lie on the sphere, which
we may imagine as the vertices of a spherical triangle. We think of centrally
projecting a face of the icosahedron onto the sphere to obtain this trian-
gle. Note that all edge angles are equal as are all vertex angles due to the
symmetry of the icosahedron.
42 CHAPTER 3. SPHERICAL TRIGONOMETRY
vertex of
icosahedron
a A
a
vertex of A
icosahedron a A vertex of
icosahedron
center of icosahedron
Figure 3.4
Using the fact that sin2 A = 1 − cos2 A, we may solve this equation for cos a,
yielding
cos 72◦ τ −1 1 1
cos a = ◦
= = (2τ − 1) = √ .
1 − cos 72 3−τ 5 5
3.3. EDGE ANGLES OF PLATONIC SOLIDS 43
A A
A A
A
Figure 3.5
−1
Now how was τ3−τ reduced to 51 (2τ − 1) above? Rather than digress too
far off task, this discussion is postponed until §3.7, where all the details are
presented. √
Continuing our discussion of the icosahedron, we saw that cos a = 1/ 5,
so that a ≈ 63.4◦ . Note that a is slightly larger than 60◦ , and thus the tetra-
hedron formed by a face of the icosahedron and the center of the icosahedron
is not quite regular.
We call the angle a just obtained E3,5 to symbolize that it is an edge
angle of the icosahedron. Similarly, we denote by Ep,q the edge angle of the
Platonic solid {p, q} as enumerated in Table 2.1. It is left to the reader to
verify that the procedure used to find E3,5 may be used to find E3,3 and E3,4 ,
resulting in
cos 120◦ 1
cos E3,3 = =−
1 − cos 120◦ 3
and
cos 90◦
cos E3,4 = = 0.
1 − cos 90◦
We now wish to tackle the task of finding E5,3 . Since the faces of the
dodecahedron are not triangles themselves, we “create” a spherical triangle
by projecting the center of a pentagonal face onto the circumscribing sphere.
Since the ends of the edges (i.e., the vertices of the dodecahedron) lie on the
sphere, the spherical triangle ∆rst in Figure 3.6 is obtained. The angle A
has measure 72◦ (see Figure 3.5), and the angles across from A have measure
60◦ as six such equal angles comprise a complete revolution about a vertex
of the dodecahedron when projected onto the sphere.
We may find E5,3 using (3.6), yielding
cos A = − cos 60◦ cos 60◦ + sin 60◦ sin 60◦ cos E5,3 .
44 CHAPTER 3. SPHERICAL TRIGONOMETRY
E5,3 60◦
s 60◦
A
r
Figure 3.6
√
Since cos 60◦ = 1
2 and sin 60◦ = 3
2 , we may rewrite this expression as
4 cos A + 1
cos E5,3 = .
3
Since A = 72◦ , we find using Table 1.1 that
√
2τ − 1 5
cos E5,3 = = .
3 3
We may likewise find cos E4,3 by similar methods, the major difference
being that A = 90◦ rather than A = 72◦ . This results in
1
cos E4,3 = .
3
Details are left to the Exercises.
It may be remarked that the methods used to find E5,3 and E4,3 may
be used to find E3,3 , E3,4 , and E5,3 as well. This results in the following
generalization: if the Platonic solid has faces with p edges and there are q
such faces about each vertex, then the edge angle Ep,q is given by
cos 2π 2π
p + cos q 1 + cos 2π 2π
q + 2 cos p 1 + cos 2π
p
cos Ep,q = = =2 − 1, (3.7)
sin2 πq 1 − cos 2π
q 1 − cos 2π
q
m
r s
ρ ρ
E E
2 2
Figure 3.7
Happily, the calculation of the edge angles in the previous section will
come in quite handy. Take, for example, Figure 3.7. Here, O is the center
of the Platonic solid, and r and s are adjacent vertices, so that ∠rOs is
the edge angle E of the polyhedron and ρ its circumradius. (We write “E”
rather than “Ep,q ” here as some of the formulas will be valid even if E does
not correspond to the edge angle of a Platonic solid, but rather some other
polyhedron.) Also, m is the midpoint of rs, so that [rm] = [ms] = 1 and
Om bisects E. It is easy to see in the right triangle ∆Oms that
1 1
sin E = .
2 ρ
Since we have data concerning cos E already, we employ the appropriate
trigonometric identity to obtain
r
1 2
ρ = csc E = . (3.8)
2 1 − cos E
46 CHAPTER 3. SPHERICAL TRIGONOMETRY
Hence knowing one of ρ and E, we may easily find the other using (3.8).
Data are summarized for reference in Table 3.1 at the end of §3.5. Note
that for typographical clarity, ρ2 is included in the table rather than ρ.
108◦
D5,3 D5,3
b 108◦ D5,3 d
c 108◦
Figure 3.8
cos 108◦ = cos 108◦ cos 108◦ + sin 108◦ sin 108◦ cos D5,3 .
q edges, all of which are equal to some angle ϕ, then with the assumption
that all vertex angles are equal to D, we obtain the following formula:
1 + cos 2π
q
cos D = 1 − 2 . (3.9)
1 + cos ϕ
A proof is given in the Exercises.
Now let us consider a Platonic solid whose faces have p sides, q of which
meet at a vertex, and whose dihedral angle is Dp,q . If we proceed as with
the dodecahedron, we obtain a spherical polygon with q edges of measure
ϕ, where ϕ represents the angle of a face of the Platonic solid. Of course,
the vertex angles of this polygon have measure Dp,q .
We recall that in a polygon of p sides, the sum (in radian measure) of
the angles of the polygon is (p − 2)π. When the polygon is regular, each of
the p angles has the same measure, so that
1 2π
ϕ= (p − 2)π = π − .
p p
Now
2π 2π 2π 2π
cos ϕ = cos π − = cos π cos + sin π sin = − cos , (3.10)
p p p p
so that substituting this value in (3.9) yields
1 + cos 2π
q
cos Dp,q = 1 − 2 . (3.11)
1 − cos 2π
p
A summary of edge and dihedral angles and circumradii for the Pla-
tonic solids is given below (where, for example, we write “D” for “Dp,q ” for
formatting purposes):
Tetrahedron 3 3 − 13 109.5◦ 1
3
√1
2
70.5◦ 3
2
√
Octahedron 3 4 0 90◦ − 31 2 109.5◦ 2
Cube 4 3 1
3 70.5◦ 0 1 90◦ 3
√
Icosahedron 3 5 √1
5
63.4◦ − 3
5
τ2 138.2◦ τ +2
√
Dodecahedron 5 3 3
5
41.8◦ − √15 τ 116.6◦ 3τ 2
Table 3.1
48 CHAPTER 3. SPHERICAL TRIGONOMETRY
lines, there is typically a shortest path between two given points. Such a
path is called a geodesic path.
On a sphere, the shortest path between two points is an arc of a great
circle, which would be a geodesic path. Note that there are usually two such
arcs, one measuring less than 180◦ and one measuring more – the geodesic
path is the one measuring less.
q y
p
G
Figure 3.9
x y
p q
Figure 3.10
We know that any line segment actually lies along a straight line. In
other words, we may extend a segment in either direction to produce a line.
Of course a line may not be extended any further – it is already as “long”
as possible. A line also has the shortest path property discussed earlier:
given any two points x and y on a line, the segment xy (that is, the geodesic
50 CHAPTER 3. SPHERICAL TRIGONOMETRY
path between x and y) lies along the line. Thus a line is called a maximal
geodesic path: maximal because it cannot be extended any further (that is,
no points may be added) while still maintaining the shortest path property.
This may all seem obvious in the plane, but remember that most ge-
ometries are non-Euclidean. So it is vitally important to be able to talk
about lines without using concepts that apply only in a Euclidean realm.
For example, although we may imagine a line as a straight line segment ex-
tended infinitely in both directions, we may also think of a line as a maximal
geodesic path. The advantage of the latter description is that it is applicable
in other domains, such as on a curved surface (like a sphere).
So what are maximal geodesic paths on the sphere? Since shortest paths
on a sphere are always arcs of great circles, it follows that maximal geodesic
paths are the great circles themselves. In our geometry, we shall refer to
maximal geodesic paths as lines.
This may seem like a long-winded way of defining a line, but it has the
advantage of being applicable on most curved surfaces usually encountered.
It is now a simple matter to define a line segment as part of a line bounded
by two distinct points. And, as we have seen, a triangle consists of three
line segments, a quadrilateral four, etc. In other words, many of the familiar
definitions may be recast in a non-Euclidean context.
But much is different. Below we highlight several differences between
geometry in the Euclidean plane and geometry on the sphere.
the sphere was not needed. We were able to calculate ρ in §3.4 only in
reference to a Platonic solid of edge length 2. Defining a Euclidean arc
length (such as for calculating distances on the surface of the Earth)
requires knowing the radius of the sphere. Most of the important
results we encounter in spherical geometry and trigonometry will be
independent of the radius of the sphere.
τ −1 1 1
= (τ + 2)(τ − 1) = (τ 2 + τ − 2)
3−τ 5 5
1 1 1
= (τ + 1 + τ − 2) = (2τ − 1) = √ .
5 5 5
Such simplifications will occur frequently, and the reader may duplicate the
above procedure in order to verify them.
3.8 Exercises
√
1. In the same manner in which it was shown that cos E5,3 = 5/3 in
§3.3, show that cos E4,3 = 1/3.
ϕ ϕ
2π 2π 2π
q q q
D D
ϕ ϕ
D D
ϕ
Figure 3.11
3. Proceed as follows to show (3.7); that is, show that if a Platonic solid
has faces with p edges and there are q such faces about each vertex,
then the edge angle Ep,q is given by
cos 2π 2π
p + cos q 1 + cos 2π 2π
q + 2 cos p 1 + cos 2π
p
cos Ep,q = = =2 − 1.
sin2 πq 1 − cos 2π
q 1 − cos 2π
q
π
q
center of face 2π Ep,q
p π
(projected q
onto sphere)
center of polyhedron
Figure 3.12
54 CHAPTER 3. SPHERICAL TRIGONOMETRY
p 3 3 4 3 5
q 3 4 3 5 3
5 5
r 4 3 3 2 2
Table 3.2
(c) Use the result from (b) with (a) to show that
π π
ρ = sin sec . (3.16)
q r
5. We now wish to derive formula (3.13) at the end of §3.5; that is,
2π 1 2π 1 π π
sin sin Dp,q = sin cos Ep,q = 2 cos cos .
p 2 q 2 p q
21
1 + cos 2π
q
sin Dp,q = .
2 1 − cos 2π
p
3.8. EXERCISES 55
1 1 + cos 2π
p
cos2 Ep,q = .
2 1 − cos 2π
q
6. Using the identity sin 2θ = 2 sin θ cos θ, conclude from the previous
exercise that
1 π π
sin Dp,q sin = cos
2 p q
and
1 π π
cos Ep,q sin = cos .
2 q p
7. By building a geometrical model, show that D3,3 + D3,4 = π.
10. Analogous to the construction in the previous exercise, build six square
pyramids which, when assembled appropriately, form a cube. Each
pyramid has as its base a face of the cube and as its apex the center
of the cube. Use data from Table 3.1 as necessary.
11. Consider the spherical pentagon shown in Figure 3.13. Show that as
a shrinks to 0, the pentagon becomes planar in the sense that 2θ gets
closer and closer to 108◦ . In other words, show that
1
lim cos 2θ = − .
a→0 2τ
a θ
θ
72◦
Figure 3.13
Chapter 4
Geodesic Structures
57
58 CHAPTER 4. GEODESIC STRUCTURES
r
q
c
b
E3,5
E3,5 E3,5
a p
p O
a
Figure 4.1
We may widen the arc pqrp by including an auxiliary arc abca and a tab
at segment pd so that this thick arc may be cut out, folded along segments
bq and cr, and then glued to form a thick spherical triangle. Since an icosa-
hedron has twenty faces, we need just twenty of these spherical triangles,
glued back-to-back, to form a geodesic icosahedron (as in Figure 4.2).
Figure 4.2
Recall (see §3.6) that given any two points on a sphere, the shortest path
between them (on the sphere) is the arc of a great circle passing through
the points. (The analogue in the plane is that the shortest path between
two points is the line segment joining them.) Since such shortest paths
are geodesics, and since such paths were used to construct our spherical
icosahedron, we describe the resulting structure as “geodesic.”
4.2. A GEODESIC DODECAHEDRON 59
r q
E5,3 60◦ p
q 60◦ r
72◦
b p
E5,3
b
b p
O O
(a) (b)
(c)
Figure 4.3
60 CHAPTER 4. GEODESIC STRUCTURES
We see from the considerations in §3.3 that ∠qOr has measure E5,3 , and
the symmetry of the configuration requires that ∠pOq and ∠rOp have the
same measure, which we shall call b. Of course, to find b is simply a matter
of applying (3.6) once again, yielding
cos 60◦ = − cos 60◦ cos 72◦ + sin 60◦ sin 72◦ cos b,
from which it follows that
√ √ 3
τ +1 τ +1 τ +2 τ +2 1 (τ + 2) 2
cos b = p = √ = √ (τ + 1) (3 − τ ) = √ .
3(τ + 2) 3 τ +2 3 5 5 3
Thus, to create a triangle as in Figure 4.3(b), we must create angles
E5,3 ≈ 41.8◦ , b ≈ 37.4◦ .
Building this model is a good first project for those interested in creating
more complex geodesic structures.
72◦
t u
q p
s
(a) (b)
Figure 4.4
4.4. 4-FREQUENCY ICOSAHEDRA 61
θ 1 + cos θ θ 1 − cos θ
cos2 = , sin2 = ,
2 2 2 2
we find after substitution and some algebra that
1 1 1 1 1 τ
cos a = 1+ √ + 1− √ · (τ − 1) = .
2 5 2 5 2 2
Thus
1
E3,5 ≈ 31.7◦ , a = 36◦ .
2
This information is sufficient to create the four smaller triangles in Figure
4.4(a), and hence the entire model (shown in Figure 4.4(b)). Note that the
purple spherical triangles are the same as those shown in Figure 4.2, while
the yellow triangles are the ones inscribed inside them.
This structure is called a 2-frequency icosahedron because it is ob-
tained by dividing the sides of an icosahedral spherical triangle into two
congruent arcs. Higher frequency icosahedra are possible, and the reader is
invited to peruse Magnus Wenninger’s excellent Spherical Models for exam-
ples of these structures and many others.
p q
a
t s
(a) (b)
Figure 4.5
t y u
AB
a c d
e x d
w z
D F
E
a b e c
q A p
a v a s
Figure 4.6
t y u
AB
a00 c d
e0 x d
w0 F
z
F
a0 b0 e0 c
q G A p (b)
a0 v 0 a 00 s
(a)
Figure 4.7
Using (3.4) and (3.6) as necessary, it may be verified that in this instance,
we have G ≈ 64.466◦ , a0 ≈ 14.545◦ , and a00 = 21 E3,5 − a0 ≈ 17.172. The
reader should take a moment to imagine exactly what spherical subdivision
was just found: nothing more than Figure 4.5(a) – inscribed on a face of the
icosahedron – projected onto the icosahedron’s circumsphere. Note that the
symmetry of the spherical triangle along qu implies that after projection, x
is in fact the midpoint of ts. But there is not sufficient symmetry to force
w0 to be the midpoint of qt. Figure 4.7(b) shows the completed 4-frequency
icosahedron.
These subdivisions of spherical triangles are called Class I subdivisions
in the literature (see the works of Joe Clinton and Hugh Kenner) because
the planar subdivisions upon which the spherical subdivisions are based are
formed by lines parallel to the sides of the original triangle. A discussion of
the case where additional lines are perpendicular to the sides of the original
triangle takes place in Chapter 10.
4.5. ANOTHER 4-FREQUENCY ICOSAHEDRON 65
n
e h l e
w x d z
a f −h f
q p
Figure 4.8
Now applying (3.4) to spherical ∆xnl yields
cos d = cos2 h + sin2 h cos 120◦ ,
which may be rewritten as
3 1
cos d = cos2 h − . (4.1)
2 2
66 CHAPTER 4. GEODESIC STRUCTURES
sin 3a sin f
sin 3a = ◦
= ,
sin 90 sin 36◦
68 CHAPTER 4. GEODESIC STRUCTURES
so that sin f = sin 36◦ sin 3a. Since a = 14 E3,5 ≈ 15.859◦ , we may solve this
equation for f . Now it is clear from Figure 4.8(b) that 2f = 3d, so that
d = 23 f . In this case, we find that d ≈ 17.143◦ . Using this method to find
b and c, we get b ≈ 18.486◦ and c = 18◦ . This technique may be applied
to a subdivision of a spherical icosahedral triangle of arbitrary frequency, as
seen in the Exercises.
a
p b q
a
c c
r s
a
d d d
t u
w
a
(a) 36◦
3a
f = 32 d
t x u
d d d
a
(b)
Figure 4.9
We again emphasize that the data obtained by this technique are not
quite exact, but rather are approximations which are sufficiently accurate
for the purpose of building paper geodesic models.
4.6. STILL ANOTHER 4-FREQUENCY ICOSAHEDRON 69
Table 4.1
a a
b
a e e a
c c
a c d d c a
e d e
a b e c c e b a
a a a a
Figure 4.10
70 CHAPTER 4. GEODESIC STRUCTURES
arcs (rather than the five of §4.4). Of course, only four different arcs were
required in §4.6, but this section produced only an approximate subdivi-
sion. The subdivisions in §4.4 and§4.5 were based on exact computations
using formulas of spherical trigonometry (although for calculation purposes,
approximations were found).
Thus finding precise values for high-frequency subdivisions of spherical
triangles is not for the mathematically timid. It is here that one comes
face-to-face with the consequences of working with non-Euclidean geometry.
