Workbook of Two Dimensional Geometries

Workbook of
two-dimensional geometries
C. Herbert Clemens and Bart Snapp
November 17, 2020

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Euclid’s postulates for plane geometry . . . . . . . . . . . . . . . . . 8
Neutral geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Sum of angles in a triangle in NG . . . . . . . . . . . . . . . . . . . . . 15
3 Rectangles and cartesian coordinates . . . . . . . . . . . . . . . . . . 23
4 Dilations and similarity in euclidean geometry . . . . . . . . . . . . 32
Dilations in EG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Similarity in EG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5 Concurrence theorems in euclidean geometry . . . . . . . . . . . . . 45
6 Central and inscribed angles in euclidean geometry . . . . . . . . . . 52
A basic fact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Central and inscribed angles . . . . . . . . . . . . . . . . . . . . . . . . 53
The Law of Sines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7 The cross-ratio and Ptolemy’s Theorem . . . . . . . . . . . . . . . . 60
The cross-ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Ptolemy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
8 Surface area and volume of the R-sphere . . . . . . . . . . . . . . . . 71
Circumference and area of circles . . . . . . . . . . . . . . . . . . . . . 71
Surface area of spheres . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Volumes of pyramids and spheres . . . . . . . . . . . . . . . . . . . . . 80
The magnification principle . . . . . . . . . . . . . . . . . . . . . . . . 82
Relation between volume and surface area of a sphere . . . . . . . . . 85
9 Spherical lunes and triangles . . . . . . . . . . . . . . . . . . . . . . 89
Spherical lunes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Spherical triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
10 Euclidean three-space as a metric space . . . . . . . . . . . . . . . . 93
11 Congruences, that is, rigid motions . . . . . . . . . . . . . . . . . . 106
12 Changing coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Bringing the North Pole of the R-sphere to (0, 0, 1) . . . . . . . . . . . 117
Formulas for euclidean lengths, angles, and areas in terms of (x, y, z)-
coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Computing length . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Computing angles . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Computing area . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
13 Rigid motions in (x, y, z)-coordinates . . . . . . . . . . . . . . . . . 129
Why use K-coordinates? . . . . . . . . . . . . . . . . . . . . . . . . . . 136
14 Lines in spherical geometry . . . . . . . . . . . . . . . . . . . . . . 138
Spherical coordinates, a shortest path from the North Pole . . . . . . 138
Shortest path between any two points . . . . . . . . . . . . . . . . . . 145
The spherical Pythagorean Theorem . . . . . . . . . . . . . . . . . . . 146
15 Lines in hyperbolic geometry . . . . . . . . . . . . . . . . . . . . . . 152
Shortest path between any two points . . . . . . . . . . . . . . . . . . 159
The hyperbolic Pythagorean Theorem . . . . . . . . . . . . . . . . . . 162
16 Central projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Central projection coordinates . . . . . . . . . . . . . . . . . . . . . . 168
Central projection dot product . . . . . . . . . . . . . . . . . . . . . . 175
When the finite is infinite and the infinite is finite . . . . . . . . . . . 180
17 Rigid motions in central projection . . . . . . . . . . . . . . . . . . 187
Rigid motions in central projection coordinates . . . . . . . . . . . . . 187
From K-rigid motions to rigid motions in central projection . . . . . . 188
18 Lines, angles, and areas in central projection . . . . . . . . . . . . . 195
Central projection preserves lines . . . . . . . . . . . . . . . . . . . . . 195
Angles in central projection coordinates . . . . . . . . . . . . . . . . . 199
Areas in central projection coordinates . . . . . . . . . . . . . . . . . . 200
Central projection unifies euclidean, spherical, and hyperbolic geometry205
19 Stereographic projection . . . . . . . . . . . . . . . . . . . . . . . . 206
Stereographic projection coordinates . . . . . . . . . . . . . . . . . . . 206
Stereographic projection dot product . . . . . . . . . . . . . . . . . . . 210
Again, when the finite is infinite and the infinite is finite . . . . . . . . 215
20 Lines, angles, and areas in stereographic projection . . . . . . . . . 223
Stereographic projection sends lines to circles (or lines) . . . . . . . . 223
Stereographic projection preserves angles . . . . . . . . . . . . . . . . 225
Areas in stereographic projection coordinates . . . . . . . . . . . . . . 227
Stereographic projection unifies euclidean, spherical, and hyperbolic
geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
21 Hyperbolic lunes and triangles . . . . . . . . . . . . . . . . . . . . . 239
Hyperbolic lunes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Hyperbolic triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
22 Connections to special relativity . . . . . . . . . . . . . . . . . . . . 246
Galilean transformations of velocity . . . . . . . . . . . . . . . . . . . 246
Hyperbolic transformations of velocity . . . . . . . . . . . . . . . . . . 253
23 The art of Escher . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
1 Introduction
1 Introduction
Seeing different geometries as a variation on a theme.
As a young mathematician I was introduced to the classic Leçons sur la Géométrie

des Espaces de Riemann, written by the great French geometer Élie Cartan.
Early in his treatise on geometries in all dimensions, the author presents the
case of two-dimensional geometries, in particular, those two-dimensional ge-
ometries that look the same at all points and in all directions. (For example, a
cylinder looks the same at each of its points but not in all directions emanating
from any one of its points, whereas a sphere looks the same at all points and
in all directions.) It turns out that there is one and only one such geometry for
each real number K, called the curvature of the geometry. The case K = 0 is
the (flat) Euclidean geometry that you learned in high school.
These twenty-five pages of Cartan’s book (Chapter VI, §i-v) so captivated me
that I have returned to them regularly throughout my career and have adapted
and taught them many times at the advanced undergraduate level. They form
the basis for this little book. To me they tell one of the most beautiful and
satisying stories in all of geometry, one which exemplifies a fundamental prin-
ciple of all great mathematics, namely that, using the tools at hand but in a
slightly novel way, the clouds part and one sees that objects and relationships
that seemed so different are in fact parts of a single elegant story!
When K > 0 it turns out that the K-geometry is the geometry of the sphere of
radius R = 1/K 1/2 that we can see as a subset of euclidean three-space R3 . But
the geometries with K < 0 are not so easy to visualize. They are the so-called
‘hyperbolic’ geometries. In fact it took mathematicians a couple thousand years
to realize that the existed at all! It turns out that the secret to understanding
all the two-dimensional geometries, including the ones with K < 0, in a unified
way is to simply rescale the third coordinate in R3 and use these ‘unusual’
coordinates to look at each two-dimensional geometry as the solution set to the
equation
K x2 + y 2 + z 2 = 1.

However the idea of changing coordinates without changing the underlying ge-
ometry described by those coordinates is a challenging one that did not come
into mathematics until a couple of centuries ago. It will require that, before
we get into the beautiful uniform study of all two-dimensional geometries, we
practice the coordinate change we are going to use, namely the rescaling of the
third coordinate in euclidean 3-space. That practice, together with a review of
some concepts from several variable calculus and linear algebra, will comprise
much of this book.
It has often been said that “mathematics is not a spectator sport.” This truism
is very much in evidence in the writing of this book. It is written so as to guide
you through the entire story, yet permit you, when possible, to construct the
5
1 Introduction
mathematical story for yourself, that is, to do some mathematics yourself rather
than just observe it done by others. This ‘doing mathematics oneself’ takes the
form of Exercises with enough help (Hints) provided so that the ‘doing’ is not
so onerous as to get in the way of the story itself.
Strong evidence has been provided by students of mathematics over many cen-
turies that such guided ‘doing’ is indispensible for understanding and retention.
In fact the very form of this book, as a loose-leaf or electronic notebook, is
intended to encourage you to write out (in correctable form) solutions to the
problems that can be inserted at the appropriate places into the text.
The second half of this book supposes familiarity with several variable calculus
and the linear algebra of matrices. In particular, it will often be useful to
consider a vector, for example V = (a, b, c), as a 1 × 3 matrix

V = a b c
with  
a
V| = b .
c
This will allow us to write the scalar product of two vectors
V • W = (a, b, c) • (d, e, f )
= ad + be + cf
as a product of matrices
 
d
V · W| = a

b c · e .
f
Furthermore we will use the notations det(A) and |A| interchangeably for the
determinant of a square matrix A.
You will also need to remember and apply the Chain Rule for differentiable
functions of several variables, written in matrix notation. Here’s the gradual
build-up to this general form of the Chain Rule using matrix notation and
matrix multiplication.
Theorem 1 (Chain Rule). We will present three different versions:
(a) Given differentiable functions y(x) and z(y) and substituting we have
z (y (x))
making z a function of x and

dz dz dy
= · .
dx dy dx
6
1 Introduction
(b) Given differentiable functions
(y1 (x) , . . . , yn (x))

z (y1 , . . . , yn )
then substituting we have
z (y1 (x) , . . . , yn (x))
making z a function of x and

   
dy1 dy1
 dx   dx 
dz ∂z ∂z  .   . 
= ... ·  ..  = ∇z · 

 .. 

dx ∂y1 ∂yn   dy   dy 
n n
dx dx
(c) Given differentiable mappings
(y1 (x1 , . . . , xm ) , . . . , yn (x1 , . . . , xm ))

(z1 (y1 , . . . , yn ) , . . . , zp (y1 , . . . , yn ))
then substituting
zk (y1 (x1 , . . . , xm ) , . . . , yn (x1 , . . . , xm ))
and fixing x1 , . . . , xi−1 and xi+1 , . . . , xm we make zk a function of xi and
∂y1 ∂y1

  
 ∂xi   ∂xi 
∂zk ∂zk ∂zk  .. 
 = ∇zk ·  ..  .
 
= ... · . .
∂xi ∂y1 ∂yn    
∂y  n
 ∂y 
n
∂xi ∂xi
(d) Putting the previous work together for all indices k and i, we have the
matrix equation
∂zk ∂zk ∂yj
= ·
∂xi ∂yj ∂xi

∂zk ∂zk
where is the p × m matrix whose (k, i)-th entry is , etc.
∂xi ∂xi
One of the Chain Rule’s important applications is the Substitution Rule for
integrals of functions of several variables.
Theorem 2 (Substitution Rule). We will present two different versions:
7
1 Introduction
(a) If f (y) is a function of y and y (x) is a function of x,

Z Z
dy
f (y) · dy = f ((y (x))) · dx.
[y(a),y(b)] [a,b] dx
(b) If f (y) is a function of y = (y1 , . . . , ym ) and the function

 
y1 (x1 , . . . , xm )
 .. 
 . 
ym (x1 , . . . , xm )
takes the region Rx to the region Ry , then

Z Z
∂yj
f (y) dy = f ((y1 (x), . . . , ym (x))) · det dx
Ry Rx ∂xi
where x = (x1 , . . . , xm ).
As a help, at some points in the text and in some of the exercises, a more
complete treatment of a particular topic can be found in one of the following
two texts:
[MJG]: Greenberg, Marvin Jay. Euclidean and Non-Euclidean Geometry: De-
velopment and History. W.H. Freeman & Co. 3rd Ed., 1994.
[DS]: Davis, H. and Snider, A.D. Introduction to Vector Analysis. Wm. C.
Brown Publishers, 7th Ed. 1994.
The corresponding topics in these texts are referenced. For example, [MJG,311]
refers to page 311 in the Greenberg book and [DS,59ff] refers to page 59 and
those pages just following page 59 in the Davis-Snider book.
Some final remarks about notation in this book. The letters ‘EG’ will always
mean euclidean (usually plane but occasionally 3-dimensional) Geometry, the
letters ‘SG’ will always mean Spherical Geometry, and the letters ‘HG’ will
always mean Hyperbolic Geometry. One further kind of geometry, which we
call Neutral Geometry, will be explained in the book and denoted by ‘NG.’
It is my hope and intention in writing this little book that you engage with
and enjoy this uniform way of understanding all two-dimensional geometries as
much as I did!
Remark 1. Special message to current or future teachers of high school ge-

ometry: Many parts of this book are especially relevant to your teaching of the
subject. Look especially closely at the treatment of congruence (rigid motion),
similarity (dilation), circles, expressing geometric properties with equations, and
geometric measurement and dimension, and compare them with the high school
geometry sections of the Common Core State Standards in Mathematics. The
latter can be found at:
8
1 Introduction
http://www.corestandards.org/Math/Content/HSG/introduction.
A useful companion course to one based on this book, one that might be called
Geometry for Teaching, would explicitly make the connections between the ma-
terial covered as in this book and what you do (or will do) in your high school
geometry classroom. The idea is not that the material we will cover will tell
you how to teach that material but rather that the treatment given here will
give you the depth and breadth of geometric understanding that will allow you
to design what you teach and bring it into your classroom in ways that those
who lack that understanding cannot.
Remark 2. This book can also be used as a bridge to a first course in Rie-
mannian geometry. It treats the case of two-dimensional geometries that are
homogeneous, that is, that look the same at all their points. But to treat these
geometries efficiently, we introduce the notion of changing coordinates for the
geometry without changing the geometry itself. It is that notion that allowed
geometers to treat surfaces and higher-dimensional smooth spaces that look dif-
ferent at different points, ones that can often not be treated at all their points
using a single set of coordinates.
9
2 Euclid’s postulates for plane geometry
2 Euclid’s postulates for plane

geometry
We begin with Euclid’s first four assumptions and explore neutral geometries.
Neutral geometry
After answering the following questions, students should be able to:
• State the axioms of neutral geoemtry.

• State the SAS congruence axiom.
• Apply SAS.
• Prove that the sum of the interior angles of a triangle in neutral geoemtry
is always less than or equal to 180◦ .
In Western civilization, the primary source of our understanding of plane geom-

etry comes from Euclid’s Elements. The treatise is of transcendent importance
well beyond geometry itself. It is among the first examples of formal logical de-
ductive reasoning. Certain fundamentals, that are called axioms, are postulated
or ‘given,’ providing the platform on which a ‘geometry’ is built. This creates a
mathematical entity modeling a physical ‘reality.’ Its properties are arrived at
by applying the laws of logic to the given fundamentals. Euclid gives five ax-
ioms for plane geometry, the first four of which seem to be ‘obvious’ reflections
of physical reality. In paraphrased form, they are:
Axiom 1 (E1). Through any point P and any other point Q, there lies a unique
line.
Axiom 2 (E2). Given any two segments AB and CD, there is a segment AE
such that B lies on AE and |CD| = |BE|.
Note: We will use both |AB| and d(A, B) to denote the distance between two
points A and B.
Axiom 3 (E3). Given a point P and any positive real number r, there exists a
(unique) circle of radius r and center P .
Said another way, if you move away from point P along a line in any direction,
you will encounter a unique point at distance r from P .
Axiom 4 (E4). All right angles are congruent.
Note: A right angle is defined as follows. Let C be the midpoint on the segment
AB. Let E be any point not equal to C. The angle ∠ACE is called a right angle
if ∠ACE is congruent to ∠ECB. [MJG,17-18]
10
Definition 1. If we are only given axioms E1–E4, we will call our geometry
neutral geometry (NG).
Definition 2. In NG, two distinct lines are called parallel if and only if they
don’t intersect.
Definition 3. We will call the set of points on a line which lie on one side of
a given point a ray. We call the given point the origin of the ray.
Definition 4. We call two rays in the plane parallel if they lie on parallel lines
and they both lie on the same side of the transversal line passing through their
origins.
Definition 5. An angle in the plane is the union of two ordered rays with
common origin and choice of one of the two connected regions into which the
union of the rays divides the plane. We often denote angles by ∠BAC where A
is the common origin and B a point along one of the rays, called the initial ray,
and C is a point along the other ray, called the final ray.
The choice of the region is either clear from the context or explicitly given.
Problem 1 Given an angle ∠BAC show by drawings the two regions into
which it divides the plane. Show how the (signed) measure of the angle depends
on:
(a) Which region you pick as the interior/exterior.

(b) Which ray is the initial ray and which is the final ray of the angle.
11
One implicit assumption of two-dimensional neutral geometry is the existence of

(a group of) rigid motions or congruences. That is, it is assumed that given any
point A and any vector ~v emanating from A and given any second point B in the
geometry and any vector w ~ emanating from B, then there is a transformation
(thought of as a matrix that we will multiply on the left) M such that:
(a) M · A = B.
(b) M · ~v is a positive scalar multiple of w.

~
(c) For all points P and Q in the geometry, M leaves the distance between
them unchanged:
d (M · P, M · Q) = d (P, Q) .
(d) For any two vectors ~u and ~v emanating from A, the angle between M · ~u
and M · ~v is the same as the angle between ~u and ~v .
Problem 2 Think back to high school days and write the congruence rules
SSS, SAS, and ASA. Translate them into the language of rigid motions.
12
Problem 3 Give an example to show that there is no universal SSA law.

Can you find a restriction that will allow for an “SSA-type” law?
13
Although it is a bit tedious to show (and we will not ask you to do it here),
using only E1–E4 you can derive the usual rules for congruent triangles (SSS,
SAS, ASA). Thus these laws hold in any neutral geometry.
Problem 4 In the diagram below, we see the intersection of BC and AE.

