i i

“MABK014-11” — 2010/6/25 — 13:48 — page 191 — #1

i i

CHAPTER 11
Geometry in Three Dimensions
Considerable obstacles generally present themselves to the
beginner, in studying the elements of Solid Geometry, from
the practice . . . of never submitting to the eye of the student,
the figures on whose properties he is reasoning.
Augustus De Morgan

The ludicrous state of solid geometry made me pass over this

branch.
Plato, The Republic (VII, 528)

Space is almost infinite. As a matter of fact, we think it is

infinite.
Dan Quayle
44th Vice President of
the United States of America
The geometry of three-dimensional Euclidean space is sometimes called
solid geometry, since it has traditionally dealt with solids such as spheres,
cylinders, cones, and polyhedra, as well as lines and planes in space. Three-
dimensional space is the space we live in, and it is the setting for some lovely
theorems and charming proofs.
In this chapter we present three different types of theorems and their
proofs. We first consider three-dimensional versions of some two-dimensional
theorems (such as the Pythagorean theorem). Second, we look at two-
dimensional theorems whose proofs are surprisingly simple when the the-
orem and its proof are viewed from a three-dimensional perspective. Then
we consider some classic theorems about polyhedra, the gems of three-
dimensional space.

191

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 192 — #2

i i

192 CHAPTER 11. Geometry in Three Dimensions

From Flatland to the Celestial Region

When Edwin Abbott Abbott (1838–1926) published his celebrated
Flatland: A Romance of Many Dimensions [Abbott, 1884], he included
the following dedication that remains today a suggestive invitation to
three (and higher) dimensional geometry:
To
The Inhabitants of S PACE IN G ENERAL
And H. C. IN PARTICULAR
This Work is Dedicated
By a Humble Native of Flatland
In the Hope that
Even as he was Initiated into the Mysteries
Of T HREE Dimensions
Having been previously conversant
With O NLY Two
So the Citizens of that Celestial Region
May aspire yet higher and higher
To the Secrets of F OUR F IVE OR E VEN S IX Dimensions
Thereby contributing
To the Enlargement of THE I MAGINATION
And the possible Development
Of that most rare and excellent Gift of M ODESTY
Among the Superior Races
Of S OLID H UMANITY

11.1 The Pythagorean theorem in

three dimensions
Some consider the three dimensional version of the Pythagorean theorem to
be the expression for the length of the diagonal of a rectangular box in terms
of the lengths of the edges. Others maintain that the real extension should
be some relationship between the areas of the faces of a right tetrahedron (a
tetrahedron with three faces perpendicular to one another at one vertex), a
three-dimensional analog of a right triangle.
We now state and prove one of these results. If one considers the three
mutually perpendicular faces of a right tetrahedron to be its legs and the other
face its hypotenuse, the theorem states that the square of the hypotenuse of

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 193 — #3

i i

11.2. Partitioning space with planes 193

a right tetrahedron equals the sum of the squares of the legs. The theorem is
sometimes called de Gua’s theorem, for the French mathematician Jean Paul
de Gua de Malves (1713–1785).
de Gua’s Theorem 11.1. In a right tetrahedron, the square of the area of
face opposite the vertex common to the three mutually perpendicular faces
is equal to the sum of the squares of the areas of the other three faces.

z
C(0,0,c)

h
B(0,b,0)
O y
g

A(a,0,0)
x
Figure 11.1.

Proof. Let O D .0; 0; 0/, A D .a; 0; 0/, B D .0; b; 0/, and C D .0; 0; c/
(with a; b; c positive) be the vertices of the right tetrahedron (See Figure
11.1). Let h denote the altitude to side AB in
ABC , and let g denote its
orthogonal projection onto the px-y plane. As such, g is the altitude to side
AB in
AOB. p Then g D ab= a2 C b 2 (since the area of
AOB is both
ab=2 and g a C b =2/ and thus h2 D g 2 C c 2 D a2 b 2 =.a2 C b 2 / C c 2 .
2 2

If K denotes the area of
ABC , then


2 2 
jABj h 2 1 a b
K2 D D .a2 C b 2 / C c 2
2 4 a2 C b 2

2
2
2
1 2 2 2 2 2 2 1 1 1
D .a b C a c C b c / D ab C ac C bc :

4 2 2 2

11.2 Partitioning space with planes

In Heinrich Dörrie’s classic book 100 Great Problems of Elementary Math-
ematics [Dörrie, 1965], one of the eight problems attributed to Jakob Steiner
is Problem 67:
What is the maximum number of parts into which a space can be di-
vided by n planes?