4.8 Exercises
1. Build a geodesic icosahedron with the template provided.
4. This exercise lays the groundwork for the following two. Here, ∆pqr
is a spherical icosahedral triangle whose sides are divided into equal
fourths. st, tu and uv are arcs of great circles, with st and uv inter-
secting at w. The goal of this exercise is to determine values of all the
labelled angles in Figure 4.11. The reader may attempt to calculate
these on his or her own, or proceed according to the following outline.
72◦
s
x
a f = ts
w c
u v
D E
a b c
◦ D
q 72 B A 72◦ p
a t a y
Figure 4.11
72 CHAPTER 4. GEODESIC STRUCTURES
5. We begin with Figure 4.11 and join x rand y by an arc of a great circle,
as shown in Figure 4.12(a). As foreshadowed in §4.4, this arc does
not intersect arcs st and uv at w, but 72◦rather creates a small triangle
shown in Figure 4.12(a) and enlarged in Figure 4.12(b). The symmetry
of the construction in Figure 4.12(a) should convince the reader that
s
this small triangle is indeed isosceles. Here, we calculate the remaining
angles and sides of this triangle. The following outline may be used.
c E
u w v
z h
w
a b g G G E
l k h=c−g
D G
q 72◦ B A F p
a t a y z
(a) (b)
Figure 4.12
(a) Recall that arc xy has measure 36◦ . Use (3.4) with spherical
∆xyq to calculate F . (Alternatively, note that F is the same
angle as A in Figure 4.6.)
(b) Use (3.6) with spherical ∆tzy in order to find G. Then apply
(3.6) again to determine g. (Note that g is less that c of Figure
4.10; it is for this reason that we know that arc xy intersects arc
tw rather than arc ws.)
(c) Note that h = c − g. Check this calculation by using (3.6) with
the triangle in Figure 4.11(b) to find h. How close is your result?
4.8. EXERCISES 73
(d) Finally, use (3.6) with this same triangle in order to calculate k.
(e) To confirm the result in (d), find l by applying (3.4) to ∆zty, and
note that k + 2l = 36◦ .
j s
x z
d
u H w v
E
H
g
t y k
Figure 4.13
(a) Square both sides of each equation, and solve both equations for
2 cos a cos b cos c.
(b) By setting the two expressions obtained in (a) for 2 cos a cos b cos c
equal to each other and replacing all cosines by sines using the
relationship cos2 θ = 1 − sin2 θ, deduce that
sin2 A sin2 B
2 = .
sin a sin2 b
(c) With the assumption that all angles are between 0◦ and 180◦ ,
extract square roots to obtain part of (4.4).
(d) Without repeating these calculations, complete a proof of (4.4).
5.1 Truncation
The process of truncation – that is, “chopping off” vertices of a polyhedron
to produce a different polyhedron – results in many interesting polyhedra
when applied to the Platonic solids. Such polyhedra are the focus of this
section.
—Partial truncation.
The first method of truncation results in the edges of the original Pla-
tonic solid remaining partially intact. Such “partial” truncation, applied to
a tetrahedron, is illustrated in Figure 5.1. As long as we truncate “evenly”
– that is, the plane of truncation cuts off equal segments from all the edges
which it intersects – we find that four small regular tetrahedra have been
removed from the original tetrahedron, leaving equilateral triangles where
the vertices once were. Note also that the truncation process transforms the
triangular faces of the original tetrahedron into irregular (although equian-
gular) hexagons.
Figure 5.1
75
76 CHAPTER 5. THE ARCHIMEDEAN SOLIDS
Figure 5.2
1
This definition is not quite precise; a more accurate definition of an Archimedean solid
would involve the notion of a symmetry group, upon which we shall not elaborate at this
point.
5.1. TRUNCATION 77
(a) (b)
Figure 5.3
As far as a partial truncation of the cube goes, we see that each vertex
of the cube becomes an equilateral triangle as three squares meet at each
vertex of the cube. However, regular octagons are not created in the square
faces by dividing
√ the edges of the cube into equal thirds, but rather the
ratios 1 : 2 : 1 (see Figure 1.7(a)). The truncated cube is shown in
Figure 5.4.
Figure 5.4
1
s s
1
1 1
(a) (b)
(c)
Figure 5.5
—Complete truncation.
A truncation process which bisects the edges of a Platonic solid yields
a complete truncation of that polyhedron – complete in the sense that
nothing remains of the original edges of the solid except their midpoints.
Because of the great symmetry of the Platonic solids, complete trunca-
tion of one of them invariably yields an Archimedean solid.
Our first example is the tetrahedron. Complete truncation results in slic-
ing off tetrahedra whose edge length is half that of the original tetrahedron
(see Figure 5.6(a)). The result is nothing more than an octahedron! This
5.1. TRUNCATION 79
(a) (b)
Figure 5.6
Let us now examine a complete truncation of the cube. Again, the ver-
tices are replaced by equilateral triangles, but the square faces are replaced
with smaller squares (see Figure 5.7(b)) rather than octagons. The resulting
polyhedron (see Figure 5.7(a)) is called the cuboctahedron.
(a) (b)
Figure 5.7
80 CHAPTER 5. THE ARCHIMEDEAN SOLIDS
Figure 5.8
Figure 5.9
2
If “icosi” is the Greek prefix for 20, why is “icosahedron” spelled with an “a”? It
happens that “icosihedron” would, in Greek, result in two consecutive “e” sounds, and
hence the prefix is modified to “icosa” in this case.
5.2. TRUNCATION DEFORMATION 81
(a) (b)
Figure 5.10
Happily, all is not lost. In this case, it is possible to deform our partially
truncated cuboctahedron into an Archimedean solid by moving the regular
hexagons in slightly toward the center of the polyhedron, thereby scrunching
the rectangles into squares and creating regular octagons in the process. The
result is the polyhedron in Figure 5.11, often called the rhombitruncated
82 CHAPTER 5. THE ARCHIMEDEAN SOLIDS
Figure 5.11
We expect a similar situation to arise upon considering a complete trun-
cation of the cuboctahedron, and indeed it does. A complete truncation
of the cuboctahedron which results in its edges being bisected is shown in
Figure √5.12(a). Here, we see eight equilateral triangles, six squares, and
twelve 2 : 1 rectangles. Like the polyhedron in Figure 5.10(a), this solid
is vertex-regular in that the same assortment of polygons is present at each
vertex, but as not all the rectangles are squares, it is not Archimedean.
Again, happily, we may deform this solid into one which is Archimedean by
moving the triangles slightly toward the center of the polyhedron, thereby
scrunching the rectangles into squares, and the six squares into six slightly
smaller squares. The result is shown in Figure 5.12(b), and is called the
rhombicuboctahedron.
(a) (b)
Figure 5.12
5.3. SNUB POLYHEDRA 83
(a) (b)
Figure 5.13
The process by which the preceding four Archimedean solids were created
inspire one to call them truncation deformations. The curious reader
may inquire as to why this process of truncation deformation was not also
applied to the partially truncated Platonic solids of §5.1. As it happens,
regardless of the effort invested, no Archimedean solids can be produced by
deforming truncations of the partially truncated Platonic solids. Such are
the perils of the ambitious geometer.
Now let us begin with a cube, expand the faces somewhat and rotate
each of the faces a bit less than 45◦ as in Figure 5.15(a) (the calculation of
the precise distance and angle is somewhat involved and is not necessary for
our purposes). The squares do not meet in this instance, but done properly,
there is just enough room to weave interlocking rings of triangles around
these squares, as in Figure 5.15(b). The result is called the snub cube,
“snub” referring to the slight twist of the square faces.
It was important that our rotations of the squares were consistent – were
six individuals small enough to stand on the outside of the square faces, they
would all have experienced a slight rotation to their right. However, had we
chosen to rotate slightly to the left, a slightly different polyhedron would
have been formed – the mirror image of the first (shown in Figure 5.15(c)).
In other words, one would see Figure 5.15(c) if one were to look at a reflection
in a mirror of Figure 5.15(b). These are different polyhedra in the sense that
if we tried to superimpose Figure 5.15(b) onto Figure 5.15(c) such that one
of the square faces of the former coincided with one of the latter, not all of
the rest of the faces would coincide in pairs. It is for this reason that we
say that the snub cube exists in enantiomorphic pairs. None of the other
Archimedean solids discussed up to this point have this property.
The other Archimedean solid with this property is the snub dodecahe-
dron, formed by a process similar to that illustrated in Figure 5.15, except
that we begin with a dodecahedron, expand outward and rotate slightly its
pentagonal faces, and surround these pentagons with rings of interlocking
triangles. The result, an enantiomorphous pair (depending upon whether
the rotation of the pentagonal faces is clockwise or counterclockwise), is
shown in Figure 5.16(a),(b).
(a) (b)
Figure 5.16
As a final remark, we note that the same process of expansion and ro-
tation may be carried out with the tetrahedron as well. The result, in-
terestingly enough, is our friend the icosahedron! In this case, the enan-
tiomorphism is not apparent unless the faces of the icosahedron are colored.
For example, the faces of the tetrahedron may be colored, say red, before
their expansion. If the surrounding rings of triangles are, say yellow, then
the clockwise and counterclockwise rotations of the red faces will yield two
enantiomorphously colored icosahedra (see Figure 5.17).
(a) (b)
Figure 5.17
86 CHAPTER 5. THE ARCHIMEDEAN SOLIDS
Figure 5.18
Archimedean solids often neglect the two infinite families discussed in this
section (or perhaps mention them in a parenthetical remark).
—A particular case.
We begin with the following question – which Archimedean solids have
faces consisting of both equilateral triangles and squares? Our analysis is
modelled after the discussion in §2.3.
Now we know that at each vertex there is an arrangement of triangles and
squares, but we do not know how many of each there are – these unknowns
we denote by t and s, respectively. We denote, as usual, the number of
vertices, edges, and faces of the Archimedean solid by V , E, and F .
How many triangles are there on this polyhedron? Let us denote this
number by T . Then the total number of sides contributed by all of the
triangles is 3T . Now at each vertex meet 2t sides of triangles (two for each),
and hence we have 2tV for the entire polyhedron. But this counts each side
twice, for each side of each triangle is incident at two vertices. Hence, tV is
also the total number of sides of all triangles of the polyhedron, and hence
3T = tV . (Note how this differs from the Platonic case – since we do not
know whether two triangles, one triangle and one square, or two squares
meet at a particular edge of the polyhedron, we cannot count the number
of edges of the polyhedron contributed by the triangular faces, but we must
rather count the number of sides of the contributing triangles.)
Denoting the number of squares on the polyhedron by S, a similar anal-
ysis yields that 4S = sV . Since our polyhedron only has triangles and
squares as faces, we see that F = T + S; using the relationships just derived,
it follows that
tV sV
F = + . (5.1)
3 4
(Note that (P 20 ) and (P 30 ) of §2.3 imply that pF = qV and hence F =
qV
p ; since we have two different polygons as faces, (5.1) is an Archimedean
analogue of this relationship.)
88 CHAPTER 5. THE ARCHIMEDEAN SOLIDS
(t + s)V tV sV
V − + + = 2,
2 3 4
or after a little algebra,
t s 2
+ + = 1. (5.4)
6 4 V
We now wish to find all possible solutions to (5.4). Since s represents the
number of squares at each vertex, then certainly s < 4 (or else the vertex
would be flat); moreover, since we assumed that our polyhedron has square
faces, it follows that s ≥ 1. We examine the three possibilities.
t 2 3
+ = .
6 V 4
Since we must have at least three polygons at each vertex, then t ≥ 2;
it also follows that t < 5 (else the sum of the angles around a vertex
would exceed 360◦ and our polyhedron would not be convex). When
t = 2, we have V = 24 5 , an impossibility. (Note that a square pyramid
has one square and two triangles at each vertex of its base, but this
results in four triangles at its apex.) When t = 3, we have V = 8,
resulting in a square antiprism. And when t = 4, we have V = 24,
giving a snub cube.
5.5. AN ALGEBRAIC ENUMERATION 89
there is no Archimedean solid possible with one triangle, one pentagon, and
one 20-gon at each of its 24 vertices. Therefore, each solution of (5.7) yields
a candidate for an Archimedean solid, not a guarantee that we have found
one. The situation is decidedly more complex than our algebraic discussion
of the Platonic solids – the ambitious reader is encouraged to consult the
Exercises for a thorough investigation.
Two particular families of solutions are noteworthy; both involve only
two types of polygons at each vertex. For the first, we assume that two
squares and one other polygon meet at each vertex. Then s1 = 4, n1 = 2,
and n2 = 1. Substituting these values into (5.7) yields V = 2s2 , which cor-
responds to the infinite family of prisms. Each polygon of s2 sides generates
a prism with 2s2 vertices. The second is the infinite family of antiprisms,
where substituting s1 = 3, n1 = 3, and n2 = 1 into (5.7) also yields V = 2s2 ,
thus giving one antiprism for each regular polygon.
5.6 Exercises
This set of Exercises enumerates the Archimedean solids by finding all al-
gebraic solutions to (5.7) which correspond to geometrically realizable poly-
hedra.
s3 s2
s2 s3
s1
s1 s2
Figure 5.19
(a) Using an argument similar to that used in (1), show that s1 must
be even.
(b) Begin with the case s1 = 4, so that s2 and s3 are different and
not equal to 4. (You should find only one solution.)
(c) Argue algebraically that when s1 ≥ 6, no other solutions to (5.7)
are possible.
4. Now consider the case where four polygons meet at a vertex, but there
are only two different types of polygons.
5. Finally, consider the case where five polygons meet at a vertex (it is
easy to show that six are impossible). Argue that at least four of these
polygons must be triangles, and find the remaining two Archimedean
solids.
Show that if the angular deficiencies are summed over all vertices of
an Archimedean solid, the result is always 720◦ . (Hint: Use (5.7).)
ft = Mt f .
(b) Find the analogous matrix MT for the process of complete trun-
cation, i.e., each edge of the original polyhedron generates just
one vertex of its complete truncation (see §5.1).
(c) Consider the following expansion procedure. As an example,
consider a cube. Now move the faces of the cube radially out-
ward (as in Figures 5.14 and 5.15) but do not rotate them. Thus,
the cube is transformed into a rhombicuboctahedron, a tetra-
hedron would become a cuboctahedron, while a dodecahedron
would generate a rhombicosidodecahedron. Analogous to (a) and
(b), find the matrix Me corresponding to the expansion process.
(d) Find the analogous matrix Md for the process of taking the dual
of a polyhedron (as described in Chapter 9).
(e) Using matrix multiplication, show that Mc Md = Mc . As a re-
sult, for any polyhedron P with f -vector f , we have
Mc Md f = Mc f .
95
96 CHAPTER 6. ANGLES AND ARCHIMEDEANS
120◦
60◦
120◦
Dt3,3
Figure 6.1
—Edge angles.
Finding the edge angles of the partially truncated Platonic solids is some-
what more involved than finding the dihedral angles. However, because we
know that all edges of such a polyhedron have the same length, and, due
to the symmetry of the truncation process, that all vertices of such a poly-
hedron lie on the same sphere, it follows that all edges of the partially
truncated Platonic solid subtend the same angle at the center of the sphere.
Thus, there is just a single edge angle to find.
One procedure is illustrated here by finding the edge angle of the trun-
cated tetrahedron. Consider Figure 6.2, where O is the center of a tetra-
hedron, ps is an edge of that tetrahedron, and hence qr is an edge of the
truncated tetrahedron (recall that [qr] = 13 [ps]). E3,3 denotes, as before,
the edge angle of the tetrahedron, and Et3,3 denotes the edge angle of the
truncated tetrahedron. The midpoint of qr is denoted by m.
6.1. PARTIALLY TRUNCATED PLATONIC SOLIDS 97
q m r
p s
Et3,3
E3,3
O
center of truncated tetrahedron
Figure 6.2
Now since
[mr] [qr] 1
= = ,
[ms] [ps] 3
it follows upon considering right triangles ∆Omr and ∆Oms that tan 12 Et3,3 =
1 1 21 t 1 21
3 tan 2 E3,3 , and hence tan 2 E3,3 = 9 tan 2 E3,3 . Now
tan2 12 θ = sec2 12 θ − 1 = 1
cos2 12 θ
−1= 1
1+cos θ −1= 2
1+cos θ − 1,
2
so that
2 1 2
t −1= −1 .
1 + cos E3,3 9 1 + cos E3,3
Solving algebraically for cos Et3,3 yields
5 cos E3,3 + 4
cos Et3,3 = . (6.2)
4 cos E3,3 + 5
Since cos E3,3 = − 31 , we calculate that cos Et3,3 = 11
7
. Of course, since partial
truncation of the octahedron and icosahedron also results in the trisection
of the edges of the original polyhedron, the same formula is valid if the
subscript “3,3” is replaced by “3,4” or “3,5”; i.e.,
5 cos E3,4 + 4
cos Et3,4 = .
4 cos E3,4 + 5
Now the edge angles of the truncated cube and the truncated dodecahe-
dron must be calculated differently, as we recall that the truncation process
does not divide the edges of the original polyhedron into three equal parts
(see Figure 1.7(a) and Figure 5.5(a)). For example, if Figure 6.3 represented
a diagram for calculating the edge angle of a truncated cube, we would have
√
[qr] 2 √
=√ = 2 − 1.
[ps] 2+2
98 CHAPTER 6. ANGLES AND ARCHIMEDEANS
q r
p s
Etp,q
Ep,q
O
center of truncated polyhedron
Figure 6.3
The derivation of this formula is identical to that of (6.2) except that “λ”
is used in place of “ 13 .” Details are left to the Exercises. Values for the
cosines of edge angles of the partially truncated Platonic solids are included
in Table 6.2 (located after the Exercises).