Suppose that |BD| = |CD| and |AD| = |ED|.
Show that triangle 4BDA and triangle 4CDE are congruent. [MJG,138]
Hint: First you should explain why ∠BDA = ∠CDE.
Hint: Next you should use one of the congruence properties above.
14
Problem 5
(a) Show in neutral geometry that, for 4ABC
B
β
α ε
A C
the exterior angle ε of the triangle at C is greater than either remote

interior angle α or β. [MJG,119]
(b) Use the previous part to show that the sum of any two angles of a triangle
is less than 180◦ .
Hint: Add line segments to the diagram above, and attempt to use the previous
problem.
15
Problem 6 Show in NG that, if two lines cut by a transversal line have a

pair of congruent alternate interior angles, then they are parallel. [MJG,117]
Hint: Suppose the assertion is false for some pair of lines. Find a triangle that
violates the conclusion of the previous problem.
16
Sum of angles in a triangle in NG

For many centuries, mathematicians attempted to prove that the sum of the
angles in a triangle was 180◦ using only E1–E4. However, the discovery of
hyperbolic geometry about two centuries ago, proved this to be a futile pursuit.
The thing that separates hyperbolic geometry from euclidean (plane) geometry
is the sum of the angles in a triangle. If you had a good geometry course in high
school, you may remember that you had to use Euclid’s fifth axiom in order to
show that the sum of the angles in a triangle was 180◦ , but more on that later.
There is one important fact about the sum of the angles in a triangle that you
can prove in NG, that is, without invoking Euclid’s fifth axiom. We will in fact
accomplish that in this section.
Problem 7 Show that the sum of the angles in 4ACE is the same as the
sum of the angles in 4ACB
Hint: Use the previous problems.
17
Problem 8 Suppose that there is a triangle 4ABC in NG for which the sum
of the angles in a triangle 4ABC is ∆◦ . Construct a new triangle, 4ACE such
that the sum of the angles in a triangle is still ∆◦ , but one of the angles of
4ACE is no more than half the size of ∠CAB.
Hint: Consider the following additions to our diagram above:
E
α1
D
α1
α2
A
Be sure to carefully explain how points D and E are constructed.
18
Problem 9 Suppose that there is a triangle 4ABC in NG for which the sum
◦
of the angles in a triangle 4ABC is (180 + x) with x > 0. Show that there
◦
is a triangle with sum of its angles still equal to (180 + x) but with one of its
angles having measure less than x. [MJG,125-127]
Hint: Use the previous problem repeatedly to construct a triangle where the sum
of the interior angles is (180 + x)◦ but one of this triangles measures is less than
∠CAB
,
2n
for any value of n.
Hint: Next, use the fact that for any positive real number M ,
M
lim = 0.
n→∞ 2n
19
On the other hand, by our previous work you cannot have a triangle with two
angles summing to more than 180◦ . Hence, from our work above, we can prove
the following theorem:
Theorem 3. In NG, the sum of the interior angles in any triangle is no greater
than 180◦ .
Proof Seeking a contradiction, suppose there exists a triangle 4ABC where

the sum of the interior angles is
◦
(180 + x)
with x > 0. We have seen that we can construct triangle 4A0 B 0 C 0 such that
the sum of the interior angles is (180 + x)◦ , and if α0 , β 0 , γ 0 are the measures of
the angles at A0 , B 0 , and C 0 respectively, we can ensure
α0 < x.
We now have two relations:
α0 + β 0 + γ 0 = 180 + x,
α0 < x.
So
β 0 + γ 0 = 180 + (x − α0 )
> 180.
As we have seen, this is impossible as two interior angles of a triangle cannot

sum to be greater than 180◦ .
Problem 10 Considering the proof above, give a sketch of the proof. This
means you should state major steps, but exclude the details.
20
Problem 11 Show the following:
(a) The sum of the interior angles in any quadrilateral is no greater than 360◦ .
(b) The sum of the interior angles of an n-gon is no greater than (n − 2)·180◦ .
Hint: Given any quadrilateral, we can break it into two triangles.
Hint: You may use the two ears theorem: For any polygon, even a non-convex
one, there are always two vertices where you can “clip off an ear”, as in the illustration.
This fact wasn’t written down and proved until 1899!
21
Problem 12 Let ABCD be a quadrilateral with ∠ABC = ∠BCD right an-

gles. (We denote polygons by naming their vertices in counterclockwise order.)
Show in NG that:
(a) • |AB| = |CD| implies that ∠BAD = ∠ADC,

• |AB| > |CD| implies that ∠BAD < ∠ADC,
• |AB| < |CD| implies that ∠BAD > ∠ADC.
(b) Use the previous part and pure logic to show that
• ∠BAD < ∠ADC implies that |AB| > |CD|,
• ∠BAD = ∠ADC implies that |AB| = |CD|,
• ∠BAD > ∠ADC implies that |AB| < |CD|.
Hint: For the first implication in a) show that quadrilateral ABCD is (self-) congru-
ent to the quadrilateral DCBA. For the second implication in the first part suppose
that |AB| > |CD|. Construct A0 on AB so that A0 B = |CD|. By a previous problem

∠BAD < ∠BA0 D.
By the (already proved) first implication
∠BA0 D = ∠CDA0 .
Finally
∠A0 DC < ∠ADC
since the segment DA0 lies between the segment DA and the segment DC. Stringing
these last three relations together, we get the second implication.
The proof of the third implication in is the same as the proof of the second implication–
just interchange A and D and interchange B and C.
For the second part, write out the contrapositive of each part and then prove that.
22
Problem 13 Explain why a) in the last Problem does not prove the existence
of rectangles in Neutral Geometry.
23
Problem 14 Summarize the results from this section. In particular, indicate

which results follow from the others.
24
3 Rectangles and cartesian coordinates

In this activity, we explore some consequences of Euclid’s fifth axiom.
Euclid’s fifth axiom, the parallel postulate

We are finally ready to introduce Euclid’s fifth and final axiom, the so-called
Parallel Postulate.
Axiom 5 (E5). Through a point not on a line there passes a unique parallel
line.
NG together with E5 is called euclidean geometry (EG). As mentioned above,

we will see later that there is another geometry called hyperbolic geometry (HG)
that satisfies all the postulates of NG but not E5. In it, the sum of the interior
angles of a triangle will always be less than 180◦ ! [MJG,134]
Problem 15 Show that if two parallel lines are cut by a transversal line, then
alternate interior angles are equal.
Hint: Draw a picture and seek a contradiction.
25
Problem 16 Show that angles ∠BAC and ∠B 0 A0 C 0 in the euclidean plane

are equal (have the same measure) if corresponding rays are parallel. Hence if
∠B 0 A0 C 0 can be rotated around A0 to obtain an angle ∠B 00 A0 C 00 with corre-
sponding rays parallel to those of ∠BAC, then ∠BAC and ∠B 0 A0 C 0 are equal.
Hint: Draw your angles and extend their legs so that you have two sets of parallel
sides.
26
Problem 17 Use the ‘uniqueness’ assertion in E5 together with what we have

established about neutral geometry to show that in EG the sum of the interior
angles of any triangle is 180◦ .
Hint: Make two parallel lines, the first being the base of a given triangle and the
second being the unique parallel line that passes through the vertex opposite to the
base.
27
Problem 18 Show that in EG the sum of the interior angles of a quadrilateral

is 360◦ .
28
Problem 19 Now we will show in EG that given any positive real numbers
a and b, there exist rectangles with adjacent side of lengths a and b. Your task
is to fill-in the details of the proof below.
Start by constructing lines `1 and `2 with `1 perpendicular to `2 at point A.
On `1 add point B so that |AB| = a. Next construct `3 perpendicular to `1
through B. On `3 add point C such that |BC| = b. Finally add `4 through C
so that `4 is perpendicular to `3 .
C
`4
`1
A a B
`2 `3
Explain why `3 is parallel to `2 .

Explain why `4 intersects `2 .
Call the intersection of `2 and `4 point D and label the two remaining sides a0
and b0 . Explain why ∠ADC = 90◦ .
Explain why `1 is parallel to `4 .
Finally, add a segment to our figure and use a triangle congruence theorem to
explain why a = a0 and b = b0 .
29
Problem 20 Show that there is a cartesian coordinate system on EG. This

means you must show that there is a bijection between the points in EG and
elements of R2 , the set of pairs of real numbers.
30
The distance formula in EG

It is the existence of a cartesian coordinate system in EG that allows us to
define distance between points
p
d ((a1 , b1 ), (a2 , b2 )) = (a2 − a1 )2 + (b2 − b1 )2
and so gives rigorous mathematical meaning to a concept that the ancient Greeks
were never able to describe precisely, namely the similarity of figures in EG. For
that we will require the notion of a dilation or magnification in EG. We need a
cartesian coordinate system to describe dilation precisely, a reality backed up by
the fact that similarities do not exist in HG or SG. (Try drawing two triangles
that are similar but not congruent on a perfectly spherical balloon!)
Problem 21 State and prove the Pythagorean theorem in EG.
Hint: In the cartesian plane, construct a square with vertices
(0, 0) , (a + b, 0) , (0, a + b) , (a + b, a + b) .
Inside that square, construct the square with vertices
(a, 0) , (a + b, a) , (b, a + b) , (0, b) .
31
Problem 22 Use the Pythagorean theorem to justify the euclidean distance

formula, p
d ((a1 , b1 ), (a2 , b2 )) = (a2 − a1 )2 + (b2 − b1 )2 .
32

33
4 Dilations and similarity in euclidean geometry
4 Dilations and similarity in euclidean

geometry
Now we explore some properties that are somewhat unique to euclidean geometry.
Dilations in EG
We start this section off by asking you to prove a basic result.
Problem 24 Let X be a set and f : X → X. Prove that f −1 exists if and

only if f is one-to-one and onto.
34
Definition 6. A dilation is a transformation of the cartesian plane to itself

that:
(a) Is one-to-one and onto.

(b) Fixes one point called the center of the dilation.
(c) Takes each line through the center of the dilation to itself.
(d) Multiplies all distances by a fixed positive real number called the magni-
fication factor of the dilation.
Definition 7. Given a point (x0 , y0 ) in the plane and a positive real number r,
we define a mapping D with center (x0 , y0 ) and magnification factor r by the
formula
D (x, y) = (x0 , y0 ) + r (x − x0 , y − y0 ) .

We will also denote the output D (x, y) of the dilation as x, y .
Problem 25 Using cartesian coordinates for the plane, show that the map-
ping D defined above is a dilation with magnification factor r and center (x0 , y0 ).
Hint: Use the parametric formula for a line:
`(t) = point + t · vector
35
Problem 26 Show that a dilation by a factor of r takes any vector to r times

itself.
Hint: View the vector as the difference between two points.
36
Problem 27 Show that a dilation takes a line to a line parallel (or equal) to
itself.
Hint: Use the parametric formula for a line.
37
Problem 28 Show that a dilation of the plane preserves angles.
38
Problem 29 Show that the inverse mapping of a dilation is again a dilation

with the same center but with magnification factor r−1 .
Hint: Solve for (x, y) in terms of (x, y).
39
Problem 30 Show using several-variable calculus that a dilation with mag-

nification factor r multiplies all areas by a factor of r2 .
Hint: Recall that if R is a region in the plane,

ZZ ZZ
f (x, y) dxdy = f (D(x, y)) |det JD (x, y)| dxdy,
D(R) R
where  
∂Dx ∂Dy
 ∂x ∂x 
JD (x, y) = 
 ∂Dx

∂Dy 
∂y ∂y
and Dx and Dy are the components of D(x, y).
40
Problem 31 Give an explanation that a middle grades student would under-

stand that a dilation with magnification factor r multiplies all areas by a factor
of r2 .
Hint: Break the region into rectangles.
41
Similarity in EG
Definition 8. Two triangles are similar if there is a dilation of the plane that
takes one to a triangle which is congruent to the other. We write
4ABC ∼ 4A0 B 0 C 0
to denote that these two triangles are similar (where the order of the vertices
tells us which vertices correspond).
Problem 32
(a) Show that, if two triangles are similar, then corresponding sides are pro-
portional with the same constant of proportionality.
(b) Show that, if corresponding sides of two triangles are proportional with
the same constant of proportionality, then the two triangles are similar.
Hint: For the first part, you have to start from the supposition that the two triangles
satisfy our definition of similar triangles.
Hint: For the second part, you have to start from the supposition that corresponding
sides of the two triangles are proportional and use SSS to show that there is a dilation
of 4ABC is congruent to 4A0 B 0 C 0 .
42
Problem 33
(a) Show that, if two triangles are similar, then corresponding angles are equal.
(b) Show that, if corresponding angles of two triangles are equal, then the two
triangles are similar.
Hint: For the first part, you have to start from the supposition that the two triangles
satisfy our definition of similar triangles.
Hint: For the second part, you have to start from the supposition that corresponding
angles of the two triangles are equal, then use a dilation with r = |A0 B 0 |/|AB| and
ASA to show that the dilation of one triangle is congruent to the other.
43
Problem 34 Show that two triangles are similar if corresponding sides are
parallel.
Hint: Use the fact that angles are equal if corresponding rays are parallel.
44
Problem 35 Show that two triangles are similar if corresponding sides are
perpendicular.
Hint: Just think about one angle at a time.
45

46
5 Concurrence theorems in euclidean geometry
5 Concurrence theorems in euclidean

geometry
Now we explore when three lines meet at a point.
Let’s look at a some concurrence theorems. Concurrence theorems deal with

situations when three or more lines (or curves) pass through the same point.
Problem 37 Denote the measure or area of a triangle 4ABC as |4ABC|.

Show that, in the diagram below,
|AF | |4AF C| |4AF X|

= = .
|F B| |4CF B| |4XF B|
E
X D
A F B
Hint: Mark the height of the relevant triangles.
47
Problem 38 Use the previous problem to show using pure algebra that
|AF | |4AXC|
= .
|F B| |4CXB|
48
Now we will present, and you will prove, Ceva’s Theorem.

Historical note: Giovanni Ceva (CHEH-vah) was an Italian mathematician
who proved this theorem in the 1600s. In fact, it was previously proved in the
1000s by Yusuf al-Mu’taman ibn-Hūd, the Muslim king of Zaragoza in Spain.
But we still call it Ceva’s theorem.
Theorem 4 (Ceva’s Theorem). Three segments AD, BE, and CF
E
X D
A F B
are concurrent if and only if
|AF | |BD| |CE|

· · = 1.
|F B| |DC| |EA|
Problem 39 Prove the forward direction of Ceva’s theorem: For three con-
current segments AD, BE and CF
E
X D
A F B
show that
|AF | |BD| |CE|
· · = 1.
|F B| |DC| |EA|
Hint: Use the previous problem repeatedly.
49
Problem 40 Prove the reverse direction of Ceva’s Theorem: If

|AF | |BD| |CE|
· · =1
|F B| |DC| |EA|
then the lines AD, BE, and CF pass through a common point.
Hint: Suppose that they do not pass through a common point.
E D
A F B
|AF |
Hint: Notice that if, for example, F moves along the segment AB, then is
|F B|
a strictly increasing function of |AF |. Now use a previous problem to determine a
position F 0 for F along the segment AB at which
|AF 0 | |BD| |CE|

· · = 1.
|F 0 B| |DC| |EA|
50
Problem 41 A median of a triangle is a line segment from a vertex to the

midpoint of the opposite side. Show that the medians of any triangle meet in a
common point.
Hint: Use Ceva’s Theorem.
51
Definition 9. An altitude of a triangle is a line segment originating at a vertex

of the triangle that meets the line containing the opposite side at a right angle.
Problem 42 Use Ceva’s theorem to show that the three lines containing al-
titudes of a triangle are concurrent.
A F B
Hint: Use all three similarities of the form 4CEB ∼ 4CDA and then apply Ceva’s
theorem.
52

53
6 Central and inscribed angles in euclidean geometry
6 Central and inscribed angles in

euclidean geometry
In this activity, we study central and inscribed angles in euclidean geometry.
A basic fact
Now we prove a basic fact about isosceles triangles.
Problem 44 Prove that given an isosceles triangle, angles opposite the con-
gruent sides are also congruent.
Hint: Use a congruence theorem.
54
Central and inscribed angles

The next topic in euclidean geometry concerns central and inscribed angles in
circles. We include this partly for its own interest, and partly because the
properties we visit here will be useful later on.
Problem 45 On the circle with center O below,
X O
B
show that
∠BXA = (1/2)(∠BOA).
Hint: 4OAX is isosceles.
55
X O
B
show that
∠BXA = (1/2)(∠BOA).
Hint: Draw the diameter through O and X and add.
56
B
X
O
show that
∠BXA = (1/2)(∠BOA).
Hint: Draw the diameter through O and X and subtract.
57
We can summarize the results of the last three problems into the following
theorem.
Theorem 5. The measure of any angle inscribed in a circle is one-half of the

measure of the corresponding central angle.
The Law of Sines
Recall the (extended) Law of Sines:

Theorem 6 (Law of Sines). Given a triangle with angles α, β, and γ, with side
a opposite α, side b opposite β, and side c opposite γ, we have
a b c
= = = 2R
sin(α) sin(β) sin(γ)
where R is the radius of a circle circumscribing the triangle.
To prove this theorem, we must check several things.