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 194 — #4

i i

194 CHAPTER 11. Geometry in Three Dimensions

Our solution to this problem follows an argument presented by George

Pólya over forty years ago. George Pólya (1887–1985) was renowned not
only for his mathematics but also for his contributions to the heuristics of
problem solving. In 1966 he filmed a session with future mathematics teach-
ers, entitled Let Us Teach Guessing, leading them to discover the answer to
Steiner’s question. The film is now a DVD [Pólya, 1966] distributed by the
Mathematical Association of America. Following Pólya’s approach, we be-
gin with n points on a line, then n lines in a plane, and finally n planes in
space.
Clearly n distinct points partition the real line into n C 1 intervals, and this
simple observation serves to establish our first theorem.
Theorem 11.2. The maximum number P .n/ of regions determined by n lines
in the plane is given by P .n/ D 1 C n.n C 1/=2.
Proof. First note that the maximum number of regions will occur when no
two lines are parallel and there is no point common to three or more lines.
Clearly P .0/ D 1; P .1/ D 2 and P .2/ D 4. Suppose k
1 lines partition
the plane into P .k
1/ regions, and we add a new line to create as many
additional regions as possible. Such a line will intersect all k
1 of the other
lines in k
1 distinct points, partitioning the line into k intervals and each
interval corresponds to a new region in the plane. See Figure 11.2, where we
illustrate the k D 4 case, a line intersecting three lines to create four new
regions in the plane.

Figure 11.2.

Hence P .k/ D P .k
1/ C k. Transposing the P .k
1/ term on the right

side to the left and summing yields
Xn Xn n.n C 1/
P .n/
P .0/ D ŒP .k/
P .k
1/ D kD ;
kD1 kD1 2
and consequently P .n/ D 1 C n.n C 1/=2 as claimed.

In the next proof it will be advantageous to recall the triangular numbers
tn D n.n C 1/=2 (see Section 1.1) and to write P .n/ as 1 C tn .

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 195 — #5

i i

11.3. Corresponding triangles on three lines 195

Theorem 11.3. The maximum number S.n/ of regions determined by n

planes in space is given by S.n/ D .n3 C 5n C 6/=6.

Proof. The maximum number of regions will occur when no two planes are
parallel, there are no two parallel intersection lines, and there is no point
common to four or more planes. Clearly S.0/ D 1, S.1/ D 2, S.2/ D 4,
and S.3/ D 8. Suppose k
1 planes partition space into S.k
1/ regions,
and we add a new plane to create as many additional regions as possible.
Such a plane will intersect all k
1 of the other planes, and those lines of
intersection points partition the new plane into P .k
1/ plane regions and
each of those plane regions corresponds to a new region in space. Hence
S.k/ D S.k
1/ C P .k
1/. Transposing the S.k
1/ term on the right
side to the left and summing yields
Xn Xn
S.n/
S.0/ D ŒS.k/
S.k
1/ D P .k
1/
kD1 kD1
Xn .n
1/n.n C 1/ n3 C 5n
D .1 C tk
1 / D n C D ;
kD1 6 6

where we have used the result in Theorem 1.8 to evaluate the final sum. Thus
S.n/ D .n3 C 5n C 6/=6 as claimed.

11.3 Corresponding triangles on three lines

Suppose two triangles ABC and A0 B 0 C 0 have their corresponding vertices
on three lines that intersect at a point, as shown in Figure 11.3 Then we have
Theorem 11.4. The three points of intersection (if they exist) of the pairs
of lines determined by corresponding pairs of sides of triangles ABC and
A0 B 0 C 0 in Figure 11.3 lie on a straight line.

B
A C
B′
A′ C′

Figure 11.3.

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 196 — #6

i i

196 CHAPTER 11. Geometry in Three Dimensions

Proof. Look at the three lines as being the edges of a triangular pyramid, or
as the legs of a tripod. Then each of the two triangles is the intersection of the
pyramid or tripod with a plane. The lines determined by corresponding pairs
of sides of the triangles lie in those two planes, and unless the two planes are
parallel, they will intersect in a straight line.