—Dihedral angles.
Let us consider the truncated cube for a moment. We see that moving
the triangular faces radially inward (i.e., furthering the truncation process)
until the vertices meet yields a cuboctahedron. However, throughout this
process, the dihedral angle between the triangular and octagonal faces does
6.2. COMPLETELY TRUNCATED PLATONIC SOLIDS 99
not change. As a result, the dihedral angle of the cuboctahedron is the same
as Dt4,3 . Viewing the cuboctahedron from the perspective of completing the
truncation process on the truncated octahedron as well, we see that
There is an alternative way to calculate DT3,4 and DT3,5 ; while more com-
putationally intense, it is of independent geometrical interest. We illustrate
by using this method to find DT3,5 .
To begin, we cut an icosidodecahedron in half as in Figure 6.4. This
exposes an equatorial decagon, with exactly one triangle, one pentagon, and
one decagon meeting at each vertex of the decagon. By centering a sphere
at one such vertex, we create a spherical triangle with edge angles 60◦ , 108◦ ,
and 144◦ – with DT3,5 being opposite the edge with measure 144◦ . Applying
(3.4) gives
cos 144◦ = cos 60◦ cos 108◦ + sin 60◦ sin 108◦ cos DT3,5 ,
Figure 6.4
—Edge angles.
Of course, consideration of these equatorial decagons and hexagons of the
icosidodecahedron and cuboctahedron immediately yields that ET3,5 = 36◦
and ET3,4 = 60◦ . We shall use spherical trigonometry to verify these results.
We illustrate by finding ET3,4 . In Figure 6.5(a), pr is an edge of a cuboc-
tahedron inscribed in a cube, q is a vertex of this cube, and O is the center
of the cube. The spherical version of ∆pqr is shown in Figure 6.5(b). Since
three squares meet at q, the dihedral angle at q must be 120◦ ; since p and
r are midpoints of edges of the cube, sides pq and qr must have measure
1
2 E4,3 . Therefore, we may apply (3.4) to yield
and hence ET3,4 = 60◦ . The same result would have been obtained were we
to have considered the cuboctahedron as being inscribed in an octahedron
rather than a cube.
q
1
2 E4,3
120◦
1
p 2 E4,3
q p
ET3,4 r
O
r
O
(a) (b)
Figure 6.5
120◦ DE
3,4
90◦
135◦
DT3,4
DE
4,3
Figure 6.6
On the other hand, the other two dihedral angles have not been encoun-
tered before. The label “DE 3,4 ” in Figure 6.6 indicates that this is the dihedral
angle between the squares and the octahedral (“{3, 4}”) faces, which happen
to be regular hexagons. The superscript “E” represents the fact that the
rhombitruncated cuboctahedron may be formed by a process of complete
expansion. In other words, the triangles of the octahedron are expanded
radially outward, then truncated to form hexagons. Octagons and squares
are then situated so that they correspond to the vertices and edges, re-
spectively, of the original octahedron. This process results in a complete
expansion of the edges of the original octahedron in that although all twelve
edges are evident in the rhombitruncated cuboctahedron, none is adjacent
to any other.
Likewise, DE4,3 is the dihedral angle between the square faces and the
cubical (octagonal) faces. A moment’s thought reveals that expanding the
cube in a manner analogous to that just described for the octahedron also
produces the rhombitruncated cuboctahedron.
Calculation of DE E
3,4 and D4,3 is a straightforward application of (3.4) to
the spherical triangle in Figure 6.6; details are left to the reader. We will
occasionally adopt a notation such as “DE 4,3 [4-8]” to reinforce the idea that
102 CHAPTER 6. ANGLES AND ARCHIMEDEANS
DE4,3 is the dihedral angle between the square and octagonal (4-sided and
8-sided) faces of the rhombitruncated cuboctahedron. The edge angle EE 3,4
is included in Table 6.2 for completeness; its derivation is left as an Exercise.
A similar methodology may be applied to the icosidodecahedron to ob-
tain the edge angle EE 3,5 of the rhombitruncated icosidodecahedron, the di-
E
hedral angle D3,5 [4-6] between the square and hexagonal faces of this poly-
hedron, and the dihedral angle DE 5,3 [4-10] between the square and decagonal
faces. Analogous to the cuboctahedral case, the dihedral angle between the
hexagonal and decagonal faces is just DT3,5 .
Figure 6.7
6.3. TRUNCATION DEFORMATION 103
t
D5,3
q α r
De5,3
p s
(a) (b)
Figure 6.8
104 CHAPTER 6. ANGLES AND ARCHIMEDEANS
6.4 Exercises
1. Derive (6.1) according to the following outline.
Consider the Platonic solid {p, q}. A vertex of this polyhedron is
truncated as in Figure 6.9. Sinceα is an angle of a polygon with q
edges, we know that α = π 1 − 2q .
α θ
Dtp,q
θ
Figure 6.9
(a) Show that θ = π 1 − p1 .
(b) Apply (3.4) to the spherical triangle formed by centering a sphere
at a vertex of the truncated polyhedron (the edge angles of this
spherical triangle are θ, θ, and α), yielding
2π π π 2π
− 1 + cos cos = sin sin cos Dtp,q .
q p p q
5. Use the following outline to find the edge angle of the rhombitruncated
cuboctahedron. Assume throughout that an edge of this polyhedron
has length 1.
r
s u 1
s
1 r O
O r
t
v
1
s
EE
3,4
(c) (e)
Figure 6.10
7. Explain all the duplications in the “cos D” and “D” columns in Table
6.1 which are not discussed in the text.
6.4. EXERCISES 107
Table 6.1
108 CHAPTER 6. ANGLES AND ARCHIMEDEANS
Table 6.2
Chapter 7
Geodesic Structures, II
a c
A
b
Figure 7.1
109
110 CHAPTER 7. GEODESIC STRUCTURES, II
so that
cos c = cos a cos b. (7.1)
We may apply (3.6) in two different ways. The first results in
cos 90◦ = − cos A cos B + sin A sin B cos c,
which after a little algebra becomes
tan A tan B cos c = 1. (7.2)
The second yields
cos A = − cos B cos 90◦ + sin B sin 90◦ cos a,
which together with the companion relationship obtained by reversing the
roles of A and B, results in
cos A = sin B cos a, cos B = sin A cos b. (7.3)
Now we turn our attention to (4.5), so that
sin a sin b sin c
= = ,
sin A sin B sin 90◦
which gives the companion relationships
sin a = sin A sin c, sin b = sin B sin c. (7.4)
Multiplying each equation in (7.3) by sin c and substituting from (7.4)
yields
cos a sin b = cos A sin c, cos b sin a = cos B sin c. (7.5)
Dividing the equations in (7.5) by (7.1), as in
cos a sin b cos A sin c
= ,
cos a cos b cos c
results in
tan b = cos A tan c, tan a = cos B tan c. (7.6)
Finally, we divide the equations in (7.4) by those in (7.5), as in
sin a sin A sin c
= ,
cos a sin b cos A sin c
to obtain the companion relationships
sin b tan A = tan a, sin a tan B = tan b. (7.7)
None of these is especially difficult to derive, but it will prove useful to
have them handy for future reference.
7.2. VERTEX ANGLES 111
A 2π
p
E
χ
Figure 7.2
2
cos E − cos 2π
p
cos χ = . (7.8)
1 − cos 2π
p
We also have
sin E2
sin χ = ,
sin πp
112 CHAPTER 7. GEODESIC STRUCTURES, II
2π
which follows immediately from bisecting the angle p in Figure 7.3(a) and
applying (7.4.).
2π
p
χ χ
E
(a) (b)
Figure 7.3
√
Thus, for our dodecahedron, where cos E5,3 = 5
and cos 2π ◦
√ 3 p = cos 72 =
5−1
4 , we find that
√ √ √ √
5 5−1
2 3 − 5+3 5+2 5
cos χ = √ 4 = √ = .
1− 5−1 3(5 − 5) 15
4
Although in a different algebraic form, this result does agree with that of
§4.2, so that we find in this case that χ ≈ 37.4◦ .
It is evident that this procedure may be applied to each face of each
Archimedean solid, for all that is needed are the values for p and E; p is
simply the number of sides on the face in question and E is the edge angle
of the Archimedean solid, whose cosine is available in Table 6.1. A complete
set of data is given in Table 7.1 (located after the Exercises). These data
may be used to construct geodesic versions of the Archimedean solids as
was done for the dodecahedron in §4.2; one example is the rhombitruncated
cuboctahedron shown in Figure 7.3(b). Since the area of the squares is so
much smaller than the areas of the octagons and hexagons, the squares on
the geodesic model are sometimes left intact to make a more aesthetically
pleasing sphere.
We remark that for building models, only numerical data are needed.
As our approach is to also find algebraic expressions for the various angles,
somewhat more effort is involved.
7.3. MIDEDGE ANGLES 113
Figure 7.4
π
p
ψ χ
m
E
2
(a) (b)
Figure 7.5
114 CHAPTER 7. GEODESIC STRUCTURES, II
E 1 + cos E
cos2 χ = cos2 ψ cos2 = cos2 ψ ,
2 2
so that
2 cos2 χ
cos2 ψ = . (7.9)
1 + cos E
Thus, for our dodecahedral example, we would have
√ √ √
5+2 5
2 2· 15 10 + 4 5 5+ 5
cos ψ = √ = √ = ,
1 + 35 15 + 5 5 10
from which we calculate ψ ≈ 31.7◦ . We may use the data in Table 7.1
for χ to calculate cos2 ψ for all of the Archimedean solids; the results are
also collected in Table 7.1. The variables “χ” and “ψ” were chosen in this
section and the last so as to be consistent with Table 3 on p. 53 of Spherical
Models. Wenninger’s “φ” in Table 3 is simply our E/2. This notation is
also consistent with Coxeter’s use of “χ,” “ψ,” and “φ” in Regular Polytopes
(see §7.9, where equation 2.52 is referenced) in describing a characteristic
simplex on a sphere.
Table 7.2
Recall that we may use (7.8) to find χ for any Archimedean solid. In
restricting our attention for the moment to Platonic solids, we may substi-
tute the last expression for cos Ep,q in (3.7) into (7.8); with a little algebra,
we obtain
1 + cos 2π
p 1 + cos 2π
q
cos2 χp,q = .
1 − cos 2π
p 1 − cos 2π
q
π
4
ψ4,3 χ4,3
π π
2 3
m b
1
2 E4,3
Figure 7.6
116 CHAPTER 7. GEODESIC STRUCTURES, II
s u
t
(b)
(a)
(c) (d)
Figure 7.7
Consider two triangles as shown in Figure 7.6, one the reflection of the
other, as they would appear drawn on a face of the cube (see Figure 7.7(a)).
We see that 8×6 = 48 such triangles would tile the cube, as shown in Figure
7.7(b). If these triangles are projected simultaneously onto the circumsphere
of the cube, we obtain a tiling of a sphere by 48 congruent right spherical
triangles, as shown in Figure 7.7(d). (The term “congruent” as used here
7.4. SCHWARZ TRIANGLES 117
includes the possibility that two triangles may have all angles the same, but
be mirror images of each other, as would triangles ∆rst and ∆rut in Figure
7.7(a) when projected onto the circumsphere.)
It is apparent that this procedure may be carried out with any of the
Platonic solids. But one case is of special interest to us at the moment: the
octahedron. In view of (7.10), we see that the result is the same tiling of
the sphere obtained for the cube! This result is not altogether surprising,
as we hope to demonstrate.
Consider the rhombic dodecahedron (see §9.1 and Figure 9.1), and di-
vide each rhombic face into four congruent right triangles as in Figure 7.8.
Now project these 48 triangles onto the circumsphere of the rhombic dodec-
ahedron (although note that only the six outermost vertices would be on
the circumsphere).
3 3
Figure 7.8
This analysis may be made for the other dual pairs of Platonic solids
(where the tetrahedron is considered to be dual to itself). In general, the
number of triangles in the spherical tiling is four times the number of edges
on the Platonic solid. The component triangles of such tilings are often
called Schwarz triangles, and are discussed much more fully in Uniform
Polyhedra by Coxeter, Longuet-Higgins, and Miller.
where the three dihedral angles are written, and the vertical slash following
π/3 indicates that it is the vertex of the dihedral angle π/3 which currently
holds our interest.
But Wythoff did not use the radian measure of the angles in notating
this phenomenon; he used the notation p | q r for an analogous polyhedron
whose generating triangle had dihedral angles π/p, π/q, and π/r, and whose
vertices were those at the dihedral angles π/p. Thus, Wythoff would have
described the cube with the expression
3 | 2 4 = 3 | 4 2.
It is perhaps clear, given the discussion in the previous section, that the
octahedron is given by
4 | 2 3 = 4 | 3 2.
7.5. WYTHOFF SYMBOLS 119
2 | 3 4 = 2 | 4 3.
Examination of Figure 7.8 suggests that points such as m (see Figure 7.6) are
projections of the centers of the rhombic faces of a rhombic dodecahedron;
thus, we suspect (see Figure 9.1), and are indeed correct in our suspicion,
that 2 | 3 4 is the cuboctahedron.
There is more to our story, however. Let us now bisect the dihedral angle
at m (see Figure 7.9) so that the point c on side ab is such that ∠amc and
∠cmb both have measure π/4. When c on side ab is determined in this way,
it may be shown that points such as c on our tiling of the sphere comprise
the vertices of yet another Archimedean solid, the rhombicuboctahedron. To
indicate that c is opposite the dihedral angle π/2 (or “2” in our symbology),
we describe the rhombicuboctahedron with the symbol
3 4 | 2 = 4 3 | 2.
2 3 | 4 = 3 2 | 4,
2 4 | 3 = 4 2 | 3.
π
4
c
π
4 π
π/4 3
m b
Figure 7.9
120 CHAPTER 7. GEODESIC STRUCTURES, II
Yes, the story continues. There is exactly one point d in the interior of
spherical ∆amb, called its incenter, such that the three arcs drawn from d
perpendicular to the three sides of the spherical triangle all have the same
measure. There is one such point for each of the 48 triangles tiling our
sphere; these 48 points are vertices of a rhombitruncated cuboctahedron,
notated by 2 3 4 | (and the five other permutations of this symbol).
And finally, we denote by | 2 3 4 (and its other five permutations) the
snub figure corresponding to spherical ∆amb, the snub cube.
Regrettably, this concludes our brief introduction. Again, the reader is
invited to peruse the reference cited above for a decidedly more thorough
discussion of the description of polyhedra using Wythoff symbols.
7.6 Exercises
1. Using the notations of §7.1, show that for any right spherical triangle
(see Figure 7.1), we have
sin a sin b tan a tan b
= = sin c tan c.
cos A cos B sin A sin B
2. In a right isosceles spherical triangle, a = b and A = B in Figure 7.1.
In this case, show that
sin 2a sin 2A
= .
sin a sin2A
3. Bisect the isosceles spherical triangles in Figures 3.11 and 3.12 to ob-
tain right triangles. Use these triangles together with (3.10) and re-
lationships from §7.1 as appropriate to derive first Exercise 6, then
Exercise 5, of Chapter 3.
4. Consider the quadrilateral as shown in Figure 7.10. Note that B > 90◦ ,
since the sum of the angles in a spherical quadrilateral is greater than
360◦ .
(a) Show that
tan b = sin a and cos B = − sin2 a.
(b) Let d1 and d2 be the measures of the two diagonals of this quadri-
lateral in the case a = 30◦ . Show that
√ √
27 + 14
cos(d2 − d1 ) = √ .
80
7.6. EXERCISES 121
b
B
a b
Figure 7.10
b b
a a a a
x d f
(a) (b)
Figure 7.11
(a) Beginning with (7.6), ultimately show that tan d = sec a tan b.
Use this to conclude that
1 cos2 a cos2 b
cos2 d = = .
1 + sec2 a tan2 b 1 − sin2 a cos2 b
cos d
cos f = .
cos b
122 CHAPTER 7. GEODESIC STRUCTURES, II
sin E2
sin χ = .
sin πp
(b) By using the results of this Exercise with (7.11), show that
1
cos Ep,q cos ψp,q = cos χp,q .
2
(c) Deduce the result in (b) by applying results from §7.1 to the
spherical right triangle in Figure 7.5. Note, as a result, that the
result in (b) (omitting the subscripts) is valid for any Archimedean
solid, not merely Platonic solids.
Table 7.1
124 CHAPTER 7. GEODESIC STRUCTURES, II
Both antiprisms and the snub polyhedra have more than three faces meeting
at each vertex. None of the techniques used so far allows us to easily calculate
their edge and dihedral angles, so we need a fresh approach. If one imagines
a sphere centered at the vertex of, say, a square antiprism (peek ahead to
Figure 8.3), it is evident that a spherical isosceles trapezoid is formed. So
we begin by deriving some useful formulas for these spherical polygons.
α α
cos δ = cos β cos + sin β sin cos D1 (8.1)
2 2
and
α α
cos β = cos δ cos + sin δ sin cos E. (8.2)
2 2
125
126 CHAPTER 8. ANTIPRISMS AND SNUB POLYHEDRA
γ γ
q
D2 D2 r
β β β δ δ β
F
D1 D1 D1 E E D1
p
α α s α
2 2
(a) (b)
Figure 8.1
We may apply (7.4) to spherical ∆qrs, which yields, since E and F are
complementary,
sin γ2 sin γ2
sin δ = = .
sin F cos E
This relationship may be substituted into (8.2) which, after slightly rear-
ranging terms, yields
α α γ
cos δ cos = cos β − sin sin .
2 2 2
Now we may multiply both sides of (8.1) by cos α2 ; equating the expression
for cos δ cos α2 thereby obtained with that of the previous equation yields
α α α α γ
cos β cos2 + sin β sin cos cos D1 = cos β − sin sin .