Definition 10. The midset of two given points is the set of points equidistant
from the given points.
Problem 48 Show that in EG, the midset of two points is given by the
perpendicular bisector of the segment connecting the two given points.
58
Problem 49 Explain why every triangle can be circumscribed by a circle.
Hint: Consider the intersection of the perpendicular bisectors of the sides.
59
Problem 50 Prove the Law of Sines.
Hint: Start by circumscribing the triangle.
Hint: Next, construct two (congruent) right triangles using the center of circle and
two vertices of the triangle.
Hint: Finally, use the definition of sine.
60

61
7 The cross-ratio and Ptolemy’s Theorem
7 The cross-ratio and Ptolemy’s

Theorem
In this activity, we do some basic projective geometry and learn Ptolemy’s The-
orem.
In this section, we will study some basic ideas of projective geometry. The
main topic, one closely related to the notion of perspective in painting, is called
the cross-ratio. In particular, we will use the cross-ratio to prove a famous
mathematical relationship, Ptolemy’s Theorem. We will then explore several
corollaries of Ptolemy’s Theorem.
The cross-ratio
Problem 52 In the diagram
α β
A
C
B
show that
|AB| sin α sin (∠AOB)
= = .
|CB| sin β sin (∠COB)
62
Problem 53 Note that if, in the above figure, B moves along the circle to the
other side of C, it is still true that
|AB| sin (∠AOB)

=
|CB| sin (∠COB) .
Carefully draw the diagram for this situation. You do not need to give the proof
again.
63
Problem 54 In the diagram
α β
A
C
B
γ δ
A0 B0 C0
show that
|A0 B 0 | sin α sin δ sin (∠A0 OB 0 ) sin (∠B 0 C 0 O)
0 0
= · = · .
|C B | sin β sin γ sin (∠C 0 OB 0 ) sin (∠B 0 A0 O)
64
Problem 55 Note that if, in the above figure, B moves along the circle to the
other side of C, it is still true that
|A0 B 0 | sin α sin δ sin (∠A0 OB 0 ) sin (∠B 0 C 0 O)

0 0
= · = · .
|C B | sin β sin γ sin (∠C 0 OB 0 ) sin (∠B 0 A0 O)
Carefully draw the diagram for this situation.
65
These problems allow us to define the cross-ratio of four points on a circle.

Definition 11. For a sequence of four (ordered) points A, B, C, and D on a
circle, we define
|AB| |CD|
(A : B : C : D) = ·
|CB| |DA|
which we call the cross-ratio of the ordered sequence of four points on
a circle. Similarly for a sequence of four (ordered) points A0 , B 0 , C 0 , and D0
on a line, we define
|A0 B 0 | |C 0 D0 |
(A0 : B 0 : C 0 : D0 ) = ·
|C 0 B 0 | |D0 A0 |
which we call the cross-ratio of the ordered sequence of the four points
on a line.
Problem 56 Show that, in the figure
A D
C
B
A0 B0 C0 D0
we have the equality
(A : B : C : D) = (A0 : B 0 : C 0 : D0 ) .
Hint: Use the previous problems.
66
We say that “Cross-ratio is invariant under stereographic projection.”
Ptolemy’s Theorem
You can easily convince yourself with a few examples that, given four non-
collinear points A, B, C and D in the plane, it is not always true that there is
a circle that passes through all four. A famous theorem of classical Euclidean
geometry gives a condition when there is a circle that passes through all four.
Theorem 7 (Ptolemy). If the ordered sequence of points A, B, C and D lies on
a circle,
A D
C
B
then
|AC| · |BD| = |AD| · |BC| + |AB| · |CD| .
That is, the product of the diagonals of the quadrilateral ABCD is the sum of
the products of pairs of opposite sides.
Proof We need to check that
|AC| · |BD| = |AB| · |CD| + |BC| · |DA|
or, what is the same, we need to check that

|AC| · |BD| |AB| · |CD|
= + 1.
|DA| · |BC| |DA| · |BC|
That is, we need to check that
(A : C : B : D) = (A : B : C : D) + 1.
67
But by a previous problem this is the same as checking that
(A0 : C 0 : B 0 : D0 ) = (A0 : B 0 : C 0 : D0 ) + 1
for the projection of the four points onto a line from a point O on the circle.
But that is the same thing as showing that
|A0 C 0 | · |B 0 D0 | |A0 B 0 | · |C 0 D0 |
0 0 0 0
= +1
|D A | · |B C | |D0 A0 | · |B 0 C 0 |
which is the same thing as showing that
|A0 C 0 | · |B 0 D0 | = |D0 A0 | · |B 0 C 0 | + |A0 B 0 | · |C 0 D0 | . (∗)
In order to finish the proof, we need to show (∗).
Problem 57 Check the equality (∗) using high-school algebra.
68
Problem 58 Use Ptolemy’s Theorem to give another proof of the Pythagorean

Theorem.
Hint: Let the four points in Ptolemy’s Theorem form a rectangle.
69
Problem 59 Prove the addition formula for sine:
sin(α + β) = sin(α) cos(β) + cos(α) sin(β)
Hint: Consider this setup,
α
B D
β
C
where BD is a unit diameter for the circle.
70
Problem 60 Prove the subtraction formula for sine:
sin(α − β) = sin(α) cos(β) − cos(α) sin(β)
Hint: Consider this setup,
A D
α β
B C
where BC is a unit diameter for the circle.
71
Both of the above formulas were crucial to Ptolemy. Ptolemy’s magnum opus
was the Almagest, a book that contained all the known information about the
stars and their motion, including information on how a potential reader could
reproduce its observations and conclusions. A key technical hurdle that needed
to be resolved was the computation of sines and cosines of lots of different angles.
The two formulas above were the key to this endeavor.

72
8 Surface area and volume of the R-sphere
8 Surface area and volume of the

R-sphere
In this activity we begin to explore spherical geometry.
We are now going to study the geometry of the R-sphere in euclidean 3-space.
This is the sphere of radius R in normal 3-space. We will show why there is a
factor of 1/3 in many formulas for volumes in 3-dimensional euclidean geometry,
just like there is a factor of 1/2 in many formulas for areas in 2-dimensional
euclidean geometry.
Circumference and area of circles

We begin with perhaps the most basic fact of all about circles in euclidean
geometry. We will use calculus to compute the perimeter of a unit circle.
Problem 62 Consider a unit circle with n equal inscribed triangles meeting

at the center. Here we have shown the nth triangle:
bn
θn
(0, 0) (1, 0)
Find θn and bn in terms of n.
73
Problem 63 Explain what the limit
lim (n · bn )
n→∞
computes and compute the limit.
74
Problem 64 The circle of radius 1 has circumference 2π. Use this to reason
that the circle of radius 1 has area π.
Hint: Approximate a rectangle by rearranging the slices in the picture. Compute

the area of the “rectangle.”
75
Not content to give just one explanation, let’s give another. The techniques we
use in this explanation will be used in later problems.
Problem 65 Again, consider the circle of radius 1 with circumference 2π.

Suppose we inscribe n triangles (with one vertex at the center of the circle),
here we have shown the nth triangle:
bn
hn
In each triangle, label bn and hn , where bn is conceptualized as a “base” and

hn is a “height.” Compute
lim hn
n→∞
and explain your reasoning.
76
Problem 66 Consider the limit

n · bn · hn
lim .
n→∞ 2
Explain what this limit is computing, and compute this limit.
77
Surface area of spheres

To compute the surface area of the sphere of radius R in 3-dimensional euclidean
space, we will show that its surface area is equal to the surface area of something
we can lay out flat. The argument for this goes way back to the great physicist
and mathematician, Archimedes of Alexandria, in the 2nd century B.C. To
follow his argument, we have to begin by computing the area of a ‘lamp shade’
or ‘collar.’ We think of a circular collar as in the figure below
as approximated by an arrangement of trapezoids. To achieve this, we approxi-

mate the bottom circle of the collar by an inscribed regular n-gon whose vertices
are the points of intersection with the slant lines in the figure. Similarly ap-
proximate the top circle by an inscribed regular n-gon positioned directly above
the bottom one, again with vertices given by the points of intersection with the
slant lines. Complete a side of the bottom n-gon and the side of the top n-gon
directly above it to a trapezoid by adjoining the two slant lines in the figure
that connect endpoints. Let
• bn denote the length of a side of the bottom regular n-gon and let
• tn denote the length of a side of the top n-gon.
Then the trapezoid has area

bn + t n
· hn
2
78
where hn is the vertical height of the trapezoid. The collar is approximated by

the union of these n trapezoids, so the area of the collar is approximated by the
sum of the areas of the n congruent trapezoids, namely

bn + t n n · bn + n · tn
n· · hn = · hn .
2 2
As n goes to infinity, the area of the approximation approaches the area of the
collar. But if
• cb is the circumference of the bottom circle and
• ct is the circumference of the top circle and

• s is the slant height of the collar as shown in the above figure,
then
lim n · bn = cb
n→∞
lim n · tn = ct
n→∞
lim hn = s.
n→∞
Problem 67 Let
• rb be the radius of the circle defining the base of the collar,
• rt be the radius of the circle defining the top of the collar,

• ra be the average of rb and rt ,
• s be the slant height of the collar.
Explain why the area of the collar is
π · (rb + rt ) · s.
(This can also be written as 2π · ra · s.)
79
Theorem 8. The surface area of the sphere of radius R is the same as the
surface area of the label of the smallest can into which the sphere will fit.
Namely the surface area of the sphere of radius R is
2πR · 2R = 4πR2 .
Problem 68 Show why the above theorem is true.
Hint: Slice the picture above into n horizontal slices. Approximate the piece of the
surface of the sphere between the ith pair of successive slices by a collar Ci . Let a (Ci )
denote the area of Ci , let ri denote its average radius and Explain why the surface
area of the sphere is
Xn
lim a(Ci ).
n→∞
i=1
Explain further how we can conclude that the area of the sphere is given by
n
X
lim 2π · ri · si .
n→∞
i=1
80
Hint: Let hi denote the vertical height of the label on the can between the ith pair
of successive slices. Explain why the area of the label is exactly
n
X
2πR · hi .
i=1
Hint: Explain why the relationship between each ri , si and hi is given by the picture
below.
si ri
hi
Now use facts about similar triangles to explain why
ri · si = hi · R.
81
Volumes of pyramids and spheres

We will start by looking at the volumes of pyramids.
Problem 69 Show that an r × r × r cube can be constructed from three equal

pyramids with an r × r square base. Conclude that the volume of each pyramid
is 1/3 the volume of the cube, namely
r3
.
3
Hint: Suppose the cube had a hollow interior and infinitely thin faces. Put your
(infinitely tiny) eye at one vertex of the cube and look inside. How many faces of the
cube can you see?
82
We next want to show why all pyramids with an r × r square base and vertical
altitude r have the same volume. That is, if we put the vertex of the pyra-
mid anywhere in a plane parallel to the base and at distance r, the volume is
unchanged.
This fact is an example of Cavalieri’s Principle: Shearing a figure parallel to
a fixed direction does not change the n-dimensional measure of an object in
euclidean n-space. Think of a stack of (very thin) books. We’ll give a proof in
euclidean 3-space, and we will use the coordinates (b
x, yb, zb) to denote this space.
Problem 70 Show that Cavalieri’s Principle is true for the pyramid using
multivariable calculus.
Hint: Put the base of the pyramid P so that its vertices are (0, 0), (r, 0), (0, r) and
(r, r) in 3-dimensional euclidean space. Consider the transformation
  
1 0 a x
T (x, y, z) = 0 1 b  y 
0 0 1 z
and use the 3-dimensional change of variables formula,

ZZZ ZZZ
f (x, y, z)dxdydz = f (T (x, y, z))|det DT (x, y, z)|dxdydz,
T (P ) P
where
∂T ∂T ∂T
DT (x, y, z) = .
∂x ∂y ∂z
(Each partial derivative is itself a vector since T (x, y, z) has three components.)
83
The magnification principle

The magnification principle: If an object in Euclidean n-space is magnified
by factors of r1 , . . . , rn , its n-dimensional measure is multiplied by r1 · · · rn .
Problem 71 Use this magnification principle to justify the volume formula
(1/3)B · h
for any pyramid with rectangular base of area B and vertical altitude h.
84
Problem 72 Prove the magnification principle using multivariable calculus.
Hint: Consider the transformation

 
r1 0 ... 0
0 r2 ... 0
J(~x) =  . ..  ~
x.
 
.. ..
 .. . . .
0 0 ... rn
and use the change of variables formula.
85
Now suppose we have any pyramid
with any shaped base of area B and any vertical altitude h. Approximate the
base as close as you want (i.e. ε-close) by tiling its interior with rectangles. Let
t denote the sum of the areas of these rectangles. Approximate the base as close
as you want (i.e. ε-close) by covering it entirely with rectangles. Let T denote
the sum of the areas of these rectangles. Why is the area B of the base of the
pyramid caught between B − ε and B + ε?
Problem 73 Show that the volume V of the pyramid is caught between (1/3)·
t · h and (1/3) · T · h. Argue that, given any positive real number ε, however
small, the volume V of the pyramid is caught between (1/3) · (B − ε) · h and
(1/3) · (B + ε) · h. Finally, show that
V = (1/3) · B · h.
86
Relation between volume and surface area of a

sphere
Think of a disco-ball. Its surface is approximately a sphere, but that surface is
made up of tiny flat mirrors.
Problem 74
(a) Explain why you can think of the disco-ball as being made up of pyramids,
with each pyramid having base one of the tiny mirrors and vertex at the
interior point O at the center of the disco-ball.
(b) Argue that the volume of the disco-ball is (1/3) times the distance h from
a mirror to O times the sum of the areas of all the mirrors.
87
Problem 75 Argue that, as the mirrors are made to be smaller and smaller,
(a) the sum of the areas of the mirrors approaches the surface area of a sphere,
(b) the distance h approaches the radius R of that sphere,
(c) the volume of the disco-ball approaches the volume of the sphere.
Conclude that, for a sphere of radius R in euclidean 3-space, the relation between
the volume V of the sphere and the surface area S of the sphere is given by the
formula
R·S
V = .
3
88
Problem 76 Use the work above to deduce a formula for the volume of a
sphere.
89

90
9 Spherical lunes and triangles

In this activity we explore the areas of lunes and triangles on the sphere. For
now we will assume that the equivalent of “lines” on the sphere are great circles:
circles that cut the sphere exactly in half. Later we will see an explanation of
why this is the right notion.
Spherical lunes
In the picture we have shaded in an ‘α-lune’ on the R-sphere in euclidean 3-
space.
The lune has two vertices. They are at opposite (antipodal) points on the R-
sphere, that is, the line in euclidean 3-space that joins the two vertices runs
through the center of the sphere. The angle at a vertex of the lune is α radians.
Problem 78 Explain why the area of the α-lune is 2α · R2 .
91
Spherical triangles
If a triangle on the sphere of radius R has interior angles with radian measures
α, β, and γ, it can be covered three times by lunes as shown in the figure below.
Notice that each lune has one vertex at a vertex of the triangle and angle equal
to that interior angle of the triangle. The other vertices of each lune are vertices
of an ‘opposite’ triangle that has the same area as the given one since it is just
the image of the given one under the rigid motion
    
−x −1 0 0 x
−y  =  0 −1 0  y  .
−z 0 0 −1 z
The three lunes cover the triangle three times. The three opposite lunes cover
the opposite triangle three times. If you take all six lunes together, they cover
each of the two triangles three times and everything else exactly once.
Problem 79 Show that the area of the spherical triangle is given by the
formula
R2 ((α + β + γ) − π) .
Hint: Use the previous problem.
92
Problem 80 Use induction to derive a formula for the area of any spherical
n-gon.
Hint: Divide the spherical n-gon into spherical triangles.
93

94
10 Euclidean three-space as a metric space
10 Euclidean three-space as a metric

space
In this activity we will work in three dimensional space and see the importance
of the dot product to geometry.
Points and vectors in euclidean 3-space

Up to this point in this book, we have studied two and three-dimensional shapes,
the objects found in two (and three) dimensional geometry. Now we will study
geometries as objects unto themselves. In particular, we will study plane, spher-
ical, and hyperbolic geometry in two-dimensions. What does it mean to study
a “whole geometry?” Well, geometry is the study of
• length,
• angle, and
• area.
Hence, if we can explain how to compute lengths, angles, and areas, we under-
stand something about the geometry as a whole. Moreover, the congruences
found in geometry are simply the transformations that preserve length, angle,
and area: the rigid motions. Thus we now seek to understand how to com-
pute length, angle, and area, along with a description of the congruences, for
euclidean, spherical, and hyperbolic geometry. Our plan of attack is as follows:
we will visualize each of the geometries as different surfaces lying in a common
three-dimensional space. This will allow us to use the techniques from calculus
and linear algebra to describe euclidean, spherical, and hyperbolic geometry in a
unified way. We start by introducing the concepts we need in the more familiar
setting in ordinary 3-dimensional euclidean space:
R3 = {(b b, yb, zb ∈ R} .
x, yb, zb) : x
We reserve the notation (x, y, z) for some new coordinates that we will put on
the ‘same’ objects later in this course. In euclidean space, there is a standard
way to measure distance between two points
X
b1 = (b
x1 , yb1 , zb1 )
X
b2 = (b
x2 , yb2 , zb2 ),
namely p
d(X
b1 , X
b2 ) = (b b1 )2 + (b
x2 − x y2 − yb1 )2 + (b
z2 − zb1 )2 .
95
When you see two point X b1 and Xb2 in what follows, the ‘hats’ mean that
distance between points is measured by the formula above. One more thing, in
euclidean three-space it will be important throughout to make the distinction
between points and vectors: Although each will be represented by a triple of
real numbers we will use
X
b = (b
x, yb, zb)
to denote points, that is, position in euclidean 3-space, and
v
b = (b
a, bb, b
c)
to denote vectors, that is, displacement by which we mean the amount and
direction a given point is being moved. So vectors always indicate motion from
an explicit (or implicit) point of reference.
The dot product determines length, angle, and area

There are various operations we can perform on one or more vectors when we
think of them as emanating from the same point in euclidean 3-space. The first
is the dot product of two vectors.
Definition 12. The dot product of two vectors
v
b1 = (b
a1 , bb1 , b
c1 ),
v
b2 = (b
a2 , b2 , b
b c2 ),
emanating from the same point in 3-dimensional euclidean space is defined as

the real number given by the formula
a1 b
b a2 + bb1bb2 + b
c1 b
c2
or in matrix notation as  
h a2
i b
ba1 bb1 c1 bb2  .
b
c2
b
It is also denoted as
b1 • v
v b2
or in matrix notation as
b2| .
b1 · v
v
Problem 82 Give the formula for the length |b v| of a vector v

b = (b
a, bb, b
c) in
3-dimensional euclidean space in terms of dot product.
96
Problem 83 Prove that
b2 + yb2 + zb2 = R2
x
is a sphere in euclidean three-space.
Hint: Remember, a sphere in euclidean three-space is the set of points equidistant

from the origin.
97
Lemma 1 (Law of Cosines). The (smaller) angle θ between two vectors v

b1 and
v
b2 emanating from O = (0, 0, 0) satisfies the relation
2 2 2
|b
v2 − v
b1 | = |b
v1 | + |b
v2 | − 2 |b
v1 | · |b
v2 | · cos θ.
Now we will prove this lemma. Your task is to fill in the details of the proof
below.
Problem 84 Start by noting that without loss of generality we can assume

that |b
v1 | ≤ |b
v2 |. Consider the triangle OP1 P2 , where P1 and P2 are the end-
points of v1 and v2 , respectively. Let P be the point on OP2 so that the segment
between P1 and P is perpendicular to OP2 .
(a) Illustrate the diagram described above.