11.4 An angle-trisecting cone

In 1896 Aubry created the following three-dimensional solution to the angle-
trisection problem [Eves, 1983]. On a piece of paper draw a circle C of
radius r with a central angle  to be trisected, as illustrated in Figure 11.4a.
Take another sheet of heavy paper or light cardboard, and cut out a 120ı
sector of a circle of radius 3r, as illustrated in Figure 11.4b.

(a) (b)
C 3r 120° 3r
r

(c) (d) θ
3
a b b
a

Figure 11.4.

Now transform the circular sector into a cone, and place it on the first
sheet of paper so that the circle C is the base of the cone (see Figure 11.4c),
and mark on the lateral surface of the cone the points a and b determined by
the angle  . Now open the cone and lay it flat, connect a and b to the vertex
of the sector, and the resulting angle measures =3 (see Figure 11.4d).
This conical approach also enables us to solve another classical problem
unsolvable using only straightedge and compass: inscribing a regular hep-
tagon in a circle. Given a circle of radius r, draw on heavy paper or light
cardboard a circle of radius 8r=7 divided into eight equal parts (see Fig-
ure 11.5). Using scissors cut out the circle, discard one slice, and form the
remainder into a cone. Since the circumference of the cone is the same as
the circumference of the circle, the cone fits exactly on top of the circle and
divides it into seven equal parts, thus inscribing a regular heptagon in the
circle.

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 197 — #7

i i

11.5. The intersection of three spheres 197

2π
r
r 8r/7

Figure 11.5.

11.5 The intersection of three spheres

If three spheres intersect one another in space, what is the set of points
common to all three?
Theorem 11.5. Three spheres (with non-collinear centers) intersecting one
another have at most two points in common.
Proof. First consider just two spheres. They intersect in a circle (or a point
if they are tangent). This circle (or point) intersects the third sphere in two
(or one) points.

A beautiful consequence is the following result about circles in the plane
[Bogomolny, 2009].
Theorem 11.6. The three chords determined by the intersection of three
circles (with non-collinear centers) meet in a point. See Figure 11.6.

Figure 11.6.

Proof. Consider Figure 11.6 embedded in space, where the three circles are
the equators of three spheres cut by the plane of this page. Then the three

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 198 — #8

i i

198 CHAPTER 11. Geometry in Three Dimensions

chords are the projections onto the page of the three circles of the pair-
wise intersections of the spheres. By Theorem 11.5 the three spheres meet in
two points, both of which project onto the point of intersection of the three
chords.

The Global Positioning System (GPS)

The geometrical principle underlying the GPS is precisely Theo-
rem 11.5 about three spheres. To locate an object equipped with a GPS
unit three satellites are required. The GPS unit receives signals from
each satellite, and thus knows the location of and distance to each satel-
lite. Hence the GPS unit is located on three spheres, each one centered at
one of the satellites, and consequently at one of the two points of inter-
section of the spheres. If we know the approximate location of the GPS
unit (e.g., on the earth’s surface), we can discard one point. Appropriate
software using elementary linear algebra enables the computation of the
location of the GPS unit on the earth’s surface.

The best viewing point

Where should one stand in front of a picture drawn in three-point per-
spective to obtain the “best” view of the picture? If a cube is well drawn,
to see it as a cube we want the angles of the faces to appear as right an-
gles. Hence our eye needs to be on each of the three spheres that have
as their diameters the line segments joining pairs of vanishing points.
These three spheres intersect in two points, but only one is in front of
the picture. See Figure 11.7.

Figure 11.7.

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 199 — #9

i i

11.7. The area of a spherical triangle 199

11.6 The fourth circle

In 1916 R. A. Johnson discovered the following result [Johnson, 1916],
which has been described as one of the few recent “really pretty theorems
at the most elementary level of geometry” [Honsberger, 1976]. The elegant
proof simply requires us to look at the configuration from a three dimen-
sional point of view.
Theorem 11.7. If three circles with the same radius r are drawn through a
point P, then the other three points of intersection A, B, and C determine a
fourth circle with the same radius. See Figure 11.8a.

(a) C (b)

P
A
B

Figure 11.8.