2 2 2 2 2
Replacing cos2 α2 by 1 − sin2 α2 in the previous equation, simplifying, and
solving for D1 yields
cos β sin α2 − sin γ2
cos D1 = . (8.3)
sin β cos α2
Of course a parallel argument may be used to obtain cos D2 , resulting in
cos β sin γ2 − sin α2
cos D2 = . (8.4)
sin β cos γ2
It will also be useful later on to find similar formulae in the case that
the diagonals of the trapezoid, rather than the sides, are given (as in Figure
8.2). Note that the symmetry of the figure implies that w is the midpoint
of arc qs and u is the midpoint of arc pt so that both ∠swv and ∠puv are
right angles. Thus, u, v and w all lie on the same great circle which bisects
Dx .
8.1. SPHERICAL ISOSCELES TRAPEZOIDS 127
γ
s w q
D2
β−δ β−δ
v
Dx
δ δ
p u D1
α t
Figure 8.2
sin γ2
= sin(β − δ) = sin β cos δ − cos β sin δ, (8.6)
sin 12 Dx
where β is the measure of the diagonals pq and st. We may solve (8.5) for
sin 12 Dx , substitute into (8.6), and rearrange terms, obtaining
Solving this equation for cos D1 and using the fact that
1 − cos α α sin α2
= tan = ,
sin α 2 cos α2
we have
(1 − cos α) cos δ sin α2 cos δ
cos D1 = = .
sin α sin δ cos α2 sin δ
Finally, we substitute from (8.7) into this equation, yielding
Using the relationship 1 + cos α = 2 cos2 α2 , we may substitute (8.8) into this
equation, which results in
Expanding this expression for cos Dx and simplifying using the relationship
2 sin2 2θ = 1 − cos θ gives
π
3
Dap
3 Dap
3
π π
3 3
Dap
p Dap
p
ϕ = π 1 − p2
Figure 8.3
and
cos π3 sin π6 − sin ϕ2 1 ϕ
cos Dap
3 = = 1 − 4 sin . (8.12)
sin π3 cos π6 3 2
We may rewrite (8.11) and (8.12) in terms of p rather than ϕ, the result
being
ap 1 π π ap 1 π
cos Dp = √ cot − csc , cos D3 = 1 − 4 cos . (8.13)
3 p p 3 p
Figure 8.4(b). Clearly, [pq] = [qr], ∠qpO and ∠qrO are both right angles,
and ∠pqr has the same measure as Dap 3 . If γ is the measure of angles ∠pOq
and ∠rOq, we see that γ = 21 (π − Dap π 1 ap
3 ) = 2 − 2 D3 , so that
π 1 ap 1
cos γ = cos − D3 = sin Dap . (8.14)
2 2 2 3
t q t̄
p q Dap
3 C
p r
r 1 ap
2E
γ γ
π
p̄ 3 q̄
O γ
O
(a) (b) (c)
Figure 8.4
2
cos Eap = − 1. (8.16)
3 − 2 cos πp
π π π
t π
3 3 3 3
D2 D2
u r
D2 D2 α
π π π π
3 3 3 β 3
D1 D1 p D1 q
ϕ ϕ
(a) (b)
Figure 8.5
π π 1 √
cos β = cos cos ϕ + sin sin ϕ cos D1 = (cos ϕ + 3 sin ϕ cos D1 ). (8.21)
3 3 2
Considering the spherical trapezoid putr, we see that D2 and β have the
same relationship as Dap
3 and ϕ of Figure 8.3, so that (8.12) becomes
1 β
cos D2 = 1 − 4 sin . (8.22)
3 2
We may substitute into sin2 β2 = 12 (1−cos β) from (8.21) and (8.22), resulting
in √
9 cos2 D2 − 6 cos D2 − 7 + 4 cos ϕ + 4 3 sin ϕ cos D1 = 0. (8.23)
Solving for cos D1 from (8.18), we have
1 ϕ ϕ
cos D1 = √ sec λ + sin . (8.24)
3 2 2
Substituting for cos D1 from (8.24) and cos D2 from (8.20) into (8.23) yields,
since λ 6= 0,
ϕ
λ3 − 2λ + 2 sin = 0. (8.25)
2
Once λ is found by solving this cubic equation, cos D1 may be found using
(8.24).
This cubic also yields a cubic equation which may be solved to obtain
cos D2 . From (8.25), we see that
ϕ 2
(λ3 − 2λ)2 = −2 sin ,
2
so that
ϕ
λ6 − 4λ4 + 4λ2 − 4 sin2 = 0.
2
8.5. EDGE ANGLES OF THE SNUB POLYHEDRA 133
Since all powers of λ are even, we may substitute for λ2 from (8.20), yielding
27 cos3 D2 − 9 cos2 D2 − 15 cos D2 + 13 − 16 cos ϕ = 0. (8.26)
Because of the relationship (8.20), it is evident that only one of (8.25) and
(8.26) needs to be solved in order to find cos D1 and cos D2 .
We use a notation for the dihedral angles of the snub polyhedra similar
to that developed in §8.2 for the antiprisms. A superscript “s” is used to
denote a snub figure, so that Ds4,3 [4-3] is the dihedral angle between the
square and triangular faces of the snub cube (the subscript denoting the six
square faces in the facial planes of the cube), and Ds3,4 [3-3] is the dihedral
angle between the triangular faces (the subscript relating to the triangular
faces, eight of which lie in the facial planes of an octahedron). Ds5,3 [5-3]
and Ds3,5 [3-3] have similar connotations as far as the snub dodecahedron is
concerned.
We may use Cardan’s formula (which is summarized immediately pre-
ceding the Exercises) to solve (8.26); for the snub cube, we have ϕ = 90◦ ,
from which we calculate
√ √
q q
s 2 3 3 1
cos D3,4 = −19 + 297 + −19 − 297 + ≈ −0.892858.
9 2
3
For the snub dodecahedron, we have ϕ = 10 , and hence
1 √ √
s 3 3 1
cos D3,5 = u+v+ u−v+ ≈ −0.962101,
3 3
where
√ √
r
2 2 2
u = − (49 + 27 5), v = (93 + 49 5).
27 3 3
The other dihedral angles may then be found as described above. Data are
summarized in Table 8.1.
Thus, these edge angles may easily be found numerically once the dihedral
angles are obtained. We proceed with a more algebraic approach.
For brevity, we write (8.27) as
4
cos E = − 1,
5 + 3 cos D
and hence we will omit subscripts and superscripts for now. This relationship
may be solved for cos D, yielding
1 5 cos E + 1
cos D = − .
3 cos E + 1
Recall, however, that cos D satisfies the cubic equation (8.26) since D2 in
that equation represents the dihedral angle between triangular faces. Thus,
we may substitute the previous expression for cos D into (8.26). After mul-
tiplying this equation by −(cos E + 1)3 , expanding, grouping like terms, and
dividing by 16, we obtain
These may be solved using Cardan’s formula, if desired. The details are left
as an Exercise.
8.6. SOLVING A CUBIC 135
Table 8.1
Also, put
β 2 α3
∆= + .
4 27
Now define
β √ β √
r r
3 3
γ= −
+ ∆, δ = − − ∆.
2 2
Then the three roots of the cubic equation are
a γ + δ γ − δ√ a γ + δ γ − δ√ a
γ+δ− , − + −3 − , − − −3 − .
3 2 2 3 2 2 3
Moreover, if a, b, and c are all real numbers, then if ∆ > 0, there is one real
root and two conjugate complex roots. If ∆ = 0, there are three real roots
with at least two equal. Finally, if ∆ < 0, there are three distinct real roots.
We illustrate with the example of finding cos Ds3,4 as given in §8.4. So
with x = cos Ds3,4 and ϕ = 90◦ , we see that (8.26) becomes
8.7 Exercises
1. Derive (8.25) as suggested in the text.
λ2 = z + 1.
5. By using Cardan’s formula to solve the cubic equation (8.25), and then
using (8.24), show that
√ √
q q
s 1 3 3
cos D4,3 = √ −54 + 6 33 + −54 − 6 33 + 3 ≈ −0.798461.
3 3
and
2 √ √ τ
cos Ds5,3 = p 3
u+v+ 3u−v+ ≈ −0.890451,
3(3 − τ ) 2
where r
τ 1 27τ − 5
u=− , v= .
2 6 3
6. With E and D1 representing either Es3,4 and Ds4,3 , or Es3,5 and Ds5,3 ,
respectively, use the following outline to show that
√ ϕ
3 cos E cos D1 cot = −1. (8.31)
2
138 CHAPTER 8. ANTIPRISMS AND SNUB POLYHEDRA
(b) Solve this relationship for 9 cos2 D2 − 6 cos D2 and substitute the
result into (8.23).
(c) Finally, solve (8.27) for cos D2 (which may be either cos Ds3,4 or
cos Ds3,5 ) and substitute into the result from (b). Using trigono-
metric identities as appropriate, simplify to obtain (8.31).
7. Use Cardan’s formula to solve (8.29), yielding
√ √
q q
s 1 3 3
cos E3,4 = 566 + 42 33 + 566 − 42 33 − 1 .
21
8. The results of this chapter may all be used to find the dihedral and
edge angles of the icosahedron by choosing ϕ = 60◦ .
(a) In this case, show that (8.25) becomes
1
(λ + τ )(λ − 1) λ − = 0,
τ
and use the
√
negative root λ = −τ with (8.24) to show that
5
cos D1 = − 3 .
(b) Show that (8.26) may be written
(3 cos D2 − 1)(9 cos2 D2 − 5) = 0,
√
5
which also yields cos D2 = − 3 , as expected, as its only negative
root.
(c) Show that (8.28) may be written
(5 cos2 E − 1)(3 cos E + 1) = 0,
which yields cos E = √1 as its only postive root.
5
9. Note that for the derivation of (8.26), the assumption that ϕ > 0
was not needed. Show that when ϕ = 0, cos D2 = − 13 is a root of
(8.26), and when ϕ = −60◦ , cos D2 = 31 is a root of (8.26). Note
that these correspond to the dihedral angles of the octahedron and
the tetrahedron. Why do you think that this is the case? Note also
that (8.27) remains valid for the octahedron and tetrahedron, with
superscripts and subscripts appropriately modified.
8.7. EXERCISES 139
π 2π
10. Note that when ϕ = 2 · 3 = 3 , then cos D2 = −1 is a root of (8.26).
Explain.
140 CHAPTER 8. ANTIPRISMS AND SNUB POLYHEDRA
Chapter 9
Duality
141
142 CHAPTER 9. DUALITY
(a) (b)
Figure 9.1
Figure 9.2
2. Upon superposing the dual on its parent (seen in Figure 9.1(b)), one
sees that each edge of the dual is perpendicular to an edge of the
parent. Moreover, each edge of the Archimedean dual bisects the cor-
responding edge of its parent. The converse, however, is not true –
the edges of the parent do not necessarily bisect those of the dual. For
example, one may show that the points where the edges of the cuboc-
tahedron meet those of the rhombic dodecahedron divide the edges
of the rhombic dodecahedron in the ratio 1 : 2 (see Figure 9.3). In
addition, it is these points (and not the midpoints, as was the case
for the cuboctahedron) on the edges of the dual which, when joined
to the center of the dual, form segments perpendicular to the edges of
the dual.
3. As a result of the process for constructing the dual, we see that the
number of edges on an Archimedean solid is the same as the number
9.2. FURTHER EXAMPLES 143
Figure 9.3
1. The Platonic solids are the only solids whose duals are among the
group as well, and
9.2. FURTHER EXAMPLES 145
2. The Platonic solids are the only solids among the group all of whose
dual edges are bisected by the parent solid.
For completeness, we include the duals of all the Platonic and Archimedean
solids below. Names for these polyhedra are given in Table 9.1. A few ex-
planatory remarks are in order, however.
1. A name such as “pentakis dodecahedron” indicates that this polyhe-
dron may be formed by placing a pentagonal (hence “pentakis”) pyra-
mid on each face of a dodecahedron (here, the triangular faces of the
pyramids are not necessarily equilateral).
2. Adjectives such as “rhombic,” “pentagonal,” and “trapezial” indicate
the shape of the faces of the polyhedron. Here, “pentagonal” does
not necessarily imply regular pentagons. A trapezium is a quadrilat-
eral with no two sides parallel, hence the adjective “trapezial.” Many
authors use “trapezoidal” rather than “trapezial;” it happens, accord-
ing to my dictionary, that the British interchange the meanings of
“trapezium” and “trapezoid.”
3. “Icositetra” is a prefix indicating “24,” “triaconta” denotes “30,” and
“hexeconta” means “60.” These prefixes refer to the number of faces
on the polyhedron.
Tetrahedron Tetrahedron
Cube Octahedron
Octahedron Cube
Icosahedron Dodecahedron
Dodecahedron Icosahedron
Truncated tetrahedron Triakis tetrahedron
Truncated octahedron Tetrakis hexahedron
Truncated icosahedron Pentakis dodecahedron
Truncated cube Triakis octahedron
Table 9.1
146 CHAPTER 9. DUALITY
q are centers, it follows that ∠qsp has the same measure as the dihedral
angle Dt3,3 of the truncated tetrahedron. Again, due to the symmetry of the
truncated tetrahedron, the segment Op is perpendicular to the hexagonal
face of which p is the center. As a result, we see that segment Op and ps in
Figure 9.7(b) are perpendicular. Similarly, Oq and qs are also perpendicular.
Finally, since rt is an edge of the triakis tetrahedron (whose center is O),
we see that ∠rOt is an edge angle of the triakis tetrahedron, which shall be
denoted by Ēt3,3 .
t
u
r s
q t
Dt3,3
(a)
Ēt3,3 p
O
(b)
Figure 9.7
It is evident from examining the quadrilateral Oqsp that Dt3,3 and Ēt3,3
are supplementary, so that Dt3,3 + Ēt3,3 = π, and hence cos Ēt3,3 = − cos Dt3,3 .
This phenomenon is not peculiar to this pair of duals; in general, we see
that
This relationship suggests the notation Ēt3,3 for the edge angle of the
triakis tetrahedron subtended by the edge rt. Thus, if D∗p,q represents a di-
hedral angle of an Archimedean solid, then Ē∗p,q represents the corresponding
148 CHAPTER 9. DUALITY
edge angle of its dual, and we have cos Ē∗p,q = − cos D∗p,q . We see, then, that
Ē3,3 is the edge angle of the edge tu of the triakis tetrahedron (see Figure
9.7(a)) which intersects an edge of the truncated tetrahedron at which the
dihedral angle is D3,3 , the same as that of the regular tetrahedron. It fol-
lows, therefore, that tu would also be an edge of the tetrahedron dual to
the tetrahedron which has been truncated in Figure 9.7(a). This gives the
triakis tetrahedron its name, for it may be imagined as a tetrahedron with
squat triangular (“triakis”) pyramids affixed to its faces.
We now proceed to investigate Figure 9.7(b) in more detail. Upon
consideration of the similar right triangles ∆Oqs and ∆Osr, we see that
[Oq]/[Os] = [Os]/[Or]; likewise, we note that the similarity of right tri-
angles ∆Ops and ∆Ost implies that [Op]/[Os] = [Os]/[Ot]. These two
relationships together imply that
Such relationships can be written for any Archimedean solid and its dual;
they can be imagined as a result of creating Archimedean duals by the
process of polar reciprocation, a topic which will not be addressed here
(but see Wenninger’s Dual Models, pp. 1-5).
s, w, and x all lie on a circle which is the intersection of the sphere and a
plane which is perpendicular to Ov (intersecting Ov at a point we call z)
and which passes through s, w, and x. This circle, whose center is z, is
shown in Figure 9.8(b). Because of the perpendicularity of the dual edges
and the parent edges, we see that the sides of the face rtu of the triakis
tetrahedron are in fact tangent to this circle at s, x, and w. Note that the
lengths of ws, sx, and xw may be easily determined as s, w, and x are
the midpoints of edges of the truncated tetrahedron (a flattened vertex of
which is shown in Figure 9.8(c)), where we assume for ease of calculation
that the edge length is 2. The triangle ∆swx is called a vertex figure of
the truncated tetrahedron.
r
w v s
t
x
u
1
w s
√
w 3
(a) x
r
w s (c)
1
z √
h 3
u t
x
(b)
Figure 9.8
To determine the shape of ∆rtu, we see that [ws] = 1 (ws being as long
as half an edge of the truncated tetrahedron), and hence from Figure 9.8(c)
150 CHAPTER 9. DUALITY
√
that [wx] = [xs] = 3. It may be shown that given a triangle whose sides
have length a, b, and c, then the radius R of the circumcircle (i.e., the circle
circumscribing the triangle) satisfies
a2 b2 c2
R2 = .
(a + b + c)(−a + b + c)(a − b + c)(a + b − c)
√ 2
√ a = b = 3 and c = 1, we find that R = 9/11 and hence R = [zx] =
With
3/ 11. It follows (see Figure 9.8(b)) that
√ √
[hx] 3/2 11
cos(∠zux) = sin(∠hzx) = = √ = √ ,
R 3/ 11 2 3
and hence
11 5
cos(∠rut) = 2 cos2 (∠zux) − 1 = 2 · −1= .
12 6
One may similarly conclude that cos(∠urt) = −7/18.
We may summarize our observations succinctly as follows:
√
2
Figure 9.9
9.5. DIHEDRAL ANGLES 151
r x
v v0
w v s z
D̄t3,3 z0
t
x
u Et3,3
v0
r0 O
(a) (b)
Figure 9.10
similarly argues that ∠Oz 0 x must likewise be a right angle. Since a tangent
to a circle is perpendicular to a radius drawn to the point of tangency, it
follows that both zx and z 0 x are perpendicular to tu (as in Figure 9.10(a)),
and therefore ∠zxz 0 (see Figure 9.10(b)) must have the same measure as the
dihedral angle of the triakis tetrahedron, which we denote by D̄t3,3 . Since O
is the center of the truncated tetrahedron as well, we see that ∠vOv 0 has
measure Et3,3 , the edge angle of the truncated tetrahedron.