(b) Now explain how the Pythagorean theorem gives
|P1 P2 |2 − |P2 P |2 = |P P1 |2
= |OP1 |2 − |OP |2 .
(c) Explain each of the following lines:
|P1 P2 |2 = |OP1 |2 + |P2 P |2 − |OP |2

(1)
2
= |OP1 | + (|P2 P | + |OP |) (|P2 P | − |OP |) (2)
2
= |OP1 | + |OP2 | (|P2 P | − |OP |) (3)
2
= |OP1 | + |OP2 | (|OP2 | − 2|OP |) . (4)
(d) Now note that

|OP | = |OP1 | · cos θ.
Explain how this completes the proof.
98
Theorem 9. The angle θ between two vectors v b1 and v

b2 emanating from the
same point in euclidean 3-space satisfies the relation
b1 • v
v b2 = |b
v1 | · |b
v2 | · cos θ. (5)
Problem 85 Prove this theorem.
Hint: Use the law of cosines together with algebraic properties of the dot product.
99
The significance of the theorem above is that the measure of angles between
vectors depends only on the definition of the dot product.
Corollary 1. The formula for the angle θ between two vectors v b1 = (b

a1 , bb1 , b
c1 )
and v
b2 = (ba2 , b2 , b
b c2 ) in 3-dimensional euclidean space depends only on the dot
products of the two vectors with themselves and with each other. Namely,

b1 • v
v b2
θ = arccos .
|b
v1 | · |b
v2 |
Similarly, the formula for the area of the parallelogram determined by two vec-
tors v
b1 and vb2 depends only on the dot products of the two vectors with them-
selves and with each other. You will see this from the next two problems.
Problem 86 Show that the area of the parallelogram determined by vb1 and
v
b2 emanating from the same point in euclidean 3-space is given by
|b
v1 | · |b
v2 | · sin θ.
100
Problem 87 Show that the area of the parallelogram determined by v b1 and

v
b2 emanating from the same point in euclidean 3-space is also given by
s
vb1 • v b2 • v
b1 v b1
det .
vb1 • v b2 • v
b2 v b2
Hint: Start from the square of |b v2 | · sin θ, substitute (1 − cos2 θ) for sin2 θ, and
v1 | · |b
use the theorem above.
101
Problem 88 Show that we have the following equality:

s s
vb1 • v b2 • v
b1 v v
b1 |
b2|
b1
det = det · vb1 v .
vb1 • v b2 • v
b2 v b2 v
b2
102
Again, the significance of the previous problems is that, to compute areas, we

only need to know how to compute dot products. Hence, it is the definition of
the dot product of the vectors completely determines the calculation of the area
of the parallelogram they generate.
Curves in euclidean 3-space and vectors tangent to them

Definition 13. A smooth curve in 3-dimensional euclidean space is
given by a differentiable mapping
b : [b, e] → R3
γ
t 7→ (b
x(t), yb(t), zb(t))
from an interval [b, e] on the real line whose tangent vector

dbγ db
x dby dbz
= , ,
dt dt dt dt
is not the zero vector for any t in [b, e]. We will sometimes use column vector
notation if it looks better:  
x
b(t)
b(t) =  yb(t) 
γ
zb(t).
Problem 89
(a) Give two examples of smooth curves,

γ
b1 (s) = (b
x1 (s), yb1 (s), zb1 (s))
γ
b2 (t) = (b
x2 (t), yb2 (t), zb2 (t))
where
• neither γ
b1 nor γ
b2 are a straight line,
• the two curves γb1 and γ b2 pass through a common point and go in
different tangent directions at that point.
• None of the components of γ b1 and γb2 are constant functions.
(b) Compute the tangent vectors of each of the two curves at each of their
points.
(c) For the two curves you defined in a), what are the coordinates of the point
in euclidean 3-space at which the two curves intersect?
(d) Use the dot product formula to compute the angle θ between (the tangent
vectors to) your two example curves in a) at the point at which the curves
intersect.
103
There are two kinds of vectors that naturally arise from curves: the displacement
between times t1 and t2 , given by
v b(t2 ) − γ
b=γ x(t2 ) − x
b(t1 ) = (b b(t1 ), yb(t2 ) − yb(t1 ), zb(t2 ) − zb(t1 )),
and the instantaneous velocity of a point moving along a curve, given by

db
γ (t) db
x(t) db
y (t) db
z (t)
v
b= = , , .
dt dt dt dt
In matrix notation we can think of the difference between two points

b2 − X
X x2 − x
b1 = (b b1 , yb2 − yb1 , zb2 − zb1 )
h i
as a 1 × 3 matrix X b2 − X
b1 . Then we can write the formula for the distance
between Xb1 and Xb2 in euclidean 3-space in terms of the dot-product:
q
d(X
b1 , X
b2 ) = b2 − X
(X b1 ) • (X
b2 − X
b1 ) (6)
or in terms of the matrix product

rh i h i|
d(X1 , X2 ) =
b b b2 − X
X b1 · Xb2 − X
b1 .
Length of a smooth curve in euclidean 3-space

Problem 90 Compute the length of the tangent vector
r
db
γ db γ
`(t) = •
dt dt
to each of your two example curves in the previous problem at each of their
points.
104
Definition 14. The length L of the curve γ b(t), t ∈ [b, e], in euclidean 3-space
is obtained by integrating the length of the tangent vector to the curve, that is,
Z e
L= `(t) dt.
b
Notice that the length of any curve only depends on the definition of the dot
product. That is, if we know the formula for the dot product, we know (the
formula for) the length of any curve.
Our first example is the path
x(t), yb(t), zb(t)) = (R · sin(t), 0, R · cos(t))

(b 0 ≤ t ≤ π.
Notice that this path lies on the sphere of radius R.
Problem 91 Write the formula for the tangent vector to the path above.
Show that the length of this path is Rπ.
105
Problem 92 (Try to) compute the length of each of your two example curves
in the previous problem.
106
Remark 3. In this last problem, you may easily be confronted with an integral
that you cannot compute. For example, if your curve γ b1 (t) happens to describe
an ellipse that is not circular, it was proved in the 19th century that no formula
involving only the standard functions from calculus will give you the length of
your path from a fixed beginning point to a variable ending point on the ellipse.
If that kind of thing occurs, go back and change the definitions of your curves in
the previous problem until you get two curves for which you can compute length
of your path from a fixed beginning point to a fixed ending point.
To study plane geometry, spherical geometry, and hyperbolic geometry in a uni-

form way we will have to change the coordinate system we use. In essence, this
means that we will have to change the distance formula slightly for each geom-
etry. These new coordinates will be chosen to keep the north and south poles
from going to infinity as the radius R of a sphere increases without bound. This
change of viewpoint will eventually let us go non-euclidean or, in the language of
Buzz Lightyear “to infinity and beyond.” The idea will be like the change from
rectangular to polar coordinates for the plane that you encountered in calculus,
only easier.

107
11 Congruences, that is, rigid motions

We explore rigid motions as matrix multiplication.
Transformations of euclidean space

Consider the following mapping of euclidean space to itself:
| |
xb yb zb = M c x b yb zb ,
c is an invertible 3 × 3 matrix. Since the matrix is invertible, this

where M
mapping is one-to-one and onto.
Notation: In this section, points and vectors will be arranged vertically.
Definition 15. Such a mapping is called a rigid motion if the distance between
any two points in euclidean space is left unchanged by the mapping, that is, for
any two points Xb1 and Xb2 in euclidean space,
d(M
cXb1 , M
cXb2 ) = d(X
b1 , X
b2 ).
It turns out that there is a special class of matrices that give rise to rigid motions.
Definition 16. A matrix M
c satisfying
 
1 0 0
c| · M
M c = I = 0 1 0 .
0 0 1
is called an orthogonal matrix.
Problem 94 Prove that a matrix M

c defines a rigid motion (a congruence)
via | |
xb yb zb = M
c x b yb zb
if and only if it is orthogonal.
Hint: Note that the square of the distance between X

b1 and X
b2 is the dot product
of the vector
v
b=X b2 − X
b1
with itself. Also recall the identity (AB)| = B | A| .
Hint: If M
c is orthogonal, write
(M
cvb ) • (M
cvb)
b•v
and deduce that this equals v b.
108
Hint: Now suppose that M

c defines a rigid motion. Explain why this means that
(M
cvb ) • (M
cv b•v
b) = v b
for every v
b . Now rewrite as:
b| · v
v b = (M
cvb )| · (M
cvb ).
Write    
a m11 m12 m13
b = b
v and M = m21
c m22 m23 
c m31 m32 m33
and view the equation
b| · v
v b = (M
cvb )| • (M
cvb)
as a polynomial equation in the variables a, b, and c.
Hint: Polynomials are equal if and only if their coefficients are equal.
109
Problem 95 Show that, if M

c is orthogonal, then the transformation
| |
xb yb zb = M c· x b yb zb .
takes the set of points (b

x, yb, zb) such that
b2 + yb2 + zb2 = R2
x

b, yb, zb such that
to the set of points x
b2 + yb2 + zb2 = R2 .
x
That is, M
c gives a one-to-one and onto mapping of the R-sphere to itself.
Hint: Write the equation

b2 + yb2 + zb2 = R2
x
as    
x
b x
b
yb • yb = R2 .
zb zb
110
Problem 96 Explain how the following diagram “proves” that function com-
position is always associative.
M2 ◦ M1 (X)
X M1 M2 M3 M3 ◦ (M2 ◦ M1 )(X)
M1 (X)
X M1 M2 M3 (M3 ◦ M2 ) ◦ M1 (X)
M1 (X) M2 (M1 (X))

X M1 M2 M3 M3 (M2 (M1 (X)))
111
Problem 97 Show that the set of orthogonal matrices M

c form a group. That
is, show that
(a) multiplication of orthogonal matrices is associative,
Hint: Use the previous problem.
(b) the product of two orthogonal matrices is orthogonal,

(c) the identity matrix is orthogonal,
c−1 of a orthogonal matrix M
(d) the inverse matrix M c is orthogonal.
112
Problem 98 Consider
 
cos θ − sin θ 0
cθ =  sin θ
M cos θ 0
0 0 1
Can you describe geometrically what this mapping is doing to the points in
euclidean space?
113
Problem 99 Show that the matrix

 
cos θ − sin θ 0
cθ =  sin θ
M cos θ 0
0 0 1
is orthogonal.
114
Problem 100 Consider

 
cos ψ 0 − sin ψ
bψ =  0
N 1 0 
sin ψ 0 cos ψ
euclidean space?
115
Problem 101 Show that the matrix

 
cos ψ 0 − sin ψ
bψ =  0
N 1 0 
sin ψ 0 cos ψ
is orthogonal.
116
Problem 102 If a curve
γ
b(t) = (b
x(t), yb(t), zb(t)), b≤t≤e
is moved by a transformation given by an orthogonal matrix M

c, show that its
length is unchanged.
Hint: Recall that the length of a curve is given by

Z er
db
γ db γ
• dt.
b dt dt
117
118
12 Changing coordinates
Now we will change coordinates.
Bringing the North Pole of the R-sphere to (0, 0, 1)

We are now ready to change coordinates on euclidean 3-space so that we can
fill up that space with plane geometry and all the spherical and hyperbolic
geometries. We have reserved the notation (x, y, z) for these new coordinates
that we will put on the ‘same’ objects we have been studying in euclidean
(b
x, yb, zb)-coordinates. These new coordinates will be chosen to keep the north
and south poles from going to infinity as the radius R of a sphere increases
without bound. In these now (x, y, z)-coordinates the sphere of radius R will
be given by the equation
K(x2 + y 2 ) + z 2 = 1
where K = 1/R2 . Notice that the above equation has solutions even when K is
negative. It is on those solution sets that hyperbolic geometries will live. So this
change of viewpoint will eventually let us go hyperbolic or, in the language of
Buzz Lightyear, will let R go ‘to infinity and beyond.’ The idea will be like the
change from rectangular to polar coordinates for the plane that you encountered
in calculus, only easier.
We are now ready to introduce this slightly different set of coordinates for R3 ,
three-dimensional euclidean space. To understand a bit better why we are doing
this, suppose we are standing at the North Pole
N = (0, 0, R)
of the sphere
b2 + yb2 + zb2 = R2
x
of radius R. As R increases, but we stay our same size, the sphere around us
becomes more and more like a flat, plane surface. However it can never get
completely flat because we are zooming out the positive zb-axis and we would
have to be ‘at infinity’ for our surface to become exactly flat. We remedy that
unfortunate situation by considering another copy of R3 , that we will call K-
warped space, whose coordinates we denote as (x, y, z). We make the following
rule in order to pass between the two R3 ’s:
x
b=x
yb = y
zb = Rz.
119
We think of the (x, y, z)-coordinates as simply being a different set of addresses

for the points in euclidean space. For example,
(x, y, z) = (0, 0, 1)
tells me that the point in euclidean space that I’m referring to is
(b
x, yb, zb) = (0, 0, R) = N.
Continuing with this “change of addresses” the sphere of radius R in euclidean

space is given by
R2 = x
b2 + yb2 + zb2
= x2 + y 2 + R 2 z 2
that is, by the equation

1
x2 + y 2 + z 2 .

1= 2
R
Definition 17. For the surface defined by
1
x2 + y 2 + z 2 .

1= 2
R
1
The quantity K = is called the curvature of the R-sphere.
R2
Problem 104 What happens to the surface when K goes to 0? How does
this relate to the colloquial sense of “curvature”?
120
Problem 105
(a) Sketch the solution set in (x, y, z)-coordinates representing the sphere
R2 = x
b2 + yb2 + zb2 = 1.
(b) Sketch the solution set in (x, y, z)-coordinates representing the sphere
R2 = x
b2 + yb2 + zb2 = 102 .
(c) Sketch the solution set in (x, y, z)-coordinates representing the sphere
b2 + yb2 + zb2 = 10−2 .

R2 = x
121
Formulas for euclidean lengths, angles, and areas

in terms of (x, y, z)-coordinates
To prepare ourselves to do hyperbolic geometry, which (in some sense) has

no satisfactory model in euclidean space, we will ‘practice’ by doing spherical
geometry (which does have a completely satisfactory model in euclidean space)
using these ‘slightly strange’ (x, y, z)-coordinates. Gradually throughout this
course we will discover that the same rules that govern spherical geometry,
expressed in (x, y, z)-coordinates, also govern flat and hyperbolic geometry! In
all three cases, the surface in (x, y, z)-coordinates that we will study is
1 = K x2 + y 2 + z 2 .