Proof. The three points A; B; C of intersection and the centers of the circles
form a hexagon divided into three rhombi, as illustrated in Figure 11.8b. The
nine dark line segments each have length r, and drawing the three dashed
line segments of length r produces a plane projection of a cube. Hence A; B,
and C lie at a distance r from another point, and hence the fourth circle also
has radius r.

11.7 The area of a spherical triangle

In plane geometry, knowing the size of the angles of a triangle tells us noth-
ing about the area of the triangle but in spherical geometry it (along with the
radius of the sphere) tells us everything.
In spherical geometry, “lines” are arcs of great circles, which are the in-
tersections of the sphere with planes through the center of the sphere. In Fig-
ure 11.9a we see two great circles that intersect in antipodal (opposite) points
to form four spherical lunes, one of which is shaded. Angles in lunes are
dihedral angles, the angles between the two planes of the great circles. If the
dihedral angle of a lune is  (in this section we measure angles in radians)

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 200 — #10

i i

200 CHAPTER 11. Geometry in Three Dimensions

(a) (b)

Figure 11.9.

and the radius of the sphere is r, then the area L( ) of the lune is the fraction
=2 of the surface area 4 r 2 of the sphere, or L. / D 2r 2  . Each pair of
great circles produces two pairs of congruent spherical lunes.
Three great circles intersect to form eight spherical triangles, which come
in congruent pairs. In Figure 11.9b we see such a pair, shaded dark and light
gray. If the (dihedral) angles of the triangle are ˛, ˇ, and  , and its area is T ,
then we have
Theorem 11.8. T D r 2 .˛ C ˇ C 
/.

Proof. Since total area of the six lunes determined by the three great circles
is the surface area of the sphere plus four times the area of the spherical
triangle, we have 4 r 2 C 4T D 2ŒL.˛/ C L.ˇ/ C L. / D 4r 2 .˛ C ˇ C  /,
and the result follows.

The result in this theorem can be easily extended to spherical polygons.

We triangulate the spherical n-gon into n
2 spherical triangles, in a man-
ner analogous to triangulation of plane n-gons. Applying the theorem to the
resulting spherical triangles yields, for the area An of a spherical polygon,
An D r 2 ŒSn
.n
2/, where Sn denotes the sum (in radians) of the
dihedral angles of the spherical n-gon. We will use this result in the next
section.

11.8 Euler’s polyhedral formula

The first important notions in topology were acquired in the
course of the study of polyhedra.
Henri Lebesgue
In 1752 Leonhard Euler (1707–1783) announced his discovery of the re-
markable formula V
E C F D 2, where V , E, and F are, respectively,
the number of vertices, edges, and faces in a convex polyhedron. Euler and
this formula were honored in 2007 when Switzerland issued a postage stamp

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 201 — #11

i i

11.8. Euler’s polyhedral formula 201

(a) (b)

Figure 11.10.

to commemorate the 300th anniversary of Euler’s birth (Figure 11.10a). The

polyhedron illustrated on the stamp is sometimes called Dürer’s solid, since
it appears in the engraving Melencolia I (Figure 11.10b) created by Albrecht
Dürer in 1514. For Dürer’s solid, V D 12, E D 18, F D 8, and 12
18 C
8 D 2.
The formula is remarkable since in applies to any convex polyhedron, re-
gardless of its size or shape. The first rigorous proof of the formula appears
to be one published in 1794 by Adrien-Marie Legendre (1752–1833), which
we present next. Legendre’s proof is based on the spherical geometry dis-
cussed in the previous section. Many other proofs are known; see [Eppstein,
2005] for a collection of nineteen different proofs.
Theorem 11.9 (Euler’s polyhedron formula). If V, E, and F are, respec-
tively, the number of vertices, edges, and faces in a convex polyhedron, then
V
E C F D 2.