These considerations lead us, upon examining the quadrilateral z 0 Ozx in
Figure 9.10(b), to the conclusion that Et3,3 and D̄t3,3 are supplementary; i.e.,
that Et3,3 + D̄t3,3 = π. But had we considered the dihedral angle at the edge
rt instead, we would nonetheless have arrived at a figure similar to Figure
9.10(b), and would have inevitably concluded that this dihedral angle was
supplementary to Et3,3 . We make the following generalization:
(D3 ) The dihedral angles of an Archimedean dual all have the same
measure, all being supplementary to the edge angle of the par-
ent Archimedean solid.
We use similar notation for the other duals; therefore, if E∗p,q represents
the edge angle of some Archimedean solid, then D̄∗p,q represents the dihedral
angle of its dual, so that E∗p,q + D̄∗p,q = π and hence cos D̄∗p,q = − cos E∗p,q .
1 + cos 2π
q
cos ϕ = 2 − 1. (9.1)
1 − cos D
In our case, q = 6 since six acute angles meet at each of four vertices
of the triakis tetrahedron, and of course D = D̄t3,3 . These values may be
substituted into (9.1) (where Table 6.1 is used to find cos D̄t3,3 = − cos Et3,3 ),
which results in cos ϕ = 5/6.
If ϕ represents the obtuse angle of the face of the triakis tetrahedron
(such as ∠urt in Figure 9.8(c)), then as three of these angles meet at each of
9.6. FACES REVISITED 153
four vertices of the triakis tetrahedron, we find upon using (9.1) with q = 3
that cos ϕ = −7/18.
Values of ϕ for the remainder of the Archimedean duals may be found
in Table 9.2.
Table 9.2
154 CHAPTER 9. DUALITY
9.7 Exercises
1. With the nets provided, build the rhombic dodecahedron, the triakis
octahedron, and the trapezial icositetrahedron.
6
[cf ] = .
5
(b) Find [ce], and then by considering right triangles ∆ced and ∆cef ,
show that
[ed]
= 5.
[ef ]
9.7. EXERCISES 155
(c) Using (b), show that the ratio of the volume of the triakis tetra-
hedron relative to its base tetrahedron is 9/5.
e
c
g
Figure 9.11
(d) Show also that ∠dcf has the same measure as D3,3 .
Figure 9.12
(Note: The squares thus formed are twelve square faces of the rhom-
bicuboctahedron, so that the procedure just described shows that a
rhombicuboctahedron may be inscribed in a rhombic dodecahedron.)
x
y
A
A/2
Figure 9.13
has faces which are pentagons – at four of its vertices, three faces
meet, while p meet at the remaining vertex. Let ϕ3 denote the interior
angles at the vertices of the pentagon where three faces meet, and let
ϕp denote the interior angle at the remaining vertex.
Since the interior angles of a convex pentagon sum to 3π, it is evident
that 4ϕ3 + ϕp = 3π, so that
Substitute for cos ϕ3 and cos ϕp from (9.1) into this expression, and
thus produce an alternate derivation of (8.24).
158 CHAPTER 9. DUALITY
Chapter 10
159
160 CHAPTER 10. GEODESIC STRUCTURES, III
a
1 1
2b 2b
c c
d
d d
c
1 c
2b d d
1
d 2b
a 1 c c a
2b
q
t
Figure 10.1
cos 2s cos A
cos A = , cos 2s = . (10.1)
1 + cos 2s 1 − cos A
First, we find the measure of the arc labelled d in Figure 10.1. It is apparent
by symmetry considerations that ∠qOr has measure 60◦ . Applying (7.7),
then, to right spherical triangle ∆Orq results in sin d tan 60◦ = tan s. In
addition, cos d may be found by employing the identity sin2 d + cos2 d = 1,
so that √
1 4 cos2 s − 1
sin d = √ tan s, cos d = √ . (10.2)
3 3 cos s
10.1. A 4-FREQUENCY ICOSAHEDRON 161
cos 2s 2 cos2 s − 1
cos e = = .
cos s cos s
We may also find sin e by employing the identity sin2 e + cos2 e = 1 and some
algebra, resulting in
2 cos2 s − 1 p
cos e = , sin e = tan s 4 cos2 s − 1. (10.3)
cos s
Again, these are valid for s ≤ 60◦ .
Since a + 2d = e, we have a = e − 2d. But d and e are fully described by
(10.2) and (10.3). Thus, we may employ the usual trigonometric relation-
ships, in addition to a good deal of algebra, to obtain
2 cos2 s + 1 1 1 p
cos a = , sin a = sin e = tan s 4 cos2 s − 1. (10.4)
3 cos s 3 3
1
cos c = 1 − tan2 s. (10.5)
6
The final piece of the puzzle is arc st. In the actual construction of the
geodesic model, the exterior edges of the equilateral triangle (such as pq)
are not present. Instead, triangles such as ∆squ are used which straddle
the exterior edges. Were the exterior edges present, they would cut such
triangles exactly in half.
So let b represent the measure of arc su. We might be tempted to find
b by extending the arc labelled c to rq to create st, and then doubling our
result. However, because of the way plane triangles project onto spherical
triangles, the angle ∠stq would in fact not be 90◦ , and we would need to
separately construct triangles ∆stq and ∆utq.
162 CHAPTER 10. GEODESIC STRUCTURES, III
Substituting for a and A from (10.4) and (10.1), a bit of algebra reveals that
22 cos4 s − 5 cos2 s + 1 1
cos b = 4
=1+ tan2 s sec2 s(1 − 4 cos2 s). (10.6)
18 cos s 18
Finally, then, the subdivision of the spherical equilateral triangle in Fig-
ure 10.1 is completely described. When this figure describes a 4-frequency
Class II icosahedron, we see that
A completed model is shown below. The red icosahedral arcs are not part
of the subdivision, but are included for reference.
Figure 10.2
a
l3
1 1
2b 2b t
c c l2
d
e e
e e l1
s
1 f 1
2h g 2h
e g g e
f f
e g r g e
c g c
1 d f f d 1
2b e e 2b
1 1 1
a 2b
c e 2h
e c 2b
a
Figure 10.3
onto arcs of a circle with radius ρ in Figure 10.4, where ρ is the radius of the
sphere circumscribing the icosahedron (see (3.8)). Recall that in deriving
the formula (3.8), it was assumed that the edge of the Platonic solid in
question had length 2. Thus, we will assume that the equilateral triangle of
Figure 10.3 has edge length 2.
g g 1
d 2h
a 1
λ λ λ λ 2λ
q p l1
w ρ
ρ
Figure 10.4
w2 + (nλ)2 = ρ2 ,
10.2. A 6-FREQUENCY ICOSAHEDRON 165
which yields p
w= ρ2 − n2 λ2 . (10.8)
So now we may find g;
λ
g = arctan .
w
Having found g, we note that in Figure 10.4,
2λ
tan(g + d) = ,
w
so that
2λ λ
d = arctan − arctan .
w w
tan α−tan β
Using the relationship tan(α − β) = 1+tan α tan β (see (0.13)), we find that
wλ
tan d = .
w2 + 2λ2
Finally, the same methodology can be used to find a, yielding
wλ
tan a = .
w2 + 6λ2
These results may readily be generalized. If instead of writing the angles
sought as g, d, a, we label them α1 , α2 , α3 , ..., αn (since there must be n such
angles), we find that if 1 ≤ k ≤ n,
wλ
tan αk = . (10.9)
w2 + k(k − 1)λ2
f
e e
1 1
λ 2λ 2λ λ
s l2 t
w0
Figure 10.5
a 10.388◦ e 13.368◦
b 12.168◦ f 14.175◦
c 12.297◦ g 14.286◦
d 12.703◦ h 13.238◦
Table 10.1
Moreover, a completed icosahedron is shown below (where, as before,
the red icosahedral arcs are only for reference).
Figure 10.6
10.3 Exercises
1. Derive the expression for cos d in (10.2) as suggested in the text.
2. Derive the expression for sin e in (10.3) as suggested in the text.
3. As indicated in the text, derive the relationships in (10.4).
4. Derive (10.5) as suggested in the text.
5. Obtain (10.6) as indicated in the text.
6. Let B represent the measure of angle ∠rsO in Figure 10.1. By con-
sidering spherical triangle ∆rOs, show that
3 cos 2s
sin B = √ .
13 cos2 2s − 1
(Hint: Use (4.4) and find sin c using (10.5).)
168 CHAPTER 10. GEODESIC STRUCTURES, III
7. Using the methods in this chapter, calculate the various angles required
to build a 4-frequency octahedron and a 6-frequency octahedron.
αk = Ak − Ak−1 ,
λ λ λ λ
···
αn
α1 α2 α3
Figure 10.7
Deltahedra
169
170 CHAPTER 11. DELTAHEDRA
ν3 + ν4 + ν5 = V, (11.3)
ν3 ν4 ν5 V E F
4 0 0 4 6 4
ν3 ν4 ν5 V E F
2 3 0 5 9 6
B γ B γ B γ
2
A/2 A/2
A A α A
B β
2
B
Figure 11.1
B B B
cos = − cos cos A + sin sin A cos γ.
2 2 2
By employing the trigonometric identity tan 2θ = sin θ
1+cos θ (see (0.19)), this
relationship readily becomes
A B
tan tan = sec γ. (11.5)
2 2
It is left as an exercise to derive from this the companion relationships
α B β A
tan = tan γ cos , tan = tan γ cos . (11.8)
2 2 2 2
The companion relationship may be derived by observations of symmetry;
that is, replace α with β and B with A in (11.7). Or apply (3.4) to the spher-
ical triangle obtained by cutting the spherical rhombus in Figure 11.1(a) in
half the other way.
11.3. THE REMAINING DELTAHEDRA 173
Finally, we have
α β
sec sec = sec γ, (11.10)
2 2
the derivation of which is left as an Exercise.
For easy reference, we include the foregoing results in the case γ = 60◦ .
A B
tan tan = 2, (11.11)
2 2
3 + 5 cos B 3 + 5 cos A
cos A = − , cos B = − , (11.12)
5 + 3 cos B 5 + 3 cos A
α √ B β √ A
tan = 3 cos , tan = 3 cos , (11.13)
2 2 2 2
1 1
cos α = (3 cos A + 1), cos β = (3 cos B + 1), (11.14)
4 4
α β
sec sec = 2. (11.15)
2 2
t s
A B
u v
Figure 11.2
Now we know three parts of spherical ∆srt: two sides and an included
dihedral angle. Thus, with the aid of (3.4) and (3.6), the other parts of this
triangle may be determined. Carrying out this procedure, or simply noting
that the three parts already known correspond exactly to the vertex figure
of a tetrahedron, we see that st also has measure 60◦ , and dihedral angles
∠rst and ∠str also have measure D3,3 .
Since st has measure 60◦ , we see that stuv is a spherical rhombus with
sides measuring 60◦ . Hence the pentavalent vertex figure in Figure 11.2 may
be thought of as a spherical equilateral triangle and a spherical rhombus,
joined by a common side.
Thus, the dihedral angles at t and s have measures D3,3 +A and D3,3 +B,
respectively, where A and B are the dihedral angles of the spherical rhombus
(as in the previous section).
Since we are considering only convex deltahedra, the vertex figure in
Figure 11.2 must also be convex, so that both D3,3 + A and D3,3 + B must
measure less than 180◦ . Our work from the previous section will show that
this is not possible, so that we may infer the impossibility of a trivalent
vertex being adjacent to a pentavalent vertex on a deltahedron.
Now let’s see why D3,3 + A and D3,3 + B cannot both be less than 180◦ .
Since the sides of our rhombus measure 60◦ , (11.11)–(11.15) are valid, so
that
A B
tan tan = 2.
2 2
√
When are A and B equal? In this case, tan A2 = tan B2 = 2, so that
employing a trigonometric relationship yields
A
1 − tan2 2 1
cos A = 2 A
=− .
1+ tan 2 3
11.3. THE REMAINING DELTAHEDRA 175
So, if A and B are equal, then A and B are both D3,4 (see Table 3.1), and our
spherical rhombus is the vertex figure of an octahedron. Since cos D3,3 = 13 ,
we see that D3,3 and D3,4 are supplementary, so that D3,3 + A and D3,3 + B
are both exactly 180◦ .
Now what happens when A and B are not equal? Then one of tan A2
√ √
and tan B2 must be greater than 2, and one must be less than 2 in order
for their product to be precisely 2. But since the function y = tan x2 is an
increasing function when x is between 0◦ and 180◦ , this means that one of
A and B is greater than D3,4 , and one is less. Finally, we conclude that one
of D3,3 + A and D3,3 + B is greater than 180◦ , and one is less than 180◦ .
Thus, both D3,3 + A and D3,3 + B cannot both be less than 180◦ , and so
the trivalent vertex p cannot be adjacent to the pentavalent vertex q.
Where does this leave us? In §11.1, we found all deltahedra where a
trivalent vertex is adjacent either to another trivalent vertex or to a tetrava-
lent vertex. We just demonstrated that a trivalent vertex may not be adja-
cent to a pentavalent vertex. Thus, we have in fact enumerated all deltahe-
dra which contain a trivalent vertex.
As a result, the remaining deltahedra contain only tetravalent and pen-
tavalent vertices. Thus, we may put ν3 = 0 in (11.1)–(11.4), giving a system
of four equation in the five unknowns V , E, F , ν4 , and ν5 .
Of course, there are many ways to enumerate the solutions to such a
system. Our approach will be to solve the system in terms of ν4 , yielding
requires at least eight additional triangular faces (since at most two faces
are incident at both of these vertices). This brings the number of triangular
faces to at least 20, but in fact F = 20 − 2ν4 = 18. This contradiction
demonstrates the impossibility of the case ν4 = 1.
When ν4 = 2, it follows from (11.16) that ν5 = 8. There is just one way
to realize this case; indeed, there is just ony way to realize each of the other
possible cases. The deltahedron obtained here may be thought of as a square
antiprism with square pyraminds affixed to each of its two square faces. This
polyhedron is called the gyroelongated square dipyramid (see Figure
11.3(a)). In 1966, Norman Johnson named all convex polyhedra with regular
polygons as faces. As it turns out, there are 92 such polyhedra (not counting
the Platonic and Archimedean solids) – a few too many to consider here.
The gyroelongated square dipyramid is number 17 in Johnson’s enumeration,
and hence is usually called J17 for brevity.
Figure 11.3
Deltahedron ν3 ν4 ν5 V E F
Tetrahedron 4 0 0 4 6 4
Triangular dipyramid (J12) 2 3 0 5 9 6
Icosahedron 0 0 12 12 30 20
Gyroelong. sq. dipyr. (J17) 0 2 8 10 24 16
Triaug. triang. prism (J51) 0 3 6 9 21 14
Snub disphenoid (J84) 0 4 4 8 18 12
Pentagonal dipyramid (J13) 0 5 2 7 15 10
Octahedron 0 6 0 6 12 8
Table 11.1
r
u t
v
w
Figure 11.4
One immediately sees that the spatial configuration of each part is com-
pletely determined by the dihedral angle at edge vw (see Figure 11.4), since
this dihedral angle determines all others at the tetravalent vertices v and w
by virtue of (11.11).
How large should this dihedral angle be? This question may be answered
by considering how the two identical parts are assembled to form the snub
disphenoid. By holding one part in each hand, the methodology is simple:
the protruding vertices r and t of one part must exactly fit into the sunken
vertices s and u of the other part.
There are at least two ways to find this dihedral angle. One method
employs spherical trigonometry, and is left as an Exercise. The second
reduces to a problem in plane geometry; it will be presented below.
So denote by A the measure of the dihedral angle at the edge vw. Because
v and w are tetravalent vertices, the dihedral angles at edges rv and tw also
have measure A. Employing the notation of §11.2, we see that the dihedral
angles at edges sv, sw, uv, and uw therefore have measure B. In addition,
the angles rvw and vwt must have measure β (as in Figure 11.1(c)).
Assume that the edges of the snub disphenoid have length 1, and denote
by x the midpoint of the edge vw. The above considerations result in the
following two diagrams:
11.4. THE SNUB DISPHENOID 179
r t s u
√ √
1 1 3 3
2 2
A
β β
v w
1 x
(a) (b)
Figure 11.5
Using the cosine law for triangles with Figure 11.5(b) results in
3
[su]2 = (1 − cos A). (11.18)
2
Finally, setting [rt]2 = [su]2 from (11.17) and (11.18) and substituting from
(11.14) for cos β yields
All that remains is to substitute from (11.12) for cos A into (11.19),
multiply through by 5 + 3 cos B, and collect like terms. This results in the
following cubic equation for cos B:
Of course, we may also substitute from (11.12) for cos B into (11.19) to ob-
tain a cubic equation for cos A. The following equation is thereby produced:
These cubic equations may be solved exactly (as in Chapter 8); approxi-
mate solutions are given by A ≈ 96.198◦ and B ≈ 121.743◦ . For further
exploration of these cubic equations, see the Exercises.
Finally, the remaining dihedral angle of the snub disphenoid, C, satisfies
the cubic equation
1 √ √
cos Da4 1 4
4 + 2 D3,4 = 1− 2−2 2 . (11.23)
3
√
r
◦ 1 ◦ 1
2
cos (60 + D3,4 ) = − 1 + 2 6 , cos 90 + 2 D3,4 =− .