If K > 0, the geometry we will be studying is the geometry of the the euclidean
sphere of radius
1
R= √ .
K
If K = 0 we will be studying flat (plane) geometry. If K < 0, we will be studying
hyperbolic geometry.
In short, we want to use (x, y, z)-coordinates to compute with, but we want
lengths and angles to be the usual euclidean ones in (b
x, yb, zb)-coordinates.
Problem 106 Suppose we have a curve γ b in euclidean space and we think of

it as a composition of a curve γ in K-warped space with a transformation. In
other words, we’re looking at a diagram
 
x
b(x,y,z)
y
b(x,y,z) 
γ z
b(x,y,z)
t (x(t), y(t), z(t)) γ
b(t) = (b
x(t), yb(t), zb(t))
γ
b
Use the chain rule to compute

db
x db
y db
z
, , ,
dt dt dt
dx dy dz ∂b
x ∂by ∂b
z ∂bx ∂b
y ∂bz ∂b
x ∂by ∂b
z
in terms of , , , , , , , , , , , and .
dt dt dt ∂x ∂x ∂x ∂y ∂y ∂y ∂z ∂z ∂z
Hint: Recall that if F is a differentiable function of x, y, and z; and if x, y, and z

are all differentiable functions of t, then the chain rule states
|
dF dx dy dz
= ∇F · .
dt dt dt dt
122
Problem 107 With the same setting as in the previous problem, rewrite the
result of your computation in matrix notation to find DK such that
   
db
x/dt dx/dt
dby /dt = DK · dy/dt
db
z /dt dz/dt
x ∂b
∂b y ∂b
z ∂bx ∂b
y ∂bz ∂b
x ∂by ∂b
z
in terms of , , , , , , , , and .
∂x ∂x ∂x ∂y ∂y ∂y ∂z ∂z ∂z
123
Problem 108 Use the previous problems and the relationship between eu-
clidean and (x, y, z)-coordinates to show that
 
1 0 0
db
γ dγ
= 0 1 0  · .
dt dt
0 0 R
124
This last computation shows that if

   
a1
b ba2
v
b 1 = b1 
b and v
b 2 = b2 
b
c1
b c2
b
are vectors tangent to a curve in (b

x, yb, zb)-coordinates and
   
a1 a2
v 1 =  b1  and v 2 =  b2 
c1 c2
are their transformations into (x, y, z)-coordinates, then

   
1 0 0 1 0 0
b1 = 0
v 1 0  v1 and b2 = 0
v 1 0  v2
0 0 R 0 0 R
and
  |   
1 0 0 1 0 0
b1 • v
v b2 = b1|
v ·v
b2 = 0 1 0  v1  0 1 0  v2 
0 0 R 0 0 R
  
1 0 0 1 0 0
= v1| 0 1 0  0 1 0  v2
0 0 R 0 0 R
 
1 0 0
= v1| 0 1 0  v2 .
0 0 K −1
Hence:
We can compute the euclidean dot product without ever referring

to euclidean coordinates!
We incorporate that fact into the following definition.
Definition 18. The K-dot-product of vectors:

 
1 0 0
v1 •K v2 = v1| 0 1 0  v2
0 0 K −1
  
1 0 0 a2
= a1 b1 c1 0 1 0   b2  .
0 0 K −1 c2
125
Computing length
Problem 109 Show that if a vector is given to us in (x, y, z)-coordinates as
v = (a, b, c)| ,
then the length of its image in euclidean space is given by

√
|v|K = v •K v.
126
Problem 110 Consider all vectors in K-warped space with their tips on the
surface
1 = K(x2 + y 2 ) + z 2
and their tails at the origin. What can you say about the length of these vectors?
What does this tell you about the surface for all values of K > 0?
127
Computing angles
Problem 111 Show that when K > 0 if two vectors are given to us in (x, y, z)-
coordinates as    
a1 a2
v 1 =  b1  and v 2 =  b2 
c1 c2
then the angle between their image in euclidean space is given by

v1 •K v2
θ = arccos .
|v1 |K · |v2 |K
128
Computing area
Problem 112 Show that if two vectors are given to us in (x, y, z)-coordinates
as    
a1 a2
v 1 =  b1  and v 2 =  b2 
c1 c2
then the area of the parallelogram spanned by the image of those two vectors
in euclidean space is
v    
s u 1 0 0
v1 •K v1 v2 •K v1 v
u
1
0  v1| v2| .

det = tdet 
u 0 1
v1 •K v2 v2 •K v2 v2
0 0 K −1
129
Moral of the story: The dot-product rules! That is, if you know the dot-product
you know everything there is to know about a geometry, lengths, areas, angles,
everything. And the set
1 = K x2 + y 2 + z 2

continues to make sense even when K is negative. And as we will see later on,
the definition of the K-dot product also makes sense for tangent vectors to that
set when K is negative. The geometry we get when the constant K is chosen
to be negative is called a hyperbolic geometry. The geometry we get, when the
constant K is just chosen to be non-zero is called a non-euclidean geometry.
In fact all the non-euclidean 2-dimensional geometries are either spherical or
hyperbolic.
130
13 Rigid motions in (x, y, z)-coordinates

Here we dig deeper to understand our different coordinates.
We now wish to figure out how to convert a transformation

   
x
b x
b
c · yb
x, yb, zb) = yb = M
Tb(b
zb zb
from euclidean space to (x, y, z)-coordinates. Recall to convert a point from

(x, y, z)-coordinates to euclidean coordinates, we write
      
x
b x 1 0 0 x
yb =  y  = · 0 1 0  y  .
zb Rz 0 0 R z
So we have the diagram

h1 00
i
01 0 ·
| 00R |
b yb zb ∈ R3 ∈ R3

x x y z
c·
M h1 0 i T =?
0
01 0 ·
| 00R |
xb yb zb x y z
Problem 114 Using the diagram above, explain why

       
x 1 0 0 1 0 0 x
y  = 0 1 0 ·Mc · 0 1 0  · y  .
z 0 0 R−1 0 0 R z
Hint: Express the transformation matrix M of T in terms of M

c.
131
Problem 115 Show that a transformation M in (x, y, z)-coordinates pre-

serves distances in K-warped space if and only if
   
1 0 0 1 0 0
M | · 0 1 0  · M = 0 1 0 . (∗)
0 0 K −1 0 0 K −1
Hint: Explain how the previous problem gives a correspondence between transfor-
mations defined by a matrix in K-warped space and transformations defined by a
matrix in euclidean space. Then show that (∗) is equivalent to the corresponding
euclidean matrix M
c being orthogonal.
132
The equation (∗) is the condition (in (x, y, z)-coordinates) which affirms that
the transformation which takes the path γ(t) = (x(t), y(t), z(t)) to the path
γM (t) = M · (x(t), y(t), z(t))| preserves lengths of tangent vectors at corre-
sponding points. Therefore, by integrating, the (total) length of the curve γM
is the same as the total length of the curve γ.
Problem 116 Verify that this is the correct condition by showing that any
3 × 3 matrix M satisfying (∗) also satisfies
(M · v) •K (M · v) = v •K v,
where
v = X2 − X1 .
That is, the transformation given in (x, y, z)-coordinates by a matrix M that
satisfies your condition preserves the K-dot product.
133
Definition 19. A K-distance-preserving transformation of K-geometry is called

a K-rigid motion or a K-congruence.
With this definition, and our work above, we make a new definition:
Definition 20. A 3 × 3 matrix M is called K-orthogonal if
   
1 0 0 1 0 0
M | · 0 1 0  · M = 0 1 0 .
0 0 K −1 0 0 K −1
Problem 117 For K 6= 0, show that if M is K-orthogonal, then the transfor-

mation    
x x
T (x, y, z) = y  = M · y 
z z
takes the set of points (x, y, z) such that
1 = K x2 + y 2 + z 2

to the set of points (x, y, z) such that
1 = K x2 + y 2 + z 2 .

That is, M gives a one-to-one and onto mapping of K-geometry to itself.
Hint: Explain how you can write the equation
1 = K x2 + y 2 + z 2 .

as    
1 0 0 x
1
x y z · 0 1 0  · y  = .
−1 K
0 0 K z
134
Problem 118 For K 6= 0, show that the set of K-orthogonal matrices M

forms a group. That is, show that
(a) multiplication of K-orthogonal matrices is associative,
Hint: Recall that function composition is always associative.
(b) the product of two K-orthogonal matrices is K-orthogonal,
(c) the identity matrix is K-orthogonal,

(d) the inverse matrix M −1 of a K-orthogonal matrix M is K-orthogonal.
135
Problem 119 Convert the orthogonal matrix

 
cos θ − sin θ 0
cθ =  sin θ
M cos θ 0
0 0 1
into its K-orthogonal counterpart Mθ . Are you surprised? Why or why not?
136
Problem 120 Convert the orthogonal matrix

 
cos ψ 0 − sin ψ
cψ =  0
M 1 0 
sin ψ 0 cos ψ
into its K-orthogonal counterpart Mψ . Are you surprised? Why or why not?
137
Why use K-coordinates?

We have seen that we could measure the usual euclidean lengths of curves γ
b in
terms of the formulas of curves γ in K-warped space using the K-dot product.
The short reason for this is that
db
γ dbγ dγ dγ
• = •K
dt dt dt dt
where  
1 0 0 |
dγ dγ dγ  dX
•K = · 0 1 0  · .
dt dt dt dt
0 0 K −1
In other words, the usual geometry of the sphere of radius R is simply the
geometry of the set
{(x, y, z) ∈ R3 : 1 = K x2 + y 2 + z 2 }

with K = 1/R2 and with lengths (and areas) given by the K-dot product. Said
another way, we can do all of spherical geometry in (x, y, z)-coordinates. All we
need is the set defined by the relation
1 = K x2 + y 2 + z 2

and the K-dot product. But the set defined by the equation above continues
to exist even if K = 0 or K < 0, and the K-dot product formula continues to
make sense even if K < 0. In short we have the following table:
Spherical (K > 0) Euclidean (K = 0) Hyperbolic (K < 0)
Surface in euclidean space b2 + yb2 + zb2 = R2

x DNE DNE
Euclidean dot product b| · w

v b DNE DNE
Surface in K-warped space 1 = K(x2 + y 2 ) + z 2 1 = K(x2 + y 2 ) + z 2 1 = K(x2 + y 2 ) + z 2
h1 0 0
i h1 0 0
i
K-dot product v| 01 0 w DNE v| 01 0 w
0 0 K −1 0 0 K −1
This table tells us that ‘there is something else out there,’ that is, some other
type of two-dimensional geometry beyond plane geometry and spherical geom-
etry. But the gap in the bottom row of the table is a bit disturbing. If we
can’t express the usual dot-product in plane geometry as the K-dot product for
138
K = 0, we can’t pass smoothly from spherical through plane geometry to hy-

perbolic geometry using (x, y, z)-coordinates. Later, we will examine two ways
to produce coordinates uniformly for spherical, plane and hyperbolic geometry
that overcome this difficulty.
139
14 Lines in spherical geometry

Here we examine “lines” in spherical geometry and prove a spherical version of
the Pythagorean Theorem.
Spherical coordinates, a shortest path from the

North Pole
We next will figure out what is the shortest path you can take between two points
on the euclidean R-sphere. However, we will do our calculation using (x, y, z)-
coordinates, as we won’t have (b x, yb, zb)-coordinates when we get to hyperbolic
geometry.
For our purposes, it will be convenient to parameterize the sphere in K-geometry:
x(σ, τ ) = R · sin σ · cos τ,
y(σ, τ ) = R · sin σ · sin τ,
z(σ, τ ) = cos σ,
where 0 ≤ σ ≤ π and 0 ≤ τ < 2π.
Problem 122 Show that these functions actually parameterize the R-sphere.
Hint: This is an exercise in “double-containment”: you must show that every point
specified in this way is a point on the sphere and vice versa. To show the one direction,
show
K x(σ, τ )2 + y(σ, τ )2 + z(σ, τ )2 = 1

140
for all (σ, τ ). To show the other direction, appeal to the diagram above.
141
If we are going to describe paths on the R-sphere by paths in the (σ, τ )-plane
we are going to need to figure out the K-dot product in (σ, τ )-coordinates so
that we can compute the lengths of paths in these coordinates.
Problem 123 Suppose we have a curve γ in K-warped space which we can

decompose as
 
x(σ,τ )
 y(σ,τ ) 
γsph z(σ,τ )
t (σ(t), τ (t)) γ(t) = (x(t), y(t), z(t))

dx dy dz
, , ,
dt dt dt
dσ dτ ∂x ∂y ∂z ∂x ∂y ∂z
in terms of , , , , , , , and .
dt dt ∂σ ∂σ ∂σ ∂τ ∂τ ∂τ
Hint: Recall that if F is a differentiable function of a and b; and if a and b are all
differentiable functions of t, then the chain rule states

dF da/dt
= ∇F · .
dt db/dt
142
result of your computation in matrix notation to find Dsph such that
 
dx/dt
dy/dt = Dsph · dσ/dt
dτ /dt
dz/dt
∂x ∂y ∂z ∂x ∂y ∂z
in terms of , , , , , and .
∂σ ∂σ ∂σ ∂τ ∂τ ∂τ
143
∂x ∂y ∂z ∂x ∂y ∂z
Problem 125 Now find Psph in terms of K, , , , , , and
∂σ ∂σ ∂σ ∂τ ∂τ ∂τ
such that
 
dσ
dx dy dz dx dy dz dσ dτ  dt 
, , •K , , = · Psph ·  dτ  .
dt dt dt dt dt dt dt dt
dt
144
Problem 126 Set
x(σ, τ ) = R · sin σ · cos τ,

y(σ, τ ) = R · sin σ · sin τ,
z(σ, τ ) = cos σ,
and show that Psph from the problem above is

2
R 0
Psph = .
0 R2 · sin2 σ
145
Definition 21. Let vsph and wsph be a vectors in (σ, τ )-coordinates originating
at the same (σ, τ )-coordinate. Define
|
vsph •sph wsph = vsph · Psph · wsph
where 2
R 0
Psph =
0 R2 · sin2 σ
and σ is determined by the coordinate that the vectors originate from.
Now notice that you can write a path on the R-sphere by giving a path (σ(t), τ (t))
in the (σ, τ )-plane. To write a path that starts at the North Pole, just write
(σ(t), τ (t)) , 0≤t≤e
and demand that σ(0) = 0. If you want the path to end on the plane y = yb = 0,
demand additionally that τ (e) = 0.
Now given a path on the R-sphere
(σ(t), τ (t)) , 0≤t≤e
satisfying σ(0) = τ (0) = 0 and τ (e) = 0, its length is given by the formula
Z e s
dσ dτ dσ dτ
L= , •sph , dt. (∗)
0 dt dt dt dt
(Why?)
Problem 127 Prove that the shortest path on the R-sphere from the North
Pole (0, 0, 1) (in (σ, τ )-coordinates, (0, 0)) to a point
(x, y, z) = (R sin e, 0, cos e)
(in σ, τ )-coordinates, (e, 0)) is the downwards path lying in the plane y = 0.
Hint: Start with some arbitrary path (σ(t), τ (t)) which begins and ends at this
point and show that it’s longer than this path, in two steps:
• First use the equation (∗) to show that it gets shorter if you keep the same σ
but make τ always zero (squishing the curve onto the plane y = 0.)
• Finish the argument by eliminating any backtracking.
146
Shortest path between any two points

We next prove the theorem that shows that shortest path on the surface of the
Earth from Rio de Janeiro to Los Angeles is the one cut on the surface of the
Earth by the plane that passes through the center of the Earth and through
Rio and through Los Angeles. That is usually the route an airplane would take
when flying between the two cities.
Theorem 10. Given any two points X1 = (x1 , y1 , z1 ) and X2 = (x2 , y2 , z2 ) in
K-geometry, the shortest path between the two points is the path cut out by the
set
K x2 + y 2 + z 2 = 1

and the plane containing (0, 0, 0), X1 , and X2 .
Problem 128 (Tricky!) Explain in words how to prove this theorem by

using K-rigid motions of the form
   
cos θ − sin θ 0 cos ψ 0 −R · sin ψ
Mθ =  sin θ cos θ 0 and Nψ =  0 1 0 .
−1
0 0 1 R · sin ψ 0 cos ψ
Hint: Mθ is a K-rigid motion that rotates around the z-axis and Nψ is a K-rigid
motion that rotates around the y-axis.
Hint: You should apply two K-rigid motions of the form Mθ (for different angles)
and one K-rigid motion of the form Nψ , though not necessarily in that order!
147
Definition 22. A line in spherical geometry will be a curve that extends in-
finitely in each direction and has the property that, given any two points X1 and
X2 on the path, the shortest path between X1 and X2 lies along that curve.
Lines in spherical geometry are usually called great circles on the R-sphere.
They are the intersections of the R-sphere with planes through (0, 0, 0).
The spherical Pythagorean Theorem

To start we need some basic facts about lengths of lines in spherical geometry.
Problem 129 Given a line in spherical geometry lying entirely in the plane
y = 0,
x(t) = R sin t,
y(t) = 0,
z(t) = cos t,
show that the length of the segment at 0 ≤ t ≤ e is exactly Re.
Hint: Use a previous problem.
148
Problem 130 Explain in words how to prove that given two points on R-
sphere, say XA and XB , the length of the spherical line connecting them is
given by
XA •K XB
R · e = R · arccos .
|XA |K · |XB |K
by using K-rigid motions of the form
   
cos θ − sin θ 0 cos ψ 0 −R · sin ψ
0 0 1 R−1 · sin ψ 0 cos ψ
149
We will now give the spherical analogue of the Pythagorean Theorem.
Theorem 11 (Spherical Pythagorean Theorem). If 4XA XB XC is a right tri-

angle on the R-sphere with right angle ∠XA XC XB , and side a opposite XA , b
opposite XB , and c opposite XC , then
c a
b
cos = cos cos .
R R R
Let’s see why this theorem is true. We may via K-rigid motions place the
triangle so that XC is at the North Pole, XA is in the plane y = 0, and XB is in
the plane x = 0 (note XA and XB may be switched—if this is the case, simply
rename them). In this case,
XA = (R · sin α, 0, cos α),