Proof. We begin by scaling and inserting the skeleton of the polyhedron

(the network of vertices and edges without the faces, as illustrated on the
stamp) into a sphere of radius 1, and project the skeleton onto the sphere by
means of a light source at the sphere’s center. This radial projection creates
a spherical polygon whose edges are arcs of great circles and with exactly
the same values of V , E, and F as the original polyhedron. Since the sphere
has radius 1 its surface area is 4, which is also the sum of the areas of the
F spherical polygon faces. Hence
X
4 D Œ.angle sum/
.number of sides/ C 2:
faces

The sum over the faces of the angle sums is 2V since the angles sur-
rounding each vertex contribute 2 to the sum.The sum of the number of

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 202 — #12

i i

202 CHAPTER 11. Geometry in Three Dimensions

sides of all the polygons is 2E since each edge is shared by two polygons,
and thus
4 D 2V
2E C 2F D 2.V
E C F /;
that is, V
E C F D 2.

It is easy to project the previous spherical configuration into the plane to

obtain a planar graph and to prove Euler’s formula by induction on the num-
ber of faces, or to draw the graph with line segments to deduce the formula
by computing angles in the configuration.

11.9 Faces and vertices in convex polyhedra

What is the domain of Euler’s polyhedral formula V
E C F D 2? Clearly
V  4 and F  4, consequently E  6. But more is true: The average
number of sides per face is 2E=F (each edge is a side of two faces), and
this average is at least 3, hence 2E  3F . Similarly computing the average
number of edges per vertex yields 2E  3V . These inequalities can rule out
particular sets of values of V , E, and F satisfying Euler’s formula and the
domain restrictions as convex polyhedra. For example, is there a convex
polyhedron with seven faces and eleven vertices? If so, then it has sixteen
edges, but then 2E D 32 < 33 D 3V , so the answer is no. See Challenge
11.3 for further inequalities for V , E, and F .
More information about the nature of the faces and vertices can be ob-
tained by introducing the numbers Fn for the number of faces that are
n-gons, and Vn for the number of vertices of degree n (the degree of a vertex
is the number of edges that meet there). Then F and V are the finite sums
P P
n3 Fn and n3 Vn , respectively. Since each edge is a side of two faces
and joins two vertices, we have
X X
2E D nFn D nVn :
n3 n3

From Challenge 11.3b we have 3F
E  6 or 6F
2E  12, and thus

X X
6 Fn
nFn  12;
n3 n3

which simplifies to
X
3F3 C 2F4 C F5  12 C .n
6/Fn :
n7

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 203 — #13

i i

11.9. Faces and vertices in convex polyhedra 203

A completely analogous calculation yields 3V3 C 2V4 C V5  12 C

P
n7 .n
6/Vn , and thus we have proved

Theorem 11.10. In any convex polyhedron, 3F3 C 2F4 C F5  12 and

3V3 C 2V4 C V5  12, so that there is always at least one triangle, quadri-
lateral, or pentagon among the faces and at least one vertex of degree 3, 4,
or 5.
A polyhedron is regular if all of its faces are congruent regular polygons
and every vertex has the same degree. Convex regular polyhedra are also
known as Platonic solids. As a consequence of Theorem 11.10, we have
Theorem 11.11. There are exactly five types of convex regular polyhedra.

Proof. We use the symbol fn; kg to denote a convex polyhedron (if it ex-
ists) all of whose faces are n-gons and whose vertices all have degree k.
From Theorem 11.10 we know that both n and k are 3, 4, or 5. Since every
edge belongs to two faces and joins two vertices, 2E D nF and 2E D kV ,
so that F D 2E=n and V D 2E=k. Substituting these into Euler’s formula
and solving for E yields E D 2nk=.2n C 2k
nk/, and consequently F D
4k=.2n C 2k
nk/ and V D 4n=.2n C 2k
nk/. Hence 2n C 2k
nk >
0, or equivalently, .n
2/.k
2/ < 4. Thus the only possible convex poly-
hedra of the form fn; kg are f3,3g, f4,3g, f3,4g, f5,3g, and f3,5g.

At this point we only know these are five potential convex polyhedra of
the form fn; kg. We have not assumed that the faces are regular, or even
equilateral, equiangular, or congruent. Thus it is rather surprising that all five
can be constructed with regular polygonal faces. Table 11.1 gives the relevant
facts about the five convex regular polyhedra, and Figure 11.11 illustrates
them.

Table 11.1. The five convex regular polyhedra

fn; kg E V F common name

f3,3g 6 4 4 tetrahedron
f4,3g 12 8 6 cube
f3,4g 12 6 8 octahedron
f5,3g 30 20 12 dodecahedron
f3,5g 30 12 20 icosahedron

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 204 — #14

i i

204 CHAPTER 11. Geometry in Three Dimensions

Figure 11.11.