6 3
Tetrahedron 70.529 — —
Triangular dipyramid 70.529 141.058 —
Icosahedron 138.190 — —
Gyroelongated square dipyramid 109.471 127.552 158.572
Triaugmented triangular prism 109.471 169.471 144.736
Snub disphenoid 96.198 121.743 166.441
Pentagonal dipyramid 138.190 74.755 —
Octahedron 109.471 — —
Table 11.2
u
60◦
A v
60◦
s
Figure 11.6
B/2 B/2
r t
E E
60◦ 60◦
v B B w
60◦
Figure 11.7
Begin with Figure 11.8. Note that ∆ryt of Figure 11.7 and ∆vwt of
Figure 11.8 are halves of spherical rhombi (as discussed in §11.2), so that
11.6. THE REMAINING DIHEDRAL ANGLE 183
α
r A t
E E− 2 A
2
60◦ β 60◦
A
B− 2
A/2 B w
v
60◦
Figure 11.8
Now again substitute for cos 12 A from (11.13), use the relationship tan 12 β =
1−cos β 1
sin β , and solve for sin E sin 2 B:
cos β − 1
sin E sin 12 B = 4 cos E cos 12 B .
2 cos β + 1
Substitute this expression for sin E sin 12 B into (11.26) and simplify, so that
5 − 2 cos β
cos C = cos E cos 21 B .
2 cos β + 1
Now substitute for cos E by applying (3.4) to ∆vrt and for cos 21 B using
(11.13) and the identity tan 12 α = 1+cos
sin α
α , yielding
Substitute for cos α and cos β from (11.14), and then substitute for cos A
from (11.12) and simplify:
1 (9 cos2 B + 24 cos B + 7)(3 − cos B)
cos C = . (11.27)
24 1 + cos B
The idea here is to get the 1 + cos B to cancel. One trick is to use (11.20).
We add the appropriate multiple of the left-hand of (11.20) (which is 0) to
the numerator of (11.27). As it happens, once does the job, so that
(9 cos2 B + 24 cos B + 7)(3 − cos B) + 27 cos3 B + 27 cos2 B − 75 cos B − 43 =
2(cos B + 1)(9 cos2 B + 6 cos B − 11),
Hence
1
cos C = (9 cos2 B + 6 cos B − 11). (11.28)
12
Thus, we have cos C as a quadratic function of cos B. Using (11.20),
then, it is possible to find a cubic equation of which cos C is a root. The
task is not too complicated; we include it for completeness. The uninterested
reader may skip directly to (11.31).
By defining
λ = 3 cos B + 1, µ = 12 cos C, (11.29)
our algebraic work will be made somewhat less cumbersome. With these
definitions, (11.28) becomes
µ = λ2 − 12, (11.30)
and with cos B = 13 (λ − 1), (11.20) becomes
λ3 − 28λ − 16 = 0.
Be rewriting this equation as λ3 − 28λ = 16 and squaring both sides of this
relation, we obtain
λ6 − 56λ4 + 784λ2 − 256 = 0.
From (11.30), we see that λ2 = µ + 12, so that
(µ + 12)3 − 56(µ + 12)2 + 784(µ + 12) − 256 = 0,
resulting in
µ3 − 20µ2 − 128µ + 2816 = 0.
Substituting from (11.29) and dividing through by 64 yields
27 cos3 C − 45 cos2 C − 24 cos C + 44 = 0. (11.31)
11.7. EXERCISES 185
11.7 Exercises
1. As suggested in the text, derive equation (11.6) from (11.5).
2. Derive equation (11.10).
3. Obtain the equations in (11.16) as suggested in the text.
4. Using the notations of §11.2, assume that γ = 60◦ so that (11.11)–
(11.15) are applicable. Show that
cos 12 α sin 12 B = cos 12 A.
α
r t
A
2
60◦ β 60◦
B 0 A/2
v B w
60◦
Figure 11.9
Since B is the dihedral angle at w where two 60◦ arcs meet, spherical
∆vtw is half a spherical rhombus, so all notations and results from
Figure 11.1 and §11.2 are applicable. That arc rt must have mea-
sure α is evident from how the two halves of the snub disphenoid are
assembled. Let β divide the angle at v into two angles, one having
measure 12 A and the other B 0 .
Now apply (3.4) to spherical triangle ∆rvt, yielding
cos α = cos 60◦ cos β + sin 60◦ sin β cos B 0 .
Using the fact that B 0 = B − 21 A, expand cos(B − 12 A) and employ
(11.11)–(11.15) and trigonometric identities as appropriate to derive
(11.19).
186 CHAPTER 11. DELTAHEDRA
1
cos B = − (9 cos2A + 60 cos A + 19). (11.32)
24
1
tan 21 D = √
4
.
2−1
Similarly, show that the dihedral angle D of the pentagonal dipyramid
described by (11.24) is also given by
√
tan 12 D = 3 − 5.
1
cos C = − (9 cos2A + 72 cos A + 31).
24
(c) Square both sides of the previous equation and use (11.21) as
necessary to conclude that
1
cos2 C = − (6 cos2A + 44 cos A − 18).
24
1
cos A = (18 cos2 C − 12 cos C − 29).
3
11.7. EXERCISES 187
a = 2 tan2 12 A, b = 2 tan2 12 B,
show that
Kepler-Poinsot Polyhedra
189
190 CHAPTER 12. KEPLER-POINSOT POLYHEDRA
Figure 12.1
is not within the scope of this text. So for this chapter, we must be content
to have the enumeration completed for us. There is still plenty to do...
As it happens, four new polyedra await us – that is, polyhedra whose
faces are regular polygons, the same number of which are incident at each
vertex, but which are not convex. Three of them are relatives of the dodec-
ahedron in the following sense. A glance at Figure 1.1 reveals that in the
center of the pentagram, we find a smaller pentagon. Thus, beginning with
a pentagon, we may extend its sides so that a pentagram is formed. If this
procedure is done to the twelve pentagons on a regular dodecahedron, we
find that the nonconvex polyhedron in Figure 12.2 is formed (20 in Polyhe-
dron Models). This polyhedron is called a small stellated dodecahedron
(“stellated” referring to the process of creating the “stars” just described).
As its faces are pentagrams and five such meet at each vertex, we refer to
this polyhedron by the Schläfli symbol { 25 , 5}.
12.3 Measurements
Now that we have created these polyhedra, the desire possesses us to calcu-
late their edge angles, dihedral angles, and circumradii. We are extremely
fortunate in that (3.7), (3.8), (3.11), (3.12), and (3.13) remain valid for the
four new pairs of {p, q} values discussed above. This is not merely lucky; the
interested reader may verify that in the derivations of the aforementioned
formulas, much use is made of the fact that the angle subtended by an edge
of a p-gon at its center is 2π/p, a property that is valid when p = 5/2 as well
(as seen in (S) in §12.1). The appropriate extension of Table 3.1 is given in
Table 12.1. Recall that we fashioned three of the Kepler-Poinsot polyhedra
12.4. DIHEDRAL ANGLES 193
Table 12.1
5 5
p 3 3 4 3 5 3 2 5 2
5 5
q 3 4 3 5 3 2 3 2 5
5 5
r 4 3 3 2 2 5 5 3 3
Table 12.2
These nine triplets of (p, q, r)-values form a set of sorts, representing the
nine polyhedra whose faces are the same regular polygon, where the same
number of such polygons meet at each vertex. Now that the set is complete,
some interesting results may be demonstrated. For example, for (p, q, r) as
in Table 12.2, we have
Dp,q + Dp,r = π. (12.1)
To see this, we begin with a result from Exercise 6 of Chapter 3; namely,
that
1 π π
sin Dp,q sin = cos .
2 p q
194 CHAPTER 12. KEPLER-POINSOT POLYHEDRA
Solving this equation for sin 12 Dp,q , substituting it into the relationship cos Dp,q =
1 − 2 sin2 21 Dp,q , and finding a common denominator yields
π π 1 − cos2 πp − 2 cos2 πq
cos Dp,q = 1 − 2 csc2 cos2 = . (12.2)
p q sin2 πp
where the latter equality follows from inspection of (12.2). Now Dp,q and
Dp,r , being dihedral angles between faces of polyhedra, are both between
0 and π. But two such angles have opposite cosines only when they are
supplementary, which gives (12.1).
1 + cos 2π
q 1 + cos 2π
p
cos Dp,q = 1 − 2 2π , cos Ep,q = 2 − 1. (12.5)
1 − cos p 1 − cos 2π
q
1 π π 1 π π
sin Dp,q = csc cos , cos Dp,q = csc cos . (12.6)
2 p q 2 p r
1 π π 1 π π
sin Ep,q = csc cos , cos Ep,q = csc cos . (12.7)
2 q p 2 q r
π π
ρ = sin sec . (12.8)
q r
Notice that (12.4) follows directly from (3.12) and (12.1). (12.5) is a
combination of (3.7) and (3.11). (12.6) and (12.7) follow from Exercise 6 of
Chapter 3 (see Exercise 2 below). (12.8) is simply (3.16).
12.6. EXERCISES 195
12.6 Exercises
1. Derive (12.1) as suggested in the text.
E 5 ,5 = π − E3,5 , E 5 ,3 = π − E5,3 .
2 2
3. Show that the results of Exercises 2–6 of Chapter 3 remain valid for
the values of (p, q, r) found in Table 12.2.
4. Using (3.12) and (12.1), show that for all values of (p, q, r) from Table
12.2, we have
Ep,r + Eq,r = π.
(a) Using (3.9), show that the angles between the sides of a spherical
pentagon derived from a pentagonal face of the icosidodecahedron
have measure D 5 ,5 .
2
(b) Again using (3.9), show that the angles between the sides of a
spherical triangle derived from a triangular face of the icosido-
decahedron have measure D 5 ,3 .
2
(c) Use (a) and (b) to argue that the twelve spikes from a small stel-
lated dodecahedron and the twenty spikes from a great stellated
dodecahedron may be assembled to form an icosidodecahedron.
(d) Build two spikes of each type – i.e., two triangular pyramids and
two pentagonal pyramids – to illustrate the result in (c).
196 CHAPTER 12. KEPLER-POINSOT POLYHEDRA
Chapter 13
Euler’s Formula
In this section, we derive the formula for finding the area of a spherical
triangle (and other spherical polygons) alluded to in §3.1. We denote by ρ
the radius of the sphere; recall that the surface area of the sphere is then
4πρ2 .
Now consider the spherical triangle pqr as shown in Figure 13.1, whose
dihedral angles (in radians) are A, B, and C, with opposite sides a, b, and
c, respectively. Here, p, q, t, and s all lie on the great circle containing the
arc pq (with measure b), so that Figure 13.1 is an aerial view of a sphere.
In addition, arcs prt and qrs are semicircles containing the edges pr and
qr, respectively, of the spherical triangle. The areas of the four regions
into which the hemisphere is divided are denoted by K, L, M , and N for
convenience.
Since precisely half of the sphere is in view in Figure 13.1, it is evident
that
1
K + L + M + N = (4πρ2 ) = 2πρ2 . (13.1)
2
197
198 CHAPTER 13. EULER’S FORMULA
p A K C q
c a
B
N r L
s t
Figure 13.1
Now A is the dihedral angle of the lune pqtrp of the sphere, whose
shape may be likened to a section of orange or grapefruit. Since A/2π is
that fraction of a revolution occupied by the dihedral angle of the spherical
triangle at p, it follows that the lune pqtrp occupies the same fraction of the
surface area of the sphere, which relationship may be algebraically expressed
as
A
K +L= (4πρ2 ) = 2Aρ2 . (13.2)
2π
Similar considerations yield that
3K + L + M + N = 2ρ2 (A + B + C).
2K = 2ρ2 (A + B + C − π),
so that
K = ρ2 (A + B + C − π). (13.4)
Thus, if A, B, and C are the dihedral angles of a spherical triangle, and we
write Σ = A + B + C, then the area of this spherical triangle is ρ2 (Σ − π).
Therefore the fraction of the sphere occupied by the triangle is
ρ2 (Σ − π) 1 1
= (Σ − π) = (Σ − 180◦ ),
4πρ 2 4π 720◦
13.1. AREA OF A SPHERICAL POLYGON 199
K1 = ρ2 (A + B1 + D1 − π)
and
K2 = ρ2 (C + B2 + D2 − π).
B1 B2
C
A K1 K2
D1 D2
Figure 13.2
Thus,
The ambitious reader may provide a straightforward argument for this result
using (13.4) and the principle of mathematical induction.
K + nπ − 2π = D1 + D2 + · · · + Dn . (13.6)
We may write such an expression for each spherical polygon, and then sum
them all. We take a moment to consider what that sum would be.
Since K is the area of the spherical polygon P, summing the “K” terms
yields the sum of the areas of the spherical polygons which tile the sphere
onto which our polyhedron is projected. Since the polygons cover the entire
surface of the sphere, this sum is therefore 4π.
13.3. DESCARTES’ RULE OF DEFICIENCY 201
4π + 2Eπ − 2F π = 2V π.
the “2π” terms gives 2π for each vertex of the polyhedron, so that this sum
is 2V π. And finally, summing α1 + α2 + · · · + αn over all vertices of the
polyhedron is just the sum of all angles in all the faces of the polyhedron,
since each angle of a face is incident at exactly one vertex of the polyhedron.
Denote by S this sum of all angles in all faces of the polyhedron. Thus, we
have so far determined that
∆ = 2V π − S. (13.7)
nπ − 2π = β1 + β2 + · · · + βn .
2Eπ − 2F π = S. (13.8)
∆ = 2V π − (2Eπ − 2F π) = 2π(V − E + F ).
the truncated cube, and the cuboctahedron are all equal. This results not
from the particular shape of the faces meeting at a given edge, but rather
from the fact a facial plane of a cube is meeting the facial plane of an oc-
tahedron along that edge of the polyhedron. When this occurs, the angle
made by two such planes is about 125.264◦ .
What is the significance of 125.264◦ ? Think for a moment about trun-
cating an octahedron. It should be clear that a half-octahedron (or square
pyramid) is being sliced off each vertex. It is also evident that the dihedral
angles at the square base of this pyramid must be 12 D3,4 . So the new dihedral
angle created by slicing off a half-octahedron must be
1
Dt3,4 = π − D3,4 ≈ 125.264◦ . (13.10)
2
Now it is also clear from Exercise 6(a) of Chapter 2 (and Table 3.1) that the
dihedral angles of the tetrahedron and octahedron are supplementary, or
3
De3,4 [4–4] = DE
3,4 [4–8] = π.
4
To account for De3,4 [4–3] = DE 3,4 [4–6], we must find the angle at which
the facial planes of a rhombic dodecahedron meet those of an octahedron.
Examine Figure 13.3, where the fact that a rhombicuboctahedron may be
inscribed in a rhombic dodecahedron is used. It is clear that if the vertices of
a rhombic dodecahedron are appropriately truncated (leaving squares as in
Figure 9.12), a rhombicuboctahedron remains. Because the dihedral angles
of the rhombic dodecahedron are 120◦ , it must be that the squat tetrahedra
sliced off the trivalent vertices of the rhombic dodecahedron are in fact one-
fourth tetrahedra. As a result, the dihedral angles at the equilateral bases of
204 CHAPTER 13. EULER’S FORMULA
C O R
π π
C 2 2 + 12 D3,3 3
4π
π
O 2 + 21 D3,3 π − D3,3 π − 12 D3,3
3
R 4π π − 12 D3,3 2
3π
Table 13.1
We may verify our geometrical observations with explicit calculation in
the event we have made an erroneous assumption along the way. (Of course,
we may always check that the angles agree to several decimal places on our
calculators, but we include a trigonometric verification for the purist.)
For example, since π2 + 12 D3,3 is an obtuse angle,
r
t π 1 1 1 − cos D3,3 1
cos D3,4 = cos + D3,3 = − sin D3,3 = − = −√ ,
2 2 2 2 3
in agreement with Table 6.1.
Figure 13.3
13.5. SOLID ANGLES 205
I D R
3
I D3,5 2π − 12 (D3,5 + D5,3 ) π
2 + 12 D3,5
3
D 2π − 21 (D3,5 + D5,3 ) D5,3 π
2 + 12 D5,3
π
R 2 + 12 D3,5 π
2 + 12 D5,3 4
5π
Table 13.2
For convenient reference, the same table is given below with all values
in degrees.
I D R
Table 13.3
of the surface area of the sphere does this spherical triangle occupy?” By
using (13.4), we see that the area of this triangle is
π π π πρ2
2
ρ + + −π = .
2 2 2 2
Since the surface area of the sphere has area 4πρ2 , this fraction is simply
πρ2
2 1
S4,3 = = .
4πρ2 8
K 1
S= = (A + B + C − π). (13.12)
4πρ2 4π
Of course, if more than three faces meet at the vertex of our polyhedron, we
generalize (13.5):
1
S= (D1 + D2 + · · · + Dn − (n − 2)π). (13.13)
4π
Using (13.12) in the case of the tetrahedron, we find that
3 1
S3,3 = D3,3 − . (13.14)
4π 4
For the octahedron, we must use (13.13), so that together with (13.11), we
have
1 1
S3,4 = − D3,3 . (13.15)
2 π
Now we may solve (13.14) for D3,3 , resulting in
4π 1
D3,3 = S3,3 + .
3 4
1 4
S3,4 = − S3,3 .
6 3
Multiplying this equation through by 6 and rearranging terms yields
Surely this must be magic! But no – this simple relationship between the
solid angles at vertices of the tetrahedron and octahedron admits an elegant
geometrical explanation.
Consider the cuboctahedron. In view of the fact that the edge angle is
◦
60 , it is evident that if a triangular face is connected to the center of the
cuboctahedron, a regular tetrahedron is formed. If a square face is joined
to the center, a square pyramid with equilateral triangular faces is formed
– in other words, a half-octahedron is created.