XB = (0, R · sin β, cos β).
Hence the length of side b is R · α. Using a rigid motion of the form

 
cos θ − sin θ 0
Mθ =  sin θ cos θ 0
0 0 1
when θ = π/2 we see that the length of side a is R · β. Set

XA •K XB
γ = arccos .
|XA |K · |XB |K
150
Since we are working on the R-sphere,
R2 · cos γ = XA •K XB
  
1 0 0 0
= R · sin α 0 cos α 0 1 0  R · sin β 
0 0 R2 cos β
= R2 · cos α · cos β.
Problem 131 Explain how to progress from the fact that
R2 · cos γ = R2 · cos α · cos β.
to the conclusion of the theorem

c a
b
cos = cos cos .
R R R
151
Problem 132 Use the Taylor series expansion of cos(x) centered around x =
0,
x2 x4 x6
cos(x) = 1 − + − + ···
2! 4! 6!
to show that for “small” triangles, the spherical Pythagorean Theorem reduces
to the euclidean Pythagorean Theorem, meaning
c2 ≈ a2 + b2 .
152
153
15 Lines in hyperbolic geometry

Here we examine “lines” in hyperbolic geometry and prove a hyperbolic version
of the Pythagorean Theorem.
Hyperbolic coordinates, a shortest path from the North

Pole
We next will figure out what is the shortest path you can take between two
points in hyperbolic geometry. Since K is negative, we must do our calculation
using only (x, y, z)-coordinates. However, this will allow us to see the full power
of working in K-warped space, since our work will be essentially the same as
when K was positive—though our parametrization will be different.
Just as (cos σ, sin σ) parametrize the unit circle, hyperbolic functions
eσ + e−σ eσ − e−σ

cosh σ = , sinh σ =
2 2
parametrize the ‘unit’ hyperbola. Hence we define
x(σ, τ ) = |K|−1/2 · sinh σ · cos τ,

y(σ, τ ) = |K|−1/2 · sinh σ · sin τ,
z(σ, τ ) = cosh σ,
where 0 ≤ σ < ∞ and 0 ≤ τ < 2π.
154
Problem 134 Show that these hyperbolic coordinates do actually parametrize

K-geometry.
Hint: Remember, K is negative.
Hint: Remember,
− sinh2 σ + cosh2 σ = 1.
Hint: This is an exercise in “double-containment.” To show the one direction, show
K x(σ, τ )2 + y(σ, τ )2 + z(σ, τ )2 = 1

for all (σ, τ ). To show the other direction, appeal to the diagram above.
155
Just as we did on the R-sphere, we can write a path on the K-surface by

giving a path in the (σ, τ )-plane. Again, we will need to figure out the K-dot
product in (σ, τ )-coordinates so that we can compute the lengths of paths in
these coordinates.

decompose as
 
x(σ,τ )
 y(σ,τ ) 
γhyp z(σ,τ )
t (σ(t), τ (t)) γ(t) = (x(t), y(t), z(t))

dx dy dz
, , ,
dt dt dt
dσ dτ ∂x ∂y ∂z ∂x ∂y ∂z
in terms of , , , , , , , and .
dt dt ∂σ ∂σ ∂σ ∂τ ∂τ ∂τ
Hint: Simply write down the answer from a previous problem.
156
result of your computation in matrix notation to find Dhyp such that
 
dx/dt
dy/dt = Dhyp · dσ/dt
dτ /dt
dz/dt
∂x ∂y ∂z ∂x ∂y ∂z
∂σ ∂σ ∂σ ∂τ ∂τ ∂τ
157
∂x ∂y ∂z ∂x ∂y ∂z
Problem 137 Now find Phyp in terms of K, , , , , , and
∂σ ∂σ ∂σ ∂τ ∂τ ∂τ
such that
 
dσ
dx dy dz dx dy dz dσ dτ  dt 
, , •K , , = · Phyp ·  dτ  .
dt dt dt dt dt dt dt dt
dt
158
Problem 138 Set
x(σ, τ ) = |K|−1/2 · sinh σ · cos τ,

y(σ, τ ) = |K|−1/2 · sinh σ · sin τ,
z(σ, τ ) = cosh σ,
and show that Phyp from the problem above is

−1
|K| 0
Phyp = .
0 |K|−1 · sinh2 σ
159
Definition 23. Let vhyp and whyp be a vectors in (σ, τ )-coordinates originating
at the same (σ, τ )-coordinate. Define
|
vhyp •hyp whyp = vhyp · Phyp · whyp
where −1
|K| 0
Phyp =
0 |K|−1 · sinh2 σ
and σ is determined by the coordinate that the vectors originate from.
Now notice that you can write a path on the K-surface by giving a path
(σ(t), τ (t)) in the (σ, τ )-plane. To write a path that starts at the North Pole,
just write
(σ(t), τ (t)) , 0≤t≤e
and demand that σ(0) = τ (0) = 0. If you want the path to end on the plane
y = yb = 0, demand additionally that τ (e) = 0.
Now given a path on the K-surface
(σ(t), τ (t)) , 0≤t≤e
satisfying σ(0) = τ (0) = 0 and τ (e) = 0, its length is given by the formula
Z e s
dσ dτ dσ dτ
L= , •hyp , dt. (∗)
0 dt dt dt dt
Problem 139 Prove that the shortest path on the K-surface from the North
Pole
N = |K|−1/2 · sinh 0 · cos 0, |K|−1/2 · sinh 0 · sin 0, cosh 0
to a point
(x, y, z) = |K|−1/2 sinh e, 0, cosh e
is the path lying in the plane y = 0.
Hint: Use the same steps you did in the sphere case.
160
Shortest path between any two points

Just as we proved in spherical geometry that the shortest path is the path cut
out by
K(x2 + y 2 ) + z 2 = 1
and the plane containing the origin and the two points in question, we will see
that a completely analogous result is true in hyperbolic geometry. Before we
start, we will need one more class of rigid motions to add to our collection.
Problem 140 Assuming K is negative, show
0 |K|−1/2 · sinh ψ
 
cosh ψ
Nψ =  0 1 0 
|K|1/2 · sinh ψ 0 cosh ψ
is a K-rigid motion.
161
Problem 141 Assuming K is negative, consider
0 |K|−1/2 · sinh ψ
 
cosh ψ
Nψ =  0 1 0 .
K-warped space?
Hint: First look at the image of the point (0, 0, 1).
162
Theorem 12. Given any two points X1 = (x1 , y1 , z1 ) and X2 = (x2 , y2 , z2 ) in

K-geometry, the shortest path between the two points is the path cut out by the
set
K x2 + y 2 + z 2 = 1

and the plane containing (0, 0, 0), X1 , and X2 .
Problem 142 Explain in words, with pictures as needed, how to prove this
theorem by using the K-rigid motions
0 |K|−1/2 · sinh ψ
   
cos θ − sin θ 0 cosh ψ
0 0 1 1/2
|K| · sinh ψ 0 cosh ψ
Hint: Mθ is a K-rigid motion that rotates around the z-axis and Nψ is a K-rigid
motion that “slides” the K-surface past the y-axis.
Hint: You should apply two K-rigid motions of the form Mθ (for different angles)
and one K-rigid motion of the form Nψ —though not necessarily in that order!
163
Definition 24. A line in hyperbolic geometry will be a curve that extends

infinitely in each direction and has the property that, given any two points
X1 and X2 on the path, the shortest path between X1 and X2 lies along that
curve. Lines in hyperbolic geometry are the intersections of the K-geometry
with planes through (0, 0, 0). The length of the shortest path between two points
in K-geometry will be called the K-distance.
The hyperbolic Pythagorean Theorem

To start we need some basic facts about lengths of lines in hyperbolic geometry.
Problem 143 Given a line in hyperbolic geometry lying entire in the plane
y = 0,
x(t) = |K|−1/2 sinh t,

y(t) = 0,
z(t) = cosh(t),
show that the length of the segment on the interval 0 ≤ t ≤ e is exactly |K|−1/2 e.
Hint: Use a previous problem.
164
Problem 144 Explain in words how to prove that given two points on the
surface
K(x2 + y 2 ) + z 2 = 1,
say XA and XB , the length of the hyperbolic line connecting them is given by

−1/2 −1/2 XA •K XB
|K| · ε = |K| · arcosh .
|XA |K · |XB |K
by using the K-rigid motions
|K|−1/2 · sinh ψ
   
cos θ − sin θ 0 cosh ψ 0
0 0 1 1/2
|K| · sinh ψ 0 cos ψ
165
We will now give the hyperbolic analogue of the Pythagorean Theorem.
Theorem 13 (Hyperbolic Pythagorean Theorem). If 4XA XB XC is a right

triangle on the surface
K(x2 + y 2 ) + z 2 where K<0
with right angle ∠XA XC XB , and side a opposite XA , b opposite XB , and c

opposite XC , then

cosh |K|1/2 · c = cosh |K|1/2 · a cosh |K|1/2 · b .
Let’s see why this theorem is true. We may via K-rigid motions place the
triangle so that XC is at the North Pole, XA is in the plane y = 0, and XB is in
the plane x = 0 (note XA and XB may be switched—if this is the case, simply
rename them). In this case,
XA = (|K|−1/2 · sinh α, 0, cosh α),

XB = (0, |K|−1/2 · sinh β, cosh β).
Hence the length of side b is |K|−1/2 · α. Using a rigid motion of the form
 
cos θ − sin θ 0
Mθ = sin θ
 cos θ 0
0 0 1
when θ = π/2 we see that the length of side a is |K|−1/2 · β. Set

XA •K XB
γ = arcosh .
|XA |K · |XB |K
166
Since we are working on the K-surface,
K −1 · cosh γ = XA •K XB
  
1 0 0 0
= |K|−1/2 · sinh α 0  |K|−1/2 · sinh β 

0 cosh α 0 1
0 0 K −1 cosh β
= K −1 · cosh α · cosh β.
Problem 145 Explain how to progress from the fact that
K −1 · cosh γ = K −1 · cosh α · cosh β.
to the conclusion of the theorem
cosh(|K|1/2 · c) = cosh(|K|1/2 · a) cosh(|K|1/2 · b).
167
Problem 146 Use the Taylor series expansion of cosh(x) centered around
x = 0,
x2 x4 x6
cosh(x) = 1 + + + + ···
2! 4! 6!
to show that for “small” triangles, the hyperbolic Pythagorean Theorem reduces
to the euclidean Pythagorean Theorem, meaning
c2 ≈ a2 + b2 .
168
169
16 Central projection
Here we start to develop unified models for our geometries.
Central projection coordinates

Now we will examine one way to produce coordinates uniformly for spherical,
plane, and hyperbolic geometry that will allow us to pass smoothly through the
three geometries. Let’s project K-geometry,
1 = K x2 + y 2 + z 2

onto the plane z = 1 using the origin O = (0, 0, 0) as the center of projection:
Let’s look at this from a different vantage point, say with our eye along the edge
of the plane z = 1:
Note, we are only projecting the top of the surface

1 = K x2 + y 2 + z 2

onto the plane z = 1. So when K > 0, this is the “Northern hemisphere” of the
sphere, when K < 0, this is the upper hyperboloid, and when K = 0, this is the
plane z = 1.
Problem 148 Use similar triangles to explain why for any given (x, y, z),
there is a number λ such that
λ · (xc , yc , 1) = (x, y, z).
170
171
Problem 149 What range of values could λ take when K > 0?
172
Problem 150 What range of values could λ take when K < 0?
173
Problem 151 What range of values could λ take when K = 0?
174
Problem 152 For the projection of the set 1 = K x2 + y 2 + z 2 onto the

z = 1 plane with center of projection O, write (xc , yc ) as a function of (x, y, z).
175

z = 1 plane with center of projection O, write (x, y, z) as a function of (xc , yc ).
176
Central projection dot product

Like we have done before, now we want to be able to find a dot product that
will allow us to compute lengths in central projection coordinates that will agree
with the K-dot product, and hence the euclidean dot product.

decompose as
 
x(xc ,yc )
 y(xc ,yc ) 
γc z(xc ,yc )
t (xc (t), yc (t)) γ(t) = (x(t), y(t), z(t))

dx dy dz
, , ,
dt dt dt
dxc dyc ∂x ∂y ∂z ∂x ∂y ∂z
in terms of , , , , , , , and .
dt dt ∂xc ∂xc ∂xc ∂yc ∂yc ∂yc
Hint: Simply write down the answer from a previous problem with some minor
changes.
177
result of your computation in matrix notation to find Dc such that
 
dx/dt
dy/dt = Dc · dxc /dt
dyc /dt
dz/dt
∂x ∂y ∂z ∂x ∂y ∂z
∂xc ∂xc ∂xc ∂yc ∂yc ∂yc
changes.
178
∂x ∂y ∂z ∂x ∂y
Problem 156 Now find Pc in terms of K, , , , , , and
∂xc ∂xc ∂xc ∂yc ∂yc
∂z
such that
∂yc
 
dxc
dx dy dz dx dy dz dxc dyc  dt 
, , •K , , = · Pc ·  dy .
dt dt dt dt dt dt dt dt c
dt
changes.
179
Before we actually compute Pc , it will help to compute the partial derivatives.
Problem 157 Set
x(xc , yc ) = λ · xc ,
y(xc , yc ) = λ · yc ,
z(xc , yc ) = λ,
and show that

∂x ∂x
= λ − Kx2c λ3 , = −Kxc yc λ3 ,
∂xc ∂yc
∂y ∂y
= −Kxc yc λ3 , = λ − Kyc2 λ3 ,
∂xc ∂yc
∂z ∂z
= −Kxc λ3 , = −Kyc λ3 .
∂xc ∂yc
Hint: Work in the following way:
(a) Remember that λ is itself a function of xc and yc , so its partial derivatives also
matter. Use the product rule:
∂x ∂λ
= λ + xc · .
∂xc ∂xc
(b) Express the partial derivative in terms of λ, K, xc , and yc .
180
Problem 158 With the same setting as above, show that Pc is
Kyc2 + 1 λ4 −Kxc ycλ4

Pc = .
−Kxc yc λ4 Kx2c + 1 λ4
Hint: When simplifying, combine the terms with the highest degree of λ and note
that
λ−2 = K x2c + yc2 + 1.

181
Definition 25. Let vc and wc be vectors in (xc , yc )-coordinates originating at

the same (xc , yc )-coordinate. Define
vc •c wc = vc| Pc wc
where

Pc =
and λ, xc , and yc are determined by the coordinate that the vectors originate
from.
When the finite is infinite and the infinite is finite

Central projection allows us to think about both “finite” areas and “infinite”
areas in new ways.
When K is positive, the finite is infinite When K > 0, we are working

with a sphere. As we know, a sphere has finite surface area.
Problem 159 As a point approaches the equator of the R-sphere from the
North Pole, where does it move to under central projection?
Hint: A picture is worth a thousand words!
182
Problem 160 Explain why one might say that central projection makes the
“finite” seem “infinite.”
183
When K is negative, the infinite is finite Notice that, if K < 0, the

equation for the K-plane becomes
z 2 − |K|(x2 + y 2 ) = 1.
This describes a 2-sheeted hyperboloid with the z-axis as major axis. We will
only consider the sheet on which z is positive as forming the K-geometry. In
plain English, this means we are doing geometry on this hyperboloid:
Of critical importance, note
this hyperboloid has infinite surface area.
The hyperboloid above is obtained by rotating the hyperbola

z 2 − |K| x2 = 1
in the (x, z)-plane around the z-axis.
184
Problem 161 Consider points of the form (x, 0, z) in K-geometry. As x and

z shoot off to infinity, what happens to the corresponding central projection
coordinate xc ? Relate this to the asymptote of the hyperbola as x → ∞.
Hint: First rewrite z 2 − |K|x2 = 1 as

p
z = 1 + |K|x2
and explain why this is acceptable.
Hint: Explain why computing this limit

p
1 + |K|x2
lim
x→∞ x
helps us in this context.
185
Problem 162 When K < 0, what is the set of points (xc , yc ) in central pro-
jection which correspond to points on the hyperboloid?
Hint: Use the answer to the previous question and the fact that both the hyper-
boloid and the process of central projection are symmetric with respect to rotation
around the z-axis.
186
Problem 163 Explain why one might say that central projection makes the
“infinite” seem “finite.”
187
The (xc , yc )-coordinates p

are called Klein coordinates for hyperbolic geometry
and the disk of radius 1/ |K| called the Klein model for hyperbolic geometry
in honor of the famous German geometer, Felix Klein.
188
17 Rigid motions in central projection

Here we study rigid motions in central projection coordinates.
Rigid motions in central projection coordinates