Folding and unfolding polyhedra

A common method for constructing a model of a polyhedron is to cut
out a plane polygon from a sheet of paper, fold it, and tape edges to-
gether to form the polyhedron. For example, the Latin cross in Fig-
ure 11.12a can be folded and taped to form a cube. This works because
the cube can be unfolded to form the Latin cross (and a variety of other
polygons consisting of six squares). An area of current research is to
find suitable unfoldings of polyhedra to flat polygons, and the comple-
mentary problem of whether or not a given polygon can be folded and
taped to form a convex polyhedron [O’Rourke, 2009].

(a) (b)

Figure 11.12.

For example, the Latin cross can be folded into at least 23 different
convex polyhedra [Demaine and O’Rourke, 2007]. Figure 11.12b illus-
trates the folds for an irregular tetrahedron.

11.10 Why some types of faces repeat

in polyhedra
Every polyhedron must have at least a pair of faces with the same number
of sides (see Challenge 11.2). But many polyhedra have three or more faces
with the same number of sides. That is true of every polyhedron that we have

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 205 — #15

i i

11.11. Euler and Descartes à la Pólya 205

(a) (b) (c)

Figure 11.13.

seen so far, but it is not true for every polyhedron. In Figure 11.13 we see
(viewed from above) three polyhedra where faces with the same number of
sides repeat at most twice.
Again letting Fn denote the number of faces that are n-gons, we have in
(a) F3 D F4 D F5 D 2, in (b) F3 D F4 D F5 D 2 and F6 D 1, and in
(c) F3 D F4 D F5 D F6 D 2. We now prove the surprising result that these
are the only cases in which a face does not repeat at least three times.
Let F D F3 C F4 C    C Fk and assume that Fi  2 for all i . Using the
inequality 6  3F
E from Challenge 11.3b yields

12  6F
2E
Xk Xk
D6 Fi
iFi
iD3 iD3
Xk
D 3F3 C 2F4 C F5 C .6
i /Fi  3  2 C 2  2 C 2 C 0 D 12;
iD7

whence the inequalities are equalities so that F3 D F4 D F5 D 2, F6 is 0,

1, or 2, and Fi D 0 for i  7.

11.11 Euler and Descartes à la Pólya

Given a convex polyhedron with V vertices v1 ; v2 ; : : : ; vV , consider the
angular defect
i of vertex vi , defined as the difference between 2 and
the sum of the plane angles around vi . Let
denote the total angular defect
for the polyhedron, i.e.,
D
1 C
2 C  C
V . For example, for the cube

i D =2 at each of the eight vertices (see Figure 11.14a) so that
D 4;
and for the icosahedron,
i D =3 at each of the twelve vertices (see
Figure 11.14b) so that again
D 4.

(a) (b) π
π
2 3

Figure 11.14.

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 206 — #16

i i

206 CHAPTER 11. Geometry in Three Dimensions

René Descartes (1596–1650) discovered that
D 4 holds for every

convex polyhedron, a result now known as Descartes’ Angular Defect The-
orem. The relationship between Descartes’ theorem and Euler’s polyhedral
formula is given in the next theorem. Our proof is from [Pólya, 1965].
Descartes’ Angular Defect Theorem 11.12.
D 2.V
E C F /.

Proof. We let S denote the sum of all the angles in all the faces of the
polygon, and evaluate S in two ways. The sum of the angles at vertex vi
is 2

i so that computing S vertex-by-vertex yields

S D .2

1 / C .2

2 / C    C .2

V / D 2V

:

Again let Fn denote the number of faces that are n-gons, so that F D
P
n3 Fn . Since the angle sum of an n-gon is .n
2/, computing S face-
by-face yields
X X X
SD .n
2/Fn D  nFn
2 Fn D .2E
2F /:
n3 n3 n3

Thus 2V

D .2E
2F /, or
D 2.V
E C F / as claimed.

As a consequence,
D 4 if and only if V
E C F D 2, establishing

the logical equivalence of Descartes’ theorem and Euler’s formula.

Was Euler aware of Descartes’ theorem?