Since a cuboctahedron has eight triangular faces and six square faces,
then these eight tetrahedra and six half-octahedra meet at the center of the
cuboctahedron. Moreover, due to our construction, there are no gaps, so
that these 14 polyhedra completely fill space at the center of the cubocta-
hedron.
But this is precisely what (13.16) validates: the fraction of space oc-
cupied by 8 tetrahedral and 6 octahedral vertices is 1, or 100%. This also
indicates that space may be filled with alternating tetrahedra and octahedra:
arrange tetrahedra and octahedra face-to-face so that no two tetrahedra are
ever adjacent, nor are two octahedra ever adjacent. (Of course those readers
already familiar with this tiling of space would understand the geometrical
content of (13.16) immediately.)
As a result, any tiling of space should be verifiable using (13.16). But
as a caution, finding a combination of solid angles which sum to 1 does not
necessarily guarantee that a face-to-face tiling of space exists. For example,
it is clear from (13.16) that
4 S3,3 + 3 S3,4 + 4 S4,3 = 1,
although tetrahedra, octahedra, and cubes can never tile space face-to-face
since the cubes can never be adjacent to the tetrahedra or octahedra.
For reference, solid angles of the vertices of the Platonic solids and the
nonsnub Archimedean solids are given in Table 13.4 (located after the Ex-
ercises). They may be used to verify that various other combinations of
polyhedra tile space; details will be left to the Exercises.
13.6 Exercises
1. Verify the angles given in Tables 13.1 and 13.2 by taking their cosines
and comparing with Table 6.1.
2. Each of the following combinations of polyhedra can arranged so that
they tile space. For each combination, determine the arrangement
208 CHAPTER 13. EULER’S FORMULA
necessary and write an analogue of (13.16) for each case. Verify your
results using Table 13.4.
Polyhedron Symbol S S
Table 13.4
210 CHAPTER 13. EULER’S FORMULA
Chapter 14
Coordinates of Polyhedra
Now that we have thoroughly analyzed the convex and nonconvex regular
and semi-regular polyhedra using spherical trigonometry, we take some time
to look at polyhedra from a few different perspectives. An entire volume
could be written on these topics, so the reader will have to be content with
the brief introductions given here.
211
212 CHAPTER 14. COORDINATES OF POLYHEDRA
p r
Figure 14.1
Let’s see how this may be shown analytically; that is, by imposing a
convenient coordinate system. It is simplest in this case to choose the center
of the circle as the origin (0, 0), and have p and r be opposite ends of a
diameter of a unit circle with coordinates (−1, 0) and (1, 0), respectively.
Thus, the circle is described by the equation x2 + y 2 = 1 in this coordinate
system. (When the choice of origin is not clear, it sometimes takes more
than one attempt to find the simplest approach.)
Now assign q the coordinates (x0 , y0 ); thus, x20 + y02 = 1. The slope
between points p and q is given by
y0 − 0 y0
mpq = = ,
x0 − (−1) x0 + 1
y0 − 0 y0
mrq = = .
x0 − 1 x0 − 1
y0 y0 y2
mpq mrq = · = 0 2 = −1.
x0 + 1 x0 − 1 −y0
Thus, the product of the slopes of the lines containing the two sides of the
triangle is −1, so that these two sides must be perpendicular at q. Hence
∆pqr is a right triangle.
This analytic solution is rarely (if ever!) encountered since the synthetic
approach is so straightforward. The point here is that both approaches yield
correct results, but for one reason or another, one approach may be preferred
to the other.
14.2. THE CUBE 213
Before giving coordinates for the cube, we review coordinates for a square
in two dimensions. One common coordinatization for a square lists vertices
as (0, 0), (1, 0), (1, 1), and (0, 1), describing a unit square with lower left
corner at the origin.
While there is nothing “wrong” with this choice of coordinates, it hap-
pens that for a discussion of symmetry and matrices in the next chapter, it
is important that the origin is the center of the square (or cube) rather than
one of its vertices.
So instead, we consider a square whose vertices are (1, 1), (−1, 1), (−1, −1),
and (1, −1). This square has sides of length 2, and the center of this square
is in fact (0, 0).
+z
(1, 2, 3)
+y
+x
Figure 14.2
Also note that two vertices are adjacent if and only if they differ in exactly
one coordinate; this coordinate indicates the axis to which the edge joining
these two vertices is parallel.
For the student more familiar with points, lines, and planes in a three-
dimensional coordinate system, it is instructive to find equations for the
lines containing the edges of the cube and the planes containing the faces of
the cube. Such discussions, however, will be relegated to the Exercises.
(−1, −1, 1)
(−1, 1, 1)
(1, −1, 1)
(1, 1, 1)
Figure 14.3
Now that we have coordinates for the vertices, how might we use them
to find the dihedral and edge angles of the cube? Some simple Euclidean
geometry is all that is needed. Consider Figure 14.4, where the ends of an
edge of the cube are connected to the origin to form the edge angle E. To
find E, all we need is the length of the sides of the given triangle; we know
that the top edge, being an edge of the cube, has length 2.
E
(0, 0, 0)
Figure 14.4
To find the other side lengths, we need the distance formula in three
dimensions. It turns out that this is a natural extension of the familiar
216 CHAPTER 14. COORDINATES OF POLYHEDRA
Applying this
√ to the triangle in Figure 14.4, we see that the other two sides
have length 3, so that the cosine law for triangles results in
√ 2 √ 2 √ √
22 = 3 + 3 − 2 3 3 cos E,
or
1
cos E = ,
3
as expected.
Finding the dihedral angle is only a little trickier. Recall that the dihe-
dral angle is the angle between adjacent faces of a Platonic solid. This angle
may be obtained as follows: slice the Platonic solid with a plane through the
midpoint of an edge and perpendicular to that edge. This creates an angle
which also passes through the centers of the two faces which meet in that
edge. So in general, the dihedral angle of a Platonic solid is always part of
a triangle whose vertices are the midpoint of a given edge and the centers
of the two faces which meet at that edge.
So let’s consider the edge given in Figure 14.4. Its midpoint is found
by averaging the respective coordinates (as in two dimensions), resulting
in (1, 0, 1). It is the edge where the top and front faces of the cube meet.
Since the positive z-axis goes through the center of the top face, its center is
(0, 0, 1), and likewise, the center of the front face is given by (1, 0, 0). Thus,
the triangle as shown in Figure 14.5 is formed. Since we have coordinates
for the three vertices of the triangle, we may find the lengths of the three
sides using (14.1) and use the cosine law for triangles to find cos D = 0, as
expected.
(0, 0, 1) (1, 0, 1)
D
(1, 0, 0)
Figure 14.5
14.3. THE TETRAHEDRON AND OCTAHEDRON 217
In the future, we will simply jump from a diagram such as Figure 14.5 to
the final result, letting the reader fill in the necessary algebraic details. We
note that the reader familiar with vector geometry and the inner product
(or dot product) may see another way to calculate cos D from Figure 14.5,
and should feel free to do so.
The reader who has completed Exercise 6(b) of Chapter 2 or studied Figure
9.4 may recall that the tetrahedron may be inscribed in a cube in two ways.
For the less thorough reader, we may see this fact by choosing a vertex of
the cube. Three square faces meet at this vertex, so that three diagonals
of these squares meet at this vertex as well. The other three ends of these
diagonals, together with the chosen vertex, form the vertices of a regular
tetrahedron. It might be instructive to build a cube at this point and draw
in the appropriate diagonals to see this.
The four remaining vertices of the cube also form a regular tetrahedron
which interpenetrates the first. The region common to these two tetrahedra
is an octahedron (see Exercise 6(b) of Chapter 2 again) whose vertices are
the centers of the faces of the cube.
For specificity, choose one of these tetrahedra; say, begin with the vertex
(1, 1, 1). The opposite ends of the diagonals meeting at this vertex have two
coordinates changed, resulting in (1, −1, −1), (−1, 1, −1), and (−1, −1, 1).
On the tetrahedron, each of these vertices is adjacent to the others, always
differing in exactly two coordinates.
The edge angle may be determined as before using a triangle with vertices
(0, 0, 0), (1, 1, 1), and (1, −1, −1). In other words, simply choose an edge of
the tetrahedron as a side of the triangle, and use the origin as the third
vertex. This results in cos E = − 31 .
To find the dihedral angle of the tetrahedron, recall that we may form
a triangle from the midpoint of an edge and the centers of the two faces
meeting at that edge.
So begin by choosing an edge of the tetrahedron, say the edge with ends
(1, 1, 1) and (1, −1, −1). The midpoint of this edge is (1, 0, 0). To find the
centers of the incident faces, we may use the following result: the center of
a regular polygon is found by averaging the coordinates of its vertices.
Since one face incident at this edge has vertices (1, 1, 1), (1, −1, −1), and
218 CHAPTER 14. COORDINATES OF POLYHEDRA
(1, 0, 0)
D
( 13 , 13 , − 13 )
q
2
3
√
2 2
3
( 13 , − 13 , 31 )
Figure 14.6
Given these relationships, it is not difficult to find the edge and dihedral
angles of the octahedron using the methods described in this chapter. Details
are left to the Exercises.
To determine the y-coordinate, recall from §1.1 that the edges and di-
agonals of a regular pentagon have lengths in the ratio 1 : τ . Since the cube
has edges of length 2 (diagonals of the pentagons), the length of the sides of
the pentagons must be τ2 . Since the y-coordinates of p and q are opposite
due to symmetry, they must be − τ1 and τ1 , respectively.
0, − τ1 , z = p q = 0, τ1 , z
(−1, 1, 1)
(−1, −1, 1)
(1, −1, 1)
(1, 1, 1)
Figure 14.7
How may z be determined? One way is to note that the distance from q
to the vertex (1, 1, 1) must be τ2 , the length of the edges of the dodecahedron.
In other words, using (14.1),
p 2
(0 − 1)2 + (τ −1 − 1)2 + (z − 1)2 = .
τ
Solving this quadratic equation for (z − 1)2 using Table 1.1 as needed, we
have
1
(z − 1)2 = 2 − τ = 2 ,
τ
so that
1 1 1
z = 1 + = τ, z = 1 − = 2 .
τ τ τ
Since z > 1, we choose the solution z = τ , the other solution being extrane-
ous.
Because of the symmetrical way that the dodecahedron is assembled by
placing roofs on the faces of a cube, the coordinates for the other vertices
may easily be found.
Since we have just found coordinates for an edge of the dodecahedron,
calculating the edge angle may be done as usual.
To find the dihedral angle of the dodecahedron, we simply take the ap-
propriate cross section of Figure 14.7, shown in Figure 14.8(a) and redrawn
in Figure 14.8(b).
220 CHAPTER 14. COORDINATES OF POLYHEDRA
(0, 0, τ )
D5,3
(1, 0, 1) (−1, 0, 1)
(a) (b)
Figure 14.8
Given the coordinates of the vertices in Figure 14.7, it is not hard to cal-
culate the midpoints of the appropriate segments to obtain the coordinates
given in Figure 14.8(b). Once these are known, it is not difficult to use the
cosine law to find D5,3 . Details are left to the Exercises.
v
p
q
u
a
Figure 14.9
x2 + τ x − τ 2 = (x − 1)(x + τ 2 ) = 0.
It is worth remarking that the icosahedron has edge length 2 while the
dual dodecahedron has edge length τ2 . These lengths are in the ratio τ : 1.
Examination of Figure 9.6 reveals that each pair of mutually bisecting edges
may be thought of as the diagonals of a rhombus on a rhombic triacon-
tahedron. Thus, the lengths of the diagonals of the faces of a rhombic
triacontahedron are in the ratio τ : 1.
a b
λ1 c λ2
Figure 14.10
Figure 14.11
14.7 Exercises
1. If you are familiar with equations for lines and planes in three dimen-
sions, find equations for the lines and planes which contain the edges
and faces, respectively, of the cube as suggested in §14.2.
2. If you are familiar with the inner product (dot product), use it to show
that the angle D in Figure 14.5 has measure 90◦ .
3. As suggested in §14.3, find the edge and dihedral angles of the octa-
hedron using the techniques described in this chapter.
4. As suggested in §14.4, find the edge and dihedral angles of the dodec-
ahedron as described in that section.
6. As suggested in §14.5, find the edge and dihedral angles of the icosa-
hedron using the techniques described in this chapter.
7. Find coordinates for the vertices of one or more of the partially trun-
cated Platonic solids as described in §14.6.
Figure 14.12
11. Thinking along the lines of the previous exercise, use Figure 6.10 to
determine coordinates for the vertices of a rhombitruncated cubocta-
hedron.
226 CHAPTER 14. COORDINATES OF POLYHEDRA
Chapter 15
15.1 Introduction
As suggested in the previous chapter, efficient rendering of polyhedral images
requires more than a knowledge of coordinates for the vertices of polyhedra.
Familiarity with matrices and symmetry groups is extremely useful.
Matrices are typically dicussed in an undergraduate course in linear al-
gebra, while symmetry groups are usually presented in an abstract algebra
course. Only the barest introduction to these topics can be given in a single
chapter. The interested reader should be cautioned, though, that many text-
books in linear algebra approach the topic of matrices from a purely algebraic
perspective. In such texts, motivation for studying matrices is the solution
of simultaneous linear equations in several variables. The reader should look
for a text, such as Banchoff and Wermer’s Linear Algebra Through Geometry
(ISBN 0-387-97586-1), which considers matrices from a geometric viewpoint.
227
228 CHAPTER 15. MATRICES AND SYMMETRY GROUPS
the point p has coordinates (1, −1, −1), we may simply refer to the point
(1, −1, −1).
Before discussing matrices in particular, we must look at functions on
R3 in general. For example, consider the function A : R3 → R3 defined by
y−z
ax + by + cz = 1
15.2. DEFINITIONS AND NOTATIONS 229
is called the identity matrix; it has the property that for any point p,
I(p) = p.
The purist may remark that a matrix simply represents a function and
is not a function itself. As it turns out, if the underlying coordinate system
changes, the matrix representing a linear function also changes. But since
we will not be deviating from the use of our three-dimensional Cartesian
coordinate system, we will not need to notationally distinguish between a
linear function and its matrix.
Thus R takes each vertex of the cube into another, not necessarily distinct,
vertex of the cube.
The rotation R is said to be a symmetry of the cube in that if we begin
with the cube as described in §14.2 and apply the linear function R, the
resulting cube is indistinguishable from the original in that it occupies the
same region of space.
Another way to say this is that if we begin with a list of the eight ver-
tices of the cube and apply R to each one, we get the same eight vertices,
although usually in a different order. Thus, R is said to be a permutation
of the vertices of the cube. It turns out that every symmetry of the cube
produces a corresponding permutation of its vertices. However, not every
permutation of the cube’s vertices yields a symmetry of the cube; for exam-
ple, a permutation which merely switches the ends of one edge of the cube
and leaves the others intact does not yield a symmetry of the cube. For a
thorough discussion of the theory of permutations, consult a textbook on
abstract algebra.
Another symmetry of the cube is given by
0 −1 0
S = 1 0 0 .
0 0 1
R ◦ R ◦ R = R3 = I.
Thus, each diagonal of the cube generates two symmetries of the cube. Since
there are four diagonals, we have found eight symmetries of the cube so far.
Technically, the identity, I, which does not move the cube at all, is
considered a symmetry of the cube. We will count it, however, only at the
end. If we counted it as being generated by a rotation about an axis, it
would be overcounted several times.
We may realize the rotation S described above by holding a cube so
that our thumb and middle finger touch the centers of two opposite faces
of the cube. If S describes a 90◦ counterclockwise rotation about the axis
passing through these facial centers, S ◦ S = S2 describes a 180◦ rotation
about this axis, and S3 described a 270◦ rotation. Of course S4 = I, since a
360◦ rotation is obtained. As we are not counting I, three rotations are thus
produced. Since three rotations are produced by each of the three pairs of
opposite faces, nine more rotations bring the count to 17.
By holding a cube by the midpoints of a pair of opposite edges, a new
rotation is generated. As this must be a 180◦ rotation, only one rotation
is generated by each of the six pairs of opposite edges. These six rotations
bring the count to 23.
Then, as promised earlier, we count the identity, I, bringing the count
up to 24.
Let us see how this works with an example. Consider the clockwise 90◦
rotation about the x-axis. That is, your viewpoint is from the positive x-
axis, and you’re turning the cube a one-quarter rotation to your right. If we
call this rotation R, what is the matrix for R?
We first examine R((1, 0, 0)). Now (1, 0, 0) is just the center of the front
face of the cube. When we rotate the cube, this point does not move as it
is on the axis of rotation. Thus,
1 1
R 0 = 0 .
0 0
What happens to (0, 1, 0)? This point is the center of the right face
(looking from our viewpoint). When we rotate, this point is moved to the
center of the bottom face of the cube. Thus,
0 0
R 1 = 0 .
0 −1
234 CHAPTER 15. MATRICES AND SYMMETRY GROUPS
Finally, we look at (0, 0, 1), the center of the top face of the cube. After
rotation, this point becomes the center of the right face of the cube, so that
0 0
R 0 = 1 .
1 0
15.5 Reflections
Not all symmetries of the cube are rotations. Consider for a moment the
xy-plane as a mirror in which we may reflect the cube. In other words, we
reflect the top face to the bottom, and vice versa, through the plane z = 0.