Suppose now we have a K-rigid motion
   
x x
y  = M y 
z z
of K-geometry, given by a K-orthogonal matrix

 
m11 m12 m13
M = m21 m22 m23  .
m31 m32 m33
Let’s convert this K-rigid motion to a rigid motion in central projection coor-
dinates. This new rigid motion will not necessarily be a mapping defined by a
matrix, so we’ll have to use some new notation.
·λ
(x, y, z) ∈ R3 (xc , yc , 1) ∈ R2 × {1} (xc , yc ) ∈ R2
M· µc0 =? µc =?
·λ
(x, y, z) (xc , yc , 1) (xc , yc )
Problem 165 Using the diagram above, explain why the formula for (xc , yc ) =
µc (xc , yc ) is

m11 xc + m12 yc + m13 m21 xc + m22 yc + m23
µc (xc , yc ) = , .
m31 xc + m32 yc + m33 m31 xc + m32 yc + m33
189
From K-rigid motions to rigid motions in central

projection
Now let’s use our new tool to convert K-rigid motions to rigid motions in central
projection.
Problem 166 For any K, consider the K-rigid motion of K-geometry

 
cos θ − sin θ 0
Mθ =  sin θ cos θ 0 .
0 0 1
central projection?
190
Problem 167 For any K, consider the K-rigid motion of K-geometry

 
cos θ − sin θ 0
Mθ =  sin θ cos θ 0 .
0 0 1
Convert this to a rigid motion in central projection.
191
Problem 168 Assuming K > 0, consider the K-rigid motion of the R-sphere
 
cos ψ 0 −R · sin ψ
Nψ =  0 1 0 .
R−1 · sin ψ 0 cos ψ
central projection?
192
Problem 169 Assuming K > 0, consider the K-rigid motion of the R-sphere
 
cos ψ 0 −R · sin ψ
Nψ =  0 1 0 .
R−1 · sin ψ 0 cos ψ
193
Problem 170 Assuming K < 0, consider the K-rigid motion of the K-surface
0 |K|−1/2 · sinh ψ
 
cosh ψ
Nψ =  0 1 0 
central projection?
194
Problem 171 Assuming K < 0, consider the K-rigid motion of the K-surface
0 |K|−1/2 · sinh ψ
 
cosh ψ
Nψ =  0 1 0 
195
196
18 Lines, angles, and areas in central projection
18 Lines, angles, and areas in central

projection
Here we look at lines, angles, and areas in central projection coordinates.
Central projection preserves lines

We all probably realize that you can’t make a perfect flat map of our spherical
world.
What this means is that either great circular arcs on the sphere won’t corre-
spond to straight lines on the map, or angles on the sphere be equal to the
corresponding angles on the map.
We do the next best thing—we make two maps of the sphere! One map has the
property that straight lines on the map correspond to shortest paths, and the
other map has the property that angles are faithfully represented. We will now
show that in central projection, straight lines correspond to shortest paths, but
angles are not faithfully represented.
Problem 173 Show that “lines” in K-geometry correspond to actual lines in

(xc , yc )-coordinates under central projection.
Hint: Referring to previous handouts, explain why intersecting the K-surface
1 = K(x2 + y 2 ) + z 2
with the plane

ax + by + cz = 0
produces a K geometry line.
Hint: Argue that the central projection to the plane z = 1 of a point on the K-
surface which lies in the plane
ax + by + cz = 0
also lies in that plane. What is the intersection of two planes?
197
Problem 174 Explain why the answer from the previous question makes per-
fect sense if K = 0.
198
Problem 175 If K < 0 explain why Euclid’s fifth axiom:
Through a point not on a line there passes a unique parallel line.
fails.
199
Problem 176 If K > 0 reconcile the fact
in spherical geometry, there are no parallel lines,
with the fact that
The lines yc = 1 and yc = −1 seem to be parallel.
200
Angles in central projection coordinates

While central projection coordinates “preserve” lines, meaning that shortest
paths in euclidean geometry, spherical geometry, and hyperbolic geometry are
all mapped to lines, it does not preserve angles. This means that the angle that
we see in central projection coordinates may or may not be the same as the
angle in euclidean geometry. To see this, we will give an example where the
appearance of the angle is changed by a rigid motion.
Problem 177 Consider the following lines in the Klein disk,
where A0 B 0 and C 0 D0 are the images of AB and CD under a K-rigid motion.

Explain how this image shows that central projection does not preserve angles.
201
Areas in central projection coordinates
We have already seen that on the R-sphere, the area of an α-lune is 2α · R2 .

Let’s compute this a different way.
Problem 178 Let K > 0 and consider the following region in central projec-
tion
Use the fact that the area of an α-lune is 2α · R2 to compute the area of the
region.
202
On the other hand, we know that the area in spherical geometry represented by
the region in central projection is given by
v
∂X ∂X ∂X ∂X
u  
ZZ u •K •K
u ∂x ∂xc ∂yc ∂xc  dx dy
t  ∂Xc
udet 
∂X ∂X ∂X  c c
Lc •K •K
∂xc ∂yc ∂yc ∂yc
Problem 179 Give a heuristic explanation of why this integral computes

what we say it computes.
Hint: Think about a tiny square in (xc , yc )-coordinates.
203
You have already shown that

v
∂X ∂X ∂X ∂X
u  
ZZ u •K •K
u ∂x
udet  c
 ∂xc ∂yc ∂xc  dx dy
t ∂X ∂X ∂X ∂X  c c
Lc •K •K
v
u 
∂X
  ↑ ↑

ZZ u ← → 1 0 0
 ∂X ∂X 
u
∂xc
= udet   0 1 0   dxc dyc
u  
∂X −1 ∂xc ∂yc
Lc ← → 0 0 K
t
∂yc ↓ ↓
Problem 180 Explain why


∂X
  ↑ ↑

← → 1 0 0
∂xc  ∂X ∂X 
 0 1 0   = Pc .
 
 ∂X ∂xc ∂yc
← → 0 0 K −1
∂yc ↓ ↓
Hint: No new computations need to be done, just look at how Pc was derived.
204
Hence now we see

v
∂X ∂X ∂X ∂X
u  
ZZ u •K •K ZZ
u ∂x
udet  c
 ∂xc ∂yc ∂xc  dx dy = p
det Pc dxc dyc
t ∂X ∂X ∂X ∂X  c c
Lc •K •K Lc
p
Problem 181 Compute det Pc .
Hint: As a gesture of friendship, we’ll remind you that

Pc = .
205
At this point we see we are interested in the following integral

ZZ
(K(x2c + yc2 ) + 1)−3/2 dxc dyc .
Lc
Problem 182 Examining the following diagram
convert ZZ
(K(x2c + yc2 ) + 1)−3/2 dxc dyc .
Lc
to polar coordinates and compute the integral.
Hint: Recall that to convert to polar coordinates, set

p
r = x2c + yc2 ,
θ = arctan(yc /xc ),
and replace dxc dyc with r dr dθ.
206
Hopefully you got the same area that you predicted in the previous problem!
Central projection unifies euclidean, spherical, and

hyperbolic geometry
Under central projection, our new dot product defined by
vc •c wc = vc| · Pc · wc
make sense when K is zero, positive, and negative. Hence this dot product
makes sense for euclidean, spherical, and hyperbolic geometry. Moreover, in
central projection, shortest paths on the K-surface
K(x2 + y 2 ) + z 2 = 1
map to lines in the plane z = 1. However, there is one trade-off: Angles are
not preserved in central projection. This means that when angles are projected
into the plane via central projection, the angle we see in the (xc , yc )-plane may
or may not be the actual angle between two vectors. Summarizing, we have:

x DNE DNE

v b DNE DNE
h1 0 0
i h1 0 0
i
0 0 K −1 0 0 K −1

(Kyc2 +1)λ4 −Kxc yc λ4
Central dot product vc| · Pc · wc = vc| −Kxc yc λ4
wc
(Kx2c +1)λ4
207
19 Stereographic projection
Here we start to develop another model for our geometry.
Stereographic projection coordinates

Now we will examine the second way to produce coordinates uniformly for spher-
ical, plane, and hyperbolic geometry that will allow us to pass smoothly through
the three geometries. Now let’s project K-geometry,
1 = K x2 + y 2 + z 2

onto the plane z = 1 using the ‘south pole’ S = (0, 0, −1) as the center of
projection:
Let’s look at this from a different vantage point, say with our eye along the edge
of the plane z = 1:
Problem 184 If K = 1 where does the ‘equator’ map to? What about the
‘Northern hemisphere?’ How about the ‘Southern hemisphere?’
208
Problem 185 Use similar triangles to explain why for any given (x, y, z),
there is a number ρ such that
ρ · (xs , ys , 2) = (x, y, z + 1).
209

z = 1 plane with center of projection S, write (xs , ys ) as a function of (x, y, z).
210

z = 1 plane with center of projection S, write (x, y, z) as a function of (xs , ys ).
Hint: Note that
ρ · (xs , ys , 2) = (x, y, z + 1) where z = 2ρ − 1.
Hence if
1 = K x2 + y 2 + z 2

we may write
1 = K (ρ · xs )2 + (ρ · ys )2 + (2ρ − 1)2

and solve for ρ.
211
Stereographic projection dot product

Once more, we want to be able to find a dot product that will allow us to
compute lengths in stereographic projection coordinates that will agree with
the K-dot product, and hence the euclidean dot product.

decompose as
 
x(xs ,ys )
 y(xs ,ys ) 
γs z(xs ,ys )
t (xs (t), ys (t)) γ(t) = (x(t), y(t), z(t))

dx dy dz
, , ,
dt dt dt
dxs dys ∂x ∂y ∂z ∂x ∂y ∂z
in terms of , , , , , , , and .
dt dt ∂xs ∂xs ∂xs ∂ys ∂ys ∂ys
changes.
212
result of your computation in matrix notation to find Ds such that
 
dx/dt
dy/dt = Ds · dxs /dt
dys /dt
dz/dt
∂x ∂y ∂z ∂x ∂y ∂z
∂xs ∂xs ∂xs ∂ys ∂ys ∂ys
changes.
213
∂x ∂y ∂z ∂x ∂y
Problem 190 Now find Ps in terms of K, , , , , , and
∂xs ∂xs ∂xs ∂ys ∂ys
∂z
such that
∂ys
 
dxs
dx dy dz dx dy dz dxs dys  dt 
, , •K , , = · Ps ·  dy .
dt dt dt dt dt dt dt dt s
dt
changes.
214
Before we actually compute Ps , it will help to compute the partial derivatives.
Problem 191 Set
x(xs , ys ) = ρ · xs ,
y(xs , ys ) = ρ · ys ,
z(xs , ys ) = 2ρ − 1,
and show that

∂x K ∂x K
=ρ− ρ2 x2s , =− ρ2 xs ys ,
∂xs 2 ∂ys 2

∂y K ∂y K
=− ρ2 xs ys , =ρ− ρ2 ys2 ,
∂xs 2 ∂ys 2
∂z ∂z
= −Kρ2 xs , = −Kρ2 ys .
∂xs ∂ys
Hint: Work in the following way:
(a) Recall x = ρ · xs .
∂x ∂ρ
(b) Note that = ρ + xs · .
∂xs ∂xs
(c) Express the partial derivative in terms of ρ, K, xs , and ys .
215
Problem 192 With the same setting as above, show that Ps is

2
ρ 0
Ps = .
0 ρ2
Hint: When simplifying, combine the terms with the highest degree of ρ and note
that
K x2s + ys2 + 4

−1
ρ = .
4
216
Definition 26. Let vs and ws be vectors in (xs , ys )-coordinates originating at

the same (xs , ys )-coordinate. Define
vs •s ws = vs| Ps ws
where 2
ρ 0
Ps =
0 ρ2
is determined by the coordinate that the vectors originate from.
Again, when the finite is infinite and the infinite

is finite
Just like central projection, stereographic projection allows us to think about
both “finite” areas and “infinite” areas in new ways.
When K is positive, the finite is infinite When K > 0, we are working

with a sphere. As we know, a sphere has finite surface area.
Problem 193 Where does the Northern hemisphere of the R-sphere map to
under stereographic projection?
217
Problem 194 Where does the Southern hemisphere of the R-sphere map to
under stereographic projection?
218
Problem 195 As a point approaches the south pole of the R-sphere where
does it move to under stereographic projection?
Hint: Simply transform the relevant coordinate.
219
Problem 196 Explain why one might say that stereographic projection makes
the “finite” seem “infinite.”
220
When K is negative, the infinite is finite Recall that, if K < 0, the

equation of K-geometry becomes
z 2 − |K|(x2 + y 2 ) = 1.
This describes a 2-sheeted hyperboloid with the z-axis as major axis. We will
only consider the sheet on which z is positive as forming the K-geometry. Recall:
• This hyperboloid has infinite surface area.
• It is obtained by rotating the hyperbola
z 2 − |K| x2 = 1
in the (x, z)-plane around the z-axis.
Problem 197 Consider points of the form (x, 0, z) in K-geometry, K < 0. As

x and z shoot off to infinity, what happens to the corresponding stereographic
projection coordinate xs ?
Hint: As you did with central projection, use similar triangles and the fact that
p
z = 1 + |K|x2
and compute a limit.
221
Problem 198 When K < 0, what is the set of points (xs , ys ) in stereographic
projection which correspond to points on the hyperboloid?
Hint: Again, use the previous question and rotational symmetry.
222
Problem 199 Explain why one might say that stereographic projection makes
the “infinite” seem “finite.”
223
The (xs , ys )-coordinates are called

p Poincaré coordinates for hyperbolic geom-
etry and the disk of radius 2/ |K| called the Poincaré model for hyperbolic
geometry in honor of the famous French geometer, Henri Poincaré.
224
20 Lines, angles, and areas in stereographic projection
20 Lines, angles, and areas in

stereographic projection
Here we look at lines, angles, and areas in stereographic projection coordinates.
Stereographic projection sends lines to circles (or

lines)
As we know, we cannot make perfect flat maps of 3D surfaces. In stereographic
projection, shortest paths are sent to either lines or circles, and (perhaps sur-
prisingly!) circles are sent to circles.
Problem 201 Show that “lines” in K-geometry correspond to either lines or

circles in (xs , ys )-coordinates under stereographic projection.
Hint: (a) Recall that intersecting the K-surface

1 = K x2 + y 2 + z 2

with the plane

ax + by + cz = 0
produces a K-geometry line.
(b) Use the projection formulas
4xs
x= ,
K(x2s + ys2 ) + 4
4ys
y= ,
K(x2s + ys2 ) + 4
4 − K(x2s + ys2 )
z= ,
4 + K(x2s + ys2 )
and note that the points generated this way automatically satisfy the condition
1 = K(x2 + y 2 ) + z 2 .
(c) If c = 0, then you will find the line

axs + bys = 0.
(d) If c 6= 0, then you will find the circle

2 2
4a2 + 4b2 + 4c2 K

2a 2b
xs − + ys − = .
cK cK (cK)2
You will need to complete the square to convert it into this form.
225
Remark 4. With entirely similar reasoning, you can show that circles in stere-
ographic projection are also sent to either circles (or lines).
When we project shortest paths, they map to circles (unless c = 0 as above).

When K < 0, the circles meet the “circle at infinity” at right angles. We won’t
make you prove this, but try to convince yourself that it’s true!
Problem 202 If K < 0 explain why Euclid’s fifth axiom:
Through a point not on a line there passes a unique parallel line.
fails.
226
Stereographic projection preserves angles

A remarkable fact about stereographic projection is that it “preserves” angles.
This means that the angles we see in stereographic projection are true represen-
tations of the angle in euclidean coordinates. Now that we are masters of dot
products, we will prove this fact with ease!
Problem 203 Show that

vs •s ws vs • ws
= .
|vs |s · |ws |s |vs | · |ws |
Hint: On the right-hand side we are using the euclidean dot product and euclidean
length formula.
227
Problem 204 Explain how the last problem shows that stereographic projec-
tion “preserves” angles.
228
Areas in stereographic projection coordinates

Let’s use the power of stereographic coordinates to compute the areas of spher-
ical, hyperbolic, and euclidean circles.
We know that the area of a region in stereographic coordinates is given by
v
∂X ∂X ∂X ∂X
u  
ZZ u •K •K
u ∂x ∂xs ∂ys ∂xs  dx dy
udet 
t  ∂Xs ∂X ∂X ∂X  s s
Cs •K •K
∂xs ∂ys ∂ys ∂ys
Problem 205 Give a heuristic explanation of why this integral computes

what we say it computes.
229
You have already shown that

v
∂X ∂X ∂X ∂X
u  
ZZ u •K •K
u ∂x ∂xs ∂ys ∂xs  dx dy
t  ∂Xs
udet 
∂X ∂X ∂X  s s
Cs •K •K
v
u 
∂X
  ↑ ↑

ZZ u ← → 1 0 0
 ∂X ∂X 
u
∂xs
= udet   0 1 0   dxs dys .
u  
∂X −1 ∂xs ∂ys
Cs ← → 0 0 K
t
∂ys ↓ ↓
Problem 206 Explain why


∂X
  ↑ ↑

← → 1 0 0
∂xs  ∂X ∂X 
 0 1 0   = Ps .
 