Descartes’ theorem appears in his work Progymnasmata de Solidorum
Elementis (Exercises on the Elements of Solids), a work that was not
published in his lifetime and, indeed, was lost until 1860 when a copy
was discovered in the papers of Gottfried Wilhelm Leibniz [Cromwell,
1997]. Since Euler was born 57 years after Descartes’ death, and died 77
years before the discovery of the Progymnasmata, he could not possibly
have known of Descartes’ result.

11.12 Squaring squares and cubing cubes

We square a square or rectangle by tiling it with smaller squares. A squared
square or rectangle is simple if it does not contain a smaller squared square or
rectangle, and compound if it does. A squared square or rectangle is perfect
if all the square tiles are of different sizes, and imperfect if they are not.
The order of a squared square or rectangle is the number of square tiles it

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 207 — #17

i i

11.12. Squaring squares and cubing cubes 207

9 10 14
8 7 4

15 18

Figure 11.15.

contains. Figure 11.15 illustrates an order 9 simple perfect squared 32  33

rectangle. The small gray square has side length 1, and the others have the
indicated side length.
It is fairly easy to create squared rectangles; squared squares are more dif-
ficult [Gardner, 1961; Honsberger, 1970]. The smallest known simple perfect
squared square is an order 21 square with side length 112.
Can we cube a cube or rectangular box in an analogous fashion?
Theorem 11.13. No cubed rectangular boxes of finite order exist.

Proof [Gardner, 1961]. Suppose a cubed box exists, and is sitting before
you on the table. The bottom face of the box is a squared rectangle, which
will contain a smallest square. Since this square cannot be on the edge of the
rectangular base, it must be the bottom face of a cube (which we denote as
cube A) that is completely surrounded by larger cubes (see Figure 11.16).

Figure 11.16.

Still smaller cubes sit on the top of cube A, forming a squared square on
its top face. In this squared square will be a smallest square, the bottom face
of cube B, the smallest cube resting on the top of cube A. Continuing, we
need a smallest cube, cube C, resting on the top face of cube B. The argument
continues, calling for an infinite collection of smaller and smaller cubes in
the box. Thus no rectangular box can be partitioned into a finite number of
cubes of different sizes.

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i
 i i

“MABK014-11” — 2010/6/25 — 13:48 — page 208 — #18

i i

208 CHAPTER 11. Geometry in Three Dimensions

11.13 Challenges
11.1 The reciprocal Pythagorean theorem (Theorem 5.1) can be extended
to right tetrahedra. In Figure 11.1 let p denote the altitude to the base

ABC . Show that .1=p/2 D .1=a/2 C .1=b/2 C .1=c/2 .
11.2 Prove that every polyhedron must have at least two faces with the same
number of sides.
11.3 Let V , E, and F denote the number of vertices, edges, and faces, re-
spectively, in a convex polyhedron. Prove that (a) 2V
F  4 and
2F
V  4, and (b) 3F
E  6 and 3V
E  6.
11.4 In a letter to Christian Goldbach (1690–1764) Euler wrote that no poly-
hedron has exactly seven edges. Prove that Euler was correct (as usual)
[Cromwell, 1997].
11.5 Prove that the number of different ways to color a Platonic solid (with
F faces and E edges) with F colors so that each face has a different
color is F Š=.2E/. Two colorings of the solid are the same if one of
them can be rotated in space so that it appears identical to the other
with corresponding faces colored alike.
11.6 Let P be a square pyramid (four equilateral triangles on a square base)
and T a regular tetrahedron with faces congruent to the triangular faces
of P . If P and T are joined by gluing one of the faces of T onto one
of the faces of P , how many faces does the resulting polyhedron have?
(Hint: it is not 7!)

a c
b
b
c
a
Figure 11.17.
11.7 An isosceles tetrahedron is one whose opposite edges are equal in
length (see Figure 11.17). Prove that the volume of an isosceles tetra-
hedron with edge lengths a, b, and c is given by
q
.a2 C b 2
c 2 /.c 2 C a2
b 2 /.b 2 C c 2
a2 /=72;
p
and that the diameter of the circumscribed sphere is .a2 C b 2 C c 2 /=2.

Downloaded from https://www.cambridge.org/core. University of Edinburgh Library, on 01 Mar 2020 at 10:55:52, subject to the Cambridge
Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.5948/UPO9781614442011.013

i i

i i