This is easy to describe algebraically as well, for it becomes a simple matter
of changing the sign of the z-coordinate. The plane z = 0 is called a plane
of symmetry because it divides the cube into two identical pieces, each a
mirror image of the other. Using (15.5), it is easy to see that the matrix
which changes the sign of the z-coordinate is given by
1 0 0
Sz = 0 1 0 . (15.7)
0 0 −1
Of course, the xz-plane and the yz-plane are also planes of symmetry for
analogous reasons. Algebraically, these change the signs of the y-coordinate
15.5. REFLECTIONS 235
1 0 0 −1 0 0
Sy = 0 −1 0 , Sx = 0 1 0 ,
0 0 1 0 0 1
respectively.
Does the cube have other planes of symmetry? (The reader may wish
to pause here and consider the question before going further.) Imagine a
plane passing through a pair of opposite edges of the cube. Such a plane
also divides the cube into two identical pieces, each the mirror image of the
other. Since there are twelve edges on the cube in six opposite pairs, this
line of thought provides six more planes of symmetry of the cube, and thus
six additional reflections.
So the cube has nine planes of symmetry, each generating a reflection.
But there are other ways to generate reflections of the cube. For example,
select any of the 24 rotations described earlier, and consider the symmetry
Sz R = Sz ◦ R.
Sz ◦ R(p) = Sz (R(p)),
Figure 15.1
Try this with the symmetry S given above. It turns out that all products
are 0 except one, making the calculation det S = −1 somewhat easier. And
now the reason for introducing the determinant in the first place: if S is a
rotation, then det S = 1, while if S is a reflection, det S = −1.
Another nice property of the determinant is that it is multiplicative. In
other words, if M1 and M2 are two matrices, then
det(M1 M2 ) = (det M1 )(det M2 ). (15.10)
One consequence of this result is that the product of two opposite symme-
tries is direct. (It would be instructive to take a moment to try and see
238 CHAPTER 15. MATRICES AND SYMMETRY GROUPS
this now.) To see this, let S1 and S2 be two reflections; that is, opposite
symmetries. Then using (15.10),
det(S1 S2 ) = (det S1 )(det S2 ) = (−1) · (−1) = 1.
But since det(S1 S2 ) = 1, then S1 S2 must be a rotation (a direct symmetry).
Of course this presupposes that if S1 and S2 are reflections, then S1 S2
is a symmetry of the cube. But the product of any two symmetries is
always another symmetry; this is one reason the symmetries form a group.
If a symmetry leaves a cube occupying the same region of space, then any
product of symmetries also leaves a cube occupying the same region of space,
and hence is also a symmetry of the cube.
0 1 0
1. R1,1,1 = 0 0 1: Clockwise 120◦ rotation about the axis.
1 0 0
0 0 1
2. R21,1,1 = 1 0 0: Counterclockwise 120◦ rotation about the axis.
0 1 0
0 0 −1
3. R1,1,−1 = 1 0 0 : Clockwise 120◦ rotation about the axis.
0 −1 0
0 1 0
4. R21,1,−1 = 0 0 −1: Counterclockwise 120◦ rotation about the
−1 0 0
axis.
0 10
5. R1,−1,1 = −1 0 0: Clockwise 120◦ rotation about the axis.
0 −1 0
240 CHAPTER 15. MATRICES AND SYMMETRY GROUPS
0 −1 0
6. R21,−1,1 = 0 −1: Counterclockwise 120◦ rotation about the
0
1 0 0
axis.
0 0 −1
7. R−1,1,1 = −1 0 0 : Clockwise 120◦ rotation about the axis.
0 1 0
0 −1 0
8. R2−1,1,1 = 0 1: Counterclockwise 120◦ rotation about the
0
−1 0 0
axis.
1 0 0
9. Rx = 0 0 1: Clockwise 90◦ rotation about the x-axis.
0 −1 0
1 0 0
10. R2x = 0 −1 0 : 180◦ rotation about the x-axis.
0 0 −1
1 0 0
11. R3x = 0 0 −1: Counterclockwise 90◦ rotation about the x-axis.
0 1 0
0 0 −1
12. Ry = 0 1 0 : Clockwise 90◦ rotation about the y-axis.
1 0 0
15.8. THE DIRECT SYMMETRIES 241
−1 0 0
13. R2y = 0 1 0 : 180◦ rotation about the y-axis.
0 0 −1
0 0 1
14. R3y = 0 1 0: Counterclockwise 90◦ rotation about the y-axis.
−1 0 0
0 1 0
15. Rz = −1 0 0: Clockwise 90◦ rotation about the z-axis.
0 0 1
−1 0 0
16. Rz = 0 −1 0: 180◦ rotation about the z-axis.
2
0 0 1
0 −1 0
17. R3z = 1 0 0: Counterclockwise 90◦ rotation about the z-axis.
0 0 1
−1 0 0
18. Rx,1,1 = 0 0 1: 180◦ rotation about the axis.
0 1 0
−1 0 0
19. Rx,1,−1 = 0 0 −1: 180◦ rotation about the axis.
0 −1 0
0 0 1
20. R1,y,1 = 0 −1 0: 180◦ rotation about the axis.
1 0 0
242 CHAPTER 15. MATRICES AND SYMMETRY GROUPS
0 0 −1
21. R1,y,−1 = 0 −1 0 : 180◦ rotation about the axis.
−1 0 0
0 1 0
22. R1,1,z = 1 0 0 : 180◦ rotation about the axis.
0 0 −1
0 −1 0
23. R1,−1,z = −1 0 : 180◦ rotation about the axis.
0
0 0 −1
1 0 0
24. I = 0 1 0: The identity.
0 0 1
15.9 Exercises
1. Decide which of the following functions from R3 to R3 are linear. For
those which are linear, give a matrix representing the function.
(a)
3 2 −1 3
1 0 1 2 ,
−1 4 5 6
(b)
x 0 1
5
3 y 7 1 ,
−1 0 2 0
(c)
1 −1 2 1 3 0
1 0 0 5 ,
0 3
−1 1 −2 1 −3 2
(d)
0 −1
1 0 0 0
−1 0 0 4 −9 3 .
0 0 8 0 0 0
(e) M1 M2 ,
(f) M2 M1 ,
(g) det M1 ,
(h) det M2 ,
(i) det M1 M2 ,
(j) det M2 M1 .
5. Describe the effects of the following symmetries on the cube in Figure
14.3.
(a)
1 0 0
0 −1 0 ,
0 0 −1
(b)
0 −1
0
0 −1 0 ,
−1 0 0
(c)
−1 0 0
0 0 1 ,
0 1 0
(d)
0 1 0
−1 0 0 .
0 0 1
6. Recall that the octahedron and cube are dual polyhedra (see Chapter
9). With this in mind, write a convincing argument that the symmetry
group of the octahedron is the same as the symmetry group of the cube.
7. Find all symmetries of the cube in Figure 14.3 with the property that
the center of the front face is taken to the center of the top face.
15.9. EXERCISES 245
8. Find all symmetries of the cube in Figure 14.3 with the property that
the vertex (1, 1, 1) is taken to the vertex (1, −1, −1).
9. Find all symmetries of the cube in Figure 14.3 with the property that
the vertex (1, 1, 1) remains fixed and the vertex (−1, 1, 1) is taken to
the vertex (1, 1, −1).
10. A symmetry T of the cube interchanges the positive x-axis and positive
y-axis, and also satisfies T((1, 1, 1)) = (1, 1, −1). Find a matrix for T.
11. It M is a matrix, there may be a matrix N with the property
MN = NM = I.
In this case, N is called the inverse of M and is denoted by M−1 .
The matrix Sz in (15.7) satisfies
Sz ◦ Sz = I,
so Sz is its own inverse; i.e., S−1
z = Sz . This makes sense, since
applying Sz twice undoes the effect of Sz : Sz reflects about a plane,
then applying Sz again reflects back.
(a) Find the inverse of the matrix R given in (15.6).
(b) Find the inverse of the matrix R given by
0 0 −1
R = 0 1 0 .
1 0 0
12. Consider the tetrahedron with vertices (1, 1, 1), (1, −1, −1), (−1, 1, −1),
and (−1, −1, 1) which is inscribed in the cube in Figure 14.3. A mo-
ment’s thought reveals that any symmetry of the tetrahedron must be
a symmetry of the cube. However, a symmetry of the cube need not
be a symmetry of the tetrahedron (for example, a 90◦ rotation about
the center of a square face takes the tetrahedron into the other regular
tetrahedron inscribable in the cube).
In the language of abstract algebra, the symmetries of the tetrahedron
form a subgroup of the symmetries of the cube. Find the symmetries
of the tetrahedron. Keeping in mind the discussion in the previous
paragraph, give a reason for the number of symmetries of the tetrahe-
dron. How many are direct? How many are opposite?
246 CHAPTER 15. MATRICES AND SYMMETRY GROUPS
Chapter 16
(a) (b)
Figure 16.1
247
248 CHAPTER 16. GRAPH THEORY AND POLYHEDRA
6 5 6 5
7 4 7 4
8 3 8 3
1 2 1 2
(a) (b)
Figure 16.2
This suggests the following question: besides having eight vertices and
twelve edges, with three edges incident at each vertex, what other properties
must a graph possess so that it corresponds exactly to the adjacency of
vertices on a cube?
Such questions belong to a fascinating branch of mathematics known as
graph theory. While some mathematicians devote their entire professional
lives to its study, we can provide only the barest introduction here.
In studying polyhedra from a graph-theoretical perspective, we restrict
our attention to one particular aspect of polyhedra; namely, the adjacency of
their vertices. Features such as the size and shape of faces are not relevant.
For example, the graph of vertex adjacency for a rectangular prism is
also given in Figure 16.1. Since vertex adjacency is the only relevant fea-
ture under discussion, edge lengths on the polyhedron are irrelevant. As a
result, Figure 16.1 represents the vertex adjacency for a parallelepiped or
the frustum of a square pyramid.
Like any other branch of mathematics, graph theory has its own “lan-
guage” of definitions and concepts which, over time, have been relevant to its
particular application. Before going further, we must significantly enlarge
our graph-theoretical vocabulary.
16.2. BASIC DEFINITIONS 249
There are many ways to draw this graph; three are given below.
4 2
1 2 3
1 5
4
4 5 6
6 3
(a) (c)
1 2
3
6 5
(b)
Figure 16.3
Figure 16.3(b). To emphasize that this graph is regular; that is, the same
number of edges are incident at each vertex, one might draw the graph as
in Figure 16.3(a).
The point to be made is that a drawing of a graph is only a convenient
representation. When there is ambiguity, or perhaps when necessary for
proving a complex theorem, a graph can be specified in terms of V(G),
E(G), and ψ as described above. For our purposes, it will usually suffice to
just give a drawing of a graph.
Note that the definition of a graph does not preclude two vertices being
joined by more than one edge, as in Figure 16.4.
v1 v2
v3 v4
Figure 16.4
Here, two distinct edges join vertices v1 and v3 . Also, an edge may begin and
end at the same vertex, such as edge e in Figure 16.4. Such an edge is called
a loop. Graphs with these features do not usually arise when representing
adjacency of vertices on a polyhedron. When a graph has no loops and
no two vertices are joined by more than one edge, the graph is said to be
simple. Most of our discussion will focus on simple graphs.
The degree of a vertex v ∈ V(G), denoted by d(v), is the number of
edge ends incident at v. In the graph shown in Figure 16.4, d(v1 ) = 3, and
d(v4 ) = 3 as well since the loop e contributes two ends to v4 .
A basic theorem results from counting edge ends. On the one hand, we
may create the sum
X
d(v),
v∈V(G)
which indicates we sum the degrees of all the vertices of G. But since each
16.2. BASIC DEFINITIONS 251
edge contributes two ends to this sum, the result must be 2E, so that
X
d(v) = 2E (16.1)
v∈V(G)
Figure 16.5
A bipartite graph is a graph whose vertices may be partitioned into
two sets, V1 (G) and V2 (G), such that each edge has one end in V1 (G) and
the other in V2 (G). Such a graph is usually drawn so that the partition of
vertices is clear, as in Figure 16.6.
Figure 16.6
252 CHAPTER 16. GRAPH THEORY AND POLYHEDRA
While this may seem like a lot of new definitions and concepts, they are
all relevant to graphs derived from polyhedra. In the next section, we’ll see
how they apply to graphs derived from the Platonic solids.
and illustrates how many of the concepts defined earlier may be used in
combination.
To see this, note that the graph must have 12 edges. This follows from
X
2E = d(v) = 8 · 3,
v∈V(G)
1 2 3 4
5 6 7 8
Figure 16.7
Due to the symmetry of the graph drawn thus far, there are two alter-
natives for vertex 6: either join 6 to vertices 1, 2, and 3, or join 6 to vertex
4 and two of the vertices 1, 2, and 3. These alternatives are drawn in Figure
16.8.
1 2 3 4 1 2 3 4
5 6 7 8 5 6 7 8
(a) (b)
Figure 16.8
254 CHAPTER 16. GRAPH THEORY AND POLYHEDRA
1 2 3 4 1 2 3 4
5 6 7 8 5 6 7 8
(a) (b)
Figure 16.9
1 2 3 4
5 6 8 7
Figure 16.10
But Figure 16.10 looks exactly like Figure 16.9(a)! This means that
Figures 16.9(a) and (b) are essentially the same graph, since one may be
obtained from the other by simply relabelling some of the vertices (in this
case, switching vertices 7 and 8). Such graphs are said to be isomorphic.
Moreover, by redrawing Figure 16.9(a); that is, by putting its vertices at
different locations on paper while maintaining vertex adjacency, the familiar
Figure 16.11 is obtained.
16.3. THE PLATONIC GRAPHS 255
6 2
4 8
7 3
1 5
Figure 16.11
This demonstrates that if we adhere to the conditions that our graph
is simple, 3–regular, bipartite, and has eight vertices, we must eventually
create a graph such as Figure 16.11, a representation of the cube. In other
words, the Theorem is proved.
–The Icosahedron and Dodecahedron
A graph of the vertex adjacency of the icosahedron must have twelve
vertices, thirty edges, and be 5–regular since five edges meet at each vertex
of the icosahedron. There are many graphs with these properties, so there
is no simple theorem analogous to our recent result about the cube. This
graph is shown in Figure 16.12.
Figure 16.12
256 CHAPTER 16. GRAPH THEORY AND POLYHEDRA
Figure 16.13
16.4 Exercises
1. Show that there is no 3–regular graph with seven vertices.
3. The graphs below show the vertex adjacency for various convex poly-
hedra. Redraw the graphs so that the polyhedra are readily apparent.
For, example, Figure 16.2 may be redrawn to yield a cube as shown
in Figure 16.1. (Hint: It may be helpful to use Euler’s formula to
determine the number of faces on the polyhedron.)
16.4. EXERCISES 257
(a)
(b)
(c)
258 CHAPTER 16. GRAPH THEORY AND POLYHEDRA
(d)
4. Prove the following statement, using logic similar to that used in prov-
ing the theorem in §16.3: There is essentially one 4–regular simple
graph on six vertices: it is the graph of the vertex adjacency of the
octahedron.
5. (a) Show that the graph of the vertex adjacency of a hexagonal prism
is simple, 3–regular, and bipartite.
(b) By giving a counterexample, show that a simple, 3–regular bipar-
titite graph with twelve vertices need not necessarily correspond
to the graph of the vertex adjacency of a hexagonal prism.
(a) no two edges in M are incident; that is, no two edges in M have
a vertex in common, and
(b) every vertex of the graph is the end of some edge in M.
Basic Constructions
p q p q p q
Figure A.1
Let the segment pq be given. Open the compass wider than half the
length of the segment (otherwise the construction will not work). With
compass point at p, draw arcs above and below the segment, as in (a).
Repeat with the compass point at q, as in (b). Finally, join the intersections
of the pairs of arcs, as in (c). This line is the perpendicular bisector of pq,
so that it intersects the segment pq at its midpoint.
259
260 APPENDIX A. BASIC CONSTRUCTIONS
p p p
Figure A.2
p p p
(a) (b) (c)
Figure A.3
Let the angle with vertex p be given. With compass point at p, draw
an arc which passes through both sides of the angle, as in (a). With the
compass point at the points of intersection of this arc and the angle, draw
arcs which meet in the interior of the angle, as in (b). (The compass may
need to be adjusted at this step.) Join the intersection of these arcs and p,
as in (c). This line bisects the given angle.
A.4. CONSTRUCTING AN EQUILATERAL TRIANGLE 261
p q p q p q
(a) (b) (c)
Figure A.4
Open the compass to the desired length of the sides of the triangle, and
draw a line. With the compass point at some point on the line, say p, mark
off a point q on the same line, as in (a). With the compass set to the same
length, create intersecting arcs by placing the compass point at p, then at
q, as in (b). Complete the equilateral triangle as in (c).
p q
original angle
p p q p q
(a) (b) (c)
Figure A.5
Draw a line and a point p on the line as in (a). With the compass point
at the vertex of the original angle, draw an arc passing through both sides
of the angle. With this same compass setting, draw an arc with compass
point at p, intersecting the line at q as in (b). Measure the width of the arc
on the original angle. With this compass setting, draw an arc with compass
point at q intersecting the arc drawn in (b). Draw a line through p and this
intersection as in (c). This produces a copy of the original angle.
262 APPENDIX A. BASIC CONSTRUCTIONS
t t
s s
r r
(a) (b) (c)
Figure A.6
To trisect segment pq, draw a ray from p as in (a). With the compass
set not too wide and the compass point at p, mark off t on the ray drawn in
(a), as in (b). With compass point at t, mark off s, and with compass point
at s, mark off r. Note that pr is divided into equal thirds.
Draw rq. Copy ∠prq at t (see above), thereby creating similar triangles.
This will create pu, which is therefore one-third of the segment pq, as in
(c). With compass set to the length [pu] and compass point at u, create uv,
finishing the construction.
To divide a segment into, say, five congruent parts, simply mark off five
points at step (b) rather than three. This method may be extended to divide
a segment into any number of congruent parts.