 ∂X ∂xs ∂ys
← → 0 0 K −1
∂ys ↓ ↓
Hint: No new computations need to be done, just look at how Ps was derived.
230
Hence now we see

v
∂X ∂X ∂X ∂X
u  
ZZ u •K •K ZZ
u ∂x
udet  s
 ∂xs ∂ys ∂xs  dx dy = p
det Ps dxs dys .
t ∂X ∂X ∂X ∂X  s s
Cs •K •K Cs
p
Problem 207 Compute det Ps in terms of K, xs , and ys .
231
For the next few problems, consider the following diagram:
Problem 208 Compute the area of Cs in K-geometry by taking the integral

ZZ p
det Ps dxs dys .
Cs
Hint: In the previous problem, you should have obtained

√ 16
det Ps = .
(K(x2s + ys2 ) + 4)2
Hint: This integral is easiest to compute in polar coordinates. Recall that to convert
to polar coordinates, you must set
p
r = x2s + ys2 ,
θ = arctan(ys /xs ),
and replace dxs dys with r dr dθ.
232
Problem 209 Explain why the radius of Cs in K-geometry is given by

Z a s
dxs dys dxs dys
r= , •s , dt
0 dt dt dt dt
where
(xs (t), ys (t)) = (t, 0).
233
Problem 210 When K > 0, find a in terms of the radius r from the previous
problem.
Hint: There are two ways of going about this:
• Compute the integral from the previous problem, using the fact that
d A
arctan(Ax) = .
dx 1 + (Ax)2
Then solve for a.
• Or, consider the following diagram in Euclidean geometry and use facts about
circles and triangles:
234
Problem 211 When K > 0, plug in and simplify to get a formula for the
area of a circle in K-geometry in terms of its radius r.
235
Problem 212 What happens as r approaches πR = πK −1/2 ?

area
When r gets very small, what happens to 2 ?
r
Explain how these computations square with our understanding of spheres.
236
Problem 213 When K < 0, find a in terms of the radius r.
Hint: There are two ways of going about this:
• Compute the integral from a previous problem, using the fact that
d A
artanh(Ax) = .
dx 1 − (Ax)2
Then solve for a.
• Or, use the fact that the K-length of the line
from (1, 0, 0) to (|K|−1/2 sinh(σ), 0, cosh(σ))
is |K|−1/2 σ, and compare that to the length of its stereographic projection onto
the plane z = 1. Use the identities
cosh(2x) = 2 cosh2 (x) − 1

sinh(2x) = 2 sinh(x) cosh(x).
to simplify.
Hint: This problem is potentially more involved than the spherical version, but in
the end you should get the same answer with an extra “h”.
237
Problem 214 When K < 0, plug in to get a formula for the area of a circle
in K-geometry in terms of its radius r.
238
Notice that for big enough x, sinh(x) ≈ ex /2, since the e−x term is pretty much
zero. So the area of a circle in hyperbolic geometry grows exponentially with
radius, unlike in Euclidean geometry where it only grows quadratically. There’s
a lot of space within radius r of a point in the hyperbolic plane.
area
Problem 215 When r gets very small, what happens to 2 ? Explain why
r
this makes sense.
239
Stereographic projection unifies euclidean, spher-

ical, and hyperbolic geometry
Under stereographic projection, our new dot product defined by
vs •s ws = vs| · Ps · ws
make sense when K is zero, positive, and negative. Hence this dot product
makes sense for euclidean, spherical, and hyperbolic geometry. However, in
stereographic projection, shortest paths on the K-surface
K(x2 + y 2 ) + z 2 = 1
map to circles (or lines) in the plane z = 1. The advantage of stereographic

projection over central projection is that angles are preserved in stereographic
projection. This means that when angles are projected into the plane via stere-
ographic projection, the angle we see in the (xs , ys )-plane is the actual angle
between two vectors. Summarizing, we have:

x DNE DNE

v b DNE DNE
h1 0 0
i h1 0 0
i
0 0 K −1 0 0 K −1

(Kyc2 +1)λ4 −Kxc yc λ4
Central dot product vc| · Pc · wc = vc| −Kxc yc λ4
wc
(Kx2c +1)λ4
h i
ρ2 0
Stereographic dot product vs| · Ps · ws = vs| 0 ρ2
ws
240
21 Hyperbolic lunes and triangles

In this activity, we explore the areas of lunes and triangles in hyperbolic geom-
etry.
We know that the area of a triangle on the R-sphere with angles α, β, and γ is
given by
R2 (α + β + γ − π).
Problem 217 Briefly sketch the line of reasoning used to deduce the formula
for the area of a triangle on the R-sphere.
241
We will use similar reasoning to deduce the formula for the area of a triangle
with angles α, β, and γ in hyperbolic geometry.
Hyperbolic lunes
Consider the following diagram on the Klein disk, that is the central projection
of hyperbolic geometry:
Notice that in both the Klein model and the Poincaré model, every line hits
“the circle at infinity” at two specific points. This means we can talk about
shapes with “vertices at infinity”.
Definition 27. We define an α-lune in hyperbolic geometry to be a triangle

with an angle of measure α where vertices opposite of α are at infinity.
We have seen that

v
∂X ∂X ∂X ∂X
u  
ZZ u •K •K ZZ p
u
udet  ∂xc ∂xc ∂yc ∂xc  dx dy = det Pc dxc dyc
t  ∂X ∂X ∂X ∂X  c c
Lc •K •K Lc
ZZ
−3/2
K x2c + yc2 + 1

= dxc dyc .
Lc
In what follows, we will assume that K = −1. This will simplify the computa-
tions somewhat. Once the mathematician is familiar with this case, the general
case when K < 0 will fall easily.
242
Problem 218 Convert

ZZ
−3/2
− x2c + yc2 + 1

dxc dyc
Lc
to polar coordinates and compute the integral.
Hint: Redraw the picture above with α = 2β and K = −1.
Hint: Recall that to convert to polar coordinates, set

p
r = x2c + yc2 ,
θ = arctan(yc /xc ),
and replace dxc dyc with r dr dθ.
Hint: At some point you may wish to use the following identities:
cos2 θ + sin2 θ = 1 cos2 β + sin2 β = 1

cos2 θ = 1 − sin2 θ cos2 β = 1 − sin2 β
So
cos2 θ − cos2 β = 1 − sin2 θ − 1 − sin2 β

= sin2 β − sin2 θ.
Hint: It may also be helpful to recall that when a > 0,

Z
1 u
√ du = arcsin √ + C.
a − u2 a
Hint: Finally, as a gesture of friendship, I will tell you that you will (hopefully!)
deduce that this integral equals π − α.
243
Hyperbolic triangles
Again assuming K = −1, we will use our knowledge that the hyperbolic α-lune
has area π − α to compute the area of an ideal triangle in hyperbolic geometry.
Note, now we are working under stereographic projection in the Poincaré disk
(because it’s more fun to draw.)
Definition 28. An ideal triangle is a triangle in hyperbolic geometry with all
of its vertices at infinity.
Problem 219 Use the diagrams below
to help compute the area of an ideal triangle in hyperbolic geometry.
Hint: Above we see a comic-strip of Poincaré disks.
Hint: Make sure your computation is general. That is, start with an arbitrary ideal
triangle and then fill in the rest of the diagram.
244
Definition 29. An ideal n-gon is an n-gon in hyperbolic geometry with its

vertices at infinity.
Problem 220 Use induction to derive a formula for the area of any ideal
n-gon in hyperbolic geometry.
245
Problem 221 Use the diagrams below
to help compute the area of a triangle in hyperbolic geometry.
Hint: Make your own drawings inspired by those above.
Hint: Make sure your computation is general: start with an arbitrary triangle and
explain how you can fill in the rest of the diagram.
246
247
22 Connections to special relativity

Here we see a connection to special relativity.
In this section we will show a connection to special relativity. Underlying all of

this work is the notion of a group action.
Definition 30. If G is a group and X is a set, then a (right) group action

is a map (written as x · g, for all g ∈ G and x ∈ X) satisfying the following
properties:
(a) (x · g1 ) · g2 = x · (g1 · g2 ), for all g1 , g2 ∈ G and x ∈ X.
(b) x · I = x for all x ∈ X where I is the identity element of the group.
As we work, you should try to see where we are using group actions.
Galilean transformations of velocity
In this case we will let points in central projection coordinates (xc , yc ) represent
“observed velocities.” We have shown that orthogonal matrices in K-warped
space form a group. That is we have shown that
(a) multiplication of orthogonal matrices is associative,
(b) the product of two orthogonal matrices is orthogonal,
(c) the identity matrix is orthogonal,
(d) the inverse matrix M −1 of a orthogonal matrix M is orthogonal.
The group of orthogonal 3 × 3 matrices is denoted as O(3). We’ll denote the

group of K-orthogonal matrices as OK (3). When K < 0, we already know three
(classes of) elements of this group:
   
1 0 0 cos θ sin θ 0
I = 0 1 0 , Mθ = − sin θ cos θ 0 ,
0 0 1 0 0 1
0 |K|1/2 · sinh ψ
 
cosh ψ
Nψ =  0 1 0 
−1/2
|K| · sinh ψ 0 cosh ψ
If we have an observer at point O = (0, 0) in central projection coordinates, we
can denote any other observed velocity in the plane as a point (xc , yc ).
248
Problem 223 With the setting described above in mind, what does the point
(.3, .3) represent?
249
Now if we have a configuration of points in this velocity space, we would like to

answer the following question:
If a new observer O0 views the original observer O as moving at a

velocity of (u, 0), how do we transform all other velocities measured
by the original observer?
Problem 224 Can you give a naive solution to this problem?
250
Even if you can give a naive solution to this problem, we will think of this
problem in central projection coordinates. Eventually we want to solve a much
harder problem where the naive solution will fail, however the solution using
central projection coordinates will prevail.
We’ll start by working when K < 0, then we will take the limit as K goes to
zero, and see what we find.
Problem 225 Convert the K-rigid motion
0 |K|1/2 · sinh ψ
 
cosh ψ
Nψ =  0 1 0 
|K|−1/2 · sinh ψ 0 cosh ψ
to a rigid motion of the (xc , yc )-plane. Call your answer νψ .
Hint: Either write down the answer from a previous exercise or recall that a rigid
motion in (x, y, z)-coordinates
 
m11 m12 m13
M = m21 m22 m23 
m31 m32 m33
converts to

xc · m11 + yc · m21 + m31 xc · m12 + yc · m22 + m32
µ(xc , yc ) = , .
xc · m13 + yc · m23 + m33 xc · m13 + yc · m23 + m33
251
We need to be translating velocities

by a constant
amount in central projection.
To do this, we must set ψ = artanh u · |K|1/2 .

Problem 226 Substitute artanh u · |K|1/2 for ψ in your formula above.
Hint:
1
cosh(artanh(x)) = √ ,
1 − x2
x
sinh(artanh(x)) = √ .
1 − x2
252
You have just written down a transformation that will work for any negative
value of K. We need this to work for K = 0.
Problem 227 Take the limit as K goes to zero in your answer above.
253
At this point we are working in the euclidean plane, that is, central projection
when K = 0. The answer to our question:

is almost too easy, as it is that we apply the transformation
νu (xc , yc ) = (xc + u, yc ).
Before you object that we have done nothing but “made mathematics difficult”
we preemptively retort:
We are building a foundation for future work.
Problem 228 Can you rephrase what we are doing in terms of group actions?
254
Hyperbolic transformations of velocity

Armed with experience from above, let’s now suppose we are again attacking
the question:

however, this time we are working not on the euclidean plane, but on the Klein
disk, when K = −1. The beauty of our work above is that it can be directly
applied in this case:
Problem 229 Show that when u ∈ [0, 1) and K = −1,

√ !
xc + u yc · 1 − u2
νu (xc , yc ) = , .
xc · u + 1 xc · u + 1
255
For those who are in-the-know, a minor miracle has just occurred. The trans-
formation √ !
xc + u yc · 1 − u2
νu (xc , yc ) = , ,
xc · u + 1 xc · u + 1
is in fact the velocity transformation for Einstein’s special theory of relativity
when one takes the speed of light to be c = 1!
Problem 230 Show that the transformation of the Klein disk νu is given by
the transformation of K-warped space
1 u · |K|
 
p
 1 − u2 · |K| 0 p
 1 − u2 · |K| 

Nu =  0 1 0 

 u 1 
p 0 p
1 − u2 · |K| 1 − u2 · |K|
when K = −1.
256
Problem 231 Show that

1 u · |K|
 
p
 1 − u2 · |K| 0 p
 1 − u2 · |K| 

Nu = 
 0 1 0 

 u 1 
p 0 p
1 − u2 · |K| 1 − u2 · |K|
is a K-rigid motion.
257
The upshot to all of these computations is that transformations of the form:

√ !
xc + u yc · 1 − u2
νu (xc , yc ) = , ,
xc · u + 1 xc · u + 1
correspond exactly to a rigid motion of the surface
−(x2 + y 2 ) + z 2 = 1
with the K-dot product. One imaginative interpretation of these computations

is that we are in fact living on a hyperbolic surface, and that we experience our
world through the lens of a central projection. What an amazing thought!
258
23 The art of Escher

Here we look for connections between mathematics and art.
The famous artist M.C. Escher is known for his mind-boggling art. Often there
is a strong mathematical component to his art. One thing that fascinated Escher
was the question of how to depict an infinitely repeating pattern in a finite space.
Problem 233 Do you know of any methods for making the “infinite finite”
from this class? If so, describe them.
259
Around 1954, H.S.M. Coxeter (the famous geometer) learned of Escher’s work.
As a source of inspiration, Coxeter sent Escher a picture similar to this one:
Apparently, Escher was “shocked” by this diagram, attempted to reproduce it

himself, and ran into some trouble. He then wrote to Coxeter for assistance.
After some time, Escher produced his own versions, three of which are provided
here for your viewing enjoyment:
Circle Limit I. M.C. Escher, 1958
260
Circle Limit III. M.C. Escher, 1959
Circle Limit IV (Heaven and Hell). M.C. Escher, 1960
261
Problem 234 Write down as many (mathematical) questions as you can re-
lated to Escher’s work above. After you have your questions, label them as
“Level 1,” “Level 2,” or “Level 3” where:
Level 1 Means you know the answer, or know exactly how to do this problem.
Level 2 Means you think you know how to do the problem, or will be able to
figure out how to do the problem.
Level 3 Means you have no idea how to do the problem.
262
Problem 235 Someone once asked, can these “distorted” images be “un-
distorted.”
(a) Answer “no” and support your answer with an argument.

(b) Answer “yes” and support your answer with an argument.
263
Problem 236 Building on ideas we learned in this course, what other sources
of inspiration could you give to artists?
264
Problem 237 Summarize the results and ideas from this course.
265

Workbook of Two Dimensional Geometries

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Workbook of Two Dimensional Geometries

Uploaded by

Copyright:

Available Formats

Workbook of

C. Herbert Clemens and Bart Snapp

November 17, 2020

As a young mathematician I was introduced to the classic Leçons sur la Géométrie

making z a function of x and

(b) Given differentiable functions

(y1 (x) , . . . , yn (x))

then substituting we have

z (y1 (x) , . . . , yn (x))

making z a function of x and

(c) Given differentiable mappings

(y1 (x1 , . . . , xm ) , . . . , yn (x1 , . . . , xm ))

zk (y1 (x1 , . . . , xm ) , . . . , yn (x1 , . . . , xm ))

and fixing x1 , . . . , xi−1 and xi+1 , . . . , xm we make zk a function of xi and

Theorem 2 (Substitution Rule). We will present two different versions:

(a) If f (y) is a function of y and y (x) is a function of x,

(b) If f (y) is a function of y = (y1 , . . . , ym ) and the function

takes the region Rx to the region Ry , then

Remark 1. Special message to current or future teachers of high school ge-

2 Euclid’s postulates for plane

• State the axioms of neutral geoemtry.

In Western civilization, the primary source of our understanding of plane geom-

(a) Which region you pick as the interior/exterior.

One implicit assumption of two-dimensional neutral geometry is the existence of

(b) M · ~v is a positive scalar multiple of w.

Problem 3 Give an example to show that there is no universal SSA law.

Problem 4 In the diagram below, we see the intersection of BC and AE.

Hint: First you should explain why ∠BDA = ∠CDE.

(a) Show in neutral geometry that, for 4ABC

the exterior angle ε of the triangle at C is greater than either remote

Problem 6 Show in NG that, if two lines cut by a transversal line have a

Sum of angles in a triangle in NG

Hint: Use the previous problems.

Hint: Consider the following additions to our diagram above:

Be sure to carefully explain how points D and E are constructed.

Proof Seeking a contradiction, suppose there exists a triangle 4ABC where

We now have two relations:

As we have seen, this is impossible as two interior angles of a triangle cannot

Problem 11 Show the following:

Hint: Given any quadrilateral, we can break it into two triangles.

This fact wasn’t written down and proved until 1899!

Problem 12 Let ABCD be a quadrilateral with ∠ABC = ∠BCD right an-

(a) • |AB| = |CD| implies that ∠BAD = ∠ADC,

∠BAD < ∠BA0 D.

By the (already proved) first implication

Problem 14 Summarize the results from this section. In particular, indicate

3 Rectangles and cartesian coordinates

Euclid’s fifth axiom, the parallel postulate

NG together with E5 is called euclidean geometry (EG). As mentioned above,

Hint: Draw a picture and seek a contradiction.

Problem 16 Show that angles ∠BAC and ∠B 0 A0 C 0 in the euclidean plane

Problem 17 Use the ‘uniqueness’ assertion in E5 together with what we have

Problem 18 Show that in EG the sum of the interior angles of a quadrilateral

Explain why `3 is parallel to `2 .

Problem 20 Show that there is a cartesian coordinate system on EG. This

The distance formula in EG

Problem 21 State and prove the Pythagorean theorem in EG.

Hint: In the cartesian plane, construct a square with vertices

Inside that square, construct the square with vertices

(a, 0) , (a + b, a) , (b, a + b) , (0, b) .

Problem 22 Use the Pythagorean theorem to justify the euclidean distance

Problem 23 Summarize the results from this section. In particular, indicate